PGFplots PDF

Manual for Package pgfplots
2D/3D Plots in LATEX, Version 1.13
http://sourceforge.net/projects/pgfplots
Dr. Christian Feuersänger

[email protected]
Revision 1.13 (2016/01/06)
Abstract
pgfplots draws high–quality function plots in normal or logarithmic scaling with

a user-friendly interface directly in TEX. The user supplies axis labels, legend en-
tries and the plot coordinates for one or more plots and pgfplots applies axis scal-
ing, computes any logarithms and axis ticks and draws the plots. It supports line
plots, scatter plots, piecewise constant plots, bar plots, area plots, mesh– and surface
plots, patch plots, contour plots, quiver plots, histogram plots, box plots, polar axes,
ternary diagrams, smith charts and some more. It is based on Till Tantau’s package
pgf/Tik Z.
Contents
1 Introduction 7
2 About PGFPlots: Preliminaries 8

2.1 Components . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.2 Upgrade remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.2.1 New Optional Features . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.2.2 Old Features Which May Need Attention . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.3 The Team . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.4 Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.5 Installation and Prerequisites . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.5.1 Licensing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.5.2 Prerequisites . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.5.3 Installation in Windows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.5.4 Installation of Linux Packages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.5.5 Installation in Any Directory - the TEXINPUTS Variable . . . . . . . . . . . . . . . . . 12
2.5.6 Installation Into a Local TDS Compliant texmf-Directory . . . . . . . . . . . . . . . . 13
2.5.7 Installation If Everything Else Fails... . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.6 Troubleshooting – Error Messages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.6.1 Problems with available Dimen-registers . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.6.2 Dimension Too Large Errors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.6.3 Restrictions for DVI-Viewers and dvipdfm . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.6.4 Problems with TEX’s Memory Capacities . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.6.5 Problems with Language Settings and Active Characters . . . . . . . . . . . . . . . . . 15
2.6.6 Other Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3 Step–by–Step Tutorials 16
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
3.2 Solving a Real Use–Case: Function Visualization . . . . . . . . . . . . . . . . . . . . . . . . . 16
3.2.1 Getting the Data Into TEX . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.2.2 Fine–Tuning of the First Picture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.2.3 Adding the Second Picture with a Di↵erent Plot . . . . . . . . . . . . . . . . . . . . . 19
3.2.4 Fixing the Vertical Alignment and Adjusting Tick Label Positions . . . . . . . . . . . 21
3.2.5 Satisfying Di↵erent Tastes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.2.6 Finishing Touches: Automatic Generation of Individual Pdf Graphics . . . . . . . . . 23
3.2.7 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
3.3 Solving a Real Use–Case: Scientific Data Analysis . . . . . . . . . . . . . . . . . . . . . . . . 24
3.3.1 Getting the Data into TeX . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.3.2 Adding the Remaining Data Files of Our Example. . . . . . . . . . . . . . . . . . . . . 25
3.3.3 Add a Legend and a Grid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.3.4 Add a Selected Fit-line . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.3.5 Add an Annotation using Tik Z: a Slope Triangle . . . . . . . . . . . . . . . . . . . . . 27
3.3.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.4 Use–Cases involving Scatter Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.4.1 Scatter Plot Use–case A . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.4.2 Scatter Plot Use–case B . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.4.3 Scatter Plot Use–case C . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.4.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2
CONTENTS 3
3.5 Solving a Real Use–Case: Functions of Two Variables . . . . . . . . . . . . . . . . . . . . . . 33

3.5.1 Surface Plot from Data File . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
3.5.2 Fine–Tuning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
3.5.3 Adding Scattered Data on Top of the Surface . . . . . . . . . . . . . . . . . . . . . . . 35
3.5.4 Computing a Contour Plot of a Math Expression . . . . . . . . . . . . . . . . . . . . . 36
3.5.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
4 The Reference 39
4.1 TEX-dialects: LATEX, ConTEXt, plain TEX . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
4.2 The Axis-Environments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
4.3 The \addplot Command: Coordinate Input . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
4.3.1 Coordinate Lists . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
4.3.2 Reading Coordinates From Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
4.3.3 Computing Coordinates with Mathematical Expressions . . . . . . . . . . . . . . . . . 53
4.3.4 Mathematical Expressions And File Data . . . . . . . . . . . . . . . . . . . . . . . . . 57
4.3.5 Computing Coordinates with Mathematical Expressions (gnuplot) . . . . . . . . . . . 59
4.3.6 Computing Coordinates with External Programs (shell) . . . . . . . . . . . . . . . . . 61
4.3.7 Using External Graphics as Plot Sources . . . . . . . . . . . . . . . . . . . . . . . . . . 62
4.3.8 Keys To Configure Plot Graphics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
4.3.9 Reading Coordinates From Files . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
4.4 About Options: Preliminaries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
4.4.1 Pgfplots and Tik Z Options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
4.5 Two Dimensional Plot Types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
4.5.1 Linear Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
4.5.2 Smooth Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
4.5.3 Constant Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
4.5.4 Bar Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
4.5.5 Histograms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
4.5.6 Box Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
4.5.7 Comb Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
4.5.8 Quiver Plots (Arrows) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
4.5.9 Stacked Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
4.5.10 Area Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
4.5.11 Closing Plots (Filling the Area Under Plots) . . . . . . . . . . . . . . . . . . . . . . . 105
4.5.12 Scatter Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
4.5.13 1D Colored Mesh Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
4.5.14 Interrupted Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
4.5.15 Patch Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
4.5.16 Image (Matrix) Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
4.5.17 Polar Coordinates / Polar Axes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
4.5.18 Tieline Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
4.5.19 Smith Charts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
4.6 Three Dimensional Plot Types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
4.6.1 Before You Start With 3D . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
4.6.2 The \addplot3 Command: Three Dimensional Coordinate Input . . . . . . . . . . . . 123
4.6.3 Line Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
4.6.4 Scatter Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
4.6.5 Mesh Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
4.6.6 Surface Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
4.6.7 Surface Plots with Explicit Color . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
4.6.8 Contour Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
4.6.9 Filled Contour Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
4.6.10 Parameterized Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
4.6.11 3D Quiver Plots (Arrows) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
4.6.12 Image (Matrix) Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
4.6.13 Patch Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169
4.6.14 About 3D Const Plots and 3D Bar Plots . . . . . . . . . . . . . . . . . . . . . . . . . 176
4.6.15 Mesh/Surface Plots with Holes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 176
4 CONTENTS
4.7 Markers, Linestyles, (Background-) Colors and Colormaps . . . . . . . . . . . . . . . . . . . . 176

4.7.1 Markers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 176
4.7.2 Line Styles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
4.7.3 Edges and Their Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
4.7.4 Font Size and Line Width . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
4.7.5 Colors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184
4.7.6 Color Maps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186
4.7.7 Cycle Lists – Options Controlling Line Styles . . . . . . . . . . . . . . . . . . . . . . . 194
4.7.8 Axis Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210
4.8 Providing Color Data - Point Meta . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
4.8.1 Point Meta Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
4.8.2 User Input Format for Point Meta . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212
4.8.3 Mapping Point Meta and Color Maps . . . . . . . . . . . . . . . . . . . . . . . . . . . 215
4.9 Axis Descriptions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219
4.9.1 Placement of Axis Descriptions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220
4.9.2 Alignment of Axis Descriptions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 226
4.9.3 Labels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231
4.9.4 Legends . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 234
4.9.5 Legend Appearance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 237
4.9.6 Legends with \label and \ref . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245
4.9.7 Legends Outside Of an Axis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 247
4.9.8 Legends with Customized Texts or Multiple Lines . . . . . . . . . . . . . . . . . . . . 249
4.9.9 Axis Lines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 250
4.9.10 Moving Axis Lines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 255
4.9.11 Two Ordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 256
4.9.12 Axis Discontinuities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 257
4.9.13 Color Bars . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 260
4.9.14 Color Bars Outside Of an Axis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 269
4.10 Scaling Options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 270
4.10.1 Common Scaling Options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 271
4.10.2 Scaling Descriptions: Predefined Styles . . . . . . . . . . . . . . . . . . . . . . . . . . . 282
4.10.3 Scaling Strategies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 285
4.11 3D Axis Configuration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 287
4.11.1 View Configuration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 287
4.11.2 Styles Used Only For 3D Axes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 289
4.11.3 Appearance Of The 3D Box . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 290
4.11.4 Axis Line Variants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 293
4.12 Error Bars . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 293
4.12.1 Input Formats of Error Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . 297
4.13 Number Formatting Options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 299
4.13.1 Frequently Used Number Printing Settings . . . . . . . . . . . . . . . . . . . . . . . . 299
4.13.2 PGFPlots-specific Number Formatting . . . . . . . . . . . . . . . . . . . . . . . . . . . 300
4.14 Specifying the Plotted Range . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 304
4.14.1 Configuration of Limits Ranges . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 304
4.14.2 Accessing Computed Limit Ranges . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 310
4.15 Tick Options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 311
4.15.1 Tick Coordinates and Label Texts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 311
4.15.2 Tick Alignment: Positions and Shifts . . . . . . . . . . . . . . . . . . . . . . . . . . . . 322
4.15.3 Tick Scaling - Common Factors In Ticks . . . . . . . . . . . . . . . . . . . . . . . . . . 324
4.15.4 Tick Fine-Tuning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 327
4.16 Grid Options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 330
4.17 Custom Annotations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 330
4.17.1 Accessing Axis Coordinates in Graphical Elements . . . . . . . . . . . . . . . . . . . . 331
4.17.2 Placing Nodes on Coordinates of a Plot . . . . . . . . . . . . . . . . . . . . . . . . . . 335
4.17.3 Placing Decorations on Top of a Plot . . . . . . . . . . . . . . . . . . . . . . . . . . . . 339
4.18 Style Options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 340
4.18.1 All Supported Styles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 340
4.18.2 (Re)Defining Own Styles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 348
CONTENTS 5
4.19 Alignment Options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 348

4.19.1 Basic Alignment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 348
4.19.2 Vertical Alignment with baseline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 351
4.19.3 Horizontal Alignment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 352
4.19.4 Alignment In Array Form (Subplots) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 353
4.19.5 Miscellaneous for Alignment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 357
4.20 The Picture’s Size: Bounding Box and Clipping . . . . . . . . . . . . . . . . . . . . . . . . . . 357
4.20.1 Bounding Box Restrictions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 357
4.20.2 Clipping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 360
4.21 Symbolic Coordinates and User Transformations . . . . . . . . . . . . . . . . . . . . . . . . . 361
4.21.1 String Symbols as Input Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . 362
4.21.2 Dates as Input Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 364
4.22 Skipping Or Changing Coordinates – Filters . . . . . . . . . . . . . . . . . . . . . . . . . . . . 366
4.23 Transforming Coordinate Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 371
4.23.1 Interaction of Transformations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 374
4.24 Fitting Lines – Regression . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 374
4.25 Miscellaneous Options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 377
4.26 Tik Z Interoperability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 383
4.27 Layers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 385
4.27.1 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 385
4.27.2 Using Predefined Layers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 386
4.27.3 Changing the Layer of Graphical Elements . . . . . . . . . . . . . . . . . . . . . . . . 388
4.28 Technical Internals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 389
5 Related Libraries 391

5.1 Clickable Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 391
5.1.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 391
5.1.2 Requirements for the Library . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 394
5.1.3 Customization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 395
5.1.4 Using the Clickable Library in Other Contexts . . . . . . . . . . . . . . . . . . . . . . 397
5.2 ColorBrewer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 397
5.2.1 Usage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 398
5.2.2 Sequential Schemes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 400
5.2.3 Diverging Schemes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 400
5.2.4 Qualitative Schemes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 401
5.2.5 Interaction with the ColorBrewer website . . . . . . . . . . . . . . . . . . . . . . . . . 401
5.2.6 External Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 401
5.3 Colormaps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 402
5.4 Dates as Input Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 407
5.5 Decoration: Soft Clipping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 407
5.6 Image Externalization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 407
5.7 Fill between . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 408
5.7.1 Filling an Area . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 408
5.7.2 Filling Di↵erent Segments of the Area . . . . . . . . . . . . . . . . . . . . . . . . . . . 410
5.7.3 Filling only Parts Under a Plot (Clipping) . . . . . . . . . . . . . . . . . . . . . . . . . 412
5.7.4 Styles Around Fill Between . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 415
5.7.5 Key Reference . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 416
5.7.6 Intersection Segment Recombination . . . . . . . . . . . . . . . . . . . . . . . . . . . . 419
5.7.7 Basic Level Reference . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 422
5.7.8 Pitfalls and Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 427
5.8 Grouping plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 427
5.8.1 Grouping options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 430
5.9 Patchplots Library . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 433
5.9.1 Additional Patch Types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 434
5.9.2 Automatic Patch Refinement and Triangulation . . . . . . . . . . . . . . . . . . . . . . 444
5.9.3 Peculiarities of Flat Shading and High Order Patches . . . . . . . . . . . . . . . . . . 446
5.9.4 Drawing Grids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 447
5.10 Polar Axes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 449
6 CONTENTS
5.10.1 Polar Axes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 449

5.10.2 Using Radians instead of Degrees . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 451
5.10.3 Mixing With Cartesian Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . 451
5.10.4 Polar Descriptions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 452
5.10.5 Special Polar Plot Types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 453
5.10.6 Partial Polar Axes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 454
5.11 Smith Charts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 455
5.11.1 Smith Chart Axes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 456
5.11.2 Size Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 458
5.11.3 Working with Prepared Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 463
5.11.4 Appearance Control and Styles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 464
5.11.5 Controlling Arcs and Their Stop Points . . . . . . . . . . . . . . . . . . . . . . . . . . 466
5.12 Statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 468
5.12.1 Box Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 469
5.12.2 Histograms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 480
5.13 Ternary Diagrams . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 486
5.13.1 Ternary Axis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 487
5.13.2 Tieline Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 496
5.14 Units in Labels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 498
5.14.1 Preset SI prefixes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 500
6 Memory and Speed considerations 502

6.1 Memory Limits of TEX . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 502
6.2 Memory Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 503
6.2.1 LuaLaTEX . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 503
6.2.2 MikTEX . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 503
6.2.3 TEXLive or similar installations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 504
6.3 Reducing Typesetting Time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 505
6.3.1 LUA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 505
6.3.2 Compiling Images Just Once . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 507
7 Import/Export From Other Formats 508

7.1 Export to pdf/eps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 508
7.1.1 Using the Automatic Externalization Framework of Tik Z . . . . . . . . . . . . . . . . 508
7.1.2 Using the Externalization Framework of PGF By Hand . . . . . . . . . . . . . . . . . 514
7.2 Importing From Matlab . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 515
7.2.1 Importing Mesh Data From Matlab To PGFPlots . . . . . . . . . . . . . . . . . . . . 515
7.2.2 matlab2pgfplots.m . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 516
7.2.3 matlab2pgfplots.sh . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 516
7.2.4 Importing Colormaps From Matlab . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 516
7.3 SVG Output . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 516
7.4 Generate pgfplots Graphics Within Python . . . . . . . . . . . . . . . . . . . . . . . . . . . 517
8 Utilities and Basic Level Commands 518

8.1 Utility Commands . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 518
8.2 Commands Inside Of PGFPlots Axes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 521
8.3 Path Operations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 521
8.4 Specifying Basic Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 522
8.5 Accessing Axis Limits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 526
8.6 Accessing Point Coordinate Values . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 526
8.7 Layer Access . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 528
Index 529
Chapter 1
Introduction
This package provides tools to generate plots and labeled axes easily. It draws normal plots, logplots and
semi-logplots, in two and three dimensions. Axis ticks, labels, legends (in case of multiple plots) can be added
with key-value options. It supports line plots, scatter plots, piecewise constant plots, bar plots, area plots,
mesh– and surface plots, patch plots, contour plots, quiver plots, histogram plots, box plots, polar axes,
ternary diagrams, smith charts and some more. It can cycle through a set of predefined line/marker/color
specifications.
In summary, its purpose is to simplify the generation of high-quality function and/or data plots, and
solving the problems of
• consistency of document and font type and font size,
• direct use of TEX math mode in axis descriptions,
• consistency of data and figures (no third party tool necessary),

• inter-document consistency using preamble configurations and styles.
Although not necessary, separate .pdf or .eps graphics can be generated using the external library devel-
oped as part of Tik Z.
You are invited to use pgfplots for visualization of medium sized data sets in two and three dimensions.
It is based on Till Tantau’s package pgf/Tik Z.
7
Chapter 2
About pgfplots: Preliminaries
This section contains information about upgrades, the team, the installation (in case you need to do it
manually) and troubleshooting. You may skip it completely except for the upgrade remarks.
pgfplots is built completely on Tik Z/pgf. Knowledge of Tik Z will simplify the work with pgfplots,
although it is not required.
However, note that this library requires at least pgf version 2.10. At the time of this writing, many
TEX-distributions still contain the older pgf version 1.18, so it may be necessary to install a recent pgf
prior to using pgfplots.
2.1 Components
pgfplots comes with two components:
1. the plotting component (which you are currently reading) and
2. the PgfplotsTable component which simplifies number formatting and postprocessing of numerical
tables. It comes as a separate package and has its own manual pgfplotstable.pdf.
2.2 Upgrade remarks

This release provides a lot of improvements which can be found in all detail in ChangeLog for interested
readers. However, some attention is useful with respect to the following changes.
2.2.1 New Optional Features

pgfplots has been written with backwards compatibility in mind: old TEX files should compile without
modifications and without changes in the appearance. However, new features occasionally lead to a di↵erent
behavior. In such a case, pgfplots will deactivate the new feature1 .
Any new features or bugfixes which cause backwards compatibility problems need to be activated man-
ually and explicitly. In order to do so, you should use
\usepackage{pgfplots}
\pgfplotsset{compat=1.13}
in your preamble. This will configure the compatibility layer.

You should have at least compat=1.3. The suggested value is printed to the .log file after running TEX.
Here is a list of changes which are inactive unless one uses a suitable compat level:
1. pgfplots 1.13 repairs axis labels in polar axis and ensures that the color chosen by shader=flat is
independent of z buffer and mesh/ordering. Furthermore, it enables stack negative=separate
for all stacked bar plots. Older compatibility levels are present to keep workarounds by end-users.
2. pgfplots 1.12 activates lua backend and defines boxplot/estimator=Excel.
3. pgfplots 1.11 changes the axis cs: it is now the default coordinate system. If you write \draw (1,2)
-- (2,2); pgfplots will automatically treat it as \draw (axis cs:1,2) -- (axis cs:2,2);.
1 In case of broken backwards compatibility, we apologize – and ask you to submit a bug report. We will take care of it.
8
2.2. UPGRADE REMARKS 9
4. pgfplots 1.10 has no di↵erences to 1.9 with respect to compatibility.

5. pgfplots 1.9 comes with a preset to combine ybar stacked and nodes near coords. Furthermore,
it suppresses empty increments in stacked bar plots. In order to activate the new preset, you have to
use compat=1.9 or higher.
6. pgfplots 1.8 comes with a new revision for alignment of label- and tick scale label alignment. Fur-
thermore, it improves the bounding box for hide axis. This revision is enabled with compat=1.8 or
higher.
The configuration compat=1.8 is nessecary to repair axis lines=center in three–dimensional axes.
7. pgfplots 1.7 added new options for bar widths defined in terms of axis units. These are enabled with
compat=1.7 or higher.
8. pgfplots 1.6 added new options for more accurate scaling and more scaling options for \addplot3
graphics. These are enabled with compat=1.6 or higher.
9. pgfplots 1.5.1 interpretes circle- and ellipse radii as pgfplots coordinates (older versions used
pgf unit vectors which have no direct relation to pgfplots). In other words: starting with
version 1.5.1, it is possible to write \draw circle[radius=5] inside of an axis. This requires
\pgfplotsset{compat=1.5.1} or higher.
Without this compatibility setting, circles and ellipses use low–level canvas units of pgf as in earlier
versions.
10. pgfplots 1.5 uses log origin=0 as default (which influences logarithmic bar plots or stacked loga-
rithmic plots). Older versions keep log origin=infty. This requires \pgfplotsset{compat=1.5} or
higher.
11. pgfplots 1.4 has fixed several smaller bugs which might produce di↵erences of about 1–2pt compared
to earlier releases. This requires \pgfplotsset{compat=1.4} or higher.
12. pgfplots 1.3 comes with user interface improvements. The technical distinction between “behavior
options” and “style options” of older versions is no longer necessary (although still fully supported).
This is always activated.
13. pgfplots 1.3 has a new feature which allows to move axis labels tight to tick labels automatically.
This is strongly recommended. It requires \pgfplotsset{compat=1.3} or higher.
Since this a↵ects the spacing, it is not enabled be default.
14. pgfplots 1.3 supports reversed axes. It is no longer necessary to use workarounds with negative
units.
Take a look at the x dir=reverse key.
Existing workarounds will still function properly. Use \pgfplotsset{compat=1.3} or higher together
with x dir=reverse to switch to the new version.
2.2.2 Old Features Which May Need Attention

1. The scatter/classes feature produces proper legends as of version 1.3. This may change the appear-
ance of existing legends of plots with scatter/classes.
2. Starting with pgfplots 1.1, \tikzstyle should no longer be used to set pgfplots options.
Although \tikzstyle is still supported for some older pgfplots options, you should replace any occu-
rance of \tikzstyle with \pgfplotsset{hstyle namei/.style={hkey-value-listi}} or the associated
/.append style variant. See Section 4.18 for more detail.
I apologize for any inconvenience caused by these changes.

10 CHAPTER 2. ABOUT PGFPLOTS: PRELIMINARIES
/pgfplots/compat=1.13|1.12|1.11|1.10|1.9|1.8|1.7|1.6|1.5.1|1.5|1.4|1.3|pre 1.3|default (initially

default)
The preamble configuration
allows to choose between backwards compatibility and most recent features.

Occasionally, you might want to use di↵erent versions in the same document. Then, provide
\begin{figure}
...
\caption{...}
\end{figure}
in order to restrict the compatibility setting to the actual context (in this case, the figure environment).
See the output of your .log file to get a suggested value for compat.
Use \pgfplotsset{compat=default} to restore the factory settings.
Although typically unnecessary, it is also possible to activate only selected changes and keep compati-
bility to older versions in general:
/pgfplots/compat/path replacement=hversioni
/pgfplots/compat/labels=hversioni
/pgfplots/compat/scaling=hversioni
/pgfplots/compat/scale mode=hversioni
/pgfplots/compat/empty line=hversioni
/pgfplots/compat/plot3graphics=hversioni
/pgfplots/compat/bar nodes=hversioni
/pgfplots/compat/BB=hversioni
/pgfplots/compat/bar width by units=hversioni
/pgfplots/compat/pgfpoint substitution=hversioni
/pgfplots/compat/general=hversioni
Let us assume that we have a document with \pgfplotsset{compat=1.3} and you want to keep
it this way.
In addition, you realized that version 1.5.1 supports circles and ellipses. Then, use
% Preamble: \pgfplotsset{width=7cm,compat=1.13}
5 % preamble:
\pgfplotsset{compat=1.3,compat/path replacement=1.5.1}
\begin{tikzpicture}
2 \begin{axis}[
extra x ticks={-2,2},
extra y ticks={-2,2},
0
extra tick style={grid=major}]
\addplot {x};
2 \draw (axis cs:0,0) circle[radius=2];
\end{axis}
\end{tikzpicture}
5
6 4 2 0 2 4 6
All of these keys accept the possible values of the compat key.
The compat/path replacement key controls how radii of circles and ellipses are interpreted.
The compat/labels key controls how axis labels are aligned: either uses adjacent to ticks or with
an absolute o↵set. As of 1.8, it also enables an entirely new revision of the axis label styles. In
most cases, you will see no di↵erence – but it repairs axis lines=center in three–dimensional
axes.
The compat/scaling key controls some bugfixes introduced in version 1.4 and 1.6: they might
introduce slight scaling di↵erences in order to improve the accuracy.
2.3. THE TEAM 11
The compat/plot3graphics controls new features for \addplot3 graphics.

The compat/scale mode allows to enable/disable the warning “The content of your 3d axis has
CHANGED compared to previous versions” because the axis equal and unit vector ratio
features where broken for all versions before 1.6 and have been fixed in 1.6.
The compat/empty line allows to write empty lines into input files in order to generate a jump.
This requires compat=1.4 or newer. See empty line for details.
The compat/BB changes to bounding box to be tight even in case of hide axis.
The compat/bar width by units allows to express bar width=1 (i.e. in terms of axis units).
The compat/bar nodes activates presets for ybar stacked and nodes near coords. In addition,
it enables stacked ignores zero for stacked bar plots.
The compat/general key controls log origin, lua backend, enable tick line clipping, and
boxplot/estimator.
The compat/pgfpoint substitution key determines if axis cs is the default coordinate system
(as of 1.11).
The detailed e↵ects can be seen on the beginning of this section.
The value hversioni can be default, a version number, and newest. The value default is the same
as pre 1.3 (up to insignificant changes). The use of newest is strongly discouraged : it might cause
changes in your document, depending on the current version of pgfplots. Please inspect your .log
file to see suggestions for the best possible version.
2.3 The Team

pgfplots has been written mainly by Christian Feuersnger with many improvements of Pascal Wolkotte
and Nick Papior Andersen as a spare time project. We hope it is useful and provides valuable plots.
If you are interested in writing something but don’t know how, consider reading the auxiliary manual
TeX-programming-notes.pdf which comes with pgfplots. It is far from complete, but maybe it is a good
starting point (at least for more literature).
2.4 Acknowledgements
I thank God for all hours of enjoyed programming. I thank Pascal Wolkotte and Nick Papior Andersen for
their programming e↵orts and contributions as part of the development team. I thank Jürnjakob Dugge
for his contribution of hist/density, matlab scripts for \addplot3 graphics, excellent user forum help
and helpful bug reports. I thank Stefan Tibus, who contributed the plot shell feature. I thank Tom
Cashman for the contribution of the reverse legend feature. Special thanks go to Stefan Pinnow for his
continuous e↵orts to test pgfplots, to discuss requirements, to request features and bug fixes which lead to
numerous quality improvements, and to adapt and integrate the colorbrewer library. Furthermore, I thank
Prof. Schweitzer for many fruitful discussions and his initial encouragement to start such a package. Thanks
to Dr. Meine for his ideas and suggestions. Special thanks go to Markus Böhning for proof-reading all the
manuals of pgf, pgfplots, and PgfplotsTable. Thanks to Vincent A. Traag for bringing colorbrewer
colors to pgfplots. Thanks as well to the many international contributors who provided feature requests
or identified bugs or simply improvements of the manual!
Last but not least, I thank Till Tantau and Mark Wibrow for their excellent graphics (and more) package
pgf and Tik Z, which is the base of pgfplots.
2.5 Installation and Prerequisites

2.5.1 Licensing
This program is free software: you can redistribute it and/or modify it under the terms of the GNU General
Public License as published by the Free Software Foundation, either version 3 of the License, or (at your
option) any later version.
This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; with-
out even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
See the GNU General Public License for more details.

A copy of the GNU General Public License can be found in the package file
doc/latex/pgfplots/gpl-3.0.txt
You may also visit http://www.gnu.org/licenses.
2.5.2 Prerequisites
pgfplots requires pgf. You should generally use the most recent stable version of PGF. pgfplots is used
with
\pgfplotsset{compat=yourversion}
in your preamble (see Section 4.1 for information about how to use it with ConTEXt and plain TEX).
The compat=hyourversioni entry should be added to activate new features, see the documentation of the
compat key for more details.
There are several ways how to teach TEX where to find the files. Choose the option which fits your needs
best.
2.5.3 Installation in Windows

Windows users often use MikTEX which downloads the latest stable package versions automatically. You do
not need to install anything manually here.
However, MikTEX provides a feature to install packages locally in its own TEX-Directory-Structure (TDS).
This is the preferred way if you like to install newer version than those of MikTEX. The basic idea is to
unzip pgfplots in a directory of your choice and configure the MikTEX Package Manager to use this specific
directory with higher priority than its default paths. If you want to do this, start the MikTEX Settings using
“Start Programs MikTEX Settings”. There, use the “Roots” menu section. It contains the MikTEX
Package directory as initial configuration. Use “Add” to select the directory in which the unzipped pgfplots
tree resides. Then, move the newly added path to the list’s top using the “Up” button. Then press “Ok”.
For MikTEX 2.8, you may need to uncheck the “Show MikTEX-maintained root directories” button to see
the newly installed path.
MikTEX complains if the provided directory is not TDS conform (see Section 2.5.6 for details), so you
can’t provide a wrong directory here. This method does also work for other packages, but some packages
may need some directory restructuring before MikTEX accepts them.
2.5.4 Installation of Linux Packages

At the time of this writing, I am unaware of pgfplots packages for recent stable Linux distributions. For
Ubuntu, there are unofficial Ubuntu Package Repositories which can be added to the Ubuntu Package Tools.
The idea is: add a simple URL to the Ubuntu Package Tool, run update and the installation takes place
automatically. These URLs are maintained as PPA on Ubuntu Servers.
The pgfplots download area on sourceforge contains recent links about Ubuntu Package Repositories, go
to http://sourceforge.net/projects/pgfplots/files and download the readme files with recent links.
2.5.5 Installation in Any Directory - the TEXINPUTS Variable

You can simply install pgfplots anywhere on your harddrive, for example into
/foo/bar/pgfplots.
Then, you set the TEXINPUTS variable to
TEXINPUTS=/foo/bar/pgfplots//:
The trailing ‘:’ tells TEX to check the default search paths after /foo/bar/pgfplots. The double slash ‘//’
tells TEX to search all subdirectories.
If the TEXINPUTS variable already contains something, you can append the line above to the existing
TEXINPUTS content.
Furthermore, you should set TEXDOCS as well,
2.6. TROUBLESHOOTING – ERROR MESSAGES 13
TEXDOCS=/foo/bar/pgfplots//:
so that the TEX-documentation system finds the files pgfplots.pdf and pgfplotstable.pdf (on some
systems, it is then enough to use texdoc pgfplots).
Starting with pgfplots 1.12, you may also need to adopt LUAINPUTS:
LUAINPUTS=/foo/bar/pgfplots//:
should usually do the job.
Please refer to your operating systems manual for how to set environment variables.
2.5.6 Installation Into a Local TDS Compliant texmf-Directory

pgfplots comes in a “TEX Directory Structure” (TDS) conforming directory structure, so you can simply
unpack the files into a directory which is searched by TEX automatically. Such directories are ~/texmf on
Linux systems, for example.
Copy pgfplots to a local texmf directory like ~/texmf. You need at least the pgfplots directories
tex/generic/pgfplots and tex/latex/pgfplots. Then, run texhash (or some equivalent path-updating
command specific to your TEX distribution).
The TDS consists of several sub directories which are searched separately, depending on what has been
requested: the sub directories doc/latex/hpackagei are used for (LATEX) documentation, the sub-directories
doc/generic/hpackagei for documentation which apply to LATEX and other TEX dialects (like plain TEX
and ConTEXt which have their own, respective sub-directories) as well.
Similarly, the tex/latex/hpackagei sub-directories are searched whenever LATEX packages are requested.
The tex/generic/hpackagei sub-directories are searched for packages which work for LATEX and other TEX
dialects.
Do not forget to run texhash.
2.5.7 Installation If Everything Else Fails...

If TEX still doesn’t find your files, you can copy all .sty and all .code.tex-files (perhaps all .def files
as well) into your current project’s working directory. In fact, you need everything which is in the
tex/latex/pgfplots and tex/generic/pgfplots sub directories.
Please refer to http://www.ctan.org/installationadvice/ for more information about package in-
stallation.
2.6 Troubleshooting – Error Messages

This section discusses some problems which may occur when using pgfplots. Some of the error messages
are shown in the index, take a look at the end of this manual (under “Errors”).
2.6.1 Problems with available Dimen-registers

To avoid problems with the many required TEX-registers for pgf and pgfplots, you may want to include
\usepackage{etex}
as first package. This avoids problems with “no room for a new dimen” in most cases. It should work with
any modern installation of TEX (it activates the e-TEX extensions).
2.6.2 Dimension Too Large Errors

The core mathematical engine of pgf relies on TEX registers to perform fast arithmetics. To compute
50 + 299, it actually computes 50pt+299pt and strips the pt suffix of the result. Since TEX registers can
only contain numbers up to ±16384, overflow error messages like “Dimension too large” occur if the result
leaves the allowed range. Normally, this should never happen – pgfplots uses a floating point unit with
data range ±10324 and performs all mappings automatically. However, there are some cases where this fails.
Some of these cases are:
1. The axis range (for example, for x) becomes relatively small. It’s no matter if you have absolutely small
ranges like [10 17 , 10 16 ]. But if you have an axis range like [1.99999999, 2], where a lot of significant
digits are necessary, this may be problematic.
I guess I can’t help here: you may need to prepare the data somehow before pgfplots processes it.
2. This may happen as well if you only view a very small portion of the data range.
This happens, for example, if your input data ranges from x 2 [0, 106 ], and you say xmax=10.
Consider using the restrict x to domain*=hmini:hmax i key in such a case, where the hmini and
hmax i should be (say) four times of your axis limits (see page 370 for details).
3. The axis equal key will be confused if x and y have a very di↵erent scale.
4. You may have found a bug – please contact the developers.
2.6.3 Restrictions for DVI-Viewers and dvipdfm

pgf is compatible with
• latex/dvips,
• latex/dvipdfm,
• pdflatex,
.
• ..
However, there are some restrictions: I don’t know any DVI viewer which is capable of viewing the output
of pgf (and therefor pgfplots as well). After all, DVI has never been designed to draw something di↵erent
than text and horizontal/vertical lines. You will need to view the postscript file or the pdf-file.
Then, the DVI/pdf combination doesn’t support all types of shadings (for example, the shader=interp
is only available for dvips, pdftex, dvipdfmx, and xetex drivers).
Furthermore, pgf needs to know a driver so that the DVI file can be converted to the desired output.
Depending on your system, you need the following options:
• latex/dvips does not need anything special because dvips is the default driver if you invoke latex.
• pdflatex will also work directly because pdflatex will be detected automatically.
• latex/dvipdfm requires to use
\def\pgfsysdriver{pgfsys-dvipdfm.def}
%\def\pgfsysdriver{pgfsys-pdftex.def}
%\def\pgfsysdriver{pgfsys-dvips.def}
%\def\pgfsysdriver{pgfsys-dvipdfmx.def}
%\def\pgfsysdriver{pgfsys-xetex.def}
\usepackage{pgfplots}.
The uncommented commands could be used to set other drivers explicitly.

Please read the corresponding sections in [6, Section 7.2.1 and 7.2.2] if you have further questions. These
sections also contain limitations of particular drivers.
The choice which won’t produce any problems at all is pdflatex.
2.6.4 Problems with TEX’s Memory Capacities

pgfplots can handle small up to medium sized plots. However, TEX has never been designed for data plots
– you will eventually face the problem of small memory capacities. See Section 6.1 for how to enlarge them.
2.6. TROUBLESHOOTING – ERROR MESSAGES 15
2.6.5 Problems with Language Settings and Active Characters

Both pgf and pgfplots use a lot of active characters – which may lead to incompatibilities with other
packages which define active characters. Compatibility is better than in earlier versions, but may still be
an issue. The manual compiles with the babel package for english and french, the german package does
also work. If you experience any trouble, let me know. Sometimes it may work to disable active characters
temporarily (babel provides such a command).
2.6.6 Other Problems

Please read the mailing list at
http://sourceforge.net/projects/pgfplots/support.
Perhaps someone has also encountered your problem before, and maybe he came up with a solution.
Please write a note on the mailing list if you have a di↵erent problem. In case it is necessary to contact
the authors directly, consider the addresses shown on the title page of this document.
Chapter 3
Step–by–Step Tutorials
3.1 Introduction
Visualization of data is often necessary and convenient in order to analyze and communicate results of
research, theses, or perhaps just results.
pgfplots is a visualization tool. The motivation for pgfplots is that you as end–user provide the data
and the descriptions as input, and pgfplots takes care of rest such as choosing suitable scaling factors,
scaling to a prescribed target dimension, choosing a good displayed range, assigning tick positions, drawing
an axis with descriptions placed at appropriate places.
pgfplots is a solution for an old problem of visualization in LaTeX: its descriptions use the same fonts
as the embedding text, with exactly the same font sizes. Its direct embedding in LaTeX makes the use
of LaTeX’s powerful math mode as easy as possible: for any kind of axis descriptions up to user–defined
annotations. It features document–wide line–styles, color schemes, markers... all that makes up consistency.
pgfplots o↵ers high–quality. At the same time, it is an embedded solution: it is largely independent of
3rd party–tools, although it features import functions to benefit from available tools.
Its main goal is: you provide your data and your descriptions – and pgfplots runs without more input.
If you want, you can customize what you want.
3.2 Solving a Real Use–Case: Function Visualization

In this section, we assume that you want to visualize two functions. The first function is given by means of
a data table. The second function is given by means of a math expression. We would like to place the two
results side–by–side, and we would like to have “proper” alignment (whatever that means).
As motivated, we have one data table. Let us assume that it is as shown below.
x_0 f(x)
# some comment line
3.16693000e-05 -4.00001451e+00
1.00816962e-03 -3.08781504e+00
1.98466995e-03 -2.88058811e+00
2.96117027e-03 -2.75205040e+00
3.93767059e-03 -2.65736805e+00
4.91417091e-03 -2.58181091e+00
5.89067124e-03 -2.51862689e+00
....
9.89226496e-01 2.29825980e+00
9.90202997e-01 2.33403276e+00
9.91179497e-01 2.37306821e+00
9.92155997e-01 2.41609413e+00
9.93132498e-01 2.46412019e+00
9.94108998e-01 2.51860712e+00
9.95085498e-01 2.58178769e+00
9.96061999e-01 2.65733975e+00
9.97038499e-01 2.75201383e+00
9.98014999e-01 2.88053559e+00
9.98991500e-01 3.08771757e+00
9.99968000e-01 3.99755546e+00
Note that parts of the data file have been omitted here because it is a bit lengthy. The data file
16
3.2. SOLVING A REAL USE–CASE: FUNCTION VISUALIZATION 17
(and all others referenced in this manual) are shipped with pgfplots; you can find them in the subfolder
doc/latex/pgfplots/plotdata.
3.2.1 Getting the Data Into TEX

Our first step is to get the data table into pgfplots. In addition, we want axis descriptions for the x and
y axes and a title on top of the plot.
Our first version looks like
Inv. cum. normal %\documentclass{article}
%\usepackage{pgfplots}
4 %\pgfplotsset{compat=1.5}
%\begin{document}
2
\begin{tikzpicture}
\begin{axis}[
0
y
title=Inv. cum. normal,

xlabel={$x$},
2 ylabel={$y$},
]
\addplot[blue] table {invcum.dat};
4 \end{axis}
\end{tikzpicture}
0 0.2 0.4 0.6 0.8 1 %\end{document}
x
The code listing already shows a couple of important aspects:

1. As usual in LATEX, you include the package using \usepackage{pgfplots}.
2. Not so common is \pgfplotsset{compat=1.5} .
A statement like this should always be used in order to (a) benefit from a more-or-less recent feature
set and (b) avoid changes to your picture if you recompile it with a later version of pgfplots.
Note that pgfplots will generate some suggested value into your log file (since 1.6.1). The minimum
suggested version is \pgfplotsset{compat=1.3} as this has great e↵ect on the positioning of axis
labels.
3. pgfplots relies on Tik Z and pgf. You can say it is a “third party package” on top of Tik Z/pgf.
Consequently, we have to write each pgfplots graph into a Tik Z picture, hence the picture environ-
ment given by \begin{tikzpicture} ... \end{tikzpicture}.
4. Each axis in pgfplots is written into a separate environment. In our case, we chose \begin{axis}
... \end{axis} as this is the environment for a normal axis.
There are more axis environments (like the \begin{loglogaxis} ... \end{loglogaxis} environ-
ment for logarithmic axes).
Although pgfplots runs with default options, it accepts keys. Lots of keys. Typically, you provide
all keys which you “want to have” in square brackets “somewhere” and ignore all other keys.
Of course, the main difficulty is to get an overview over the available keys and to find out how to use
them. This reference manual and especially its Section 4 has been designed for online browsing: it
contains hundreds of cross-referenced examples. Opening the manual in a pdf viewer and searching it
for keywords will hopefully jump to a good match from which you can jump to the reference section
(for example about tick labels, tick positions, plot handlers etc). It is (and will always be) the most
reliable source of detail information about all keys.
Speaking about the reference manual: note that most pdf viewers also have a function to “jump back
to the page before you clicked on a hyperlink” (for Acrobat Reader, open the menu View / Toolbars
/ More Tools and activate the “Previous View” and “Next View” buttons which are under “Page
Navigation Toolbar”).
Note that the code listing contains two sets of keys: the first is after \begin{axis}[ .... ] and
the second right after \addplot[...]. Note furthermore that the option list after the axis has been
18 CHAPTER 3. STEP–BY–STEP TUTORIALS
indented: each option is on a separate line, and each line has a tab stop as first character. This is a
good practice. Another good practice is to place a comma after the last option (in our case, after the
value for ylabel). This allows to add more keys easily – and you won’t forget the comma. It does
not hurt at all. The second “set” of keys after \addplot shows that indentation and trailing comma
a really just a best-practice: we simply said \addplot[blue], meaning that the plot will be placed in
blue color, without any plot mark. Of course, once another option would be added here, it would be
best to switch to indentation and trailing comma:
\addplot[
blue,
mark=*,
]
table {invcum.dat};
5. Inside of an axis, pgfplots accepts an \addplot ... ; statement (note the final semicolon).
In our case, we use \addplot table: it loads a table from a file and plots the first two columns.
There are, however, more input methods. The most important available inputs methods are \addplot
expression (which samples from some mathematical expression) and \addplot table (loads data
from tables), and a combination of both which is also supported by \addplot table (loads data from
tables and applies mathematical expressions). Besides those tools which rely only on builtin methods,
there is also an option to calculated data using external tools: \addplot gnuplot which uses gnuplot
as “desktop calculator” and imports numerical data, \addplot shell (which can load table data from
any system call), and the special \addplot graphics tool which loads an image together with meta
data and draws only the associated axis.
In our axis, we find a couple of tokens: the first is the mandatory \addplot token. It “starts” a further
plot. The second is the option list for that plot, which is delimited by square brackets (see also the
notes about best-practices above). The name “option list” indicates that this list can be empty. It
can also be omitted completely in which case pgfplots will choose an option list from its current
cycle list (more about that in a di↵erent lecture). The next token is the keyword “table”. It tells
pgfplots that table data follows. The keyword “table” also accepts an option list (for example, to
choose columns, to define a di↵erent col sep or row sep or to provide some math expression which
is applied to each row). More on that in a di↵erent lecture. The next token is {invcum.dat}: an
argument in curly braces which provides the table data. This argument is interpreted by “plot table”.
Other input types would expect di↵erent types of arguments. In our case, the curly braces contain a
file name. Plot table expects either a file name as in our case or a so-called “inline table”. An inline
table means that you would simply insert the contents of your file inside of the curly braces. In our
case, the table is too long to be inserted into the argument, so we place it into a separate file. Finally,
the last (mandatory!) token is a semicolon. It terminates the \addplot statement.
6. Axis descriptions can be added using the keys title, xlabel, ylabel as we have in our example
listing.
pgfplots accepts lots of keys — and sometimes it is the art of finding just the one that you were
looking for. Hopefully, a search through the table of contents of the reference manual and/or a keyword
search through the entire reference manual will show a hit.
3.2.2 Fine–Tuning of the First Picture

While looking at the result of Section 3.2.1, we decide that we want to change something. First, we decide
that the open ends on the left and on the right are disturbing (perhaps we have a strange taste – or perhaps
we know in advance that the underlying function is not limited to any interval). Anyway, we would like to
show it only in the y interval from 3 to +3.
We can do so as follows:
Inv. cum. normal \begin{tikzpicture}
\begin{axis}[
2 xlabel={$x$},
ylabel={$y$},
ymin=-3, ymax=3,
minor y tick num=1,
]
0
y

\end{axis}
\end{tikzpicture}
0 0.2 0.4 0.6 0.8 1

x
We added three more options to the option list of the axis. The first pair is ymin=-3 and ymax=3. Note
that we have placed them on the same line although we said the each should be on a separate line. Line
breaks are really optional; and in this case, the two options appear to belong together. They define the
display limits. Display limits define the “window” of the axis. Note that any \addplot statements might
have more data (as in our case). They would still generate graphics for their complete set of data points!
The keys ymin,ymax,xmin,xmax control only the visible part, i.e. the axis range. Everything else is clipped
away (by default). The third new option is minor y tick num=1 which allows to customize minor ticks.
Note that minor ticks are only displayed if the major ticks have the same distance as in our example.
Note that we could also have modified the width and/or height of the figure (the keys have these
names). We could also have used one of the predefined styles like tiny or small in order to modify not
just the graphics, but also use di↵erent fonts for the descriptions. We could also have chosen to adjust the
unspecified limits: either by fixing them explicitly (as we did for y above) or by modifying the enlargelimits
key (for example using enlargelimits=false).
We are now satisfied with the first picture and we would like to add the second one.
3.2.3 Adding the Second Picture with a Di↵erent Plot

As motivated, our goal is to have two separate axes placed side–by–side. The second axis should show
a function given as math expression. More precisely, we want to show the density function of a normal
distribution here (which is just a special math expression).
We simply start a new tikzpicture and insert a new axis environment (perhaps by copy-pasting our
existing one). The \addplot command is di↵erent, though:
400 \begin{tikzpicture}
\begin{axis}[
]
300 % density of Normal distribution:
\addplot[
red,
200 domain=-3e-3:3e-3,
samples=201,
]
100 {exp(-x^2 / (2e-3^2)) / (1e-3 * sqrt(2*pi))};
\end{axis}
\end{tikzpicture}
2 0 2
3
·10
We see that it has an axis environment with an empty option list. This is quite acceptable: after all, it
is to be expected that we will add options eventually. Even if we don’t: it does not hurt. Then, we find the
expected \addplot statement. As already explained, \addplot statements initiate a new plot. It is followed
by an (optional) option list, then by some keyword which identifies the way input coordinates are provided,
then arguments, and finally a semicolon. In our case, we find an option list which results in a red plot.
The two keys domain and samples control how our math expression is to be evaluated: domain defines the
sampling interval in the form a:b and samples=N expects the number of samples inserted into the sampling
interval. Note that domain merely controls which samples are taken; it is independent of the displayed axis
range (and both can di↵er significantly). If the keyword defining how coordinates are provided is missing,
pgfplots assumes that the next argument is a math expression. Consequently, the first token after the
option list is a math expression in curly braces. We entered the density function of a normal distribution
here (compare Wikipedia).
Note that the axis has an axis multiplier: the x tick labels have been chosen to be 2, 0, and 2 and an
extra x tick scale label of the form ‘·10 3 ’. These tick scale labels are quite convenient are are automatically
deduced from the input data. We will see an example with the e↵ects of scaled x ticks=false at the end
of this tutorial.
Inside of the math expression, you can use a lot of math functions like exp, sin, cos, sqrt, you
can use exponents using the a^b syntax, and the sampling variable is x by default. Note, however,
that trigonometric functions operate on degrees by default! If you need to sample the sinus func-
tion, you can use \addplot[domain=0:360] {sin(x)};. This is quite uncommon. You can also use
\addplot[domain=0:2*pi] {sin(deg(x)};. This samples radians (which is more common). But since
the math parser expects degrees, we have to convert x to degrees first using the deg() function. See also
trig format plots=rad. The math parser is written in TEX(it does not need any third-party tool). It
supports the full range of a double precision number, even though the accuracy is about that of a single
precision number. This is typically more than sufficient to sample any function accurately. If you ever en-
counter difficulties with precision, you can still resort to \addplot gnuplot in order to invoke the external
tool gnuplot as “coordinate calculator”.
The experienced reader might wonder about constant math expressions domain=-3e-3:3e-3, 2e-3^2,
and 1e-3 rather than some variable name like “mu” or ‘sigma’. This is actually a matter of taste: both is
supported and we will switch to variable names in the next listing.
The main part of our step here is still to be done: we wanted to place two figures side–by–side. This can
be done as follows:
Inv. cum. normal
400
2
0 200
y
2
0
0 0.2 0.4 0.6 0.8 1 2 0 2
x ·10 3
\begin{tikzpicture}
\begin{axis}[
xlabel={$x$},
ylabel={$y$},
ymin=-3, ymax=3,
minor y tick num=1,
]
\end{axis}
\end{tikzpicture}% - avoid white space
%
\hskip 10pt % insert a non-breaking space of specified width.
%
\begin{tikzpicture}
\begin{axis}[
]
% density of Normal distribution:
\newcommand\MU{0}
\newcommand\SIGMA{1e-3}
\addplot[
red,
domain=-3*\SIGMA:3*\SIGMA,
samples=201,
]
{exp(-(x-\MU)^2 / 2 / \SIGMA^2) / (\SIGMA * sqrt(2*pi))};
\end{axis}
\end{tikzpicture}
The listing above shows the two separate picture environments: the first is simply taken as-is from the
previous step and the second is new. Note that both are simply placed adjacent to each other: we only
inserted comment signs to separate them. This approach to place graphics side–by–side is common in TEX:
it works for \includegraphics in the same way. You could, for example, write
\includegraphics{image1}%
%
%
\includegraphics{image2}
to place two graphics next to each other. This here is just the same (except that our graphics occupy
more code in the .tex file).
Note that there is also a comment sign after \end{tikzpicture}. This is not just a best-practice: it
is necessary to suppress spurious spaces! In TEX, every newline character is automatically converted to a
white space (unless you have an empty line, of course). In our case, we want no white spaces.
In our second picture, we see the e↵ects of switch our math expression to constant definitions as promised
earlier. The interesting part starts with two constants which are defined by means of two \newcommand s:
we define \MU to be 0 and \SIGMA to be 1e-3. This is one way to define constants (note that such a
definition of constants should probably introduce round braces if numbers are negative, i.e. something like
\newcommand\negative{(-4)}).
3.2.4 Fixing the Vertical Alignment and Adjusting Tick Label Positions
Note that even though our individual pictures look quite good, the combination of both is not properly
aligned. The experienced reader identifies the weak point immediately: the bounding box of the two images
di↵ers, and they are aligned at their baseline (which is the bottom edge of the picture). In particular,
the xlabel=$x$ of the left picture and the automatically inserted scaling label \cdot 10^{-3} of the right
picture cause an unwanted vertical shift. We want to fix that in the next step.
Besides the bad alignment, we find it a little bit misleading that the axis descriptions of the second
picture are between both pictures. We would like to move them to the right.
Let us present the result first:
Inv. cum. normal
400
2
0 200
y
2
0
0 0.2 0.4 0.6 0.8 1 2 0 2
x 3
·10
\begin{tikzpicture}[baseline]
\begin{axis}[
xlabel={$x$},
ylabel={$y$},
ymin=-3, ymax=3,
minor y tick num=1,
]
\end{axis}
\end{tikzpicture}%
%
%
\begin{axis}[
yticklabel pos=right,
]
\newcommand\MU{0}
\addplot[red,domain=-3*\SIGMA:3*\SIGMA,samples=201]
{exp(-(x-\MU)^2 / 2 / \SIGMA^2) / (\SIGMA * sqrt(2*pi))};
\end{axis}
\end{tikzpicture}
This listing has a couple of modifications. The most important one is the we added an option list to the
tikzpicture environment: the baseline option. This option shifts the picture up or down such that the
canvas coordinate y = 0 is aligned at the baseline of the surrounding text. In pgfplots, the y = 0 line is
the lower edge of the box. This simple feature allows both axes to be aligned vertically: now, their boxes
are aligned rather than the lower edges of their bounding boxes. The option baseline needs to be provided
to all pictures for which this shifting should be done – in our case, to all which are to be placed in one row.
Keep in mind that it is an option for \begin{tikzpicture}.
The second change is rather simple: we only added the option yticklabel pos=right to the second
axis. This moves all tick labels to the right, without changing anything else.
Note that there is much more to say about alignment and bounding box control. After all, we did not
really change the bounding box - we simply moved the pictures up or down. There is also the use-case where
we want horizontal alignment: for example if the two pictures should be centered horizontally or if they should
be aligned with the left- and right end of the margins. The associated keys \begin{tikzpicture}[trim axis
left, trim axis right] and \centering are beyond the scope of this tutorial, please refer to Section 4.19
for details.
3.2.5 Satisfying Di↵erent Tastes

We are now in a position where the figures as such are in a good shape.
However, an increase in knowledge will naturally lead to an increase in questions. Some of these questions
will be part of other how-to lectures. But the most commonly asked questions are addressed here (feel free
to email some more if you believe that I should include another hotspot):
3
1. How can I get rid of that 10 -label?
2. How can I modify the number printing?
3. How can I have one single line per axis rather than a box?
This here gives brief hints where to look in this reference manual for more details. We modify the
appearance of the second picture according to the questions above:
\begin{tikzpicture}
\begin{axis}[
300 axis lines=left,
scaled ticks=false,
xticklabel style={
rotate=90,
200 anchor=east,
/pgf/number format/precision=3,
/pgf/number format/fixed,
100 /pgf/number format/fixed zerofill,
},
]
0.002
0.000
0.002
\newcommand\MU{0}
\addplot[red,domain=-3*\SIGMA:3*\SIGMA,samples=201]
{exp(-(x-\MU)^2 / 2 / \SIGMA^2)
/ (\SIGMA * sqrt(2*pi))};
\end{axis}
\end{tikzpicture}
The appearance of the axes as such can be controlled by means of the axis lines key. It accepts
the values left, right, box, center, none (and also top, bottom, middle which are aliases). The
xticklabel style key modifies a predefined style (note the use of indentation here!). A style is a collection
of keys which are applied in a specific context. Styles are very useful and are widely used by pgfplots. In
our case, we adjust a couple of options like rotation, alignment (the anchor option), and number printing
options. The precise details of these individual options is beyond the scope of this tutorial. The keys actually
belong to Tik Z - and the Tik Z manual is the reference for these keys (although pgfplots also covers most
of the topics). The complete set of number printing options is available in both the Tik Z manual [6] and the
manual for PgfplotsTable which is shipped with pgfplots. A brief extract can be found in Section 4.13.
3.2.6 Finishing Touches: Automatic Generation of Individual Pdf Graphics

As last step in this lecture, I would like to talk about one technical topic. Typically, a TEXdocument starts
quite simple: a little bit of text, perhaps one or two pictures. But they tend to grow. And eventually, you
will encounter one of the weak points of pgfplots: the graphics are involved and TEXconsumes a lot of
time to generate them. Especially if it keeps regenerating them even though they did not change at all. The
fact that we need to rerun the pdflatex processor all the time makes things worse.
Fortunately, there are solutions. A simple solution is: why can’t we write each individual graphics into
a separate .pdf file and use \includegraphics to include it!? The answer is: yes, we can. And it is
surprisingly simple to do so.
In order to convert every tikzpicture environment automatically to an external graphics without chang-
ing any line of code in the TEXfile, we can simply write the following two lines into the document’s preamble:
\usepgfplotslibrary{external}
\tikzexternalize
...
\begin{document}
...
\end{document}
But now, we have to provide a command-line switch to pdflatex :

pdflatex -shell-escape myfile.tex
This works out-of-the box with pdflatex. If you use latex/dvips, lualatex, dvipdfm or any other
TEXderivate, you need to modify the option \tikzexternalize[system call=....] (which is, unfortu-
nately, system-dependent, especially for the postscript variants).
It might be too much to discuss how to define individual file names or how to modify the file name
generation strategy. There is also the \tikzexternalize[mode=list and make] feature which generates
a GNU Make file to allow \label/\ref to things inside of the external graphics and which supports the
generation of all images in parallel (if you have a multi-core PC).
Details of the external library can be found in Section 7.1 (but only a brief survey) and, in all depth,
in the Tik Z reference manual [6].
3.2.7 Summary
We learned how to create a standard axis, and how to assign basic axis descriptions. We also saw how to
plot functions from a data table (in our case a tab-separated file, but other delimiters as in csv files are also
supported) and from math expressions. We saw that pgfplots does a reasonable good job at creating a
fully-featured axis automatically (like scaling the units properly, choosing tick positions and labels). We also
learned how to improve vertical alignment and how to customize the appearance of an axis.
Next steps might be how to draw multiple plots into the same axis, how to employ scatter plots of
pgfplots, how to generate logarithmic axes, or how to draw functions of two variables. Some of these
aspects will be part of further how-to lectures.
3.3 Solving a Real Use–Case: Scientific Data Analysis

In this section, we assume that you did some scientific experiment. The scientific experiment yielded three
input data tables: one table for each involved parameter d = 2, d = 3, d = 4. The data tables contain “degrees
of freedom” and some accuracy measurement “l2_err”. In addition, they might contain some meta-data
(in our case a column “level”). For example, the data table for d = 2 might be stored in data_d2.dat and
may contain
dof l2_err level
5 8.312e-02 2
17 2.547e-02 3
49 7.407e-03 4
129 2.102e-03 5
321 5.874e-04 6
769 1.623e-04 7
1793 4.442e-05 8
4097 1.207e-05 9
9217 3.261e-06 10
The other two tables are similar, we provide them here to simplify the reproduction of the examples.
The table for d = 3 is stored in data_d3.dat, it is
dof l2_err level
7 8.472e-02 2
31 3.044e-02 3
111 1.022e-02 4
351 3.303e-03 5
1023 1.039e-03 6
2815 3.196e-04 7
7423 9.658e-05 8
18943 2.873e-05 9
47103 8.437e-06 10
Finally, the last table is data_d4.dat

dof l2_err level
9 7.881e-02 2
49 3.243e-02 3
209 1.232e-02 4
769 4.454e-03 5
2561 1.551e-03 6
7937 5.236e-04 7
23297 1.723e-04 8
65537 5.545e-05 9
178177 1.751e-05 10
What we want is to produce three plots, each dof versus l2_err, in a loglog plot. We expect that
the result is a line in a loglog plot, and we are interested in its slope log e(N ) = a log(N ) because that
characterizes our experiment.
3.3. SOLVING A REAL USE–CASE: SCIENTIFIC DATA ANALYSIS 25
3.3.1 Getting the Data into TeX

Our first step is to get one of our data tables into pgfplots. In addition, we want axis descriptions for the
x and y axes and a title on top of the plot.
Our first version looks like
Convergence Plot %\documentclass{article}
%\usepackage{pgfplots}
1
10 %\pgfplotsset{compat=1.5}
2 %\begin{document}
10
L2 Error
\begin{tikzpicture}
3
10 \begin{loglogaxis}[
title=Convergence Plot,
4 xlabel={Degrees of freedom},
10 ylabel={$L_2$ Error},
]
5
10 \addplot table {data_d2.dat};
\end{loglogaxis}
\end{tikzpicture}
101 102 103 104
%\end{document}
Degrees of freedom
Our example is similar to that of the lecture in Section 3.2.1 in that it defines some basic axis descriptions
by means of title, xlabel, and ylabel and provides data using \addplot table. The only di↵erence is
that we used \begin{loglogaxis} instead of \begin{axis} in order to configure logarithmic scales on both
axes. Note furthermore that we omitted any options after \addplot. As explained in Section 3.2.1, this tells
pgfplots to consult its cycle list to determine a suitable option list.
3.3.2 Adding the Remaining Data Files of Our Example.

pgfplots accepts more than one \addplot ... ; command – so we can just add our remaining data files:
Convergence Plot \begin{tikzpicture}
\begin{loglogaxis}[
1
10 title=Convergence Plot,
xlabel={Degrees of freedom},
10 ]
L2 Error
\addplot table {data_d2.dat};

3
4 \end{loglogaxis}
10 \end{tikzpicture}
5
10
101 102 103 104 105

Degrees of freedom
You might wonder how pgfplots chose the di↵erent line styles. And you might wonder how to modify
them. Well, if you simply write \addplot without options in square brackets, pgfplots will automatically
choose styles for that specific plot. Here “automatically” means that it will consult its current cycle list:
a list of predefined styles such that every \addplot statement receives one of these styles. This list is
customizable to a high degree.
Instead of the cycle list, you can easily provide style options manually. If you write
\addplot[hoptionsi] ...,
pgfplots will only use hoptionsi and will ignore its cycle list. If you write a plus sign before the
square brackets as in
\addplot+[hoptionsi] ...,
pgfplots will append hoptionsi to the automatically assigned cycle list.
3.3.3 Add a Legend and a Grid

A legend is a text label explaining what the plots are. A legend can be provided for one or more \addplot
statements using the legend entries key:
Convergence Plot \begin{tikzpicture}
\begin{loglogaxis}[
1
10 d=2 title=Convergence Plot,
xlabel={Degrees of freedom},
d=3
10 d=4 grid=major,
L2 Error
legend entries={$d=2$,$d=3$,$d=4$},
3
10 ]
\end{loglogaxis}
5
101 102 103 104 105

Degrees of freedom
Here, we assigned a comma-separated list of text labels, one for each of our \addplot instruc-
tions. Note the use of math mode in the text labels. Note that if any of your labels contains a
comma, you have to surround the entry by curly braces. For example, we could have used legend
entries={{$d=2$},{$d=3$},{$d=4$}} — pgfplots uses these braces to delimit arguments and strips
them afterwards (this holds for any option, by the way).
Our example also contains grid lines for which we used the grid=major key. It activates major grid lines
in all axes.
You might wonder how the text labels map to \addplot instructions. Well, they are mapped by index.
The first label is assigned to the first plot, the second label to the second plot and so on. You can exclude
plots from this counting if you add the forget plot option to the plot (using \addplot+[forget plot],
for example). Such plots are excluded from both cycle lists and legends.
3.3.4 Add a Selected Fit-line

Occasionally, one needs to compute linear regression lines through input samples. Let us assume that we
want to compute a fit line for the data in our fourth data table (data_d4.dat). However, we assume that
the interesting part of the plot happens if the number of degrees of freedom reaches some asymptotic limit
(i.e. is very large). Consequently, we want to assign a high uncertainty to the first points when computing
the fit line.
pgfplots o↵ers to combine table input and mathematical expressions (note that you can also type pure
mathematic expressions, although this is beyond the scope of this example). In our case, we employ this
feature to create a completely new column - the linear regression line:
3.3. SOLVING A REAL USE–CASE: SCIENTIFIC DATA ANALYSIS 27
Convergence Plot %\usepackage{pgfplotstable}
%...
100 d=2 \begin{tikzpicture}
\begin{loglogaxis}[
d=3
d=4 xlabel={Degrees of freedom},
L2 Error
10 grid=major,
]
10
\addplot table[
10 6 x=dof,
101 102 103 104 105 y={create col/linear regression={y=l2_err,
variance list={1000,800,600,500,400,200,100}}}]
Degrees of freedom {data_d4.dat};
\end{loglogaxis}
\end{tikzpicture}
Note that we added a further package: pgfplotstable. It allows to postprocess tables (among other
things. It also has a powerful table typesetting toolbox which rounds and formats numbers based on your
input CSV file).
Here, we added a fourth plot to our axis. The first plot is also an \addplot table statement as before
– and we see that it loads the data file data_d4.dat just like the plot before. However, it has special keys
which control the coordinate input: x=dof means to load x coordinates from the column named “dof”. This
is essentially the same as in all of our other plots (because the “dof” column is the first column). It also
uses y={create col/.......}. This lengthy statement defines a completely new column. The create
col/linear regression prefix is a key which can be used whenever new table columns can be generated.
As soon as the table is queried for the first time, the statement is evaluated and then used for all subsequent
rows. The argument list for create col/linear regression contains the column name for the function
values y=l2_err which are to be used for the regression line (the x arguments are deduced from x=dof as
you guessed correctly). The variance list option is optional. We use it to assign variances (uncertainties)
to the first input points. More precisely: the first encountered data point receives a variance of 1000, the
second 800, the third 600, and so on. The number of variances does not need to match up with the number
of points; pgfplots simply matches them with the first encountered coordinates.
Note that since our legend entries key contains only three values, the regression line has no legend
entry. We could easily add one, if we wanted. We can also use \addplot+[forget plot] table[....] to
explicitly suppress the generation of a legend as mentioned above.
Whenever pgfplots encounteres mathematical expressions, it uses its built-in floating point unit. Con-
sequently, it has a very high data range – and a reasonable precision as well.
3.3.5 Add an Annotation using Tik Z: a Slope Triangle

Often, data requires interpretation – and you may want to highlight particular items in your plots. This
“highlight particular items” requires to draw into an axis, and it requires a high degree of flexibility. Users
of Tik Z would say that Tik Z is a natural choice – and it is.
In our use-case, we are interested in slopes. We may want to compare slopes of di↵erent experiments.
And we may want to show selected absolute values of slopes.
Here, we use Tik Z to add custom annotations into a pgfplots axis. We choose a particular type of a
custom annotation: we want to mark two points on a line plot. One way to do so would be to determine the
exact coordinates and to place a graphical element at this coordinate (which is possible using \draw ....
(1e4,1e-5) ... ). Another (probably simpler) way is to use the pos feature to identify a position “25%
after the line started”.
Based on the result of Section 3.3.4, we find
Convergence Plot %\usepackage{pgfplotstable}
%...
\begin{loglogaxis}[
d=3
1.15 d=4 xlabel={Degrees of freedom},
L2 Error
10 grid=major,
]
10
\addplot table[
10 6 x=dof,
101 102 103 104 105 y={create col/linear regression={y=l2_err,
variance list={1000,800,600,500,400,200,100}}}]
Degrees of freedom {data_d4.dat}
% save two points on the regression line
% for drawing the slope triangle
coordinate [pos=0.25] (A)
coordinate [pos=0.4] (B)
;
% save the slope parameter:
\xdef\slope{\pgfplotstableregressiona}
% draw the opposite and adjacent sides

% of the triangle
\draw (A) -| (B)
node [pos=0.75,anchor=west]
{\pgfmathprintnumber{\slope}};
\end{loglogaxis}
\end{tikzpicture}
The example is already quite involved since we added complexity in every step. Before we dive into the
details, let us take a look at a simpler example:
1
\begin{loglogaxis}
2 \addplot table {data_d2.dat}
10
coordinate [pos=0.25] (A)
coordinate [pos=0.4] (B)
3
10 ;
4 Special. \draw[-stealth] (A) -| (B);

10
\node[pin=0:Special.] at (769,1.623e-04) {};
5
10
\end{loglogaxis}
\end{tikzpicture}
101 102 103 104
Here, we see two annotation concepts o↵ered by pgfplots : the first is to insert drawing commands
right after an \addplot command (but before the closing semicolon). The second is to add standard Tik Z
commands, but use designated pgfplots coordinates. Both are Tik Z concepts. The first is what we want
here: we want to identify two coordinates which are “somewhere” on the line. In our case, we define two
named coordinates: coordinate A at 25% of the line and coordinate B at 40% of the line. Then, we use
\draw (A) -| (B) to draw a triangle between these two points. The second is only useful if we know some
absolute coordinates in advance.
Coming back to our initial approach with the regression line, we see that it uses the first con-
cept: it introduces named coordinates after \addplot, but before the closing semicolon. The statement
\xdef\slope introduces a new macro. It contains the (expanded due to the “eXpanded DEFinition”) value
of \pgfplotstableregressiona which is the slope of the regression line. In addition to the slope triangle,
we also add a node in which we typeset that value using \pgfmathprintnumber.
Note that the example above is actual a “happy case”: it can happen easily that labels which are added
inside of the axis environment are clipped away:
3.4. USE–CASES INVOLVING SCATTER PLOTS 29
1 % Preamble: \pgfplotsset{width=7cm,compat=1.13}
10
10 2 \begin{tikzpicture}
10
4
tiny,
10
5
]
10
Special.
\addplot table {data_d2.dat}
101 102 103 104 node [pos=1,pin=0:Special.] {}
;
\end{loglogaxis}
\end{tikzpicture}
The example above combines the pos label placement with the node’s label. Note that the small style
tiny installs a pgfplots preset which is better suited for very small plots – it is one of the many supported
scaling parameters. The problem here is apparent: the text of our extra node is clipped away. Depending
on your data, you have a couple of solutions here:
• use clip=false to disable clipping of plot paths at all,

• use clip mode=individual to enable clipping only for plot paths,
• draw the node outside of the axis environment but inside of the picture environment.
The first attempt works quite well for most figures:
10
10
4
tiny,
10
5
clip=false,
10
Special. ]
101 102 103 104 \addplot table {data_d2.dat}
node [pos=1,pin=0:Special.] {}
;
\end{loglogaxis}
\end{tikzpicture}
Note that this approach in which the nodes are placed before the closing semicolon implies that nodes
inherit the axis line style and color.
3.3.6 Summary
We learned how to define a (logarithmic) axis, and how to assign basic axis descriptions. We also saw once
more how to use one or more\addplot table commands to load table data into pgfplots . We took a
brief look into regression and Tik Z drawing annotations.
We also encountered the tiny style which is one of the ways to customize the size of an axis. Others
are width, height, the other predefined size styles like normalsize, small, or footnotesize, and the two
di↵erent scaling modes /pgfplots/scale and /tikz/scale (the first scales only the axis, the second also
the labels).
Next steps might be how to visualize functions using line plots, how to align adjacent graphics properly
(even if the axis descriptions vary), how to employ scatter plots of pgfplots , or how to draw functions of
two variables.
3.4 Use–Cases involving Scatter Plots

Assuming that we are more–or–less familiar with the basics of the preceding tutorials, we would like to draw
a scatter plot. A scatter plot is one in which markers indicate the important information.
There are many di↵erent kinds of scatter plots and this section covers a couple of them.
3.4.1 Scatter Plot Use–case A

In this subsection, we address the following scatter plot use–case: assume that we are given a couple of
special (x, y) coordinates along with color data at every vertex. We would like to draw markers at the
positions and choose individual colors depending on the color data.
Importing the Data File

We assume that our input data is given as a table containing much more columns than we need. The first
couple of rows are as follows:
# ordering = colwise , basis = BASIS _ HAT _ HIER , number points =1657 ,
# ( max ) level ( s ): (7 ,8) , domain =[0 ,1] x [0 ,6.2831853]
x _0 x _1 f(x) l _0 i _0
0 0 0 0 0
0 0.024543693 -0.00054758064 0 0
0 0.049087385 -0.0021903226 0 0
0 0.073631078 -0.00054758064 0 0
0 0.09817477 -0.0095006552 0 0
What we need is the first and second column to get the x and y coordinate values, respectively, and the
third column f(x) to choose color values. The color values are very small and have a high range: there are
values of order 10 6 and there are values of order 1. Such ranges are best shown on a logarithmic scale, which
is why we will resort to some logarithmic scale on the absolute values of this column. Thus, a requirement
will be to accept a math expression (involving logs) on the color data column.
Note that the data file (and all others referenced in this manual) are shipped with pgfplots; you can
find them in the subfolder doc/latex/pgfplots/plotdata.
We learned already how to read table data from a file, so our first step is relatively straight–forward.
\begin{axis}
\addplot+[only marks] table
{concat_VV_together_grid.dat};
4 \end{axis}
\end{tikzpicture}
0
0 0.2 0.4 0.6 0.8 1
Here, the only non–trivial variation is the option only marks which is given after the plus sign. Keep in
mind that \addplot+[hoptionsi] means that pgfplots shall combine the set of options of its cycle list
with hoptionsi. In our case, only marks does what it says. The only marks plot handler is the most simple
scatter plot: it uses the same color for every marker.
Note that \addplot table takes the first column as x and the second as y (which matches our input file
perfectly).
Fine Tuning
We agree that our initial import has unsuitable displayed limits: there is too much white space around
the interesting plot area. In addition, the markers overlap because they are too large. We can modify the
appearance as follows:
3.4. USE–CASES INVOLVING SCATTER PLOTS 31
6
\begin{tikzpicture}
\begin{axis}[
enlargelimits=false,
]
4 \addplot+[only marks,mark size=0.6pt]
table {concat_VV_together_grid.dat};
\end{axis}
\end{tikzpicture}
2
0
0 0.2 0.4 0.6 0.8 1
As before, we assume that we add more options after \begin{axis}. Consequently, we introduced
suitable indentation and a trailing comma after the option. Note that enlargelimits is typically active;
it means that pgfplots increases the displayed range by 10% by default. Deactivating it produces tight
limits according to the input data.
Our second option is mark size – using an absolute size (about the radius or half size of the marker).
Color Coding According To Input Data

We are quite close to our goal, except for the colors. As discussed, our input file contains three columns and
the third one should be used to provide color information. In our case, the data file has a column named
f(x).
0
6
4 2
2 4
0 6
0 0.2 0.4 0.6 0.8 1
\begin{tikzpicture}
\begin{axis}[
colorbar,
]
\addplot+[
only marks,
scatter,
point meta={ln(1e-6+abs(\thisrow{f(x)}))/ln(10)},
mark size=0.6pt,
]
table
\end{axis}
\end{tikzpicture}
We added a couple of options to our example: the options scatter, and point meta, colorbar. The
option scatter has a slightly misleading name as we already had a scatter plot before we added that option.
It activates scatter plots with individual appearance: without further options, it chooses individual colors for
every marker. The “individual colors” are based on something which is called “point meta” in pgfplots.
The point meta is typically a scalar value for every input coordinate. In the default configuration , it
is interpreted as “color data” for the coordinate in question. This also explains the other option: point
meta=... tells pgfplots which values are to be used to determine colors. Note that the default value
of point meta is to use the y coordinate. In our case, we have a complicated math expression which is
related to our input file: it contains small quantities in column f(x) which are based shown on a logarithmic
scale as their di↵er over a huge range. Since a logarithm must not have a non-positive argument, we have
10 6 + abs(· · · ) as expression which ensures that the argument is never smaller than 10^{-6} and that is
is positive. The divider /ln(10) means that we have logarithms base 10. But the key point of the whole
complicated expression can be summarized as follows:
1. We can use \thisrow{hcolumn namei} to refer to table columns. Here, “this row” means to evaluate
the table for the “data point which is being read from the current row”.
2. We can combine \thisrow with any complicated math expression.
The third new option colorbar activates the color bar on the right hand side (as you guessed correctly).
We see that the smallest value is 6 which corresponds to our value 1e-6 in the math expression.
You might wonder how a scalar value (the number stored in the f(x) column) results in a color. pgfplots
computes the minimum and maximum value of all such numbers. Then, it maps every number into a
colormap. A colormap defines a couple of colors and interpolates linearly between such colors. That means
that the smallest value of point meta is mapped to the first color in a colormap whereas the largest value of
point meta is mapped to the last color in the colormap. All others are mapped to something in–between.
More information about colormap and point meta can be found in Section 4.7.6 and in Section 4.8.
3.4.2 Scatter Plot Use–case B

As already mentioned, there are various use–cases for scatter plots. The default configuration of the scatter
key is to read numeric values of point meta and choose colors by mapping that value into the current
colormap.
A di↵erent application would be to expect symbolic input (some string) and choose di↵erent markers
depending on that input symbol.
Suppose that you are given a sequence of input coordinates of the form (x, y) hclass label i and that you
want to choose marker options depending on the hclass label i. A pgfplots solution could be
\begin{tikzpicture}
\begin{axis}
0.4 \addplot[
scatter,
only marks,
point meta=explicit symbolic,
scatter/classes={
0.2
a={mark=square*,blue},%
b={mark=triangle*,red},%
c={mark=o,draw=black}},
]
table[meta=label] {
0 0.2 0.4 0.6 0.8 x y label
0.1 0.15 a
0.45 0.27 c
0.02 0.17 a
0.06 0.1 a
0.9 0.5 b
0.5 0.3 c
0.85 0.52 b
0.12 0.05 a
0.73 0.45 b
0.53 0.25 c
0.76 0.5 b
0.55 0.32 c
};
\end{axis}
\end{tikzpicture}
As in our previous use–case in Section 3.4.1, we have the options scatter, only marks, and a configura-
tion how to retrieve the point meta values by means of the meta key. One new key is point meta=explicit
symbolic: it tells pgfplots that any encountered values of point meta are to be interpreted as string
symbols. Furthermore, it tells pgfplots that the every input coordinate comes with an explicit value (as
opposed to a common math expression, for example). The other di↵erent option is scatter/classes. As
3.5. SOLVING A REAL USE–CASE: FUNCTIONS OF TWO VARIABLES 33
you guessed from the listing, it is a map from string symbol to marker option list. This allows to address
such use–cases in a simple way.
This example has actually been replicated from the reference manual section for scatter/classes.
3.4.3 Scatter Plot Use–case C

Finally, this tutorial sketches a further use–case for scatter plots: given a sequence of coordinates (x, y) with
individual string labels, we want to draw the string label at the designated positions.
This can be implemented by means of the nodes near coords feature of pgfplots, which is actually
based on scatter:
0.6 \begin{axis}[
enlargelimits=0.2,
]
Y4 \addplot+[nodes near coords,only marks,
0.4 point meta=explicit symbolic]
table[meta=label] {
x y label
1 0.5 0.2 1
0.2
0.2 0.1 t2
t2 5 0.7 0.6 3
0.35 0.4 Y4
0 0.65 0.1 5
0.2 0.4 0.6 0.8 };
\end{axis}
\end{tikzpicture}
In this case, we have point meta=explicit symbolic in order to express the fact that our labels are of
textual form (see the reference manual section for nodes near coords for applications of numeric labels).
The remaining stu↵ is done by the implementation of nodes near coords. Note that enlarged the axis
limits somewhat in order to include the text nodes in the visible area.
There is much more to say about scatter plots, and about nodes near coords. Please consider this
subsection as a brief pointer to Section 4.5.12 in the reference manual.
3.4.4 Summary
We learned how to generate scatter plots with single color using only marks, scatter plots with individually
colored markers using the scatter key, scatter plots with specific marker styles depending on some class
label using scatter/classes and text nodes using nodes near coords.
Furthermore, we introduced the concept of “point meta data”: once as (scalar valued) color data, once
as symbolic class label and once as text label.
There is much more to say, especially about point meta which is introduced and explained in all depth
in Section 4.8.
There is also more to say about scatter plots, for example how to generate scatter plots with individually
sized markers and/or colors (by relying on \pgfplotspointmetatransformed, see the reference manual
section for visualization depends on). In addition, scatter plots can be customized to a high degree
which is explained in Section 4.5.12.
3.5 Solving a Real Use–Case: Functions of Two Variables

In this tutorial, we assume that we have two functions for which we seek a plot: the first is a sampled
function given by a huge data file and the second is the math expression g(x, y) = exp( x2 y 2 ) · x.
Our first function actually consists of two data files: the first file contains some scattered data which is
used to discretized (“sample”) a function and the second file contains data for the function as such. Our
requirement here is two draw two graphs into the same axis: one in which the function is plotted as a smooth,
colored surface and one in which the scattered data file should be on top of the surface because it provides
more detail how the function was represented in the computer.
The second function which is given as math expression should be visualized using a contour plot.
3.5.1 Surface Plot from Data File

Our first step is to load the data file and to plot a surface.
Clearly, functions of two variables require a more sophisticated input format: they are typically sampled
on a unified grid with n ⇥ m points, i.e. n points for x and m points for y, resulting in a total of matrix
with n · m values fij = f (xi , yi ). How can we read matrix data? And what if you have more than just the
z value? A standard way is to write the matrix to a table, either in row–wise ordering or in column–wise
ordering (both are common).
Here, we assume that our function values are written to a table in which the y values vary from line to
line. Here is an extract of the data file (which is too large to list it here):
# ordering = colwise , number points =1089 ,
# ( max ) level ( s ): (5 ,5) , domain =[0 ,1] x [0 ,6.2831853]
x _0 x _1 f(x) l _0 i _0
0 0 0 0 0
0 0.19634954 0.038239818 0 0
0 0.39269908 0.14662768 0 0
..
.
0 5.8904862 0.14662768 0 0
0 6.0868358 0.038239818 0 0
0 6.2831853 6.9388939 e -18 0 0
0.03125 0 3.0517578 e -05 5 1

0.03125 0.19634954 0.030093496 5 1
0.03125 0.39269908 0.1146012 5 1
0.03125 0.58904862 0.24129102 5 1
0.03125 0.78539816 0.38946059 5 1
0.03125 0.9817477 0.53949733 5 1
..
.
Note that the data file (and all others referenced in this manual) are shipped with pgfplots; you can
find them in the subfolder doc/latex/pgfplots/plotdata.
The input file contains x0 , x1 , and f (x0 , x1 ) in columns named x_0, x_1, and f(x), respectively. In
addition, it contains some meta data which is irrelevant for us here.
Note that our input file contains empty lines whenever x_0 changes. This is a common data format which
simplifies the detection of “scanline length”. A scanline is one line in the input matrix, for example the line
consisting of all points with x0 = 0. With such scanlines, pgfplots can automatically deduce the size of
the input matrix.
In order to plot the file as a surface, we proceed as in the previous example by using \addplot table.
However, we have to use \addplot3 to indicate that a three–dimensional result is expected:
\begin{tikzpicture}
\begin{axis}
\addplot3[surf,mesh/ordering=y varies]
table {concat_VV_together.dat};
1 \end{axis}
\end{tikzpicture}
0.5
6
0 4
0 0.2 0.4 2
0.6 0.8
1 0
The example looks familiar compared to our results of the preceding tutorials: a tikzpicture environ-
ment containing an axis environment and the mentioned \addplot3 command. The option list contains
surf, which tells pgfplots how to visualize the input data. The key mesh/ordering=y varies tells pgf-
plots how to decode the input matrix. This is important; otherwise pgfplots would have chosen x varies
which does not match our file.
Note that we there is no need to configure either mesh/rows=hN i or mesh/cols=hN i here because these
parameters are automatically deduced from the scan line lengths marked by empty lines in our input file.
Since our \addplot3 table statement does not contain any hints which columns should be plotted,
pgfplots simply plots the first three columns against each other.
The colors of a surf plot are chosen from the function values (unless you configure some other value for
point meta; this is similar to the scatter plot example). In case of a function of two variables, the function
value is the third column.
3.5.2 Fine–Tuning
In order to stress how colors are to be mapped to values, we add a color bar. In addition, we rotate the view
a little bit and add axis labels. Furthermore, we would like to have a smooth color mapping.
We end up at
\begin{tikzpicture}
\begin{axis}[
view/h=40,
1 colorbar horizontal,
xlabel=$x$, ylabel=$y$,
]
0.5 \addplot3[surf,mesh/ordering=y varies,
shader=interp]
0 \end{axis}
0 6 \end{tikzpicture}
4
0.5 2
1 0 y
x
0 0.2 0.4 0.6 0.8 1
Here, view/h rotates the “horizontal” parts of the view (only). It chooses a new view angle for the
orthographic projection. As you guessed, there is also a view/v key and a view={hhi}{hv i} variant.
The key colorbar horizontal is a style which activates a colorbar and configures it to be displayed
horizontally. The labels are placed using xlabel and ylabel as we saw it before for visualizations of
one–dimensional functions. A Colorbar uses the current colormap and adds axis descriptions to show how
values are mapped to colors.
The shader=interp key activates a smooth color interpolation.
3.5.3 Adding Scattered Data on Top of the Surface

As motivated earlier, we have a second data set, one which characterizes how the function has been repre-
sented in some computer simulation. We would like to add the second data set as scatter plot on top of the
function.
The data set as such is the very same as the one used in Section 3.4.1, so we do not need to list it here
again. However, we have to include the two–dimensional scatter data into the three–dimensional axis in a
suitable way. We chose to place it on a fixed z value as follows:
\begin{tikzpicture}
\begin{axis}[
view/h=40,
colorbar horizontal,
1 xlabel=$x$, ylabel=$y$,
]
0.5 \addplot3[surf,mesh/ordering=y varies,
shader=interp]
0
0 6 \addplot3[blue,mark=*,
4 mark options={fill=blue!80!black},
0.5 2 only marks,mark size=0.6pt]
1 0 y
x table[z expr=1.2]
\end{axis}
\end{tikzpicture}
0 0.2 0.4 0.6 0.8 1
Now, we have two \addplot3 table statements in the same axis. None of them uses the cycle list
as we used explicit option lists. The first is our surface plot. Note that it is plotted before the scatter
plot: pgfplots cannot handle depth information between adjacent \addplot statements. It does, however,
handle z buffer information for data of a single \addplot statement. The second plot is our scatter plot:
we recognize only marks and mark size from Section 3.4.1. In addition, we configured some color and
marker options.
An important aspect is \addplot3 table[z expr=1.2] – it tells pgfplots how to choose z values for
the input file (otherwise, pgfplots would have used the third column of that file). This is a convenient way
to insert two–dimensional data into a three–dimensional axis, provided you have table data. There is also a
di↵erent way which works for both tables and math expressions (or other input types). This di↵erent way
is to install a z filter, but that is beyond the scope of this tutorial for now.
3.5.4 Computing a Contour Plot of a Math Expression

This section addresses the second part of our use–case example: a function of two variables given by a math
expression.
Our function of interest is x exp( x2 y 2 ). We start as in our tutorial for one–dimensional functions
given by a math expression (compare Section 3.2.3): by using an \addplot statement which is followed by
a math expression in curly braces. However, we rely on \addplot3 as in the preceding section:
x exp( x2 y2 ) \begin{tikzpicture}
\begin{axis}[
title={$x \exp(-x^2-y^2)$},
small,
]
0.4 \addplot3
0.2 {exp(-x^2-y^2)*x};
0 \end{axis}
\end{tikzpicture}
0.2
5
0.4
4 0
2 0 2 4 y
5
x
Our example contains a basic axis environment with title, xlabel, ylabel and the small key which
are already known from the preceding tutorials. The \addplot3 has no options and is immediately followed
by the math expression. The absence of options tells pgfplots to rely on its cycle list. This, in turn
configures mark=* with blue color – and a line plot. A line plot combined with \addplot3 is of limited use;
it merely connects all incoming points. Since points are sampled as a matrix (line–by–line). Our next step
will be to define a suitable plot handler.
Note, however, that our math expression depends on x and y. These two variables are the sampling
variables of pgfplots in its default configures: both are sampled in the domain of interest using the correct
number of samples. The \addplot3 statement takes care of computing N · M points in the correct sequence
where N is the number of samples for x and M is samples y, the number of samples used for y.
We can see that our sampling domain is too large. Switching to a smaller domain focusses on the
interesting parts of our function:
\begin{axis}[
small,
0.5 ]
\addplot3[
surf,
0 domain=-2:2,
domain y=-1.3:1.3,
]
1 {exp(-x^2-y^2)*x};
0.5 0 \end{axis}
2 1 0 \end{tikzpicture}
1 2 1 y
x
Here, we introduced an option list after \addplot3. Since we provided the option list without the leading
plus sign ‘+’, pgfplots does not consider its cycle list at all (and switches o↵ marks and the default color
settings). We added domain and domain y in order to restrict the sampling domain in a suitable way. If we
would have omitted domain y, the y domain would use the same value as the x domain.
As you might have guessed, the surf key has the main use–case of providing a connection to the previous
tutorial section: it is one of the natural visualizations for functions of two variables. As in the preceding
section, the color has been deduced from the function value z = f (x, y) (more precisely, by relying on the
default configuration point meta=f(x)).
The next step is to switch to contour plots by replacing ‘surf‘ by ‘contour gnuplot’:
\begin{axis}[
0 .4 small,
]
0.4
\addplot3[
0.2 0.2 contour gnuplot,
0 domain=-2:2,
0. 0 domain y=-1.3:1.3,
2
0.
0.2 ]
4
1 {exp(-x^2-y^2)*x};
0.4
0 \end{axis}
1
0 \end{tikzpicture}
1 1 y
x
Now, we have a contour plot – although it is not quite what we had in mind. First, there are so few
contour lines that it is hard to see anything (especially since the line width is too small). Furthermore,
the view direction is unfamiliar.
We add the view option with the argument for “view from top” and configure the number of contour
lines using the contour/number key and the line width using the thick style:
\begin{axis}[
1
enlarge x limits,
0.3 view={0}{90},
0.3
0.5 xlabel=$x$, ylabel=$y$,
0.1
0 .4 small,
0.1
0.
0.1
4
]
0.1
0
0
y
\addplot3[
0.2
0.2
0 .2
0 .2
0. domain=-2:2,
0.5 0. 3
3 domain y=-1.3:1.3,
contour gnuplot={number=14},
1 thick,
0 .1 0.1 ]
2 1 0 1 2 {exp(-x^2-y^2)*x};
x \end{axis}
\end{tikzpicture}
This is what we wanted to achieve. Note that contour gnuplot accepts options which have the key
prefix contour/. In this context, the prefix is optional.
Note that contour gnuplot is di↵erent from almost all other plot handlers of pgfplots with respect to
one aspect: it relies on an external tool to compute coordinates whereas all other pgfplots plot handlers
depend on TEX alone and do not need 3rd party tools. The non–linear algorithm to compute contour lines
is currently unavailable in pgfplots which is stressed by the name ‘contour gnuplot’. Consequently, you
can only reproduce the example if you have gnuplot installed. pgfplots invokes the executable name
‘gnuplot’, i.e. the executable must be on your search path (the PATH environment variable must contain
it). And, more importantly, you have to tell LATEX that it is allowed to launch 3rd party executables while
processing your .tex file. Typically, you have to add the argument -shell-escape to your TEX executable,
i.e. one of
latex -shell-escape htexfilenamei
or
pdflatex -shell-escape htexfilenamei
or
lualatex -shell-escape htexfilenamei
or
xelatex -shell-escape htexfilenamei.
Note that it is occasionally named in a di↵erent way like “-enable-write18”. The interaction with gnuplot
is controlled by means of temporary input- and output files.
Note that contour gnuplot and \addplot gnuplot are two ways to extend the built-in capabilities of
pgfplots by means of gnuplot’s math library, although their use is optional.
3.5.5 Summary
We have sketched how to load a data table containing a sampled function of two variables, and we learned
how to visualize such data as surface plot. We learned how to rotate the view, how to change the color
shader of surface plots, how to enabled colorbars, and how to add scatter plots on top of surface plots.
Furthermore, we encountered the first contour plot as an example for how to sample a function of two
variables by means of builtin methods of pgfplots.
It should be stressed that pgfplots needs no external tool to generate such plots (except for contour
gnuplot which is the only exception): every computer with a decent version of pgfplots can regenerate
these plots.
There is more to say about three–dimensional axes, in particular regarding mesh/ordering, parametric
plots, perhaps line plots in three dimensions or other plot types. Furthermore, there are some limitations
regarding the z buffering, i.e. how pgfplots decides which parts of the figure are in front of others. These
items can be read in Section 4.6 and its sub–sections.
You might also be interested in styles to change the appearance of a three–dimensional axis, compare
Section 4.11.
Chapter 4
The Reference
4.1 TEX-dialects: LATEX, ConTEXt, plain TEX

The starting point for pgfplots is an axis enviroment like axis or the logarithmic variants semilogxaxis,
semilogyaxis or loglogaxis.
Each environment is available for LATEX, ConTEXt and plain TEX:
LATEX: \usepackage{pgfplots} \pgfplotsset{compat=1.13} and
\begin{tikzpicture} \begin{tikzpicture}
\begin{axis} \begin{semilogxaxis}
... ...
\end{axis} \end{semilogxaxis}
\end{tikzpicture} \end{tikzpicture}
Here, the \pgfplotsset{compat=1.13} key should be set to at least version 1.3. Otherwise pgf-
plots assumes that your document has been generated years ago and attempts to run in backwards
compatibility mode as good as it can.
Since LATEX is the default for many people, this manual only shows LATEX examples. A full document
sceleton can be found below this enumeration.
ConTEXt: \usemodule[pgfplots] \pgfplotsset[compat=1.13] and
\starttikzpicture \starttikzpicture
\startaxis \startsemilogxaxis
... ...
\stopaxis \stopsemilogxaxis
\stoptikzpicture \stoptikzpicture
A complete ConTEXt–example file can be found in
doc/context/pgfplots/pgfplotsexample.tex.
plain TEX: \input pgfplots.tex \pgfplotsset{compat=1.13} and
\tikzpicture \tikzpicture
\axis \semilogxaxis
... ...
\endaxis \endsemilogxaxis
\endtikzpicture \endtikzpicture
A complete plain–TEX–example file can be found in
doc/plain/pgfplots/pgfplotsexample.tex.
39
40 CHAPTER 4. THE REFERENCE
For LATEX, a complete example will look somehow like this:

\documentclass[a4paper]{article}
% for dvipdfm:
%\def\pgfsysdriver{pgfsys-dvipdfm.def}
\pgfplotsset{compat=1.6}% <-- moves axis labels near ticklabels (respects tick label widths)
\begin{document}
\begin{figure}
\centering
\begin{tikzpicture}
\begin{loglogaxis}[xlabel=Cost,ylabel=Error]
\addplot coordinates {
(5, 8.31160034e-02)
(17, 2.54685628e-02)
(49, 7.40715288e-03)
(129, 2.10192154e-03)
(321, 5.87352989e-04)
(769, 1.62269942e-04)
(1793, 4.44248889e-05)
(4097, 1.20714122e-05)
(9217, 3.26101452e-06)
};
(7, 8.47178381e-02)
(31, 3.04409349e-02)
(111, 1.02214539e-02)
(351, 3.30346265e-03)
(1023, 1.03886535e-03)
(2815, 3.19646457e-04)
(7423, 9.65789766e-05)
(18943, 2.87339125e-05)
(47103, 8.43749881e-06)
};
\legend{Case 1,Case 2}
\end{loglogaxis}
\end{tikzpicture}
\caption{A larger example}
\end{figure}
\end{document}
If you use latex / dvips or pdflatex, no further modifications are necessary. For dvipdfm, you should
use the \def\pgfsysdriver line as indicated above in the examples (see also Section 2.6.3).
4.2 The Axis-Environments

There is an axis environment for linear scaling, two for semi-logarithmic scaling and one for double-
logarithmic scaling.
\begin{tikzpicture}[hoptions of tikz i]
henvironment contentsi
\end{tikzpicture}
This is the graphics environment of Tik Z. It produces a single picture and encloses also every axis.
Instead of using the environment version, there is also a shortcut command
\tikz{hpicture contenti}
which can be used alternatively.
\begin{axis}[hoptionsi]
\end{axis}
The axis environment for normal plots with linear axis scaling.
The ‘every linear axis’ style key can be modified with
\pgfplotsset{every linear axis/.append style={...}}

4.3. THE \ADDPLOT COMMAND: COORDINATE INPUT 41
to install styles specifically for linear axes. These styles can contain both Tik Z- and pgfplots options.
\begin{semilogxaxis}[hoptionsi]
\end{semilogxaxis}
The axis environment for logarithmic scaling of x and normal scaling of y. Use
\pgfplotsset{every semilogx axis/.append style={...}}
to install styles specifically for the case with xmode=log, ymode=normal.

The logarithmic scaling means to apply the natural logarithm (base e) to each x coordinate. Further-
more, ticks will be typeset as 10hexponenti , see Section 4.13 for more details.
\begin{semilogyaxis}[hoptionsi]
\end{semilogyaxis}
The axis environment for normal scaling of x and logarithmic scaling of y,
The style ‘every semilogy axis’ will be installed for each such plot.
The same remarks as for semilogxaxis apply here as well.
\begin{loglogaxis}[hoptionsi]
\end{loglogaxis}
The axis environment for logarithmic scaling of both, x and y axes. As for the other axis possibilities,
there is a style ‘every loglog axis’ which is installed at the environment’s beginning.
The same remarks as for semilogxaxis apply here as well.
They are all equivalent to

\begin{axis}[
xmode=log|normal,
ymode=log|normal]
...
\end{axis}
with properly set variables ‘xmode’ and ‘ymode’ (see below).
4.3 The \addplot Command: Coordinate Input

1
\begin{tikzpicture}
\begin{axis}[ymin=0,ymax=1,enlargelimits=false]
0.8 \addplot
[blue!80!black,fill=blue,fill opacity=0.5]
coordinates
0.6 {(0,0.1) (0.1,0.15) (0.2,0.5) (0.3,0.62)
(0.4,0.56) (0.5,0.58) (0.6,0.65) (0.7,0.6)
(0.8,0.58) (0.9,0.55) (1,0.52)}
0.4
|- (0,0) -- cycle;
0.2 \addplot
[red,fill=red!90!black,opacity=0.5]
coordinates
0 {(0,0.25) (0.1,0.27) (0.2,0.24) (0.3,0.24)
0 0.2 0.4 0.6 0.8 1 (0.4,0.26) (0.5,0.3) (0.6,0.23) (0.7,0.2)
(0.8,0.15) (0.9,0.1) (1,0.1)}
|- (0,0) -- cycle;
\addplot[green!20!black] coordinates
{(0,0.4) (0.2,0.75) (1,0.75)};
\end{axis}
\end{tikzpicture}
100
4x2 5 \begin{tikzpicture}
\begin{axis}
\addplot+[id=parable,domain=-5:5]
gnuplot{4*x**2 - 5}
node[pin=180:{$4x^2-5$}]{};
50 \end{axis}
\end{tikzpicture}
6 4 2 0 2 4 6
\begin{tikzpicture}
\begin{axis}
\addplot3[surf,domain=0:360,samples=40]
{sin(x)*sin(y)};
1 \end{axis}
\end{tikzpicture}
0
1
0 200
100 200 300 0
0.5
1
0
0
0.5
1
0 200
100 200 300 0
\begin{tikzpicture}
\begin{axis}[colormap/redyellow,colorbar]
\addplot3[surf,
domain=0:360,samples=40]
{sin(x)*sin(y)};
\end{axis}
\end{tikzpicture}
\begin{tikzpicture}
\begin{axis}[view={60}{30}]
0
\addplot3[surf,shader=flat,
samples=20,
0.5 domain=-1:0,y domain=0:2*pi,
z buffer=sort]
({sqrt(1-x^2) * cos(deg(y))},
{sqrt( 1-x^2 ) * sin(deg(y))},
1 x);
\end{axis}
\end{tikzpicture}
0 0.5
0
1 0.5
Inside of an axis environment, the \addplot command is the main user interface. It comes in two variants:
\addplot for two–dimensional visualization and \addplot3 for three–dimensional visualization.
\addplot[hoptionsi] hinput datai htrailing path commandsi;

This is the main plotting command, available within each axis environment. It can be used one or more
times within an axis to add plots to the current axis. There is also an \addplot3 command which is
described in Section 4.6.
It reads point coordinates from one of the available input sources specified by hinput datai, updates
limits, remembers hoptionsi for use in a legend (if any) and applies any necessary coordinate transfor-
mations (or logarithms).
The hoptionsi can be omitted in which case the next entry from the cycle list will be inserted as
hoptionsi. These keys characterize the plot’s type like linear interpolation with sharp plot, smooth
plot, constant interpolation with const plot, bar plot, mesh plots, surface plots or whatever and
define colors, markers and line specifications1 . Plot variants like error bars, the number of samples or
a sample domain can also be configured in hoptionsi.
The hinput datai is one of several coordinate input tools which are described in more detail below.
Finally, if \addplot successfully processed all coordinates from hinput datai, it generates Tik Z paths
to realize the drawing operations. Any htrailing path commandsi are appended to the final drawing
command, allowing to continue the Tik Z path (from the last plot coordinate).
Some more details:
• The style /pgfplots/every axis plot will be installed at the beginning of hoptionsi. That means
you can use
\pgfplotsset{every axis plot/.append style={...}}
to add options to all your plots - maybe to set line widths to thick. Furthermore, if you have more
than one plot inside of an axis, you can also use
\pgfplotsset{every axis plot no 3/.append style={...}}
to modify options for the plot with number 3 only. The first plot in an axis has number 0.
• The hoptionsi are remembered for the legend. They are available as ‘current plot style’ as long
as the path is not yet finished or in associated error bars.
• See Subsection 4.7 for a list of available markers and line styles.
• For log plots, pgfplots will compute the natural logarithm log(·) numerically using a floating
point unit developed for this purpose2 . For example, the following numbers are valid input to
\addplot.
1 In version 1.2.2 and earlier, there was an explicit distinction between “behaviour” options like error bars, domain, number
of samples etc. and “style options” like color, line width, markers etc. This distinction is obsolete now, simply collect everything
into hoptionsi.
2 This floating point unit is available as Tik Z library as part of Tik Z.
10
\begin{loglogaxis}
(769, 1.6227e-04)
4 (1793, 4.4425e-05)
10
(4097, 1.2071e-05)
(9217, 3.2610e-06)
(2.2e5, 2.1E-6)
10 5 (1e6, 0.00003341)
(2.3e7, 0.00131415)
};
\end{loglogaxis}
\end{tikzpicture}
103 104 105 106 107
You can represent arbitrarily small or very large numbers as long as its logarithm can be represented
as a TEX-length (up to about 16384). Of course, any coordinate x  0 is not possible since the
logarithm of a non-positive number is not defined. Such coordinates will be skipped automatically
(using the initial configuration unbounded coords=discard).
• For normal (non–logarithmic) axes, pgfplots applies floating point arithmetics to support large
or small numbers like 0.00000001234 or 1.234 · 1024 . Its number range is much larger than TEX’s
native support for numbers. The relative precision is between 4 and 7 significant decimal digits for
the mantissa.
As soon as the axes limits are completely known, pgfplots applies a transformation which maps
these floating point numbers into TEX-precision using transformations
Tx (x) = 10sx · x ax and Ty (y) = 10sy · y ay and (for 3D plots) Tz (y) = 10sz · z az
with properly chosen integers sx , sy , sz 2 Z and shifts ax , ay , az 2 R. Section 4.25 contains a

description of disabledatascaling and provides more details about the transformation.
• Some of the coordinate input routines use the powerful \pgfmathparse feature of pgf to read their
coordinates, among them plot coordinates, plot expression and plot table. This allows
to use mathematical expressions as coordinates which will be evaluated using the floating point
routines (this applies to logarithmic and linear scales).
• pgfplots automatically computes missing axis limits. The automatic computation of axis limits
works as follows:
1. Every coordinate will be checked. Care has been taken to avoid TEX’s limited numerical
capabilities.
2. Since more than one \addplot command may be used inside of \begin{axis}...\end{axis},
all drawing commands will be postponed until \end{axis}.
\addplot+[hoptionsi] ...;
Does the same like \addplot[hoptionsi] ...; except that hoptionsi are appended to the arguments
which would have been taken for \addplot ... (the element of the default list).
Thus, you can combine cycle list and hoptionsi.
1 1
0 0
1 1
6 4 2 0 2 4 6 6 4 2 0 2 4 6
\begin{tikzpicture}
\begin{axis}
\addplot {sin(deg(x))};
\end{axis}
\end{tikzpicture}
\begin{tikzpicture}
\begin{axis}
\addplot+[only marks] {sin(deg(x))};
\end{axis}
\end{tikzpicture}
The distinction is as follows: \addplot ... (without options) lets pgfplots select colors, markers
and linestyles automatically (using cycle list). The variant \addplot+[hoptioni] ... will use the
same automatically determined styles, but in addition it uses hoptionsi. Finally, \addplot[hoptionsi]
(without the +) uses only the manually provided hoptionsi.
/pgfplots/empty line=auto|none|scanline|jump (initially auto)

Controls how empty lines in the input coordinate stream are to be interpreted. You should ensure that
you have \pgfplotsset{compat=1.4} or newer in your preamble and leave this key at its default empty
line=auto.
Empty lines can occur between the coordinates of \addplot coordinates or successive rows of the
data file input routines \addplot table (and \addplot file).
The choice auto checks if the current plot type is mesh or surf. If so, it uses scanline. If the current
plot type is some other plot type (like a standard line plot), it uses jump. Note that the value auto
for non-mesh plots results in none if compat=1.3 or older is used. In other words: you have to write
\pgfplotsset{compat=1.4} or newer to let pgfplots interpret empty lines as jump in standard line
plots:
Ignored: compat=1.3 Jump: compat=1.4
\begin{tikzpicture}
2 2
\begin{axis}[tiny,
1.5 1.5
title={Ignored: compat=1.3},
1 1 compat=1.3]
0.5 0.5 \addplot table {
0 0 A B
0 0.5 1 1.5 2 0 0.5 1 1.5 2 0 0
1 1
1 2
2 2
};
\end{axis}
\end{tikzpicture}
\begin{tikzpicture}
\begin{axis}[tiny,
title={Jump: compat=1.4},
compat=1.4]
\addplot table {
A B
0 0
1 1
1 2
2 2
};
\end{axis}
\end{tikzpicture}
The choice scanline is only useful for mesh and surf: it is used to decode a matrix from a coordinate
stream. If an empty line occurs once every N data points, the “scanline” length is N . This information,
together with mesh/ordering and the total number of points, allows to deduce the matrix size. However,
the distance between empty lines has to be consistent: if the first two empty lines have a distance of 2
and the next comes after 5, pgfplots will ignore the information and will expect explicit matrix sizes
using mesh/rows and/or mesh/cols. The choice scanline is ignored if mesh input=patches. It has
no e↵ect for other plot types.
The choice none will silently discard any empty line in the input stream.
The choice jump tells pgfplots to generate a jump.
4.3.1 Coordinate Lists

\addplot coordinates {hcoordinate listi};
\addplot[hoptionsi] coordinates {hcoordinate listi} htrailing path commandsi;
\addplot3 . . .
The ‘\addplot coordinates’ command is like that provided by Tik Z and reads its input data from a
sequence of point coordinates, encapsulated in round braces.
\begin{axis}
(0,0)
(0.5,1)
(1,2)
1
};
\end{axis}
\end{tikzpicture}
0
0 0.2 0.4 0.6 0.8 1
You should only use this input format if you have short diagrams and you want to provide mathematical
expressions for each of the involved coordinates. Any data plots are typically easier to handle using a
table format and \addplot table.
The coordinates can be numbers, but they can also contain mathematical expressions like sin(0.5) or
\h*8 (assuming you defined \h somewhere). However, expressions which involve round braces need to
be encapsulated in a further set of curly braces, for example ({sin(0.5)},{cos(0.1)}).
You can also supply error coordinates (reliability bounds) if you are interested in error bars. Simply
append the error coordinates with ‘+- (hex,eyi)’ (or +- (hex,ey,ez i)) to the associated coordinate:
\begin{tikzpicture}
\begin{axis}
4 \addplot+[error bars/.cd,x dir=both,x explicit]
coordinates {
(0,0) +- (0.1,0)
(0.5,1) +- (0.4,0.2)
(1,2)
2 (2,5) +- (1,0.1)
};
\end{axis}
\end{tikzpicture}
0
0 1 2 3
or
(900,1e-6) +- (0.1,0.2)
(2600,5e-7) +- (0.2,0.5)
(4000,7e-8) +- (0.1,0.01)
};
These error coordinates are only used in case of error bars, see Section 4.12. You will also need to
configure whether these values denote absolute or relative errors.
The coordinates as such can be numbers as +5, -1.2345e3, 35.0e2, 0.00000123 or 1e2345e-8. They
are not limited to TEX’s precision.
Furthermore, coordinates allows to define “meta data” for each coordinate. The interpretation of
meta data depends on the visualization technique: for scatter plots, meta data can be used to define
colors or style associations for every point (see page 112 for an example). Meta data (if any) must be
provided after the coordinates and after error bar bounds (if any) in square brackets:
6
·10 % Preamble: \pgfplotsset{width=7cm,compat=1.13}
\begin{tikzpicture}
1 \begin{axis}
\addplot+[scatter,scatter src=explicit] coordinates {
(900,1e-6) [1]
(2600,5e-7) [2]
(4000,7e-8) [3]
};
0.5 \end{axis}
\end{tikzpicture}
0
1,000 2,000 3,000 4,000
Please refer to the documentation of point meta on page 211 for more information about per point
meta data.
The coordinate stream can contain empty lines to tell pgfplots that the function has jumps. To use
it, simply insert an empty line (and ensure that you have \pgfplotsset{compat=1.4} or newer in your
preamble). See the documentation of empty line for details.
/pgfplots/plot coordinates/math parser=true|false (initially true)
Allows to turn o↵ support for mathematical expressions in every coordinate inside of plot coordinates.
This might be necessary if coordinates are not in numerical form (or if you’d like to improve speed).
It is necessary to disable plot coordinates/math parser if you use some sort of symbolic transfor-
mations (i.e. text coordinates).
4.3.2 Reading Coordinates From Tables

\addplot table [hcolumn selectioni]{hfile or inline tablei};
\addplot[hoptionsi] table [hcolumn selectioni]{hfile or inline tablei} htrailing path commandsi;
\addplot3 . . .
This input method is the main input format for any data–based function. It accepts either a file
containing data or an inline table provided in curly braces.
Given a data file like
dof L2 Lmax maxlevel

5 8.31160034e-02 1.80007647e-01 2
17 2.54685628e-02 3.75580565e-02 3
49 7.40715288e-03 1.49212716e-02 4
129 2.10192154e-03 4.23330523e-03 5
321 5.87352989e-04 1.30668515e-03 6
769 1.62269942e-04 3.88658098e-04 7
1793 4.44248889e-05 1.12651668e-04 8
4097 1.20714122e-05 3.20339285e-05 9
9217 3.26101452e-06 8.97617707e-06 10
one may want to plot ‘dof’ versus ‘L2’ or ‘dof’ versus ‘Lmax’. This can be done by
\begin{tikzpicture}
\begin{loglogaxis}[
xlabel=Dof,
ylabel=$L_2$ error]
\addplot table[x=dof,y=L2] {datafile.dat};
\end{loglogaxis}
\end{tikzpicture}
or, for the Lmax column, using

\begin{tikzpicture}
\begin{loglogaxis}[
xlabel=Dof,
ylabel=$L_\infty$ error]
\addplot table[x=dof,y=Lmax] {datafile.dat};
\end{loglogaxis}
\end{tikzpicture}
It is also possible to provide the data inline, i.e. directly as argument in curly braces:
\begin{tikzpicture}
\begin{loglogaxis}[
xlabel=Dof,
ylabel=$L_\infty$ error]
\addplot table[x=dof,y=Lmax] {
dof L2 Lmax maxlevel
5 8.31160034e-02 1.80007647e-01 2
17 2.54685628e-02 3.75580565e-02 3
49 7.40715288e-03 1.49212716e-02 4
129 2.10192154e-03 4.23330523e-03 5
321 5.87352989e-04 1.30668515e-03 6
769 1.62269942e-04 3.88658098e-04 7
1793 4.44248889e-05 1.12651668e-04 8
4097 1.20714122e-05 3.20339285e-05 9
9217 3.26101452e-06 8.97617707e-06 10
};
\end{loglogaxis}
\end{tikzpicture}
Inline table may be convenient together with ‘\\’ and row sep=\\, see below for more information.
Alternatively, you can load the table once into an internal structure and use it multiple times3 :
\pgfplotstableread{datafile.dat}\loadedtable % use any custom name in place of ‘\loadedtable’

...
\addplot table[x=dof,y=L2] {\loadedtable};
...
\addplot table[x=dof,y=Lmax] {\loadedtable};
...
I am not really sure how much time can be saved, but it works anyway. The \pgfplotstableread
command is documented in all detail in the manual for PgfplotsTable. As a rule of thumb, decide
as follows:
1. If tables contain few rows and many columns, the h\macroi framework will be more efficient.
2. If tables contain more than 200 data points (rows), you should always use file input (and reload if
necessary).
Occasionally, it might be handy to load a table, apply manual preparation steps (for example
\pgfplotstabletranspose) and plot the result tables afterwards.
If you do prefer to access columns by column indices instead of column names (or your tables do not
have column names), you can also use
\addplot table[x index=2,y index=3] {datafile.dat};

\addplot table[x=dof,y index=2] {datafile.dat};
Summary and remarks:

• Use \addplot table[x={hcolumn namei},y={hcolumn namei}] to access column names. Those
names are case sensitive and need to exist.
• Use \addplot table[x index={hcolumn index i},y index={hcolumn index i}] to access column
indices. Indexing starts with 0. You may also use an index for x and a column name for y.
• Use \addplot table[x expr=\coordindex,y={hcolumn namei}] to plot the coordinate index
versus some y data.
3 In
earlier versions, there was an addition keyword ‘from’ before the argument like \addplot table from {\loadedtable}.
This keyword is still accepted, but no longer required.
• Use \addplot table[header=false] {hfile namei} if your input file has no column names. Oth-
erwise, the first non-comment line is checked for column names: if all entries are numbers, they
are treated as numerical data; if one of them is not a number, all are treated as column names.
• It is possible to read error coordinates from tables as well. Simply add options ‘x error’, ‘y error’
or ‘x error index’/‘y error index’ to hsource columnsi. See Section 4.12 for details about error
bars.
• It is possible to read per point meta data (usable in scatter src, see page 109) as has been
discussed for plot coordinates and plot file above. The meta data column can be provided
using the meta key (or the meta index key).
• Use \addplot table[hsource columnsi] {h\macroi} to use a pre–read table. Tables can be read
using
\pgfplotstableread{datafile.dat}\macroname.
If you like, you can insert the optional keyword ‘from’ before \macroname.
• The accepted input format of tables is as follows:
– Rows are separated by new line characters.
Alternatively, you can use row sep=\\ which enables ‘\\’ as row separator. This might become
necessary for inline table data, more precisely: if newline characters have been converted to
white spaces by TEX’s character processing before pgfplots had a chance to see them. This
happens if inline tables are provided inside of macros. Use row sep=\\ and separate the rows
by ‘\\’ if you experience such problems.
– Columns are usually separated by white spaces (at least one tab or space).
If you need other column separation characters, you can use the
col sep=space|tab|comma|colon|semicolon|braces|&|ampersand
option documented in all detail in the manual for PgfplotsTable which is part of pgfplots.
– Any line starting with ‘#’ or ‘%’ is ignored.
– The first line will be checked if it contains numerical data. If there is a column in the first
line which is no number, the complete line is considered to be a header which contains column
names. Otherwise it belongs to the numerical data and you need to access column indices
instead of names.
– There is future support for a second header line which must start with ‘$flags ’. Currently,
such a line is ignored. It may be used to provide number formatting hints like precision and
number format if those tables shall be typeset using \pgfplotstabletypeset (see the manual
for PgfplotsTable).
– The accepted number format is the same as for ‘plot coordinates’, see above.
– If you omit column selectors, the default is to plot the first column against the second. That
means plot table does exactly the same job as plot file for this case.
– If you need unbalanced columns, simply use nan as “empty cell” placeholder. These coordinates
will be skipped in plots.
• It is also possible to use mathematical expressions together with ‘plot table’. This is docu-
mented in all detail in Section 4.3.4, but the key idea is to use one of x expr, y expr, z expr or
meta expr as in ‘plot table[x expr=\thisrow{maxlevel}+3,y=L2]’.
• The PgfplotsTable package coming with pgfplots has a the feature “Postprocessing Data in
New Columns” (see its manual).
This allows to compute new columns based on existing data. One of these features is create
col/linear regression (described in Section 4.24).
You can invoke all the create col/hkey namei features directly in \addplot table using
\addplot table[x={create col/hkey namei=hargumentsi}].
In this case, a new column will be created using the functionality of hkey namei. This column
generation is described in all detail in PgfplotsTable. Finally, the resulting data is available as
x coordinate (the same holds for y= or z=).
One application (with several examples how to use this syntax) is line fitting with create
col/linear regression, see Section 4.24 for details.
• The table can contain empty lines to tell pgfplots that the function has jumps. To use it,
simply insert an empty line (and ensure that you have \pgfplotsset{compat=1.4} or newer in
your preamble). See the documentation of empty line for details.
• Technical note: every opened file will be protocolled into your log file.
Keys To Configure Table Input

The following list of keys allow di↵erent methods to select input data or di↵erent input formats. Note that
the common prefix ‘table/’ can be omitted if these keys are set after \addplot table[hoptionsi]. The
/pgfplots/ prefix can always be omitted when used in a pgfplots method.
/pgfplots/table/header=true|false (initially true)

Allows to disable header identification for plot table. See above.
/pgfplots/table/x={hcolumn namei}
/pgfplots/table/y={hcolumn namei}
/pgfplots/table/z={hcolumn namei}
/pgfplots/table/x index={hcolumn index i}
/pgfplots/table/y index={hcolumn index i}
/pgfplots/table/z index={hcolumn index i}
These keys define the sources for plot table. If both column names and column indices are given,
column names are preferred. Column indexing starts with 0. The initial setting is to use x index=0
and y index=1.
Please note that column aliases will be considered if unknown column names are used. Please refer to
the manual of PgfplotsTable which comes with this package.
/pgfplots/table/x expr={hexpressioni}
/pgfplots/table/y expr={hexpressioni}
/pgfplots/table/z expr={hexpressioni}
These keys allow to combine the mathematical expression parser with file input. They are listed here
to complete the list of table keys, but they are described in all detail in Section 4.3.4.
The key idea is to provide an hexpressioni which depends on table data (possibly on all columns in one
row). Only data within the same row can be used where columns are referenced with \thisrow{hcolumn
namei} or \thisrowno{hcolumn index i}.
Please refer to Section 4.3.4 for details.
/pgfplots/table/x error={hcolumn namei}

/pgfplots/table/y error={hcolumn namei}
/pgfplots/table/z error={hcolumn namei}
/pgfplots/table/x error index={hcolumn index i}
/pgfplots/table/y error index={hcolumn index i}
/pgfplots/table/z error index={hcolumn index i}
/pgfplots/table/x error expr={hmath expressioni}
/pgfplots/table/y error expr={hmath expressioni}
/pgfplots/table/z error expr={hmath expressioni}
These keys define input sources for error bars with explicit error values.
The x error method provides an input column name (or alias), the x error index method provides
input column indices and x error expr works just as table/x expr: it allows arbitrary mathematical
expressions which may depend on any number of table columns using \thisrow{hcol namei}.
Please see Section 4.12 for details about the usage of error bars.
/pgfplots/table/x error plus={hcolumn namei}

/pgfplots/table/y error plus={hcolumn namei}
/pgfplots/table/z error plus={hcolumn namei}
/pgfplots/table/x error plus index={hcolumn index i}
/pgfplots/table/y error plus index={hcolumn index i}
/pgfplots/table/z error plus index={hcolumn index i}
/pgfplots/table/x error plus expr={hmath expressioni}

/pgfplots/table/y error plus expr={hmath expressioni}
/pgfplots/table/z error plus expr={hmath expressioni}
/pgfplots/table/x error minus={hcolumn namei}
/pgfplots/table/y error minus={hcolumn namei}
/pgfplots/table/z error minus={hcolumn namei}
/pgfplots/table/x error minus index={hcolumn index i}
/pgfplots/table/y error minus index={hcolumn index i}
/pgfplots/table/z error minus index={hcolumn index i}
/pgfplots/table/x error minus expr={hmath expressioni}
/pgfplots/table/y error minus expr={hmath expressioni}
/pgfplots/table/z error minus expr={hmath expressioni}
These keys define input sources for error bars with asymmetric error values, i.e. di↵erent values for
upper and lower bounds.
They are to be used in the same way as x error. In fact, x error is just a style which sets both x
error plus and x error minus to the same value.
Please see Section 4.12 for details about the usage of error bars.
/pgfplots/table/meta={hcolumn namei}
/pgfplots/table/meta index={hcolumn index i}
/pgfplots/table/meta expr={hexpressioni}
These keys define input sources for per point meta data. Please see page 109 for details about meta
data or the documentation for plot coordinates and plot file for further information.
These keys are only useful in conjunction with point meta=explicit or point meta=explicit
symbolic. Note that
\addplot[point meta=explicit] table[meta=colname] .... ;
is equivalent to
\addplot[point meta=\thisrow{colname}] table[] .... ;
If the value of point meta is neither explicit nor explicit symbolic, the choice table/meta (and
its friends) are ignored.
However, if point meta is one of explicit or explicit symbolic, the choice table/meta (or one of
its friends) is mandatory.
/pgfplots/table/row sep=newline|\\ (initially newline)

Configures the character to separate rows.
The choice newline uses the end of line as it appears in the table data (i.e. the input file or any inline
table data).
The choice \\ uses ‘\\’ to indicate the end of a row.
Note that newline for inline table data is “fragile”: you can’t provide such data inside of TEX macros
(this does not apply to input files). Whenever you experience problems, proceed as follows:
1. First possibility: call \pgfplotstableread{hdatai}\yourmacro outside of any macro declaration.
2. Use row sep=\\.
The same applies if you experience problems with inline data and special col sep choices (like col
sep=tab).
The reasons for such problems is that TEX scans the macro bodies and replaces newlines by white
spaces. It does other substitutions of this sort as well, and these substitutions can’t be undone (maybe
not even found).
/pgfplots/table/col sep=space|tab|comma|semicolon|colon|braces|&|ampersand (initially space)

Allows to choose column separators for plot table. Please refer to the manual of PgfplotsTable
which comes with this package for details about col sep.
/pgfplots/table/read completely={hauto,true,falsei} (initially auto)

Allows to customize \addplot table{hfile namei} such that it always reads the entire table into mem-
ory.
This key has just one purpose, namely to create postprocessing columns on-the-fly and to plot those
columns afterwards. This “lazy evaluation” which creates missing columns on-the-fly is documented in
the PgfplotsTable manual (in section “Postprocessing Data in New Columns”).
The initial configuration auto checks whether one of the keys table/x, table/y, table/z or table/meta
contains a create on use column. If so, it enables read completely, otherwise it prefers to load the
file in the normal way.
Attention: Usually, \addplot table only picks required entries, requiring linear runtime complexity.
As soon as read completely is activated, tables are loaded completely into memory. Due to datastruc-
tures issues (“macro append runtime”), the runtime complexity for read completely is O(N 2 ) where
N is the number of rows. Thus: use this feature only for “small” tables4 .
/pgfplots/table/ignore chars={hcomma-separated-listi} (initially empty)

Allows to silently remove a set of single characters from input files. The characters are separated by
commas. The documentation for this command, including cases like ‘\%,\#,\ ’ or binary character
codes like ‘\^^ff’ can be found in the manual for PgfplotsTable.
This setting applies to \addplot file as well.
/pgfplots/table/white space chars={hcomma-separated-listi} (initially empty)

Allows to define a list of single characters which are actually treated like white spaces (in addition to
tabs and spaces). Please refer to the manual of PgfplotsTable for details.
This setting applies to \addplot file as well.
/pgfplots/table/comment chars={hcomma-separated-listi} (initially empty)

Allows to add one or more additional comment characters. Each of these characters has a similar e↵ect
as the # character, i.e. all following characters of that particular input line are skipped.
For example, comment chars=! uses ‘!’ as additional comment character (which allows to parse
Touchstone files).
Please refer to the manual of PgfplotsTable for details.
/pgfplots/table/skip first n={hinteger i} (initially 0)

Allows to skip the first hinteger i lines of an input file. The lines will not be processed.
Please refer to the manual of PgfplotsTable for details.
/pgfplots/table/search path={hcomma-separated-listi} (initially .)

Allows to provide a search path for input tables. This variable is evaluated whenever pgfplots attempts
to read a data file. This includes both \pgfplotstableread and \addplot table; its value resembles
a comma-separated list of path names. The requested file will be read from the first matching location
in that list.
Use ‘.’ to search using the normal TEX file searching procedure. This standard search procedure will
typically use the current working directory and the environment variable TEXINPUTS as for any other
\input or \include statements.
An entry in hcomma-separated-listi can be relative to the current working directory, i.e. something like
search path={.,../datafiles} is accepted.
/pgfplots/table/search path/implicit .=true|false (initially true)

PgfplotsTable allows to add ‘.’ to the value of search path implicitly as this is typically assumed
in many applications of search paths.
The initial configuration search path/implicit .=true will ensure that ‘.’ is added in front of the
search path if the user value does not contain a ‘.’.
4 This remark might be deprecated; many of the slow routines have been optimized in the meantime to have at least
pseudo-linear runtime.
The value search path/implicit .=false will not add ‘.’.

Keep in mind that ‘.’ means “let TEX search for the file on its own”. This will typically find files in the
current working directory, but it will also include processing of the environment variable TEXINPUTS.
4.3.3 Computing Coordinates with Mathematical Expressions

\addplot {hmath expressioni} ;
\addplot[hoptionsi] {hmath expressioni} htrailing path commandsi;
\addplot3 . . .
This input method allows to provide mathematical expressions which will be sampled. But unlike plot
gnuplot, the expressions are evaluated using the math parser of pgf, no external program is required.
Plot expression samples x from the interval [a, b] where a and b are specified with the domain key. The
number of samples can be configured with samples=hN i as for plot gnuplot.
\begin{tikzpicture}
500 \begin{axis}
\addplot {x^2 + 4};
\addplot {-5*x^3 - x^2};
\end{axis}
0 \end{tikzpicture}
500
6 4 2 0 2 4 6
Please note that pgf’s math parser is configured to use trig format=degrees by default whenever
trigonometric functions are involved:
\begin{axis}
\addplot+[domain=0:360]
{sin(x)};
\end{axis}
\end{tikzpicture}
0
1
0 100 200 300
If you want to use radians, use

\begin{axis}
\addplot+[domain=-pi:pi]
{sin(deg(x))};
\end{axis}
\end{tikzpicture}
0
1
2 0 2
to convert the radians to degrees (also see trig format and trig format plots).
The plot expression parser also accepts some more options like samples at={hcoordinate listi} or
domain=hfirsti:hlasti which are described below.
Remarks
1. What really goes on is a loop which assigns the current sample coordinate to the macro \x.
pgfplots defines a math constant x which always has the same value as \x.
In short: it is the same whether you write \x or just x inside of math expressions.
The variable name can be customized using variable=t. Then, t will be the same as \t.
2. The complete set of math expressions can be found in the pgf manual. The most important
mathematical operations are +, -, *, /, abs, round, floor, mod, <, >, max, min, sin, cos, tan, deg
(conversion from radians to degrees), rad (conversion from degrees to radians), atan, asin, acos,
cot, sec, cosec, exp, ln, sqrt, the constants pi and e, ^ (power operation), factorial5 , rand
(random between 1 and 1), rnd (random between 0 and 1), number format conversions hex, Hex,
oct, bin and some more. The math parser has been written by Mark Wibrow and Till Tantau [6],
the FPU routines have been developed as part of pgfplots. The documentation for both parts
can be found in [6].
Please note, however, that trigonometric functions are defined in degrees (see trig format). The
character ‘^’ is used for exponentiation (not ‘**’ as in gnuplot).
3. If the x axis is logarithmic, samples will be drawn logarithmically.
4. Please note that plot expression does not allow separate per point meta data (color data). You
can, of course, use point meta=f(x) or point meta=x.
About the precision and number range: Starting with version 1.2, plot expression uses a
floating point unit. The FPU provides the full data range of scientific computing with a relative
precision between 10 4 and 10 6 . The /pgf/fpu key provides some more details.
In case the fpu does not provide the desired mathematical function or is too slow6 , you should consider
using the plot gnuplot method which invokes the external, freely available program gnuplot as desktop
calculator.
x2 \begin{tikzpicture}
10
title={$\frac{1}{x^2}$}]
\addplot[blue,domain=1:1e30]
{x^-2};
\end{loglogaxis}
18
42
10
66
10 3
10 106 1015 1024 1033
5 Starting with pgf versions newer than 2.00, you can use the postfix operator ! instead of factorial.
6 Or in case you find a bug. . .
ex logarithmically plotted % Preamble: \pgfplotsset{width=7cm,compat=1.13}

\begin{tikzpicture}
10334 \begin{semilogyaxis}[
title={$e^x$ logarithmically plotted}]
\addplot[blue,domain=1:700]
{exp(x)};
\end{semilogyaxis}
1092
29
10
0 200 400 600
\addplot expression {hmath expr i};

\addplot[hoptionsi] expression {hmath expr i} htrailing path commandsi;
\addplot3 . . .
The syntax
\addplot {hmath expressioni};
as short-hand equivalent for
\addplot expression {hmath expressioni};
\addplot (hx expressioni,hy expressioni) ;

\addplot[hoptionsi] (hx expressioni,hy expressioni) htrailing path commandsi;
\addplot3 . . .
A variant of \addplot expression which allows to provide di↵erent coordinate expressions for the x
and y coordinates. This can be used to generate parametrized plots.
Please note that \addplot (x,x^2) is equivalent to \addplot expression {x^2}.
Note further that since the complete point expression is surrounded by round braces, round braces for
either hx expressioni or hy expressioni need special attention. You will need to introduce curly braces
additionally to allow round braces:
\addplot ({hx expr i}, {hy expr i}, {hz expr i});
/pgfplots/domain=hx1 i:hx2 i (initially [-5:5])

/pgfplots/y domain=hy1 i:hy2 i
/pgfplots/domain y=hy1 i:hy2 i
Sets the function’s domain(s) for plot expression and plot gnuplot. Two dimensional plot expres-
sions are defined as functions f : [x1 , x2 ] ! R and hx1 i and hx2 i are set with domain. Three dimensional
plot expressions use functions f : [x1 , x2 ] ⇥ [y1 , y2 ] ! R and hy1 i and hy2 i are set with y domain. If
y domain is empty, [y1 , y2 ] = [x1 , x2 ] is assumed for three dimensional plots (see page 126 for details
about three dimensional plot expressions).
The keys y domain and domain y are the same.
The domain key will be ignored if samples at is specified; samples at has higher precedence.
Please note that domain is not necessarily the same as the axis limits (which are configured with the
xmin/xmax options).
The domain keys are only relevant for gnuplot and plot expression. In case you’d like to plot only a
subset of other coordinate input routines, consider using the coordinate filter restrict x to domain.
Remark for Tik Z-users: /pgfplots/domain and /tikz/domain are independent options. Please
prefer the pgfplots variant (i.e. provide domain to an axis, \pgfplotsset or a plot command). Since
older versions also accepted something like \begin{tikzpicture}[domain=. . . ], this syntax is also
accepted as long as no pgfplots domain key is set.
/pgfplots/samples={hnumber i} (initially 25)

/pgfplots/samples y={hnumber i}
Sets the number of sample points for plot expression and plot gnuplot. The samples key defines the
number of samples used for line plots while the samples y key is used for mesh plots (three dimensional
visualisation, see page 126 for details). If samples y is not set explicitly, it uses the value of samples.
The samples key won’t be used if samples at is specified; samples at has higher precedence.
The same special treatment of /tikz/samples and /pgfplots/samples as for the domain key applies
here. See above for details.
/pgfplots/samples at={hcoordinate listi}
Sets the x coordinates for plot expression explicitly. This overrides domain and samples.
The hcoordinate listi is a \foreach expression, that means it can contain a simple list of coordinates
(comma–separated), but also complex ... expressions like7
\pgfplotsset{samples at={5e-5,7e-5,10e-5,12e-5}}
\pgfplotsset{samples at={-5,-4.5,...,5}}
\pgfplotsset{samples at={-5,-3,-1,-0.5,0,...,5}}
The same special treatment of /tikz/samples at and /pgfplots/samples at as for the domain key
applies here. See above for details.
Attention: samples at overrides domain, even if domain has been set after samples at! Use
samples at={} to clear hcoordinate listi and re-activate domain.
/pgfplots/variable={hvariable namei} (initially x)
/pgfplots/variable y={hvariable namei} (initially y)
Defines the variables names which will be sampled in domain (with variable) and in domain y (with
variable y).
The same variables are used for parametric and for non-parametric plots. Use variable=t to change
them if you like (for gnuplot, there is such a distinction; see parametric/var 1d).
Technical remark: Tik Z also uses the variable key. However, it expects a macro name, i.e. \x instead
of just x. Both possibilities are accepted here.
/pgfplots/trig format plots=default|deg|rad (initially default)
Allows to reconfigure the input format for trigonometric functions like sin, cos, tan, and their friends.
This key reconfigures trigonometric functions inside of plot expressions, point meta arguments, and
other items which are directly related to the evaluation of plot coordinates.
Note that this does not apply to Tik Z drawing instructions like \node, \draw, \fill, etc.
Attention: this feature is experimental: it has been tested for default axes only and is known to fail
for polaraxis and smithchart.
plots in radians \begin{tikzpicture}
\begin{axis}[view={60}{30},trig format plots=rad,
title=plots in radians]
\addplot3+[domain=0:4*pi,samples=19,samples y=0]
2 ({sin(x)},
{cos(x)},
{2*x/(4*pi)});
1
% drawing instructions still use PGF’s default
\node[fill=white,draw=black,anchor=center] at
0 X ({sin(90)},{cos(90)},1) {X};
\end{axis}
\end{tikzpicture}
0.5 1
0
0.5 0
7 Unfortunately, the ... is somewhat restrictive when it comes to extended accuracy. So, if you have particularly small or
large numbers (or a small distance), you have to provide a comma–separated list (or use the domain key).
/pgf/trig format=deg|red (initially deg)

Allows to reconfigure the trigonometric format for all user arguments.
This a↵ects all user arguments including view, Tik Z polar coordinates, pins of \nodes, start/end angles
for edges, etc.
At the time of this writing, this feature is in experimental state: it can happen that it breaks Tik Z
internals. Please handle with care and report any bugs.
all in radians \begin{tikzpicture}
\begin{axis}[view={rad(60)}{rad(30)},trig
format=rad,
title=all in radians]
2 \addplot3+[domain=0:4*pi,samples=19,samples y=0]
({sin(x)},
{cos(x)},
1 {2*x/(4*pi)});
% drawing instructions now also use radians

0 X
\node[fill=white,draw=black,anchor=center] at
({sin(pi/2)},{cos(pi/2)},1) {X};
\end{axis}
0.5 1
0 \end{tikzpicture}
0.5 0
4.3.4 Mathematical Expressions And File Data

pgfplots allows to combine ‘plot table’ and ‘plot expression’ to get both file input and modifications
by means of mathematical expressions.
\addplot table [hcolumn selection and expressionsi]{hfilei};

\addplot[hoptionsi] table [hcolumn selection and expressionsi]{hfilei} htrailing path commandsi;
\addplot3 . . .
Besides the already discussed possibility to provide a column selection by means of column names
(x=hnamei or x index=hindex i, see Section 4.3.2), it is also possible to provide mathematical expressions
as arguments.
Mathematical expressions are specified with x expr=hexpressioni inside of hcolumn selection and
expressionsi. They can depend on zero, one or more columns of the input file. A column is referenced
using the special command ‘\thisrow{hcolumn namei}’ within hexpressioni (or \thisrownohcolumn
index i).
maxlevel L2
2
2 2.97 · 10
2
2 2.97 · 10
3
4 5.27 · 10
3
5 3.8 · 10
4
6 8.41 · 10
4
6 5.01 · 10
4
7 1.11 · 10
5
8 5.41 · 10
5
9 1.25 · 10
6
10 6.01 · 10
6
11 1.11 · 10
7
11 5.9 · 10
7
12 1.03 · 10
1
10
3
10
5
10
7
10
12 14 16 18 20 22
maxlevel+10
\pgfplotstabletypeset[columns={maxlevel,L2}]{plotdata/newexperiment1.dat}
\begin{tikzpicture}
\begin{semilogyaxis}[
xlabel=\texttt{maxlevel}$ + 10$
]
\addplot table
[x expr=\thisrow{maxlevel}+10, y=L2]
{plotdata/newexperiment1.dat};
\end{semilogyaxis}
\end{tikzpicture}
Besides x expr, there are keys y expr, z expr and meta expr where the latter allows to provide point
meta data (which is used as scatter src or color data for surface plots etc.).
Inside of hexpressioni, the following macros can be used to access numerical data cells inside of the
input file:
\thisrow{hcolumn namei}
Yields the value of the column designated by hcolumn namei. There is no limit on the number of
columns which can be part of a mathematical expression, but only values inside of the currently
processed table row can be used.
It is possible to provide column aliases for hcolumn namei as described in the manual of Pgfplot-
sTable.
The argument hcolumn namei has to denote either an existing column or one for which a column
alias exists (see the manual of PgfplotsTable). If it can’t be resolved, the math parser yields an
“Unknown function” error message.
\thisrowno{hcolumn index i}
Similar to \thisrow, this command yields the value of the column with index hcolumn index i
(starting with 0).
\coordindex
Yields the current index of the table row (starting with 0). This does not count header or comment
lines.
\lineno
Yields the current line number (starting with 0). This does also count header and comment lines.
If x index, x and x expr (or the corresponding keys for y, z or meta) are combined, this is how they
interact:
1. Column access via x has higher precedence than index access via x index.
2. Even if x expr is provided, the values of x index and x are still checked. Any value found using
column name access or column index access is made available as \columnx (or \columny, \columnz,
\columnmeta, resp.). However, the result of x expr is used as plot coordinate.
This allows to access the cell values identified by x or x index using the “pointer” \columnx. I
am not sure if this yields any advantage, but it is possible nevertheless. If in doubt, prefer using
\thisrow{hcolumn namei}.
Attention: If your table has less than two rows, you may need to set x index={},y index={} ex-
plicitly. This is a consequence of the fact that column name/index access is still applied even if an
expression is provided.
4.3.5 Computing Coordinates with Mathematical Expressions (gnuplot)

\addplot gnuplot [hfurther optionsi]{hgnuplot codei};
\addplot[hoptionsi] gnuplot [hfurther optionsi]{hgnuplot codei} htrailing path commandsi;
\addplot3 . . .
In contrast to plot expression, the plot gnuplot command8 employs the external program gnuplot
to compute coordinates. The resulting coordinates are written to a text file which will be plotted
with plot file. pgf checks whether coordinates need to be re-generated and calls gnuplot whenever
necessary (this is usually the case if you change the number of samples, the argument to plot gnuplot
or the plotted domain9 ).
The di↵erences between plot expression and plot gnuplot are:
• plot expression does not require any external programs and requires no additional command
line options.
• plot expression does not produce a lot of temporary files.
• plot gnuplot uses radians for trigonometric functions while plot expression has degrees (unless
pgf is configured for trig format=rad).
• plot gnuplot is faster.
• plot gnuplot has a larger mathematical library.
• plot gnuplot has a higher accuracy. However, starting with version 1.2, this is no longer a great
problem. The new floating point unit for TEX provides reasonable accuracy and the same data
range as gnuplot.
Since system calls are a potential danger, they need to be enabled explicitly using command line options,
for example
pdflatex -shell-escape filename.tex.
Sometimes it is called shell-escape or enable-write18. Sometimes one needs two hyphens – that all
depends on your TEX distribution.
8 Note that plot gnuplot is actually a re-implementation of the plotfunction method known from pgf. It also invokes pgf
basic layer commands.
9 Please note that pgfplots produces slightly di↵erent files than Tik Z when used with plot gnuplot (it configures high
precision output). You should use di↵erent id for pgfplots and Tik Z to avoid conflicts in such a case.
\begin{axis}
\addplot
gnuplot[id=sin]{sin(x)};
\end{axis}
\end{tikzpicture}
0
1
6 4 2 0 2 4 6
\begin{tikzpicture}
104 \begin{semilogyaxis}
\addplot gnuplot
[id=exp,domain=0:10]{exp(x)};
103
\end{semilogyaxis}
\end{tikzpicture}
102
101
100
0 2 4 6 8 10
The hoptionsi determine the appearance of the plotted function; these parameters also a↵ect the legend.
There is also a set of options which are specific to the gnuplot interface. These options are described in
all detail in [6, section 18.6]. A short summary is shown below.
Some remarks:
• The independent variable for one-dimensional plots can be changed with the variable option, just
as for plot expression. Similarly, the second variable for two dimensional plots can be changed
with variable y.
For parametric plots, the variable names need to be adjusted with parametric/var 1d and
parametric/var 2d (since gnuplot uses t and u,v as initial values for parametric plots).
• Please note that plot gnuplot does not allow separate per point meta data (color data for each
coordinate). You can, however, use point meta=f(x) or point meta=x.
• The generated output file name can be customized with id, see below.
Please refer to [6, section 18.6] for more details about plot function and the gnuplot interaction.
\addplot function {hgnuplot codei};

\addplot[hoptionsi] function {hgnuplot codei} htrailing path commandsi;
\addplot3 . . .
Use
\addplot function {hgnuplot codei};
as alias for
\addplot gnuplot {hgnuplot codei};
/pgfplots/translate gnuplot=true|false (initially true)

Enables or disables automatic translation of the exponentiation operator ‘^’ to ‘**’.
This features allows to use ^ in plot gnuplot instead of gnuplot’s **.
/pgfplots/parametric=true|false (initially false)

Set this to true if you’d like to use parametric plots with gnuplot. Parametric plots use a comma
separated list of expressions to make up x(t), y(t) for a line plot or x(u, v), y(u, v) z(u, v) for a mesh
plot (refer to the gnuplot manual for more information about its input methods for parametric plots).
/pgfplots/parametric/var 1d={hvariable namei} (initially t)

/pgfplots/parametric/var 2d={hvariable name,variable namei} (initially u,v)
Allows to change the dummy variables used by parametric gnuplot plots. The initial setting is the
one of gnuplot: to use the dummy varialbe ‘t’ for parametric line plots and ‘u,v’ for parametric mesh
plots.
These keys are quite the same as variable and variable y, only for parametric plots. If you like to
change variables for non-parametric plots, use variable and/or variable y.
In case you don’t want the distinction between parametric and non-parametric plots, use
\pgfplotsset{parametric/var 1d=,parametric/var 2d=}.
/tikz/id={hunique string identifier i}

A unique identifier for the current plot. It is used to generate temporary filenames for gnuplot output.
/tikz/prefix={hfile name prefix i}

A common path prefix for temporary filenames (see [6, section 18.6] for details).
/tikz/raw gnuplot (no value)

Disables the use of samples and domain.
4.3.6 Computing Coordinates with External Programs (shell)

\addplot shell [hfurther optionsi]{hshell commandsi};
\addplot[hoptionsi] shell [hfurther optionsi]{hshell commandsi} htrailing path commandsi;
\addplot3 . . .
An extension by Stefan Tibus
In contrast to plot gnuplot, the plot shell command allows execution of arbitrary shell commands
to compute coordinates. The resulting coordinates are written to a text file which will be plotted with
plot file. pgf checks whether coordinates need to be re-generated and executes the hshell commandsi
whenever necessary.
Since system calls are a potential danger, they need to be enabled explicitly using command line options,
for example
pdflatex -shell-escape filename.tex.
Sometimes it is called shell-escape or enable-write18. Sometimes one needs two slashes – that all
depends on your TEX distribution.
\begin{axis}
\addplot
shell[prefix=pgfshell_,id=cos]{awk ’BEGIN{
pi=3.14159; N=10;
for(i=0;i<=N;i++) print i,cos(i/N*pi);}’};
0
\end{axis}
\end{tikzpicture}
1
0 2 4 6 8 10
\begin{axis}
\addplot+[prefix=pgfshell_,id=replot]
shell{cat pgfshell_cos.out};
% just reprint the result from above
\end{axis}
0
\end{tikzpicture}
1
0 2 4 6 8 10
The hoptionsi determine the appearance of the plotted function; these parameters also a↵ect the legend.
There is also a set of options which are specific to the gnuplot and the shell interface. These options
are described in all detail in [6, section 19.6]. A short summary is shown below.
/tikz/id={hunique string identifier i}

A unique identifier for the current plot. It is used to generate temporary filenames for shell output.
/tikz/prefix={hfile name prefix i}

A common path prefix for temporary filenames (see [6, section 19.6] for details).
4.3.7 Using External Graphics as Plot Sources

\addplot graphics {hfile namei};
\addplot[hoptionsi] graphics {hfile namei} htrailing path commandsi;
\addplot3 . . .
This plot type allows to extend the plotting capabilities of pgfplots beyond its own limitations. The
idea is to generate the graphics as such (for example, a contour plot, a complicated shaded surface10 or
a large point cluster) with an external program like Matlab (®) or gnuplot. The graphics, however,
should not contain an axis or descriptions. Then, we use \includegraphics and a pgfplots axis
which fits exactly on top of the imported graphics.
Of course, one could do this manually by providing proper scales and such. The operation plot
graphics is intended so simplify this process. However the main difficulty is to get images with correct
bounding box. Typically, you will have to adjust bounding boxes manually.
Let’s start with an example: Suppose we use, for example, matlab to generate a surface plot like
[X,Y] = meshgrid( linspace(-3,3,500) );

surf( X,Y, exp(-(X - Y).^2 - X.^2 ) );
shading flat; view(0,90); axis off;
print -dpng external1
which is then found in external1.png. The surf command of Matlab generates the surface, the
following commands disable the axis descriptions, initialise the desired view and export it. Viewing
the image in any image tool, we see a lot of white space around the surface – Matlab has a particular
weakness in producing tight bounding boxes, as far as I know. Well, no problem: use your favorite
image editor and crop the image (most image editors can do this automatically). We could use the free
ImageMagick command
convert -trim external1.png external1.png
to get a tight bounding box. Then, we use
10 See also Section 4.6.6 for an overview of pgfplots methods to draw shaded surfaces.
\begin{tikzpicture}
2 \begin{axis}[enlargelimits=false,axis on top]
\addplot graphics
[xmin=-3,xmax=3,ymin=-3,ymax=3]
{external1};
\end{axis}
0
\end{tikzpicture}
2 0 2
to load the graphics11 just as if we would have drawn it with pgfplots. The axis on top simply tells
pgfplots to draw the axis on top of any plots (see its description).
Please note that pgfplots o↵ers support for smaller surface plots as well which might be an option –
unless the number of samples is too large. See Section 4.6.6 for details.
However, external programs have the following advantages here: they are faster, allow more complexity
and provide real z bu↵ering which is currently only simulated by pgfplots. Thus, it may help to
consider plot graphics for complicated surface plots.
Our first test was successful – and not difficult at all because graphics programs can automatically
compute the bounding box. There are a couple of free tools available which can compute tight bounding
boxes for .eps or .pdf graphics:
1. The free vector graphics program inkscape can help here. Its feature “File Document Proper-
ties: Fit page to selection” computes a tight bounding box around every picture element.
However, some images may contain a rectangular path which is as large as the bounding box
(Matlab (®) computes such .eps images). In this case, use the “Ungroup” method (context menu
of inkscape) as often as necessary and remove such a path.
Finally, save as .eps.
However, inkscape appears to have problems with postscript fonts – it substitutes them. This
doesn’t pose problems in this application because fonts shouldn’t be part of such images – the
descriptions will be drawn by pgfplots.
2. The tool pdfcrop removes surrounding whitespace in .pdf images and produces quite good bound-
ing boxes.
Adjusting bounding boxes manually
In case you don’t have tools at hand to provide correct bounding boxes, you can still use TEX to set
the bounding box manually. Some viewers like gv provide access to low–level image coordinates. The
idea is to determine the number of units which need to be removed and communicate these units to
\includegraphics.
I am aware of the following methods to determine bounding boxes manually:
inkscape I am pretty sure that inkscape can do it.
gv The ghost script viewer gv always shows the postscript units under the mouse cursor.
gimp The graphics program gimp usually shows the cursor position in pixels, but it can be configured
to display postscript points (pt) instead.
Let’s follow this approach in a further example.
We use gnuplot to draw a (relatively stupid) example data set. The gnuplot script
11 Please note that I don’t have a Matlab license, so I used gnuplot to produce an equivalent replacement graphics.
set samples 30000

set parametric
unset border
unset xtics
unset ytics
set output "external2.eps"
set terminal postscript eps color
plot [t=0:1] rand(0),rand(0) with dots notitle lw 5
generates external2.eps with a uniform random sample of size 30000. As before, we import this scatter
plot into pgfplots using plot graphics. Again, the bounding box is too large, so we need to adjust
it (gnuplot can do this automatically, but we do it anyway to explain the mechanisms):
Using gv, I determined that the bounding box needs to be shifted 12 units to the left and 9 down.
Furthermore, the right end is 12 units too far o↵ and the top area has about 8 units space wasted. This
can be provided to the trim option of \includegraphics, and we use clip to clip the rest away:
Graphics Import \begin{tikzpicture}
\begin{axis}[axis on top,title=Graphics Import]
1 \addplot graphics
[xmin=0,xmax=1,ymin=0,ymax=1,
% trim=left bottom right top
includegraphics={trim=12 9 12 8,clip}]
{external2};
\addplot coordinates {(0,0) (1,1)};
0.5 \end{axis}
\end{tikzpicture}
0
0 0.2 0.4 0.6 0.8 1
So, plot graphics takes a graphics file along with options which can be passed to \includegraphics.
Furthermore, it provides the information how to embed the graphics into an axis. The axis can contain
any other \addplot command as well and will be resized properly.
Details about plot graphics:
The loaded graphics file is drawn with

\node[/pgfplots/plot graphics/node] {\includegraphics[hoptionsi]{hfile namei}};
where the node style is a configurable style. The node is placed at the coordinate designated by xmin,
ymin.
The hoptionsi are any arguments provided to the includegraphics key (see below) and width and
height determined such that the graphics fits exactly into the rectangle denoted by the xmin, ymin and
xmax, ymax coordinates.
The scaling will thus ignore the aspect ratio of the external image and prefer the one used by pgfplots.
You will need to provide width and height to the pgfplots axis to change its scaling. Use the scale
only axis key in such a case.
Legends in plot graphics:
A legend for plot graphics uses the current plot handler and the current plot mark:
1 Scatter % provide options for the legend:
\addplot[red,only marks,mark=*,mark size=1pt]
Line
graphics
[xmin=0,xmax=1,ymin=0,ymax=1,
0.5 {external2};
\legend{Scatter,Line}
0 \end{axis}
\end{tikzpicture}
0 0.2 0.4 0.6 0.8 1
4.3.8 Keys To Configure Plot Graphics

The following list of keys configure \addplot graphics. Note that the common prefix ‘plot graphics/’
can be omitted if these keys are set after \addplot graphics[hoptionsi]. The /pgfplots/ prefix can always
be omitted when used in a pgfplots method.
/pgfplots/plot graphics/xmin={hcoordinatei}
/pgfplots/plot graphics/ymin={hcoordinatei}
/pgfplots/plot graphics/zmin={hcoordinatei}
/pgfplots/plot graphics/xmax={hcoordinatei}
/pgfplots/plot graphics/ymax={hcoordinatei}
/pgfplots/plot graphics/zmax={hcoordinatei}
These keys are required for plot graphics and provide information about the external data range.
The graphics will be squeezed between these coordinates. The arguments are axis coordinates; they are
only useful if you provide each of them.
Alternatively, you can also use the plot graphics/points feature to provide the external data range,
see below.
/pgfplots/plot graphics/points={hlist of coordinatesi} (initially empty)

This key also allows to provide the external data range. It constitutes an alternative to plot graph-
ics/xmin (and its variants): simply provide at least two coordinates in hlist of coordinatesi. Their
bounding box is used to determine the external data range, and the graphics is squeezed between these
coordinates.
The example from above can be written equivalently as
1 \addplot graphics
% instead of the min/max things:
[points={(0,1) (1,0)},
{external2};
0.5 \addplot coordinates {(0,0) (1,1)};
\end{axis}
\end{tikzpicture}
0
0 0.2 0.4 0.6 0.8 1
The hlist of coordinatesi is a sequence of the form (x,y) for two–dimensional plots and (x,y,z) for
three–dimensional ones, the ordering is irrelevant. The single elements are separated by white space.
It is possible to mix plot graphics/xmin and variants with plot graphics/points.

The plot graphics/points key has further functionality for inclusion of three–dimensional graphics
which is discussed at the end of this section (on page 66). Here is a short reference on the accepted syntax
for three–dimensional plot graphics: in addition to the (x,y,z) syntax, you can provide arguments of
the form (x,y,z) => (X,Y). Here, the first (three–dimensional) coordinate is a logical coordinate and
the second (two–dimensional) coordinate denotes the coordinates of the very same point, but inside of
the included image (relative to the lower left corner of the image). Applications and examples for this
syntax can be found in the section for three–dimensional plot graphics (see page 66).
/pgfplots/plot graphics/includegraphics={hoptionsi}
A list of options which will be passed as–is to \includegraphics. Interesting options include the
trim=hlefti hbottomi hrighti htopi key which reduces the bounding box and clip which discards every-
thing outside of the bounding box. The scaling options won’t have any e↵ect, they will be overwritten
by pgfplots.
/pgfplots/plot graphics/includegraphics cmd={h\macroi} (initially \includegraphics)
Allows to use a di↵erent graphics routine. A possible choice could be \pgfimage. The macro should
accept the width and height arguments (in brackets) and the file name as first argument.
/pgfplots/plot graphics/node (style, no value)
A predefined style used for the Tik Z node containing the graphics. The predefined value is
\pgfplotsset{
plot graphics/node/.style={
transform shape,
inner sep=0pt,
outer sep=0pt,
every node/.style={},
anchor=south west,
at={(0pt,0pt)},
rectangle
}
}
/pgfplots/plot graphics (no value)

This key belongs to the public low–level plotting interface. You won’t need it in most cases.
This key is similar to sharp plot or smooth or const plot: it installs a low–level plot–handler which
expects exactly two points: the lower left corner and the upper right one. The graphics will be drawn
between them. The graphics file name is expected as value of the /pgfplots/plot graphics/src key.
The other keys described above need to be set correctly (excluding the limits, these are ignored at this
level of abstraction). This key can be used independently of an axis.
/pgfplots/plot graphics/lowlevel draw={hwidthi}{hheighti}
A low–level interface for plot graphics which actually invokes \includegraphics. But there is no
magic involved: the command is simply expected to draw a box of dimensions hwidthi ⇥ hheighti. The
coordinate system has already been shifted correctly.
The initial configuration is
\includegraphics[hvalue of “ plot graphics/includegraphics” i,width=#1,height=#2]
{hvalue of “ plot graphics/src” i}.
Thus, you can tweak plot graphics to place any TEX box of the desired dimensions into an axis
between the provided minimum and maximum coordinates. It is not necessary to make use of the
graphics file name or the options in the ‘includegraphics’ key if you overwrite this lowlevel interface
with
plot graphics/lowlevel draw/.code 2 args={hcode which depends on #1 and #2 i}.
Support for External Three-Dimensional Graphics

pgfplots o↵ers several visualization techniques for three dimensional graphics. Nevertheless, complex
visualizations or specialized applications are beyond the scope of pgfplots and you might want to use
other tools to generate such figures.
Figure 4.1: Using Matlab to extract image coordinates (left) and Gimp to measure distances (right).
The plot graphics tool of pgfplots allows to include three–dimensional external graphics: it generates
a three–dimensional axis on its own. The idea is to provide a graphics (without descriptions) and use
pgfplots to overlay a three–dimensional axis automatically. This allows to maintain document consistency
(making it unnecessary to use di↵erent programs within the same document).
You are probably wondering how this is possible. Well, it needs more user input than two–dimensional
external graphics. The cost to include external three dimensional images into pgfplots is essentially control
of a graphics program like gimp: you need to identify the 3D coordinates of a couple of points in your image.
pgfplots will then squeeze the graphics correctly, and it reconfigures the axis to ensure a correct display
of the result.
Matlab versus other tools: Although this section is based on Matlab images, the technique to import
three–dimensional graphics is independent of Matlab. Thus, if you have a di↵erent tool, you need to read all
that follows. However, users of Matlab can use a simplified export mechanism which has been contributed
by Jürnjakob Dugge. Please skip to section 4.3.8 on page 71 if you use Matlab to generate the graphics files
(although you may want to take a brief look at the examples on the following pages to learn about flexibility
or legends).
Let’s start with two examples. Suppose you generate a surface plot with Matlab and want to include it
in pgfplots. We have the matlab script
[x,y]=meshgrid(linspace(0,1,120));
surf(x,y,sin(8*pi*x).* exp(-20*(y-0.5).^2) + exp(-(x-0.5).^2*30 - (y-0.25).^2 - (x-0.5).*(y-0.25)))
xlabel(’x’), ylabel(’y’)
axis off
print -dpng plotgraphics3dsurf
which generates the figure in question.

After automatically computing a tight bounding box for plotgraphics3dsurf.png (I used gimp’s
Image Autocrop feature), and making the background color transparent (gimp: select the outer white
space with the magic wand, then use12 Layer Transparency Color to Transparency) we get:
The key idea is now to identify several points in the image, and assign both their logical three–dimensional
coordinates and the corresponding two–dimensional canvas coordinates in image coordinates. How? Well,
the three–dimensional coordinates are known to Matlab, it can display them for you if you click somewhere
into the image, compare Figure 4.1 (left).
The two–dimensional canvas coordinates need work; they need to be provided relative to the lower left
corner of the image. I used gimp and activated “Points” as units (lower left corner). The lower left corner
now displays the image coordinates in pt which is compatible with pgfplots. An alternative to pointing
onto coordinates is a measurement tool; compare Figure 4.1 (right) for the “Measure” tool in gimp which
allows to compute the length of a line (in our case, the length of the lower left corner to the point of interest).
12 I have a german version, I am not sure if the translation is correct.
I selected four points in the graphics and noted their 2d image coordinates and their 3d logical coordinates
as follows:
\begin{tikzpicture}
\begin{axis}[
grid=both,minor tick num=1,
2
xlabel=$x$,ylabel=$y$,
]
\addplot3 graphics[
points={% important
0 (0,1,0) => (0,207-112)
(1,0,0) => (446,207-133)
(0.5546,0.5042,1.825) => (236,207)
1 (0,0,0) => (194,207-202)
1
0.5 }] {plotdata/plotgraphics3dsurf.png};
0.5 \end{axis}
y 0 0 x \end{tikzpicture}
Here, the points key gets our collected coordinates as argument. It accepts a sequence of maps of
the form h3d logical coordinatei => h2d canvas coordinatei. In our case, (0,1,0) has been found in the
.png file at (0,207-112). Note that I introduced the di↵erence since gimp counts from the upper left, but
pgfplots counts from the lower left.
Once these four point coordinates are gathered, we find Matlab’s surface plot in a pgfplots axis. You
can modify any appearance options, including di↵erent axis limits or further \addplot commands:
Graphics \begin{tikzpicture}
\begin{axis}[
Scatter xmax=1.5,% extra limits
2 xlabel=$x$,ylabel=$y$,
]
\addplot3[surf] % ’surf’ is only used for the legend.
0 graphics[
points={
(0,1,0) => (0,207-112)
1 1.5 (1,0,0) => (446,207-133)
1 (0.5546,0.5042,1.825) => (236,207)
0.5 0.5 (0,0,0) => (194,207-202)
y 0 0 x }]
{plotdata/plotgraphics3dsurf.png};
\addlegendentry{Graphics}
\addplot3+[only marks] coordinates {

(0,1,0) (1,0,0)
(0.5546,0.5042,1.825) (0,0,0)
};
\addlegendentry{Scatter}
\end{axis}
\end{tikzpicture}
pgfplots uses the four input points to compute appropriate x, y and z unit vectors (and the origin in
graphics coordinates). These four vectors (with two components each) can be computed as a result of a
linear system of size 8 ⇥ 8, that is why you need to provide four input points (each has two coordinates).
pgfplots computes the unit vectors of the imported graphics, and afterwards it rescales the result such that
it fits into the specified width and height. This rescaling respects the unit vector ratio (more precisely, it
uses scale mode=scale uniformly instead of scale mode=stretch to fill). Consequently, the freedom
to change the view of a three–dimensional axis which contains a projected graphics is considerably smaller
than before. Surprisingly, you can still change axis limits and width and height – pgfplots will take care
of a correct display of your imported graphics. Since version 1.6, you can also change zmin and/or zmax –
pgfplots will respect your changes as good as it can.
Here is a further example. Suppose we are given the three–dimensional visualization
It has been generated by matlab (I only added transparency to the background with gimp). Besides
advanced visualization techniques, it uses axis equal, i.e. unit vector ratio=1 1 1. As before, we need
to identify four points, each with its 3d logical coordinates (from matlab) and the associated 2d canvas
coordinates relative to the lower left corner of the graphics (note that there is a lot of white space around
the graphics). Here is the output of pgfplots when you import the resulting graphics:
Geometry provided by Sven Groß, Bonn % Preamble: \pgfplotsset{width=7cm,compat=1.13}
http://www.igpm.rwth-aachen.de/DROPS \begin{tikzpicture}
\begin{axis}[
3 xlabel=$x$,ylabel=$y$,
·10
title={\centering
Geometry provided by Sven Gro\ss, Bonn\\
2 \url{http://www.igpm.rwth-aachen.de/DROPS}\\},
0 title style={text width=6cm,font=\tiny},
]
2 \addplot3 graphics[
points={
5 (-0.002625,0.002625,0) => (140,234)
5 (0,0.00263,0.00263) => (230,364)
3 0 (0,-0.00263,-0.00263) => (366,81)
·10 0 (0,-0.00263,0.00263) => (366,276)
3
y ·10 (0.002625,0.002625,0.002625)
5 5 x
}
]
{plotdata/risingdrop3d.png};
\end{axis}
\end{tikzpicture}
Note that I provided five three–dimensional coordinates here, but the last entry has no => mapping to
two–dimensional canvas coordinates. Thus, it is only used to update the bounding box (see the reference
manual for the points key for details).
The example above leads to a relatively small image and much “empty space”. This is due to the scale
mode=scale uniformly implementation of pgfplots: it decided that the best way is to enlarge the involved
axis limits. Here, “best way” means to satisfy width/height constraints combined with minimally enlarged
(never shrinked) axis limits. The remaining degrees of freedom are width, height, and the axis limits. In
our case, changing the ratio between width and height improves the display:
Geometry provided by Sven Groß, Bonn % Preamble: \pgfplotsset{width=7cm,compat=1.13}

http://www.igpm.rwth-aachen.de/DROPS \begin{tikzpicture}
\begin{axis}[
3 height=8cm,width=7cm,% improve scaling manually
·10
title={\centering
Geometry provided by Sven Gro\ss, Bonn\\
2 \url{http://www.igpm.rwth-aachen.de/DROPS}\\},
title style={text width=6cm,font=\tiny},
]
0 \addplot3 graphics[
points={
(-0.002625,0.002625,0) => (140,234)
(0,0.00263,0.00263) => (230,364)
2 (0,-0.00263,-0.00263) => (366,81)
(0,-0.00263,0.00263) => (366,276)
(0.002625,0.002625,0.002625)
2 }
2 ]
3 0 {plotdata/risingdrop3d.png};
·10 0 \end{axis}
3
y 2 2 ·10 \end{tikzpicture}
x
What happens is that pgfplots selects a single scaling factor which is applied to all units as they have
been deduced from the points key. This ensures that the imported graphics fits correctly into the axis. In
addition, pgfplots does its best to satisfy the remaining constraints.
The complete description of how pgfplots scales the axis can be found in the documentation for scale
mode=scale uniformly. Here is just a brief summary: pgfplots assumes that the prescribed width and
height have to be satisfied. To this end, it rescales the projected unit vectors (i.e. the space which is
taken up for each unit in x, y, and z) and it can modify the axis limits. In the default configuration scale
uniformly strategy=auto, pgfplots will never shrink axis limits.
Compatibility remark: Note that the scaling capabilities have been improved for pgfplots version 1.6.
In previous versions, only scale uniformly strategy=change vertical limits was available which lead
to clipped axes. In short: please consider writing \pgfplotsset{compat=1.6} or newer into your document
to benefit from the improved scaling. If you have \pgfplotsset{compat=1.5} or older, the outcome for
\addplot3 graphics will be di↵erent.
We consider a third example which has been generated by the Matlab code
clear all
close all
seed = sum(clock)
rand(’seed’,seed);
X = rand(10,10,10);
data = smooth3(X,’box’,5);
p1 = patch(isosurface(data,.5), ...
’FaceColor’,’blue’,’EdgeColor’,’none’);
p2 = patch(isocaps(data,.5), ...
’FaceColor’,’interp’,’EdgeColor’,’none’);
isonormals(data,p1)
daspect([1 2 2])
view(3); axis vis3d tight
camlight; lighting phong
% print -dpng plotgraphics3withaxis
axis off
print -dpng plotgraphics3
save plotgraphics3.seed seed -ASCII % to reproduce the result
I only added background transparency with gimp and got the following graphics:
We proceed as before and collect four points, each with 3d logical coordinates (by clicking into the matlab
figure) and their associated 2d canvas (graphics) coordinates using the measure tool of gimp. The result is
shown in the code example below.
\begin{tikzpicture}
\begin{axis}[
10 3d box,
]
\addplot3 graphics[
5 points={
(1,1,1) => (205,48)
10 (10,1,10) => (503,324)
(1,1,4.044)=> (206,102)
10 5 (10,10,10) => (390,398)
5
x }
y 0 0 ]
{plotdata/plotgraphics3.png};
\end{axis}
\end{tikzpicture}
Note that it has non-standard data aspect ratio which is respected by pgfplots automatically.
External Three-Dimensional Graphics and Matlab

An extension by Jürnjakob Dugge
The procedure to map three–dimensional logical coordinates to two–dimensional canvas coordinates is

tedious.
Jürnjakob Dugge contributed a script which does most of the logic and your work is reduced to a
copy–paste job. With his permission, I post the contribution here.
The idea is to start a simple script which records mappings for any coordinates which have been clicked
by the user. It works as follows:
1. Create the Matlab plot, say, using
hist3(randn(10000,2)) % some random data

set(get(gca,’child’),’FaceColor’,’interp’,’CDataMode’,’auto’); % colors
% make sure the "print" paper format is the same as the screen paper format:
set(gcf,’PaperPositionMode’,’auto’)
2. Save the following code as pgfplotscsconversion.m:

function pgfplotscsconversion
% Hook into the Data Cursor "click" event

h = datacursormode(gcf);
set(h,’UpdateFcn’,@myupdatefcn,’SnapToDataVertex’,’off’);
datacursormode on
% select four points in plot using mouse
% The function that gets called on each Data Cursor click

function [txt] = myupdatefcn(obj,event_obj)
% Get the screen resolution, in dots per inch

dpi = get(0,’ScreenPixelsPerInch’);
% Get the click position in pixels, relative to the lower left of the
% screen
screen_location=get(0,’PointerLocation’);
% Get the position of the plot window, relative to the lower left of
% the screen
figurePos = get(gcf,’Position’);
% Get the data coordinates of the cursor

pos = get(event_obj,’Position’);
% Format the data and figure coordinates. The factor "72.27/dpi" is

% necessary to convert from pixels to TeX points (72.27 poins per inch)
display([’(’,num2str(pos(1)),’,’,num2str(pos(2)),’,’,num2str(pos(3)),’) => (’, ...
num2str((screen_location(1)-figurePos(1))*72.27/dpi),’,’, ...
num2str((screen_location(2)-figurePos(2))*72.27/dpi),’)’])
% Format the tooltip display

txt = {[’X: ’,num2str(pos(1))],[’Y: ’,num2str(pos(2))],[’Z: ’,num2str(pos(3))]};
Run pgfplotscsconversion, click on four points in your plot. Preferably select non-colinear points
near the edges of the plot. Copy and paste the four lines that were written to the Matlab command
window.
Make sure that the first two points have di↵erent X and Y values on screen (i.e. image canvas coordi-
nates).
3. Export the plot as an image
axis off
print -dpng matlabout -r400 % PNG called "matlabout.png" with 400 dpi resolution
If you want to export vectors graphics, you should note that pdf output of Matlab is clumsy. It might
be best to export to eps first, followed by a conversion from eps to pdf.
If you really want to use pdf output of Matlab, you may need to set the paper size to match the figure
size by yourself, since the PDF driver does not automatically adjust the size:
% It might be better to use print -depsc followed by epstopdf.

% Use this if you (really) want to use print -dpdf:
currentScreenUnits=get(gcf,’Units’) % Get current screen units
currentPaperUnits=get(gcf,’PaperUnits’) % Get current paper units
set(gcf,’Units’,currentPaperUnits) % Set screen units to paper units
plotPosition=get(gcf,’Position’) % Get the figure position and size
set(gcf,’PaperSize’,plotPosition(3:4)) % Set the paper size to the figure size
set(gcf,’Units’,currentScreenUnits) % Restore the screen units
print -dpdf matlabout % PDF called "matlabout.pdf"
4. Include the image in your pgfplots axis. If you selected points on the plot corners, your xmin,
xmax, ymin and ymax should be set automatically, otherwise you may want to provide those yourself.
Also, adjustments of width and height might be of interest to get the right vertical placement of the
plot. Consider changing zmin and/or zmax to fit your needs (preferrably only one of them; otherwise
pgfplots may be unable to fix the height).
This contribution is from

http://tex.stackexchange.com/questions/52987/3-dimensional-histogram-in-pgfplots .
Summary: External Three-Dimensional Graphics

As has been shown in the previous sections, \addplot3 graphics allows to include three-dimensional graphics
and pgfplots overlays a flexible axis with all its power. The cost to do so is
1. collect both logical three–dimensional coordinates and image–internal two–dimensional coordinates for
four points of your graphics.
In Matlab, this can be simplified by the tool mentioned on page 71.
2. If your axes form a right–handed–coordinate system, that is all. If not, also add x dir=reverse for
any reversed axes.
Consider the following list of you encounter problems while working with \addplot3 graphics:
• It must be possible to deduce the origin and the three (two–dimensional) unit vectors from the four
provide points; otherwise the algorithm will fail.
The algorithm should detect any deficiancies. However, if you encounter strange “Dimension too large”
messages here, you can try other arguments in points. Take a look into your log file, it will probably
indicate the source of problems (or use the debug key).
• Ensure that the external graphics has an orthogonal axis. In fact, the axis may be skewed (just like a
pgfplots axis can be created by means of custom x, y, and z vectors). However, the external image
must not have perspective projection as this is unsupported by pgfplots. The points command
needs to receive four points which belong to linearly independent position vectors.
• pgfplots uses the first two points to squeeze the graphics into the desired coordinates (which implies
that they should not have the same canvas X or Y coordinates). It verifies that the remaining points
arguments are projected correctly.
• The resulting scaling by means of scale mode=scale uniformly will try to satisfy all scaling con-
straints. You can change these constraints by modifying width, height, xmin, xmax, ymin, ymax, zmin,
zmax and/or any combination of these parameters. See also unit rescale keep size which controls
the flexibility of limit changes. There is also a key scale uniformly strategy which allows to select
a di↵erent scaling strategy.
• The image should have a “right–handed–coordinate system”: you should be able to take your right
hand, point your thumb in direction of the x axis, your first finger in direction of y, and your second
finger in direction of the z axis. If that is impossible, once of your axes is reversed and you need to
communicate that to pgfplots explicitly by means of the x dir=reverse key (and its variants).
• Note that this feature has been verified with standard cartesian axes only.
• There is a debug key to investigate what the algorithm is doing:
/pgfplots/plot graphics/debug=true|false|visual (initially false)
If you provide \addplot3 graphics[debug,points={...}], pgfplots will provide debug in-
formation onto your terminal and into the logfile. It will also generate extra files containing
the determined unit vectors and the linear system used to derive them (one such file for every
\addplot3 graphics statement, the filename will be the graphics file name and .dat appended).
Without the debug key, only the log file will contain brief information what pgfplots is doing
behind the scenes.
The choice true activates log messages. The choice visual activates log messages and places
some filled circles at the provided points. The choice false disables all debug features.
4.3.9 Reading Coordinates From Files

\addplot file {hnamei};
\addplot[hoptionsi] file {hnamei} htrailing path commandsi;
\addplot3 . . .
Deprecation note: If you have data files, you should generally use \addplot table. The input
type \addplot file is almost the same, but considerably less powerful. It is only kept for backwards
compatibility.
The \addplot file input mechanism is similar to the Tik Z-command ‘plot file’. It is to be used
like
\addplot file {datafile.dat};
where hnamei is a text file with at least two columns which will be used as x and y coordinates. Lines
starting with ‘%’ or ‘#’ are ignored. Such files are often generated by gnuplot:
#Curve 0, 20 points
#x y type
0.00000 0.00000 i
0.52632 0.50235 i
1.05263 0.86873 i
1.57895 0.99997 i
...
9.47368 -0.04889 i
10.00000 -0.54402 i
This listing has been copied from [6, section 16.4].

Plot file accepts one optional argument,
\addplot file[skip first] {datafile.dat};
which allows to skip over a non-comment header line. This allows to read the same input files as plot
table by skipping over column names. Please note that comment lines do not count as lines here.
The input method plot file can also read meta data for every coordinate. As already explained for
plot coordinates (see above), meta data can be used to change colors or other style parameters for
every marker separately. Now, if point meta is set to explicit or to explicit symbolic and the
input method is plot file, one further element will be read from disk – for every line. Meta data is
always the last element which is read. See page 109 for information and examples about per point meta
data and page 112 for an application example using scatter/classes.
Plot file is very similar to plot table: you can achieve the same e↵ect with
\addplot table[x index=0,y index=1,header=false] {datafile.dat};
Due to its simplicity, plot file is slightly faster while plot table allows higher flexibility.
Technical note: every opened file will be protocolled into your log file.
The file can contain empty lines to tell pgfplots that the function has jumps. To use it, simply insert
an empty line (and ensure that you have \pgfplotsset{compat=1.4} or newer in your preamble). See
the documentation of empty line for details.
/pgfplots/plot file/skip first=true|false (initially false)

/pgfplots/plot file/ignore first=true|false (initially false)
The two keys can be provided as arguments to \addplot file[hoptionsi] {hfilenamei}; to skip the
first non-comment entry in the file. They are equivalent. If you provide them in this context, the prefix
/pgfplots/plot file can be omitted.
4.4 About Options: Preliminaries

pgfplots knows a whole lot of key–value options which can be (re)defined to activate desired features or
modified to apply some fine-tuning.
A key usually has a value (like a number, a string, or perhaps some macro code). You can assign values
to keys (“set keys”) in many places in a LATEX document. The value will remain e↵ective until it is changed
or until the current TEX scope ends (which happens after a closing curly brace ‘}’, after \end{hnamei} or,
for example, after \addplot).
Most keys can be used like
4.4. ABOUT OPTIONS: PRELIMINARIES 75
\begin{tikzpicture}
\begin{axis}[key=value,key2=value2] % axis-wide keys
...
\end{axis}
\end{tikzpicture}
which changes them for the complete axis. A key in this context can be any option defined in this manual,
no matter if it has the /pgfplots/ or the /tikz/ key prefix. Note that key prefixes can be omitted in almost
all cases.
A value can usually be provided without curly braces. For example, if the manual contains something
like ‘xmin={hx coordinatei}’, you can safely skip the curly braces. The curly braces are mandatory if values
contain something which would otherwise confuse the key setup (for example an equal sign ‘=’ or a comma
‘,’).
Some keys can be changed individually for each plot:
\begin{tikzpicture}
\begin{axis}
% keys valid for single plots:
\addplot ...; % uses the "cycle list" to determine keys
\addplot[key=value,key2=value2] ... ; % uses the provided keys (not the "cycle list")
\addplot+[key=value,key2=value2] ... ; % appends something to the "cycle list"
\end{axis}
\end{tikzpicture}
Besides these two possibilities, it is also possible to work with document-wide keys:
\chapter{My Section}
\pgfplotsset{
key=value,
key2=value2,
}
This section has a common key configuration:
\begin{tikzpicture}
\begin{axis}% uses the key config from above
...
\end{axis}
\end{tikzpicture}
In the example above, the \pgfplotsset command changes keys. The changes are permanent and will be
used until
• you redefine them or
• the current environment (like \end{figure}) is ended or
• TEX encounters a closing brace ‘}’.
This includes document–wide preamble configurations like
\documentclass{article}
\pgfplotsset{
xticklabel={$\mathsf{\pgfmathprintnumber{\tick}}$},
every axis/.append style={
font=\sffamily,
},
}
...
The basic engine to manage key–value pairs is pgfkeys which is part of pgf. This engine always has
a key name and a key “path”, which is somehow similar to file name and directory of files. The common
“directory” (key path) of pgfplots is ‘/pgfplots/’. Although the key definitions below provide this full
path, it is always (well, almost always) safe to skip this prefix – pgfplots uses it automatically. The same
holds for the prefixes ‘/tikz/’ which are common for all Tik Z drawing options and ‘/pgf/’ which are for
the (more or less) low–level commands of pgf. All these prefixes can be omitted.
One important concept is the concept of styles. A style is a key which contains one or more other keys.
It can be redefined or modified until it is actually used by the internal routines. Each single component of
Tik Z and pgfplots can be configured with styles.
For example,
\pgfplotsset{legend style={line width=1pt}}
sets the line width for every legend to 1pt by appending ‘line width=1pt’ to the existing style for legends.
There are keys like legend style, ticklabel style, and label style which allow to modify the
predefined styles (in this case the styles for legends, ticklabels and axis labels, respectively). They are, in
general, equivalent to a hstyle namei/.append style={} command (the only di↵erence is that the /.append
style thing is a little bit longer). There is also the possibility to define a new style (or to overwrite an
already existing one) using /.style={}.
There are several other styles predefined to modify the appearance, see Section 4.18.
\pgfplotsset{hkey-value-listi}
Defines or sets all options in hkey-value-listi. The hkey-value-listi can contain any of the options in this
manual which have the prefix /pgfplots/ (however, you do not need to type that prefix).
Inside of hkey-value-listi, the prefixes ‘/pgfplots/’ which are commonly presented in this manual can
be omitted (they are checked automatically).
This command can be used to define default options for the complete document or a part of the
document. For example,
\pgfplotsset{
cycle list={%
{red, mark=*}, {blue,mark=*},
{red, mark=x}, {blue,mark=x},
{red, mark=square*}, {blue,mark=square*},
{red, mark=triangle*}, {blue,mark=triangle*},
{red, mark=diamond*}, {blue,mark=diamond*},
{red, mark=pentagon*}, {blue,mark=pentagon*}
},
legend style={
at={(0.5,-0.2)},
anchor=north,
legend columns=2,
cells={anchor=west},
font=\footnotesize,
rounded corners=2pt,
},
xlabel=$x$,ylabel=$f(x)$
}
can be used to set document-wise styles for line specifications, the legends’ style and axis labels. The
settings remain in e↵ect until the end of the current environment (like \end{figure}) or until you
redefine them or until the next closing curly brace ‘}’ (whatever comes first).
You can also define new styles (collections of key–value–pairs) with /.style and /.append style.
\pgfplotsset{
My Style 1/.style={xlabel=$x$, legend entries={1,2,3} },
My Style 2/.style={xlabel=$X$, legend entries={4,5,6} }
}
The /.style and /.append style key handlers are described in Section 4.18 in more detail.
Key handler hkeyi/.code={hTEX codei}
Occasionally, the pgfplots user interface o↵ers to replace parts of its routines. This is accomplished
using so called “code keys”. What it means is to replace the original key and its behavior with new
hTEX codei. Inside of hTEX codei, any command can be used. Furthermore, the #1 pattern will be the
argument provided to the key.
I’ve been invoked with ‘this here’ \pgfplotsset{

My Code/.code={I’ve been invoked with ‘#1’}}
\pgfplotsset{My Code={this here}}
The example defines a (new) key named My Code. Essentially, it is nothing else but a \newcommand,
plugged into the key-value interface. The second statement “invokes” the code key.
Key handler hkeyi/.code 2 args={hTEX codei}
As /.code, but this handler defines a key which accepts two arguments. When the so defined key is
used, the two arguments are available as #1 and #2.
4.5. TWO DIMENSIONAL PLOT TYPES 77
Key handler hkeyi/.cd

Each key has a fully qualified name with a (long) prefix, like /pgfplots/xmin. However, if the “current
directory” is /pgfplots, it suffices to write just xmin. The /.cd key handler changes the “current
directory” in this way.
The prefixes /tikz/ and /pgfplots/ are checked automatically for any argument provided to \be-
gin{axis}[hoptionsi] or \addplot. So, you won’t need to worry about them, just omit them – and
look closer in case the package doesn’t identify the option.
4.4.1 Pgfplots and Tik Z Options

This section is more or less technical and can be skipped unless one really wants to know more about this
topic.
Tik Z options and pgfplots options can be mixed inside of the axis arguments and in any of the associated
styles. For example,
\pgfplotsset{every axis legend/.append style={
legend columns=3,font=\Large}}
assigns the ‘legend columns’ option (a pgfplots option) and uses ‘font’ for drawing the legend (a Tik Z
option). The point is: legend columns needs to be known before the legend is typeset whereas font needs
to be active when the legend is typeset. pgfplots sorts out any key dependencies automatically:
The axis environments will process any known pgfplots options, and all ‘every’–styles will be parsed
for pgfplots options. Every unknown option is assumed to be a Tik Z option and will be forwarded to the
associated Tik Z drawing commands. For example, the ‘font=\Large’ above will be used as argument to the
legend matrix, and the ‘font=\Large’ argument in
\pgfplotsset{every axis label/.append style={
ylabel=Error,xlabel=Dof,font=\Large}}
will be used in the nodes for axis labels (but not the axis title, for example).
It is an error if you assign incompatible options to axis labels, for example ‘xmin’ and ‘xmax’ can’t be set
inside of ‘every axis label’.
4.5 Two Dimensional Plot Types

pgfplots supports several two-dimensional line plots like piecewise linear line plots, piecewise constant
plots, smoothed plots, bar plots and comb plots. Most of them use the pgf plot handler library directly, see
[6, section 18.8].
Plot types are part of the plot style, so they are set with options. Most of the basic 2d plot types are
part of Tik Z, see [6, section 18.8], and are probably known to users of Tik Z. They are documented here as
well.
4.5.1 Linear Plots

/tikz/sharp plot (no value)
\addplot+[sharp plot]
Linear (‘sharp’) plots are the default. Point coordinates are simply connected by straight lines.
\begin{axis}
\addplot+[sharp plot] coordinates
{(0,0) (1,2) (2,3)};
2 \end{axis}
\end{tikzpicture}
0
0 0.5 1 1.5 2
The ‘+’ here means to use the normal plot cycle list and append ‘sharp plot’ to its option list.
4.5.2 Smooth Plots

/tikz/smooth (no value)
\addplot+[smooth]
Smooth plots interpolate smoothly between successive points.
\begin{axis}
\addplot+[smooth] coordinates
{(0,0) (1,2) (2,3)};
2 \end{axis}
\end{tikzpicture}
0
0 0.5 1 1.5 2
As described in [6] in all detail, this plot handler results in a “smooth” outline. However, it “not very
intelligent” (compare [6]) and is unrelated to common plot-based interpolation schemes.
/tikz/tension={htensioni} (initially 0.55)

A parameter which controls how the remaining degrees of freedom are fixed: it controls the smoothing
e↵ect. Higher values results in more “rounded” corners whereas low values result in sharp corners.
Please refer to [6] for details.
4.5.3 Constant Plots

Constant plots draw lines parallel to the x-axis to connect coordinates. The discontinuous edges may be
drawn or not, and marks may be placed on left or right ends.
/tikz/const plot (no value)
\addplot+[const plot]
Connects all points with horizontal and vertical lines. Marks are placed left-handed on horizontal line
segments, causing the plot to be right-sided continuous at all data points.
\begin{tikzpicture}
0.6 \begin{axis}
coordinates
{(0,0.1) (0.1,0.15) (0.2,0.5) (0.3,0.62)
0.4 (0.4,0.56) (0.5,0.58) (0.6,0.65) (0.7,0.6)
(0.8,0.58) (0.9,0.55) (1,0.52)};
\end{axis}
\end{tikzpicture}
0.2
0 0.2 0.4 0.6 0.8 1
1
\begin{tikzpicture}
\begin{axis}[ymin=0,ymax=1,enlargelimits=false]
0.8 \addplot
[const plot,fill=blue,draw=black]
coordinates
0.6 {(0,0.1) (0.1,0.15) (0.2,0.5) (0.3,0.62)
(0.4,0.56) (0.5,0.58) (0.6,0.65) (0.7,0.6)
(0.8,0.58) (0.9,0.55) (1,0.52)}
0.4
\closedcycle;
\end{axis}
0.2 \end{tikzpicture}
0
0 0.2 0.4 0.6 0.8 1
/tikz/const plot mark left (no value)
\addplot+[const plot mark left]

An alias for ‘const plot’.
/tikz/const plot mark right (no value)
\addplot+[const plot mark right]

A variant which places marks on the right of each line segment, causing plots to be left-sided continuous
at the given coordinates.
\begin{tikzpicture}
0.6 \begin{axis}
\addplot+[const plot mark right]
coordinates
{(0,0.1) (0.1,0.15) (0.2,0.5) (0.3,0.62)
0.4 (0.4,0.56) (0.5,0.58) (0.6,0.65) (0.7,0.6)
(0.8,0.58) (0.9,0.55) (1,0.52)};
\end{axis}
\end{tikzpicture}
0.2
0 0.2 0.4 0.6 0.8 1
/tikz/const plot mark mid (no value)
\addplot+[const plot mark mid]

A variant which places marks in the middle of each line segment, causing plots to be symmetric around
its data points.
\begin{tikzpicture}
0.6 \begin{axis}
\addplot+[const plot mark mid]
coordinates
{(0,0.1) (0.1,0.15) (0.2,0.5) (0.3,0.62)
0.4 (0.4,0.56) (0.5,0.58) (0.6,0.65) (0.7,0.6)
(0.8,0.58) (0.9,0.55) (1,0.52)};
\end{axis}
\end{tikzpicture}
0.2
0 0.2 0.4 0.6 0.8 1
Note that “symmetric” is only true for constant mesh width: if the x–distances between adjacent
data points di↵er, const plot mark mid will produce vertical lines in the middle between each pair of
consecutive points.
/tikz/jump mark left (no value)
\addplot+[jump mark left]
A variant of ‘const plot mark left’ which does not draw vertical lines.
\begin{axis}[samples=8]
\addplot+[jump mark left,domain=-5:0]
{4*x^2 - 5};
50
\addplot+[jump mark right,domain=-5:0]
{0.7*x^3 + 50};
\end{axis}
0 \end{tikzpicture}
50
4 2 0
/tikz/jump mark right (no value)

\addplot+[jump mark right]
A variant of ‘const plot mark right’ which does not draw vertical lines.
/tikz/jump mark mid (no value)
\addplot+[jump mark mid]
A variant of ‘const plot mark mid’ which does not draw vertical lines.
\begin{tikzpicture}
0.6 \begin{axis}
\addplot+[jump mark mid]
coordinates
{(0,0.1) (0.1,0.15) (0.2,0.5) (0.3,0.62)
0.4 (0.4,0.56) (0.5,0.58) (0.6,0.65) (0.7,0.6)
(0.8,0.58) (0.9,0.55) (1,0.52)};
\end{axis}
\end{tikzpicture}
0.2
0 0.2 0.4 0.6 0.8 1

4.5.4 Bar Plots

Bar plots place horizontal or vertical bars at coordinates. Multiple bar plots in one axis can be stacked on
top of each other or aligned next to each other.
/tikz/xbar (no value)
\addplot+[xbar]
Places horizontal bars between the (y = 0) line and each coordinate.
This option is used on a per-plot basis and configures only the visualization of coordinates. The figure-
wide style /pgfplots/xbar also sets reasonable options for ticks, legends and multiple plots.
\begin{tikzpicture}
\begin{axis}
4 \addplot+[xbar] coordinates
{(4,0) (1,1) (2,2)
(5,3) (6,4) (1,5)};
\end{axis}
\end{tikzpicture}
2
0
2 4 6
Bars are centered at plot coordinates with width bar width. Using bar plots usually means more than
just a di↵erent way of how to connect coordinates, for example to draw ticks outside of the axis, change
the legend’s appearance or introduce shifts if multiple \addplot commands appear.
There is a preconfigured style for xbar which is installed automatically if you provide xbar as argument
to the axis environment which provides this functionality.
\begin{tikzpicture}
\begin{axis}[xbar,enlargelimits=0.15]
\addplot
20 [draw=blue,pattern=horizontal lines light blue]
coordinates
{(10,5) (15,10) (5,15) (24,20) (30,25)};
\addplot
10 [draw=black,pattern=horizontal lines dark blue]
coordinates
{(3,5) (5,10) (15,15) (20,20) (35,25)};
\end{axis}
\end{tikzpicture}
0 10 20 30
Here xbar yields /pgfplots/xbar because it is an argument to the axis, not to a single plot.
For bar plots, it is quite common to provide textual coordinates or even descriptive nodes near the bars.
This can be implemented using the keys symbolic y coords and nodes near coords, respectively:
yes 7
no 3
0 1 2 3 4 5 6 7
#participants
\begin{tikzpicture}
\begin{axis}[
xbar, xmin=0,
width=12cm, height=3.5cm, enlarge y limits=0.5,
xlabel={\#participants},
symbolic y coords={no,yes},
ytick=data,
nodes near coords, nodes near coords align={horizontal},
]
\addplot coordinates {(3,no) (7,yes)};
\end{axis}
\end{tikzpicture}
The symbolic y coords defines a dictionary of accepted coordinates which are then expected in y
coordinates and the nodes near coords key displays values as extra nodes (see their reference docu-
mentations for details). The example employs enlarge y limits in order to get some more free space
since the default spacing is not always appropriate for bar plots.
Note that it might be quite important to include xmin=0 explicitly as in the example above. Without
it, the lower bound will be used:
Uses lowest x coords for xmin
yes 9
no 1
1 2 3 4 5 6 7 8 9
#participants
\begin{tikzpicture}
\begin{axis}[
title=Uses lowest $x$ coords for xmin,
xbar,
width=12cm, height=3.5cm, enlarge y limits=0.5,
xlabel={\#participants},
symbolic y coords={no,yes},
ytick=data,
nodes near coords, nodes near coords align={horizontal},
]
\addplot coordinates {(1,no) (9,yes)};
\end{axis}
\end{tikzpicture}
Besides line, fill, and colorstyles, bars can be configured with bar width and bar shift, see below.
/pgfplots/xbar={hshift for multiple plotsi} (style, default 2pt)

This style sets /tikz/xbar and some commonly used options concerning horizontal bars for the complete
axis. This is automatically done if you provide xbar as argument to an axis argument, see above.
The xbar style defines shifts if multiple plots are placed into one axis. It draws bars adjacent to each
other, separated by hshift for multiple plotsi. Furthermore, it sets the style bar cycle list and sets
tick and legend appearance options.
The style is defined as follows.
\pgfplotsset{
/pgfplots/xbar/.style={
/tikz/xbar,
bar cycle list,
tick align=outside,
xbar legend,
/pgfplots/bar shift auto={#1},
},
}
/pgfplots/bar shift auto={hshift for multiple plotsi} (default 2pt)

The formula for bar shift assigns shifts dependent on the total number of plots and the current
plot’s number. It is designed to fill a total width of n·bar width+(n 1)·hshift for multiple plotsi.
The 0.5 compensates for centering.
The style is defined as
\pgfplotsset{
/pgfplots/bar shift auto/.style={%
/pgf/bar shift={%
% total width = n*w + (n-1)*skip
% -> subtract half for centering
-0.5*(\numplotsofactualtype*\pgfplotbarwidth + (\numplotsofactualtype-1)*(#1)) +
% the ’0.5*w’ is for centering
(.5+\plotnumofactualtype)*\pgfplotbarwidth + \plotnumofactualtype*(#1)},%
},
}
/tikz/ybar (no value)
\addplot+[ybar]
Like xbar, this option generates bar plots. It draws vertical bars between the (x = 0) line and each
input coordinate.
\begin{axis}
\addplot+[ybar] plot coordinates
{(0,3) (1,2) (2,4) (3,1) (4,2)};
3 \end{axis}
\end{tikzpicture}
1
0 1 2 3 4
The example above simply changes how input coordinates shall be visualized. As mentioned for xbar,
one usually needs modified legends and shifts for multiple bars in the same axis.
There is a predefined style which installs these customizations when provided to the axis environment:

\begin{tikzpicture}
\begin{axis}[
x tick label style={
6 /pgf/number format/1000 sep=},
ylabel=Population,
Population
enlargelimits=0.15,
Houses legend style={at={(0.5,-0.15)},
4 anchor=north,legend columns=-1},
ybar,
bar width=7pt,
]
2
\addplot
coordinates {(1930,50e6) (1940,33e6)
(1950,40e6) (1960,50e6) (1970,70e6)};
1930 1940 1950 1960 1970 \addplot
Far Near Here Annot coordinates {(1930,38e6) (1940,42e6)
(1950,43e6) (1960,45e6) (1970,65e6)};
\addplot
coordinates {(1930,15e6) (1940,12e6)
(1950,13e6) (1960,25e6) (1970,35e6)};
\addplot[red,line legend,sharp plot,

update limits=false]
coordinates {(1910,4.3e7) (1990,4.3e7)}
node[above] at (1950,4.3e7) {Houses};
\legend{Far,Near,Here,Annot}
\end{axis}
\end{tikzpicture}
Here, ybar yields /pgfplots/ybar because it is an argument to the axis, not to a single plot. The
style a↵ects the first three \addplot commands. Note that it shifts them horizontally around the plot
coordinates. The fourth \addplot command is some kind of annotation which doesn’t update limits.
The ybar style can be combined with symbolic x coords in a similar way as described for xbar:
10
\begin{axis}[
8 ybar,
7
#participants
enlargelimits=0.15,
legend style={at={(0.5,-0.15)},
ylabel={\#participants},
4 4 4 4
4 symbolic x coords={tool8,tool9,tool10},
xtick=data,
nodes near coords,
2 nodes near coords align={vertical},
1 1 1
]
0 \addplot coordinates {(tool8,7) (tool9,9) (tool10,4)};
\addplot coordinates {(tool8,4) (tool9,4) (tool10,4)};
tool8 tool9 tool10 \addplot coordinates {(tool8,1) (tool9,1) (tool10,1)};
used understood not understood \legend{used,understood,not understood}
\end{axis}
\end{tikzpicture}
As for xbar, the bar width and shift can be configured with bar width and bar shift. However, the
bar shift is better provided as argument to /pgfplots/ybar since this style will overwrite the bar shift.
Thus, prefer /pgfplots/ybar=4pt to set the bar shift.
Sometimes it is useful to write the y values directly near the bars. This can be realized using the nodes
near coords method:

\begin{tikzpicture}
7 \begin{axis}[
6.5 x tick label style={
/pgf/number format/1000 sep=},
6 ylabel=Population,
Population
enlargelimits=0.15,
5 5 legend style={at={(0.5,-0.15)},
anchor=north,legend columns=-1},
4.5 ybar=5pt,% configures ‘bar shift’
4.2 4.3
4 bar width=9pt,
4 3.8 nodes near coords,
3.3 point meta=y *10^-7 % the displayed number
]
\addplot
coordinates {(1930,50e6) (1940,33e6)
1930 1940 1950 1960 1970 (1950,40e6) (1960,50e6) (1970,70e6)};
Far Near
\addplot
coordinates {(1930,38e6) (1940,42e6)
(1950,43e6) (1960,45e6) (1970,65e6)};
\legend{Far,Near}
\end{axis}
\end{tikzpicture}
Any support style changes are possible, of course. A useful example for bar plots might be to use rotated
tick labels:
8 \begin{axis}[
ybar,
#participants
6 enlargelimits=0.15,
4 ylabel={\#participants},
symbolic x coords={excellent,good,neutral,%
2 not good,poor},
2
xtick=data,
0 0 0 nodes near coords,
0 nodes near coords align={vertical},
x tick label style={rotate=45,anchor=east},
]
or
od
od
nt
l
ra
\addplot coordinates {(excellent,0) (good,8)

po
lle
go
go
ut
(neutral,2) (not good,0) (poor,0)};

ce
ne
t
no
ex
\end{axis}
\end{tikzpicture}
/pgfplots/ybar={hshift for multiple plotsi} (style, default 2pt)

As /pgfplots/xbar, this style sets the /tikz/ybar option to draw vertical bars, but it also provides
commonly used options for vertical bars.
If you supply ybar to an axis environment, /pgfplots/ybar will be chosen instead of /tikz/ybar.
It changes the legend, draws ticks outside of the axis lines and draws multiple \addplot arguments
adjacent to each other; block–centered at the x coordinate and separated by hshift for multiple plotsi.
It will also install the bar shift for every node near coord. Furthermore, it installs the style bar
cycle list. It is defined similarly to /pgfplots/xbar.
/pgfplots/bar cycle list (no value)

A style which installs cycle lists for multiple bar plots.
\pgfplotsset{
/pgfplots/bar cycle list/.style={/pgfplots/cycle list={%
{blue,fill=blue!30!white,mark=none},%
{red,fill=red!30!white,mark=none},%
{brown!60!black,fill=brown!30!white,mark=none},%
{black,fill=gray,mark=none},%
}
},
}
/pgf/bar width={hdimension or uniti} (initially 10pt)

Configures the width used by xbar and ybar. It is accepted to provide mathematical expressions.
As of pgfplots 1.7, it is allows to provide an unit as bar width. In this case, the bar width will be
interpreted as axis unit:
\begin{tikzpicture}
\begin{axis}[
25 xbar=0pt,% space of 0pt between adjacent bars
bar width=2,
width=7cm,
height=12cm,
minor y tick num=4,
ytick=data,
20 enlargelimits=0.15]
\addplot
coordinates
{(10,5) (15,10) (5,15) (24,20) (30,25)};
\addplot
coordinates
15 {(3,5) (5,10) (15,15) (20,20) (35,25)};
\end{axis}
\end{tikzpicture}
10
0 10 20 30
In order to interprete arguments as units, you have to write \pgfplotsset{compat=1.7} (or newer)
into your preamble. Older versions will implicitly append the pt suffix if the argument is no dimension.
\pgfplotbarwidth
A mathematical expression which results in the fully computed value of bar width (i.e. it includes
any unit computations).
Note that you may need to enlargelimits in order to see the complete bar – pgfplots will not
automatically update the axis limits to respect bar width.
/pgf/bar shift={hdimension or uniti} (initially 0pt)

Configures a shift for xbar and ybar. Use bar shift together with bar width to draw multiple bar
plots into the same axis. It is accepted to provide mathematical expressions.
As of pgfplots 1.7, it is allows to provide an unit as bar shift. In this case, the bar shift will be
interpreted as axis unit.
\pgfplotbarshift
A mathematical expression which results in the fully computed value of bar shift (i.e. it includes
any unit computations).
Note that you may need to enlargelimits in order to see the complete bar – pgfplots will not
automatically update the axis limits to respect bar shift.
/pgfplots/bar direction=auto|x|y (initially auto)

If pgfplots encounters a value bar width=1 (i.e. without dimension like 1pt), it attempts to evaluate
the bar’s direction.
The default configuration auto assumes that you write something like ybar,bar width=1. In this case,
it is clear that you have a y bar and pgfplots assumes bar direction=y.
However, this context information is unavailable. In this case, you can use the choice x if pgfplots in
unaware that it works on an xbar plot or y if pgfplots is unaware that you meant an ybar plot.
/tikz/ybar interval (no value)
\addplot+[ybar interval]
This plot type produces vertical bars with width (and shift) relatively to intervals of coordinates.
There is one conceptional di↵erence when working with intervals: an interval is defined by two coordi-
nates. Since ybar has one value for each interval, the ith bar is defined by
1. the y value of the ith coordinates,
2. the x value of the ith coordinate as left interval boundary,
3. the x value of the (i + 1)th coordinate as right interval boundary.
Consequently, there is one coordinate too much: the last coordinate will only be used to determine the
interval width; its y value doesn’t influence the bar appearance.
It is installed on a per-plot basis and configures only the visualization of coordinates. See the style
/pgfplots/ybar interval which configures the appearance of the complete figure.
\begin{axis}
\addplot+[ybar interval] plot coordinates
3 {(0,2) (0.1,1) (0.3,0.5) (0.35,4) (0.5,3)
(0.6,2) (0.7,1.5) (1,1.5)};
\end{axis}
2 \end{tikzpicture}
0 0.2 0.4 0.6 0.8 1

\begin{axis}[ybar interval,
xtick=data,
xticklabel interval boundaries,
3 x tick label style=
{rotate=90,anchor=east}
]
2 \addplot coordinates
{(0,2) (0.1,1) (0.3,0.5) (0.35,4) (0.5,3)
(0.6,2) (0.7,1.5) (1,1.5)};
1 \end{axis}
\end{tikzpicture}
0 – 0.1
0.1 – 0.3
0.3 – 0.35
0.35 – 0.5
0.5 – 0.6
0.6 – 0.7
0.7 – 1

\begin{tikzpicture}
\begin{axis}[
6 /pgf/number format/1000 sep=},
ylabel=Population,
Population
enlargelimits=0.05,
ybar interval=0.7,
]
\addplot
2 coordinates {(1930,50e6) (1940,33e6)
(1950,40e6) (1960,50e6) (1970,70e6)};
\addplot
1930 1940 1950 1960 coordinates {(1930,38e6) (1940,42e6)
Far Near Here (1950,43e6) (1960,45e6) (1970,65e6)};
\addplot
coordinates {(1930,15e6) (1940,12e6)
(1950,13e6) (1960,25e6) (1970,35e6)};
\legend{Far,Near,Here}
\end{axis}
\end{tikzpicture}
/pgfplots/ybar interval={hrelative widthi} (style, default 1)

A style which is intended to install options for ybar interval for a complete figure. This includes tick
and legend appearance, management of multiple bar plots in one figure and a more adequate cycle
list using the style bar cycle list.
/tikz/xbar interval (no value)
\addplot+[xbar interval]
As ybar interval, just for horizontal bars.
\begin{tikzpicture}
50 – 70 \begin{axis}[
xmin=0,xmax=53,
ylabel=Age,
xlabel=Quantity,
30 – 50
Age
ytick=data,
yticklabel interval boundaries,
25 – 30 xbar interval,
21 – 25
18 – 21 ]
13 – 18 \addplot
10 – 13 coordinates {(10,5) (10.5,10) (15,13)
5 – 10 (24,18) (50,21) (23,25) (10,30)
0 10 20 30 40 50 (3,50) (3,70)};
Quantity \end{axis}
\end{tikzpicture}
/pgfplots/xbar interval={hrelative widthi} (style, default 1)

A style which is intended to install options for xbar interval for a complete figure, see the style
/pgfplots/ybar interval for details.
/pgfplots/xticklabel interval boundaries (no value)

/pgfplots/yticklabel interval boundaries (no value)
/pgfplots/zticklabel interval boundaries (no value)
These are style keys which set x tick label as interval (see page 320 for details) and configure the
tick appearance to be hstarti – hend i for each tick interval.
4.5.5 Histograms
This section has been moved to the statistics library, see Section 5.12.2 on page 480.
4.5.6 Box Plots

This section has been moved to the statistics library, see Section 5.12.1 on page 469.
4.5.7 Comb Plots

Comb plots are very similar to bar plots except that they employ single horizontal/vertical lines instead of
rectangles.
/tikz/xcomb (no value)
\addplot+[xcomb]
\begin{tikzpicture}
\begin{axis}
4 \addplot+[xcomb] coordinates
{(4,0) (1,1) (2,2)
(5,3) (6,4) (1,5)};
\end{axis}
\end{tikzpicture}
2
0
2 4 6
/tikz/ycomb (no value)

\addplot+[ycomb]
\begin{axis}
\addplot+[ycomb] plot coordinates
{(0,3) (1,2) (2,4) (3,1) (4,2)};
3 \end{axis}
\end{tikzpicture}
1
0 1 2 3 4
4.5.8 Quiver Plots (Arrows)

/pgfplots/quiver={hoptions with ‘quiver/’ prefixi}
\addplot+[quiver={hoptions with ‘quiver/’ prefixi}]
A plot type which draws small arrows, starting at (x, y), in direction of (u, v).
\begin{tikzpicture}
\begin{axis}
30
\addplot[blue,
quiver={u=1,v=2*x},
-stealth,samples=15] {x^2};
20 \end{axis}
\end{tikzpicture}
10
6 4 2 0 2 4 6
The base point (x, y) is provided as before; in the example above, it is generated by plot expression
and yields (x, x2 ). The vector direction (u, v) needs to be given in addition. Our example with quiv-
er/u=1 and quiver/v=2*x results in u = 1 and v = 2x. Thus, we have defined and visualized a vector
field for the derivative of f (x) = x2 .
A common example is to visualize the gradient (@x f, @y f )(x, y) of a two–dimensional function f (x, y):
x exp( x2 y 2 ) and its gradient \begin{tikzpicture}
2 \begin{axis}[
title={$x \exp(-x^2-y^2)$ and its gradient},
domain=-2:2,
view={0}{90},
1 axis background/.style={fill=white},
]
\addplot3[contour gnuplot={number=9,
0 labels=false},thick]
{exp(0-x^2-y^2)*x};
\addplot3[blue,
1 quiver={
u={exp(0-x^2-y^2)*(1-2*x^2)},
v={exp(0-x^2-y^2)*(-2*x*y)},
scale arrows=0.3,
2 },
2 1 0 1 2
-stealth,samples=15]
{exp(0-x^2-y^2)*x};
\end{axis}
\end{tikzpicture}
The example visualizes f (x, y) = x exp( x2 y 2 ) using contour gnuplot as first step. The options
contour/number and contour/labels provide fine-tuning for the contour and are not of interest here
(so is the axis background which just improves visibility of contour lines). What we are interested
in is the quiver= style: it defines u and v to some two–dimensional expressions. Furthermore, we
used quiver/scale arrows to reduce the arrow size. The -stealth is a Tik Z style which configures
outgoing arrow styles of type ‘stealth’. The samples=15 key configures how we get our input data. In
our case, we have input data (xi , yj , f (xi , yj )) with 15 samples for each, i and j.
It is also possible to place quiver plots on a prescribed z value:
\begin{tikzpicture}
\begin{axis}[
domain=0:1,
xmax=1,
1 ymax=1,
]
\addplot3[surf] {x*y};
0.5 \addplot3[blue,/pgfplots/quiver,
quiver/u=y,
1 quiver/v=x,
0 quiver/w=0,
0 0.2 0.4 0.5 quiver/scale arrows=0.1,
0.6 0.8 -stealth,samples=10] {1};
1 0 \end{axis}
\end{tikzpicture}
Here, the quiver plots is placed on top of a surf. It visualizes the gradient (using a common scale factor of
1/10 to reduce the arrow lengths). The quiver/w=0 means that arrows have no z di↵erence, and the {1}
argument indicates that all start at (xi , yj , 1). Here, the values (xi , yj ) are sampled in the domain=0:1
argument (with samples=10), i.e. arrows start at (xi , yj , 1) and end at (xi + yj /10, yj + xi /10, 1).
So far, quiver plots do not assume a special sequence of input points. This has two consequences: first,
you can plot any vector field by considering just (x, y) + (u, v) (or (x, y, z) + (u, v, w)) – the data doesn’t
necessarily need to be a two–dimensional function (as opposed to surf etc). On the other hand, you
need to provide quiver/scale arrows manually since quiver doesn’t know the mesh width in case you
provide matrix data13 .
Note that quiver plots are currently not available together with logarithmic axes.
/pgfplots/quiver/u=hexpressioni (initially 0)
/pgfplots/quiver/v=hexpressioni (initially 0)
/pgfplots/quiver/w=hexpressioni (initially 0)
These keys define how the vector directions (u, v) (or, for three dimensional plots, (u, v, w)) shall
be set.
The hexpressioni can be a constant expression like quiver/u=1 or quiver/u=42*5. It may also
depend on the final base point values using the values x, y or z as in the example above. In this
context, x yields the x coordinate of the point where the vector starts, y the y coordinate and so
on.
Attention: the fact that x refers to the final x coordinate means that parametric plots should
use t as variable 14 . Consider the following example:
13 Actually, I might add something like quiver/scale arrows=auto in the future, I don’t know yet. Loops through input data
are slow in TEX, automatic mesh widths computation even more...

14 Sorry for this usability issue.
\begin{tikzpicture}
\begin{axis}[axis equal,
axis lines=middle,
axis line style={->},
tick style={color=black},
xtick=\empty,
ytick=\empty
]
\addplot[samples=20, domain=0:2*pi,
% the default choice ’variable=\x’ leads to
% unexpected results here!
variable=\t,
quiver={
u={-sin(deg(t))},
v={cos(deg(t))},
scale arrows=0.5},
->,blue]
({cos(deg(t))}, {sin(deg(t))});
\addplot[samples=100, domain=0:2*pi]
({cos(deg(x))}, {sin(deg(x))});
\end{axis}
\end{tikzpicture}
Here, a parametric plot is used to draw a circle and tangent vectors. The choice variable=\t plays
a functional role besides naming conventions: it allows to access the parametric variable within the
expressions for both u and v. On the other hand, we could have used u=y and v=-x since x expands
to the x coordinate with value sin(deg(t)) and y expands to the y coordinate cos(deg(t)).
Another important application is to use table column references like quiver/u=\thisrow{col} in
conjunction with \addplot table:
Quiver and plot table \begin{tikzpicture}
\begin{axis}[title=Quiver and plot table]
\addplot[blue,
quiver={u=\thisrow{u},v=\thisrow{v}},
20 -stealth]
table
{
x y u v
0 0 1 0
10 1 1 1 1
2 4 1 4
3 9 1 6
4 16 1 8
0 };
\end{axis}
Here, the hexpressioni employs \thisrow which always refers to the actual row of \addplot table.
Note that hexpressioni should always be of numeric type (no symbolic input extensions are sup-
ported currently).
/pgfplots/quiver/u value={hvaluei} (initially 0)

/pgfplots/quiver/v value={hvaluei} (initially 0)
/pgfplots/quiver/w value={hvaluei} (initially 0)
These keys have the same function as quiver/u and its variants. However, they don’t call the
math parser, so only single values are allowed (including something like \thisrow{columnname}).
/pgfplots/quiver/colored (no value)

/pgfplots/quiver/colored={hcolor i}
Allows to define an individual color for each arrow. Omitting the argument ‘hcolor i’ is identical to
quiver/colored=mapped color which uses the point meta to get colors for every arrow.
If you just want to set the same color for every arrow, prefer using \addplot[blue,quiver] which
is more efficient.
/pgfplots/quiver/scale arrows={hscalei} (initially 1)

Allows to rescale the arrows by a factor. This may be necessary if the arrow length is longer
than the distance between adjacent base points (xi , yi ). There may come a feature to rescale them
automatically.
/pgfplots/quiver/update limits=true|false (initially true)

A boolean indicating whether points (x, y) + (u, v) shall contribute to the axis limits.
/pgfplots/quiver/every arrow (style, initially empty)

Allows to provide individual arrow styles.
The style can contain any Tik Z drawing option. It will be evaluated for every individual arrow and
may depend upon anything which is available at visualization time.
In particular, this includes point meta data, typically using \pgfplotspointmetatransformed
2 [0, 1000] where 0 corresponds to point meta min and 1000 corresponds to point meta max:
Thickness indicates “strength”. \begin{tikzpicture}
% define some constants:
\def\U{1}
\def\V{2*x}
\def\LEN{(sqrt((\U)^2 + (\V)^2)}
20
\begin{axis}[axis equal image,
title=Thickness indicates ‘‘strength’’.
]
10 \addplot[blue,
point meta={\LEN},
quiver={
u={(\U)/\LEN}, v={(\V)/\LEN},
0 scale arrows=2,
every arrow/.append style={
5 0 5 line width=2pt*\pgfplotspointmetatransformed/1000
},
},
-stealth,samples=15,
] {x^2};
\end{axis}
\end{tikzpicture}
In the example, we have some 2d vector field stored in helper constants \U and \V. The length of
each vector is stored in \LEN here. The quiver plot as such contains unit length vectors – and the
\LEN enters an every arrow style to get varying line width.
An every arrow style might also depend upon mapped color (provided point meta has been set).
Again, if you do not need individual arrow styles, prefer using a plot style (cycle list or argument
to \addplot) which is more efficient.
/pgfplots/quiver/before arrow/.code={h... i}
/pgfplots/quiver/after arrow/.code={h... i}
Advanced keys for more fine tuning of the display. They allow to install some TEX code manually
before or after the drawing operations for single arrows. Both are initially empty.
/pgfplots/quiver/quiver legend (style, no value)

A style which redefines legend image code in order to produce a suitable legend for quiver plots.
It is implicitly activated whenever quiver plot handlers are selected.
Legend
0
\begin{axis}[tiny]
\addplot[blue,
100 quiver={u=1,v=3*x},
-stealth,
6 4 2 0 2 4 6 samples=15]
{x^3};
\addlegendentry{Legend}
\end{axis}
\end{tikzpicture}
4.5.9 Stacked Plots

/pgfplots/stack plots=x|y|false (initially false)
Allows stacking of plots in either x or y direction. Stacking means to add either x- or y coordinates of
successive \addplot commands on top of each other.
\begin{axis}[stack plots=y]
\addplot coordinates
{(0,1) (1,1) (2,2) (3,2)};
{(0,1) (1,1) (2,2) (3,2)};
{(0,1) (1,1) (2,2) (3,2)};
\end{axis}
2 \end{tikzpicture}
0 1 2 3
The current implementation for stack plots does not interpolate missing coordinates. That means
stacking will fail if the plots have di↵erent grids.
/pgfplots/ybar stacked=plus|minus (style, default plus)

The plot handler stack plots is particularly useful for bar plots. There are two possible modes of
operation: the first is to set stack plots=y,/tikz/ybar. It activates just these two features without
making them aware of each other. The second is to set ybar stacked which activates the two features
and makes them aware of each other.
If you use stack plots together with /tikz/ybar, you have kind of a low–level implementation which
is kind of “raw”:
\begin{tikzpicture}
8 \begin{axis}[stack plots=y,/tikz/ybar]
{(0,1) (1,1) (2,3) (3,2) (4,1.5)};
{(0,1) (1,1) (2,3) (3,2) (4,1.5)};
4 {(0,1) (1,1) (2,3) (3,2) (4,1.5)};
\end{axis}
\end{tikzpicture}
2
0 1 2 3 4
Using ybar stacked enables stacked vertical bars (i.e. ybar and stack plots=y) and it also adjusts
the legend and tick appearance and assigns a useful cycle list. To this end, it should be given as
option to the axis:
\begin{tikzpicture}
8 \begin{axis}[ybar stacked]
{(0,1) (1,1) (2,3) (3,2) (4,1.5)};
{(0,1) (1,1) (2,3) (3,2) (4,1.5)};
4 {(0,1) (1,1) (2,3) (3,2) (4,1.5)};
\end{axis}
\end{tikzpicture}
2
0 1 2 3 4
\begin{axis}[
ybar stacked,
#participants
enlargelimits=0.15,
5
symbolic x coords={tool1, tool2, tool3, tool4,
tool5, tool6, tool7},
xtick=data,
0 ]
\addplot+[ybar] plot coordinates {(tool1,0) (tool2,2)
1
2
3
4
5
6
7
(tool3,2) (tool4,3) (tool5,0) (tool6,2) (tool7,0)};

ol
ol
ol
ol
ol
ol
ol
to
to
to
to
to
to
to

never rarely sometimes often \addplot+[ybar] plot coordinates {(tool1,6) (tool2,6)
\legend{never, rarely, sometimes, often}
\end{axis}
\end{tikzpicture}
/pgfplots/xbar stacked=plus|minus (style, default plus)

The same remarks as for ybar stacked hold for xbar stacked as well: xbar stacked is a figure-wide
style which enables stacked horizontal bars (i.e. xbar and stack plots=x). It also adjusts the legend
and tick appearance and assigns a useful cycle list.
Consequently, one can have a “raw” picture which combines stacking and bars as in the following picture
(i.e. without xbar stacked):
\begin{axis}[stack plots=x,/tikz/xbar]
{(1,0) (2,1) (2,2) (3,3)};
{(1,0) (2,1) (2,2) (3,3)};
1 {(1,0) (2,1) (2,2) (3,3)};
\end{axis}
\end{tikzpicture}
0
2 4 6 8
Alternatively, one activates xbar stacked right after \begin{axis} and benefits from several style
adoptions.
\begin{axis}[xbar stacked]
{(1,0) (2,1) (2,2) (3,3)};
{(1,0) (2,1) (2,2) (3,3)};
1 {(1,0) (2,1) (2,2) (3,3)};
\end{axis}
\end{tikzpicture}
0
2 4 6 8
/pgfplots/stack dir=plus|minus (initially plus)

Configures the direction of stack plots. The value plus will add coordinates of successive plots while
minus subtracts them.
/pgfplots/reverse stacked plots=true|false (initially true, default true)

Configures the sequence in which stacked plots are drawn. This is more or less a technical detail which
should not be changed in any normal case.
The motivation is as follows: suppose multiple \addplot commands are stacked on top of each other
and they are processed in the order of appearance. Then, the second plot could easily draw its lines (or
fill area) on top of the first one - hiding its marker or line completely. Therefor, pgfplots reverses the
sequence of drawing commands.
This has the side-e↵ect that any normal Tik Z-paths inside of an axis will also be processed in reverse
sequence.
/pgfplots/stacked ignores zero=true|false (initially true, default true)

Configures stacked plots to ignore “zero” increments.
In this context, “ignore” means to suppress visualization if an increment vanishes.
Configuring \pgfplotsset{compat=1.9} (or higher) activates this feature for xbar stacked and ybar
stacked.
/pgfplots/stacked ignores zero/default=true|false (initially true)

/pgfplots/stacked ignores zero/markers=true|false (initially true)
/pgfplots/stacked ignores zero/errorbars=true|false (initially false)
A detail key which allows to customize how and where to apply stacked ignores zero. This key is
ignored unless stacked ignores zero=true. Its default configuration is to suppress visualization
of an empty increment for the standard visualization and for markers. Error bars will be displayed,
though (the error bar is typically non-empty even if the increment is 0).
/pgfplots/xbar interval stacked=plus|minus (style, default plus)

A style similar to /pgfplots/xbar stacked for the interval based bar plot variant.
/pgfplots/ybar interval stacked=plus|minus (style, default plus)

A style similar to /pgfplots/ybar stacked for the interval based bar plot variant.
Stacked Bar Plots and Nodes Near Coords

It is possible to combine ybar stacked and xbar stacked with nodes near coords. In contrast to non-
stacked plots, it appears to be of limited use to draw a node near the top of the stack. Instead, one typically
wants to see the di↵erence added in each stacking step. To this end, pgfplots will automatically reconfigure
nodes near coords to display the added values in the center of each new bar:
% needs \pgfplotsset{compat=1.9} or newer!
3 \begin{axis}[ybar stacked,nodes near coords,
bar width=0.4]
{(0,1) (1,1) (2,3) (3,2) (4,1.5)};
2
4 1.5 {(0,1) (1,1) (2,3) (3,2) (4,1.5)};
1 1 1.5 {(0,1) (1,1) (2,3) (3,2) (4,1.5)};
2 \end{axis}
1 1 3
2 1.5 \end{tikzpicture}
1 1
0 1 2 3 4
10 % needs \pgfplotsset{compat=1.9} or newer!

2 2 2 \begin{tikzpicture}
4 3 4 \begin{axis}[
#participants
2 ybar stacked,
bar width=15pt,
8 5 nodes near coords,
5 6 3 enlargelimits=0.15,
6 legend style={at={(0.5,-0.20)},
6 1 6 anchor=north,legend columns=-1},
2 2 3 2
0 1 symbolic x coords={tool1, tool2, tool3, tool4,
tool5, tool6, tool7},
xtick=data,
1
2
3
4
5
6
7

ol
ol
ol
ol
ol
ol
ol
to
to
to
to
to
to
to
]
never rarely sometimes often (tool3,2) (tool4,3) (tool5,0) (tool6,2) (tool7,0)};
\legend{never, rarely, sometimes, often}
\end{axis}
\end{tikzpicture}
Note that the preceding example contains no nodes for coordinates with value 0. This is due to the key
stacked ignores zero which is active if compat=1.9 or newer: empty increments will be discarded.
This automatic reconfiguration is essentially part of the styles xbar stacked or ybar stacked: both
reconfigure nodes neard coords. Note that this feature has been introduced in version 1.9. In order to
maintain backwards compatible to previous work-arounds, you have to write compat=1.9 to get these e↵ects.
3 3 3 3 \begin{tikzpicture}
\begin{axis}[xbar stacked,nodes near coords,
bar width=0.5]
2 2 2 2 {(1,0) (2,1) (2,2) (3,3)};
{(1,0) (2,1) (2,2) (3,3)};
1 2 2 2 \addplot coordinates
{(1,0) (2,1) (2,2) (3,3)};
\end{axis}
\end{tikzpicture}
0 1 1 1
2 4 6 8
Stacked Plots with Negative Values

The meaning of negative values in xbar stacked or ybar stacked and their visualization depends on the
use-case, and they need to be defined. pgfplots o↵ers two di↵erent interpretations of negative values in
the context of stacked plots:
/pgfplots/stack negative=on previous|separate (initially separate)

The choice on previous will simply sum any values as they are found in the input data. Thus, if you
have four plots which are stacked on top of each other, and the first has value 10, the second value 50,
and the third value 10, and the fourth has 30, the final value will be 10 + 50 10 + 30 = 80. However,
visualizing a bar of negative size as a usual xbar stacked graph is hard to read and requires vertical
o↵sets in this context. This can be accomplished as follows:
\begin{tikzpicture}
60
\pgfplotstableread{
15
Year Cat1 Cat2 Cat3 Cat4
2007 60 2005 10 50 -10 30
20 2006 -40 60 -15 90
2007 -20 60 -15 60
}\mytable
90
15 \begin{axis}[
2006 60
xbar stacked,
stack negative=on previous,
40
%
xmajorgrids,
30 legend style={at={(0.5,-0.1)},anchor=north,legend
10
columns=-1},
2005 bar width=10pt,
50
bar shift auto,
10 nodes near coords,
nodes near coords style={font=\tiny},
y=60pt,
50 0 50 100 ytick distance=1,
enlarge y limits=0.3,
Cat1 Cat2 Cat3 Cat4 /pgf/number format/1000 sep=,
extra x ticks={0},
extra x tick style={grid
style={black},xticklabel=\empty},
]
\addplot table [x index=1,y=Year] {\mytable};
\legend{Cat1,Cat2,Cat3,Cat4}
\end{axis}
\end{tikzpicture}
In this case, the values for y = 2005 resemble our input case of 10 + 50 10 + 30 = 80. It is a
“normal” xbar stacked with the configuration stack negative=on previous – but with the special
key bar shift auto which is also used in order to create grouped xbar charts: every \addplot receives
a vertical o↵set.
Despite the similiarity with “waterfall charts”, this is not indented to be a waterfall chart. At the time
of this writing, pgfplots has no support for waterfall charts.
The choice stack negative=on previous is the initial value for
• all but stacked bar plots,
• all plots if compat is less than 1.13.
As of compat=1.13, the initial value of stack negative is separate for xbar stacked, ybar stacked,
and their * interval based variants.
The alternative choice separate tracks the start points of bars separately for negative and non-negative
values:
2007 15 20 60 60
\begin{tikzpicture}
\pgfplotstableread{
Year Cat1 Cat2 Cat3 Cat4
2005 10 50 -10 30
2006 -40 60 -15 90
2007 -20 60 -15 60
2006 15 40 60 90
}\mytable
\begin{axis}[
xbar stacked,
% is default anyway:
2005 10
10 50 30 stack negative=separate,
%
50 0 50 100 150 /pgf/number format/1000 sep=,
Cat1 Cat2 Cat3 Cat4 xmajorgrids,
nodes near coords,
nodes near coords style={font=\tiny},
ytick distance=1,
legend style={at={(0.5,-0.1)},anchor=north,legend
columns=-1},
extra x ticks={0},
extra x tick style={grid
style={black},xticklabel=\empty},
]
\legend{Cat1,Cat2,Cat3,Cat4}
\end{axis}
\end{tikzpicture}
In this case, positive contributions extend into the positive x axis whereas negative contributions extend
into the negative x axis. There is no final sum (or better: there are two final sums). The e↵ect is as if
you would have defined two separate axes, one for the positive contributions and one for the negative
ones and aligned them at the origin.
As of compat=1.13, stack negative=separate is the initial setting for stacked bar plots. Older com-
patibility levels use stack negative=on previous.
4.5.10 Area Plots

Area plots means two–dimensional plots in which an area is filled between a couple of curves. Note that this
is a rather incomplete characterization as mesh, surface, and patch plots are, of course, also “area plots”
of some sort.
This section covers two types of area plots, namely those which are defined by stack plots and those
which are defined using \addplot fill between.
Filling Using Stacked Plots

The first (and older) option to fill areas between plots is a combination of \closedcycle and stack plots.
They can be combined with any other plot type.
\begin{axis}[
stack plots=y,
area style,
4 enlarge x limits=false]
{(0,1) (1,1) (2,2) (3,2)}
\closedcycle;
2 {(0,1) (1,1) (2,2) (3,2)}
\closedcycle;
{(0,1) (1,1) (2,2) (3,2)}
0 1 2 3 \closedcycle;
\end{axis}
\end{tikzpicture}
The main property of this kind of area visualization is that all plots of an axis are taken together, and since
they are stacked, they form areas.
Area plots may need modified legends, for example using the area legend key. Furthermore, one may want
to consider the axis on top key such that filled areas do not overlap ticks and grid lines since such plots
typically cover huge areas of the axis.
Note that Area plots which rely on stack plots have one severe limitation: stacking works if and only
if each plot has the same coordinates (in our case of stack plots=y, each plot has the same x coordinates).
/pgfplots/area style (style, no value)

A style which sets
\pgfplotsset{
/pgfplots/area style/.style={%
area cycle list,
area legend,
axis on top,
}}
/pgfplots/area cycle list (style, no value)

A style which installs a cycle list suitable for area plots. The initial configuration of this style simply
invokes the bar cycle list which does also provide filled plot styles.
\begin{axis}[
const plot,
stack plots=y,
4 area style,
enlarge x limits=false]
{(0,1) (1,1) (2,2) (3,2)}
\closedcycle;
{(0,1) (1,1) (2,2) (3,2)}
\closedcycle;
0 1 2 3 {(0,1) (1,1) (2,2) (3,2)}
\closedcycle;
\end{axis}
\end{tikzpicture}
\begin{axis}[
smooth,
stack plots=y,
4 area style,
enlarge x limits=false]
{(0,1) (1,1) (2,2) (3,2)}
\closedcycle;
{(0,1) (1,1) (2,2) (3,2)}
\closedcycle;
0 1 2 3 {(0,1) (1,1) (2,2) (3,2)}
\closedcycle;
\end{axis}
\end{tikzpicture}
time 1minload nodes cpus processes memused memcached membuf memtotal

0 18 100 200 20 15 45 1 150
1 25 100 200 30 20 45 2 150
2 25 100 200 30 21 42 2 150
3 30 100 200 30 20 40 2 150
4 30 100 200 30 19 40 1 150
5 80 100 200 30 20 40 3 150
6 120 100 200 10 3 40 3 150
7 180 100 200 10 4 41 3 150
8 183 100 200 10 3 42 2 150
9 178 100 200 10 2 41 1 150
10 180 100 200 20 15 45 2 150
11 184 100 200 20 20 45 3 150
12 170 100 200 20 22 47 4 150
13 164 100 200 20 24 50 4 150
14 150 100 200 20 25 52 3 150
15 148 100 200 20 26 53 2 150
16 149 100 200 30 30 54 2 150
17 154 100 200 30 35 55 1 150
\pgfplotstableread{pgfplots.timeseries.dat}\loadedtable
\pgfplotstabletypeset\loadedtable
200
150
100
50
0
0 5 10 15
1min load nodes cpus processes

\pgfplotstableread
{pgfplots.timeseries.dat}
{\loadedtable}
\begin{tikzpicture}
\begin{axis}[
ymin=0,
minor tick num=4,
enlarge x limits=false,
axis on top,
every axis plot post/.append style=
{mark=none},
const plot,
legend style={
area legend,
at={(0.5,-0.15)},
anchor=north,
legend columns=-1}]
\addplot[draw=blue,fill=blue!30!white]
table[x=time,y=1minload] from \loadedtable
\closedcycle;
\addplot table[x=time,y=nodes] from \loadedtable;
\addplot table[x=time,y=cpus] from \loadedtable;
\addplot table[x=time,y=processes]
from \loadedtable;
\legend{1min load,nodes,cpus,processes}
\end{axis}
\end{tikzpicture}
150
Mem [GB]
100
50
0
0 5 10 15
Memory used Memory cached

Memory bu↵ered Total memory
\pgfplotstableread{pgfplots.timeseries.dat}\loadedtable
\begin{tikzpicture}
\begin{axis}[
ymin=0,
minor tick num=4,
const plot,
axis on top,
stack plots=y,
cycle list={%
{blue!70!black,fill=blue},%
{blue!60!white,fill=blue!30!white},%
{draw=none,fill={rgb:red,138;green,82;blue,232}},%
{red,thick}%
},
ylabel={Mem [GB]},
legend style={
area legend,
at={(0.5,-0.15)},
anchor=north,
legend columns=2}]
\addplot table[x=time,y=memused] from \loadedtable \closedcycle;

\addplot table[x=time,y=memcached] from \loadedtable \closedcycle;
\addplot table[x=time,y=membuf] from \loadedtable \closedcycle;
\addplot+[stack plots=false]
table[x=time,y=memtotal] from \loadedtable;
\legend{Memory used,Memory cached,Memory buffered,Total memory}
\end{axis}
\end{tikzpicture}
Filling Under/Between Named Plots

The second way to fill areas under or between plots is the fillbetween library shipped with pgfplots.
Tik Z allows to name a plot or a path using name path. The fillbetween library o↵ers the special “plot
type” \addplot fill between which expects two such named paths on input. Then, it generates filled
areas between these named paths.
The biggest di↵erence to Area plots by means of stack plots as discussed in the previous section is
that \addplot fill between accepts arbitrary input plots: the coordinates can vary, the ranges can vary,
even smooth and sharp plot can be combined. In addition, \addplot fill between allows to select only
sub–regions between the two input paths by means of its soft clip feature, and it accepts individually
styled fill segments using its styles every even segment, every segment no hindex i, etc.
The reference documentation is Section 5.7, but the following examples attempt to cover basic use–cases
around fill–between. Please refer to the reference documentation in Section 5.7 for an in–depth discussion
of every involved key.
The following advanced example selects two partial areas using soft clip in order to fill these areas:
% requires \usepgfplotslibrary{fillbetween}
\begin{tikzpicture}
20 \begin{axis}
\addplot+[name path=A,domain=0:5] {x^2};
\path[name path=B]
(\pgfkeysvalueof{/pgfplots/xmin},0) --
10 (\pgfkeysvalueof{/pgfplots/xmax},0);
\addplot[gray] fill between[of=A and B,

soft clip={domain=3:4},
0 ];
0 2 4 \addplot[gray!30] fill between[of=A and B,

];
\end{axis}
\end{tikzpicture}
The preceding example is characterized by two paths, each of which is named using name path. The
path named ‘A’ is a plot of x2 . The other path named ‘B’ is actually nothing but the visible x axis, the
\path instruction means that it is not drawn at all. Note that \pgfkeysvalueof{/pgfplots/xmin} expands
to the visible lower x axis limit. Finally, we see two \addplot fill between instructions, both with the
selection of=A and B and both with (di↵erent) values of soft clip. The soft clip key restricts the filled
segments to the bounding rectangle defined by some path hlower corner i rectangle hupper corner i where
the two arguments are the lower– and upper corner of an invisible rectangular bounding box. In our case,
this bounding box has been defined by some domain restriction and the upper and lower axis limits. Only
those parts of the filled segment makes up the result of \addplot fill between.
This restriction to a specific interval (or better: rectangle) is also available when stroking the involved
paths ‘A’ or ‘B’: pgfplots comes with decoration=soft clip which can be combined with postaction as
follows.
\begin{tikzpicture}
20 \begin{axis}
% define some common macro
\def\clippath{
(3,-1)
rectangle
10 (4,100)
}
\draw[help lines] \clippath;
0 \addplot+[name path=A,domain=0:5,
mark=none,
0 2 4 postaction={decorate,red,line width=2pt,
decoration={
soft clip,
soft clip path={\clippath},
},
},
] {x^2};
\path[name path=B]
(\pgfkeysvalueof{/pgfplots/xmax},0);
\addplot[orange] fill between[of=A and B,

soft clip={\clippath},
];
\end{axis}
\end{tikzpicture}
The preceding example has a \def\clippath which defines a re-usable bounding box. Here, the fill
between statement is similar to the preceding example, with the only di↵erence that our macro is used as
value. The main di↵erence is the postaction in plot ‘A’. A postaction is a Tik Z construct which allows to
draw the same path twice. The second time the plot is drawn makes use of the option list after postaction.
In our case, it is a decoration which will be drawn in red with thick lines. The decoration applies a soft
clip path. In this context, soft clip is applied to one path only, but the idea is the same: only the parts
within the soft clip argument are displayed.
Note that fill between is also possible for arbitrary plots, not just the x axis. In fact, one can combine
any two plots.
\begin{tikzpicture}
20 \begin{axis}
\addplot[blue,name path=A,domain=0:5] {x^2};
\addplot[red,name path=B,smooth] table {

x y
10 0 10
1 8
2 6
3 1
0 4 5
};
0 2 4
\addplot[gray!50] fill between[of=A and B,
];
\addplot[gray!10,draw=black] fill between[

of=A and B,
soft clip={domain=1:2.5},
];
\end{axis}
\end{tikzpicture}
The preceding example has two plots with a di↵erent number of samples, a di↵erent range, and a di↵erent
input method. Nevertheless, the area between these two plots can be filled just as in the previous examples,
including any intersections.
It is also possible to define individual styles for the filled segments. To this end, one has to set the split
key which activates separate output paths. This enables the use of special styles which is shown in in the
following example. Here, we have two styles with explicitly numbered segment indices:
\begin{tikzpicture}
20 \begin{axis}
\addplot[blue,name path=A,domain=0:5] {x^2};
\addplot[red,name path=B,smooth] table {

x y
10 0 10
1 8
2 6
3 1
0 4 5
};
0 2 4
split,
every segment no 0/.style=
{pattern color=gray!50,
pattern=north east lines},
{pattern=north west lines},
];
\end{axis}
\end{tikzpicture}
Note that the example has no soft clip argument. Consequently, it fills between the start– and end points
of the involved paths. The split argument combined with the styles yields individually drawn fill segments.
Further reading: see Section 5.7.
4.5.11 Closing Plots (Filling the Area Under Plots)

\closedcycle
Provide \closedcycle as htrailing path commandsi after \addplot to draw a closed line from the last
plot coordinate to the first one.
Use \closedcycle whenever you intend to fill the area under a plot.
\begin{tikzpicture}
\begin{axis}
\addplot {x^2+2} \closedcycle;
20 \end{axis}
\end{tikzpicture}
10
0
6 4 2 0 2 4 6
\begin{tikzpicture}
\begin{axis}
\addplot+[fill] {x^2+2} \closedcycle;
20 \end{axis}
\end{tikzpicture}
10
0
6 4 2 0 2 4 6
In case of stacked plots, \closedcycle connects the current plot with the previous plot instead of
connecting with the x axis15 .
\begin{axis}[stack plots=y]
\addplot+[fill] coordinates
{(0,1) (1,1) (2,2) (3,2)} \closedcycle;
3 \addplot+[fill] coordinates
{(0,1) (1,1) (2,2) (3,2)} \closedcycle;
\end{axis}
2 \end{tikzpicture}
1
0 1 2 3
Closing a plot is also possible for three–dimensional axes, see Section 4.6.3 on page 129.
Note that \closedcycle has been designed for functions (i.e. for a plot where every x has at most one y
value). For arbitrary curves, you can safely use the Tik Z path --cycle instead which simply connects the
last and the first path element:
15 The implementation for stacked plots requires some additional logic to determine the filled area: \closedcycle will produce
a plot coordinates command with reversed coordinates of the previous plot. This is usually irrelevant for end users, but it
assumes that the plot’s type is symmetric. Since constant plots are inherently asymmetric, \closedcycle will use const plot
mark right as reversed sequence for const plot mark left.
\begin{axis}
{(0,1) (1,2) (0,3) (-1,2)};
\end{axis}
\end{tikzpicture}
2
1
1 0.5 0 0.5 1
\begin{axis}
{(0,1) (1,2) (0,3) (-1,2)} --cycle;
\end{axis}
\end{tikzpicture}
2
1
1 0.5 0 0.5 1
\begin{axis}
\addplot+[fill] coordinates
{(0,1) (1,2) (0,3) (-1,2)} --cycle;
\end{axis}
\end{tikzpicture}
2
1
1 0.5 0 0.5 1
The --cycle is actually a path instruction of [6]; it connects the first and the last coordinate of one path.
Note that this is automatically done for filled paths.
4.5.12 Scatter Plots

The most simple scatter plots produce the same as the line plots above – but they contain only markers.
They are enabled using the only marks key of Tik Z.
/tikz/only marks (no value)
\addplot+[only marks]
Draws a simple scatter plot: all markers have the same appearance.
\begin{tikzpicture}
\begin{axis}[enlargelimits=false]
0.5 \addplot+[only marks,samples=400]
{rand};
\end{axis}
\end{tikzpicture}
0
0.5
4 2 0 2 4
The only marks visualization style simply draws marks at every coordinate. Marks can be set with
mark=hmark namei and marker options like size and color can be specified using the mark options=hstyle
optionsi key (or by modifying the every mark style). The available markers along with the accepted
style options can be found in Section 4.7 on page 176.
More sophisticated scatter plots change the marker appearance for each data point. An example is that
marker colors depend on the magnitude of function values f (x) or other provided coordinates. The term
“scatter plot” will be used for this type of plot in the following sections.
Scatter plots require “source” coordinates. These source coordinates can be the y coordinate, or explicitly
provided additional values.
/pgfplots/scatter (no value)
\addplot+[scatter]
Enables marker appearance modifications. The default implementation acquires “source coordinates”
for every data point (see scatter src below) and maps them linearly into the current color map. The
resulting color is used as draw and fill color of the marker.
\begin{axis}
\addplot+[scatter,only marks,
samples=50,scatter src=y]
10 {x-x^2};
\end{axis}
\end{tikzpicture}
20
30
6 4 2 0 2 4 6
The key scatter is simply a boolean variable which enables marker modifications. It applies only to
markers and it can be combined with any other plot type.
\begin{tikzpicture}
100 \begin{axis}
\addplot+[scatter,
samples=50,scatter src=y]
{x^3};
\end{axis}
0
\end{tikzpicture}
100
6 4 2 0 2 4 6
Scatter plots can be configured using a set of options. One of them is mandatory, the rest allows fine
grained control over marker appearance options.
/pgfplots/scatter src=none|hexpression i|x|y|z|f(x)|explicit|explicit symbolic (initially none)

This key is necessary for any scatter plot and it is set to f(x) as soon as scatter is activated and
no di↵erent choice has been made. It needs to be provided as hoptioni for \addplot to configure the
value used to determine marker appearances. Actually, scatter src is nothing but an alias for point
meta, so the main documentation for this key is on page 212. However, we summarize the choices here
together with scatter plot examples.
Usually, scatter src provides input data (numeric or string type) which is used to determine colors
and other style options for markers. The default configuration expects numerical data which is mapped
linearly into the current color map.
The value of scatter src determines how to get this data: the choices x, y and z will use either the
x, y or z coordinates to determine marker options. Any coordinate filters, logarithms or stacked-plot
computations have already been applied to these values (use rawx, rawy and rawz for unprocessed
values). The special choice f(x) is the same as y for two dimensional plots and the same as z for three
dimensional plots. The choice explicit expects the scatter source data as additional coordinate from
the coordinate input streams (see Section 4.3.1 for how to provide input meta data or below for some
small examples). They will be treated as numerical data. The choice explicit symbolic also expects
scatter source data as additional meta information for each input coordinate, but it treats them as
strings, not as numerical data. Consequently, no arithmetics is performed. It is the task of the scatter
plot style to do something with it. See, for example, the scatter/classes style below. Finally, it is
possible to provide an arbitrary mathematical expression which involves zero, one or more of the values
x (the current x coordinate), y (the current y coordinate) or z (the current z coordinate, if any).
If data is read from tables, mathematical expressions might also involve \thisrow{hcolumn namei} or
\thisrowno{hcolumn index i} to access any of the table cells in the current row.
Here are examples for how to provide data for the choices explicit and explicit symbolic.
\begin{tikzpicture}
\begin{axis}
% provide color data explicitly using [<data>]
% behind coordinates:
\addplot+[scatter,scatter src=explicit]
coordinates {
(0,0) [1.0e10]
(1,2) [1.1e10]
(2,3) [1.2e10]
(3,4) [1.3e10]
% ...
};
% Assumes a datafile.dat like

% xcolname ycolname colordata
% 0 0 0.001
% 1 2 0.3
% 2 2.1 0.4
% 3 3 0.5
% ...
% the file may have more columns.
table[x=xcolname,y=ycolname,meta=colordata]
{datafile.dat};
% Same data as last example:
\addplot+[scatter,scatter src=\thisrow{colordata}+\thisrow{ycolname}]
table[x=xcolname,y=ycolname]
{datafile.dat};

% 0 0 0.001
% 1 2 0.3
% 2 2.1 0.4
% 3 3 0.5
% ...
% the first three columns will be used here:
file {datafile.dat};
\end{axis}
\end{tikzpicture}
Please note that scatter src6=none results in computational work even if scatter=false.
/pgfplots/scatter/use mapped color={hoptions for each marker i} (style, initially draw=mapped

color!80!black,fill=mapped color)
This style is installed by default. When active, it recomputes the color mapped color for every processed
point coordinate by transforming the scatter src coordinates into the current color map linearly.
Then, it evaluates the options provided as hoptions for each marker i which are expected to depend on
mapped color.
The user interface for color maps is described in Section 4.7.6.
Default arguments \begin{tikzpicture}
\begin{axis}[title=Default arguments]
\addplot+[scatter,scatter src=y]
{2*x+3};
10 \end{axis}
\end{tikzpicture}
6 4 2 0 2 4 6
Black fill color and varying draw color \begin{tikzpicture}
\begin{axis}[
title=Black fill color and varying draw color,
scatter/use mapped color=
10 {draw=mapped color,fill=black}]
{2*x+3};
\end{axis}
\end{tikzpicture}
0
6 4 2 0 2 4 6
Black draw color and varying fill color \begin{tikzpicture}
\begin{axis}[
title=Black draw color and varying fill color,
10 {draw=black,fill=mapped color}]
{2*x+3};
\end{axis}
\end{tikzpicture}
0
6 4 2 0 2 4 6
0.5
1
0
0
5 0.5
1 0
4 2 0 2 4 5
\begin{tikzpicture}
\begin{axis}[colormap/bluered,colorbar]
\addplot3+[samples=4,
z buffer=sort,
% enable scatter:
scatter,
% redefine appearance:
scatter/use mapped color={ball color=mapped color},
% configure input:
scatter src=rand,
only marks,
mark=ball,
mark size=3pt,
]
{0};
\end{axis}
\end{tikzpicture}
This key is actually a style which redefines @pre marker code and @post marker code (see below).
Remark: The style use mapped color redefines @pre marker code and @post marker code. There
is a starred variant use mapped color* which appends the functionality while keeping the old marker
code.
/pgfplots/scatter/classes={hstyles for each class namei}

A scatter plot style which visualizes points using several classes. The style assumes that every point
coordinate has a class label attached, that means the choice scatter src=explicit symbolic is as-
sumed16 . A class label can be a number, but it can also be a symbolic constant. Given class labels for
every point, hstyles for each class namei contains a comma-separated list which associates appearance
options to each class label.
\begin{tikzpicture}
\begin{axis}[scatter/classes={
0.4 b={mark=triangle*,red},%
c={mark=o,draw=black}}]
% \addplot[] is better than \addplot+[] here:

% it avoids scalings of the cycle list
0.2
\addplot[scatter,only marks,
scatter src=explicit symbolic]
coordinates {
(0.1,0.15) [a]
(0.45,0.27) [c]
0 0.2 0.4 0.6 0.8 (0.02,0.17) [a]
(0.06,0.1) [a]
(0.9,0.5) [b]
(0.5,0.3) [c]
(0.85,0.52) [b]
(0.12,0.05) [a]
(0.73,0.45) [b]
(0.53,0.25) [c]
(0.76,0.5) [b]
(0.55,0.32) [c]
};
\end{axis}
\end{tikzpicture}
In this example, the coordinate (0.1,0.15) has the associated label ‘a’ while (0.45,0.27) has the
label ‘c’ (see Section 4.3 for details about specifying point meta data). Now, the argument to scat-
ter/classes contains styles for every label – for label ‘a’, square markers will be drawn in color blue.
The generation of a legend works as for a normal plot – but scatter/classes requires one legend entry
for every provided class. It communicates the class labels to the legend automatically. It works as if
there had been di↵erent \addplot commands, one for every class label.
It is also possible to provide scatter/classes as argument to a single plot, allowing di↵erent scatter
plots in one axis.
16 If scatter src is not explicit symbolic, we expect a numeric argument which is rounded to the nearest integer. The
resulting integer is used a class label. If that fails, the numeric argument is truncated to the nearest integer. If that fails as
well, the point has no label.
\begin{tikzpicture}
\begin{axis}[legend pos=south east]
% The data file contains:
0.4 % x y label
% 0.1 0.15 a
% 0.45 0.27 c
% 0.02 0.17 a
Class 1 % 0.06 0.1 a
0.2
Class 2 % 0.9 0.5 b
% 0.5 0.3 c
Class 3 % 0.85 0.52 b
Line % 0.12 0.05 a
% 0.73 0.45 b
0 0.2 0.4 0.6 0.8 % 0.53 0.25 c
% 0.76 0.5 b
% 0.55 0.32 c
\addplot[
% clickable coords={\thisrow{label}},
scatter/classes={
c={mark=o,draw=black,fill=black}%
},
scatter,only marks,
table[x=x,y=y,meta=label]
{plotdata/scattercl.dat};
{(0.1,0.1) (0.5,0.3) (0.85,0.5)};
\legend{Class 1,Class 2,Class 3,Line}
\end{axis}
\end{tikzpicture}
In general, the format of hstyles for each class namei is a comma separated list of hlabel i={hstyle
optionsi}.
Attention: The keys every mark and mark options have no e↵ect when used inside of hstyles for
each class namei! So, instead of assigning mark options, you can simply provide the options directly.
They apply only to markers anyway.
Remark: To use \label and \ref in conjunction with scatter/classes, you can provide the class
labels as optional arguments to \label in square brackets:
\addplot[
scatter/classes={
c={mark=o,draw=black,fill=black}%
},
scatter,only marks,
% [and coordinate input here... ]
;
\label[a]{label:for:first:class}
\label[b]{label:for:second:class}
\label[c]{label:for:third:class}
...
First class is \ref{label:for:first:class}, second is \ref{label:for:second:class}.
Remark: It is possible to click into the plot to display labels with mouse popups, see the clickable
coords key of the clickable library.
Remark: The style scatter/classes redefines @pre marker code and @post marker code. There
is a starred variant scatter/classes* which appends the functionality while keeping the old marker
code.
/pgfplots/nodes near coords={hcontenti} (default \pgfmathprintnumber\pgfplotspointmeta)

/pgfplots/nodes near coords*={hcontenti} (default \pgfmathprintnumber\pgfplotspointmeta)
A scatter plot style which places text nodes near every coordinate.
0.6
0.6 \begin{tikzpicture}
\begin{axis}[nodes near coords]
\addplot+[only marks] coordinates {
(0.5,0.2) (0.2,0.1) (0.7,0.6)
0.4 (0.35,0.4) (0.65,0.1)};
0.4 \end{axis}
\end{tikzpicture}
0.2
0.2
0.1 0.1
0.2 0.4 0.6
The hcontenti is, if nothing else has been specified, the content of the “point meta”, displayed using
the default hcontenti=\pgfmathprintnumber{\pgfplotspointmeta}. The macro \pgfplotspoint-
meta contains whatever has been selected by the point meta key, it defaults to the y coordinate for
two dimensional plots and the z coordinate for three dimensional plots.
Since point meta=explicit symbolic allows to treat string data, you can provide textual descriptions
which will be shown inside of the generated nodes17 :
(3) \begin{tikzpicture}
0.6 \begin{axis}[nodes near coords,enlargelimits=0.2]
\addplot+[only marks,
point meta=explicit symbolic]
(4) coordinates {
0.4 (0.5,0.2) [(1)]
(0.2,0.1) [(2)]
(1) (0.7,0.6) [(3)]
0.2 (0.35,0.4) [(4)]
(2) (5) (0.65,0.1) [(5)]
};
\end{axis}
0 \end{tikzpicture}
0.2 0.4 0.6 0.8
The square brackets are the way to provide explicit point meta for plot coordinates. Please refer to
the documentation of plot file and plot table for how to get point meta from files.
The hcontenti can also depend on something di↵erent than \pgfplotspointmeta. But since hcontenti
is evaluated during \end{axis}, pgfplots might not be aware of any special information inside of
hcontenti – you’ll need to communicate it to pgfplots with the visualization depends on key as
follows:
17 In this case, the \pgfmathprintnumber will be skipped automatically.
( 34 )
0.6
( 18 )
0.4
( 14 )
0.2
(1 12 ) (2)
0
0.2 0.4 0.6 0.8
\begin{tikzpicture}
\begin{axis}[enlargelimits=0.2]
\addplot[
scatter,mark=*,only marks,
% we use ’point meta’ as color data...
point meta=\thisrow{color},
% ... therefore, we can’t use it as argument for nodes near coords ...
nodes near coords*={$(\pgfmathprintnumber[frac]\myvalue)$},
% ... which requires to define a visualization dependency:
visualization depends on={\thisrow{myvalue} \as \myvalue},
]
table {
x y color myvalue
0.5 0.2 1 0.25
0.2 0.1 2 1.5
0.7 0.6 3 0.75
0.35 0.4 4 0.125
0.65 0.1 5 2
};
\end{axis}
\end{tikzpicture}
The example uses a scatter plot to get di↵erent colors, where the scatter src (or, equivalently,
point meta) is already used to define the markers color. In addition to the colored scatter plot, we’d
like to add nodes near coords, where the displayed nodes should contain \thisrow{myvalue}. To
do so, we define scatter,point meta=\thisrow{color} (just as described in the previous sections).
Furthermore, we use nodes near coords* in order to combine di↵erent scatter styles (see below for
details). The value for nodes near coords* depends on \thisrow{myvalue}, but we can’t use \pgf-
plotspointmeta (which is already occupied). Thus, we communicate the additional input data by
means of visualization depends on={\thisrow{myvalue} \as \myvalue}. The statement defines
a new macro, \myvalue, and assigns the value \thisrow{myvalue}. Furthermore, it configures pgf-
plots to remember this particular macro and its contents until \end{axis} (see the documentation for
visualization depends on for details).
The style nodes near coords might be useful for bar plots, see ybar for an example of nodes near
coords.
Remarks and Details:

• nodes near coords uses the same options for line styles and colors as the current plot. This may
be changed using the style every node near coord, see below.
• nodes near coords is actually one of the scatter plot styles. It redefines scatter/@pre marker
code to generate several Tik Z \node commands.
In order to use nodes near coords together with other scatter plot styles (like scatter/use
mapped color or scatter/classes), you may append a star to each of these keys. The variant
nodes near coords* will append code to scatter/@pre marker code without overwriting the
previous value.
• Consider using enlargelimits together with nodes near coords if text is clipped away.
• Currently nodes near coords does not work satisfactorily for ybar interval or xbar interval,
sorry.
/pgfplots/every node near coord (style, no value)
A style used for every node generated by nodes near coords. It is initially empty.
/pgfplots/nodes near coords style={hkey-value-listi}
An abbreviation for every node near coord/.append style={hkey-value-listi}.
It appends options to the already existing style every node near coord.
/pgfplots/node near coords style={hkey-value-listi}
/pgfplots/node near coord style={hkey-value-listi}
/pgfplots/nodes near coords align={halignment method i} (initially auto)
Specifies how to align nodes generated by nodes near coords.
Possible choices for halignment method i are
auto uses horizontal if the x coordinates are shown or vertical in all other cases. This checks the
current value of point meta.
horizontal uses left if \pgfplotspointmeta < 0 and right otherwise.
vertical uses below if \pgfplotspointmeta < 0 and above otherwise.
It is also possible to provide any Tik Z alignment option such as anchor=north east, below or some-
thing like that. It is also allowed to provide multiple options.
/pgfplots/scatter/position=absolute|relative (initially relative)
Allows to choose how to position scatter plot markers. This applies only if the scatter option is true.
The choice relative is the initial configuration, it means that the scatter marker is placed at the given
point’s coordinates. Technically, it means that the transformation matrix is shifted such that (0,0) is
right at the current point’s coordinates. This is typically what you want.
The choice absolute allows to position a scatter plot marker absolutely, for example by means of axis
cs. This can be combined with nodes near coords and a suitable set of hoptionsi in every node
near coord/.style={hoptionsi}: one could say at={(\coordindex,\coordindex*5)} or something
like that. Note that this choice necessarily needs an at key to specify the node’s position, otherwise its
position will be undefined18 .
/pgfplots/scatter/@pre marker code/.code={h... i}
/pgfplots/scatter/@post marker code/.code={h... i}
These two keys constitute the public interface which determines the marker appearance depending on
scatter source coordinates.
Redefining them allows fine grained control even over marker types, line styles and colors.
The scatter plot algorithm works as follows:
1. The scatter source coordinates form a data stream whose data limits are computed additionally to
the axis limits. This step is skipped for symbolic meta data.
2. Before any markers are drawn, a linear coordinate transformation from these data limits to the
interval [0.0, 1000.0] is initialised.
3. Every scatter source coordinate19 will be transformed linearly and the result is available as macro
\pgfplotspointmetatransformed 2 [0.0, 1000.0].
18 Well, it will be the origin of the canvas. Which is not necessarily the same as the origin of your axis.
19 During the evaluation, the public macros \pgfplotspointmeta and \pgfplotspointmetarange indicate the source coordinate
and the source coordinate range in the format a : b (for log–axis, they are given in fixed-point representation and for linear axes
in floating point).
The decision is thus based on per thousands of the data range. The transformation is skipped for
symbolic meta data (and the meta data is simply contained in the mentioned macro).
4. The pgf coordinate system is translated such that (0pt,0pt) is the plot coordinate.
5. The code of scatter/@pre marker code is evaluated (without arguments).
6. The standard code which draws markers is evaluated.
7. The code of scatter/@post marker code is evaluated (without arguments).
The idea is to generate a set of appearance keys which depends on \pgfplotspointmetatransformed.
Then, a call to \scope[hgenerated keysi] as @pre code and the associated \endscope as @post code
will draw markers individually using [hgenerated keysi].
A technical example is shown below. It demonstrates how to write user defined routines, in this case a
three–class system20 .
% Low-Level scatter plot interface Example:
% use three different marker classes
% 0% - 30% : first class
% 30% - 60% : second class
% 60% - 100% : third class
0
\begin{axis}[
scatter/@pre marker code/.code={%
\ifdim\pgfplotspointmetatransformed pt<300pt
\def\markopts{mark=square*,fill=blue}%
\else
1 \ifdim\pgfplotspointmetatransformed pt<600pt
\def\markopts{mark=triangle*,fill=orange}%
6 4 2 0 2 4 6 \else
\def\markopts{mark=pentagon*,fill=red}%
\fi
\fi
\expandafter\scope\expandafter[\markopts]
},%
scatter/@post marker code/.code={%
\endscope
}]
\addplot+[scatter,scatter src=y,
samples=40]
{sin(deg(x))};
\end{axis}
\end{tikzpicture}
Please note that \ifdim compares TEX lengths, so the example employs the suffix pt for any number
used in this context. That doesn’t change the semantics. The two (!) \expandafter constructions make
sure that \scope is invoked with the content of \markopts instead of the macro name \markopts.
4.5.13 1D Colored Mesh Plots

/pgfplots/mesh (no value)
\addplot+[mesh]
Uses the current color map to determine colors for each fixed line segment. Each line segment will get
the same color.
20 Please note that you don’t need to copy this particular example: the multiple–class example is also available as predefined
style scatter/classes.
\begin{axis}
\addplot[mesh] {x+sin(deg(x))};
2 \end{axis}
\end{tikzpicture}
0
4
6 4 2 0 2 4 6
The color data is per default the y value of the plot. It can be reconfigured using the point meta key
(which is actually the same as scatter src). The following example provides the color data explicitly
for plot coordinates, using the square bracket notation.
\begin{axis}
\addplot[mesh,point meta=explicit]
coordinates {
0.2 (0,0) [0]
(1,0.1) [1]
(2,0.1) [2]
0.1 (3,0.3) [3]
(4,0.3) [4]
};
\end{axis}
0 \end{tikzpicture}
0 1 2 3 4
This one-dimensional mesh plot is actually a special case of the twodimensional mesh plots, so more
detailed configuration, including how to change the color data, can be found in Section 4.6.5.
4.5.14 Interrupted Plots

Sometimes it is desirable to draw parts of a single plot separately, without connection between the parts
(discontinuities). pgfplots o↵ers three ways to generate interrupted plots:
1. empty lines in the input file or
2. by providing unbounded coords or

3. by providing unbounded point meta.
Interrupted Plots using Empty Lines

The first way to generate interrupted plots is simple; it needs no extra key (only \pgfplotsset{compat=1.4}
or newer in your preamble):
Interrupted data plot \begin{tikzpicture}
\begin{axis}[
1,000 title=Interrupted data plot]
(0,0) (10,50) (20,100) (30,200)
(50,600) (60,800) (80,1000)

500 };
\end{axis}
\end{tikzpicture}
0
0 20 40 60 80
Here, pgfplots runs with the default configuration empty line=auto which interpretes empty lines as
“jump” markers. This works for any data input method, i.e. using \addplot coordinates, \addplot
table, and \addplot file.
Note that empty line is useful for two–dimensional plots only since it has a special meaning for three–
dimensional plots (in which case empty line=scanline decodes matrices).
Interrupted Plots using Unbounded Coordinate Values

The second way to generate interrupted plots addresses the case where empty lines are unavailable or
impossible (due to limitations of the tool generating the data file, for example). In this case, interrupted
plots can be achieved using the unbounded coords key combined with coordinate values nan, inf or -inf.
/pgfplots/unbounded coords=discard|jump (initially discard)

This key configures what to do if one or more coordinates of a single point are unbounded. Here,
unbounded means it is either ±1 (+inf or -inf) or it has the special “not–a–number” value nan.
The initial setting discard discards the complete point and a warning is issued in the log file21 . This
setting has the same e↵ect as if the unbounded point did not occur: pgfplots will interpolate between
the bounded adjacent points.
The alternative jump allows interrupted plots: it provides extra checking for these coordinates and does
not interpolate over them; only those line segments which are adjacent to unbounded coordinates will
be skipped.
Discarding unbounded coords Jumps at unbounded coords
1,000 1,000
500 500
0 0
0 20 40 60 80 0 20 40 60 80
21 The warning can be disabled with filter discard warning=false.

\begin{tikzpicture}
\begin{axis}[
title=Discarding unbounded coords,
unbounded coords=discard]
(0,0) (10,50) (20,100) (30,200)
(40,inf) (50,600) (60,800) (80,1000)
};
\end{axis}
\end{tikzpicture}
\begin{tikzpicture}
\begin{axis}[
title=Jumps at unbounded coords,
unbounded coords=jump]
(0,0) (10,50) (20,100) (30,200)
(40,inf) (50,600) (60,800) (80,1000)
};
\end{axis}
\end{tikzpicture}
For plot expression and its friends, it is more likely to get very large floating point numbers instead of
inf. In this case, consider using the restrict x to domain key described on page 370.
The unbounded coords=jump method does also work for mesh/surface plots: every face adjacent to
an unbounded coordinate will be discarded in this case. The following example sets up a (cryptic)
coordinate filter which cuts out a quarter of the domain and replaces its values with nan:
\begin{tikzpicture}
\begin{axis}[
unbounded coords=jump,
% A technical filter to cut out
% the x<0 and y<0 edge.
0.5 filter point/.code={%
\pgfmathparse
{\pgfkeysvalueof{/data point/x}<0}%
\ifpgfmathfloatcomparison
5 \pgfmathparse
0 {\pgfkeysvalueof{/data point/y}<0}%
4 0 \ifpgfmathfloatcomparison
2 0 \pgfkeyssetvalue{/data point/x}{nan}%
2 4 5 \fi
\fi
},
]
\addplot3[surf] {exp(-sqrt(x^2 + y^2))};
\end{axis}
\end{tikzpicture}
This example requires unbounded coords=jump and an unbounded value (like nan) in order to work.
Choosing either unbounded coords=discard or an empty string to discard values would result in a
broken lattice which would confuse pgfplots. Note that the same example can be achieved using the
method in the following chapter, see the associated example below.
Interrupted Plots using Unbounded Point Meta Values

The third way to generate interrupted plots is designed for input data which forms a mesh, i.e. for three–
dimensional plots: in this case, point meta is expected (by mesh or surf and its variants). If one of the
involved coordinates has an unbounded value for its point meta, all adjacent patches which require the
point meta will be omitted.
The following example has one of its point meta values set to nan (not a number):
\begin{axis}[y dir=reverse]
\addplot[surf,point meta=explicit] coordinates {

(0,0) [0] (1,0) [1] (2,0) [2] (3,0) [3]
1
(0,1) [4] (1,1) [5] (2,1) [nan] (3,1) [7]
(0,2) [8] (1,2) [9] (2,2) [10] (3,2) [11]

};
2 \end{axis}
\end{tikzpicture}
0 1 2 3
Note that all four adjacent cells are omitted since the default shader is shader=flat mean, an averaging
scheme which fails to compute the averages.
If we use flat corner, only one cell becomes invisible:
\begin{axis}[y dir=reverse,shader=flat corner]
\addplot[surf,point meta=explicit] coordinates {

(0,0) [0] (1,0) [1] (2,0) [2] (3,0) [3]
1
(0,1) [4] (1,1) [5] (2,1) [nan] (3,1) [7]
(0,2) [8] (1,2) [9] (2,2) [10] (3,2) [11]

};
2 \end{axis}
\end{tikzpicture}
0 1 2 3
The case shader=flat corner choose the color of the first vertex of each patch. For mesh input=lat-
tice as in our example above, this applies to the rectangle with vertices (2, 1), (3, 1), (2, 2), (3, 2).
The feature can be used to choose point meta accordingly, for example to strip certain pieces of the
input:
\begin{tikzpicture}
\begin{axis}[shader=flat corner,domain=-1:1]
\addplot3[surf,
point meta={
1 abs(x-y)<0.1 || abs(x-0.5)<0.1 ? nan : z
}
]
0.5 {exp(-x^2-y^2)};
\end{axis}
1 \end{tikzpicture}
1 0
0.5 0 0.5 1 1
The expression removes all points where x == y or x == 12 (approximately) and assigns the color value
f (x, y) for all other coordinates.
This approach can also be applied to the example of the previous subsection, i.e. to
\begin{tikzpicture}
\begin{axis}[
point meta={x<0 && y<0 ? nan : z},
]
1 \addplot3[surf] {exp(-sqrt(x^2 + y^2))};
\end{axis}
\end{tikzpicture}
0.5
5
0
4 0
2 0 2 4 5
Limitations of this approach: coordinate filtering by means of point meta still includes the omitted patch
segments into the limit computation. This is di↵erent than x filter and its variants which also updates
axis limits accordingly.
More about this coordinate filtering can be found in Section 4.22 “Skipping Or Changing Coordinates –
Filters”.
4.5.15 Patch Plots

Patch Plots visualize a sequence of one or more triangles (or other sorts of patches). These triangles
can be drawn with a single color (shader=flat and shader=faceted interp) or with interpolated colors
(shader=interp).
There are both two- and three-dimensional patch plots, both with the same interface and the same
keys. Therefore, the reference documentation for patch plots can be found in Section 4.6.13 together with
three–dimensional patch plots.
4.5.16 Image (Matrix) Plots

A matrix plot consists of an input matrix and one colored cell for each element of the input matrix. Such
plots are a special case of a surface plot and are described in all detail in section 4.6.12 together with other
options about surface and mesh plots.
4.5.17 Polar Coordinates / Polar Axes

Polar coordinates can be provided in any kind of axis by means of data cs=polar, please refer to its reference
documentation of page 371.
Polar axes are available by means of \usepgfplotslibrary{polar}; please refer to Section 5.10.
4.5.18 Tieline Plots

Tieline plots are a special application in ternaryaxis. They are part of \usepgfplotslibrary{ternary};
please refer to Section 5.13.2 on page 496.
4.5.19 Smith Charts

Smith charts have a special coordinate system and a part of \usepgfplotslibrary{smitchart}; please
refer to Section 5.11.
4.6 Three Dimensional Plot Types

pgfplots provides three dimensional visualizations like scatter, line, mesh or surface plots. This section
explains the methods to provide input coordinates and how to use the di↵erent plot types.
4.6. THREE DIMENSIONAL PLOT TYPES 123
4.6.1 Before You Start With 3D

Before we delve into the capabilities of pgfplots for three dimensional visualization, let me start with some
preliminary remarks. The reason to use pgfplots for three dimensional plots are similar to those of normal,
two dimensional plots: the possibility to get consistent fonts and document consistent styles combined with
high–quality output.
While this works very nice for (not too complex) two dimensional plots, it requires considerably more
e↵ort than non–graphical documents. This is even more so for three dimensional plots. In other words:
pgfplots’ three dimensional routines are slow. There are reasons for this and some of them may vanish
in future versions. But one of these reasons is that TEX has never been designed for complex visualisation
techniques. Consider the image externalization routines mentioned in Section 7.1, in particular the external
library to reduce typesetting time. Besides the speed limitations, three dimensional plots reach memory limits
easily. Therefore, the plot complexity of three dimensional plots is limited to relatively coarse resolutions.
Section 7.1 also discusses methods to extend the initial TEX memory limits.
Another issue which arises in three dimensional visualization is depth: it is necessary to decide which items
are to be drawn in front of others. pgfplots supports z bu↵ering techniques up to a certain extend: it works
pretty well for single scatter plots (z buffer=sort), mesh or surface plots (z buffer=auto) or parametric
mesh and surface plots (z buffer=sort). However, it cannot combine di↵erent \addplot commands, those
will be drawn in the order of appearance. You may encounter the limitations sometimes. Maybe it will be
improved in future versions.
If you decide that you need high complexity, speed and 100% reliable z bu↵ers (depth information), you
should consider using other visualization tools and return to pgfplots in several years. If you can wait for a
complex picture and you do not even see the limitations arising from z bu↵ering limitations, you should use
pgfplots. Again, consider using the automatic picture externalization with the external library discussed
in Section 7.1.
4.6.2 The \addplot3 Command: Three Dimensional Coordinate Input

\addplot3[hoptionsi] hinput datai htrailing path commandsi;
The \addplot3 command is the main interface for any three dimensional plot. It works in the same
way as its two dimensional variant \addplot which has been described in all detail in Section 4.3 on
page 43.
The \addplot3 command accepts the same input methods as the \addplot variant, including expression
plotting, coordinates, files and tables. However, a third coordinate is necessary for each of these methods
which is usually straight–forward and is explained in all detail in the following.
Furthermore, \addplot3 has a way to decide whether a line visualization or a mesh visualization has
to be done. The first one is a map from one dimension into R3 and the latter one a map from two
dimensions to R3 . Here, the keys mesh/rows and mesh/cols are used to define mesh sizes (matrix
sizes). Usually, you don’t have to care about that because the coordinate input routines already allow
either one- or two-dimensional structure.
The precise rules how pgfplots distinguishes between line visualization and mesh visualization is as
follows:
1. If the key mesh input=patches has been used, pgfplots relies on the current patch type (which
has an inherent dimension).
2. For all other values of mesh input (like mesh input=lattice and mesh input=image), pgfplots
proceeds as follows:
• It auto-completes the values for mesh/rows and mesh/cols (or uses their prescribed values).
• If these keys describe a matrix of size 1 ⇥ N or N ⇥ 1, pgfplots assumes that it is a line.
The “auto-completion” depends on how you provide your data.
• If you use \addplot3 expression, mesh/rows and mesh/cols are computed from the values
samples and samples y. By default, \addplot3 expression always samples a mesh. If you
want it to sample a line, set samples y=1 (or, equivalently, y domain=0:0).
• If you use \addplot3 table, \addplot3 coordinates, or \addplot3 file, you have basi-
cally two options: either you provide at least one of mesh/rows or mesh/cols explicitly, or you
format the input table by means of end–of–scanline markers as outlined in empty line=auto or
empty line=scanline. In this case, pgfplots counts scanline lengths and uses mesh/order-
ing to conclude how to determine the matrix values.
Thus, if you have empty lines in your input table, pgfplots will automatically identify the
matrix size.
If you do not have empty lines, pgfplots expects at least one of mesh/rows or mesh/cols.
\addplot3 coordinates {hcoordinate listi};

\addplot3[hoptionsi] coordinates {hcoordinate listi} htrailing path commandsi;
The \addplot3 coordinates method works like its two–dimensional variant, \addplot coordinates
which is described in all detail on page 46:
A long list of coordinates (hx i,hyi,hz i) is expected, separated by white spaces. The input list can be
either an unordered series of coordinates, for example for scatter or line plots. It can also have matrix
structure, in which case an empty line (which is equivalent to “\par”) marks the end of one matrix
row. Matrix structure can also be provided if one of mesh/rows or mesh/cols is provided explicitly.
\begin{tikzpicture}
\begin{axis}
% this yields a 3x4 matrix:
\addplot3[surf] coordinates {
(0,0,0) (1,0,0) (2,0,0) (3,0,0)
0.5 (0,1,0) (1,1,0.6) (2,1,0.7) (3,1,0.5)
(0,2,0) (1,2,0.7) (2,2,0.8) (3,2,0.5)

2 };
0 \end{axis}
1
2
3 0
Here, \addplot3 reads a matrix with three rows and four columns. The empty lines separate one row
from the following.
As for the two–dimensional plot coordinates, it is possible to provide (constant) mathematical ex-
pressions inside of single coordinates. The syntax (hx i,hyi,hz i) [hmetai] can be used just as for two
dimensional plot coordinates to provide explicit color data; error bars are also supported.
\addplot3 file {hnamei};

\addplot3[hoptionsi] file {hnamei} htrailing path commandsi;
The \addplot3 file input method is the same as \addplot file – it only expects one more coordinate.
Thus, the input file contains xi in the first column, yi in the second column and zi in the third.
A further column is read after zi if point meta=explicit has been requested, see the documentation
of \addplot file on page 73 for details.
As for \addplot3 coordinates, an empty line in the file marks the end of one matrix row.
\begin{tikzpicture}
\begin{axis}
% We have ‘plotdata/first3d.dat’ with
%---------
% 0 0 0.8
% 1 0 0.56
% 2 0 0.5
0.5 % 3 0 0.75
%
% 0 1 0.6
2 % 1 1 0.3
0 % 2 1 0.21
1 % 3 1 0.3
2
3 0 %
% 0 2 0.68
% 1 2 0.22
% 2 2 0.25
% 3 2 0.4
%
% 0 3 0.7
% 1 3 0.5
% 2 3 0.58
% 3 3 0.9
% -> yields a 4x4 matrix:
\addplot3[surf] file {plotdata/first3d.dat};
\end{axis}
\end{tikzpicture}
For matrix data in files, it is important to specify the ordering in which the matrix entries have been writ-
ten. The default configuration is mesh/ordering=x varies, so you need to change it to mesh/order-
ing=y varies in case you have columnwise ordering.
\addplot3 table [hcolumn selectioni]{hfilei};

\addplot3[hoptionsi] table [hcolumn selectioni]{hfilei} htrailing path commandsi;
The \addplot3 table input works in the same way as its two dimensional counterpart \addplot
table. It only expects a column for the z coordinates. Furthermore, it interprets empty input lines as
end–of–row (more generally, end–of–scanline) markers, just as for plot file. The remark above about
the mesh/ordering applies here as well.
/pgfplots/mesh/rows={hinteger i}
/pgfplots/mesh/cols={hinteger i}
For visualization of mesh or surface plots which need some sort of matrix input, the dimensions of the
input matrix need to be known in order to visualize the plots correctly. The matrix structure may be
known from end–of–row marks (empty lines as general end–of–scanline markers in the input stream)
as has been described above.
If the matrix structure is not yet known, it is necessary to provide at least one of mesh/rows or mesh/cols
where mesh/rows indicates the number of samples for y coordinates whereas mesh/cols is the number
of samples used for x coordinates (see also mesh/ordering).
Thus, the following example is also a valid method to define an input matrix.
\begin{tikzpicture}
\begin{axis}
% this yields also a 3x4 matrix:
\addplot3[surf,mesh/rows=3] coordinates {
(0,0,0) (1,0,0) (2,0,0) (3,0,0)
(0,1,0) (1,1,0.6) (2,1,0.7) (3,1,0.5)
0.5 (0,2,0) (1,2,0.7) (2,2,0.8) (3,2,0.5)
};
\end{axis}
2 \end{tikzpicture}
0
0 1
1
2
3 0
It is enough to supply one of mesh/rows or mesh/cols – the missing value will be determined automat-
ically.
If you provide one of mesh/rows or mesh/cols, any end–of–row marker seen inside of input files or
coordinate streams will be ignored.
/pgfplots/mesh/scanline verbose=true|false (initially false)

Provides debug messages in the LATEX output about end–of–scanline markers.
The message will tell whether end–of–scanlines have been found and if they are the same.
/pgfplots/mesh/ordering=x varies|y varies|rowwise|colwise (initially x varies)

Allows to configure the sequence in which matrices (meshes) are read from \addplot3 coordinates,
\addplot3 file or \addplot3 table.
Here, x varies means a sequence of points where n=mesh/cols successive points have the y coordinate
fixed. This is intuitive when you write down a function because x is horizontal and y vertical. Note
that in matrix terminology, x refers to column indices whereas y refers to row indices. Thus, x varies
is equivalent to rowwise ordering in this sense. This is the initial configuration.
\begin{tikzpicture}
\begin{axis}[mesh/ordering=x varies]
% this yields a 3x4 matrix in ‘x varies’
% ordering:
(0,0,0) (1,0,0) (2,0,0) (3,0,0)
0.5
(0,1,0) (1,1,0.6) (2,1,0.7) (3,1,0.5)
2 (0,2,0) (1,2,0.7) (2,2,0.8) (3,2,0.5)

0 };
0 1 \end{axis}
1 \end{tikzpicture}
2
3 0
Note that mesh/ordering is mandatory, even though the size of the matrix can be provided in di↵erent
ways. The example above uses empty lines to mark scanlines. One could also say mesh/rows=3 and
omit the empty lines.
Consequently, mesh/ordering=y varies provides points such that successive m=mesh/rows points
form a column, i.e. the x coordinate is fixed and the y coordinate changes. In this sense, y varies is
equivalent to colwise ordering, it is actually a matrix transposition.
\begin{tikzpicture}
\begin{axis}[mesh/ordering=y varies]
% this yields a 3x4 matrix in colwise ordering:
(0,0,0) (0,1,0) (0,2,0)
0.5 (1,0,0) (1,1,0.6) (1,2,0.7)
(2,0,0) (2,1,0.7) (2,2,0.8)

2
0 (3,0,0) (3,1,0.5) (3,2,0.5)
0 1 };
1 \end{axis}
2
Again, note the subtle di↵erence to the common matrix indexing where a column has the second index
fixed. pgfplots refers to the way one would write down a function on a sheet of paper (this is consistent
with how Matlab (®) displays discrete functions with matrices).
\addplot3 {hmath expressioni} ;

\addplot3[hoptionsi] {hmath expressioni} htrailing path commandsi;
Expression plotting also works in the same way as for two dimensional plots. Now, however, a two
dimensional mesh is sampled instead of a single line, which may depend on x and y.
The method \addplot3 {hmath expr i} visualizes the function f (x, y) =hmath expr i where f : [x1 , x2 ] ⇥
[y1 , y2 ] ! R. The interval [x1 , x2 ] is determined using the domain key, for example using domain=0:1.
The interval [y1 , y2 ] is determined using the y domain key. If y domain is empty, [y1 , y2 ] = [x1 , x2 ] will
be assumed. If y domain=0:0 (or any other interval of length zero), it is assumed that the plot does
not depend on y (thus, it is a line plot).
The number of samples in x direction is set using the samples key. The number of samples in y direction
is set using the samples y key. If samples y is not set, the same value as for x is used. If samples
y  1, it is assumed that the plot does not depend on y (meaning it is a line plot).
% requires \usepgfplotslibrary{colorbrewer}
\begin{tikzpicture}
\begin{axis}[colormap/PuBu]
\addplot3[surf] {y};
5 \end{axis}
\end{tikzpicture}
0
5
5
4 0
2 0 2 4 5
0.5
1
0
0
1 0.5
1
0 0
0.2 0.4
0.6 0.8 1
1 1
\begin{tikzpicture}
\begin{axis}[colorbar]
\addplot3
[surf,faceted color=blue,
samples=15,
domain=0:1,y domain=-1:1]
{x^2 - y^2};
\end{axis}
\end{tikzpicture}
Expression plotting sets mesh/rows and mesh/cols automatically; these settings don’t have any e↵ect
for expression plotting.
\addplot3 expression {hmath expressioni};
\addplot3[hoptionsi] expression {hmath expressioni} htrailing path commandsi;
The syntax
\addplot3 {hmath expressioni};
as short-hand equivalent for
\addplot3 expression {hmath expressioni};
\addplot3 (hx expressioni,hy expressioni,hz expressioni) ;
\addplot3[hoptionsi] (hx expressioni,hy expressioni,hz expressioni) htrailing path commandsi;
A variant of \addplot3 expression which allows to provide di↵erent coordinate expressions for the x,
y and z coordinates. This can be used to generate parametrized plots.
Please note that \addplot3 (x,y,x^2) is equivalent to \addplot3 expression {x^2}.

Note further that since the complete point expression is surrounded by round braces, round braces
inside of hx expressioni, hy expressioni or hz expressioni need to be treated specially. Surround the
expressions (which contain round braces) with curly braces:
\addplot3 ({hx expr i}, {hy expr i}, {hz expr i});
4.6.3 Line Plots

Three dimensional line plots are generated if the input source has no matrix structure. Line plots take the
input coordinates and connect them in the order of appearance.
\begin{tikzpicture}
\begin{axis}[xlabel=$x$,ylabel=$y$]
\addplot3 coordinates {(0,0,0) (0,0.5,1) (0,1,0)};
\addplot3 coordinates {(0,1,0) (0.5,1,1) (1,1,0)};
1 \end{axis}
\end{tikzpicture}
0.5
1
0
0 0.2 0.4 0.5
0.6 0.8 y
1 0
x
If there is no value for neither mesh/rows nor mesh/cols or if one of them is 1, pgfplots will draw a
line plot. This is also the case if there is no end–of–scanline marker (empty line) in the input stream.
For \addplot3 expression, this requires to set samples y=0 to disable the generation of a mesh.
\begin{tikzpicture}
2
({sin(deg(x))},
{cos(deg(x))},
1 {2*x/(5*pi)});
\end{axis}
\end{tikzpicture}
0
0.5 1
0
0.5 0
1 1
The example above is a parametric plot by expression, i.e. it has three distinct expressions for x, y, and z.
Line plots in three dimensions are also possible for data plots (tables). The most simple case is if you
simply provide a series of three–dimensional coordinates which will be connected in the order of appearance:
\begin{tikzpicture}
\begin{axis}
\addplot3 table {
x y z
1 0 0 0
0.1 0.1 0.1
0.1 0.2 0.2
0.5 0.3 0.3 0.3
1 1 1
1 };
0 \end{axis}
0 0.2 0.4 0.5 \end{tikzpicture}
0.6 0.8
1 0
Note that this plot implicitly has mesh/rows=1 because it has no end–of–scanline markers (empty lines).
If in doubt, you can set mesh/rows=1 explicitly to tell pgfplots that you have one–dimensional data (and
not a matrix).
Line plots from data files are also possible if the data files only contains two coordinates – and the third
should be provided somehow. In this case, the table/x expr feature comes into play: it allows to combine
data plots and math expressions:
\begin{tikzpicture}
\begin{axis}[
xmin=3,xmax=6,
extra x ticks={4,5},
extra x tick style={xticklabel=\empty,grid=major}
]
5 \addplot3 table[x expr=4,y=a,z=b] {
a b
-3 9
-2 4
0 2 -1 1
3 0 0 0
4 1 1
5 2
6 2 4
3 9
};
\addplot3[red,domain=-3:3,samples y=0] (5,x,x^2);
\end{axis}
\end{tikzpicture}
Here, we have two plots in one axis: one data plot from a data table with just two coordinates and one
parametric plot. Both denote the same two functions. For the data plot, x expr=4 assigns the x coordinate,
and y=a,z=b define how the input columns map to coordinates. Again, the plot implicitly uses mesh/rows=1
since there is no end–of–scanline marker. The second plot does the same with the short–handed notation
(5,x,x^2). It only samples one–dimensional data due to samples y=0. Finally, extra x ticks configures
two additional ticks for the x axis; this is used to display grid lines for these specific ticks. The xtickla-
bel=\empty argument avoids overprinted x tick labels at positions x 2 {4, 5}.
Three dimensional line plots will usually employ lines to connect points (i.e. the initial sharp plot
handler of Tik Z). The smooth method of Tik Z might also prove be an option. Note that no piecewise
constant plot, comb or bar plot handler is supported for three dimensional axes.
Filled Lined Plots in 3D

Closing a line plot is also possible for three–dimensional axes. This works in the same way as outlined in
Section 4.5.11 for two–dimensional axes: by using \closedcycle.
\begin{tikzpicture}
3 \pgfplotstableread{
·10 plot1 plot2 plot3 plot4
0.0045 0.0029 0.0089 0.0001
0.0024 0.0023 0.0050 0.0016
0.0007 0.0012 0.0010 0.0001
5 0.0000 0.0004 -0.0000 -0.0015
0.0001 0.0001 0.0007 -0.0021
0.0003 0.0000 0.0015 -0.0020
plot4 0.0003 0.0001 0.0017 -0.0018
0 0.0003 0.0001 0.0016 -0.0015
0 plot3 0.0003 0.0001 0.0016 -0.0013
2 4 plot2
6 0.0003 0.0002 0.0015 -0.0012
8 }\tabledata
plot1
\begin{axis}[
zmin=-0.001,
area plot/.style={
fill opacity=0.75,
draw=orange!80!black,thick,
fill=orange,
mark=none,
},
ytick={1,...,4},
yticklabel=plot\pgfmathprintnumber{\tick},
]
\pgfplotsinvokeforeach{4,3,...,1}{
\addplot3 [area plot] table
[x expr=\coordindex, y expr=#1, z=plot#1]
{\tabledata} \closedcycle;
}
\end{axis}
\end{tikzpicture}
The di↵erence here is that pgfplots will connect the first and last coordinates on the z = 0 plane, or,
if z = 0 is outside of the axis limits, on the plane zmin .
4.6.4 Scatter Plots

Three dimensional scatter plots have the same interface as for two dimensional scatter plots, so all examples
of Section 4.5.12 can be used for the three dimensional case as well. The key features are to use only marks
and/or scatter as plot styles.
We provide some more examples which are specific for the three dimensional case.
Our first example uses only marks to place the current plot mark at each input position:
A Scatter Plot Example \begin{tikzpicture}
\begin{axis}[
xlabel=$x$,
ylabel=$y$,
zlabel={$f(x,y) = x\cdot y$},
title=A Scatter Plot Example]
1 % ‘pgfplotsexample4_grid.dat’ contains a
f (x, y) = x · y
% large sequence of input points of the form

% x_0 x_1 f(x)
0.5 % 0 0 0
% 0 0.03125 0
1 % 0 0.0625 0
0 % 0 0.09375 0
0 0.5 % 0 0.125 0
0.2 0.4 % 0 0.15625 0
0.6 0.8 y
1 0 \addplot3+[only marks] table
x {plotdata/pgfplotsexample4_grid.dat};
\end{axis}
\end{tikzpicture}
If we add the key scatter, the plot mark will also use the colors of the current colormap:
A Scatter Plot Example \begin{tikzpicture}
\begin{axis}[
xlabel=$x$,
ylabel=$y$,
title=A Scatter Plot Example]
1 \addplot3+[only marks,scatter] table
f (x, y) = x · y
{plotdata/pgfplotsexample4_grid.dat};
\end{axis}
1
0
0 0.2 0.4 0.5
0.6 0.8 y
1 0
x
A more sophisticated example is to draw the approximated function as a surf plot (which requires matrix
data) and the underlying grid (which is scattered data) somewhere into the same axis. We choose to place
the (x, y) grid points at z = 1.4. Furthermore, we want the grid points to be colored according to the value
of column f(x) in the input table:
Using Coordinate Filters to fix z = 1.4

1
0.8
0.6
f (x, y) = x · y
1
0.4
1
0 0.2
0 0.2 0.4 0.5
0.6 0.8 y 0
1 0
x
\begin{tikzpicture}
\begin{axis}[
3d box,
zmax=1.4,
colormap/viridis,
colorbar,
xlabel=$x$,
ylabel=$y$,
title={Using Coordinate Filters to fix $z=1.4$}]
% ‘pgfplotsexample4.dat’ contains similar data as in
% ‘pgfplotsexample4_grid.dat’, but it uses a uniform
% matrix structure (same number of points in every scanline).
% See examples above for extracts.
\addplot3[surf,mesh/ordering=y varies]
table {plotdata/pgfplotsexample4.dat};
\addplot3[scatter,scatter src=\thisrow{f(x)},only marks, z filter/.code={\def\pgfmathresult{1.4}}]
table {plotdata/pgfplotsexample4_grid.dat};
\end{axis}
\end{tikzpicture}
We used z filter to fix the z coordinate to 1.4. We could also have used the table/z expr=1.4 feature
\addplot3[scatter,scatter src=\thisrow{f(x)},only marks]
table[z expr=1.4] {plotdata/pgfplotsexample4_grid.dat};
to get exactly the same e↵ect. Choose whatever you like best. The z filter works for every coordinate
input routine, the z expr feature is only available for plot table.
The following example uses mark=cube* and z buffer=sort to place boxes at each input coordinate.
The color for each box is determined by point meta={x+y+3}. The remaining keys are just for pretty
printing.
9
7
5
l3
3
1 1
1
1 3
1
1 5
3
5 7
7 9 l1
9
l2
\begin{tikzpicture}
\begin{axis}[
view={120}{40},
width=220pt,
height=220pt,
grid=major,
z buffer=sort,
xmin=-1,xmax=9,
ymin=-1,ymax=9,
zmin=-1,zmax=9,
enlargelimits=upper,
xtick={-1,1,...,19},
ytick={-1,1,...,19},
ztick={-1,1,...,19},
xlabel={$l_1$},
ylabel={$l_2$},
zlabel={$l_3$},
point meta={x+y+z+3},
colormap={summap}{
color=(black); color=(blue);
color=(black); color=(white)
color=(orange) color=(violet)
color=(red)
},
scatter/use mapped color={
draw=mapped color,fill=mapped color!70},
]
% ‘pgfplots_scatter4.dat’ contains a large sequence of
% the form
% l_0 l_1 l_2
% 1 6 -1
% -1 -1 -1
% 0 -1 -1
% -1 0 -1
% -1 -1 0
% 1 -1 -1
% 0 0 -1
% 0 -1 0
\addplot3[only marks,scatter,mark=cube*,mark size=7]
table {plotdata/pgfplots_scatterdata4.dat};
\end{axis}
\end{tikzpicture}
4.6.5 Mesh Plots

/pgfplots/mesh (no value)
\addplot+[mesh]
A mesh plot uses di↵erent colors for each mesh segment. The color is determined using a “color
coordinate” which is also called “meta data” throughout this document. It is the same data which
is used for surface and scatter plots as well, see Section 4.8.2. In the initial configuration, the “color
coordinate” is the z axis (or the y axis for two dimensional plots). This color coordinate is mapped
linearly into the current color map to determine the color for each mesh segment. Thus, if the smallest
occurring color data is, say, 1 and the largest is 42, points with color data 1 will get the color at the
lower end of the color map and points with color data 42 the color of the upper end of the color map.
\begin{tikzpicture}
\begin{axis}
\addplot3[mesh] {x^2};
\end{axis}
\end{tikzpicture}
20
5
0
4 0
2 0 2 4 5
A mesh plot can be combined with markers or with the scatter key which also draws markers in
di↵erent colors.
·10 \begin{axis}[colormap/PuOr]
\addplot3+[mesh,scatter,samples=10,domain=0:1]
{x*(1-x)*y*(1-y)};
5 \end{axis}
\end{tikzpicture}
1
0
0 0.2 0.4 0.5
0.6 0.8
1 0
\begin{tikzpicture}
\begin{axis}[grid=major,view={210}{30}]
\addplot3+[mesh,scatter,samples=10,domain=0:1]
{5*x*sin(2*deg(x)) * y*(1-y)};
1 \end{axis}
\end{tikzpicture}
0.5
0 0
1 0.5
0.5
0 1
Occasionally, one may want to hide the background mesh segments. This can be achieved using the
surf plot handler (see below) and a specific fill color:
With background Without background
1 1
0.5 0.5
2 2
0 0
2 0 2 0
1 0 1 0
1 2 1 2
2 2
\begin{tikzpicture}
\begin{axis}[title=With background]
\addplot3[mesh,domain=-2:2] {exp(-x^2-y^2)};
\end{axis}
\end{tikzpicture}
\begin{tikzpicture}
\begin{axis}[title=Without background]
\addplot3[surf,fill=white,domain=-2:2] {exp(-x^2-y^2)};
\end{axis}
\end{tikzpicture}
The fill color needs to be provided explicitly.
Details:
• A mesh plot uses the same implementation as shader=flat to get one color for each single segment.
Thus, if shader=flat mean, the color for a segment is determined using the mean of the color
data of adjacent vertices. If shader=flat corner, the color of a segment is the color of the first
adjacent vertex.22
• As soon as mesh is activated, color=mapped color is installed. This is necessary unless one needs
a di↵erent color – but mapped color is the only color which reflects the color data.
It is possible to use a di↵erent color using the color=hcolor namei as for any other plot.
• It is easily possible to add mark=hmarker namei to mesh plots, scatter is also possible. Scatter
plots will use the same color data as for the mesh.
\begin{tikzpicture}
\begin{axis}[view/az=14]
\addplot3[mesh,draw=red,samples=10] {x^2-y^2};
\end{axis}
\end{tikzpicture}
20
0
5
20 0
4 2 0 2 4 5
Mesh plots use the mesh legend style to typeset legend images.
/pgfplots/mesh/check=false|warning|error (initially error)

Allows to configure whether an error is generated if mesh/rows ⇥ mesh/cols does not equal the total
number of coordinates.
22 Starting with pgfplots 1.13 and the associated compatibility level, this holds even in the presence of z bu↵ering.
If you know exactly what you are doing, it may be useful to disable the check. If you are unsure, it is
best to leave the initial setting.
/pgfplots/z buffer=default|none|auto|sort|reverse x seq|reverse y seq|reverse xy seq (initially

default)
This key allows to choose between di↵erent z bu↵ering strategies. A z bu↵er determines which parts of
an image should be drawn in front of other parts. Since both, the graphics packages pgf and the final
document format .pdf are inherently two dimensional, this work has to be done in TEX. Currently,
several (fast) heuristics can be used which work reasonably well for simple mesh- and surface plots.
Furthermore, there is a (time consuming) sorting method which also works if the fast heuristics fails.
The z bu↵ering algorithms of pgfplots apply only to a single \addplot command. Di↵erent \addplot
commands will be drawn on top of each other, in the order of appearance.
The choice default checks if we are currently working with a mesh or surface plot and uses auto in
this case. If not, it sets z buffer=none.
The choice none disables z bu↵ering. This is also the case for two dimensional axes which don’t need z
bu↵ering.
The choice auto is the initial value for any mesh or surface plot: it uses a very fast heuristics to decide
how to execute z bu↵ering for mesh and surface plots. The idea is to reverse either the sequence of
all x coordinates, or those of all y coordinates, or both. For regular meshes, this suffices to provide z
bu↵ering. In other words: the choice auto will use one of the three reverse strategies reverse * seq
(or none at all). The choice auto, applied to patch plots, uses z buffer=sort since patch plots have
no matrix structure.
The choice sort can be used for scatter, line, mesh, surface and patch plots. It sorts according to the
depth of each point (or mesh segment). Sorting in TEX uses a slow algorithm and may require a lot of
memory (although it has the expected runtime asymptotics O(N log N )). The depth of a mesh segment
is just one number, currently determined as mean over the vertex depths. Since z buffer=sort is
actually just a more intelligent way of drawing mesh segments on top of each other, it may still fail.
Failure can occur if mesh segments are large and overlap at di↵erent parts of the segment (see Wikipedia
“Painter’s algorithm”). If you experience problems of this sort, consider reducing the mesh width (the
mesh element size) such that they can be sorted independently (for example automatically using patch
refines=2, see the patchplots library).
The remaining choices apply only to mesh/surface plots (i.e. for matrix data) and do nothing more
then their name indicates: they reverse the coordinate sequences of the input matrix (using quasi linear
runtime). They should only be used in conjunction by z buffer=auto.
4.6.6 Surface Plots

/pgfplots/surf (no value)
\addplot+[surf]
A surface plot visualizes a two dimensional, single patch using di↵erent fill colors for each patch segment.
Each patch segment is a (pseudo) rectangle, that means input data is given in form of a data matrix as
is discussed in the introductory section about three dimensional coordinates, 4.6.2.
\begin{tikzpicture}
\begin{axis}
\addplot3[surf,shader=interp] {x*y};
\end{axis}
20 5
4 0
2 0 2 4 5
The simplest way to generate surface plots is to use the plot expression feature, but – as discussed in
Section 4.6.2 – other input methods like \addplot3 table or \addplot3 coordinates are also possible.
The appearance can be configured using colormaps, the value of the shader, faceted color keys and
the current color and/or draw/fill color. As for mesh plots, the special color=mapped color is
installed for the faces. The stroking color for faceted plots can be set with faceted color (see below
for details).
\begin{tikzpicture}
\begin{axis}[
grid=major,
colormap/viridis]
4 \addplot3[surf,samples=30,domain=0:1]
{5*x*sin(2*deg(x)) * y};
\end{axis}
2 \end{tikzpicture}
1
0
0 0.2 0.4 0.5
0.6 0.8
1 0
\begin{tikzpicture}
\begin{axis}
\addplot3[surf,faceted color=blue] {x+y};
\end{axis}
5
10
4 0
2 0 2 4 5
2 2
·10 ·10
5 5
1 1
0 0
0 0.2 0.4 0.5 0 0.2 0.4 0.5
0.6 0.8 0.6 0.8
1 0 1 0
\begin{tikzpicture}
\begin{axis}[colormap/cool]
\addplot3[surf,samples=10,domain=0:1,
shader=interp]
{x*(1-x)*y*(1-y)};
\end{axis}
\end{tikzpicture}
\begin{tikzpicture}
\begin{axis}[colormap/cool]
\addplot3[surf,samples=25,domain=0:1,
shader=flat]
{x*(1-x)*y*(1-y)};
\end{axis}
\end{tikzpicture}
\begin{tikzpicture}
\begin{axis}[grid=major]
\addplot3[surf,shader=interp,
samples=25,domain=0:2,y domain=0:1]
1 {exp(-x) * sin(pi*deg(y))};
\end{axis}
\end{tikzpicture}
0.5
1
0
0 0.5 0.5
1 1.5 2 0
\begin{tikzpicture}
\addplot3[surf,shader=faceted,
samples=25,domain=0:2,y domain=0:1]
1 {exp(-x) * sin(pi*deg(y))};
\end{axis}
\end{tikzpicture}
0.5
1
0
0 0.5 0.5
1 1.5 2 0
Details about the shading algorithm are provided below in the documentation of shader.
Surface plots use the mesh legend style to create legend images.
/pgfplots/shader=flat|interp|faceted|flat corner|flat mean|faceted interp (initially faceted)

Configures the shader used for surface plots. The shader determines how the color data available at
each single vertex is used to fill the surface patch.
The simplest choice is to use one fill color for each segment, the choice flat.
\begin{tikzpicture}
\begin{axis}
samples=10,domain=0:1]
1 {x^2*y};
\end{axis}
\end{tikzpicture}
0.5
1
0
0 0.2 0.4 0.5
0.6 0.8
1 0
There are (currently) two possibilities to determine the single color for every segment:
flat corner Uses the color data of one vertex to color the segment. The color of the first vertex
determines the color of the entire segment. Note that pgfplots ensures that the outcome is
always the same, even if the vertices are reordered due to z buffering or mesh/ordering.
Note that pgfplots versions up to and including 1.12 chose one of them without respectiving
reordering. In order to have the ordering independent of other features, you have to write com-
pat=1.13 or newer.
flat mean Uses the mean of all four color data values as segment color. This is the initial value as it
provides symmetric colors for symmetric functions.
The choice flat is actually the same as flat mean. Please note that shader=flat mean and
shader=flat corner also influence mesh plots – the choices determine the mesh segment color.
Another choice is shader=interp which uses Goraud shading (smooth linear interpolation of two tri-
angles approximating rectangles) to fill the segments.
\begin{tikzpicture}
\begin{axis}
1 {x^2*y};
\end{axis}
\end{tikzpicture}
0.5
1
0
0 0.2 0.4 0.5
0.6 0.8
1 0
The shader=interp employs a low–level shading implementation which is currently available for the
following drivers:
• the postscript driver \def\pgfsysdriver{pgfsys-dvips.def},
• the pdflatex driver \def\pgfsysdriver{pgfsys-pdftex.def},
• the lualatex driver \def\pgfsysdriver{pgfsys-pdftex.def} (yes, it is the same),
• the dvipdfmx driver \def\pgfsysdriver{pgfsys-dvipdfmx.def}.
For other drivers, the choice shader=interp will result in a warning and is equivalent to shader=flat
mean. See also below for detail remarks.
Note that shader=interp,patch type=bilinear allows real bilinear interpolation, see the patchplots
library.
The choice shader=faceted uses a constant fill color for every mesh segment (as for flat) and the
value of the key /pgfplots/faceted color to draw the connecting mesh elements:
\begin{tikzpicture}
\begin{axis}
1 {x^2*y};
\end{axis}
\end{tikzpicture}
0.5
1
0
0 0.2 0.4 0.5
0.6 0.8
1 0
The last choice is shader=faceted interp. As the name suggests, it is a mixture of interp and
faceted in the sense that each element is shaded using linear triangle interpolation (see also the patch-
plots library for bilinear interpolation) in the same way as for interp, but additionally, the edges are
colored in faceted color:
\begin{tikzpicture}
\begin{axis}
\addplot3[surf,shader=faceted interp,
1 {x^2*y};
\end{axis}
\end{tikzpicture}
0.5
1
0
0 0.2 0.4 0.5
0.6 0.8
1 0
In principle, there is nothing wrong with the idea as such, and it looks quite good – but it enlarges
the resulting pdf document considerably and might take a long time to render. It works as follows: for
every mesh element (either triangle for patch plots or rectangle for lattice plots), it creates a low level
shading. It then fills the single mesh element with that shading, and strokes the edges with faceted
color. The declaration of that many low level shadings is rather inefficient in terms of pdf objects
(large output files) and might render slowly23 . For orthogonal plots (like view={0}{90}), the e↵ect of
faceted interp can be gained with less cost if one uses two separate \addplot commands: one with
surf and one with mesh. Handle this choice with care.
Details:
• All shaders support z buffer=sort (starting with version 1.4)
• The choice shader=faceted is the same as shader=flat – except that it uses a special draw color.
So, shader=faceted has the same e↵ect as
shader=flat,draw=\pgfkeysvalueof{/pgfplots/faceted color}.
• The flat shader uses the current draw and fill colors. They are set with color=mapped color
and can be overruled with draw=hdraw color i and fill=hfill color i. The mapped color always
contains the color of the color map.
• You easily add mark=hplot mark i to mesh and/or surface plots or even colored plot marks with
scatter. The scatter plot feature will use the same color data as for the surface.
But: Markers and surfaces do not share the same depth information. They are drawn on top of
each other.
• Remarks on shader=interp:
– It uses the current color map in any case, ignoring draw and fill.
– For surface plots with lots of points, shader=interp produces smaller pdf documents, re-
quires less compilation time in TEX and requires less time to display in Acrobat Reader than
shader=flat.
– The postscript driver truncates coordinates to 24 bit – which might result in a loss of precision
(the truncation is not very intelligent). See the surf shading/precision key for details.
To improve compatibility, this 24 bit truncation algorithm is enabled by default also for pdf
documents.
– The choice shader=interp works well with either Acrobat Reader or recent versions of free
viewers24 . However, some free viewers show colors incorrectly (like evince). I hope this mes-
sage will soon become outdated... if not, provide bug reports to the Linux community to
communicate the need to improve support for Type 4 (patch) and Type 5 pdf (surf) and
Type 7 (patch and elements of the patchplots library) shadings.
– The interp shader yields the same outcome as faceted interp,faceted color=none, al-
though faceted interp requires much more ressources.
23 My experience is as follows: Acrobat reader can efficiently render huge interp shadings. But it is very slow for faceted
interp shadings. Linux viewers like xpdf are reasonably efficient for interp (at least with my bugfixes to libpoppler) and are
also fast for faceted interp shadings.
24 The author of this package has submitted bugfixes to xpdf/libpoppler which should be part of the current stable versions
of many viewers.
\begin{tikzpicture}
\begin{axis}[colormap/viridis]
draw=black,
1 samples=10,domain=0:1]
{x^2*y};
\end{axis}
1
0
0 0.2 0.4 0.5
0.6 0.8
1 0
\begin{tikzpicture}
\begin{axis}[colormap/viridis]
scatter,mark=*,
1 samples=10,domain=0:1]
{x^2*y};
\end{axis}
1
0
0 0.2 0.4 0.5
0.6 0.8
1 0
Note that the shader always interacts with colormap access. In particular, colormap access=piece-
wise constant combined with shader=interp results in a filled contour plot, see Section 4.6.9 for
details.
/pgfplots/faceted color={hcolor namei} (initially mapped color!80!black)

Defines the color to be used for meshes of faceted surface plots.
Set faceted color=none to disable edge colors.
/pgfplots/mesh/interior colormap={hmap namei}{hcolormap specificationi}

/pgfplots/mesh/interior colormap name={hmap namei}
Allows to use a di↵erent colormap for the “other side” of the surface.
Each mesh has two sides: one which “points” to the view’s origin and one which points away from it.
This key allows to define a second colormap for the side which points away from the view’s origin. The
motivation is to distinguish between the outer side and the interior parts of a surface:
\begin{tikzpicture}
1 z \begin{axis}[
axis lines=center,
axis on top,
0.5 xlabel={$x$}, ylabel={$y$}, zlabel={$z$},
domain=0:1,
y
y domain=0:2*pi,
1 xmin=-1.5, xmax=1.5,
1 x ymin=-1.5, ymax=1.5, zmin=0.0,
1 mesh/interior colormap=
1 {blueblack}{color=(black) color=(blue)},
colormap/blackwhite,
samples=10,
samples y=40,
z buffer=sort,
]
\addplot3[surf]
({x*cos(deg(y))},{x*sin(deg(y))},{x});
\end{axis}
\end{tikzpicture}
The interior colormap is often the one for the “‘inner side”. However, the orientation of the surface
depends on its normal vectors: pgfplots computes them using the right-hand-rule. The right-hand-rule
applied to a triangle means to take the first encountered point, point the thumb in direction of the second
point and the first finger in direction of the third point. Then, the normal for that triangle is the third
finger (i.e. the cross–product of the involved oriented edges). For rectangular patches, pgfplots uses
the normal of one of its triangles25 . Consequently, mesh/interior colormap will only work if the
involved patch segments are consistently oriented.
A patch whose normal vector points into the same direction as the view direction uses the standard
colormap name. A patch whose normal vector points into the opposite direction (i.e. in direction of the
viewport) uses mesh/interior colormap.
\begin{tikzpicture}
\begin{axis}[
hide axis,
mesh/interior colormap name=hot,
]
\addplot3[domain=-1.5:1.5,surf]
{-exp(-x^2-y^2)};
\end{axis}
\end{tikzpicture}
The implementation of mesh/interior colormap works well for most examples; in particular, if the
number of samples is large enough to resolve the boundary between inner and outer colormap. However,
it might still produce spurious artifacts:
Example needing fine-tuning \begin{tikzpicture}
\begin{axis}[
title=Example needing fine-tuning,
xlabel=$x$,
ylabel=$y$]
2 \addplot3[surf,
mesh/interior colormap=
{blueblack}{color=(black) color=(blue)},
domain=0:1]
0 {sin(deg(8*pi*x))* exp(-20*(y-0.5)^2)
1 + exp(-(x-0.5)^2*30
- (y-0.25)^2 - (x-0.5)*(y-0.25))};
0 0.5 \end{axis}
0.2 0.4 \end{tikzpicture}
0.6 0.8 y
1 0
x
The previous example has need for improvement with respect to a couple of aspects: first, it has small
overshoots near some of the meshes vertices (especially on top of the hills). These can be fixed using
miter limit=1. Second, the boundary between blue and black is incorrect. This can be improved by
means of an increased sampling density (samples=31). In addition, we can configure pgfplots to move
the boundary between the two colormaps in favor of the blue region using mesh/interior colormap
thresh as follows:
25 This may change in future versions.
Example of before with fine-tuning \begin{tikzpicture}
\begin{axis}[
title=Example of before with fine-tuning,
xlabel=$x$,
ylabel=$y$]
2 \addplot3[surf,
mesh/interior colormap=
{blueblack}{color=(black) color=(blue)},
% slightly increase sampling quality (was 25):
samples=31,
0 % avoids overshooting corners:
1 miter limit=1,
% move boundary between inner and outer:
0 0.5 mesh/interior colormap thresh=0.1,
0.2 0.4 colormap/blackwhite,
0.6 0.8 y
1 0 domain=0:1]
x {sin(deg(8*pi*x))* exp(-20*(y-0.5)^2)
+ exp(-(x-0.5)^2*30
- (y-0.25)^2 - (x-0.5)*(y-0.25))};
\end{axis}
\end{tikzpicture}
This improves the display.
Call for volunteers: it would be nice if the fine–tuning of these keys would be unnecessary. If
someone has well–founded suggestions (like knowledge and perhaps exhaustive experiments) on how to
improve the feature, let me know.
Note that mesh/interior colormap cannot be combined with mesh/refines currently.
Note that mesh/interior colormap will increase compilation times due to the computation of normal
vectors.
/pgfplots/mesh/interior colormap thresh={hNumber between 1.0 and +1.0i} (initially 0)

A threshold which moves the boundary between the colormap and interior colormap in favor of
colormap (if the value is negative) or in favor of interior colormap (if the value is positive).
The extreme value 1 essentially deactivates interior colormap whereas the other extreme +1 deac-
tivates colormap.
See above for an example.
/pgfplots/surf shading/precision=pdf|postscript|ps (initially postscript)

A key to configure how the low level driver for shader=interp writes its data. The choice pdf uses
32 bit binary coordinates (which is lossless). The resulting .pdf files appear to be correct, but they
can’t be converted to postscript – the converter software always complains about an error.
The choice postscript (or, in short, ps) uses 24 bit truncated binary coordinates. This results in both,
readable .ps and .pdf files. However, the truncation is lossy.
If anyone has ideas how to fix this problem: let me know. As far as I know, Postscript should accept
32 bit coordinates, so it might be a mistake in the shading driver.
4.6.7 Surface Plots with Explicit Color

The surface plots described in Section 4.6.6 are all based on colormaps. This section introduces a di↵erent
type of surface plot. In fact, it uses the very same plot handlers: it applies to mesh, surf, and patch plots.
However, the way colors are provided and the way pgfplots interpolates colors is substantially di↵erent.
This section describes surface plots with explicit colors. These expect colors like red, green, or
rgb=(0.5,0.2,1) for every vertex of the mesh – and interpolates smoothly between these vertices. This
appears to be simpler, perhaps even more straight-forward than surface plots based on colormaps. It is
not. Surface plots with explicit color are more difficult to define, and they are more difficult to read.
If you are in doubt of whether to use a surface colored by a colormap or explicit colors, you should prefer
colormaps for reasons discussed below.
/pgfplots/mesh/color input=colormap|explicit|explicit mathparse (initially colormap)

Allows to configure how pgfplots expects color input for surface plots.
The choice colormap uses the standard colormaps. This particular choice expects scalar values of point
meta which are mapped linearly into the colormap. It resembles the surface plots which are explained
in more detail in Section 4.6.6. It is the default configuration and covers (probably) most common
use–cases.
The choice explicit expects explicitly provided point meta of symbolic form: every coordinate of your
input coordinate stream is supposed to have an explicitly defined color as point meta. Here, “explicitly
provided” refers to point meta=explicit symbolic. This choice and the available color formats are
explained in all detail in the following sub–sections.
The choice explicit mathparse is similar to explicit, but it allows to provide just one math expres-
sion which is evaluated for every coordinate. The math expression can be of the form rgb=x,y,0.5 in
which case x is used to define the “red” component, y is used to define the “green” component and the
“blue” component is fixed to 0.5. This key is also explained in more detail in the following sub–sections.
The main use–case of mesh/color input=colormap is to allow a map between the interpolated colors
and some value of interest. This map can be shown as colorbar:
1 1
% \usepgfplotslibrary{patchplots}
\begin{tikzpicture}
0.8 0.8 \begin{axis}[small,view={0}{90}, colorbar]
\addplot3[surf,shader=interp,patch type=bilinear]
0.6 0.6 coordinates {
(0,0,0) (1,0,0)
0.4 0.4
(0,1,0) (1,1,1)
};
0.2 0.2 \end{axis}
\end{tikzpicture}
0 0
0 0.2 0.4 0.6 0.8 1
Note that the preceding example is a standard surf plot except for patch type=bilinear which controls
how color is to be interpolated. There is a 2 ⇥ 2 matrix, and its z values are used as color data. Clearly,
value z = 0 corresponds to blue and z = 1 corresponds to red – and all other colors in–between are not
directly related to blue and red; they are taken from the colormap. The colormap defines which colors
appear: those which make up the colormap and those which can occur as interpolated colors between the
colors of the color map. The pairwise mixture of colors is a property of mesh/color input=colormap,
not of mesh/color input=explicit (where more than two colors are mixed together). Furthermore,
the surface indicates contours of constant z level. Take, for example, the yellow contour. We know
that it has some value between 0.3 and 0.4, say 0.35. Since these shadings are continuous, we know
that the point z = 0.35 occurs between z = 0 and z = 1 – at every point of the surface. Due to the
colormap, each point on the surface which has z = 0.35 will receive the yellow color. This is because
the interpolation is carried out on the scalar point meta value, which is afterwards mapped into the
colormap. This contour–property is also unique for colormap surfaces.
The other two choices are explained in all detail in the next sub–sections.
Providing Colors Explicitly For Each Coordinate
Here is a simple approach with the same vertices as the colormap–example above, but with explicit
colors:
1
\begin{tikzpicture}
0.8 \begin{axis}[small,view={0}{90}]
\addplot3[surf,shader=interp,patch type=bilinear,
0.6 mesh/color input=explicit]
coordinates {
(0,0,0) [color=blue] (1,0,0) [color=green]
0.4
(0,1,0) [color=yellow] (1,1,1) [color=red]
0.2 };
\end{axis}
0 \end{tikzpicture}
0 0.2 0.4 0.6 0.8 1
The coordinates and the view is the same, even the way colors are being interpolated bilinearly. However,
we have four di↵erent colors in the corners. We see these corners in the output, and we see that they
are smoothly mixed together. However, the mix contains all four colors, not just two. As a direct
consequence, there are no contour lines.
The absence of direct information how to map color information to “some information of the data
visualization” implies that if you want to use surface plots with explicit color, you have to state clearly
what you want to show. This is considerably simpler for colormaps.
The following example configures z = 0 to receive blue and z = 1 to receive red as in our preceding
colormap example (see above) using mesh/color input=explicit.
1
\begin{tikzpicture}
0.8 \begin{axis}[small,view={0}{90}]
\addplot3[surf,shader=interp,patch type=bilinear,
0.6 mesh/color input=explicit]
coordinates {
(0,0,0) [color=blue] (1,0,0) [color=blue]
0.4
(0,1,0) [color=blue] (1,1,1) [color=red]
0.2 };
\end{axis}
0 \end{tikzpicture}
0 0.2 0.4 0.6 0.8 1
We see the bilinear nature of the interpolation; it is related to that of mesh/color input=colormap
above (compare the contour lines in–between). In most cases, you simply want to show some contour
lines. And for such cases, a colormap is the way to go.
There might be cases where mesh/color input=explicit is adequate. However, you will need to think
it through properly. And you need to explain clearly what you did because your audience will also have
to think a lot before they make sense of any data visualization based on explicit color interpolation.
The choice mesh/color input=explicit expects a choice of point meta which results in symbolic
values. In this context, “symbolic” refers to a special color definition:
\begin{axis}[minor x tick num=1]
\addplot[
patch,
shader=interp,
mesh/color input=explicit,
0.5
]
table[meta=c] {
x y c
0 0 color=green
% default color model is rgb:
0 1 1 1,0,0
2 0 1,1,0
0 2 4
1.5 1 cmyk=1,0,0,0
2.5 0 gray=0.5
3.5 1 color=red!80!black
3 0 1,0,1
4 1 0,0,1
5 0 rgb255=0,128,128
};
\end{axis}
\end{tikzpicture}
The previous example defines a patch plot with three triangle patches, each made up of three vertices
which are placed as–is into the input coordinate stream. Each vertex has its color data in column c.
The format of color specifications is explained in more detail in the following paragraph.
As soon as you write mesh/color input=explicit, pgfplots checks the current value of point meta.
If the current value of point meta is none, it is set to point meta=explicit symbolic (that is what
happened in our example above). If the current value of point meta is some choice which yields
numeric output (like point meta=x or point meta=\thisrow{x}+1), it is set to point meta=explicit
symbolic. If the current value of point meta is already of symbolic form, it is left unchanged.
Consequently, our example above sets point meta=explicit symbolic as soon as it encounters mesh/-
color input=explicit. The explicit symbolic input handler in turn expects the coordinate stream to
provide point meta data for every streamed coordinate. We use a table here, and a table reads its color
data from the column name provided in the table/meta key.
The accepted format of colors is quite similar to that of Colormap definitions (compare Section 4.7.6
on page 187). The common format is hcolor model i=harguments i. In contrast to the similar input
format inside of colormap definitions, the syntax here has no round braces and does not have the hlengthi
argument. Nevertheless, the same hcolor model is with the same hargumentsi are accepted. The choices
are
rgb=hred i,hgreeni,hbluei where each component is in the interval [0, 1],

rgb255=hred i,hgreeni,hbluei is similar to rgb except that each component is expected in the interval
[0,255],
gray=hvaluei with hvaluei in the interval [0, 1],
color=hnamed color i where hnamed color i is a predefined (named) color like ‘red’ or a color expression
like ‘red!50’,
cmyk=hcyani,hmagentai,hyellow i,hblack i where each component is in the interval [0, 1],
cmyk255=hcyani,hmagentai,hyellow i,hblack i is the same as cmyk but expects components in the interval
[0, 255],
cmy=hcyani,hmagentai,hyellow i where each component is in the interval [0, 1],
hsb=hhuei,hsaturationi,hbrightnessi where each component is in the interval [0, 1],
Hsb=hhuei,hsaturationi,hbrightnessi is the same as hsb except that hhuei is accepted in the interval
[0, 360] (degree),
HTML=hhex red ihhex greenihhex bluei where component is expected to be a hex number between 00 and
FF (a variant of rgb255),
wave=hwave lengthi which expects a single wave length as numeric argument in the range [363, 814].
/pgfplots/mesh/colorspace explicit color input=rgb|rgb255|cmy|cmyk|cmyk255|gray|wave|hsb

|Hsb|HTML (initially rgb)
If the input color has no color model, the color components are interpreted as color in the color
model specified as argument to this key.
This key has just one purpose: to omit the hcolor model i if it is the same for lots of points anyway.
/pgfplots/mesh/colorspace explicit color output=rgb|cmyk|gray (initially rgb)

Any color which is encountered by the survey phase (i.e. while inspecting the point meta value)
is immediately transformed into the color space configured as mesh/colorspace explicit color
output.
Note that this option has no e↵ect if you told xcolor to override the color space globally. More
precisely, the use of
\usepackage[cmyk]{xcolor}
or, alternatively,
\selectcolormodel{cmyk}
will cause all colors to be converted to cmyk, and pgfplots honors this configuration. Consequently,
both these statements cause all colors to be interpolated in the desired color space, and all output
colors will use this colorspace. This is typically exactly what you need.
The transformed color is used for any color interpolation. In most cases, this is done by the shader,
but it applies to patch refines and patch to triangles as well.
Because the transformed color is used for color interpolation, the list of available output color
spaces is considerably smaller than the available input color spaces. Only the device color spaces
rgb, gray, and cmyk are available as value for mesh/colorspace explicit color output.
Any necessary colorspace transformations rely on the xcolor package26 . Note that colorspace
transformations are subject to nearest–color–matching, i.e. they are less accurate. This is typically
far beyond pure rounding issues; it is caused by the fact that these color spaces are actually so
di↵erent that transformations are hard to accomplish. If you can specify colors immediately in the
correct color space, you can eliminate such transformations.
Here is the same shading, once with CMYK output and once with RGB output. Depending on
your output media (screen or paper), you will observe slightly di↵erent colors when comparing the
pictures.
RGB shading \begin{tikzpicture}
\begin{axis}[title=RGB shading]
4 \addplot[
patch,
shader=interp,
2 mesh/colorspace explicit color output=rgb,
data cs=polar,
]
coordinates {
0 (90,4) [color=red]
(210,4) [color=green]
(-30,4) [color=blue]
2 };
\end{axis}
4 2 0 2 4 \end{tikzpicture}
26 Colorspace transformations are unavailable for plain T X and ConT Xt. For these cases, you have to ensure that input
E E
and output color model are the same.
CMYK shading \begin{tikzpicture}
\begin{axis}[title=CMYK shading]
4 \addplot[
patch,
shader=interp,
2 mesh/colorspace explicit color output=cmyk,
data cs=polar,
]
coordinates {
0 (90,4) [color=red]
(210,4) [color=green]
(-30,4) [color=blue]
2 };
\end{axis}
This key is similar to the related key colormap default colorspace, although the values can be
chosen independently.
Providing Color Components as Table
The previous section shows how to provide a single symbolic color expression for each coordinate, namely
using point meta=explicit symbolic.
Another use–case might be to provide a table containing both the coordinate values and one column by
color component. In order to assemble the color specification from the input table, you can provide a
symbolic expression:
\begin{axis}
\addplot[
patch,
shader=interp,
0.5
point meta={symbolic=
{\thisrow{R},\thisrow{G},\thisrow{B}}
},
]
table {
0 x y R G B
0 0 0 1 0
0 0.5 1 1.5 2 1 1 1 0 0
2 0 0 0 1
};
\end{axis}
\end{tikzpicture}
The preceding example employs a patch plot with triangular elements as we have seen before. Fur-
thermore, it uses explicit color input – but combined with point meta={symbolic={hvaluei}} (note
the extra pair of braces). This choice accepts arbitrary symbols on input which will be reevaluated
(expanded) for every coordinate. In our example, we simply read the values from table columns using
\thisrow. Since the default input colorspace is RGB, this results in the expected triangle with red,
green, and blue corners. The result has to form a valid color specification.
Note that the symbolic expression is purely string–based in this context. If you plan to use math
expressions, you have to use mesh/color input=explicit mathparse as explained in the following
section.
Providing Colors as Math Expression
The key mesh/color input has two choices for explicit color input. The choice explicit has been
discussed in the preceding paragraphs, it expects one color specification for every node for which col-
ors are needed. It also accepts a kind of string–based expressions to concatenate the expected color
specification in a suitable way.
The second choice mesh/color input=explicit mathparse is almost the same – with one major dif-
ference: it allows to provide math expressions inside of the point meta value. However, the provided
math expressions need to form a color specification which typically has more than one color component.
With this choice, the value of point meta is of symbolic form, but the color components are reevaluated
with the math parser for every input point which has color data. The most convenient way to provide
such expressions is point meta={symbolic={hRi,hGi,hB i}} (again, note the extra set of braces for the
argument).
\begin{axis}
\addplot[
surf,
shader=interp,
mesh/color input=explicit mathparse,
0.5
point meta={symbolic={x,\thisrow{y},0}},
]
table {
x y
0 0
0 1 0
0 0.2 0.4 0.6 0.8 1 0 1

1 1
};
\end{axis}
\end{tikzpicture}
The preceding example is a surface plot with a 2 ⇥ 2 input matrix. Note that the table has no explicit
point meta data. The point meta data is acquired from a common math expression which uses the
final x coordinate as hred i component, the value seen in the current row and column y as hgreeni value
and constant value hbluei= 0. Consequently, the output is black in the lower left corner since black is
(0, 0, 0), red in the lower right corner, green in the upper left corner, and a mixture of both along the
diagonal.
The value provided as point meta={symbolic={hvaluei}} is of the same form as for mesh/color in-
put=explicit, i.e. it is supposed to be of the form hcolor model i=hcolor componentsi. If the hcolor
model i is omitted, it defaults to mesh/colorspace explicit color input (which is rgb by default).
Since the math expression can be anything, it can safely be combined with plot by expression.
A potential use–case could be to show a surface with (x, y, z) and some two–dimensional quantity which
is encoded as mixture of red and green. The following example relies on mesh/color input=explicit
mathparse and point meta=symbolic to provide a vector of math expressions:
\begin{tikzpicture}
\begin{axis}
% this example burns colors if opacity
2 % is active in the document.
\addplot3[
patch,
0 patch type=bilinear,
1 domain=0:1,
2 samples=30,
0 0.2 0.4 0.5 point meta={symbolic={%
0.6 0.8 (sin(deg(x*pi*2))+1)/2,% R
1 0 (sin(deg(y*pi*2))+1)/2,% G
0% B
}
},
]
{sin(deg(x*pi*2))+sin(deg(y*pi*2))};
\end{axis}
\end{tikzpicture}
Note that the preceding example su↵ers from color burning27 : the green areas become too bright and the
black areas become too dark. Note that the picture is entirely acceptable if it is written as stand–alone
picture. But as soon as you import the picture (either as .pdf or as .png) into a document for which
opacity is active, it su↵ers from burned colors.
The color burning is caused by the combination of RGB colorspace, the special color set in this example,
and the color blending which is activated by opacity. Note that it is enough to activate opacity
somewhere in the document.
In order to repair the problem for the picture at hand, one has to choose a di↵erent output colorspace:
\begin{tikzpicture}
\begin{axis}
\addplot3[
2 patch,
patch type=bilinear,
0 %
% CMYK produces better quality here
1 % since the manual has opacity enabled
2 mesh/colorspace explicit color output=cmyk,
0 0.2 0.4 0.5 domain=0:1,
0.6 0.8 samples=30,
1 0 point meta={symbolic={%
(sin(deg(x*pi*2))+1)/2,% R
(sin(deg(y*pi*2))+1)/2,% G
0% B
}
},
]
{sin(deg(x*pi*2))+sin(deg(y*pi*2))};
\end{axis}
\end{tikzpicture}
The key mesh/colorspace explicit color output transforms every input RGB color to a matching
CMYK color. This, in turn, is a lossy transformation which seems to lack a trivial solution28 .
Math expressions can also be used for complicated input color spaces, for example the wave colorspace.
\begin{tikzpicture}
\begin{axis}
\addplot3[
surf,
1 shader=interp,
domain=0:1,
0.5 samples y=2,
point meta={symbolic={wave=363+x*(814-363)}},
1 ]
0 {x*y};
0 0.2 0.4 0.5 \end{axis}
0.6 0.8 \end{tikzpicture}
1 0
Note that you have to take care that the color components are within the expected bounds.
4.6.8 Contour Plots

pgfplots supports visualization of contour plots whose coordinates have been computed by external tools.
The contour prepared plot handler coming with pgfplots takes precomputed contour line coordinates
and handles their visualization (contour/draw color, contour/labels etc.). The contour gnuplot style
takes matrix input in the same format as for mesh or surf (that includes any of the pgfplots matrix input
methods). It then writes the matrix data to a file and invokes gnuplot (or other, user customizable external
programs) to compute contour coordinates. Finally, the computed contours are visualized with the contour
27 At least in Acrobat Reader.
28 If some expert in color space operations can contribute best–practices here, feel free to contact me.
prepared algorithm. Thus, external programs need to compute the contour coordinates and pgfplots
visualizes the result.
We discuss the high level interface to external programs first and continue with contour prepared
later-on.
/pgfplots/contour gnuplot={hoptions with ‘contour/’ or ‘contour external/’ prefixi}
\addplot+[contour gnuplot={hoptions with ‘contour/’ or ‘contour external/’ prefixi}]

This is a high level contour plot interface. It expects matrix data in the same way as two dimensional
surf or mesh plots do. It then computes contours and visualizes them.
20 20
\addplot3[contour gnuplot]
2 10 10
{x*y};
\end{axis}
0
\end{tikzpicture}
0 0 0
4
0
10
20 20
10
4 2 0 2 4
The example uses \addplot3 together with expression plotting, that means the input data is of the
form (xi , yi , f (xi , yi )). The view={0}{90} flag means “view from top”, otherwise the contour lines
would have been drawn as z value:
\begin{tikzpicture}
\begin{axis}
\addplot3[contour gnuplot]
{exp(0-x^2-y^2)};
1 \end{axis}
0.6
0.8
\end{tikzpicture}
0.6
0 . 0 .2
0.5 4
0 .4
1
0 .2 0 .2
1 0
0
1 1
As mentioned, you can use any of the pgfplots input methods as long as it yields matrix output.
Thus, we can re-use our introductory example of matrix data, this time with inline data:
% Preamble: \pgfplotsset{width=7cm,compat=1.13}%
2
\begin{tikzpicture}%
\begin{axis}[view={0}{90}]%
\addplot3[contour gnuplot]%
coordinates {
0.2
0.4
0.6
0.6
1.5
(0,0,0) (1,0,0) (2,0,0) (3,0,0)
(0,1,0) (1,1,0.6) (2,1,0.7) (3,1,0.5)

1 0.6
(0,2,0) (1,2,0.7) (2,2,0.8) (3,2,0.5)
};
0 .4
0.
0.4 \end{axis}%
2
0.5 \end{tikzpicture}%
0.2
0.5 1 1.5 2 2.5 3
What happens behind the scenes is that pgfplots takes the input matrix and writes all encountered co-
ordinates to a temporary file, including the end–of–scanline markers. Then, it generates a small gnuplot
script and invokes gnuplot to compute the contour coordinates, writing everything into a temporary
output file. Afterwards, it includes gnuplot’s output file just as if you’d write \addplot3[contour
prepared] file {htemporaryfilei};.
All this invocation of gnuplot, including input/output file management is transparent to the user. It
only requires two things: first of all, it requires matrix data as input29 . Second, it requires you to enable
system calls. Consider the documentation for plot gnuplot for how to enable system calls.
Note that the z coordinate of the data which is communicated to gnuplot is the current value of point
meta. This allows to generate contours on two columns only and has more freedom. See also the contour
external/output point meta key.
The resulting data can be projected onto a separate slice, for example using z filter.
\begin{axis}%
1 \addplot3[surf,shader=interp]%
1 coordinates {
1 1
(0,0,0) (1,0,0) (2,0,0) (3,0,0)
(0,1,0) (1,1,0.6) (2,1,0.7) (3,1,0.5)

0.5
(0,2,0) (1,2,0.7) (2,2,0.8) (3,2,0.5)
2 };
0 \addplot3[
0 1 contour gnuplot,
1 z filter/.code={\def\pgfmathresult{1}},
2
3 0 ]%
coordinates {
(0,0,0) (1,0,0) (2,0,0) (3,0,0)
(0,1,0) (1,1,0.6) (2,1,0.7) (3,1,0.5)
(0,2,0) (1,2,0.7) (2,2,0.8) (3,2,0.5)

};
\end{axis}%
\end{tikzpicture}%
An unexpected side–e↵ect of a z filter combined with contour gnuplot is that the color information
and the label sizes are essentially lost. To componsate this e↵ect, we have to assign a new value for
point meta. But since point meta defines the values for contour levels (see above), there is a special
key named output point meta which can be used here:
29 Note that contour gnuplot processes the input stream only once. Consequently, the temporary file will contain only
information which was available before the first point has been seen. The example above works because it contains emptylines
as end-of-scanline markers. If you do not provide such markers, you may need to provide two of the three options mesh/rows,
mesh/cols, or mesh/num points.
\begin{axis}%
0 .6 \addplot3[surf,shader=interp]%
0 .4 coordinates {
1 0.2
(0,0,0) (1,0,0) (2,0,0) (3,0,0)
(0,1,0) (1,1,0.6) (2,1,0.7) (3,1,0.5)

0.5
(0,2,0) (1,2,0.7) (2,2,0.8) (3,2,0.5)
2 };
0 \addplot3[
0 1 contour gnuplot={
1 output point meta=rawz,
2
3 0 },
z filter/.code={\def\pgfmathresult{1}},
]%
coordinates {
(0,0,0) (1,0,0) (2,0,0) (3,0,0)
(0,1,0) (1,1,0.6) (2,1,0.7) (3,1,0.5)
(0,2,0) (1,2,0.7) (2,2,0.8) (3,2,0.5)

};
\end{axis}%
\end{tikzpicture}%
Here, output point meta=rawz allows to assign point meta to the output of the contouring algorithm,
i.e. something which is handed over to contour prepared. The choice rawz means to use the unfiltered
z coordinate which is what we want here.
There are several fine-tuning parameters of the input/output file management, and it is even possible
to invoke di↵erent programs than gnuplot (even matlab). These details are discussed at the end of this
section, see below at page 161.
/pgfplots/contour/number={hinteger i} (initially 5)
Configures the number of contour lines which should be produced by any external contouring algorithm.
2 \begin{axis}[
domain=-2:2,enlarge x limits,
0. view={0}{90},
1 1 ]
0.1
\addplot3[contour gnuplot={number=14},thick]
0 .2
0 .4 0 .4 {exp(-x^2-y^2)*x};
0
0.2
0 \end{axis}
0.1
0.1
3 \end{tikzpicture}
0 .3 0.
0 .2
0.
2
1 1
0 .1 0.
2
0
2 1 0 1 2
It is also possible to change the /pgf/number format settings, see the documentation for the con-
tour/every contour label style below.
Note that contour/number has no e↵ect on contour prepared.
/pgfplots/contour/levels={hlist of levelsi} (initially empty)

Configures the number of contour lines which should be produced by any external contouring algorithm
by means of a list of discrete levels.
\begin{axis}[
0 .1
1 domain=-2:2,
.1 0.2 enlargelimits,
0
view={0}{90},
]
0.2
\addplot3[
0.2
0 contour gnuplot={levels={-0.1,-0.2,-0.6}},
thick]
0.1
{exp(-x^2-y^2)*x};
0.
\end{axis}
1
0 .2
\end{tikzpicture}
1
0.1
1.5 1 0.5 0
It is also possible to change the /pgf/number format settings, see the documentation for the con-
tour/every contour label style below.
This key has higher precedence than contour/number, i.e. if both are given, contour/levels will be
active.
Note that contour/levels has no e↵ect on contour prepared.
/pgfplots/contour/contour dir=x|y|z (initially z)
Allows to generate contours with respect to another direction.
\begin{axis}[
domain=-2:2,
]
\addplot3[
0.4 contour gnuplot={
contour dir=x,labels=false,number=15},
0.2 thick]
0 {exp(-x^2-y^2)*x};
\end{axis}
0.2 2 \end{tikzpicture}
0.4
0
1
0
1 2
The input data is the same as before – it has to be given in matrix form. The key contour dir
configures the algorithm to compute contours along the provided direction.
\begin{axis}[
domain=-2:2,
]
\addplot3[
0.5 contour gnuplot={
contour dir=y,labels=false,number=15},
thick]
0 {exp(-x^2-y^2)*x};
\end{axis}
\end{tikzpicture}
1
0.5 0
2 1 0 1 1
2
This function is also available for parameterized surfaces.

\begin{axis}[view={60}{30}, axis equal image]
\addplot3[
0 contour gnuplot={
contour dir=x,
1 number=20,
labels=false,
0.50 0.5 },
0.5 0.50 samples=30,domain=-1:1,y domain=0:2*pi]
({sqrt(1-x^2) * cos(deg(y))},
x);
\end{axis}
\end{tikzpicture}
Note, however, that each contour line receives a single color. This is what one expects for a contour plot:
it has a single style, and a single contour level. Note furthermore that the color which is assigned to a
contour plot with contour dir=x is di↵erent compared with the color assigned to a contour plot with
contour dir=z: the argument of contour dir implicitly defines the argument for point meta (also
known as color data). More precisely, a contour plot with contour dir=x has point meta=x whereas
a contour plot with contour dir=z uses point meta=z.
If you would like to have individually colored segments inside of contours, you have to use a di↵erent
plot handler. There is a simple alternative which works well in many cases: you can use a standard
mesh plot combined with patch type=line:
\begin{axis}[
domain=-2:2,
]
0.4 \addplot3[
samples y=10, samples=25,
0.2 mesh, patch type=line,
0 thick]
{exp(-x^2-y^2)*x};
0.2 \end{axis}
2
2 0
1 0 1 2 y
2
x
Here, we did not generate a contour plot. We generated a mesh plot with patch type=line. The choice
patch type=line causes an inherently one–dimensional plot as opposed to the default matrix–style
visualization which would be generated by mesh in di↵erent cases. Since a mesh plot uses one color for
every patch segment, we have a lot of freedom to color the segments. In the example above, we have
the default configuration point meta=z, i.e. the z value defines the color.
The fact that a mesh plot with patch type=line yields almost the same output as contour dir=y
is an artifact of the scanline encoding. Our example uses \addplot3 expression which relies on
mesh/ordering=x varies. If we visualize the resulting matrix by means of patch type=line, the
visualization follows the scanlines which vary along the x axis. In our example, we used samples y=10
to control the number of “contour lines”.
A consequence of the previous paragraph is that we have a more challenging task at hand if we want
to get the same e↵ect as contour dir=x: we would need mesh/ordering=y varies. In our case, we
would need to transpose the data matrix. For \addplot3 expression, this is relatively simple: we can
exchange the meaning of x and y to get a transposition:
\begin{axis}[
domain=-2:2,
]
0.5 \addplot3[
samples y=10, samples=25,
mesh, patch type=line,
0 thick]
(y,x,{exp(-x^2-y^2)*y});
2 \end{axis}
\end{tikzpicture}
0.5 0
2 1 0 1 2 2 y
x
This is the same example as above – but as you noted, the meaning of x and y in the expression has
been exchanged and the notation has been switched to a parametric plot. Such an approach is also
possible for data files, but pgfplots cannot transpose the matrix ordering on its own.
Coming back to contour dir, we can also use its output to generate several contour projections using
coordinate filters.
\begin{tikzpicture}
\begin{axis}[
20 3d box=complete]
\addplot3[surf]
{x^2-y^2};
0
\addplot3[
5 contour gnuplot={contour dir=y,
20 draw color=red,labels=false},
4 0 y filter/.expression={-5}
2 0 ] {x^2-y^2};
2 4 y
5
x \addplot3[
contour gnuplot={contour dir=x,
draw color=blue,labels=false},
x filter/.expression={5}
] {x^2-y^2};
\addplot3[
contour gnuplot={contour dir=z,
draw color=black,labels=false},
z filter/.expression={25}
] {x^2-y^2};
\end{axis}
\end{tikzpicture}
The preceding example uses a fixed draw color combined with x filter, y filter, and z filter to
fix the contours in one of the axis planes.
Technical background: This section is probably unnecessary and can be skipped. The key contour
dir is implemented by means of coordinate permutations. Since contouring algorithms always support
contour dir=z, it is relatively simple to compute z–contour lines from input matrixes X, Y, Z. The
choice contour dir=z key simply takes the input as–is. The choice contour dir=x reorders the input
coordinates to yzx. The choice contour dir=y reorders the input coordinates to xzy. All this reordering
is applied before coordinates are handed over to the contouring algorithm (see contour external) and
is undone when reading the results back from the contouring algorithm. That means that contour
dir is also available for contour prepared. In this context, contour prepared is supposed to be the
output of some contouring algorithm. Its input coordinates are automatically reordered according to
the inverse permutation. This allows to draw x or y contours which are given in the prepared format.
/pgfplots/contour prepared={hoptions with ‘contour/’ prefixi}
\addplot+[contour prepared={hoptions with ‘contour/’ prefixi}]

A plot handler which expects already computed contours on input and visualizes them. It cannot
compute contours on its own.
/pgfplots/contour prepared format=standard|matlab (initially standard)

There are two accepted input formats. The first is a long sequence of coordinates of the form
(x, y, z) where all successive coordinates with the same z value make up a contour level (this is only
part of complete truth, see below). The end–of–scanline markers (empty lines in the input) mark
an interruption in one contour level.
For example, contour prepared format=standard could be30
\begin{axis}
\addplot[contour prepared]
0.2
0.4
0.6
0.6
1.5 table {
2 2 0.8
1 0.857143 2 0.6
0 .6 1 1 0.6
2 0.857143 0.6
0.4 2.5 1 0.6
0.5 2.66667 2 0.6
0.2 0.2
0.571429 2 0.4
1 2 3 0.666667 1 0.4
1 0.666667 0.4
2 0.571429 0.4
3 0.8 0.4
0.285714 2 0.2
0.333333 1 0.2
1 0.333333 0.2
2 0.285714 0.2
3 0.4 0.2
};
\end{axis}
\end{tikzpicture}
Note that the empty lines are not necessary in this case: empty lines make only a di↵erence
if they occur within the same contour level (i.e. if the same z value appears above and below of
them).
The choice contour prepared format=matlab expects two–dimensional input data where the con-
tour level and the number of elements of the contour line are provided as x and y coordinates,
respectively, of a leading point. Such a format is used by matlab’s contour algorithms, i.e. it
resembles the output of the matlab commands data=contour(...) or data=contourc(...).
30 This is actually the output from our \addplot3[contour gnuplot] coordinates example from above.
\begin{axis}
\addplot[contour prepared,
0.2
0.4
0.6
0.6
1.5 contour prepared format=matlab]
table {
% (0.2,5) ==> contour ‘0.2’ (x), 5 points follow (y):
1 2.0000000e-01 5.0000000e+00
0.6 3.0000000e+00 4.0000000e-01
2.0000000e+00 2.8571429e-01
0.
0.4 1.0000000e+00 3.3333333e-01

2
0.5 3.3333333e-01 1.0000000e+00

0.2 2.8571429e-01 2.0000000e+00
% (0.4,5) ==> contour ‘0.4’, consists of 5 points
1 2 3 4.0000000e-01 5.0000000e+00
3.0000000e+00 8.0000000e-01
2.0000000e+00 5.7142857e-01
1.0000000e+00 6.6666667e-01
6.6666667e-01 1.0000000e+00
5.7142857e-01 2.0000000e+00
% (0.6,6) ==> contour ‘0.6’, has 6 points
6.0000000e-01 6.0000000e+00
2.6666667e+00 2.0000000e+00
2.5000000e+00 1.0000000e+00
2.0000000e+00 8.5714286e-01
1.0000000e+00 1.0000000e+00
1.0000000e+00 1.0000000e+00
8.5714286e-01 2.0000000e+00
};
\end{axis}
\end{tikzpicture}
In case you use matlab, you can generate such data with
[x,y]=meshgrid(linspace(0,1,15));
data=contour(x,y,x.*y);
data=data’;
save ’exporteddata.dat’ data -ASCII
As already mentioned in the beginning, the z coordinate is not necessarily the coordinate used to
delimit contour levels. In fact, the point meta data is acquired here, i.e. you are free to use whatever
z coordinate you want as long as you have a correct point meta value. The example from above could
be modified as follows:
Separating z from Color Value \begin{tikzpicture}
\begin{axis}[
title=Separating $z$ from Color Value,
xlabel=$x$,
ylabel=$y$,
0.6 ]
\addplot3[contour prepared,
0.5 point meta=\thisrow{level}]
0 .4
table {
x y z level
0.2
0.857143 2 0.4 0.6
2 1 1 0.6 0.6
0 0 .2 2 0.857143 0.6 0.6
2.5 1 0.6 0.6
1 1 2.66667 2 0.4 0.6
2 y
x 3
0.571429 2 0.2 0.4
0.666667 1 0.4 0.4
1 0.666667 0.4 0.4
2 0.571429 0.4 0.4
3 0.8 0.2 0.4
0.285714 2 0 0.2
0.333333 1 0.2 0.2
1 0.333333 0.2 0.2
2 0.285714 0.2 0.2
3 0.4 0 0.2
};
\end{axis}
\end{tikzpicture}
The example above uses di↵erent z coordinates for each first and each last point on contour lines. The
contour lines as such are defined by the level column since we wrote point meta=\thisrow{level}.
Such a feature also allows contour prepared for nonstandard axes, compare the examples for the
ternary lib on page 492.
/pgfplots/contour/draw color={hcolor i} (initially mapped color)

Defines the draw color for every contour. Note that only mapped color actually depends on the
contour level.
/pgfplots/contour/labels={htrue,falsei} (initially true)

Configures whether contour labels shall be drawn or not.
/pgfplots/contour/label distance={hdimensioni} (initially 70pt)

Configures the distance between adjacend contour labels within the same contour level.
/pgfplots/contour/every contour plot (style, no value)

A style which is installed as soon as either contour or contour external is set.
The initial value is
\pgfplotsset{
contour/every contour plot/.style={
/pgfplots/legend image post style={sharp plot},
},
}
/pgfplots/contour/every contour label (style, no value)

Allows to customize contour labels. The preferred way to change this style is the contour label
style={hoptionsi} method, see below.
\pgfplotsset{
contour/every contour label/.style={
sloped,
transform shape,
inner sep=2pt,
every node/.style={mapped color!50!black,fill=white},
/pgf/number format/relative={\pgfplotspointmetarangeexponent},
}
}
Note that \pgfplotspointmetarangeexponent= e where ±m · 10e is the largest occuring label

value (technically, it is the largest occurring value of point meta).
The following example modifies the /pgf/number format styles for contour labels:
2 \begin{axis}[
domain=-2:2,enlarge x limits,
0. view={0}{90},
1 1 ]
0 .1
\addplot3[
0 .2
0 .4 0 .4 contour gnuplot={
0
0.2
0 scanline marks=required,
0.1
0.1
3 number=14,
0.3 0.
0 .2
contour label style={

0.
2
1 1 /pgf/number format/fixed,
0 .1 0.
},
},thick
2
0
2 1 0 1 2 ]
{exp(0-x^2-y^2)*x};
\end{axis}
\end{tikzpicture}
/pgfplots/contour/contour label style={hkey-value-listi}

An abbreviation for contour/every contour label/.append style={hkey-value-listi}.
It appends options to the already existing style contour/every contour label.
/pgfplots/contour/labels over line (style, no value)

A style which changes every contour label such that labels are right over the lines, without fill
color.
20 20
4 15
\begin{tikzpicture}
15 \begin{axis}[view={0}{90}]
10 10 \addplot3[contour gnuplot={
2 5 5 labels over line,number=9}]
{x*y};
0
0 0 \end{axis}
0
\end{tikzpicture}
2
5
5
4
0
10
20 20
15
10
15
4 2 0 2 4
/pgfplots/contour/handler (style, no value)

Allows to modify the plot handler which connects the points of a single contour level.
\pgfplotsset{contour/handler/.style={/tikz/sharp plot}}
but a useful alternative might be the smooth handler:

2
\begin{tikzpicture}
\addplot3[
1 contour gnuplot={
levels={0.1,0.2,0.5,1,2,5},
labels=false,
0 handler/.style=smooth,
},
]
1 {2*sqrt(x^2+y^2+(-0.5*(x^2-y^2)))};
\end{axis}
\end{tikzpicture}
2
2 0 2
Unfortunately, the Tik Z handler smooth has its limitations for such plots, you may need to play
around with the number of samples.
/pgfplots/contour/label node code/.code={h... i}

A lowlevel interface to modify how contour labels are placed.
\pgfplotsset{
contour/label node code/.code={\node {\pgfmathprintnumber{#1}};}
}
/pgfplots/contour external={hoptions with ‘contour/’ or ‘contour external/’ prefixi}
\addplot+[contour external={hoptions with ‘contour/’ or ‘contour external/’ prefixi}]

This handler constitutes a generic interface to external programs to compute contour lines. The contour
gnuplot method is actually a special case of contour external.
/pgfplots/contour external/file={hbase file namei} (initially empty)

The initial configuration is to automatically generate a unique file name.
/pgfplots/contour external/scanline marks=false|if in input|required|true (initially if in

input)
Controls how contour external writes end-of-scanline markers.
The choice false writes no such markers at all. In this case, script should contain mesh/rows
and/or mesh/cols.
The choice if in input generates end-of-scanline markers if they appear in the provided input
data (either as empty lines or if the user provided at least two of the three options mesh/rows,
mesh/cols, or mesh/num points explicitly).
The choice required works like if in input, but it will fail unless there really was such a marker.
The choice true is an alias for required.
/pgfplots/contour external/script={hCode for external programi} (initially empty)

Provides template code to generate a script for the external program. Inside of hCode for external
programi, the placeholder \infile will expand to the temporary input file and \outfile to the
temporary output file. The temporary \infile is a text file containing one point on each line, in
the form x y meta meta, separated by tabstops. Whenever a scanline is complete, an empty line
is issued (but only if these scanline markers are found in the input stream as well). The complete
set of scanlines forms a matrix. There are no additional comments or extra characters in the file.
The macro \ordering will expand to 0 if the matrix is stored in mesh/ordering=x varies and
\ordering will be 1 for mesh/ordering=y varies.
Inside of hCode for external programi, you can also use \pgfkeysvalueof{/pgfplots/mesh/rows}
and \pgfkeysvalueof{/pgfplots/mesh/cols}; they expand to the matrix’ size. Similarly,
\pgfkeysvalueof{/pgfplots/mesh/num points} expands to the total number of points.
Inside of hCode for external programi, the macro \thecontournumber is defined to be the
value \pgfkeysvalueof{/pgfplots/contour/number} and \thecontourlevels contains the value
\pgfkeysvalueof{/pgfplots/contour/levels}. These two macros simplify conditional code.
If you need one of the characters ["|;:#’‘] and some macro package already uses the character for
other purposes, you can prepend them with a backslash, i.e. write \" instead of ".
/pgfplots/contour external/script extension={hextensioni} (initially script)

The file name extension for the temporary script.
/pgfplots/contour external/cmd={hsystem call i} (initially empty)

A template to generate system calls for the external program. Inside of hsystem call i, you may use
\script as placeholder for the filename which contains the result of contour external/script.
/pgfplots/contour external/output point meta={hpoint meta read from result of external tool i}
(initially empty)
Allows to customize the point meta configuration which is applied to the result of the external
tool.
In contour external, the value of point meta is used to generate the input z coordinate for the
external tool.
As soon as the external tool computed contour lines, its output is read and interpreted as contour
lines – and the value of output point meta determines the value of point meta which will be
used to visualize the result.
An empty value means to use the z coordinate returned by the external tool.
Any other value is interpreted as a valid choice of point meta.
/pgfplots/contour gnuplot (style, no value)

\pgfplotsset{
contour gnuplot/.style={
contour external={
script={
unset surface;
\ifx\thecontourlevels\empty
set cntrparam levels \thecontournumber;
\else
set cntrparam levels discrete \thecontourlevels;
\fi
set contour;
set table \"\outfile\";
splot \"\infile\";
},
cmd={gnuplot \"\script\"},
#1,%
},
}
}
Note that contour gnuplot requires explicit scanline markers in the input stream, and it assumes
mesh/ordering=x varies.
Note that contour external lacks the intelligence to detect changes; it will always re-generate the
output (unless the -shell-escape feature is not active).
4.6.9 Filled Contour Plots

At the time of this writing, pgfplots comes with rudimentary support for filled contour plots. It is based
on colormap access=piecewise constant combined with shader=interp:
2 \begin{axis}[
domain=-2:2,
view={0}{90},
1 colormap={CM}{
samples of colormap=(14 of hot)},
colormap access=piecewise constant,
0 colorbar horizontal,
]
\addplot3[surf,shader=interp]
1 {exp(-x^2-y^2)*x};
\end{axis}
\end{tikzpicture}
2
2 1 0 1 2
0.4 0.2 0 0.2 0.4
The key idea is to use a colormap which contains as much colors as we want di↵erent contour levels. In our
example, we define a colormap called “CM” consisting of 14 colors sampled from colormap name=hot. These
14 colors resemble contour levels, and the filling is done by colormap access=piecewise constant. Note
that shader=interp eliminates any rectangular interpolation artifacts due to its smooth color interpolation.
This kind of filled contour plot has high quality and is a smooth vector graphics. However, the current
version of pgfplots comes with limitations in this context:
• Only uniform colormaps are available, meaning that only a uniform distance between adjacent contour
levels is possible.
• The only way to select individual contour levels is to use a suitable colormap definition, for example
using colors of colormap as outlined in Section 4.7.6 on page 192. However, this approach only
accepts mapped numbers. These are typically in the range [0, 1000] (maybe up to [0, 16000]) on input.
• This approach o↵ers no contour labels. One would need to overlay contour gnuplot in order to add
contour labels.
There are plans to eliminate these limitations eventually.
4.6.10 Parameterized Plots

Parameterized plots use the same plot types as documented in the preceding sections: both mesh and surface
plots are actually special parametrized plots where x and y are on cartesian grid points.
Parameterized plots just need a special way to provide the coordinates:
\begin{tikzpicture}
2
({sin(deg(x))},
{cos(deg(x))},
1 {2*x/(5*pi)});
\end{axis}
\end{tikzpicture}
0
0.5 1
0
0.5 0
1 1
The preceding example uses samples y=0 to indicate that a line shall be sampled instead of a matrix. The
curly braces are necessary because TEX can’t nest round braces. The single expressions here are used to
parametrize the helix.
Another example follows. Note that z buffer=sort is a necessary method here.

\begin{tikzpicture}
0
\addplot3[mesh,z buffer=sort,
samples=20,domain=-1:0,y domain=0:2*pi]
0.5 ({sqrt(1-x^2) * cos(deg(y))},
x);
\end{axis}
1 \end{tikzpicture}
0 0.5
0
1 0.5
\begin{tikzpicture}
1
\addplot3[mesh,z buffer=sort,
scatter,only marks,scatter src=z,
0 ({sqrt(1-x^2) * cos(deg(y))},
x);
1 \end{axis}
\end{tikzpicture}
0.5
0 0.5
0.5 0
1 0.5
\begin{tikzpicture}
\begin{axis}[view={60}{30},colormap/viridis]
0
\addplot3[surf,shader=interp,z buffer=sort,
0.5 ({sqrt(1-x^2) * cos(deg(y))},
x);
\end{axis}
1 \end{tikzpicture}
0 0.5
0
1 0.5
http://en.wikipedia.org/wiki/Klein_bottle
\begin{tikzpicture}
\begin{axis}[
xlabel=$x$,
0.5 ylabel=$y$,
view/h=-10,
0 title=\footnotesize
\url{http://en.wikipedia.org/wiki/Klein_bottle},
0.5 ]
\addplot3[
surf,
z buffer=sort,
4 colormap={periodic}{%
color=(blue)
2 color=(yellow)
y color=(orange)
1 0 1
0 color=(red)
x color=(orange)
color=(yellow)
color=(blue)},
domain=0:180, domain y=0:360,
samples=41, samples y=25,
variable=\u, variable y=\v,
point meta=u,
]
({-2/15 * cos(u) * (
3*cos(v) - 30*sin(u)
+ 90 *cos(u)^4 * sin(u)
- 60 *cos(u)^6 * sin(u)
+ 5 * cos(u)*cos(v) * sin(u))
},
{-1/15 * sin(u) * (3*cos(v)
- 3*cos(u)^2 * cos(v)
- 48 * cos(u)^4*cos(v)
+ 48*cos(u)^6 *cos(v)
- 60 *sin(u)
+ 5*cos(u)*cos(v)*sin(u)
- 5*cos(u)^3 * cos(v) *sin(u)
- 80*cos(u)^5 * cos(v)*sin(u)
+ 80*cos(u)^7 * cos(v) * sin(u))
},
{2/15 * (3 + 5*cos(u) *sin(u))*sin(v)});
\end{axis}
\end{tikzpicture}
4.6.11 3D Quiver Plots (Arrows)

Three dimensional quiver plots are possible with the same interface as their two-dimensional counterparts,
simply provide the third coordinate using quiver/w. Please refer to Section 4.5.8 for details and examples.
4.6.12 Image (Matrix) Plots

/pgfplots/matrix plot (no value)
\addplot+[matrix plot]
A matrix plot is very similar to a surface plot in that it produces a filled lattice of shaded rectangles.
The key di↵erence is that a surface plot expects the corners as input whereas a matrix plot expects
the cell mid points as input:
0 \begin{axis}[enlargelimits=false]
\addplot[
matrix plot,
3 4 5 mark=*,nodes near coords=\coordindex,
1
mesh/color input=explicit]
coordinates {
(0,0) [color=red] (1,0) [color=blue]
6 7 8 (2,0) [color=yellow]
2
(0,1) [color=black] (1,1) [color=brown]
(2,1) [color=magenta]
0 1 2
(0,2) [color=green] (1,2) [color=red]
(2,2) [color=white]
};
\end{axis}
\end{tikzpicture}
The example above shows a matrix plot along with the associated node near coords. The input
coordinates resemble the cell mid points, and the color given according to the format mesh/color
input=explicit defines the cell’s color. Note that a surface plot would connect the input vertices, but
here we have one cell for every mid point.
The size of each cell is determined according to the distance of adjacent cell mid points and cells are
always connected with each other; it is an interpolation scheme for adjacent cells.
A matrix plot is implemented using the same routines as surf, which means that it supports all of
the options associated with a surface plot. A standard application is to visualize the matrix using a
colormap:
8
\begin{tikzpicture}
0 1 2 \begin{axis}[enlargelimits=false,colorbar]
0
6 \addplot[matrix plot,
nodes near coords=\coordindex,mark=*,
3 4 5 point meta=explicit]
1 4 coordinates {
(0,0) [0] (1,0) [1] (2,0) [2]
(0,1) [3] (1,1) [4] (2,1) [5]

6 7 8 2
2 (0,2) [6] (1,2) [7] (2,2) [8]
};
0 \end{axis}
Note that matrix plot uses shader=flat corner by default. This ensures that the colors are taken
from the inputs, i.e. from the cell mid points. Use shader=faceted or shader=interp to interpolate
the colors. Similarly, other options which apply to surface plots (like faceted color, mesh/color
input, or mesh/ordering) can be combined with matrix plot as well.
Note that a matrix plot has one major di↵erence to almost all other plot handlers: providing matrix
plot to one of the \addplot commands in an axis will reconfigure the entire axis such that its display
fits. More precisely, a matrix plot implicitly sets the options
\pgfplotsset{
y dir=reverse,
axis on top
}
in the assumption that the matrix “coordinate” (0, 0) is typically the upper left corner (as opposed to
the lower left corner as for standard axes). This reconfiguration of the axis overrides any user options
for y dir and axis on top. If you prefer other values, you can and should use the starred version
matrix plot* which does not reconfigure the axis in any way.
A matrix plot automatically updates axis limits to match those of the cell corners.
\addplot[
matrix plot,
3 4 5 nodes near coords=\coordindex,mark=*,
1
mesh/cols=3,
point meta=explicit]
table[meta=C] {
6 7 8 x y C
2 0 0 0
1 0 1
2 0 2
0 1 2 0 1 3
1 1 4
2 1 5
0 2 6
1 2 7
2 2 8
};
\end{axis}
\end{tikzpicture}
Limitations: Due to current implementational restrictions, matrix plot can only update axis limits
successfully if the length of the matrix scanlines is known in advance. The example above contains
mesh/cols=3, and a “scanline” is a row (with 3 columns each). Consequently, the example above works
fine. However, If mesh/cols=3 would be unknown in advance, the current implementation assumes that
end–of–scanline markers are given in the input file according to empty line=scanline. As a rule of
thumb, you have to follow the following guidelines when using matrix plot:
1. The input matrix should have an empty line whenever one line of input is finished.
2. If there is no empty line, the length of each scanline must be given. For the standard con-
figuration mesh/ordering=x varies, pgfplots expects mesh/cols. For the alternative choice
mesh/ordering=y varies, pgfplots expects mesh/rows to be set.
3. The input matrix must have at least 2 rows and at least 2 columns. This allows pgfplots to
interpolate/extrapolate the vertices in a well-defined way.
If both of the first two conditions are violated, pgfplots fails to update the limits correctly and
generates the warning
Package pgfplots Warning: Automatic computation of axis limits for ’matrix plot’ is INACCURATE.
Please add the key ’mesh/cols=3’. You can also ignore this warning and deal with axis limits manually
If you encounter this warning, you should proceed according to the guidelines above or ignore the
warning and configure ymin and ymax manually.
A matrix plot expects a complete matrix on input; it is forbidden to leave holes in it. However, it
supports “invisible” cells: provide unbounded values for point meta in order to produce a “hole” in
the lattice:
8
\begin{tikzpicture}
0 1 2 \begin{axis}[enlargelimits=false,colorbar]
0
1 4 coordinates {
(0,0) [0] (1,0) [1] (2,0) [2]
(0,1) [3] (1,1) [nan] (2,1) [5]

6 7 8 2
2 (0,2) [6] (1,2) [7] (2,2) [8]
};
0 \end{axis}
Here, “unbounded” means a color value which is nan (not a number), inf, or -inf. In case of
mesh/color input=explicit, you can simply provide an empty explicit color to achieve the same
e↵ect:
\addplot[
matrix plot,
3 4 5 mark=*,nodes near coords=\coordindex,
1
mesh/color input=explicit]
coordinates {
(0,0) [color=red] (1,0) [color=blue]
6 7 8 (2,0) [color=yellow]
2
(0,1) [color=black] (1,1) []
(2,1) [color=magenta]
0 1 2
(0,2) [color=green] (1,2) [color=red]
(2,2) [color=white]
};
\end{axis}
\end{tikzpicture}
This feature works for shader=flat mean or shader=interp as well. However, these shaders require
color interpolation in order to determine the correct color for the visible cells. In this case, the missing cell
will be replaced by the lowest possible point meta (or by black in case of mesh/color input=explicit),
resulting in the following output:
8
\begin{tikzpicture}
\begin{axis}[enlargelimits=false,colorbar]
0
shader=interp,
point meta=explicit]
1 4 coordinates {
(0,0) [0] (1,0) [1] (2,0) [2]
(0,1) [3] (1,1) [nan] (2,1) [5]

2
2 (0,2) [6] (1,2) [7] (2,2) [8]
};
0 \end{axis}
A matrix plot will always be connected, even if the vertices form a distorted mesh. The input must be
a complete matrix, but its x and y coordinates (or maybe even z coordinates) can still be di↵erent from
what one expects from a matrix. The resulting cells are produced using interpolation/extrapolation:
8
\begin{tikzpicture}
0 1 \begin{axis}[enlargelimits=0.2,colorbar]
0 2
6 \addplot[
matrix plot,
3 4 5 nodes near coords=\coordindex,mark=*,
1
4 point meta=\coordindex]
table {
6 7 x y
2 2 0 0
8 1 -0.1
2 0.2
3 0 -0.2 1
1 0 1 2 3 1 1
2 1
0 2
1.3 2
2 2.5
};
\end{axis}
\end{tikzpicture}
Note that the expected usage of matrix plot is to have something which is at least close to a matrix,
and the axis limits may become inaccurate if the input forms a highly irregular mesh. In particular, the
axis limits in z direction are highly inaccurate for matrix plots, but the visualization is precise. Applying
matrix plot to a three–dimensional input might also need z buffer=auto since matrix plot disables
z buffer by default (in order to ensure correct color associations, i.e. to avoid reorderings). If you really
need three–dimensional matrix plots, you may want to specify surf,mesh input=image directly, and
you may need to adopt axis limits manually. The main use–case of matrix plot is for input which
looks roughly like an image.
Technically, matrix plot is a style which configures y dir=reverse and axis on top as outlined
above, and which installes matrix plot* afterwards.
See also colormap access=direct for examples how to use matrix plot in conjunction with bitmap
graphics.
/tikz/matrix plot* (no value)
\addplot+[matrix plot*]
The same as matrix plot except that matrix plot* does not reconfigure the axis. As outlined above,
matrix plot reconfigures the axis for
y dir=reverse,
axis on top,
These options are missing for matrix plot*:
2 \begin{axis}[enlargelimits=false, axis on top]
\addplot[matrix plot*,
1
coordinates {
(0,0) [0] (1,0) [1] (2,0) [2]
0 1 2 (0,1) [3] (1,1) [4] (2,1) [5]

0
(0,2) [6] (1,2) [7] (2,2) [8]
};
0 1 2 \end{axis}
\end{tikzpicture}
Actually, matrix plot* is a style which is implemented as

\pgfplotsset{
matrix plot*/.style={%
surf,
mesh input=image,
shader=flat corner,
z buffer=none,
},
}
Where the most important key is mesh input. Note that z buffer=none avoids reorderings which
might a↵ect the color association. Use z buffer=auto if you have a need to apply reorderings. Be
aware that z buffer reorders the color data.
/tikz/imagesc (no value)
\addplot+[imagesc]
A pure alias for matrix plot because imagesc is the reference to this type of plot in other applications.
/tikz/imagesc* (no value)
\addplot+[imagesc*]
A pure alias for matrix plot*.
4.6.13 Patch Plots

/pgfplots/patch (no value)
\addplot+[patch]
Patch plots are similar to mesh and surf plots in that they describe a filled area by means of a geometry.
However, patch plots are defined by explicitly providing the elements of the geometry: they expect a
sequence of triangles (or other patch types) which make up the mesh.
There are two dimensional and three dimensional patch plots, both with the same interfaces which are
explained in the following sections.
The standard input format (constituted by mesh input=patches) is to provide a sequence of coordinates
(either two– or three–dimensional) as usual. Each consecutive set of points makes up a patch element,
which is often a triangle:
\begin{axis}
\addplot[patch]
table {
x y
0 0
0.5
1 1
2 0
% empty lines do not hurt, they are ignored here:
1 1
0 2 0
3 1
0 1 2 3 4
2 0
3 1
4 0
};
\end{axis}
\end{tikzpicture}
Patch plots use point meta to determine fill colors. In its initial configuration, point meta will be set
to the y coordinate (or the z coordinate for three dimensional patch plots). Set point meta somehow
to color the patches:
\begin{axis}
\addplot[patch]
table[point meta=\thisrow{c}] {
x y c
0 0 0.2
0.5
1 1 0
2 0 1
1 1 0
2 0 1
0 3 1 0
0 1 2 3 4 2 0 1
3 1 0
4 0 0.5
};
\end{axis}
\end{tikzpicture}
Patch plots make use of the mesh configuration, including the shader. Thus, the example above uses the
initial shader=faceted (which uses the mean color data to determine a triangle’s color and a related
stroke color). The shader=interp yields the following result:
\begin{axis}
\addplot[patch,shader=interp]
x y c
0 0 0.2
0.5
1 1 0
2 0 1
1 1 0
2 0 1
0 3 1 0
0 1 2 3 4 2 0 1
3 1 0
4 0 0.5
};
\end{axis}
\end{tikzpicture}
For triangles, shader=interp results in linearly interpolated point meta values throughout each indi-
vidual triangle, which are then mapped to the color map (a technique also known as Gouraud shading).
The color data does not need to be continuous, it is associated to triangle vertices. Thus, changing some
of the color values allows individually shaded regions:
\begin{axis}
\addplot[patch,shader=interp]
x y c
0 0 0.2
0.5
1 1 0
2 0 1
1 1 0
2 0 -1
0 3 1 0
0 1 2 3 4 2 0 0.5
3 1 1
4 0 0.5
};
\end{axis}
\end{tikzpicture}
Two dimensional patch plots simply draw triangles in their order of appearance. In three dimensions,
single elements are sorted according to their view depth, with foreground elements drawn on top of
background elements (“Painter’s algorithm”, see z buffer=sort).
/pgfplots/patch table={htable file name or inline tablei} (initially empty)

/pgfplots/patch table with point meta={htable file name or inline tablei} (initially empty)
/pgfplots/patch table with individual point meta={htable file name or inline tablei} (initially
empty)
Allows to provide patch connectivity data stored in an input table.
A non–empty argument for patch table enables patch input mode. Now, the standard input
stream is a long list of vertices which are stored in an array using their \coordindex as key. Each
row of htable file name or inline tablei makes up one patch, defined by indices into the vertex array:
\begin{axis}
\addplot[patch,table/row sep=\\,patch table={%
0 1 2\\
1 2 3\\
4 3 5\\
0.5
}]
table[row sep=\\,point meta=\thisrow{c}] {
x y c \\
0 0 0.2\\% 0
1 1 0 \\% 1
0 2 0 1 \\% 2
3 1 0 \\% 3
0 1 2 3 4 2 0 0.5\\% 4
4 0 0.5\\% 5
};
\end{axis}
\end{tikzpicture}
The example consists of two separate tables. The patch table argument is a table, provided inline
where rows are separated by \\ (which is the purpose of the row sep=\\ key as you guessed31 ).
The patch table here declares three triangles: the triangle made up by vertex #0, #1 and #2,
the triangle made up by #1, #2 and #3 and finally the one using the vertices #4, #3 and #5.
The vertices as such are provided using the standard input methods of pgfplots; in our case
using a table as well. The standard input simply provides coordinates (and point meta) which are
stored in the vertex array; you could also have used plot coordinates to provide them (or plot
expression).
The argument to patch table needs to be a table – either a file name or an inline table as in the
example above. The first n columns of this table are assumed to contain indices into the vertex array
(which is made up using all vertices of the standard input as explained in the previous paragraph).
The entries in this table can be provided in floating point, just make sure they are not rounded.
The variable n is the number of vertices required to make up a single patch. For triangular patches,
it is n = 3, for patch type=bilinear it is n = 4 and similar for other choices of patch type.
The alternative patch table with point meta is almost the same as patch table – but it allows
to provide (a single) point meta (color data) per patch instead of per vertex. Here, a further column
of the argument table is interpreted as color data:
31 Note that the choice row sep=\\ is much more robust here: newlines would be converted to spaces by T X before pgfplots
E
had a chance to see them.
\begin{axis}
% this uses per-patch color data:
\addplot[patch,table/row sep=\\,
patch table with point meta={%
0 1 2 100\\
0.5
1 2 3 10\\
4 3 5 0\\
}]
table[row sep=\\] {
x y \\
0 0 0 \\% 0
1 1 \\% 1
0 1 2 3 4 2 0 \\% 2
3 1 \\% 3
2 0 \\% 4
4 0 \\% 5
};
\end{axis}
\end{tikzpicture}
The patch table with point meta always prefers point meta data from the provided table ar-
gument. However, it is still supported to write point meta=\thisrow{hcolnamei} or similar con-
structs – but now, hcolnamei refers to the provided table argument. More precisely, point meta is
evaluated in a context where the patch connectivity has been resolved and the patch table with
point meta is loaded.
The other alternative patch table with individual point meta is very similar, but instead of
a flat color per patch, it allows to write one color value for every patch:
\begin{axis}
% this uses n per-patch color values:
\addplot[patch,shader=interp,
table/row sep=\\,
patch table with individual point meta={%
0.5
0 1 2 100 100 100\\% V_0 V_1 V_2 C_0 C_1 C_2
1 2 3 10 0 50\\
4 3 5 0 0 100\\
}]
table[row sep=\\] {
0 x y \\
0 0 \\% 0
0 1 2 3 4 1 1 \\% 1
2 0 \\% 2
3 1 \\% 3
2 0 \\% 4
4 0 \\% 5
};
\end{axis}
\end{tikzpicture}
To find the point meta data for vertex #i, i = 0, 1, 2, pgfplots searches in column i + n where n
is the number of vertices for patch type (in our case, n = 3).
Technical remark: The key patch table with individual point meta automatically installs
point meta=explicit as well. It might be confusing to override the value of point meta here
(although it is allowed).
The patch table input type allows to reduce the size of geometries since vertices are stored just
once. pgfplots unpacks them into memory into the redundant format in order to work with
single patch elements32 . In case you experience TEX memory problems with this connectivity
input, consider using the redundant format. It uses other types of memory limits.
A more involved example is shown below; it uses \addplot3[patch] to visualize a three dimensional
patch plot, provided by means of a long sequence of patches:
32 The reason for such an approach is that T X doesn’t really know what an array is – and according to my experience, arrays
E
implemented by macros tend to blow up TEX’s memory limits even faster than the alternative.
\begin{tikzpicture}
\begin{axis}[axis equal]
% FokkerDrI_layer_0.patches.dat contains:
% # each row is one vertex; three consecutive
% # vertices make one triangle (patch)
50 % 105.577 -19.7332 2.85249
% 88.9233 -21.1254 13.0359
0 % 89.2104 -22.1547 1.46467
100 % # end of facet 0
50 % 105.577 -19.7332 2.85249
0 % 105.577 -17.2161 12.146
100 % 88.9233 -21.1254 13.0359
100 100 \addplot3[patch]
file
{plotdata/FokkerDrI_layer_0.patches.dat};
\end{axis}
\end{tikzpicture}
The ordering in which triangles are specified is irrelevant, three–dimensional patch plots use z
buffer=sort to sort patches according to their depth (defined as mean depth over each vertex), where
foreground patches are drawn on top of background patches. This so–called “Painter’s algorithm” works
well for most meshes. If it fails, consider using patch refines=1 or patch refines=2 to split larger
elements into small ones automatically.
The drawing color associated to single vertices can be changed using the point meta key (which
is the common method to configure color data in pgfplots). The initial configuration is point
meta=z for three dimensional patch plots, i.e. to use the z coordinate also as color data. Use point
meta=\thisrow{hcolnamei} in conjunction with \addplot3[patch] table to load a selected table col-
umn.
Patch plots are (almost) the same as mesh or surf plots, they only have more freedom in their in-
put format (and a more complicated geometry). Actually, “patch” is just a style for surf,mesh
input=patches. In other words, patch is the same as surf, it even shares the same internal implemen-
tation. Thus, most of the keys to configure mesh or surf plots apply to patch as well, especially shader
and z buffer. As already mentioned, \addplot3[patch] automatically activates z buffer=sort to
ensure a good drawing sequence. The shader can be used to modify the appearance:
\begin{tikzpicture}
\begin{axis}
% FokkerDrI_layer_0.facetIdx.dat contains:
% # each row makes up one facet; it
50 % # consists of 0-based indices into
% # the vertex array
% 0 1 2 % triangle of vertices #0,#1 and #2
0
% 0 3 1 % triangle of vertices #0,#3 and #1
% 3 4 1
% 5 6 7
50 100 % 6 8 7
100 0 % 8 9 7
0 % 8 10 9
100 100 % ...
% while FokkerDrI_layer_0.vertices.dat contains
% 105.577 -19.7332 2.85249 % vertex #0
% 88.9233 -21.1254 13.0359 % vertex #1
% 89.2104 -22.1547 1.46467 % vertex #2
% 105.577 -17.2161 12.146
% 105.577 -10.6054 18.7567
% 105.577 7.98161 18.7567
% 105.577 14.5923 12.146
% ...
\addplot3[patch,shader=interp,
patch table=
{plotdata/FokkerDrI_layer_0.facetIdx.dat}]
file
{plotdata/FokkerDrI_layer_0.vertices.dat};
\end{axis}
\end{tikzpicture}
See the description of shader=interp for details and remarks. The example above makes use of the
alternative syntax to provide a geometry: the patch table input. It allows to provide vertices separate
from patch connectivity, where each patch is defined using three indices into the vertex array as discussed
above.
\begin{tikzpicture}
50 \begin{axis}[view/h=70]
% FokkerDrI_layer_0.patches.dat contains:
% # each row is one vertex; three consecutive
0 % # vertices make one triangle (patch)
% 105.577 -19.7332 2.85249
% 88.9233 -21.1254 13.0359
50
% 89.2104 -22.1547 1.46467
% # end of facet 0
% 105.577 -19.7332 2.85249
100 % 105.577 -17.2161 12.146
0 % 88.9233 -21.1254 13.0359
100
100 100 \addplot3[patch,mesh]
file
{plotdata/FokkerDrI_layer_0.patches.dat};
\end{axis}
\end{tikzpicture}
/pgfplots/mesh input=lattice|patches|image
This key controls how input coordinates are decoded to get patches. It is used only if patch table
is empty (patch table has its own way to decode input coordinates). Usually, you won’t need to
bother with this key as it is set implicitly.
The choice mesh input=lattice is the initial configuration for mesh and surf plots: it expects
input in a compact matrix form as described at the beginning of this section starting with page 122
and requires a mesh/ordering and perhaps end–of–scanline markers. It yields patches with exactly
four corners and is compatible with patch type=rectangle and patch type=bilinear (the latter
requiring to load the patchplots library). In this case, the input stream contains the vertices.
The choice mesh input=patches is implicitly set when you use the patch style (remember that
surf is actually some sort of patch plot on its own). It expects the input format as described for
patch plots, i.e. n consecutive coordinates make up the vertices of a single patch where n is the
expected number of vertices for the configured patch type.
The choice mesh input=image implements matrix plot: it expects input in the same technical
format as for mesh input=lattice (i.e. using end–of–scanline markers, mesh/ordering, etc), but
the coordinates are interpreted as patch mid points. It yields patches with exactly four corners and
is compatible with patch type=rectangle and patch type=bilinear. The patches are generated
by interpolating the coordinates between adjacent cell mid points. Vertices which have only one
adjacent cell mid point are extrapolated. This choice is unsupported for one–dimensional inputs,
i.e. for lines or for 1 ⇥ n or n ⇥ 1 matrices. See matrix plot for more details.
Note that a non–empty patch table implies mesh input=patches.
/pgfplots/patch type=default|rectangle|triangle|line (initially default)

Defines the type of patch.
The initial configuration patch type=default checks the configuration of mesh input: for mesh
input=patches, it uses triangle. For mesh input=lattice, it checks if there is just one row or
just one col and uses patch type=line in such a case, otherwise it uses patch type=rectangle.
The choice patch type=rectangle expects n = 4 vertices. The vertices can be either encoded as
a matrix or, using mesh input=patches, in the sequence in which you would connect the vertices:
Rectangle from matrix input \begin{tikzpicture}
(2) (3) \begin{axis}[nodes near coords={(\coordindex)},
1 title=Rectangle from matrix input]
% note that surf implies ’patch type=rectangle’
\addplot[surf,mesh/rows=2,patch type=rectangle]
coordinates {
(0,0) (1,0)
(0,1) (1,1)
0.5 };
\end{axis}
\end{tikzpicture}
(0) (1)
0
0 0.2 0.4 0.6 0.8 1
Rectangle from patch input \begin{tikzpicture}
(3) (2) \begin{axis}[nodes near coords={(\coordindex)},
1 title=Rectangle from patch input]
\addplot[patch,patch type=rectangle]
coordinates {
(0,0) (1,0) (1,1) (0,1)
};
\end{axis}
(0) (1)
0
0 0.2 0.4 0.6 0.8 1
As for all other patch type values, the vertices can be arbitrary two– or three–dimensional points,
there may be even two on top of each other (resulting in a triangle). When used together with
shader=interp, patch type=rectangle is visualized using two Gouraud shaded triangles (see
below for triangle). It is the most efficient representation for interpolated shadings together
with mesh input=lattice since the input lattice is written directly into the pdf. Use patch
type=rectangle if you want rectangular elements and perhaps “some sort” of smooth shading. Use
patch type=bilinear of the patchplots library in case you need real bilinear shading. Examples
of such shadings can be found in Section 5.9.1.
The choice patch type=triangle expects n = 3 vertices which make up a triangle. The ordering
of the vertices is irrelevant:
(2) \begin{tikzpicture}
1 \begin{axis}[nodes near coords={(\coordindex)}]
\addplot[patch,patch type=triangle]
coordinates {
(0,0) (1,0) (0,1)
};
0.5 \end{axis}
\end{tikzpicture}
(0) (1)
0
0 0.2 0.4 0.6 0.8 1
The use of shader=interp is implemented by means of linear interpolation of the three color values
(specified with the point meta key) between the corners; the resulting interpolated point meta
values are then mapped into the actual colormap. This type of interpolation is called Gouraud
shading. Examples of such shadings can be found in Section 5.9.1.
The choice patch type=line expects n = 2 vertices which make up a line. It is used for one-
dimensional mesh plots (see Section 4.5.13 for examples).
There are more values for patch type like bilinear, triangle quadr, biquadratic, coons,
polygon and tensor bezier. Please refer to the separate patchplots library in Section 5.9.
/pgfplots/every patch (style, no value)

This style will be installed as soon as the patch plot handler is activated.
\pgfplotsset{
every patch/.style={miter limit=1}
}
which improves display of sharp triangle corners significantly (see the Tik Z manual for details about
miter limit and line join parameters).
There is much more to say about patch plots, like patch type which allows triangles, bilinear elements,
quadratic triangles, biquadratic quadrilaterals, coons patches; the patch refines key which allows
automatic refinement, patch to triangles which triangulates higher order elements; how matrix data
can be used for rectangular shapes and more. These details are subject of the patchplots library in
Section 5.9.
4.6.14 About 3D Const Plots and 3D Bar Plots

There are currently no equivalents of const plot and its variants or the bar plot types like ybar for three
dimensional axes, sorry.
4.6.15 Mesh/Surface Plots with Holes

Please refer to Section 4.5.14 for interrupted plots.
4.7 Markers, Linestyles, (Background-) Colors and Colormaps

The following options of Tik Z are available to plots.
4.7.1 Markers
This list is copied from [6, section 29]:
mark=*
mark=x
mark=+
And with \usetikzlibrary{plotmarks}:
mark=
mark=|
mark=o
mark=asterisk
mark=star
4.7. MARKERS, LINESTYLES, (BACKGROUND-) COLORS AND COLORMAPS 177
mark=10-pointed star
mark=oplus
mark=oplus*
mark=otimes
mark=otimes*
mark=square
mark=square*
mark=triangle
mark=triangle*
mark=diamond
mark=diamond*
mark=halfdiamond*
mark=halfsquare*
mark=halfsquare right*
mark=halfsquare left*
mark=Mercedes star
mark=Mercedes star flipped
mark=halfcircle
One half is filled with white (more precisely, with mark color).
mark=halfcircle*
One half is filled with white (more precisely, with mark color) and the other
half is filled with the actual fill color.
mark=pentagon
mark=pentagon*
mark=ball
This marker is special and can easily generate big output files if there are lots
of them. It is also special in that it needs ball color to be set (in our case,
it is ball color=yellow!80!black.
p p
mark=text p p
This marker is special as it can be configured freely. The character (or even
text) used is configured by a set of variables, see below.
mark=cube
This marker is only available inside of a pgfplots axis, it draws a cube with
axis parallel faces. Its dimensions can be configured separately, see below.
mark=cube*
User defined It is possible to define new markers with \pgfdeclareplotmark, see below.
All these options have been drawn with the additional options
\draw[
gray,
thin,
mark options={%
scale=2,fill=yellow!80!black,draw=black
}
]
Please see Section 4.7.5 for how to change draw and fill colors. Note that each of the provided marks can
be rotated freely by means of mark options={rotate=90} or every mark/.append style={rotate=90}.
/tikz/mark size={hdimensioni} (initially 2pt)

This Tik Z option allows to set marker sizes to hdimensioni. For circular markers, hdimensioni is the
radius, for other plot marks it is about half the width and height.
/pgfplots/cube/size x={hdimensioni} (initially \pgfplotmarksize=2pt)

/pgfplots/cube/size y={hdimensioni} (initially \pgfplotmarksize=2pt)
/pgfplots/cube/size z={hdimensioni} (initially \pgfplotmarksize=2pt)
Sets the size for mark=cube separately for every axis.
/tikz/every mark (no value)

This Tik Z style can be reconfigured to set marker appearance options like colors or transformations like
scaling or rotation. pgfplots appends its cycle list options to this style.
1
\begin{tikzpicture}
\begin{axis}[y=2cm]
0.5
{(-2,0) (-1,1) (0,0) (1,1) (2,0)};
\end{axis}
\end{tikzpicture}
0
2 1 0 1 2
1
\tikzset{every mark/.append style={scale=2}}
\begin{tikzpicture}
\begin{axis}[y=2cm]
0.5
{(-2,0) (-1,1) (0,0) (1,1) (2,0)};
\end{axis}
0
\end{tikzpicture}
2 1 0 1 2
1
\begin{tikzpicture}
\begin{axis}[y=2cm]
\addplot+[
0.5
mark=halfcircle*,
every mark/.append style={rotate=90}]
coordinates
0
{(-2,0) (-1,1) (0,0) (1,1) (2,0)};
2 1 0 1 2
\addplot+[
mark=halfcircle*,
every mark/.append style={rotate=180}]
coordinates
{(-2,-0.1) (-1,0.9) (0,-0.1) (1,0.9) (2,-0.1)};
\end{axis}
\end{tikzpicture}
Note that every mark is kind of static in the sense that it is evaluated once only. If you need individually
colored markes as part of a scatter plot, you will need to resort to scatter/use mapped color.
1
\begin{tikzpicture}
\begin{axis}[y=2cm]
\addplot+[
0.5
mark=ball,
mark size=4pt,
scatter,% enable scatter
0
scatter src=rand,% the "color data"
2 1 0 1 2 % configure individual appearances:
{ball color=mapped color}]
coordinates
{(-2,0) (-1,1) (0,0) (1,1) (2,0)};
\end{axis}
\end{tikzpicture}
/pgfplots/no markers (style, no value)

A key which overrides any mark value set by cycle list of option lists after \addplot.
If this style is provided as argument to a complete axis, it is appended to every axis plot post such
that it disables markers even for cycle lists which contain markers.
/tikz/mark repeat={hinteger i} (initially empty)

Allows to draw only each nth mark where n is provided as hinteger i.
/tikz/mark phase={hinteger pi} (initially 1)

This option allows to control which markers are drawn. It is primarily used together with the Tik Z
option mark repeat=r: it tells Tik Z that the first mark to be draw should be the pth, followed by the
(p + r)th, then the (p + 2r)th, and so on.
1 1 % Preamble: \pgfplotsset{width=7cm,compat=1.13}
0.5 0.5
\begin{tikzpicture}
\begin{axis}[tiny]
0 0
\addplot+[scatter] {sin(deg(x))};
0.5 0.5
\end{axis}
6 4 2 0 2 4 6 6 4 2 0 2 4 6 \begin{tikzpicture}
\begin{axis}[tiny]
\addplot+[scatter,
mark repeat=3,mark phase=2]
{sin(deg(x))};
\end{axis}
\end{tikzpicture}
Here, p = 1 is the first point (the one with \coordindex= 0).
/tikz/mark indices={hindex listi} (initially empty)

Allows to draw only the marker whose index numbers are in the argument list.
/pgf/mark color={hcolor i} (initially empty)

Defines the additional fill color for the halfcircle, halfcircle*, halfdiamond* and halfsquare*
markers. An empty value uses white (which is the initial configuration). The value none disables filling
for this part.
These markers have two distinct fill colors, one is determined by fill as for any other marker and the
other one is mark color.
1
\begin{tikzpicture}
\begin{axis}[y=2cm]
\addplot[
0.5
blue,mark color=blue!50!white,
mark=halfcircle*]
coordinates
0
{(-2,0) (-1,1) (0,0) (1,1) (2,0)};
2 1 0 1 2
\addplot[
red,mark color=red!50!white,
mark=halfsquare*]
coordinates
{(-2,-0.1) (-1,0.9) (0,-0.1) (1,0.9) (2,-0.1)};
\end{axis}
\end{tikzpicture}
Note that this key requires pgf 2.10 or later.
/tikz/mark options={hoptionsi}
Resets every mark to {hoptionsi}.
/pgf/text mark={htexti} (initially p)

Changes the text shown by mark=text.
m m
With /pgf/text mark=m: m m
A A
With /pgf/text mark=A: A A
There is no limitation about the number of characters or whatever. In fact, any TEX material can be
inserted as htexti, including images.
/pgf/text mark style={hoptions for mark=text i}

Defines a set of options which control the appearance of mark=text.
If /pgf/text mark as node=false (the default), hoptionsi is provided as argument to \pgftext –
which provides only some basic keys like left, right, top, bottom, base and rotate.
If /pgf/text mark as node=true, hoptionsi is provided as argument to \node. This means you can
provide a very powerful set of options including anchor, scale, fill, draw, rounded corners etc.
/pgf/text mark as node=true|false (initially false)

Configures how mark=text will be drawn: either as \node or as \pgftext.
The first choice is highly flexible and possibly slow, the second is very fast and usually enough.
\pgfdeclareplotmark{hplot mark namei}{hcodei}

Defines a new marker named hplot mark namei. Whenever it is used, hcodei will be invoked. It is
supposed to contain (preferrable pgf basic level) drawing commands. During hcodei, the coordinate
system’s origin denotes the coordinate where the marker shall be placed.
Please refer to [6] section “Mark Plot Handler” for more detailed information.
/pgfplots/every axis plot post (style, initially )

The every axis plot post style can be used to overwrite parts (or all) of the drawing styles which
are assigned for plots.
1
% Overwrite any cycle list:
\pgfplotsset{
every axis plot post/.append style={
0.5
mark=triangle,
every mark/.append style={rotate=90}}}
\begin{tikzpicture}
0
\begin{axis}[y=2cm]
2 1 0 1 2 \addplot coordinates
{(-2,0) (-1,1) (0,0) (1,1) (2,0)};
\end{axis}
\end{tikzpicture}
Markers paths are not subjected to clipping as other parts of the figure. Markers are either drawn
completely or not at all.
Tik Z o↵ers more options for marker fine tuning, please refer to [6] for details.
4.7.2 Line Styles

The following line styles are predefined in Tik Z.
/tikz/solid (style, no value)
/tikz/dotted (style, no value)
/tikz/densely dotted (style, no value)
/tikz/loosely dotted (style, no value)
/tikz/dashed (style, no value)
/tikz/densely dashed (style, no value)
/tikz/loosely dashed (style, no value)
/tikz/dashdotted (style, no value)

/tikz/densely dashdotted (style, no value)
/tikz/loosely dashdotted (style, no value)
/tikz/dashdotdotted (style, no value)
/tikz/densely dashdotdotted (style, no value)
/tikz/loosely dashdotdotted (style, no value)
since these styles apply to markers as well, you may want to consider using
\pgfplotsset{
every mark/.append style={solid}
}
in marker styles.
Besides linestyles, pgf also o↵ers (a lot of) arrow heads. Please refer to [6] for details.
4.7.3 Edges and Their Parameters

When pgfplots connects points, it relies on pgf drawing parameters to create proper edges (and it only
changes them in the every patch style).
It might occasionally be necessary to change these parameters:
/tikz/line cap=round|rect|butt (initially butt)

/tikz/line join=round|bevel|miter (initially miter)
/tikz/miter limit=hfactor i (initially 10)
These keys control how lines are joined at edges. Their description is beyond the scope of this manual,
so interested readers should consult [6].
Here is just an example illustrating why it might be of interest to study these parameters:
1 % requires \usetikzlibrary{spy}
1
\begin{tikzpicture}[spy using outlines=
{circle, magnification=6, connect spies}]
\begin{axis}[no markers,grid=major,
every axis plot post/.append style={thick}]
0.5
{(0, 0.0) (0, 0.9) (1, 0.9) (2, 1) (3, 0.9) (80, 0)};
\addplot +[line join=round] coordinates
{(0, 0.0) (0, 0.9) (2, 0.9) (3, 1) (4, 0.9) (80, 0)};
\addplot +[line join=bevel] coordinates
{(0, 0.0) (0, 0.9) (3, 0.9) (4, 1) (5, 0.9) (80, 0)};
0 \addplot +[miter limit=5] coordinates
{(0, 0.0) (0, 0.9) (4, 0.9) (5, 1) (6, 0.9) (80, 0)};
0 20 40 60 80
\coordinate (spypoint) at (3,1);
\coordinate (magnifyglass) at (60,0.7);
\end{axis}
\spy [blue, size=2.5cm] on (spypoint)

in node[fill=white] at (magnifyglass);
\end{tikzpicture}
4.7.4 Font Size and Line Width

Often, one wants to change line width and font sizes for plots. This can be done using the following options
of Tik Z.
/tikz/font={hfont namei} (initially \normalfont)

Sets the font which is to be used for text in nodes (like tick labels, legends or descriptions).
A font can be any LATEX argument like \footnotesize or \small\bfseries33 .
It may be useful to change fonts only for specific axis descriptions, for example using
\pgfplotsset{
tick label style={font=\small},
label style={font=\small},
legend style={font=\footnotesize}
}
See also the predefined styles normalsize, small and footnotesize in Section 4.10.2.
/tikz/line width={hdimensioni} (initially 0.4pt)

Sets the line width. Please note that line widths for tick lines and grid lines are predefined, so it may
be necessary to override the styles every tick and every axis grid.
The line width key is changed quite often in Tik Z. You should use
\pgfplotsset{every axis/.append style={line width=1pt}}
or
\pgfplotsset{every axis/.append style={thick}}
to change the overall line width. To also adjust ticks and grid lines, one can use
\pgfplotsset{every axis/.append style={

line width=1pt,
tick style={line width=0.6pt}}}
or styles like

thick,
tick style={semithick}}}
The ‘every axis plot’ style can be used to change line widths for plots only.
/tikz/thin (no value)

/tikz/ultra thin (no value)
/tikz/very thin (no value)
/tikz/semithick (no value)
/tikz/thick (no value)
/tikz/very thick (no value)
/tikz/ultra thick (no value)
These Tik Z styles provide di↵erent predefined line widths.
33 ConT
EXt and plain TEX users need to provide other statements, of course.
1
10 \pgfplotsset{every axis/.append style={
font=\large,
2 line width=1pt,
10 tick style={line width=0.8pt}}}
L2 Error
\begin{tikzpicture}
3
10 d=2 \begin{loglogaxis}[
d=3 legend style={at={(0.03,0.03)},
4 anchor=south west},
10 d=4 xlabel=\textsc{Dof},
5 d=5 ylabel=$L_2$ Error
10 ]
d=6 \addplot coordinates {
(5,8.312e-02) (17,2.547e-02) (49,7.407e-03)
101 102 103 104 105 106 (129,2.102e-03) (321,5.874e-04) (769,1.623e-04)
(1793,4.442e-05) (4097,1.207e-05) (9217,3.261e-06)
Dof };
\addplot coordinates{
(7,8.472e-02) (31,3.044e-02) (111,1.022e-02)
(351,3.303e-03) (1023,1.039e-03) (2815,3.196e-04)
(7423,9.658e-05) (18943,2.873e-05)
(47103,8.437e-06)};
(9,7.881e-02) (49,3.243e-02) (209,1.232e-02)
(769,4.454e-03) (2561,1.551e-03)
(7937,5.236e-04) (23297,1.723e-04)
(65537,5.545e-05) (178177,1.751e-05)};
(11,6.887e-02) (71,3.177e-02) (351,1.341e-02)
(1471,5.334e-03) (5503,2.027e-03)
(18943,7.415e-04) (61183,2.628e-04)
(187903,9.063e-05) (553983,3.053e-05)};
(13,5.755e-02) (97,2.925e-02) (545,1.351e-02)
(2561,5.842e-03) (10625,2.397e-03)
(40193,9.414e-04) (141569,3.564e-04)
(471041,1.308e-04) (1496065,4.670e-05)};
\legend{$d=2$,$d=3$,$d=4$,$d=5$,$d=6$}
\end{loglogaxis}
\end{tikzpicture}
The preceding example defines data which is used a couple of times throughout this manual; it is refer-
enced by \plotcoords.
10 1 d=2 \pgfplotsset{every axis/.append style={
d=3 font=\footnotesize,
10 2 d=4 thin,
d=5 tick style={ultra thin}}}
L2 Error
d=6 \begin{tikzpicture}
10 3
\begin{loglogaxis}[
xlabel=\textsc{Dof},
10 4 ylabel=$L_2$ Error
]
% see above for this macro:
10 5
\plotcoords
\legend{$d=2$,$d=3$,$d=4$,$d=5$,$d=6$}
\end{loglogaxis}
101 102 103 104 105 106 \end{tikzpicture}
Dof
4.7.5 Colors
pgf uses the color support of xcolor. Therefore, the main reference for how to specify colors is the xcolor
manual [3]. The pgf manual [6] is the reference for how to select colors for specific purposes like drawing,
filling, shading, patterns etc. This section contains a short overview over the specification of colors in [3]
(which is not limited to pgfplots).
The package xcolor defines a set of predefined colors, namely red, green, blue, cyan,
magenta, yellow, black, gray, white, darkgray, lightgray, brown, lime,
olive, orange, pink, purple, teal, violet.
\begin{axis}[enlarge x limits=false]
\addplot[red,samples=500] {sin(deg(x))};
\addplot[orange,samples=7] {sin(deg(x))};
0
\addplot[teal,const plot,
samples=14] {sin(deg(x))};
\end{axis}
\end{tikzpicture}
1
4 2 0 2 4
Besides predefined colors, it is possible to mix two (or more) colors. For example, red!30!white
contains 30% of red and 70% of white. Consequently, one can build red!70!white to get 70% red
and 30% white or red!10!white for 10% red and 90% white. This mixing can be done with any color,
for example red!50!green, blue!50!yellow or green!60!black.
A di↵erent type of color mixing is supported, which allows to take 100% of each component. For
example, rgb,2:red,1;green,1 will add 1/2 part red and 1/2 part green and we repro-
duced the example from above. Using the denominator 1 instead of 2 leads to rgb,1:red,1;green,1
which uses 1 part red and 1 part green. Many programs allow to select pieces between
0, . . . , 255, so a denominator of 255 is useful. Consequently, rgb,255:red,231;green,84;blue,121
uses 231/255 red, 84/255 green and 121/255. This corresponds to the standard RGB color (231, 84, 121).
Other examples are rgb,255:red,32;green,127;blue,43, rgb,255:red,178;green,127;blue,43,
rgb,255:red,169;green,178;blue,43.
It is also possible to use RGB values, the HSV color model, the CMY (or CMYK) models, or the HTML
color syntax directly. However, this requires some more programming. I suppose this is the fastest (and
probably the most uncomfortable) method to use colors. For example,
\definecolor{color1}{rgb}{1,1,0}
\tikz \fill[color1]
(0,0) rectangle (1em,0.6em);
creates the color with 100% red, 100% green and 0% blue;
\definecolor{color1}{cmyk}{0.6,0.9,0.5,0.1}
\tikz \fill[color1]
creates the color with 60% cyan, 90% magenta, 50% yellow and 10% black;
\definecolor{color1}{HTML}{D0B22B}
\tikz \fill[color1]
creates the color with 208/255 pieces red, 178/255 pieces green and 43 pieces blue, specified in standard
HTML notation. Please refer to the xcolor manual [3] for more details and color models.
The xcolor package provides even more methods to combine colors, among them the prefix ‘-’ (minus)
which changes the color into its complementary color ( -black, -white, -red) or color wheel
calculations. Please refer to the xcolor manual [3].
/tikz/color={ha color i}
/tikz/draw={hstroke color i}
/tikz/fill={hfill color i}
These keys are (generally) used to set colors. Use color to set the color for both drawing and filling.
Instead of color={hcolor namei} you can simply write hcolor namei. The draw and fill keys only set
colors for stroking and filling, respectively.
Use draw=none to disable drawing and fill=none to disable filling34 .

Since these keys belong to Tik Z, the complete documentation can be found in the Tik Z manual [6,
Section “Specifying a Color”].
Color Spaces
Since pgfplots relies on xcolor, all mechanisms of xcolor to define color spaces apply here as well.
One of the most useful approaches is global color space conversion: if you want a document which contains
only colors in the cmyk color spaces, you can say
in order to convert all colors of the entire document (including all shaded) to cmyk.
The same can be achieved by means of the xcolor statement \selectcolormodel.
4.7.6 Color Maps

/pgfplots/colormap name={hcolor map namei} (initially hot)
Changes the current color map to the already defined map named hcolor map namei. The predefined
color maps are
viridis
hot
The definition can be found in the documentation for colormap/hot and colormap/viridis, respec-
tively. These, and further color maps, are described below.
Colormaps can be used, for example, in scatter plots and surface plots.
You can use colormap to create new color maps (see below).
/pgfplots/colormap={hnamei}{hcolor specificationi}
Defines a new colormap named hnamei according to hcolor specificationi and activates it using colormap
name={hnamei}.
The hcolor specificationi is a sequence of positions and associated colors where linear interpolation is
applied in-between. The syntax is very similar as the one used for pgf shadings described in [6, VIII –
Shadings]: it is a semicolon–separated series of
hcolor typei(ho↵seti)=(hcolor valuei); :
% possibility 1: like PGF shadings:

rgb(0cm)=(1,0,0); rgb(1cm)=(0,1,0); rgb255(2cm)=(0,0,255); gray(3cm)=(0.3); color(4cm)=(green)
If the distance between successive colors is the same, the ho↵seti can be omitted. The ‘;’ separators
are not necessary either:
% (simplified) possibility 2: skip ‘;’ and length arguments:

rgb=(1,0,0) rgb=(0,1,0) rgb255=(0,0,255) gray=(0.3) color=(green)
34 Up to now, plot marks always have a stroke color (some also have a fill color). This restriction may be lifted in upcoming
versions.
It is also possible to provide non-uniform distances between the di↵erent colors – if all single positions
can be projected onto a uniform grid. pgfplots will perform this interpolation automatically:
% non uniform spacing example: the mesh width is provided as first

% part of the specification.
\pgfplotsset{colormap={violetnew}
{[1cm] rgb255(0cm)=(25,25,122) color(1cm)=(white) rgb255(5cm)=(238,140,238)}}
In this last example, the mesh width has been provided explicitly and pgfplots interpolates the missing
grid points on its own. It is an error if the provided positions are no multiple of the mesh width. The
\pgfplotsset employs the public user interface to create a new color map named ‘violetnew’.
The single colors can be separated by semicolons ‘;’. The (optional) length describes how much of the
bar is occupied by the interval, it is interpreted relative to the complete length. If the length argument
is missing, it is taken to be the last specified length plus the last length di↵erence (the first color defaults
to 1cm in this case).
Colormap Input Format
Each entry in hcolor specificationi has the form

hcolor model i(hlengthi)=(hargumentsi) or
hspecial modei(hlengthi)=(hargumenti of hcolormap namei)
where the most common form is to specify a color using a hcolor model i like rgb=(0,0.5,1). The
hspecial modeis are discussed later in Section 4.7.6 on page 192; they are useful to access colors of
existing colormaps. The hlengthi argument is optional as discussed above. The entries can be separated
by semicolons ‘;’ or by white spaces. If the leftmost entry is given, it must have hlengthi=0pt. As
discussed, all entries will be placed on a uniform grid, i.e. the distance between adjacent hlengthi
arguments has to be the same (see the previous paragraph for automatic generation of intermediate
points). The complete length of a color map is irrelevant: it will be mapped linearly to an internal
range anyway (for efficient interpolation). The only requirement is that the left end must be at 0.
Available choices for hcolor model i are
rgb which expects hargumentsi of the form (hred i,hgreeni,hbluei) where each component is in the
interval [0, 1],
rgb255 which is similar to rgb except that each component is expected in the interval [0,255],
gray in which case hargumentsi is a single number in the interval [0, 1],
color in which case hargumentsi contains a predefined (named) color like ‘red’ or a color expression
like ‘red!50’,
cmyk which expects hargumentsi of the form (hcyani,hmagentai,hyellow i,hblack i) where each com-
ponent is in the interval [0, 1],
cmyk255 which is the same as cmyk but expects components in the interval [0, 255],
cmy which expects hargumentsi of the form (hcyani,hmagentai,hyellow i) where each component is in
the interval [0, 1],
hsb which expects hargumentsi of the form (hhuei,hsaturationi,hbrightnessi) where each component
is in the interval [0, 1],
Hsb which is the same as hsb except that hhuei is accepted in the interval [0, 360] (degree),
HTML which is similar to rgb255 except that each component is expected to be a hex number between
00 and FF,
wave which expects a single wave length as numeric argument in the range [363, 814].
3,000 \begin{tikzpicture}
\begin{axis}[
colormap={bw}{gray(0cm)=(0); gray(1cm)=(1)}]
\addplot+[scatter,only marks,
2,000
domain=0:8,samples=100]
{exp(x)};
\end{axis}
1,000 \end{tikzpicture}
0
0 2 4 6 8
The Colorspace of a Colormap
Attention: this section is essentially superfluos if you have configured the xcolor package to override
color spaces globally (for example by means of \usepackage[cmyk]{xcolor} before loading pgfplots),
see the end of this sub–section.
Even though a colormap accepts lots of color spaces on input (in fact, it accepts most or all that xcolor
provides), the output color of a colorspace has strict limitations. The output colorspace is the one in
which pgfplots interpolates between two other colors. To this end, it transforms input colors to the
output color space. The output colorspace is also referred to as “the colorspace of a colormap”.
There are three supported color spaces for a colormap: the GRAY, RGB, and CMYK color spaces.
Each access into a colormap requires linear interpolation which is performed in its color space. Color
spaces make a di↵erence: colors in di↵erent color spaces may be represented di↵erently, depending
on the output device. Many printers use CMYK for color printing, so providing CMYK colors might
improve the printing quality on a color printer. The RGB color space is often used for display devices.
The predefined colormaps in pgfplots all use RGB.
Whenever a new colormap is created, pgfplots determines an associated color space. Then, each color
in this specific colormap will be represented in its associated color space (converting colors automatically
if necessary). Furthermore, every access into the colormap will be performed in its associated color space
and every returned mapped color will be represented with respect to this color space. Furthermore,
every shading generated by shader=interp will be represented with respect to the colormap’s associated
color space.
The color space is chosen as follows: in case colormap default colorspace=auto (the initial configu-
ration), the color space depends on the first encountered color in hcolor specificationi. For rgb or gray
or color, the associated color space will be RGB (as it was in all earlier versions of pgfplots). For
cmyk, the associated color space will be CMYK. If colormap default colorspace is either gray, rgb
or cmyk, this specific color space is used and every color is converted automatically.
/pgfplots/colormap default colorspace=auto|gray|rgb|cmyk (initially auto)
Allows to set the color space of every newly created colormap. The choices are explained in the
previous paragraph.
It is (not yet) possible to change the color space of an existing colormap; re-create it if conversion
is required.
The macro \pgfplotscolormapgetcolorspace{hnamei} defines \pgfplotsretval to contain the
color space of an existing colormap name, if you are in doubt.
Note that this option has no e↵ect if you told xcolor to override the color space globally. More
precisely, the use of
or, alternatively,
will cause all colors to be converted to cmyk, and pgfplots honors this configuration. Consequently,
both these statements cause all colors to be interpolated in the desired color space, and all output
colors will use this colorspace. This is typically exactly what you need.
Predefined Colormaps
Available color maps are shown below.
/pgfplots/colormap/viridis (style, no value)

A style which installs the colormap “viridis” which has been defined by Stéfan van der Walt and
Nathaniel Smith for Matplotlib. It is designated to be the default colormap for Matplotlib starting
with version 2.0 and is released under the CC035 .
The choice viridis is a downsampled copy included in pgfplots.
\pgfplotsset{
colormap name=viridis,
}
This colormap has considerably better properties compared to other choices:

• its color distribution is perceptually uniform (compare the definition in the link below),
• it is suitable for moderate forms of color blindness,
• it is still good when printed in black and white.
Details about these properties can be found in http://bids.github.io/colormap.
Please use colormap name=viridis as this makes uses of the predefined colormap whereas
colormap/viridis will redefine it.
There is also a high resolution copy of viridis which is called colormap/viridis high res in the
colormaps library. It resembles the original resolution of the authors, but it is visually almost identically
to viridis and requires less resources in TEX.
/pgfplots/colormap/hot (style, no value)

A style which installs the colormap
\pgfplotsset{
colormap={hot}{color(0cm)=(blue); color(1cm)=(yellow); color(2cm)=(orange); color(3cm)=(red)}
}
This is a preconfigured color map.
/pgfplots/colormap/hot2 (style, no value)

A style which is equivalent to
\pgfplotsset{
/pgfplots/colormap={hot2}{[1cm]rgb255(0cm)=(0,0,0) rgb255(3cm)=(255,0,0)
rgb255(6cm)=(255,255,0) rgb255(8cm)=(255,255,255)}
}
Note that this particular choice ships directly with pgfplots, you do not need to load the colormaps
library for this value.
This colormap is similar to one shipped with Matlab (®) under a similar name.
35 https://creativecommons.org/about/cc0
/pgfplots/colormap/jet (style, no value)

\pgfplotsset{
/pgfplots/colormap={jet}{rgb255(0cm)=(0,0,128) rgb255(1cm)=(0,0,255)
rgb255(3cm)=(0,255,255) rgb255(5cm)=(255,255,0) rgb255(7cm)=(255,0,0) rgb255(8cm)=(128,0,0)}
}
/pgfplots/colormap/blackwhite (style, no value)

\pgfplotsset{
colormap={blackwhite}{gray(0cm)=(0); gray(1cm)=(1)}
}
/pgfplots/colormap/bluered (style, no value)

\pgfplotsset{
colormap={bluered}{
rgb255(0cm)=(0,0,180); rgb255(1cm)=(0,255,255); rgb255(2cm)=(100,255,0);
rgb255(3cm)=(255,255,0); rgb255(4cm)=(255,0,0); rgb255(5cm)=(128,0,0)}
}
\begin{axis}[colormap/bluered]
\addplot+[scatter,
scatter src=x,samples=50]
{sin(deg(x))};
\end{axis}
0
\end{tikzpicture}
1
6 4 2 0 2 4 6
Remark: The style bluered (re-)defines the color map and activates it. TEX will be slightly faster
if you call \pgfplotsset{colormap/bluered} in the preamble (to create the color map once) and use
colormap name=bluered whenever you need it. This remark holds for every color map style which
follows. But you can simply ignore this remark.
/pgfplots/colormap/cool (style, no value)

\pgfplotsset{
colormap={cool}{rgb255(0cm)=(255,255,255); rgb255(1cm)=(0,128,255); rgb255(2cm)=(255,0,255)}
}
/pgfplots/colormap/greenyellow (style, no value)

\pgfplotsset{
colormap={greenyellow}{rgb255(0cm)=(0,128,0); rgb255(1cm)=(255,255,0)}
}
/pgfplots/colormap/redyellow (style, no value)

\pgfplotsset{
colormap={redyellow}{rgb255(0cm)=(255,0,0); rgb255(1cm)=(255,255,0)}
}
/pgfplots/colormap/violet (style, no value)

\pgfplotsset{
colormap={violet}{rgb255=(25,25,122) color=(white) rgb255=(238,140,238)}
}
\pgfplotscolormaptoshadingspec{hcolormap namei}{hright end sizei}{h\macroi}

A command which converts a colormap into a pgf shading’s color specification. It can be used in
commands like \pgfdeclare*shading (see the pgf manual [6] for details).
The first argument is the name of a (defined) colormap, the second the rightmost dimension of the
specification. The result will be stored in h\macroi.
% convert ‘hot’ -> \result

\pgfplotscolormaptoshadingspec{hot}{8cm}\result
% define and use a shading in pgf:
\def\tempb{\pgfdeclarehorizontalshading{tempshading}{1cm}}%
% where ‘\result’ is inserted as last argument:
\expandafter\tempb\expandafter{\result}%
\pgfuseshading{tempshading}%
The usage of the result h\macroi is a little bit low–level.
Attention: pgf shadings are always represented with respect to the RGB color space. Consequently,
even CMYK hcolormap nameis will result in an RGB shading specification when using this method36 .
36 In case pgf should someday support CMYK shadings and you still see this remark, you can add the macro definition
\def\pgfplotscolormaptoshadingspectorgb{0} to your preamble.

\pgfplotscolormaptodatafile[hoptionsi]{hcolormap namei}{houtput filei}

Allows to export colormap data to a file.
1
\begin{tikzpicture}
0.5 \begin{axis}[y=1cm,table/col sep=comma]
\pgfplotscolormaptodatafile{hot}{hot.dat}
0
\addplot[red,mark=|] table[y index=1] {hot.dat};
0 200 400 600 800 1,000
\addplot[green,mark=|] table[y index=2] {hot.dat};
\addplot[blue,mark=|] table[y index=3] {hot.dat};
\end{axis}
\end{tikzpicture}
1
\begin{tikzpicture}
0.5 \begin{axis}[y=1cm,table/col sep=comma]
\pgfplotscolormaptodatafile{viridis}{viridis.dat}
0
\addplot[red] table[y index=1] {viridis.dat};
0 200 400 600 800 1,000
\addplot[green] table[y index=2] {viridis.dat};
\addplot[blue] table[y index=3] {viridis.dat};
\end{axis}
\end{tikzpicture}
Valid hoptionsi are

/pgfplots//pgfplots/colormap/output each nth=hnumi (initially 1)
Allows to downsample the color map by writing only each hnumi’s entry.
/pgfplots//pgfplots/colormap/output format=cvs|native (initially csv)

The choice csv generates a csv file where the first column is the o↵set of the color map and all
following are the color components.
The choice native generates a TEX file which can be \input in order to define the colormap for
use in pgfplots.
Note that there more available choices in the associated libraries, in particular, in the
colorbrewer library and in the colormaps library which need to be loaded by means of
\usepgfplotslibrary{colorbrewer,colormaps}.
Building Colormaps based on other Colormaps

A colormap definition of sorts colormap={hnamei}{hcolor specificationi} typically consists of hcolor
specificationsi made up from single colors, each with its own hcolor model i. As outlined in Section 4.7.6,
each entry in hcolor specificationi has the form
hcolor model i(hlengthi)=(hargumentsi) or
hspecial modei(hlengthi)=(hargumenti of hcolormap namei).
This section explains how to make use of hspecial modei in order to build colormaps based on existing
colormaps. To this end, pgfplots o↵ers the following values inside of a hcolor specificationi:
1. index of colormap(hlengthi)=(hindex i of hcolormap namei)

This key allows to identify a single color of hcolormap namei and use it as part of hcolor specificationi.
The hindex i is a numeric index 0, 1, . . . , N 1 where N is the size of hcolormap namei.
The hlengthi argument is optional as usual.
Note that index of colormap is also available as key for drawing operations.
2. color of colormap(hlengthi)=(hvaluei of hcolormap namei).
This key allows to interpolate a color within hcolormap namei and use the result as part of hcolor
specificationi. The interpolation point is a floating point number in the range [0, 1000] and is interpo-
lated using colormap access=map (i.e. using piecewise linear interpolation).
The hlengthi argument is optional as usual.
Note that color of colormap is also available as key for drawing operations.
3. colors of colormap(hlengthi)=(hlisti of hcolormap namei)

A convenience key which is equivalent to a sequence of color of colormap(hlengthi), one for each
element in hlisti. The hlisti is evaluated using \foreach. Note that you need round braces around the
argument.
The hlengthi argument is optional as usual. Note that only the first hlengthi can be set explicitly, all
following ones are deduced automatically.
Note that colors of colormap is also available during cycle list definitions.
4. samples of colormap(hlengthi)=(hnumber i of hcolormap namei) or
A convenience key which is equivalent to a sequence of hnumber i occurences of color of
colormap(hlengthi). Note that you need round braces around the argument.
5. indices of colormap(hlengthi)=(hlist of indicesi of hcolormap namei) or

A convenience key which is equivalent to a sequence of index of colormap(hlengthi), one for each
element in hlisti. The hlisti is evaluated using \foreach. Note that you need round braces around the
argument.
Choosing a Colormap Entry as Normal Color

/pgfplots/color of colormap=hvaluei
/pgfplots/color of colormap=hvaluei of hcolormap namei
[See also Section 4.7.6 on page 192 for how to employ this within colormap definitions]
Defines the Tik Z color to be the hvaluei of hcolormap namei. If hcolormap namei is omitted, the value
of colormap name is evaluated (i.e. the current colormap is used).
The argument hvaluei is expected to be a number in the range [0, 1000] where 0 resembles the lower end
of the color map and 1000 the upper end.
Current colormap (hot):
\tikz\fill[color of colormap={800}, thick,

draw=.!60!black]
(0,0) -- (1,1) -- (2,1) circle (10pt);
The key computes the requested color and calls color=.. Keep in mind that the magic color name ‘.’
always reflects the “current color”, i.e. the result of color=hsome color i. Also keep in mind that color
is a Tik Z command which merely defines the color, you also have to provide one of ‘draw’ or ‘fill’ such
that it has an e↵ect. Since ‘.’ is a normal color, we can write draw=.!60!black to combine it with
another color.
viridis:
\pgfplotsset{colormap name=viridis}
\tikz\fill[color of colormap={1000},thick,
draw=.!60!black]
(0,0) -- (1,1) -- (2,1) circle (10pt);
\tikz\fill[color of colormap={300 of viridis},thick,

draw=.!60!black]
(0,0) -- (1,1) -- (2,1) circle (10pt);
The argument hcolormap namei is either a valid argument of colormap name or a style name like
colormap/cool:
cool:
\tikz\fill[color of colormap={300 of colormap/cool},

thick,draw=.!60!black]
(0,0) -- (1,1) -- (2,1) circle (10pt);
This last syntax allows to evaluate colormaps lazily. However, if you have many references to the same
colormap, it makes sense to write \pgfplotsset{colormap/cool} first followed by many references to
color of colormap={... of cool} in order to avoid unnecessary lazy evaluations.
It is possible to write lots of invocations without an explicit hcolormap namei, i.e. lots of invocations of
sorts color of colormap=hvaluei. They will all use the colormap name which is active at that time.
Note that there are actually keys with two key prefixes: /pgfplots/color of colormap and an alias
/tikz/color of colormap. This allows to use the keys both for plain Tik Z graphics and for pgfplots.
See also colormap access=map.
/pgfplots/index of colormap=hindex i
/pgfplots/index of colormap=hindex i of hcolormap namei
[See also Section 4.7.6 on page 192 for how to employ this within colormap definitions]
A variant of color of colormap which accesses the hcolormap namei by index. Consequently, the
argument hindex i is an integer number in the range 0, . . . , N 1 where N is the number of colors which
define the hcolormap namei. A hindex i outside of this range is automatically clipped to the upper
bound.
\pgfplotsset{colormap/jet}
\foreach \i in {0,...,\pgfplotscolormaplastindexof{jet}}{
\tikz\fill[index of colormap={\i of jet},
thick,
draw=.!60!black] (0,0) rectangle (10pt,6pt);
}
\pgfplotscolormapsizeof{hcolormap namei}
Expands to the number of colors which make up hcolormap namei.
If the argument hcolormap namei is an unknown colormap, it expands to 0.
\pgfplotscolormaplastindexof{hcolormap namei}
Expands to the last index of hcolormap namei, i.e. it is a convenience method to access N 1.
If the argument hcolormap namei is an unknown colormap, it expands to 1.
See also colormap access=direct.
4.7.7 Cycle Lists – Options Controlling Line Styles

/pgfplots/cycle list={hlisti}
/pgfplots/cycle list name={hnamei}
Allows to specify a list of plot specifications which will be used for each \addplot command without
explicit plot specification. Thus, the currently active cycle list will be used if you write either
\addplot+[hkeysi] ...; or if you don’t use square brackets as in \addplot[hexplicit plot specificationi]
...;.
The list element with index i will be chosen where i is the index of the current \addplot command (see
also the cycle list shift key which allows to use i + n instead). This indexing does also include plot
commands which don’t use the cycle list.
There are several possibilities to change the currently active cycle list:
Predefined Lists
Use one of the predefined lists37 ,

• color (from top to bottom)
0
\begin{tikzpicture}
\begin{axis}[
stack plots=y,stack dir=minus,
cycle list name=color]
5 \addplot coordinates {(0,1) (0.5,1) (1,1)};
\addplot coordinates {(0,1) (0.5,1) (1,1)};
0 0.2 0.4 0.6 0.8 1 \addplot coordinates {(0,1) (0.5,1) (1,1)};
\end{axis}
\end{tikzpicture}
• exotic (from top to bottom)

0
\begin{tikzpicture}
\begin{axis}[
cycle list name=exotic]
\end{axis}
\end{tikzpicture}
• black white (from top to bottom)

37 In an early version, these lists were called \coloredplotspeclist and \blackwhiteplotspeclist which appeared to be
unnecessarily long, so they have been renamed. The old names are still accepted, however.
0
\begin{tikzpicture}
\begin{axis}[
cycle list name=black white]
\end{axis}
\end{tikzpicture}
• mark list (from top to bottom)
0
\begin{tikzpicture}
\begin{axis}[
cycle list name=mark list]
5 \addplot+[blue] coordinates {(0,1) (0.5,1) (1,1)};
\addplot+[blue] coordinates {(0,1) (0.5,1) (1,1)};
p p p \addplot+[blue] coordinates {(0,1) (0.5,1) (1,1)};
10 a a a \addplot+[blue] coordinates {(0,1) (0.5,1) (1,1)};
0 0.2 0.4 0.6 0.8 1 \addplot+[blue] coordinates {(0,1) (0.5,1) (1,1)};
\end{axis}
\end{tikzpicture}
The mark list always employs the current color, but it doesn’t define one (the \addplot+ state-
ment explicitly sets the current color to blue).
The mark list is especially useful in conjunction with cycle multi list which allows to combine
it with other lists (for example linestyles or a list of colors).
• mark list* A list containing only markers. In contrast to mark list, all these markers are filled.
They are defined as (from top to bottom)
0
\begin{tikzpicture}
\begin{axis}[
cycle list name=mark list*]
0 0.2 0.4 0.6 0.8 1 \addplot+[blue] coordinates {(0,1) (0.5,1) (1,1)};
\end{axis}
\end{tikzpicture}
Similar to mark list, the mark list* always employs the current color, but it doesn’t define one
(see above for the \addplot+).
• color list (from top to bottom)
0
\begin{tikzpicture}
\begin{axis}[
cycle list name=color list]
\end{axis}
\end{tikzpicture}
The cycle list name=color choice also employs markers whereas color list uses only colors.
• linestyles (from top to bottom)
0
\begin{tikzpicture}
\begin{axis}[
cycle list name=linestyles]
\end{axis}
\end{tikzpicture}
• linestyles* contains more dotted line styles than linestyles (from top to bottom)
0
\begin{tikzpicture}
\begin{axis}[
cycle list name=linestyles*]
\end{axis}
\end{tikzpicture}
• auto The cycle list name=auto always denotes the most recently used cycle list activated by
cycle list or cycle list name.
The definitions of all predefined cycle lists follow (see the end of this paragraph for a syntax description).
\pgfplotscreateplotcyclelist{color}{%
blue,every mark/.append style={fill=blue!80!black},mark=*\\%
red,every mark/.append style={fill=red!80!black},mark=square*\\%
brown!60!black,every mark/.append style={fill=brown!80!black},mark=otimes*\\%
black,mark=star\\%
blue,every mark/.append style={fill=blue!80!black},mark=diamond*\\%
red,densely dashed,every mark/.append style={solid,fill=red!80!black},mark=*\\%
brown!60!black,densely dashed,every mark/.append style={
solid,fill=brown!80!black},mark=square*\\%
black,densely dashed,every mark/.append style={solid,fill=gray},mark=otimes*\\%
blue,densely dashed,mark=star,every mark/.append style=solid\\%
red,densely dashed,every mark/.append style={solid,fill=red!80!black},mark=diamond*\\%
}
\pgfplotscreateplotcyclelist{black white}{%
every mark/.append style={fill=gray},mark=*\\%
every mark/.append style={fill=gray},mark=square*\\%
every mark/.append style={fill=gray},mark=otimes*\\%
mark=star\\%
every mark/.append style={fill=gray},mark=diamond*\\%
densely dashed,every mark/.append style={solid,fill=gray},mark=*\\%
densely dashed,every mark/.append style={solid,fill=gray},mark=square*\\%
densely dashed,every mark/.append style={solid,fill=gray},mark=otimes*\\%
densely dashed,every mark/.append style={solid},mark=star\\%
densely dashed,every mark/.append style={solid,fill=gray},mark=diamond*\\%
}
\pgfplotscreateplotcyclelist{exotic}{%
teal,every mark/.append style={fill=teal!80!black},mark=*\\%
orange,every mark/.append style={fill=orange!80!black},mark=square*\\%
cyan!60!black,every mark/.append style={fill=cyan!80!black},mark=otimes*\\%
red!70!white,mark=star\\%
lime!80!black,every mark/.append style={fill=lime},mark=diamond*\\%
red,densely dashed,every mark/.append style={solid,fill=red!80!black},mark=*\\%
yellow!60!black,densely dashed,
every mark/.append style={solid,fill=yellow!80!black},mark=square*\\%
black,every mark/.append style={solid,fill=gray},mark=otimes*\\%
blue,densely dashed,mark=star,every mark/.append style=solid\\%
red,densely dashed,every mark/.append style={solid,fill=red!80!black},mark=diamond*\\%
}
% note that "." is the currently defined Tikz color.

\pgfplotscreateplotcyclelist{mark list}{%
every mark/.append style={solid,fill=\pgfplotsmarklistfill},mark=*\\%
every mark/.append style={solid,fill=\pgfplotsmarklistfill},mark=square*\\%
every mark/.append style={solid,fill=\pgfplotsmarklistfill},mark=triangle*\\%
every mark/.append style={solid},mark=star\\%
every mark/.append style={solid,fill=\pgfplotsmarklistfill},mark=diamond*\\%
every mark/.append style={solid,fill=\pgfplotsmarklistfill!40},mark=otimes*\\%
every mark/.append style={solid},mark=|\\%
every mark/.append style={solid,fill=\pgfplotsmarklistfill},mark=pentagon*\\%
every mark/.append style={solid},mark=text,text mark=p\\%
every mark/.append style={solid},mark=text,text mark=a\\%
}
In this context, a common fill color expression can be customized using mark list fill:
/pgfplots/mark list fill={hcolor i} (initially .!80!black)
Allows to customize the fill color for the mark list and mark list*.
For example, if you have black as color, the alternative choice mark list fill=.!50!white will
produce much better results.
\pgfplotsmarklistfill
Expands to \pgfkeysvalueof{/pgfplots/mark list fill}.
% note that "." is the currently defined Tikz color.

\pgfplotscreateplotcyclelist{mark list*}{%
every mark/.append style={solid,fill=\pgfplotsmarklistfill},mark=*\\%
every mark/.append style={solid,fill=\pgfplotsmarklistfill},mark=square*\\%
every mark/.append style={solid,fill=\pgfplotsmarklistfill},mark=triangle*\\%
every mark/.append style={solid,fill=\pgfplotsmarklistfill},mark=halfsquare*\\%
every mark/.append style={solid,fill=\pgfplotsmarklistfill},mark=pentagon*\\%
every mark/.append style={solid,fill=\pgfplotsmarklistfill},mark=halfcircle*\\%
every mark/.append style={solid,fill=\pgfplotsmarklistfill,rotate=180},mark=halfdiamond*\\%
every mark/.append style={solid,fill=\pgfplotsmarklistfill!40},mark=otimes*\\%
every mark/.append style={solid,fill=\pgfplotsmarklistfill},mark=diamond*\\%
every mark/.append style={solid,fill=\pgfplotsmarklistfill},mark=halfsquare right*\\%
every mark/.append style={solid,fill=\pgfplotsmarklistfill},mark=halfsquare left*\\%
}
\pgfplotscreateplotcyclelist{color list}{%
red,blue,black,yellow,brown,teal,orange,violet,cyan,green!70!black,magenta,gray}
\pgfplotscreateplotcyclelist{linestyles}{solid,dashed,dotted}
\pgfplotscreateplotcyclelist{linestyles*}{solid,dashed,dotted,dashdotted,dashdotdotted}
Defining Own Cycle Lists
The second choice for cycle lists is to provide each entry directly as argument to cycle list,
1
d=3 \begin{loglogaxis}[cycle list={%
2 {blue,mark=*},
10 d=4 {red,mark=square},
d=5 {dashed,mark=o},
3
10 d=6 {loosely dotted,mark=+},
{brown!60!black,
10 4 mark options={fill=brown!40},
mark=otimes*}}
]
5
10 \plotcoords
\legend{$d=2$,$d=3$,$d=4$,$d=5$,$d=6$}
\end{loglogaxis}
101 102 103 104 105 106 \end{tikzpicture}
(This example list requires \usetikzlibrary{plotmarks}).

The input format is described below in more detail.
Defining and Labeling Own Cycle Lists
The last method for cycle lists is to combine the define named cycle lists in the preamble and use them
with ‘cycle list name’:
\pgfplotscreateplotcyclelist{hnamei}{hlisti}
\pgfplotscreateplotcyclelist{mylist}{%
{blue,mark=*},
{red,mark=square},
{dashed,mark=o},
{loosely dotted,mark=+},
{brown!60!black,mark options={fill=brown!40},mark=otimes*}}
...
\begin{axis}[cycle list name=mylist]
...
\end{axis}
Defining Cycle Lists: Input Format
A cycle list is defined by key-value pairs of sorts cycle list={hlisti} or by the equivalent macro
outlined above, \pgfplotscreateplotcyclelist{hnamei}{hlisti}.
In this context, the argument hlisti is usually a comma separated list of lists of style keys like colors,
line styles, marker types and marker styles. This “comma list of comma lists” structure requires to
encapsulate the inner list using curly braces:
\pgfplotscreateplotcyclelist{mylist}{%
{blue,mark=*},
{red,mark=square},
{dashed,mark=o},
{loosely dotted,mark=+},
{brown!60!black,mark options={fill=brown!40},mark=otimes*}}
Alternatively, one can terminate the inner lists (i.e. those for one single plot) with ‘\\’:
\begin{axis}[cycle list={%
blue,mark=*\\%
red,mark=square\\%
dashed,mark=o\\%
loosely dotted,mark=+\\%
brown!60!black,mark options={fill=brown!40},mark=otimes*\\%
}
]
...
\end{axis}
In this case, the last entry also needs a terminating ‘\\’, but one can omit braces around the single
entries.
Manipulating Associations of Cycle Lists to Plots

/pgfplots/cycle list shift={hinteger i} (initially empty)
Allows to shift the index into the cycle list. If hinteger i is n, the list element i + n will be taken
instead of the ith one. Remember that i is the index of the current \addplot command (starting
with 0).
Since a cycle list is queried immediately when \addplot (or \addplot+) is called, you can adjust
the cycle list shift for selected plots:
\pgfplotsset{cycle list shift=3}

\addplot ....
\pgfplotsset{cycle list shift=-1}

\addplot ....
Special case: If the result is negative, i + n < 0, the list index (i + n) will be taken. For
example, cycle list shift=-10 and i < 10 will result in list index 10 i. Note that you can use
reverse legend to reverse legends, so this feature is probably never needed.
Defining Cycle Lists based on Color Maps
In addition to defining cycle lists from scratch, pgfplots supports dedicated input definitions of
cycle list=hlisti which allow to acquire values from an existing colormap. In this case, hlisti contains
keys enclosed in square brackets:
The first syntax, of colormap, allows to convert the colors of a colormap to a cycle list. It can
be specified without argument by means of cycle list={[of colormap]} in order to take the value
of the most recently assigned colormap name (i.e. the current colormap). It can also be specified as
cycle list={[of colormap=name]} in which case it will use the specified colormap name=name. In
both cases, the definition merely converts the colors as they are found in the colormap into the cycle
list, i.e. there is no interpolation involved. Applying this to the default colormap name=hot which has
4 colors results in the following example:
1 \begin{axis}[
2 colormap name=hot,colorbar,
3 cycle list={[of colormap]},
%
4 cycle from colormap manual style,
5 ]
0 0.5 1 10 \addplot coordinates {(0,1) (0.5,1) (1,1)};
\end{axis}
\end{tikzpicture}
Note that since hot has 4 colors, the cycle list also contains 4 entries which are repeated every 4
plots.
The second possibility is similar to color of colormap: it expects colors of colormap={hlisti} or
colors of colormap={hlisti} of hcolormap namei. This choice interpolates colors and expects a hlisti
of values in the range [0, 1000] where 0 is the lowest element in the colormap and 1000 is its highest
element:
1 \begin{axis}[
3 cycle list={
[colors of colormap={0,100,...,1000}]},
4 %
6 ]
\end{axis}
\end{tikzpicture}
In this case, we specified 11 colors and have 11 plots. Clearly, interpolated colors are of limited use and
are only applicable for special use-cases. Use only cycle lists of this sort if the colormap allows a
suitable distinction of adjacent plot lines! Note that colors of colormap is quite similar to the related
way to build colormaps based on existing colormaps as outlined in Section 4.7.6 on page 192.
Finally, there is the choice indices of colormap={hlisti}. As above, it accepts an optional ‘of’ clause
of the form indices of colormap={hlisti} of hcolormap namei. The main argument is a list of indices
0  Ni < N where N is the number colors in the colormap definition (compare the documentation of
index of colormap). Indices which are out of range are clipped to the nearest index. For example,
viridis comes with 18 elements and we can write
1 \begin{axis}[
2 colormap name=viridis,colorbar,
3 cycle list={
[indices of colormap={0,4,8,12,17} of viridis]},
4 %
6 ]
\end{axis}
\end{tikzpicture}
Note that ‘ of viridis’ is actually redundant as viridis was already selected in this case. In general,
the syntax allows to select any (defined) colormap. Note that indices of colormap is quite similar to
the related way to build colormaps based on existing colormaps as outlined in Section 4.7.6 on page 192.
Note that all these special lists are valid arguments for \pgfplotscreateplotcyclelist and can also
appear as sub-lists in cycle multi list and its variants.
Since creating a cycle list from a colormap necessarily results in plots without markers and line style
variations, it makes sense to combine the result with cycle multiindex* list, i.e. to join two existing
lists. The following example joins a pure color list with markers:
1 \begin{axis}[
3 cycle multiindex* list={
[colors of colormap={0,100,...,1000}]\nextlist
4 mark list\nextlist
5 },
6 %
p p p 7 cycle from colormap manual style,
a a a p 8 ]
a 9 \addplot coordinates {(0,1) (0.5,1) (1,1)};
\end{axis}
\end{tikzpicture}
Please refer to the next subsection for details about cycle multiindex* list.
Note that the preceding examples all use the following style.
/pgfplots/cycle from colormap manual style (style, no value)
A style defined in this manual. It has the value
\pgfplotsset{
cycle from colormap manual style/.style={
x=3cm,y=10pt,ytick=\empty,
colorbar style={x=,y=,ytick=\empty},
point meta min=0,point meta max=1,
stack plots=y,
y dir=reverse,colorbar style={y dir=reverse},
every axis plot/.style={line width=2pt},
legend entries={0,...,20},
legend pos=outer north east,
}
}
Remark: It is possible to call \pgfplotsset{cycle list={ha listi}} or cycle list name between
plots. Such a setting remains e↵ective until the end of the current TEX group (that means curly braces).
Every \addplot command queries the cycle list using the plot index; it doesn’t hurt if cycle lists
have changed in the meantime.
/pgfplots/cycle list/.define={hnamei}list
A command which merely calls \pgfplotscreateplotcyclelist{hnamei}{hlisti} without actually se-
lecting it as the current list.
Note that pgfplots uses this to implement its cycle list key as follows:
\pgfplotsset{
cycle list/.style={%
cycle list/.define={@internal@}{#1},%
cycle list name={@internal@}%
},
}
Building Block to Combine Di↵erent Cycle Lists

The following keys allow to combined di↵erent cycle lists in order to build more complex ones.
/pgfplots/cycle multi list=hlist 1 i\nextlisthlist 2 i\nextlist· · ·

This is one of two ways to employ more than one cycle list in order to determine the plot style (see
also cycle multiindex list for the other one). This is probably best explained using an example:
0
1
0 2
3
4
5
6
10
7
8
9
10
6 4 2 0 2 4 6
11
\begin{tikzpicture}
\begin{axis}[
cycle multi list={
red,blue\nextlist
solid,{dotted,mark options={solid}}\nextlist
mark=*,mark=x,mark=o
},
samples=3,
legend pos=outer north east
]
\addplot {x};
\addplot {x-1};
\addplot {x-2};
\addplot {x-3};
\addplot {x-4};
\addplot {x-5};
\addplot {x-6};
\addplot {x-7};
\addplot {x-8};
\addplot {x-9};
\addplot {x-10};
\addplot {x-11};
\end{axis}
\end{tikzpicture}
The provided cycle multi list consists of three lists. The style for a single plot is made up us-
ing elements of each of the three lists: the first plot has style red,solid,mark=*, the second has
red,solid,mark=x, the third has red,solid,mark=o. The fourth plot restarts the third list and uses
the next one of list 2: it has red,dotted,mark options={solid},mark=* and so on.
The last list will always be advanced for a new plot. The list before the last (in our case the second
list) will be advanced after the last one has been reset. In other words: cycle multi list allows a
composition of di↵erent cycle list in a lexicographical way38 .
The argument for cycle multi list is a sequence of arguments as they would have been provided for
cycle list, separated by \nextlist. In addition to providing a new cycle list, the hlist ii elements can
also denote cycle list name values (including the special auto cycle list which is the most recently
assigned cycle list or cycle list name). The final \nextlist is optional.
The list in our example above could have been written as
\begin{axis}[
cycle multi list={
red\\blue\\\nextlist
solid\\dotted,mark options={solid}\\\nextlist
mark=*\\mark=x\\mark=o\\
}]
as well (note the terminating \\ commands!).

38 For those who prefer formulas: The plot with index 0  i < N will use cycle list o↵sets i , i , . . . , i , 0  i
0 1 k m < Nm where k
is the number of arguments provided to cycle multi list and Nm is the number of elements in the mth cycle list. The o↵sets
im are computed in a loop { int tmp=i; for( int m=k-1; m>=0; m=m-1 ) { i m = tmp%N m; tmp = tmp/N m; }}.
Cycle color between successive plots, then marks
0
1
0 2
3
4
5
6
10
7
8
9
10
6 4 2 0 2 4 6
11
\begin{tikzpicture}
\begin{axis}[
title={Cycle color between successive plots, then marks},
cycle multi list={
mark list\nextlist
blue,red%
},
samples=3,
]
\addplot {x};
\addplot {x-1};
\addplot {x-2};
\addplot {x-3};
\addplot {x-4};
\addplot {x-5};
\addplot {x-6};
\addplot {x-7};
\addplot {x-8};
\addplot {x-9};
\addplot {x-10};
\addplot {x-11};
\end{axis}
\end{tikzpicture}
Using Sub–Lists The hlist ii entry can also contain just the first n elements of an already known
cycle list name using the syntax [hnumber i of]hcycle list namei. For example [2 of]mark list will
use the first 2 elements of mark list:
Cycle 2 marks between successive plots, then colors
0
1
0 2
3
4
5
6
10
7
8
9
10
6 4 2 0 2 4 6
11
\begin{tikzpicture}
\begin{axis}[
title={Cycle 2 marks between successive plots, then colors},
cycle multi list={%
color list\nextlist
[2 of]mark list
},
samples=3,
]
\addplot {x};
\addplot {x-1};
\addplot {x-2};
\addplot {x-3};
\addplot {x-4};
\addplot {x-5};
\addplot {x-6};
\addplot {x-7};
\addplot {x-8};
\addplot {x-9};
\addplot {x-10};
\addplot {x-11};
\end{axis}
\end{tikzpicture}
/pgfplots/cycle multiindex list=hlist 1 i\nextlisthlist 2 i\nextlist· · ·

This is one of two ways to employ more than one cycle list in order to determine the plot style (see
also cycle multi list for the other one). The di↵erence between the two choices is how the list index
is mapped into the sub lists. Let us start with our example:
0
1
0 2
3
4
5
6
10
7
8
9
10
6 4 2 0 2 4 6
11
\begin{tikzpicture}
\begin{axis}[
cycle multiindex list={
red,blue,teal\nextlist
solid\\dotted,mark options={solid}\\only marks\\\nextlist
mark=*,mark=x,mark=oplus\nextlist
},
samples=3,
]
\addplot {x};
\addplot {x-1};
\addplot {x-2};
\addplot {x-3};
\addplot {x-4};
\addplot {x-5};
\addplot {x-6};
\addplot {x-7};
\addplot {x-8};
\addplot {x-9};
\addplot {x-10};
\addplot {x-11};
\end{axis}
\end{tikzpicture}
The provided cycle multiindex list consists of three lists. The style for a single plot is made up
using elements of each of the three lists: the first plot has style red,solid,mark=*, the second has
blue,dotted,mark options={solid},mark=x, the third has teal,only marks,oplus. The fourth
plot restarts all lists and uses the same as the first plot, i.e. red,solid,mark=*.
Note that the second list uses the list-separator ‘\\’ which requires a final terminator as defined for
cycle list.
Thus, this style uses the same index into every list (a “multi index”). Consequently, it has consid-
erably less di↵erent choices than cycle multi list (which results in all possible variations), but its
combination method addresses di↵erent use-cases.
The argument for cycle multiindex list has the very same format as the one for cycle multi list,
including the special [2 of]mark list syntax and providing other cycle lists by name:
Same index of each input list
0
1
0 2
3
4
5
6
10
7
8
9
10
6 4 2 0 2 4 6
11
\begin{tikzpicture}
\begin{axis}[
title={Same index of each input list},
cycle multiindex list={
[3 of]mark list\nextlist
blue,red\nextlist
linestyles\nextlist
},
samples=3,
]
\addplot {x};
\addplot {x-1};
\addplot {x-2};
\addplot {x-3};
\addplot {x-4};
\addplot {x-5};
\addplot {x-6};
\addplot {x-7};
\addplot {x-8};
\addplot {x-9};
\addplot {x-10};
\addplot {x-11};
\end{axis}
\end{tikzpicture}
Note that cycle multiindex list accepts lists of di↵erent sizes. The size of a cycle multiindex
list is the size of the largest input list, all smaller input lists are padded with empty option lists. That
is why the previous example uses the color black for every third plot: there is no color in the second
list, and omitting the color results in black. As soon as the last item of the largest sub-list has been
used, the list is restarted.
/pgfplots/cycle multiindex* list=hlist 1 i\nextlisthlist 2 i\nextlist· · ·

A variant of cycle multiindex list which behaves in the same way – except for sub-lists of di↵erent
sizes.
As documented above, the unstarred version cycle multiindex list pads missing entries with empty
options lists until all list elements have the same size.
The starred key cycle multiindex* list restarts sub-lists independently whenever they reach their
end:
Same index of each input list, lists restarted
0
1
0 2
3
4
5
6
10
7
8
9
10
6 4 2 0 2 4 6
11
\begin{tikzpicture}
\begin{axis}[
title={Same index of each input list, lists restarted},
cycle multiindex* list={
[3 of]mark list\nextlist
blue,red\nextlist
linestyles\nextlist
},
samples=3,
]
\addplot {x};
\addplot {x-1};
\addplot {x-2};
\addplot {x-3};
\addplot {x-4};
\addplot {x-5};
\addplot {x-6};
\addplot {x-7};
\addplot {x-8};
\addplot {x-9};
\addplot {x-10};
\addplot {x-11};
\end{axis}
\end{tikzpicture}
This is the very same example as documented for the unstarred variant cycle multiindex list. How-
ever, the second sub-list has fewer elements – and while the unstarred variant resulted in black, the
starred variant restarts the second sub-list as soon as its two existing colors are consumed.
This style allows to concatenate lists in complex ways:
Same index of each input list
0
1
0 2
3
4
5
6
10
7
8
9
10
6 4 2 0 2 4 6
11
\begin{tikzpicture}
\begin{axis}[
title={Same index of each input list},
red,blue,teal,brown\nextlist
mark=*,mark=square*,mark=triangle*\nextlist
dashed,every mark/.append style={
solid,draw=.!50!black,fill=.,}\\
\nextlist
},
samples=3,
]
\addplot {x};
\addplot {x-1};
\addplot {x-2};
\addplot {x-3};
\addplot {x-4};
\addplot {x-5};
\addplot {x-6};
\addplot {x-7};
\addplot {x-8};
\addplot {x-9};
\addplot {x-10};
\addplot {x-11};
\end{axis}
\end{tikzpicture}
We see that the four di↵erent colors appear periodically as expected. We also see the three di↵erent
markers with their own period (which restarts every fourth plot as expected). But the third sub-list
contains just one element as we can see by its separator character ‘\\’ which appears just once at the end
of the list! Consequently, this list is restarted for every plot such that every plot receives its arguments.
4.7.8 Axis Background

/pgfplots/axis background (initially empty)
This is a style to configure the appearance of the axis as such. It can be defined and/or changed using the
axis background/.style={hoptionsi} method. A background path will be generated with hoptionsi,
which may contain fill colors or shadings.
\begin{tikzpicture}
\begin{axis}[
axis background/.style={fill=blue!10}]
0.2 \addplot3[surf,y domain=0:1]

{sin(deg(x)) * y*(1-y)};
0 \end{axis}
\end{tikzpicture}
0.2 1
4 0.5
2 0 2 4 0
Please note that legends are filled with white in the default configuration.
4.8. PROVIDING COLOR DATA - POINT META 211
x
108 e \begin{tikzpicture}
4x \begin{semilogyaxis}[
e
axis background/.style={
shade,top color=gray,bottom color=white},
102 legend style={fill=white}]
\addplot {exp(-x)};
\addplot {exp(-4*x)};
4
10 \legend{$e^{-x}$,$e^{-4x}$}
\end{semilogyaxis}
\end{tikzpicture}
10
10
6 4 2 0 2 4 6
Details about fill and shade can be found in the Tik Z manual, [6].
4.8 Providing Color Data - Point Meta

pgfplots provides features which modify plots depending on a special coordinate, the “point meta data”.
For example, scatter plots may vary marker colors, size or appearance depending on this special data. Surface
and mesh plots are another example: here, the color of a surface patch (or mesh part) depends on “point
meta”. The “meta data” of a node is not the position (which is given as (x, y) or (x, y, z)). It is “some
information” about that node; you could say: it is a specific property of the node. This is commonly called
“meta data”.
4.8.1 Point Meta Overview

In pgfplots, every node has its coordinate and its meta data. Thus, two–dimensional plots have three
values: x, y, and the point meta data. Three–dimensional plots have four values for each coordinate: x, y,
z, and the point meta.
In many cases, point meta is interpreted to be color data. To be more precise: it is interpreted to be
scalar color data which is mapped into the colormap (more about this approach in the next paragraphs).
However, point meta can be anything. Often, it is a single number as in the case of color data. That
number is mapped linearly to the interval [0, 1000] such that 0 corresponds to the smallest encountered
value and 1000 corresponds to the largest encountered scalar value. The mapped value is available as
\pgfplotspointmetatransformed. This special value allows to define some property of the plot: it can
be the color (together with colormap). It can also be the line width in a mesh plot (more precisely: the
line width could be defined to depend on the transformed meta data). An example for this approach
is on page 93 where a quiver plot’s arrow line with depends on \pgfplotspointmetatransformed. The
value can also be used to define mark size as in the example for scatter/@pre marker code on page 117.
However, point meta data does not necessarily need to be a number. It can be a text label (any text that
you like). This is used by nodes near coords, for example. It could also contain a tuple like RGB color
information (compare mesh/color input=explicit). Thus, point meta is really some abstract information
about individual coordinates.
Note that there is only one point meta per point. See the key visualization depends on if you need
more than one meta data value per coordinate.
The common idea is to tell pgfplots how to get the meta data. It is not necessary to provide data
explicitly – in many cases, the data which is used to color surface patches or marker colors is the plot’s y or
z coordinate. The method used to tell pgfplots where to find “point meta data” is the point meta key.
The most common use-case of point meta is color information: if the point meta data is in the interval
[mmin , mmax ], the point meta coordinate for the smallest encountered value m = mmin will get the lowest color
provided by the colormap while the largest encountered value m = mmax will get the highest color provided
by the colormap. As already mentioned, this is accomplished using \pgfplotspointmetatransformed
2 [0, 1000] (per convention). Any coordinate between the smallest and largest values will be mapped linearly:
for example, the mean m = 1/2(mmax + mmin ) will get the middle color of the color map (it will have
\pgfplotspointmetatransformed = 500). This is why “point meta” is sometimes called “color data” in
this manual.
20
20
10 10
0
0
6 4 2 0 2 4 6
\begin{tikzpicture}
\addplot[mesh,point meta=y,thick] {x^2};
\end{axis}
\end{tikzpicture}
4.8.2 User Input Format for Point Meta

/pgfplots/point meta=none|hexpression i|x|y|z|f(x)|explicit|explicit symbolic (initially none)
The point meta key tells pgfplots where to get the special point meta data. Please note that point
meta and scatter src is actually the same – scatter src is an alias for point meta. Thus, the
summary provided for scatter src on page 109 covers the same topics. However, the main reference
for point meta is here.
none The initial choice none disables point meta data, resulting in no computational work. Any other
choice will activate the computation of upper and lower ranges for point meta data, i.e. the com-
putation of [mmin , mmax ].
x The choice x uses the already available x coordinates as point meta data. This does always refer to
the final x coordinates after any user transformations, logarithms, stacked plot computations etc.
have been applied. Consider using rawx if you need the unprocessed coordinate value here.
y
z The choices y and z are similar: they use the y or z coordinates respectively as point meta data.
Consequently, these three choices do not need any extra data. As for x, there are math constants
rawy and rawz which yield the unprocessed y and z value, respectively.
f(x) This will use the last available coordinate, in other words: it is the same as y for two dimensional
plots and z for three dimensional ones.
explicit This choice tells pgfplots to expect numerical point meta data which is provided explicitly
in the coordinate input streams. This data will be transformed linearly into the current color map
as it has been motivated above.
How point meta data is provided for plot coordinates, plot table and the other input methods
is described in all detail in Section 4.3.1 – but we provide small examples here to summarize the
possibilities:
% for ’coordinates’:
% provide color data explicitly using [<data>]
% behind coordinates:
\addplot+[point meta=explicit]
coordinates {
(0,0) [1.0e10]
(1,2) [1.1e10]
(2,3) [1.2e10]
(3,4) [1.3e10]
% ...
};
% for ’table’:
% xcolname ycolname colordata
% 0 0 0.001
% 1 2 0.3
% 2 2.1 0.4
% 3 3 0.5
% ...
table[x=xcolname,y=ycolname,meta=colordata]
{datafile.dat};
% or, equivalently (perhaps a little bit slower):
\addplot+[point meta=\thisrow{colordata}]
{datafile.dat};
% for ’file’:
% 0 0 0.001
% 1 2 0.3
% 2 2.1 0.4
% 3 3 0.5
% ...
% the first three columns will be used here as x,y and meta,
% resp.
file {datafile.dat};
% ’table’ using expressions which may depend on all

% columns:
% xcolname ycolname anything othercol
% 0 0 4 15
% 1 2 5 20
% 2 2.1 8 30
% 3 3 42 40
% ...
\addplot+[point meta={0.5*(\thisrow{anything} + sqrt(\thisrow{othercol}))}]
{datafile.dat};
Thus, there are several methods to provide point meta (color data). The key for the choice
explicit is that some data is provided explicitly – although point meta does not know how. The
data is expected to be of numerical type and is mapped linearly into the range [0, 1000] (maybe
for use in the current color map).
explicit symbolic The choice explicit symbolic is very similar to explicit in that it expects
extra data by the coordinate input routines. However, explicit symbolic does not necessarily
expect numerical data: you can provide any sort of symbols. One might provide a set of styles,
one for each class in a scatter plot. This is implemented using scatter/classes, see page 112.
Input data is provided in the same fashion as mentioned above for the choice explicit.
This choice is usedful for nodes near coords with textual labels (see page 114) and for surface
plots with explicit color (see mesh/color input=explicit for details).
hexpressioni This choice allows to compute point meta data using a mathematical expression. The
hexpressioni may depend on x, y, z which yield the current x, y or z coordinate, respectively. The
coordinates are completely processed (transformations, logs) as mentioned above for the choice
x. Furthermore, the hexpressioni may depend on commands which are valid during \addplot like
\plotnum or \coordindex (see Section 4.25 for details). If coordinates are provided using \addplot
table, the macro \thisrow{hcolnamei} is particularly useful as it allows to access the value of the
provided hcolnamei of the “current” row. Computations are performed using the floating point
unit of pgf, and all supported arithmetical operations can be used.
In essence, the hexpressioni may depend on everything which is known to all \addplot commands:
the x, y and (if any) z coordinates. In addition, it may depend upon rawx, rawy or rawz. These
three expressions yield the unprocessed x, y or z value as it has been found in the input stream
(no logs, no user transformations)39 . If used together with plot table, you may also access other
table columns (for example with \thisrow{hcolnamei}).
TeX code=hcodei A rather low level choice which allows to provide TEX hcodei to compute a numerical
value. The hcodei should define the macro \pgfplotspointmeta. It is evaluated in a locally
scoped environment (it’s local variables are freed afterwards). It may depend on the same values
as described for hexpressioni above, especially on \thisrow{hcolnamei} for table input.
Note that the math parser will be configured to use the fpu at this time, so \pgfmathparse yields
floats.
Note that you need an extra pair of braces to provide this key, i.e.
point meta={TeX code={hcodei}}.
TeX code symbolic=hcodei Just as TeX code, you can provide hcodei which defines the macro
\pgfplotspointmeta, but the result is not interpreted as a number. It is like the explicit
symbolic choice.
point meta={TeX code symbolic={hcodei}}.
symbolic=hsymbol i A choice which accepts some arbitrary hsymbol i which is used for every coordinate.
As explicit symbolic and TeX code symbolic, this choice yields symbolic representations, i.e.
it is kept as–is without mapping the result.
The di↵erence to explicit symbolic is that hsymbol i is a common symbol for all points whereas
explicit symbolic expects the input data stream (like a table) to provide individual symbols.
The di↵erent to TeX code symbolic is marginal: symbolic is actually the same as
point meta/TeX code symbolic={\def\pgfplotspointmeta{hsymbol i}}.
A use-case for symbolic is mesh/color input=explicit mathparse, see the documentation there-
in.
point meta={symbolic={hsymbol i}}.
As already mentioned, a main application of point meta data is to determine (marker/face/edge) colors
using a linear map into the range [0, 1000] (maybe for use in the current color map). This map works
as follows: it is a function
: [mmin , mmax ] ! [0, 1000]
with
mminm
(m) =
1000
such that (mmin ) = 0 and (mmax ) = 1000. The value 1000 is – per convention – the upper limit of all
color maps. Now, if a coordinate (or edge/face) has the point meta data m, its color will be determined
using (m): it is the color at (m)‰ of the current color map.
This transformation depends on the interval [mmin , mmax ] which, in turn, can be modified using the
keys point meta rel, point meta min and point meta max described below.
The untransformed point meta data is available in the macro \pgfplotspointmeta (only in the correct
context, for example the scatter plot styles or the scatter/@pre marker code interface). This macro
contains a low level floating point number (unless it is non-parsed string data). The transformed data
will be available in the macro \pgfplotspointmetatransformed and is in fixed point representation.
It is expected to be in the range [0, 1000].
/pgfplots/set point meta if empty={hpoint meta sourcei}

Sets point meta=hpoint meta sourcei, but only if point meta=none currently. This is used for scatter,
mesh and surf with set point meta if empty=f(x).
39 In rare circumstances, it might be interesting to apply a math expression to another source of point meta (one of the other
choices. To this end, the hexpressioni is checked after the other possible choices have already been evaluated. In other words,
the statement point meta=explicit, point meta=meta*meta+3 will evaluate the expression with meta set to whatever data has
been provided explicitly.
4.8.3 Mapping Point Meta and Color Maps

/pgfplots/point meta rel=axis wide|per plot (initially axis wide)
As already explained in the documentation for point meta, one application for point meta data is to
determine colors using the current color map and a linear map from point meta data into the current
color map. The question is how this linear map is computed.
The key point meta rel configures whether the interval of all point meta coordinates, [mmin , mmax ]
is computed as maximum over all plots in the complete axis (the choice axis wide) or only for one
particular plot (the choice per plot).
Axis wide color mapping Per Plot color mapping
2 2
2 2
0 0 0 0
2 2
2 2
6 4 2 0 2 4 6 6 4 2 0 2 4 6
\begin{tikzpicture}
\begin{axis}[
title=Axis wide color mapping,
colorbar,
samples=50,point meta rel=axis wide,
point meta=y]
\addplot[mesh,thick] {sin(deg(x))};
\addplot[mesh,thick] {3*tanh(x)};
\end{axis}
\end{tikzpicture}
~
\begin{tikzpicture}
\begin{axis}[
title=Per Plot color mapping,
colorbar,
samples=50,
point meta rel=per plot,
point meta=y]
\addplot[mesh,thick] {sin(deg(x))};
\addplot[mesh,thick] {3*tanh(x)};
\end{axis}
\end{tikzpicture}
Note that a colorbar will still use the axis wide point meta limits. Consider the colorbar source
key if you want the color data limits of a particular plot for your color bar. The point meta rel key
configures how point meta maps to colors in the colormap.
/pgfplots/point meta min={hnumber i}

/pgfplots/point meta max={hnumber i}
These keys allow to define the range required for the linear map of point meta data into the range
[0, 1000] (for example, for current maps) explicitly. This is necessary if the same mapping shall be used
for more than one axis.
Remarks about special cases:

• It is possible to provide limits partially; in this case, only the missing limit will be computed.
• If point meta data falls outside of these limits, the linear transformation is still well defined which
is acceptable (unless the interval is of zero length). However, color data can’t be outside of these
limits, so color bars perform a truncation.
• This key can be provided for single plots as well as for the complete axis (or for both).
• If meta limits are provided for a single plot, these limits may also contribute to the axis wide meta
interval.
/pgfplots/colormap access=map|direct|piecewise constant|piecewise linear (initially map)

This key configures how point meta data is used to determine colors from a color map. The initial
configuration map performs the linear mapping operation explained above.
Using Colormaps as Indexed Color Space (Palette)
The choice direct does not perform any transformation; it interpretes the value of point meta as
integer indices into the current color map. If a number is no integer, it will be truncated by throwing
away any fraction parts of the number.
Suppose we have the color definitions RdPu-K, RdPu-I, RdPu-G, RdPu-E, and RdPu-B
(which are actually taken from \usetikzlibrary{colorbrewer}).
We will use these colors in order to define a colormap:
0 1 2 NaN % requires \usetikzlibrary{colorbrewer}

\begin{axis}[
colormap={brewer}{
color=(RdPu-B) color=(RdPu-E)
3 4 5 99 color=(RdPu-G) color=(RdPu-I)
1
color=(RdPu-K)},
2 enlargelimits=false,
colorbar as palette,
3.5 2.99 3 ]
1 nodes near coords,mark=*,
0 point meta=explicit,
0 1 2 3 colormap access=direct,
]
coordinates {
(0,0) [0] (1,0) [1] (2,0) [2] (3,0) [NaN]
(0,1) [3] (1,1) [4] (2,1) [5] (3,1) [99]
(0,2) [3.5] (1,2) [-1] (2,2) [2.99] (3,2) [3]

};
\end{axis}
\end{tikzpicture}
The example has a 3 ⇥ 4 matrix, each with explicitly provided values of point meta. The first point has
a point meta value of 0 and uses the first color of the colormap (RdPu-B, ). The second point has
value 1 and uses the second color, RdPu-E , and so on. Note that NaN is interpreted as ‘hole’ in the
matrix and is left empty (you see the background color here). Note furthermore that coordinate (1,2)
[-1] is mapped to the first color since values outside of the allowed range are automatically clipped
to the next correct color. The same holds for the point with [99]: it is mapped to the last available
color . Finally, the point with [2.99] is mapped to index 2 ( ) since colormap access=direct
truncates to the next integer index.
/pgfplots/colorbar as palette (no value)

The example above makes use of colorbar as palette, a dedicated style useful for such palette-
type applications.
It is defined as
\pgfplotsset{
colorbar as palette/.style={
colorbar sampled={
surf,
colormap access=direct,
shader=flat corner,
samples=\pgfplotscolormapsizeof
{\pgfkeysvalueof{/pgfplots/colormap name}}+1,
},
colorbar style={
point meta min=0,
point meta max=\pgfplotscolormapsizeof
{\pgfkeysvalueof{/pgfplots/colormap name}},
}}}
The idea is to show only the colors of the “palette”, i.e. without interpolating values of the colormap.
Furthermore, it is supposed to show all available colors.
Note that colormap access=direct changes the way how point meta is translated to colors. However,
it does not a↵ect shader. In other words, it is perfectly sane to use shader=interp or shader=flat
mean. This, however, will interpolate the colors resulting from colormap access=direct! Our example
above uses shader=flat corner which results in no color interpolation at all.
The choice colormap access=direct is actually like a color palette for images with indexed colorspaces:
suppose we use the famous “Lena” image [4], downsample it to 128 ⇥ 128, and export it to coordinates
of the form
1 1 194
1 2 194
1 3 170
1 4 193
1 5 188
1 6 188
...
with an indexed colorspace. Furthermore, we import its color palette as pgfplots colormap. To this
end, a small Matlab script serves as utility to convert a tiff image to coordinates:40
[Z,cm] = imread(’4.2.04_128.tiff’);
[X,Y]=meshgrid(1:size(Z,1), 1:size(Z,2));
data = [ X(:) Y(:) Z(:) ];
fid = fopen(’lena.dat’, ’w’);
fprintf(fid, "%d %d %d\n", data’);
fclose(fid);
fid = fopen(’lena_cm.tex’, ’w’);

fprintf(fid, "\\pgfplotsset{colormap={lena palette}{\n");
fprintf(fid, "rgb=(%f,%f,%f)\n", cm’);
fprintf(fid, "}}\n");
fclose(fid);
This allows us to import the graphics into pgfplots as matrix plot:
\begin{tikzpicture}
% load and activate the CM:
\input lena_cm.tex
\begin{axis}[view={0}{90},axis equal]
50 \addplot3[matrix plot,
colormap access=direct,
mesh/rows=128,
mesh/ordering=colwise,
]
100 table {lena.dat};
\end{axis}
\end{tikzpicture}
0 50 100
40 Also see Section 7.2.1, “Importing Mesh Data From Matlab To PGFPlots” on page 515
The files lena_cm.tex and lena.dat are shipped with pgfplots and can be downloaded from CTAN.
The keys to decode the table data into a mesh are specific to the Matlab export format of our script.
Note that rendering images from coordinates is considerably less efficient than \includegraphics in
terms of
• computational time required by pgfplots,
• space consumption in the resulting pdf document,
• and time to render the resulting pdf in the viewer(s).
Depending on the use-case, \addplot graphics may help to reduce typesetting times.
Some more details:

• If there are m colors in the color map and the color data falls outside of [0, m 1], it will be pruned
to either the first or the last color.
• If color data is a real number, it will be truncated to the next smaller integer.
• If colormap access=direct is combined with a shader6=flat corner, color interpolation takes
place. However, the color interpolation is not carried out in terms of the scalar point meta (as
would have been done for colormap access=map). Instead, it is carried out on the resulting colors
in their own colorspace.
• Note that patch refines or patch to triangles is unsupported in this context (at the time of
this writing, it will interpolate the color indices, resulting in badly defined behavior).
See also the related keys color of colormap and index of colormap.
The choice colormap access=piecewise linear is an alias for colormap access=map.
Mapping Point Meta using Piecewise Constant Interpolation
The choice colormap access=piecewise constant is very similar to colormap access=map: it accepts
the same point meta values as input, maps them linearly into [0, 1000] and uses the first color of the
colormap if the value of point meta is point meta min and the last color of the colormap if the value
of point meta is point meta max. However, values of point meta which fall between two colormap
entries are interpolated using piecewise constant interpolation, i.e. just like const plot: the interpolated
color resembles the value of the first color of the interval. As a consequence, the result contains at most
as many colors as the colormap provides:
\begin{tikzpicture}
\begin{axis}[
20 colormap/RdPu-6,
%
0 ]
\addplot3[
20 5 surf,
] {x*y};
4 0 \end{axis}
2 4 5
20 10 0 10 20
The choice colormap access=piecewise constant is compatible with all available shaders, but its
best results are possible using shader=interp as this will avoid rectangular artifacts at mesh boundaries.
4.9. AXIS DESCRIPTIONS 219
\begin{tikzpicture}
\begin{axis}[
20 colormap/RdPu-6,
%
0 ]
\addplot3[
20 5 surf,
shader=interp,
4 0 ] {x*y};
2 0 \end{axis}
20 10 0 10 20
The choice colormap access=piecewise constant requires colormaps with a suitable number of col-
ors. This is quite di↵erent to colormap access=map (or piecewise linear) where the actual number
of colors in the colormap cannot be seen in the resulting plot. In order to get the desired e↵ects, one
may need the same colormap with a di↵erent number of samples. To this end, one can make use of the
special colormap syntax which is based on samples of colormap and its variants:
\begin{tikzpicture}
\begin{axis}[
colormap={my sampled version of viridis}{
20 samples of colormap=(10 of viridis)},
%
0 ]
\addplot3[
20 5 surf,
shader=interp,
4 0 ] {x*y};
2 0 \end{axis}
20 10 0 10 20
The example takes the existing colormap name=viridis and chooses 10 samples of it. The result is
called my sampled version of viridis and is used as “current colormap”. Details about this process
can be read in Section 4.7.6 on page 192.
Note that colormap access=piecewise constant combined with shader=interp actually is a filled
contour plot – except that the contours are implicitly encoded by the samples of the colormap. Please
refer to Section 4.6.9 for details about this approach.
Technical side-note: colormap access=piecewise constant internally ensures that each color in the
colormap receives its own interval (i.e. is actually visible in the output). To this end, it adds an artifical
end-point at the right end of the colormap. This operation is done without any user action and it is
specific to colormap access=piecewise constant.
4.9 Axis Descriptions

Axis descriptions are labels for x and y axis, titles, legends and the like. Axis descriptions are drawn after
the plot is finished and they are not subjected to clipping.
4.9.1 Placement of Axis Descriptions

This section describes how to modify the placement of titles, labels, legends and other axis descriptions. It
may be skipped at first reading.
There are di↵erent methods to place axis descriptions. One of them is to provide coordinates relative
to the axis’ rectangle such that (0,0) is the lower left corner and (1,1) is the upper right corner – this is
very useful for figure titles or legends. Coordinates of this type, i.e. without unit like (0,0) or (1.03,1),
are called axis description cs (the cs stands for “coordinate system”). One other method is of primary
interest for axis labels – they should be placed near the tick labels, but it a way that they don’t overlap or
obscure tick labels. Furthermore, axis labels shall be placed such that they are automatically moved if the
axis is rotated (or tick labels are moved to the right side of the figure). There is a special coordinate system
to realize these two demands, the ticklabel cs.
In the following, the two coordinate systems axis description cs and ticklabel cs are described
in more detail. It should be noted that axis description cs is used automatically, so it might never be
necessary to use it explicitly.
Coordinate system axis description cs

A coordinate system which is used to place axis descriptions. Whenever the option ‘at={(hx i,hyi)}’
occurs in label style, legend style or any other axis description, (hx i,hyi) is interpreted to be a
coordinate in axis description cs.
The point (0, 0) is always the lower left corner of the tightest bounding box around the axes (without
any descriptions or ticks) while the point (1, 1) is the upper right corner of this bounding box.
In most cases, it is not necessary to explicitly write axis description cs as it is the default coordinate
system for any axis description. An example for how coordinates are placed is shown below.
How axis description cs works
5 (1, 1)
(0.5, 0.5)
0
(1.03, 0.5)
(0, 0)
5
6 4 2 0 2 4 6
% [See the TikZ manual if you’d like to learn about nodes and pins]
\begin{tikzpicture}
\tikzset{
every pin/.style={fill=yellow!50!white,rectangle,rounded corners=3pt,font=\tiny},
small dot/.style={fill=black,circle,scale=0.3}
}
\begin{axis}[
clip=false,
title=How \texttt{axis description cs} works
]
\addplot {x};
\node[small dot,pin=120:{$(0,0)$}] at (axis description cs:0,0) {};

\node[small dot,pin=-30:{$(1,1)$}] at (axis description cs:1,1) {};
\node[small dot,pin=-90:{$(1.03,0.5)$}] at (axis description cs:1.03,0.5) {};
\node[small dot,pin=125:{$(0.5,0.5)$}] at (axis description cs:0.5,0.5) {};
\end{axis}
\end{tikzpicture}
Axis descriptions are Tik Z nodes, that means all placement and detail options of [6] apply. The point
on the node’s boundary which is actually shifted to the at coordinate needs to be provided with an
anchor (cf [6, Nodes and Edges]):
20
x
10
x2
6 4 2 0 2 4 6
\begin{tikzpicture}
\begin{axis}[
legend entries={$x$,$x^2$},
legend style={
at={(1.03,0.5)},
anchor=west
}
]
\addplot {x};
\addplot {x^2};
\end{axis}
\end{tikzpicture}
Standard anchors of nodes are north, east, south, west and mixed components like north east.
Please refer to [6] for a complete documentation of anchors.
Remarks:
• Each of the anchors described in Section 4.19 can be described by axis description cs as well.
• The axis description cs is independent of axis reversals or skewed axes. Only for the default
configuration of boxed axes is it the same as rel axis cs, i.e. (0,0) is the same as the smallest
axis coordinate and (1,1) is the largest one in case of standard boxed axes41 .
• Even for three dimensional axes, the axis description cs is still two-dimensional: it always refers
to coordinates relative to the tightest bounding box around the axis (without any descriptions or
ticks).
How axis description cs works in 3D
(1, 1)
5 (0.5, 0.5)
0
(1.03, 0.5)
5
(0, 0) 5
4 0
2 0 2 4 5
41 This was di↵erent in versions before 1.3: earlier versions did not have the distinction between axis description cs and
rel axis cs.

% the same as above for 3D ...
% [See the TikZ manual if you’d like to learn about nodes and pins]
\begin{tikzpicture}
\tikzset{
}
\begin{axis}[
clip=false,
title=How \texttt{axis description cs} works in 3D
]
\addplot3 coordinates {(-5,-5,-5) (5,5,5)};
\draw[black!15] (axis description cs:0,0) rectangle (axis description cs:1,1);
\node[small dot,pin=120:{$(0,0)$}] at (axis description cs:0,0) {};

\node[small dot,pin=-30:{$(1,1)$}] at (axis description cs:1,1) {};
\node[small dot,pin=-90:{$(1.03,0.5)$}] at (axis description cs:1.03,0.5) {};
\node[small dot,pin=125:{$(0.5,0.5)$}] at (axis description cs:0.5,0.5) {};
\end{axis}
\end{tikzpicture}
• Since the view does not influence these positions, axis description cs might not be a good
choice for axis labels in 3D. The ticklabel cs is used in this case.
Coordinate system xticklabel cs

Coordinate system yticklabel cs
Coordinate system zticklabel cs
Coordinate system ticklabel cs
Coordinate system xticklabel* cs
Coordinate system yticklabel* cs
Coordinate system zticklabel* cs
Coordinate system ticklabel* cs
A set of special coordinate systems intended to place axis descriptions (or any other drawing operation)
besides tick labels, in a way such that neither tick labels nor the axis as such are obscured.
See also xlabel near ticks as one main application of ticklabel cs.
The xticklabel cs (and its variants) always refer to one, uniquely identified axis: the one which is (or
would be) annotated with tick labels.
The ticklabel cs (without explicit x, y or z) can only be used in contexts where the axis character is
known from context (for example, inside of xlabel style – there, the ticklabel cs is equivalent to
xticklabel cs).
The starred variants xticklabel* cs and its friends do not take the size of any tick labels into account.
Each of these coordinate systems allows to specify points on a straight line which is placed parallel to
an axis containing tick labels, moved away just far enough to avoid overlaps with the tick labels:
Positioning with xticklabel cs

yticklabel cs:1
yticklabel cs:0.5 0
5
yticklabel cs:0
6 4 2 0 2 4 6
xticklabel cs:0 xticklabel cs:0.5 xticklabel cs:1

\tikzset{
}
\begin{tikzpicture}
\begin{axis}[
clip=false,
ticklabel style={draw=red},
title=Positioning with \texttt{xticklabel cs}]
\addplot {x};
\node[small dot,pin=-90:{\texttt{xticklabel cs:0}}] at (xticklabel cs:0) {};
\node[small dot,pin=-90:{\texttt{xticklabel cs:0.5}}] at (xticklabel cs:0.5) {};
\node[small dot,pin=180:{\texttt{yticklabel cs:0}}] at (yticklabel cs:0) {};

\node[small dot,pin=180:{\texttt{yticklabel cs:0.5}}] at (yticklabel cs:0.5) {};
\node[small dot,pin=180:{\texttt{yticklabel cs:1}}] at (yticklabel cs:1) {};
\end{axis}
\end{tikzpicture}
The basic idea is to place coordinates on a straight line which is parallel to the axis containing tick
labels – but shifted such that the line does not cut through tick labels.
Note that an axis description which has been placed with xticklabel cs or its friends is also useful for
skewed axes or the axis x line variants – it is often the same value for all these variants. In particular,
it is useful for three–dimensional axes, see below.
Typically, xticklabel cs places nodes exactly at the position where the largest associated tick label
is finished. While this is very useful, it might be undesired – for example if one wants to move into the
opposite direction (here, the special anchor near ticklabel opposite might be of interest). To this
end, there are the starred variants, i.e. xticklabel* cs and its friends:
Starred variant xticklabel* cs

yticklabel* cs:1
yticklabel* cs:0.5 0
5
yticklabel* cs:0
6 4 2 0 2 4 6
xticklabel* cs:0 xticklabel* cs:0.5 xticklabel* cs:1
\tikzset{
}
\begin{tikzpicture}
\begin{axis}[
clip=false,
title=Starred variant \texttt{xticklabel* cs}]
\addplot {x};
\node[small dot,pin=-90:{\texttt{xticklabel* cs:0}}] at (xticklabel* cs:0) {};
\node[small dot,pin=-90:{\texttt{xticklabel* cs:0.5}}] at (xticklabel* cs:0.5) {};
\node[small dot,pin=-90:{\texttt{xticklabel* cs:1}}] at (xticklabel* cs:1) {};
\node[small dot,pin=180:{\texttt{yticklabel* cs:0}}] at (yticklabel* cs:0) {};

\node[small dot,pin=180:{\texttt{yticklabel* cs:0.5}}] at (yticklabel* cs:0.5) {};
\node[small dot,pin=180:{\texttt{yticklabel* cs:1}}] at (yticklabel* cs:1) {};
\end{axis}
\end{tikzpicture}
The preceding example places all the additional anchors precisely onto the axis on which tick labels are
drawn. The starred version xticklabel* cs ignores the size of tick labels.
Of course, it is relatively simple to get the same coordinates as in the two dimensional example above with
axis description cs, except that ticklabel cs always respects the tick label sizes appropriately.
However, ticklabel cs becomes far superior when it comes to three dimensional positioning:
Positioning with ticklabel cs in 3D
zticklabel cs:1
5
zticklabel cs:0.5 0
5
5
zticklabel cs:0
4 0
2 0 yticklabel cs:1
2 4 5 yticklabel cs:0.5
xticklabel cs:0
xticklabel cs:0.5
yticklabel cs:0
xticklabel cs:1
% the same as above for 3D ...
\begin{tikzpicture}
\tikzset{
}
\begin{axis}[
clip=false,
title=Positioning with \texttt{ticklabel cs} in 3D
]

\node[small dot,pin=-90:{\texttt{xticklabel cs:0.5}}] at (xticklabel cs:0.5) {};
\node[small dot,pin=-45:{\texttt{yticklabel cs:0}}] at (yticklabel cs:0) {};

\node[small dot,pin=-45:{\texttt{yticklabel cs:0.5}}] at (yticklabel cs:0.5) {};
\node[small dot,pin=-45:{\texttt{yticklabel cs:1}}] at (yticklabel cs:1) {};
\node[small dot,pin=180:{\texttt{zticklabel cs:0}}] at (zticklabel cs:0) {};

\node[small dot,pin=180:{\texttt{zticklabel cs:0.5}}] at (zticklabel cs:0.5) {};
\node[small dot,pin=180:{\texttt{zticklabel cs:1}}] at (zticklabel cs:1) {};
\end{axis}
\end{tikzpicture}
The coordinate ticklabel cs:0 is associated with the lower axis limit while ticklabel cs:1 is near
the upper axis limit. The value 0.5 is in the middle of the axis, any other values (including negative
values or values beyond 1) are linearly interpolated inbetween.
All coordinate systems like ticklabel cs also accepts a second (optional) argument: a shift “away”
from the tick labels. The shift points to a vector which is orthogonal42 the associated axis, away from
the tick labels. A shift of 0pt is directly at the edge of the tick labels in direction of the normal vector,
positive values move the position away and negative closer to the tick labels.
ticklabel cs and its optional shift
5
5
4 0
2 0 2 4 5
xticklabel cs:1,0
xticklabel cs:1,15pt
42 Actually, the outer normal has the impression of being “orthogonal” to its axis, which appears to be sufficient.
\tikzset{
}
\begin{tikzpicture}
\begin{axis}[
xticklabel style={draw=red},
clip=false,
title=\texttt{ticklabel cs} and its optional shift
]
\draw[blue,thick,->] (xticklabel cs:0,0) -- (xticklabel cs:1,0);

\draw[red,thick,->] (xticklabel cs:0,5pt) -- (xticklabel cs:1,5pt);
\draw[magenta,thick,->] (xticklabel cs:0,10pt) -- (xticklabel cs:1,10pt);
\draw[green,thick,->] (xticklabel cs:0,15pt) -- (xticklabel cs:1,15pt);
\node[small dot,pin=0:{\texttt{xticklabel cs:1,0}}] at (xticklabel cs:1,0) {};
\node[small dot,pin=0:{\texttt{xticklabel cs:1,15pt}}] at (xticklabel cs:1,15pt) {};
\draw[blue,thick,->] (xticklabel cs:0,0) -- (xticklabel cs:0,15pt);

\draw[blue,thick,->] (xticklabel cs:1,0) -- (xticklabel cs:1,15pt);
\end{axis}
\end{tikzpicture}
Whenever the ticklabel cs is used, the anchor should be set to anchor=near ticklabel (see below).
Whenever the starred version ticklabel* cs is used, both anchors anchor=near ticklabel and
anchor=near ticklabel opposite are useful choices.
There is one specialty: if you reverse an axis (with x dir=reverse), points provided by ticklabel cs
will be una↵ected by the axis reversal. This is intented to provide consistent placement even for reversed
axes. Use allow reversal of rel axis cs=false to disable this feature.
The purpose of ticklabel cs is to place nodes “next to tick labels”. The position of tick labels as
such is determined in a similar way to ticklabel* cs with a customized shift along the outer normal
vector. The shift is typically
\pgfkeysvalueof{/pgfplots/major tick length} (or half of it for centered axes).
Besides the mentioned positioning methods, there is also the predefined node current axis. The anchors
of current axis can also be used to place descriptions: At the time when axis descriptions are drawn, all
anchors which refer to the axis origin (that means the “real” point (0, 0)) or any of the axis corners can be
referenced using current axis.hanchor namei. Please see Section 4.19, Alignment, for further details.
4.9.2 Alignment of Axis Descriptions

This section describes how to modify the default alignment of axis descriptions. It can be skipped at first
reading.
The two topics positioning and alignment always work together: positioning means to select an appro-
priate coordinate and alignment means to select an anchor inside of the description which will actually be
moved to the desired position.
Tik Z uses many anchors to provide alignment; most of them are named like north, north east etc.
These names hold for any axis description as well (as axis descriptions are Tik Z nodes). Readers can learn
details about this topic in the Tik Z manual [6] or some more advice in Section 4.19.
When it comes to axis descriptions, pgfplots o↵ers some specialized anchors and alignment methods
which are described below.
Anchor near xticklabel
Anchor near yticklabel
Anchor near zticklabel
Anchor near ticklabel
These anchors can be used to align at the part of a node (for example, an axis description) which is
nearest to the tick labels of a particular axis (or nearest to the position where tick labels would have
been drawn if there were any).
These anchors are used for axis labels, especially for three dimensional axes. Furthermore, they are
used for every tick label.
Maybe it is best to demonstrate it by example:
Without near ticklabel With near ticklabel
5 5
f (x) = x
f (x) = x
0 0
5 5
6 4 2 0 2 4 6 6 4 2 0 2 4 6
\begin{tikzpicture}
\begin{axis}[
title=Without \texttt{near ticklabel},
ylabel={$f(x)=x$},
every axis y label/.style=
{at={(ticklabel cs:0.5)},rotate=90,anchor=center},
clip=false,% to display the \path below
ylabel style={draw=red},
yticklabel style={draw=red}
]
\addplot {x};
% visualize the position:

\fill (yticklabel cs:0.5) circle(2pt);
\end{axis}
\end{tikzpicture}%
~
\begin{tikzpicture}
\begin{axis}[
title=With \texttt{near ticklabel},
ylabel={$f(x)=x$},
every axis y label/.style=
{at={(ticklabel cs:0.5)},rotate=90,anchor=near ticklabel},
clip=false,
ylabel style={draw=red},
yticklabel style={draw=red}
]
\addplot {x};
\fill (yticklabel cs:0.5) circle(2pt);
\end{axis}
\end{tikzpicture}
The motivation is to place nodes such that they are anchored next to the tick label, regardless of the
node’s rotation or the position of ticks. The special anchor near ticklabel is only available for axis
labels (as they have a uniquely identified axis, either x, y or z).
In more detail, the anchor is placed such that first, the node’s center is on a line starting in the node’s
at position going in direction of the inwards normal vector of the axis line which contains the tick labels
and second, the node does not intrude the axis (but see also the key near ticklabel align and the
details in the lengthy elaboration in the documentation for near xticklabel opposite below). This
normal vector is the same which is used for the shift argument in ticklabel cs: it is orthogonal to
the tick label axis. Furthermore, near ticklabel inverts the transformation matrix before it computes
this intersection point.
The near ticklabel anchor and its friends will be added temporarily to any shape used inside of
an axis. This includes axis descriptions, but it is not limited to them: it applies to every Tik Z
\node[anchor=near xticklabel] ... setting.
Note that it is not necessary at all to have tick labels in an axis. The anchor will be placed such that
it is near the axis on which tick labels would be drawn. In fact, every tick label uses anchor=near
ticklabel as initial configuration.
Anchor near xticklabel opposite
Anchor near yticklabel opposite
Anchor near zticklabel opposite
Anchor near ticklabel opposite
These anchors are similar to near xticklabel and its variants, except that they align at the opposite
direction.
Mathematically speaking, the only di↵erence to near xticklabel and its variants is the sign in front
of the normal vector.
But it is probably best explained by means of an example.
near ticklabel (opposite)
0.5
1
near xticklabel opposite.
0
0 0.2 0.5
0.4 0.6 0.8 1 0 near yticklabel.
near xticklabel.
\begin{tikzpicture}
\begin{axis}[
clip=false,
small,
title=\texttt{near ticklabel} (\texttt{opposite}),
min=0, max=1
]
\node[draw=yellow,anchor=near xticklabel,font=\small]
(namex) at (xticklabel cs:0.2) {\texttt{near xticklabel}.};
\fill (xticklabel cs:0.2) circle(2pt);
\draw[green,-stealth] (xticklabel* cs:0.2) -- (xticklabel cs:0.2);
\draw[blue,-stealth] (xticklabel cs:0.2) -- (namex.center);
\draw[red,-stealth] (namex.north east) -- (xticklabel cs:0.2);
\node[draw=yellow,anchor=near xticklabel opposite,font=\small]

(namexx) at (xticklabel* cs:0.6) {\texttt{near xticklabel opposite}.};
\fill[red] (xticklabel* cs:0.6) circle(2pt);
\draw[blue,-stealth] (xticklabel* cs:0.6) -- (namexx.center);
\draw[red,thick,-stealth] (namexx.south west) -- (xticklabel* cs:0.6);
\node[draw=yellow,anchor=near yticklabel,font=\small]
(name) at (yticklabel cs:1) {\texttt{near yticklabel}.};
\fill (yticklabel cs:1) circle(2pt);
\draw[green,-stealth] (yticklabel* cs:1) -- (yticklabel cs:1);
\draw[blue,-stealth] (yticklabel cs:1) -- (name.center);
\draw[red,-stealth] (name.north west) -- (yticklabel cs:1);
\end{axis}
\end{tikzpicture}
The figure is a boxed three–dimensional axis with standard ranges. It has three manually placed nodes.
The nodes are placed at (xticklabel cs:0.2), at (xticklabel* cs:0.6), and at (yticklabel
cs:1), respectively. The ‘at’ locations are visually emphasized using filled circles.
Despite the di↵erent locations, we clearly see the e↵ect of near xticklabel opposite: it causes the
node to be aligned into the box rather than outside of the box. Note that this is the only essential
di↵erence between the two nodes ‘near xticklabel opposite’ and ‘near xticklabel’.
The figure also shows the di↵erence between xticklabel cs and xticklabel* cs when we compare
the ‘at’ locations. Take, for example, the two nodes on the x axis. The position at (xticklabel
cs:0.2) is shifted by the green arrow. The length of this arrow is precisely the length of the largest
x tick label. The position at (xticklabel* cs:0.6) is exactly on the axis; it ignores the size of any
tick labels. Note that the direction of the green arrow is the “outer normal vector in x direction” (in
our case, it is the vector sum of the z and y unit vectors with appropriate signs).
The di↵erence between near xticklabel opposite and near xticklabel is that the direction of the
green arrow (the “outer normal”) is flipped.
The nodes also highlight how the anchoring works. This technique is almost the same for both
anchor=near xticklabel and anchor=near xticklabel opposite. Let us discuss the technique for
the node with text ‘near yticklabel’. The black circle is placed at (yticklabel cs:1). This posi-
tion has been computed by starting at 100% of the y axis43 and moving along the green vector whose
magnitude is the size of the bounding box of the largest y tick label. As soon as the ‘at’ location
is fixed, the algorithm for near yticklabel starts. First, it computes the anchor inside of the node
for which we do not penetrate the y axis. To this end, it checks the direction of the green vector. It
came up with north west. Then, it considers the red line which starts at (name.north west) and has
the same direction as the y axis. Note that the red line and the y axis are parallel. It also considers
the blue line. This line points into the direction of the green line and is fixed by the current node’s
center. The precise intersection point of the red line and the blue line are the result of anchor=near
yticklabel. The same applies for the other two nodes as well: the red line is always parallel to the
axis under consideration and is anchored at the “snap–to–nearest–anchor” of the node. The blue line
is always parallel to the outer normal vector of the axis under consideration, and is anchored at the
current node’s center.
/pgfplots/near ticklabel align=inside|center|outside (initially center)

Allows to change the alignment algorithm of anchor=near ticklabel.
In the default configuration, anchor=near ticklabel and its variants take the node’s center in order
to derive the final anchor. This corresponds to the target of the blue line in the illustration for near
xticklabel opposite, see above.
This key changes the node’s anchor which is used here. The choice center can move half of the node’s
bounding box beyond the associated axis. The choice inside moves the complete node’s bounding box
in a way such that it is not beyond the associated axis. The choice outside moves the entire node’s
bounding box in a way that it is beyond the associated axis:
near ticklabel align
0.5
1
outside.
0
0 0.2 0.5
0.4 0.6 inside.
0.8 1 0
center.
43 In our example, percentages and absolute values are accidentally the same.
\begin{tikzpicture}
\begin{axis}[
clip=false,
small,
title=\texttt{near ticklabel align},
min=0, max=1,
/tikz/node style/.style={
draw=yellow,
anchor=near yticklabel,
font=\tiny,
},
]
\node[node style,
/pgfplots/near ticklabel align=center]
(C) at (yticklabel cs:0) {\texttt{center}.};
\node[node style,
/pgfplots/near ticklabel align=inside]
(I) at (yticklabel cs:1) {\texttt{inside}.};
\node[node style,
/pgfplots/near ticklabel align=outside]
(O) at (yticklabel cs:1) {\texttt{outside}.};

\fill[blue] (C.center) circle(1pt);

\fill[blue] (I.south west) circle(1pt);
\fill[blue] (O.north east) circle(1pt);

\draw[blue] (yticklabel cs:0) -- (C.center);
\draw[red,-stealth] (C.north west) -- (yticklabel cs:0);

\draw[blue] (yticklabel cs:1) -- (O.north east);
\draw[red,-stealth] (O.north west) -- (yticklabel cs:1);
\draw[red,-stealth] (I.north west) -- (yticklabel cs:1);
\end{axis}
\end{tikzpicture}
The example is similar to the one above: it generates an empty axis with a default range. Then, it
creates three nodes: one for choice center, one for choice inside, and one for choice outside. The
node for center is placed at (yticklabel cs:0) (lower black circle). We see that its bounding box
extends the size of the y axis. This is because its anchor C.center is used for the alignment. Both
the node for inside and the node for outside are placed at (yticklabel cs:1) (upper black circle).
Their only di↵erence is the choice for near ticklabel align. The node for inside does not extend
the size of the y axis; it is placed within its boundaries – because its internal anchor (I.north east)
(blue) has automatically been used in order to align the node. The node for outside is completely
outside of the extends for the y axis because its (blue) anchor (O.south west) has been chosen.
Note that the red lines are always the same. They are the “snap–to–nearest” anchor such that the node
is outside of the axis. Only the location of the blue anchors is a↵ected by this key.
Note that near ticklabel align always results in the same alignment, independent of the actual
position of the node. This is because an anchor is independent of the at location of a node. In this
context, the names “inside” and “outside” might be a bad choice: they stress the intended meaning if
the node is chosen at the upper end of the axis. However, if you say at (yticklabel cs:0), near
ticklabel align=inside, it will actually end up outside of the axis. This is because the “inside”
anchor has been computed without considering where the node is.
/pgfplots/near ticklabel at={hcoordinatei} (initially empty)
Occasionally, anchor=near ticklabel results in a di↵erent anchor depending on where the node is
placed. This can happen for a polaraxis.
If this key is needed, pgfplots will spit out the warning “The anchor ‘near ticklabel’ cannot be
computed correctly because the position is missing. Please add ‘near ticklabel at=coordinate’ with a
suitable coordinate”.
pgfplots uses context information to determine the correct information, so this may be unnecessary.
If you encounter this warning, you should add near ticklabel at=hvaluei with a suitable value.
/tikz/sloped like x axis (no value)

/tikz/sloped like y axis (no value)
/tikz/sloped like z axis (no value)
/tikz/sloped like x axis={hoptionsi}
/tikz/sloped like y axis={hoptionsi}
/tikz/sloped like z axis={hoptionsi}
A key which replaces the rotational / scaling parts of the transformation matrix such that the node is
sloped like the provided axis. For two dimensional plots, sloped like y axis is e↵ectively the same
as rotate=90. For a three dimensional axis, this will lead to a larger di↵erence:
\begin{tikzpicture}
\begin{axis}[
xlabel=Variable 1,
ylabel=Variable 2,
zlabel=value,
100 xlabel style={sloped like x axis},
ylabel style={sloped}
0
value
100 5 \addplot3[surf] {y*x*(1-x)};

\end{axis}
2 2
0 2 b le
4
Variable 5 ria
1 Va
Inside of axis labels, sloped is an alias for sloped like hchar i axis with the correct hchar i chosen
automatically.
Please note that rotated text might not look very good (neither on screen nor printed).
It is possible to customize sloped like x axis by means of the following keys, which need to be
provided as hoptionsi (simply ignore the lengthy gray key prefixes):
/pgfplots/sloped/allow upside down=true|false (initially false)
Use sloped like x axis=allow upside down to enable upside down labels.
/pgfplots/sloped/execute for upside down=code (initially empty)

Use sloped like x axis={execute for upside down=\tikzset{anchor=north}} or something
like that to handle upside down text nodes in a customized way (this is used by the smithchart
library).
/pgfplots/sloped/reset nontranslations=true|false (initially true)

Use sloped like x axis={reset nontranslations=false} to append the transformations to the
actual transformation matrix (instead of replacing it).
/pgfplots/sloped/at position={hpositioni} (initially empty)

This key can either be empty or a position. It is merely of interest for axes in which the transfor-
mation depends on the axis position as for polaraxis.
4.9.3 Labels
/pgfplots/xlabel={htexti}
/pgfplots/ylabel={htexti}
/pgfplots/zlabel={htexti}
These options set axis labels to htexti which is any TEX text.
To include special characters, you can use curly braces: “xlabel={, = characters}”. This is necessary
if characters like ‘=’ or ‘,’ need to be included literally.
Use xlabel/.add={hprefix i}{hsuffix i} to modify an already assigned label.
Labels are Tik Z-nodes which are placed with
% for x:
\node
[style=every axis label,
style=every axis x label]
% for y:
\node
[style=every axis label,
style=every axis y label]
so their position and appearance can be customized.

For example, a multiline xlabel can be configured using
\begin{axis}[xlabel style={align=right,text width=3cm},xlabel=A quite long label with a line break]

...
\end{axis}
See [6] to learn more about align and text width.
Upgrade notice: Since version 1.3, label placement can respect the size of adjacent tick labels. Use
\pgfplotsset{compat=1.3} (or newer) in the preamble to activate this feature. See xlabel near
ticks for details.
/pgfplots/xlabel shift={hdimensioni} (initially 0pt)

/pgfplots/ylabel shift={hdimensioni} (initially 0pt)
/pgfplots/zlabel shift={hdimensioni} (initially 0pt)
/pgfplots/label shift={hdimensioni}
Shifts labels in direction of the outer normal vector of the axis by an amount of hdimensioni. The
label shift sets all three label shifts to the same value.
Attention: This does only work if \pgfplotsset{compat=1.3} (or newer) has been called (more
precisely: if xlabel near ticks is active for the respective axis).
/pgfplots/xlabel near ticks (no value)

/pgfplots/ylabel near ticks (no value)
/pgfplots/zlabel near ticks (no value)
/pgfplots/compat=1.3
These keys place axis labels (like xlabel) near the tick labels. If tick labels are small, labels will
move closer to the axis. If tick labels are large, axis labels will move away from the axis. This is
the default for every three dimensional plot, but it won’t be used initially for two–dimensional plots
for backwards compatibility. Take a look at the definition of near ticklabel on page 226 for an
example.
The definition of these styles is
\pgfplotsset{
/pgfplots/xlabel near ticks/.style={
/pgfplots/every axis x label/.style={
at={(ticklabel cs:0.5)},anchor=near ticklabel
}
},
/pgfplots/ylabel near ticks/.style={
/pgfplots/every axis y label/.style={
at={(ticklabel cs:0.5)},rotate=90,anchor=near ticklabel
}
}
}
It is encouraged to write
\pgfplotsset{compat=1.3} % or newer
in your preamble to install the styles document-wide – it leads to the best output (it avoids un-
necessary space). It is not activated initially for backwards compatibility with older versions which
used fixed distances from the tick labels.
/pgfplots/xlabel absolute (no value)

/pgfplots/ylabel absolute (no value)
/pgfplots/zlabel absolute (no value)
/pgfplots/compat=pre 1.3
Installs placement styles for axis labels such that xlabel yields a description of absolute, fixed
distance to the axis. This is the initial configuration (for backwards compatibility with versions
before 1.3). Use compat=1.3 to get the most recent, more flexible configuration. Take a look at
the definition of near ticklabel on page 226 for an example.
These styles are defined by
\pgfplotsset{
/pgfplots/xlabel absolute/.style={%
/pgfplots/every axis x label/.style={at={(0.5,0)},below,yshift=-15pt},%
/pgfplots/every x tick scale label/.style={
at={(1,0)},yshift=-2em,left,inner sep=0pt
},
},
/pgfplots/ylabel absolute/.style={%
/pgfplots/every axis y label/.style={at={(0,0.5)},xshift=-35pt,rotate=90},
/pgfplots/every y tick scale label/.style={
at={(0,1)},above right,inner sep=0pt,yshift=0.3em
},
}
}
There is no predefined absolute placement style for three dimensional axes.
Whenever possible, consider using /.append style instead of overwriting the default styles to ensure
compatibility with future versions.
\pgfplotsset{every axis label/.append style={...}}

\pgfplotsset{every axis x label/.append style={...}}
\pgfplotsset{every axis y label/.append style={...}}
/pgfplots/title={htexti}
Adds a caption to the plot. This will place a Tik Z-node with
\node[every axis title] {text};
to the current axis.
µ = 0.1, = 0.2
\begin{tikzpicture}
\begin{loglogaxis}[
1
10 xlabel=Dof,ylabel=Error,
title={$\mu=0.1$, $\sigma=0.2$}]
2
10 \addplot coordinates {
(5, 8.312e-02)
Error
3
10 (17, 2.547e-02)
(49, 7.407e-03)
4 (129, 2.102e-03)
10 (321, 5.874e-04)
(769, 1.623e-04)
5
10 (1793, 4.442e-05)
(4097, 1.207e-05)
(9217, 3.261e-06)
101 102 103 104 };
\end{loglogaxis}
Dof \end{tikzpicture}%
The title’s appearance and/or placement can be reconfigured with

\pgfplotsset{title style={at={(0.75,1)}}}
% or, equivalently,
\pgfplotsset{every axis title/.append style={at={(0.75,1)}}}
This will place the title at 75% of the x-axis. The coordinate (0, 0) is the lower left corner and (1, 1)
the upper right one (see axis description cs for details).
Use title/.add={hprefix i}{hsuffix i} to modify an already assigned title.
/pgfplots/extra description/.code={h... i}
Allows to insert hcommandsi after axis labels, titles and legends have been typeset.
As all other axis descriptions, the code can use (0, 0) to access the lower left corner and (1, 1) to access
the upper right one. It won’t be clipped.
extra description/.code={
20 \node at (0.5,0.5) {Center!};
}}}
\begin{tikzpicture}
\begin{axis}
Center! \addplot {x^2};
10 \end{axis}
\end{tikzpicture}
0
6 4 2 0 2 4 6
4.9.4 Legends
Legends can be generated in two ways: the first is to use \addlegendentry or \legend inside of an axis.
The other method is to use the key legend entries.
\addlegendentry[hoptionsi]{hnamei}
Adds a single legend entry to the legend list. This will also enable legend drawing.
3 Case 1 \begin{tikzpicture}
Case 2 \begin{axis}
\addplot[smooth,mark=*,blue] coordinates {
(0,2)
2 (2,3)
(3,1)
};
1 \addlegendentry{Case 1}
\addplot[smooth,color=red,mark=x]
coordinates {
0 (0,0)
(1,1)
0 1 2 3 (2,1)
(3,2)
};
\addlegendentry{Case 2}
\end{axis}
\end{tikzpicture}
It does not matter where \addlegendentry commands are placed, only the sequence matters. You will
need one \addlegendentry for every \addplot command (unless you prefer an empty legend).
The optional hoptionsi a↵ect how the text is drawn; they apply only for this particular description text.
For example, \addlegendentry[red]{Text} would yield a red legend text. Behind the scenes, the text
is placed with \node[hoptionsi] {hnamei};, so hoptionsi can be any Tik Z option which a↵ects nodes.
Using \addlegendentry disables the key legend entries.
\addlegendentryexpanded[hoptionsi]{hTEX texti}
A variant of \addlegendentry which provides a method to deal with macros inside of hTEX texti.
Suppose hTEX texti contains some sort of parameter which varies for every plot. Moreover, you like to
use a loop to generate the plots. Then, it is simpler to use \addlegendentryexpanded:
100 x2 \begin{axis}
\foreach \p in {1,2,3} {
x3 \addplot {x^\p};
\addlegendentryexpanded{$x^\p$}
}
0
\end{axis}
\end{tikzpicture}
100
6 4 2 0 2 4 6
Note that this example wouldn’t have worked with \addlegendentry{$x^\p$} because the macro \p
is no longer defined when pgfplots attempts to draw the legend.
The invocation \addlegendentryexpanded{$x^\p$} is equivalent to calling \addlegendentry{$x^2$}
if \p expands to 2.
The argument hTEX texti is expanded until nothing but un-expandable material remains (i.e. it uses
the TEX primitive \edef). Occasionally, hTEX texti contains parts which should be expanded (like \p)
and other parts which should be left unexpanded (for example \pgfmathprintnumber{\p}). Then, use
\noexpand\pgfmathprintnumber{\p}
or, equivalently
\protect\pgfmathprintnumber{\p}
to avoid expansion of the macro which follows the \protect immediately.
\legend{hlisti}
You can use \legend{hlisti} to assign a complete legend.
\legend{$d=2$,$d=3$,$d=4$,$d=5$,$d=6$}
The argument of \legend is a list of entries, one for each plot.

Two di↵erent delimiters are supported:
1. There are comma–separated lists like
\legend{$d=2$,$d=3$,$d=4$,$d=5$,$d=6$}
These lists are processed using the pgf \foreach command and are quite powerful.
The \foreach command supports a dots–notation to denote ranges like \legend{1,2,...,5} or
even \legend{$x^1$,$x^...$,$x^d$}.
Attention with periods: to avoid confusion with the dots ... notation, you may need to encap-
sulate a legend entry containing periods by curly braces: \legend{{ML spcm.},{CW spcm.},{ML
AC}} (or use the \\ delimiter, see below).
2. It is also possible to delimit the list by ‘\\’. In this case, the last element must be terminated by
\\ as well:
\legend{$a=1, b=2$\\,$a=2, b=3$\\$a=3, b=5$\\}
This syntax simplifies the use of ‘,’ inside of legend entries, but it does not support the dots–
notation.
The short marker/line combination shown in legends is acquired from the hstyle optionsi argument of
\addplot.
Using \legend overwrites any other existing legend entries.
/pgfplots/legend entries={hcomma separated listi}

This key can be used to assign legend entries just like the commands \addlegendentry and \legend.
Again, the positioning is relative to the axis rectangle (unless units like cm or pt are specified explicitly).
x \begin{tikzpicture}
x2 \begin{axis}[legend entries={$x$,$x^2$}]
20 \addplot {x};
\addplot {x^2};
\end{axis}
\end{tikzpicture}
10
6 4 2 0 2 4 6
The commands for legend creation take precedence: the key legend entries is only considered if there
is no legend command in the current axis.
a \begin{tikzpicture}
b \begin{axis}[legend entries={$x$,$x^2$}]
20 \addplot {x};
\addplot {x^2};
\legend{$a$,$b$}% overrides the option
\end{axis}
10
\end{tikzpicture}
6 4 2 0 2 4 6
Please be careful with whitespaces in hcomma separated listi: they will contribute to legend entries.
Consider using ‘%’ at the end of each line in multiline arguments (the end of line character is also a
whitespace in TEX).
Just as for \addlegendentry, it is possible to provide [hoptionsi] to single descriptions. To do so, place
the options in square brackets right before the text:
100 x2 \begin{axis}[legend entries={$x$,[red]$x^2$,$x^3$}]
\addplot {x};
x3 \addplot {x^2};
\addplot {x^3};
\end{axis}
0
\end{tikzpicture}
100
6 4 2 0 2 4 6
If the square brackets contain a comma, you can enclose the complete entry in curly braces like
{[red,font=\Huge]Text} (or you can use the ‘\\’ delimiters).
4.9.5 Legend Appearance

/pgfplots/every axis legend (style, no value)
The style “every axis legend” determines the legend’s position and outer appearance:

at={(0,0)},
anchor=south west}}
will draw it at the lower left corner of the axis while

at={(1,1)},
anchor=north east}}
means the upper right corner. The ‘anchor’ option determines which point of the legend will be placed
at (0, 0) or (1, 1).
The legend is a Tik Z-matrix, so one can use any Tik Z option which a↵ects nodes and matrices (see [6,
section 13 and 14]). The matrix is created by something like
\matrix[style=every axis legend] {

draw plot specification 1 & \node{legend 1}\\
draw plot specification 2 & \node{legend 2}\\
...
};
3 l1 \begin{tikzpicture}
l2 \begin{axis}[
2 % this modifies ’every axis legend’:
l3 legend style={font=\large}
1 ]
0 \addplot coordinates {(0,1) (1,2)};
0 0.5 1 \addplot coordinates {(0,2) (1,3)};
\legend{$l_1$,$l_2$,$l_3$}
\end{axis}
\end{tikzpicture}
3 l1
\begin{tikzpicture}
legend 2 \begin{axis}[
2 l3 % align right:
legend style={
1 cells={anchor=east},
0 }
0 0.5 1 ]
\legend{$l_1$, legend $2$,$l_3$}
\end{axis}
\end{tikzpicture}
3 l1
% similar placement as previous example:
l2 \pgfplotsset{every axis legend/.append style={
2 l3 at={(1.02,1)},
anchor=north west}}
\begin{axis}
\legend{$l_1$,$l_2$,$l_3$}
\end{axis}
\end{tikzpicture}
Use legend columns={hnumber i} to configure the number of horizontal legend entries.
l1 l2 l3
\begin{tikzpicture}
3 \pgfplotsset{every axis legend/.append style={
at={(0.5,1.03)},
2 anchor=south}}
\begin{axis}[legend columns=4]
0 \legend{$l_1$,$l_2$,$l_3$}
0 0.5 1 \end{axis}
\end{tikzpicture}
Instead of the /.append style, it is possible to use legend style as in the following example. It has
the same e↵ect.
3
\begin{tikzpicture}
l1 \begin{axis}[
2 legend style={
l2
at={(1,0.5)},
1 l3 anchor=east}]
\legend{$l_1$,$l_2$,$l_3$}
\end{axis}
\end{tikzpicture}
The default every axis legend style is
\pgfplotsset{every axis legend/.style={

cells={anchor=center},% Centered entries
inner xsep=3pt,inner ysep=2pt,nodes={inner sep=2pt,text depth=0.15em},
anchor=north east,
shape=rectangle,
fill=white,
draw=black,
at={(0.98,0.98)}
}
}
Whenever possible, consider using /.append style to keep the default styles active. This ensures
compatibility with future versions.
\pgfplotsset{every axis legend/.append style={...}}
Note that in order to disable drawing of the legend box, you can use draw=none as style argument:
With legend box
10 x \begin{axis}[tiny,title=With legend box]
8 2x
\addplot[blue]{x};
6
\addplot[red]{2*x};
4
2
\legend{$x$,$2x$}
0 \end{axis}
2 \end{tikzpicture}
4
6
8
10
12
6 5 4 3 2 1 0 1 2 3 4 5 6
Without legend box
10 x \begin{axis}[tiny,title=Without legend box,
8 2x
legend style={draw=none}]
6
\addplot[blue]{x};
4
2
\addplot[red]{2*x};
0 \legend{$x$,$2x$}
2 \end{axis}
4 \end{tikzpicture}
6
8
10
12
6 5 4 3 2 1 0 1 2 3 4 5 6
/pgfplots/legend style={hkey-value-listi}
An abbreviation for every axis legend/.append style={hkey-value-listi}.
It appends options to the already existing style every axis legend.
/pgfplots/legend pos=south west|south east|north west|north east|outer north east

A style which provides shorthand access to some commonly used legend positions.
Each of these styles appends at={(hx i,hyi)},anchor=hnamei values to every axis legend.
3
\begin{tikzpicture}
\begin{axis}[legend pos=south west]
l1 \addplot coordinates {(0,1) (1,2)};
1 l2 \addplot coordinates {(0,2) (1,3)};
l3 \legend{$l_1$,$l_2$,$l_3$}
0 \end{axis}
0 0.5 1 \end{tikzpicture}
3
\begin{tikzpicture}
\begin{axis}[legend pos=south east]
l1 \addplot coordinates {(0,1) (1,2)};
l3 \legend{$l_1$,$l_2$,$l_3$}
0 \end{axis}
l2 \begin{axis}[legend pos=north east]
\legend{$l_1$,$l_2$,$l_3$}
0 \end{axis}
l2 \begin{axis}[legend pos=north west]
\legend{$l_1$,$l_2$,$l_3$}
0 \end{axis}
3 l1
\begin{tikzpicture}
l2 \begin{axis}[legend pos=outer north east]
\legend{$l_1$,$l_2$,$l_3$}
0 \end{axis}
/pgfplots/legend cell align=left|right|center (initially center)

These keys provide horizontal alignment of legend cells.
3 a
\begin{tikzpicture}
fine
\begin{axis}[legend cell align=left,
2 legend legend pos=outer north east]
0 \legend{a,fine,legend}
0 0.5 1 \end{axis}
\end{tikzpicture}
3 a
\begin{tikzpicture}
fine
\begin{axis}[legend cell align=center,
0 0.5 1 \end{axis}
\end{tikzpicture}
3 a
\begin{tikzpicture}
fine
\begin{axis}[legend cell align=right,
0 0.5 1 \end{axis}
\end{tikzpicture}
They are actually just styles for commonly used alignment choices: the choice left is equiva-
lent to legend style={cells={anchor=west}}; the second choice right is equivalent to legend
style={cells={anchor=east}}, and center to legend style={cells={anchor=center}}. Using dif-
ferent values allows more control over cell alignment.
/pgfplots/legend columns={hnumber i} (default 1)

Allows to configure the maximum number of adjacent legend entries. The default value 1 places legend
entries vertically below each other.
Use legend columns=-1 to draw all entries horizontally.
/pgfplots/legend plot pos=left|right|none (initially left)

Configures where the small line specifications will be drawn: left of the description, right of the descrip-
tion or not at all.
/pgfplots/every legend image post (style, no value)

A style which can be used to provide drawing options to every small legend image. These options apply
after current plot style has been set, allowing users di↵erent line styles for legends than for plots.
For example, suppose you have a line plot and you plot selected markers on top of it (in the same color).
Then, you may want to draw just a single legend entry (which should contain both the line and the
markers). The following example shows a solution:
1 Parabola \begin{tikzpicture}
\begin{axis}[legend image post style={mark=*}]
\addplot+[only marks,forget plot]
coordinates {(0.5,0.75) (1,1) (1.5,0.75)};
\addplot+[mark=none,smooth,domain=0:2]
{-x*(x-2)};
0.5
\addlegendentry{Parabola}
\end{axis}
\end{tikzpicture}
0
0 0.5 1 1.5 2
The example has two \addplot commands, one for the line and one for markers. Due to the forget
plot option, the marker plot (the first one) doesn’t advance the cycle list. The axis has only one
legend entry, and since legend image post style={mark=*} has been used, the legend has a plot mark
as well. Due to the forget plot option, the marker plot will not get a separate legend label.
/pgfplots/legend image post style={hkey-value-listi}

An abbreviation for every legend image post/.append style={hkey-value-listi}.
It appends options to the already existing style every legend image post.
/pgfplots/legend image code/.code={h... i}

Allows to replace the default images which are drawn inside of legends. When this key is evaluated, the
current plot specification has already been activated (using \begin{scope}[current plot style])44 ,
so any drawing operations use the same styles as the \addplot command.
The default is the style line legend.
Technical note: At the time when legend images are drawn, the style every axis legend is in e↵ect
– which have unwanted side-e↵ects due to changed parameters (especially those concerning node place-
ment, alignment, and shifting). It might be necessary to reset these parameters manually (pgfplots
also attempts to reset the fill color).
44 This was di↵erent in versions before 1.3. The new scope features allow plot styles to change legend image code.
/pgfplots/line legend (style, no value)

A style which sets legend image code (back) to its initial value.
Its initial value is
\pgfplotsset{
/pgfplots/line legend/.style={
legend image code/.code={
\draw[mark repeat=2,mark phase=2,##1]
plot coordinates {
(0cm,0cm)
(0.3cm,0cm)
(0.6cm,0cm)
};%
}
}
}
The style line legend can also be used to apply a di↵erent legend style to one particular plot (see the
documentation on area legend for an example).
/pgfplots/empty legend (style, no value)

A style which clears legend image code, thereby omitting the legend image.
/pgfplots/area legend (style, no value)

A style which sets legend image code to
\pgfplotsset{
legend image code/.code={%
\draw[#1] (0cm,-0.1cm) rectangle (0.6cm,0.1cm);
}
}
1
p % \usetikzlibrary{patterns}
0.8 x2 \begin{axis}[area legend,
x axis x line=bottom,
axis y line=left,
0.6 domain=0:1,
legend style={at={(0.03,0.97)},
anchor=north west},
0.4
axis on top,xmin=0]
\addplot[pattern=crosshatch dots,
0.2 pattern color=blue,draw=blue,
samples=500]
{sqrt(x)} \closedcycle;
0
0 0.2 0.4 0.6 0.8 1 \addplot[pattern=crosshatch,
pattern color=blue!30!white,
draw=blue!30!white]
{x^2} \closedcycle;
\addplot[red,line legend] coordinates {(0,0) (1,1)};

\legend{$\sqrt x$,$x^2$,$x$}
\end{axis}
\end{tikzpicture}
/pgfplots/xbar legend (no value)

/pgfplots/ybar legend (no value)
/pgfplots/zbar legend (no value)
/pgfplots/xbar interval legend (no value)
/pgfplots/ybar interval legend (no value)
/pgfplots/zbar interval legend (no value)
These style keys redefine legend image code such that legends use xbar, ybar or the xbar interval
and ybar interval handlers.
first \begin{tikzpicture}
second \begin{axis}[legend pos=north west]
\addplot {x^3};
0 \addplot[ybar,fill=red,draw=red!60,
ybar legend,mark=none,samples=5]
{-30*(x +4)};
\legend{first,second}
\end{axis}
\end{tikzpicture}
200
6 4 2 0 2 4 6
The initial values for these styles might be interesting if someone wants to modify them. Here they are:
\pgfplotsset{
/pgfplots/xbar legend/.style={
/pgfplots/legend image code/.code={%
\draw[##1,/tikz/.cd,bar width=3pt,yshift=-0.2em,bar shift=0pt]
plot coordinates {(0cm,0.8em) (2*\pgfplotbarwidth,0.6em)};},
},
/pgfplots/ybar legend/.style={
\draw[##1,/tikz/.cd,bar width=3pt,yshift=-0.2em,bar shift=0pt]
plot coordinates {(0cm,0.8em) (2*\pgfplotbarwidth,0.6em)};},
},
/pgfplots/xbar interval legend/.style={%
\draw[##1,/tikz/.cd,yshift=-0.2em,bar interval width=0.7,bar interval shift=0.5]
plot coordinates {(0cm,0.8em) (5pt,0.6em) (10pt,0.6em)};},
},
/pgfplots/ybar interval legend/.style={
\draw[##1,/tikz/.cd,yshift=-0.2em,bar interval width=0.7,bar interval shift=0.5]
plot coordinates {(0cm,0.8em) (5pt,0.6em) (10pt,0.6em)};},
},
}
/pgfplots/mesh legend (no value)

Redefines legend image code such that it is compatible with mesh and surf plot handlers (for three
dimensional visualization mainly).
x y
x
y
1
0.5
1
0
0 0.2 0.4 0.5
0.6 0.8
1 0
\begin{tikzpicture}
\begin{axis}[legend pos=outer north east]
\addplot3[surf,samples=9,domain=0:1]
{(1-abs(2*(x-0.5))) * (1-abs(2*(y-0.5)))};
\addlegendentry{$\phi_x \phi_y$}
\addplot3+[ultra thick] coordinates {(0,0,0) (0.5,0,1) (1,0,0)};

\addlegendentry{$\phi_x $}
\addplot3+[ultra thick] coordinates {(1,0,0) (1,0.5,1) (1,1,0)};

\addlegendentry{$\phi_y $}
\end{axis}
\end{tikzpicture}
/pgfplots/reverse legend=true|false (initially false)

/pgfplots/legend reversed=true|false (initially false)
Allows to reverse the order in which the pairs (legend entry, plot style) are drawn.
100 x2 \begin{axis}[reverse legend]
\addplot {x};
x \addlegendentry{$x$}
\addplot {x^2};
\addlegendentry{$x^2$}
0
\addplot {x^3};
\addlegendentry{$x^3$}
\end{axis}
\end{tikzpicture}
100
6 4 2 0 2 4 6
/pgfplots/transpose legend=true|false (initially false)

/pgfplots/legend transposed=true|false (initially false)
Allows to transpose the order in which the pairs (legend entry, plot style) are drawn.
Consider a set of 3 experiments, each consisting of 2 parameters. We might want to draw them together
as in the following example:
A1 A2
B1 B2
C1 C2
10
6 4 2 0 2 4 6
\begin{tikzpicture}
\begin{axis}[
legend columns=2,
cycle multi list={%
color list\nextlist
[2 of]mark list
}]
\addplot {-x}; \addlegendentry{A1}
\addplot {-x+1}; \addlegendentry{A2}
\addplot {-1.2*x + 4}; \addlegendentry{B1}

\addplot {-1.3*x + 9}; \addlegendentry{C1}

\end{axis}
\end{tikzpicture}
An alternative might be to draw them horizontally – then, we’d like to use transpose legend to get a
flat legend:
\begin{tikzpicture}
\begin{axis}[
transpose legend,
legend columns=2,
10
legend style={at={(0.5,-0.1)},anchor=north},
cycle multi list={%
color list\nextlist
[2 of]mark list
}]
0 \addplot {-x}; \addlegendentry{A1}
\addplot {-x+1}; \addlegendentry{A2}

6 4 2 0 2 4 6 \addplot {-1.2*x + 5}; \addlegendentry{B2}
A1 B1 C1
A2 B2 C2 \addplot {-1.4*x + 10}; \addlegendentry{C2}
\end{axis}
\end{tikzpicture}
Thus, legend columns defines the input columns, before the transposition (in other words, legend
columns indicates the rows of the resulting legend).
Transposing legends has only an e↵ect if legend columns> 1. Note that reverse legend has higher
precedence: it is applied first.
4.9.6 Legends with \label and \ref

pgfplots o↵ers a \label and \ref feature for LATEX to assemble a legend manually, for example as part of
the figure caption. These references work as usual LATEX references: a \label remembers where and what
needs to be referenced and a \ref expands to proper text. In context of plots, a \label remembers the plot
specification of one plot and a \ref expands to the small image which would also be used inside of legends.
20
\begin{axis}
\addplot+[only marks,
10
samples=15,
error bars/y dir=both,
error bars/y fixed=2.5]
0
{3*x+2.5*rand};
\label{pgfplots:label1}
10
\addplot+[mark=none] {3*x};
\label{pgfplots:label2}
20
\addplot {4*cos(deg(x))};
6 4 2 0 2 4 6 \label{pgfplots:label3}
\end{axis}
\end{tikzpicture}
The picture shows the estimations \ref{pgfplots:label1} which are subjected to noise.
It appears the model \ref{pgfplots:label2} fits the data appropriately.
Finally, \ref{pgfplots:label3} is only here to get three examples.
The picture shows the estimations which are subjected to noise. It appears the model fits the data
appropriately. Finally, is only here to get three examples.
\label{hlabel namei}
\label[hreferencei]{hlabel namei}
When used after \addplot, this command creates a LATEX label named hlabel namei45 . If this label is
cross-referenced with \ref{hlabel namei} somewhere, the associated plot specification will be inserted.
Label3 = ; Label2 = Label3 = \ref{pgfplots:label3};

Label2 = \ref{pgfplots:label2}
The label is assembled using legend image code and the plot style of the last plot. Any pgfplots
option is expanded until only Tik Z (or pgf) options remain; these options are used to get an independent
label.
More precisely, the small image generated by \ref{hlabel namei} is
\tikz[/pgfplots/every crossref picture] {...}
where the contents is determined by legend image code and the plot style.
The second syntax, \label[hreferencei]{hlabel namei} allows to label particular pieces of an \addplot
command. It is (currently) only interesting for scatter/classes: there, it allows to reference particular
classes of the scatter plot. See page 112 for more details.
Note that \label information, even the small Tik Z pictures here, can be combined with the external
library for image externalization, see Section 7.1 for details (in particular, the external/mode key). In
other words, references remain valid even if the defining axis has been externalized.
\ref{hlabel namei}
Can be used to reference a labeled, single plot. See the example above.
This will also work together with hyperref links and \pageref46 .
/pgfplots/refstyle={hlabel namei}
Can be used to set the styles of a labeled, single plot. This allows to write
\addplot[/pgfplots/refstyle={pgfplots:label2}]
somewhere. Please note that it may be easier to define a style with .style.
45 This feature is only available in LATEX, sorry.
46 Older versions of pgfplots required the use of \protect\ref when used inside of captions or section headings. This is no
longer necessary.
/pgfplots/every crossref picture (style, no value)

A style which will be used by the cross-referencing feature for plots. The default is
\pgfplotsset{every crossref picture/.style={baseline,yshift=0.3em}}
/pgfplots/invoke before crossref tikzpicture={hTEX codei}

/pgfplots/invoke after crossref tikzpicture={hTEX codei}
Code which is invoked just before or just after every cross reference picture. This applies to legend
images generated with \ref, legend to name and colorbar to name images.
The initial configuration checks if the external library is in e↵ect. If so, it modifies the generated figure
names by means of \tikzappendtofigurename{_crossref}.
4.9.7 Legends Outside Of an Axis

Occasionally, one has multiple adjacent plots, each with the same legend – and just one legend suffices.
But where shall it be placed? And how? One solution is to use the overlay key to exclude the legend
from bounding box computations, and place it absolutely such that it fits. Another is the legend to name
feature:
/pgfplots/legend to name={hnamei} (initially empty)

Enables a legend export mode: instead of drawing the legend, a self–contained, independent set of
drawing commands will be stored using the label hnamei. The definition is done using \label{hnamei},
just like any other LATEX label. The name can be referenced using
\ref{hnamei}.
Thus, typing \ref{hnamei} somewhere outside of the axis, maybe even outside of any picture, will
cause the legend to be drawn.
k=1 k=2 k=3

8
60
6 400
4
40
2 200
0
20
2 0
4
0
6
6 4 2 0 2 4 6 6 4 2 0 2 4 6 6 4 2 0 2 4 6
(x + 0)k ; (x + 1)k ; (x + 2)k ; (x + 3)k

\pgfplotsset{footnotesize,samples=10}
\begin{center}% note that \centering uses less vspace...
\begin{tikzpicture}
\begin{axis}[
legend columns=-1,
legend entries={$(x+0)^k$;,$(x+1)^k$;,$(x+2)^k$;,$(x+3)^k$},
legend to name=named,
title={$k=1$}]
\addplot {x};
\addplot {x+1};
\addplot {x+2};
\addplot {x+3};
\end{axis}
\end{tikzpicture}
%
\begin{tikzpicture}
\begin{axis}[title={$k=2$}]
\addplot {x^2};
\addplot {(x+1)^2};
\addplot {(x+2)^2};
\addplot {(x+3)^2};
\end{axis}
\end{tikzpicture}
%
\begin{tikzpicture}
\begin{axis}[title={$k=3$}]
\addplot {x^3};
\addplot {(x+1)^3};
\addplot {(x+2)^3};
\addplot {(x+3)^3};
\end{axis}
\end{tikzpicture}
\\
\ref{named}
\end{center}
Note that only the first plot has legend entries. Thus, its legend will be created as usual, and
stored under the name ‘named’, but it won’t be drawn. The stored legend can then be drawn with
\ref{named} below the three plots. Since there is no picture in this context, a \tikz picture is created
and a \matrix[/pgfplots/every axis legend] path is drawn inside of it, resulting in the legend as
if it had been placed inside of the axis.
The stored legend will contain the currently active values of legend- and plot style related options. This
includes legend image code, every axis legend, and any plot style options (and some more). The
algorithm works in the same way as for \label and \ref, i.e. it keeps any options with /tikz/ prefix
and expands those with /pgfplots/ prefix.
Note that the legend is drawn with every axis legend, even though the placement options might be
chosen to fit into an axis. You may want to adjust the style in the same axis in which the stored legend
has been defined (the value will be copied and restored as well).
About \ref{hnamei} The \ref{hnamei} command retrieves a stored legend (one defined by legend
to name) and draws it.
(x + 0)k ; (x + 1)k ; (x + 2)k ; (x + 3)k
\ref{named}:
If you want the legend to be exported and drawn inside of the current axis, consider using extra
description/.append code={\ref{hnamei}}.
Note that \ref can be combined with the external library for image externalization. In other words,
the legend will work even if the defining axis has been externalized, see Section 7.1 for details (in
particular the external/mode key).
Note furthermore that this .aux file related stu↵ is (currently) only supported, if pgfplots is run by
means of LATEX, sorry.
\pgfplotslegendfromname{hnamei}
This command poses an equivalent alternative for \ref{hnamei}: it has essentially the same e↵ect,
but it does not create links when used with the hyperref package47 .
/pgfplots/every legend to name picture (style, no value)

A style which is installed when \ref is used outside of a picture: a new picture will be created with
\tikz[/pgfplots/every legend to name picture].
Thus, you can redefine this style to set alignment options (such as baseline). For example, the
initialization
\pgfplotsset{
legend style={matrix anchor=west,at={(0pt,0pt)}},
every legend to name picture/.style={baseline},
}
...
will cause the legend to be positioned such that its west anchor is at y=0pt. The baseline option
will align this point of the legend with the text baseline (please refer to the documentation for
baseline in Section 4.19 for details).
4.9.8 Legends with Customized Texts or Multiple Lines

\addlegendimage{hoptionsi}
Adds a further legend image for legend creation.
Each \addplot command appends its plot style options to a list, and \addlegendimage adds hoptionsi
to the very same list.
Thus, the e↵ect is as if you had provided \addplot[hoptionsi], but \addlegendimage bypasses all the
logic usually associated with a plot. In other words: except for the legend, the state of the axis remains
as if the command would not have been issued. Not even the current plot’s index is advanced.
2
10 x2 \begin{semilogyaxis}[
domain=0:4,
x3
101 — ]
\addplot {x}; \addlegendentry{$x$}
x 1 \addplot {x^2}; \addlegendentry{$x^2$}
100 \addplot {x^3}; \addlegendentry{$x^3$}
x 2
\addlegendimage{empty legend}
10 1 x 3 \addlegendentry{---}
\addplot {x^(-1)}; \addlegendentry{$x^{-1}$}
2 \addplot {x^(-2)}; \addlegendentry{$x^{-2}$}
10
\end{semilogyaxis}
The example above has six plots, each with its legend entry. Furthermore, it has an \addlegendimage
command and its separate legend entry. We see that \addlegendimage needs its own legend entry, but
it is detached from the processing of plots as such. In our case, we chose empty legend as style for the
separator.
Use \addlegendimage to provide custom styles into legends, for example to document custom \draw
commands inside of an axis.
You can call \label after \addlegendimage just as for a normal style.
Occasionally, one may want multiple lines for legend entries. That is possible as well using a fixed text
width:
47 Since this manual uses colored links, the text in \ref would usually be blue. Using \pgfplotslegendfromname avoids link
text colors in the legend (this has been applied to the manual styles here).
2
domain=0:4,
x3
101 ]
Neg. \addplot {x}; \addlegendentry{$x$}
100 Sign: \addplot {x^2}; \addlegendentry{$x^2$}
\addplot {x^3}; \addlegendentry{$x^3$}
x 1 \addlegendimage{empty legend}
1
10 x 2 \addlegendentry[text width=25pt,text depth=]
x 3 {Neg. Sign:}
2 \addplot {x^(-1)}; \addlegendentry{$x^{-1}$}
10
0 1 2 3 4 \end{semilogyaxis}
\end{tikzpicture}
The example provides options for the single multiline element. Note that the initial configuration of legend
style employs text depth=0.15em, which needs to be reset manually to text depth={}48 .
There are two approaches with the same e↵ect which are subject of the following example:
domain=0:4,
x3
101 legend entries={%
Neg. $x$,$x^2$,$x^3$,%
100 Sign: {[text width=25pt,text depth=]Neg. Sign:},%
$x^{-1}$,$x^{-2}$,$x^{-3}$},
x 1 % same effect:
1
10 x 2 % legend style={
x 3 % nodes={text width=25pt,text depth=}}
2 ]
10
\addplot {x};
\addplot {x^2};
0 1 2 3 4 \addplot {x^3};
\addlegendimage{empty legend}
\addplot {x^(-1)};
\addplot {x^(-2)};
\addplot {x^(-3)};
\end{semilogyaxis}
\end{tikzpicture}
Here, the legend entries are provided using the single key syntax. Note that the special options are
provided as part of the legend entry, using square brackets right before the text as such. The comments
indicate that you could also add the text width stu↵ to legend style, in which case it would hold for
every node.
Note that legend texts are realized using \node[hoptionsi] {htexti};, so anything which produces a valid
Tik Z node is permitted (this includes minipage or tabular environments inside of htexti).
4.9.9 Axis Lines

An extension by Pascal Wolkotte
By default the axis lines are drawn as a box, but it is possible to change the appearance of the x and y axis
lines.
/pgfplots/axis x line=box|top|middle|center|bottom|none (initially box)

/pgfplots/axis x line*=box|top|middle|center|bottom|none (initially box)
/pgfplots/axis y line=box|left|middle|center|right|none (initially box)
/pgfplots/axis y line*=box|left|middle|center|right|none (initially box)
/pgfplots/axis lines=box|left|middle|center|right|none
/pgfplots/axis lines*=box|left|middle|center|right|none
These keys allow to choose the locations of the axis lines. The last one, axis lines sets the same value
for every axis.
48 Perhaps I can reset text depth automatically in the future.
Ticks and tick labels are placed according to the chosen value as well. The choice bottom will draw the
x line at y = ymin , middle will draw the x line at y = 0, and top will draw it at y = ymax . Finally,
box is a combination of options top and bottom. The choice axis x line=none is an alias for hide x
axis. The y- and z variants work in a similar way.
The case center is a synonym for middle, both draw the line through the respective coordinate 0. If
this coordinate is not part of the axis limit, the lower axis limit is chosen instead.
The starred versions . . . line* only a↵ect the axis lines, without correcting the positions of axis labels,
tick lines or other keys which are (possibly) a↵ected by a changed axis line. The non-starred versions
are actually styles which set the starred key and some other keys which also a↵ect the figure layout:
• In case axis x line=box, the style every boxed x axis will be installed immediately.
• In case axis x line6=box, the style every non boxed x axis will be installed immediately. Fur-
thermore, some of these choices will modify axis label positions.
The handling of axis y line and axis z line is similar. The default styles are defined as
\pgfplotsset{
every non boxed x axis/.style={
xtick align=center,
x axis line style={-stealth}
},
every boxed x axis/.style={}
}
In addition, conditional modifications of axis label styles will be taken. For example, axis x
line=middle will set
\pgfplotsset{every axis x label/.style={at={(current axis.left of origin)},anchor=south west}}
if the matching y style has value axis y line=right and
\pgfplotsset{every axis x label/.style={at={(current axis.right of origin)},anchor=south east}}
if axis y line6=right.
Feel free to overwrite these styles if the default doesn’t fit your needs or taste. Again, these styles will
not be used for axis line*.
\begin{axis}[
xlabel=$x$,ylabel=$\sin x$]
\addplot[blue,mark=none,
sin x
domain=-10:0,samples=40]
0
{sin(deg(x))};
\end{axis}
\end{tikzpicture}
1
10 8 6 4 2 0
x
1
\begin{tikzpicture}
\begin{axis}[
axis x line=middle,
0.5 axis y line=right,
ymax=1.1, ymin=-1.1,
x
sin x
xlabel=$x$,ylabel=$\sin x$
0 ]
10 8 6 4 2
0.5 {sin(deg(x))};
\end{axis}
\end{tikzpicture}
1
2
\begin{tikzpicture}
\begin{axis}[
axis x line=bottom,
1.5 axis y line=left,
xlabel=$x$,ylabel=$\sqrt{|x|}$
]
|x|
1 \addplot[blue,mark=none,
p
{sqrt(abs(x))};
\end{axis}
0.5
\end{tikzpicture}
0
4 2 0 2 4
x
1
sin x \begin{tikzpicture}
\begin{axis}[
minor tick num=3,
0.5 axis y line=center,
axis x line=middle,
x xlabel=$x$,ylabel=$\sin x$
]
4 2 2 4 \addplot[smooth,blue,mark=none,
0.5 {sin(deg(x))};
\end{axis}
\end{tikzpicture}
1
\begin{tikzpicture}
\begin{axis}[
minor tick num=3,
0.5 axis y line=left,
axis x line=middle,
x xlabel=$x$,ylabel=$\sin x$
sin x
0 ]
4 2 0 2 4 \addplot[smooth,blue,mark=none,
0.5 {sin(deg(x))};
\end{axis}
\end{tikzpicture}
In case middle, the style every inner axis x line allows to adjust the appearance.
Note that three dimensional axes only support to use the same value for every axis, i.e. three dimensional
axes support only the axis lines key (or, preferably for 3D axes, the axis lines* key – check what
looks best). See Section 4.11.4 for examples of three dimensional axis line variations.
/pgfplots/every inner x axis line (no value)

/pgfplots/every inner y axis line (no value)
/pgfplots/every inner z axis line (no value)
A style key which can be redefined to customize the appearance of inner axis lines. Inner axis lines are
those drawn by the middle (or center) choice of axis x line, see above.
This style a↵ects only the line as such.
y3 \begin{tikzpicture}
100 \begin{axis}[
minor tick num=1,
axis x line=middle,
axis y line=middle,
50 every inner x axis line/.append style=
{|->>},
every inner y axis line/.append style=
{|->>},
x xlabel=$x$,ylabel=$y^3$
]
2 2 4 \addplot[blue,domain=-3:5] {x^3};
\end{axis}
\end{tikzpicture}
/pgfplots/every outer x axis line (no value)

/pgfplots/every outer y axis line (no value)
/pgfplots/every outer z axis line (no value)
Similar to every inner x axis line, this style configures the appearance of all axis lines which are
part of the outer box.
\begin{tikzpicture}
\begin{axis}[
0.2 separate axis lines, % important !
every outer x axis line/.append style=
{-stealth},
every outer y axis line/.append style=
{-stealth},
0 ]
\addplot[blue,id=DoG,
samples=100,
domain=-15:15]
0.2 gnuplot{1.3*exp(-x**2/10) - exp(-x**2/20)};
\end{axis}
/pgfplots/axis line style={hkey-value-listi}

A command which appends hkey-value-listi to all axis line appearance styles.
/pgfplots/inner axis line style={hkey-value-listi}

A command which appends hkey-value-listi to both, every inner x axis line and the y variant.
/pgfplots/outer axis line style={hkey-value-listi}

A command which appends hkey-value-listi to both, every outer x axis line and the y variant.
/pgfplots/x axis line style={hkey-value-listi}

/pgfplots/y axis line style={hkey-value-listi}
/pgfplots/z axis line style={hkey-value-listi}
A command which appends hkey-value-listi to all axis lines styles for either x or y axis.
/pgfplots/every boxed x axis (no value)

/pgfplots/every boxed y axis (no value)
/pgfplots/every boxed z axis (no value)

A style which will be installed as soon as axis x line=box (y) is set.
The default is simply empty.
/pgfplots/every non boxed x axis (no value)
/pgfplots/every non boxed y axis (no value)
/pgfplots/every non boxed z axis (no value)
A style which will be installed as soon as axis x line (y) will be set to something di↵erent than box.
The default is
\pgfplotsset{
every non boxed x axis/.style={
xtick align=center,
x axis line style={-stealth}}}
with similar values for the y-variant. Feel free to redefine this style to your needs and taste.
/pgfplots/separate axis lines={htrue,falsei} (default true)
Enables or disables separate path commands for every axis line. This option a↵ects only the case if axis
lines are drawn as a box.
Both cases have their advantages and disadvantages, I fear there is no reasonable default (suggestions
are welcome).
The case separate axis lines=true allows to draw arrow heads on each single axis line, but it can’t
close edges very well – in case of thick lines, unsatisfactory edges occur.
\begin{tikzpicture}
\begin{axis}[
0.2 separate axis lines,
{-stealth,red},
{-stealth,green!30!black},
0 ]
\addplot[blue,
samples=100,
domain=-15:15]
0.2 {1.3*exp(0-x^2/10) - exp(0-x^2/20)};
% Unfortunately, there is a bug in PGF 2.00
10 0 10 % something like exp(-10^2)
% must be written as exp(0-10^2) :-(
\end{axis}
\end{tikzpicture}
The case separate axis lines=false issues just one path for all axis lines. It draws a kind of
rectangle, where some parts of the rectangle may be skipped over if they are not wanted. The advantage
is that edges are closed properly. The disadvantage is that at most one arrow head is added to the path
(and yes, only one drawing color is possible).
\begin{tikzpicture}
\begin{axis}[
0.2 separate axis lines=false,
{-stealth,red},
{-stealth,green!30!black},
0 ]
\addplot[blue,id=DoG,
samples=100,
domain=-15:15]
0.2 gnuplot{1.3*exp(-x**2/10) - exp(-x**2/20)};
\end{axis}
4.9.10 Moving Axis Lines

pgfplots o↵ers styles to shift axis lines along their normal vector:
/pgfplots/axis x line shift={hdimension or uniti} (initially empty)
/pgfplots/axis y line shift={hdimension or uniti} (initially empty)
/pgfplots/axis z line shift={hdimension or uniti} (initially empty)
/pgfplots/axis line shift={hdimension or uniti} (initially empty)
Allows to shift one or more axis lines together with all their descriptions.
\begin{tikzpicture}
5 \begin{axis}[
axis line shift=10pt,
grid=major]
\addplot+[samples=3] {x};
\end{axis}
\end{tikzpicture}
0
6 4 2 0 2 4 6
The option takes the axis lines and shifts them along their “outer normal vectors”. This direction is
known for every axis; it is the direction which points away from the center.
The operation is often combined with the axis lines variations, for example
\begin{axis}[
axis lines=left,
2 axis line shift=10pt,
grid=major]
0 \end{axis}
\end{tikzpicture}
2
4 2 0 2 4
Keep in mind that axis lines=left also reconfigures the axis such that it has an arrow head and
enlargelimits=false (see also axis lines*=left). Customizations are, of course possible, for exam-
ple an arrow tip which shows the end of the axis:
\begin{axis}[
axis lines=left,
2 axis line style={|-stealth},
axis line shift=10pt,
grid=major]
0 \addplot+[samples=3] {x};
\end{axis}
2 \end{tikzpicture}
4 2 0 2 4
The argument hdimension or uniti can be a dimension like 10pt (as in the previous examples). However,
it can also be given in terms of axis units:
\begin{tikzpicture}
100 \begin{axis}[
axis lines=center,
axis y line shift=2,
axis x line shift=100,
grid=major]
0 \addplot+[samples=3] {30*x};
\end{axis}
\end{tikzpicture}
100
4 2 0 2 4
In this context, axis y line shift=2 shifts the y axis by 2 units along the x axis since the x axis
(with negative sign) is the “outer normal”. Shifting by axis units is only possible for two–dimensional
plots since the outer normal is well-defined here: it is the other available axis.
The feature axis line shift can also be applied to three dimensional axes (with a hdimensioni argu-
ment, not a unit). In this case, the shift is also along the “outer normal vector”. Note that this vector
is a combination of the two other axes, so the resulting axis is no longer aligned with grid lines! Handle
this option with care as it easily results in confusion when one tries to align the axis descriptions with
grid lines.
\begin{tikzpicture}
\begin{axis}[
axis lines=left,
10 axis line shift=10pt,
grid=major]
\addplot3[surf,shader=interp,samples=2] {x+y};
\end{axis}
0 \end{tikzpicture}
5
10
0
4 2 0 2 4 5
The option axis line shift is a style which configures a shifts of hdimensioni for each axis.
Note that axis line shift implies separate axis lines and will implicitly configure hide obscured
x ticks=false and its variants..
4.9.11 Two Ordinates (y axis) or Multiple Axes

In some applications, more than one y axis is used if the x range is the same. This section demonstrates
how to create them. The idea in pgfplots is to draw two axes on top of each other, one with descriptions
only on the left and the second with descriptions only on the right:
\begin{tikzpicture}
% let both axes use the same layers
Second ordinate
20 10
First ordinate
\pgfplotsset{set layers}
\begin{axis}[
scale only axis,
0 xmin=-5,xmax=5,
10 axis y line*=left,% the ’*’ avoids arrow heads
xlabel=$x$,
10 ylabel=First ordinate]
\addplot {x^2};
0
\end{axis}
4 2 0 2 4
x \begin{axis}[
scale only axis,
xmin=-5,xmax=5,
axis y line*=right,
axis x line=none,
ylabel=Second ordinate]
\addplot[red] {3*x};
\end{axis}
\end{tikzpicture}
Thus, the two axes are drawn “on top” of each other – one, which contains the x axis and the left y axis,
and one which has only the right y axis. Since pgfplots does not really know what it’s doing here, user
attention in the following possibly non-obvious aspects is required:
1. Scaling. You should set scale only axis because this forces equal dimensions for both axis, without
respecting any labels.
2. Same x limits. You should set those limits explicitly.
3. You need to tell pgfplots that it should share the same graphics layers for both axes. In this case,
pgfplots will draw plots of the first axis and of the second axis onto the same layer. It will also
draw background(s) into the background layer and descriptions into the foreground layer. Use the
key \pgfplotsset{set layers} in front of the first axis to prepare the complete picture for layered
graphics.
You may want to consider di↵erent legend styles. It is also possible to use only the axis, without any plots:
1,000‰ % \usepackage{textcomp}
\begin{tikzpicture}
800‰ % let both axes use the same layers
20
per thousand
Absolute
600‰ \begin{axis}[
scale only axis,
10 xmin=-5,xmax=5,
400‰ axis y line*=left,%’*’ avoids arrow heads
xlabel=$x$,
200‰ ylabel=Absolute]
0 \addplot {x^2};
0‰ \end{axis}
4 2 0 2 4
x \begin{axis}[
scale only axis,
xmin=-5,xmax=5,
ymin=0,ymax=1000,
yticklabel=
{$\pgfmathprintnumber{\tick}$\textperthousand},
axis y line*=right,
axis x line=none,
ylabel=per thousand]
\end{axis}
\end{tikzpicture}
4.9.12 Axis Discontinuities

An extension by Pascal Wolkotte
In case the range of either of the axis do not include the zero value, it is possible to visualize this with a
discontinuity decoration on the corresponding axis line.
/pgfplots/axis x discontinuity=crunch|parallel|none (initially none)

/pgfplots/axis y discontinuity=crunch|parallel|none (initially none)
/pgfplots/axis z discontinuity=crunch|parallel|none (initially none)
Insert a discontinuity decoration on the x (or y, respectively) axis. This is to visualize that the y axis
does cross the x axis at its 0 value, because the minimum x axis value is positive or the maximum value
is negative.
The description applies to axis y discontinuity and axis z discontinuity as well, simply substi-
tute x by y or z, respectively.
\begin{tikzpicture}
6 \begin{axis}[
axis x line=bottom,
axis x discontinuity=parallel,
axis y line=left,
4 xmin=360, xmax=600,
ymin=0, ymax=7,
enlargelimits=false
2 ]
(420,2)
(500,6)
0 (590,4)
400 450 500 550 600 };
\end{axis}
\end{tikzpicture}
\begin{tikzpicture}
120 \begin{axis}[
axis x line=bottom,
axis y line=center,
tick align=outside,
110 axis y discontinuity=crunch,
ymin=95, enlargelimits=false
]
100 {x*x+x+104};
\end{axis}
\end{tikzpicture}
4 2 0 2 4
A problem might occur with the placement of the ticks on the axis. This can be solved by specifying the
minimum or maximum axis value for which a tick will be placed.
/pgfplots/xtickmin={hcoord i} (default axis limits)

/pgfplots/ytickmin={hcoord i} (default axis limits)
/pgfplots/ztickmin={hcoord i} (default axis limits)
/pgfplots/xtickmax={hcoord i} (default axis limits)
/pgfplots/ytickmax={hcoord i} (default axis limits)
/pgfplots/ztickmax={hcoord i} (default axis limits)
The options xtickmin, xtickmax and ytickmin, ytickmax allow to define the axis tick limits, i.e. the
axis values before respectively after no ticks will be placed. Everything outside of the axis tick limits
will be not drawn. Their default values are equal to the axis limits.
\begin{tikzpicture}
120 \begin{axis}[
axis x line=bottom,
axis y line=center,
tick align=outside,
110 axis y discontinuity=crunch,
xtickmax=3,
ytickmin=110,
ymin=95, enlargelimits=false
]
{x*x+x+104};
4 2 0 2 \end{axis}
\end{tikzpicture}
/pgfplots/hide x axis=true|false (initially false)

/pgfplots/hide y axis=true|false (initially false)
/pgfplots/hide z axis=true|false (initially false)
/pgfplots/hide axis=true|false (initially false)
Allows to hide either a selected axis or all of them. No outer rectangle, no tick marks and no labels will
be drawn. Only titles and legends will be processed as usual.
Axis scaling and clipping (!) will be done as if you did not use hide axis.
x2 cos(x) \begin{tikzpicture}
\begin{axis}[
hide x axis,
hide y axis,
title={$x^2\cos(x)$}]
\addplot {cos(x)*x^2};
\end{axis}
\end{tikzpicture}% <- eliminate space.
x2 cos(x) \begin{tikzpicture}
\begin{axis}[
hide x axis,
axis y line=left,
20 title={$x^2\cos(x)$}]
\addplot {cos(x)*x^2};
\end{axis}
\end{tikzpicture}% <- eliminate space
10
Note that a hidden axis contributes nothing to the resulting picture’s bounding box49 , see clip has
bounding box.
49 Since version 1.8.
% requires \usepgfplotslibrary{patchplots}
% and compat=1.8 or newer
\begin{tikzpicture}
\begin{axis}[
hide axis,
]
\addplot3[patch,
patch type=biquadratic,
shader=interp]
coordinates {
(0,0,1) (6,1,0) (5,5,0) (-1,5,0)
(3,1,0) (6,3,0) (2,6,0) (0,3,0)
(3,3.75,0)
};
\end{axis}
\end{tikzpicture}% <- eliminate space
This can be used to embed a pgfplots path which needs an axis to a standard Tik Z picture. See also
Section 4.26 for details how to synchronize the alignment between a pgfplots figure (which typically
rescales its coordinates) to that of a standard tikzpicture.
Note that pgfplots uses the input coordinates to determine the bounding box of the picture. In this
case, the bounding box is slightly smaller than the shading. A cure would be to increase the bounding
box manually.
You may want to disable the clip path using the option clip=false.
4.9.13 Color Bars

pgfplots supports mesh, surface and scatter plots which can use color maps. While color maps can be
chosen as described in Section 4.7.6, they can be visualized using color bars.
/pgfplots/colorbar=true|false (initially false)

Activates or deactivates color bars.
5 4
0 0
5 4
6 4 2 0 2 4 6
\begin{tikzpicture}
\addplot[mesh,ultra thick] {x};
\end{axis}
\end{tikzpicture}
5 4
0 0
5 4
6 4 2 0 2 4 6
\begin{tikzpicture}
\begin{axis}[colorbar,colormap/greenyellow]
\end{axis}
\end{tikzpicture}
\begin{axis}[colorbar horizontal]
\end{axis}
\end{tikzpicture}
0
5
6 4 2 0 2 4 6
4 2 0 2 4
A color bar is only useful for plots with non–zero color data range, more precisely, for which minimum
and maximum point meta data is available. Usually, this is the case for scatter, mesh or surf (or
similar) plots, but you can also set point meta min and point meta max manually in order to draw a
colorbar.
Color bars are just normal axes which are placed right besides their parent axes. The only di↵erence
is that they inherit several styles such as line width and fonts and they contain a bar shaded with the
color map of the current axis.
Color bars are drawn internally with
\axis[every colorbar,colorbar shift,colorbar=false]

\addplot graphics {};
\endaxis
where the placement, alignment, appearance and other options are done by the two styles every
colorbar and colorbar shift. These styles and the possible placement and alignment options are
described below.
Remarks for special cases:

• Since there is always only one color bar per plot, this color bar uses the axis wide configurations
of color map and color data. Consider using colorbar source to select color data limits of a
particular \addplot command instead.
• If someone needs more than one color bar, the draw command above needs to be updated. See the
key colorbar/draw/.code for this special case.
/pgfplots/colorbar right (style, no value)
A style which redefines every colorbar and colorbar shift such that color bars are placed right of
their parent axis.
This is the initial configuration.
10
10
8
6
5
4
2
0
0 1 2 3
\begin{tikzpicture}
\begin{axis}[colorbar right]
\addplot[mesh,thick,samples=150,domain=0.1:3]
{1/x};
\end{axis}
\end{tikzpicture}
The style colorbar right is defined as

\pgfplotsset{
colorbar right/.style={
/pgfplots/colorbar=true,
/pgfplots/colorbar shift/.style={xshift=0.3cm},
/pgfplots/every colorbar/.style={
title=,
xlabel=,
ylabel=,
zlabel=,
legend entries=,
axis on top,
at={(parent axis.right of north east)},
anchor=north west,
xmin=0,
xmax=1,
ymin=\pgfkeysvalueof{/pgfplots/point meta min},
ymax=\pgfkeysvalueof{/pgfplots/point meta max},
plot graphics/xmin=0,
plot graphics/xmax=1,
plot graphics/ymin=\pgfkeysvalueof{/pgfplots/point meta min},
plot graphics/ymax=\pgfkeysvalueof{/pgfplots/point meta max},
scale only axis,
height=\pgfkeysvalueof{/pgfplots/parent axis height},
x=\pgfkeysvalueof{/pgfplots/colorbar/width},
xtick=\empty,
colorbar vertical/lowlevel,
}
},
/pgfplots/colorbar vertical/lowlevel/.style={
plot graphics/lowlevel draw/.code 2 args={%
\pgfuseshading{...} % some advanced basic level shading operations
}
},
}
Attention: colorbar right redefines every colorbar. That means any user customization must
take place after colorbar right:
% correct:
\begin{axis}[colorbar right, colorbar style={<some customization>}]
% wrong, colorbar right resets the customization:
\begin{axis}[colorbar style={<some customization>}, colorbar right]
/pgfplots/colorbar left (style, no value)

A style which re-defines every colorbar and colorbar shift such that color bars are placed left of
their parent axis.
4 5
0 0
4 5
6 4 2 0 2 4 6
\begin{tikzpicture}
\begin{axis}[colorbar left]
\addplot[mesh,thick,samples=150]
{x*sin(deg(4*x))};
\end{axis}
\end{tikzpicture}
The style colorbar left is defined as
\pgfplotsset{
colorbar left/.style={
/pgfplots/colorbar right,
/pgfplots/colorbar shift/.style={xshift=-0.3cm},
/pgfplots/every colorbar/.append style={
at={(parent axis.left of north west)},
anchor=north east,
yticklabel pos=left,
}
}
}
Attention: colorbar left redefines every colorbar. That means any user customization must
take place after colorbar left (see also the documentation for colorbar right).
/pgfplots/colorbar horizontal (style, no value)

A style which re-defines every colorbar and colorbar shift such that color bars are placed below
their parent axis, with a horizontal bar.
\begin{axis}[colorbar horizontal]
\addplot[only marks,scatter,
scatter src={mod(\coordindex,15)},samples=150]
{rand};
\end{axis}
0 \end{tikzpicture}
1
6 4 2 0 2 4 6
0 5 10 15
This style is defined as
\pgfplotsset{
colorbar horizontal/.style={
/pgfplots/colorbar=true,
/pgfplots/colorbar shift/.style={yshift=-0.3cm},
/pgfplots/every colorbar/.style={
title=,
xlabel=,
ylabel=,
zlabel=,
legend entries=,
axis on top,
at={(parent axis.below south west)},
anchor=north west,
ymin=0,
ymax=1,
xmin=\pgfkeysvalueof{/pgfplots/point meta min},
xmax=\pgfkeysvalueof{/pgfplots/point meta max},
plot graphics/ymin=0,
plot graphics/ymax=1,
plot graphics/xmin=\pgfkeysvalueof{/pgfplots/point meta min},
plot graphics/xmax=\pgfkeysvalueof{/pgfplots/point meta max},
scale only axis,
width=\pgfkeysvalueof{/pgfplots/parent axis width},
y=\pgfkeysvalueof{/pgfplots/colorbar/width},
xticklabel pos=left,
ytick=\empty,
colorbar horizontal/lowlevel,
}%
},%
/pgfplots/colorbar horizontal/lowlevel/.style={%
plot graphics/lowlevel draw/.code 2 args={%
\pgfuseshading{...} % some advanced basic level shading operations
},%
},%
}
Attention: colorbar horizontal re-defines every colorbar. That means any user customization
must take place after colorbar horizontal:
% correct:
\begin{axis}[colorbar horizontal, colorbar style={<some customization>}]
% wrong, colorbar horizontal resets the customization:
\begin{axis}[colorbar style={<some customization>}, colorbar horizontal]
/pgfplots/every colorbar (style, no value)

This style governs the placement, alignment and appearance of color bars. Any desired detail changes
for color bars can be put into this style. Additionally, there is a style colorbar shift which is set
after every colorbar. The latter style is intended to contain only shift transformations like xshift or
yshift (making it easier to overwrite or deactivate them).
While a color bar is drawn, the predefined node parent axis can be used to align at the parent axis.
Predefined node parent axis
A node for the parent axis of a color bar. It is only valid for color bars.
Thus,
\pgfplotsset{
colorbar style={
at={(parent axis.right of north east)},
anchor=north west,
},
colorbar shift/.style={xshift=0.3cm}
}
places the colorbar in a way that its top left (north west) corner is aligned right of the top right corner
(right of north east) of its parent axis. Combining this with the colorbar shift is actually the
same as the initial setting.
Since color bars depend on some of its parent’s properties, these properties are available as values of
the following keys:
/pgfplots/point meta min (no value)
/pgfplots/point meta max (no value)
The values of these keys contain the lower and upper bound of the color map, i.e. the lower and
upper limit for the color bar.
The value is \pgfkeysvalueof{/pgfplots/point meta min} inside of every colorbar.
The value is usually determined using the axis wide point meta limits, i.e. they are computed as
minimum and maximum value over all plots (unless the user provided limits manually). Consider
the colorbar source key if you’d like to select point meta limits of one specific \addplot command.
/pgfplots/colorbar source={htrue,falsei} (initially false)

Allows to select a specific \addplot command whose point meta limits are taken as upper and
lower limit of a colorbar’s data range. This a↵ects the tick descriptions of the colorbar. It needs
to be provided as argument to \addplot, i.e. using
\addplot[...,colorbar source] ...

% or
\addplot+[colorbar source] ...
or as key inside of a cycle list.

Using colorbar source automatically implies point meta rel=per plot for that specific plot.
If there are more than one \addplot commands with colorbar source, the last one is selected.
/pgfplots/parent axis width (no value)

/pgfplots/parent axis height (no value)
The values of these keys contain the size of the parent axis. They can be used as width and/or
height arguments for every colorbar with \pgfkeysvalueof{/pgfplots/parent axis width}.
These values are only valid inside of color bars.
Besides these values, each color bar inherits a list of styles of its parent axis, namely
• every tick,
• every minor tick,
• every major tick,
• every axis grid,
• every minor grid,
• every major grid,

• every tick label.
This can be used to inherit line width and/or fonts.
Customization: “colorbar top” \begin{tikzpicture}
\begin{axis}[
0.5 1 1.5 2 2.5 3
colorbar style={
at={(0.5,1.03)},anchor=south,
3 xticklabel pos=upper
},
title style={yshift=1cm},
title=Customization: ‘‘colorbar top’’]
2
{x};
\end{axis}
1 \end{tikzpicture}
0
0 1 2 3
More Customization: “colorbar top” \begin{tikzpicture}
\begin{axis}[
1 2 3
colorbar style={
at={(1,1.03)},anchor=south east,
3 width=0.5*
\pgfkeysvalueof{/pgfplots/parent axis width},
xticklabel pos=upper,
},
2 title style={yshift=1cm},
title=More Customization: ‘‘colorbar top’’]
1 {x};
\end{axis}
\end{tikzpicture}
0
0 1 2 3
Please take a look at the predefined styles colorbar right, colorbar left and colorbar horizontal
for more details about configuration possibilities for every colorbar.
Remark: A color bar is just a normal axis. That means every colorbar can contain specifications
where to place tick labels, extra ticks, scalings and most other features of a normal axis as well (except
nested color bars).
/pgfplots/colorbar style={hkey-value listi}

A shortcut for every colorbar/.append style={hkey-value listi}. It appends options to the colorbar
style.
/pgfplots/colorbar/width={hdimensioni} (initially 0.5cm)

Sets the width of a color bar.
0.2
0.2
0 0
0.2
0 2 0.2
0.5 0
1 2
\begin{tikzpicture}
\begin{axis}[
view/az=45,
colorbar,
colorbar/width=2cm,
colormap/blackwhite]
\addplot3[surf,domain=0:1,y domain=-3:3] {x*(1-x)*tanh(y)};

\end{axis}
\end{tikzpicture}
For horizontal color bars, this sets the height.
/pgfplots/colorbar shift (style, no value)

This style is installed after every colorbar. It is intended to contain only shift transformations like
xshift and/or yshift. The reason to provide two separate styles is to allow easier deactivation of shift
transformations.
\pgfplotsset{
colorbar shift/.style={xshift=1cm}
}
Predefined node current colorbar axis

A predefined node for the color bar of an axis. After \end{axis}, this node can be used to align further
graphical elements at the color bar. Note that current axis refers to the axis as such while current
colorbar axis refers to the color bar (which is an axis itself).
/pgfplots/colorbar/draw/.code={h... i}
This code key belongs to the low level interface of color bars. It is invoked whenever a color bar needs
to be drawn. Usually, it won’t be necessary to use or modify this key explicitly.
When this key is invoked, the styles inherited from the parent axis are already set and the required
variables (see the documentation of every colorbar) are initialized.
This code key can be replaced if one needs more than one color bar (or other wrinkles).
\pgfplotsset{colorbar/draw/.code={%
\axis[every colorbar,colorbar shift,colorbar=false]
\addplot graphics {};
\endaxis
}
}
Please note that a color bar axis is nothing special as such – it is just a normal axis with one plot
graphics command and it is invoked with a special set of options. The only special thing is that a set
of styles and some variables are inherited from its parent axis.
/pgfplots/colorbar sampled={hoptional optionsi} (style, default surf,mark=none,shader=flat)

A style which installs a discretely sampled color bar.
1
1
0.5
0 0
0.5
1
1
6 4 2 0 2 4 6
\begin{tikzpicture}
\begin{axis}[colorbar sampled]
\addplot[mesh,samples=40] {sin(deg(x))};
\end{axis}
\end{tikzpicture}
Attention: this style merely changes the appearance of the colorbar. The way colormaps are used
to determine plot colors is unrelated to this key. See also colormap access=direct for ways to a↵ect
the color selection.
The style uses \addplot3[hoptionsi] to draw the colorbar, with domain set to the color range and
the current value of the samples key to determine the number of samples. In other words: it uses
plot expression and a surface plot to visualize the colorbar. Use colorbar style={samples=10}
to change the number of samples.
1
1
0.5
0 0
0.5
1
1
6 4 2 0 2 4 6
\begin{tikzpicture}
\begin{axis}[colorbar sampled,colorbar style={samples=8}]
\addplot[mesh,samples=40] {sin(deg(x))};
\end{axis}
\end{tikzpicture}
The hoptionsi can be used to change the \addplot3 options used for the colorbar visualization. For
example, colorbar sampled={surf,shader=interp} will use Gouraud shading which has visually the
same e↵ect as the standard color bar.
/pgfplots/colorbar sampled line={hoptional optionsi} (style, default scatter,only marks)

A style which draws a discrete colorbar. In contrast to colorbar sampled, it visualizes the colorbar
using a line plot, not a surf plot.
0.5
0 0
0.5
1
6 4 2 0 2 4 6
\begin{tikzpicture}
\begin{axis}[colorbar sampled line]
\addplot+[scatter] {sin(deg(x))};
\end{axis}
\end{tikzpicture}
Attention: this style merely changes the appearance of the colorbar. The way colormaps are used
to determine plot colors is unrelated to this key. See also colormap access=direct for ways to a↵ect
the color selection.
The initial configuration uses a scatter plot to visualize the colorbar, it can be changed by specifying
hoptionsi.
Furthermore, the axis appearance is changed using axis y line*=left|right, depending on the posi-
tion of the color bar (or axis x line*=bottom for colorbar horizontal).
Consider the tick align=outside feature if you prefer tick lines outside of the colorbar instead of
inside.
/pgfplots/every colorbar sampled line (style, no value)

A style which is used by colorbar sampled line to change the color of the line without ticks.
It is initially set to help lines.
4.9.14 Color Bars Outside Of an Axis

Occasionally, one has multiple adjacent plots, each with the same colormap and the same point meta min
and point meta max values and we’d like to show a single colorbar. pgfplots supports the colorbar to
name feature which is similar to the related method for legends, legend to name:
/pgfplots/colorbar to name={hnamei} (initially empty)

Enables to detach a colorbar from its parent axis: instead of drawing the colorbar, a self–contained,
independent set of drawing commands will be stored using the label hnamei. The label is defined using
\label{hnamei}, just as for any other LATEX label. The name can be referenced using
\ref{hnamei}.
Thus, typing \ref{hnamei} somewhere outside of the axis, maybe even outside of any picture, will
cause the colorbar to be drawn.
1
1 1
0.8 0.8
0.8
0.6 0.6
0.6
0.4 0.4
0.4 0.2 0.2
0.2 0
0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
0 0.2 0.4 0.6 0.8 1
\pgfplotsset{footnotesize,samples=10, domain=0:1,point meta min=0, point meta max=1}
\begin{center}% note that \centering uses less vspace...
\begin{tikzpicture}
\begin{axis}[colorbar,colorbar horizontal,colorbar to name={storedcolorbar}]
\addplot[scatter,only marks,mark=*] {rnd};
\end{axis}
\end{tikzpicture}
%
\begin{tikzpicture}
\begin{axis}
\addplot+[domain=0:1,mark=none,mesh] {x^2};
\end{axis}
\end{tikzpicture}
%
\begin{tikzpicture}
\end{axis}
\end{tikzpicture}
\\
\ref{storedcolorbar}
\end{center}
The feature works in the same way as described for legend to name, please refer to its description on
page 247 for the details. We only summarize the di↵erences here.
\pgfplotscolorbarfromname{hnamei}
This command poses an equivalent alternative for \ref{hnamei}: it has essentially the same e↵ect,
but it does not create links when used with the hyperref package.
/pgfplots/every colorbar to name picture (style, no value)

A style which is installed when \ref is used outside of a picture: a new picture will be created with
\tikz[/pgfplots/every colorbar to name picture].
See also the every legend to name picture style.
4.10 Scaling Options

There are a various options which control or change the scaling of an axis. Here, “scaling” typically means
two aspects: first, the unit vector size in each direction and second, the displayed limits in each direction.
Both together control the size of the axis box. In addition, axis descriptions change the size. However,
pgfplots scales only units and determines limits. It does not scale axis descriptions by default in order to
keep consistent font sizes between the text and the figure.
Often, one wishes to provide the target size only and let pgfplots do the rest. This is the default; it
scales the axis to width and height. If you provide one of these options, the axis will be rescaled while
keeping the aspect ratio.
Occasionally, one wants to provide unit vectors explicitly to ensure that one unit takes a prescribed
amount of space. This is possible by means of the x, y, and z keys. In such a case, the width and height
options will be ignored (or only applied for the unspecified unit vectors).
4.10. SCALING OPTIONS 271
Another common approach is to enforce specific unit vector ratios: for example by specifying that
each unit should take the same amount of space using axis equal. Here, pgfplots scales the lengths of
all vectors uniformly, i.e. it applies the same scale to each vector. This is done by means of the scale mode
configuration which is basically one of scale mode=stretch to fill or scale mode=scale uniformly:
pgfplots tries to satisfy the prescribed width and height arguments by finding a common scaling factor.
In addition, it attempts to enlarge the limits individually to fit into the prescribed dimensions.
In addition, you can use the option /pgfplots/scale to simply scale all final units up by some prescribed
factor. This does not change text labels.
If needed, you can also supply /tikz/scale to an axis. This will scale the complete resulting image,
including all text labels.
4.10.1 Common Scaling Options

All common options mentioned in the previous paragraphs are described here.
/pgfplots/width={hdimeni} (initially empty)

Sets the width of the final picture to {hdimeni}.
Any non-empty dimension like width=5cm sets the desired target width. Any TEX unit is accepted (like
200pt or 5in).
An empty value width={} means “use default width or rescale proportionally to height”. In this
case, pgfplots uses the value of \axisdefaultwidth as target quantity. However, if the height key
has been set, pgfplots will rescale the \axisdefaultwidth in a way which keeps the ratio between
\axisdefaultwidth and \axisdefaultheight. This allows to specify just one of width and height
and keep aspect ratios.
Consequently, if you specify just width=5cm but leave the default height={}, the scaling will respect
the initial aspect ratio.
The scaling a↵ects the unit vectors for x, y, and z. It does not change the size of text labels or axis
descriptions.
5
\begin{tikzpicture}
\begin{axis}[width=5cm]
\addplot {x};
0 \end{axis}
\end{tikzpicture}
5
5 0 5
Please note that pgfplots only estimates the size needed for axis- and tick labels. The estimate
assumes a fixed amount of space for anything which is outside of the axis box. This has the e↵ect that
the final images may be slightly larger or slightly smaller than the prescribed dimensions. However, the
fixed amount is always the same; it is set to 45pt. That means that multiple pictures with the same
target dimensions will have the same size for their axis boxes – even if the size for descriptions varies.
It is also possible to scale the axis box to the prescribed width/height. In that case, the total width will
be larger due to the axis descriptions. However, the axis box fills the desired dimensions exactly.
\begin{axis}[width=5cm,scale only axis]
\addplot {x};
\end{axis}
\end{tikzpicture}
0
5
6 4 2 0 2 4 6
Note: changing width and/or height changes only the unit vector sizes. In particular, it does not
change the font size for any axis description, nor does it change the default spacing between adjacent
tick labels. It is best-practice to use width/height for “small” changes, i.e. changes for which the font
size should remain the same. Consider using one of the styles normalsize, small, footnotesize, or
tiny which are described in Section 4.10.2 on page 282, and then change to your desired dimensions if
you need a di↵erent “quality” of scaling.
\axisdefaultwidth
This macro defines the default width. It is preset to 240pt.
This default width defines the aspect ratio which will be used whenever just one of width or height
is specified: the aspect ratio is the ratio between \axisdefaultwidth and \axisdefaultheight.
You can change it using
\def\axisdefaultwidth{10cm}
/pgfplots/height={hdimeni} (initially empty)

Works in the same way as width except that an empty value height={} defaults to “use either
\axisdefaultheight or scale proportionally if just width has been changed”.
\axisdefaultheight
This macro defines the default height. It is preset to 207pt.
See \axisdefaultwidth.
/pgfplots/scale only axis=true|false (initially false)

If scale only axis is enabled, width and height apply only to the axis rectangle. Consequently, the
resulting figure is larger that width and height (because of any axis descriptions). However, the axis
box has exactly the prescribed target dimensions.
If scale only axis=false (the default), pgfplots will try to produce the desired width including
labels, titles and ticks.
/pgfplots/x={hdimeni} (initially empty)

/pgfplots/y={hdimeni} (initially empty)
/pgfplots/z={hdimeni} (initially empty)
/pgfplots/x={(hx i,hyi)}
/pgfplots/y={(hx i,hyi)}
/pgfplots/z={(hx i,hyi)}
Allows to assign zero, one, two, or three of the target unit vectors.
In this context, a “unit vector” is a two–dimensional vector which defines the projection onto the canvas:
every logical plot coordinate (x, y) is drawn at the canvas position
 
exx e
x· + y · yx .
exy eyy
The unit vectors ex and ey determine the paper position in the current (always two dimensional) image.
For a standard three–dimensional axis, a plot coordinate (x, y, z) is drawn at
  
exx eyx e
x· +y· + z · zx .
exy eyy ezy
The initial setting assigns empty values to each of these keys, i.e. x={},y={},z={}. In this case,
pgfplots is free to choose these vectors as best. To this end, it uses width, height, scale mode, plot
box ratio, unit vector ratio, view, and the axis limits.
The key x={hdimeni} simply sets ex = (hdimeni, 0)T while y={hdimeni} sets ey = (0, hdimeni)T . Using
z={hdimeni} results in ez = (hdimeni, hdimeni)T . In this context, hdimeni is any TEX size like 1mm,
2cm or 5pt. Note that you should not use negative values for hdimeni (consider using x dir and its
variants to reverse axis directions).
\begin{tikzpicture}
6 \begin{axis}[x=1cm,y=1cm]
\addplot expression[domain=0:3] {2*x};
\end{axis}
5 \end{tikzpicture}
0 1 2 3
\begin{axis}[x=1cm,y=0.5cm,y dir=reverse]
2 \end{axis}
\end{tikzpicture}
4
6
0 1 2 3
Note that if you change the unit vector for just one direction, the other vector(s) will be chosen by
pgfplots – and scaled in order to fill the prescribed width and height as best as pgfplots can (but
see remarks for three–dimensional plots at the end of this key).
Height is deduced from height option \begin{tikzpicture}
\begin{axis}[x=1cm,
6 title=Height is deduced from height option]
\end{axis}
\end{tikzpicture}
4
0
0 1 2 3
The second syntax, x={(hx i,hyi)} sets ex = (hx i, hyi)T explicitly50 . The corresponding keys for y and
z work in a similar way. This allows to define skewed or rotated axes.
\begin{tikzpicture}
\begin{axis}[x={(1cm,0.1cm)},y=1cm]
6 \addplot expression[domain=0:3] {2*x};
\end{axis}
\end{tikzpicture}
5
0
2 3
0 1
\begin{tikzpicture}
\begin{axis}[
0 x={(5pt,1pt)},
y={(-4pt,4pt)}]
\addplot {1-x^2};
\end{axis}
20
0 5
5
Setting x and/or y for logarithmic axis will set the dimension used for 1 · e ⇡ 2.71828 (or whatever has
been set as log basis x).
Please note that it is not possible to specify x as argument to tikzpicture. The option
50 Please note that you need extra curly braces around the vector. Otherwise, the comma will be interpreted as separator for
the next key-value pair.

\begin{tikzpicture}[x=1.5cm]
\begin{axis}
...
\end{axis}
\end{tikzpicture}
does not have any e↵ect because an axis rescales its coordinates (see the width option).
Note that providing unit vectors explicitly usually causes pgfplots to ignore any other scaling options.
In other words: if you say y=0.1cm, pgfplots will use (0cm, 0.1cm) as y projection vector. However,
if you add scale mode=scale uniformly, you allow pgfplots to change the lengths of your vectors.
Of course, it will keep their relative directions and relative sizes. In this case, pgfplots will try to
determine a good common scaling factor and it will try to change the axis limits in order to fill the
prescribed width and height (see the documentation for scale mode for details).
Allow to rescale lengths \begin{tikzpicture}
\begin{axis}[
title=Allow to rescale \emph{lengths},
x={(0.1cm,-0.05cm)},
y=0.1cm,
z=0cm,
5 axis on top,
scale mode=scale uniformly,
]
\addplot3[surf,shader=interp] {x*y};
0 \end{axis}
5 \end{tikzpicture}
5 5
In the example above, pgfplots decided that it should only rescale units – at the expensive of the
width constraint.
Changes to font sizes: see also Section 4.10.2 if you want to change font sizes or the density of tick
labels in a simple way.
Explicit units for 3D axes: As of version 1.5, it is also possible to supply unit vectors to three-
dimensional axes. In this case, the following extra assumptions need to be satisfied:
1. If you want to control three-dimensional units, you need to provide all of x, y, and z keys. For
two–dimensional axes, it is also supported to supply just one of x or y.
2. Any provided three-dimensional unit vectors are assumed to form a right–handed coordinate system.
In other words: take your right hand, let the thumb point into the x direction, the index finger in
y direction and the middle finger in z direction. If that is impossible, the pgfplots output will
be wrong. The reason for this assumption is that pgfplots needs to compute the view direction
out of the provided units (see below).
Consider using x dir=reverse or its variants in case you want to reverse directions.
3. For three-dimensional axes, pgfplots computes a view direction out of the provided unit vectors.
The view direction is required to allow the z buffer feature (i.e. to decide about depths)51 .
This feature is used to for the \addplot3 graphics feature, compare the examples in Section 4.3.8 on
page 66.
Limitations: Unfortunately, skewed axes are not available for bar plots.
/pgfplots/xmode=normal|linear|log (initially normal)

/pgfplots/ymode=normal|linear|log (initially normal)
51 pgfplots provides a debug option called view dir={hx i}{hy i}{hz i} to override the view direction, should that ever be
interesting.
/pgfplots/zmode=normal|linear|log (initially normal)

Allows to choose between linear (=normal) or logarithmic axis scaling or logplots for each
x, y, z-combination.
Logarithmic plots use the current setting of log basis x and its variants to determine the basis (default
is e).
/pgfplots/x dir=normal|reverse (initially normal)

/pgfplots/y dir=normal|reverse (initially normal)
/pgfplots/z dir=normal|reverse (initially normal)
Allows to reverse axis directions such that values are given in decreasing order.
This key is documented in all detail on page 306.
/pgfplots/axis equal={htrue,falsei} (initially false)

Each unit vector is set to the same length while the axis dimensions stay constant. Afterwards, the size
ratios for each unit in x and y will be the same.
Axis limits will be enlarged to compensate for the scaling e↵ect.
2
0.5
0 0
0.5
2
0 2 4 6 0 2 4 6
\begin{tikzpicture}
\begin{axis}[axis equal=false,grid=major]
\addplot[blue] expression[domain=0:2*pi,samples=300] {sin(deg(x))*sin(2*deg(x))};
\end{axis}
\end{tikzpicture}
\hspace{1cm}
\begin{tikzpicture}
\begin{axis}[axis equal=true,grid=major]
\end{axis}
\end{tikzpicture}
2 2
10 10
5 5
10 10
8 8
10 10
100 101 102 103 104 10 3
10 1
101 103 105 107
\begin{tikzpicture}
\begin{loglogaxis}[axis equal=false,grid=major]
\addplot expression[domain=1:10000] {x^-2};
\end{loglogaxis}
\end{tikzpicture}
\hspace{1cm}
\begin{tikzpicture}
\begin{loglogaxis}[axis equal=true,grid=major]
\end{loglogaxis}
\end{tikzpicture}
\begin{tikzpicture}
\begin{axis}[axis equal,small,view={45}{35.26}]
\addplot3[mark=cube, blue, mark size=1cm]
1 coordinates {(0,0,0)};
\end{axis}
0 \end{tikzpicture}
1
2 2
0 0
2 2
The configuration axis equal=true is actually just a style which sets unit vector ratio=1 1 1,unit
rescale keep size=true.
/pgfplots/axis equal image={htrue,falsei} (initially false)

Similar to axis equal, but the axis limits will stay constant as well (leading to smaller images).
0.5
0.5 0.5
0
0.5
0 2 4 6 0 2 4 6
\begin{tikzpicture}
\begin{axis}[axis equal image=false,grid=major]
\end{axis}
\end{tikzpicture}
\hspace{1cm}
\begin{tikzpicture}
\begin{axis}[axis equal image=true,grid=major]
\end{axis}
\end{tikzpicture}
2 2
10 10
5 5
10 10
8 8
10 10
100 101 102 103 104 100 102 104
\begin{tikzpicture}
\begin{loglogaxis}[axis equal image=false,grid=major]
\end{loglogaxis}
\end{tikzpicture}
\hspace{1cm}
\begin{tikzpicture}
\begin{loglogaxis}[axis equal image=true,grid=major]
\end{loglogaxis}
\end{tikzpicture}
The configuration axis equal image=true is actually just a style which sets unit vector ratio=1 1
1,unit rescale keep size=false.
/pgfplots/unit vector ratio={hrx ry rz i} (initially empty)

Allows to provide custom unit vector ratios.
The key allows to tell pgfplots that, for example, one unit in x direction should be twice as long as
one unit in y direction:
\begin{tikzpicture}
\begin{axis}[unit vector ratio=2 1,small]
\addplot table[row sep=\\,col sep=&] {
x & y \\
0.5 0 & 1 \\
1 & 0 \\
};
0 \end{axis}
\end{tikzpicture}
0 0.2 0.4 0.6 0.8 1
Providing unit vector ratio=2 1 means that eexy = 2 where each coordinate (x, y) is placed at xex +
yey 2 R2 (see the documentation for x and y options). Note that axis equal is nothing but unit
vector ratio=1 1 1.
The arguments hrx i, hryi, and hrz i are ratios for x, y and z vectors, respectively. For two–dimensional
axes, only hrx i and hryi are considered; they are provided relative to the y axis. In other words: the x
unit vector will be hrx i / hryi times longer than the y unit vector. For three-dimensional axes, all three
arguments can be provided; they are interpreted relative to the z unit vector. Thus, a three dimensional
axis with unit vector ratio=1 2 4 will have an x unit which is 1/4 the length of the z unit, and a y
unit which is 2/4 the length of the z unit.
Trailing values of 1 can be omitted, i.e. unit vector ratio=2 1 is the same as unit vector ratio=2;
and unit vector ratio=3 2 1 is the same as unit vector ratio=3 2. An empty value unit vector
ratio={} disables unit vector rescaling.
Note that an active unit vector ratio will implicitly set scale mode=scale uniformly52 .
/pgfplots/unit vector ratio*={hrx ry rz i}

/pgfplots/unit rescale keep size=true|false|unless limits declared (initially unless
limits declared)
In the default configuration, pgfplots maintains the original axis dimensions even though unit
vector ratio involves di↵erent scalings.
It does so by enlarging the limits.
1
1 1
0
0 0
2 4
1 1 1 2
0.5 0 0
0 2 10 5
0.5 0 1 0 0 2
0.5 0.5 1 5 4
2 2 y 10 y
y
x x x
\begin{tikzpicture}
\begin{axis}[footnotesize,xlabel=$x$,ylabel=$y$,unit vector ratio=]
\addplot3[surf,z buffer=sort,samples=15,
domain=0:180, y domain=0:360]
({cos(u)*sin(v)}, {sin(u)*sin(v)}, {cos(v)});
\end{axis}
\end{tikzpicture}
\begin{tikzpicture}
\begin{axis}[footnotesize,xlabel=$x$,ylabel=$y$,unit vector ratio=1 1 1]
\end{axis}
\end{tikzpicture}
\begin{tikzpicture}
\begin{axis}[footnotesize,xlabel=$x$,ylabel=$y$,unit vector ratio=0.25 0.5]
\end{axis}
\end{tikzpicture}
The example above has the same plot, with three di↵erent unit ratios. The first has no limitations
(it is the default configuration). The second uses the same length for each unit vector and enlarges
the limits in order to maintain the same dimensions. The third example has an x unit which is 1/4
the length of a z unit, and an y unit which is 1/2 the length of a z unit.
pgfplots does its best to respect the involved scaling options (the prescribed width and height,
the unit vector ratio, and any specified axis limits). In the case above, it enlarged the horizontal
limits and kept the z limit as-is. See scale mode and its documentation for details about the
involved algorithm and its parameters.
The unit rescale keep size=false key, or, equivalently, unit vector ratio*=..., does not
enlarge limits:
52 This has been introduced in version 1.6. For older versions, the axis equal feature produced wrong results for three–
dimensional axes.
0 1 1
0 0
1 0.5 1
0 1 1.5
0.5 0 0.500.5 00.5 01 00.5
0.5
0.5 0.5 0.5
y y x y
x x
\begin{tikzpicture}
\begin{axis}[footnotesize,xlabel=$x$,ylabel=$y$,unit vector ratio=]
\end{axis}
\end{tikzpicture}
\begin{tikzpicture}
\begin{axis}[footnotesize,xlabel=$x$,ylabel=$y$,
unit rescale keep size=false,
unit vector ratio=1 1 1]
\end{axis}
\end{tikzpicture}
\begin{tikzpicture}
\begin{axis}[footnotesize,xlabel=$x$,ylabel=$y$,
unit vector ratio*=0.25 0.5, % the ’*’ implies ’unit rescale keep size=false’
]
\end{axis}
\end{tikzpicture}
The key unit rescale keep size also a↵ects scale mode=scale uniformly (which is closely
related to axis equal).
Here is the reference of the value of unit rescale keep size: the value true means that pgf-
plots will enlarge limits in order to keep the size. It will try to respect user provided limits, but if
the user provided all limits, it will override the user-provided limits and will rescale them. Thus,
true gives higher priority to the axis size than to user-provided limits. The choice false will never
rescale axis limits. The choice unless limits declared is a mixture: it will enlarge limits unless
the user provided them. If the user provides all limits explicitly, this choice is the same as false.
/pgfplots/x post scale={hscalei} (initially empty)

/pgfplots/y post scale={hscalei} (initially empty)
/pgfplots/z post scale={hscalei} (initially empty)
/pgfplots/scale={hscalei} (initially empty)
Lets pgfplots compute the axis scaling based on width, height, view, plot box ratio, axis equal
or explicit unit vectors with x, y, z and rescales the resulting vector(s) according to hscalei.
The scale key sets all three keys to the same huniform scalei value. This is e↵ectively the same as if
you rescale the complete axis (without changing sizes of descriptions).
The other keys allow individually rescaled axes.
0
5
5
6
6 4 2 0 2 4 6 6 4 2 0 2 4 6
\begin{tikzpicture}
\begin{axis}[y post scale=1]
\addplot {x};
\end{axis}
\end{tikzpicture}
\begin{tikzpicture}
\begin{axis}[y post scale=2]
\addplot {x};
\end{axis}
\end{tikzpicture}
Thus, the axis becomes larger. This overrules any previous scaling.
20
0
20
0
20
20 5 5
4 0 4 0
2 0 2 0
2 4 2 4
5 5
\begin{tikzpicture}
\begin{axis}[z post scale=1]
\end{axis}
\end{tikzpicture}
\begin{tikzpicture}
\begin{axis}[z post scale=2]
\end{axis}
\end{tikzpicture}
4.10.2 Scaling Descriptions: Predefined Styles

It is reasonable to change font sizes, marker sizes etc. together with the overall plot size: Large plots should
also have larger fonts and small plots should have small fonts and a smaller distance between ticks.
/tikz/font=\normalfont|\small|\tiny|. . .
/pgfplots/max space between ticks={hinteger i}
/pgfplots/try min ticks={hinteger i}
/tikz/mark size={hinteger i}
These keys should be adjusted to the figure’s dimensions. Use
\pgfplotsset{tick label style={font=\footnotesize},

legend style={font=\small}
}
to provide di↵erent fonts for di↵erent descriptions.

The keys max space between ticks and try min ticks are described on page 328 and configure the
approximate distance and number of successive tick labels (in pt). Please omit the pt suffix here.
There are a couple of predefined scaling styles which set some of these options:
/pgfplots/normalsize (style, no value)

Re-initialises the standard scaling options of pgfplots.
A “normalsize” figure
40
Leg
20
The y axis
20
6 4 2 0 2 4 6
The x axis
\begin{tikzpicture}
\begin{axis}[normalsize,
title=A ‘‘normalsize’’ figure,
xlabel=The $x$ axis,
ylabel=The $y$ axis,
minor tick num=1,
legend entries={Leg}]
\addplot {max(4*x,7*x)};
\end{axis}
\end{tikzpicture}
The initial setting is
\pgfplotsset{
normalsize/.style={
/pgfplots/width=240pt,
/pgfplots/height=207pt,
/pgfplots/max space between ticks=35
}
}
/pgfplots/small (style, no value)

Redefines several keys such that the axis is “smaller”.
A “small” figure \begin{tikzpicture}
\begin{axis}[small,
Leg title=A ‘‘small’’ figure,
20 ylabel=The $y$ axis,
minor tick num=1,
The y axis
legend entries={Leg}]
\addplot {x^2};
10 \end{axis}
\end{tikzpicture}
0
6 4 2 0 2 4 6
The x axis
\pgfplotsset{
small/.style={
width=6.5cm,
height=,
tick label style={font=\footnotesize},
max space between ticks=25,
}
}
Feel free to redefine the scaling – the option may still be useful to get more ticks without typing too
much. You could, for example, set small,width=6cm.
/pgfplots/footnotesize (style, no value)

Redefines several keys such that the axis is even smaller. The tick labels will have \footnotesize.
A “footnotesize” figure
\begin{tikzpicture}
10 Leg \begin{axis}[footnotesize,
8 title=A ‘‘footnotesize’’ figure,
The y axis
4 minor tick num=1,
2 legend entries={Leg}]
0 coordinates {
0 1 2 3 4 5 (0,0) (1,1) (3,3) (5,10)
};
The x axis
\end{axis}
\end{tikzpicture}
\pgfplotsset{
footnotesize/.style={
width=5cm,
height=,
legend style={font=\footnotesize},
tick label style={font=\footnotesize},
title style={font=\small},
every axis title shift=0pt,
every mark/.append style={mark size=8},
major tick length=0.1cm,
minor tick length=0.066cm,
},
}
As for small, it can be convenient to set footnotesize and set width afterwards.
You will need compat=1.3 or newer for this to work.
/pgfplots/tiny (style, no value)
Redefines several keys such that the axis is very small. Most descriptions will have \tiny as fontsize.
A “tiny” figure
\begin{tikzpicture}
10
Leg \begin{axis}[tiny,
The y axis
8
6
title=A ‘‘tiny’’ figure,
4
0 minor tick num=1,
0 1 2 3 4 5 legend entries={Leg}]
The x axis
coordinates {
(0,0) (1,1) (3,3) (5,10)
};
\end{axis}
\end{tikzpicture}
\pgfplotsset{
tiny/.style={
width=4cm,
height=,
legend style={font=\tiny},
tick label style={font=\tiny},
label style={font=\tiny},
title style={font=\footnotesize},
every mark/.append style={mark size=6},
major tick length=0.1cm,
minor tick length=0.066cm,
every legend image post/.append style={scale=0.8},
},
}
As for small, it can be convenient to use tiny,width=4.5cm to adjust the width.

You will need compat=1.3 or newer for this to work.
4.10.3 Scaling Strategies

The content of this section is quite involved – and its knowledge is typically unnecessary because by default,
pgfplots controls the involved stu↵ automatically. You may want to skip this section.
/pgfplots/scale mode=auto|none|stretch to fill|scale uniformly (initially auto)

Specifies how to choose the (individual) unit vector scaling factors, their length ratios, and perhaps the
axis limits in order to fill the prescribed width and height.
The scale mode implementation expects some “initial” set of unit vectors. This initial set of unit vectors
is determined as follows: for standard two–dimensional axes, it is simply the unit cube ex = (1pt, 0pt)T ,
ey = (0pt, 1pt)T , ez = 0. For three–dimensional axes, it is the outcome of the two keys view and plot
box ratio. If you provided units explicitly by means of one of x, y, or z, this value is the initial unit
vector.
In addition, it expects “initial” axis limits (i.e. values of xmin, xmax, etc.). The initial axis limits are those
limits which have been deduced from your data or which have been provided explicitly. Furthermore,
the initial axis limits already include changes of the enlargelimits key.
Given the initial set of unit vectors and the initial axis limits, the scale mode implementation is a kind
of “post–processor” which creates modified unit vectors and modified axis limits in order to satisfy all
specified constraints. These constraints are width, height, and unit vector ratio.
The initial choice auto tells pgfplots to take full control over this key. It chooses one of the other
possible choices depending on the actual context. The choice auto evaluates to scale uniformly if
unit vector ratio is set. Otherwise it evaluates to stretch to fill.
The choice none does not apply any rescaling at all. Use this if prescribed lengths of x, y (and perhaps
z) should be used. In other words: it ignores width and height. In this case, you may want to set x
post scale and its variants to rescale units manually. See also disabledatascaling.
The choice stretch to fill takes the initial unit vectors and rescales the unit vectors with two separate
scales: one which results in the proper width and one which results in the proper height. As a
consequence, the unit vectors are modified and distorted such that the final image fits into the prescribed
dimensions. This is usually what one expects unless one provides unit directions explicitly. This mode
does not change axis limits. Note that if one of the unit vectors has been provided explicitly, pgfplots
will not change it. It will only change the remaining axis limits. This mode contradicts axis equal or
unit vector ratio.
The choice scale uniformly takes the initial unit vectors and applies only one scaling factor to all
units. In this case, there is just one common scaling factor for both width and height. Naturally,
this will result in unsatisfactory results because either the final width or the final height will not be
met. Therefore, this choice will adjust axis limits to get the desired dimensions. Thus, the unit vectors
have exactly the same size relations and angles as they had before the scaling; only their magnitude
is changed uniformly. In addition, axis limits may be changed (with individual scaling factors for each
axis limit). Note that if unit vectors have been provided explicitly, pgfplots can still rescale it with
this choice – it will keep the relative directions and size ratios. The choice scale uniformly tries its
best to modify the degrees of freedom in a “useful” way. The precise meaning of “useful” is the scale
uniformly strategy key.
/pgfplots/scale uniformly strategy=auto|units only|change vertical limits|

change horizontal limits (initially auto)
The scale uniformly method requires to determine one common scaling factor which rescales
every axis unit. In addition, it allows one scaling factor for each axis limit, i.e. up to three.
The constraints for this search are that we want to satisfy the width/height constraint, have as
few rescaling as possible and that we do not want to reduce limits (as this could possibly hide data
points).
The choice auto chooses one of the other possibilities automatically. Depending on whether we have
two dimensions or three dimensions, it compares the available methods and chooses the one which
does not reduce limits and which involves the fewest rescaling (i.e. it may compare the outcome of
the other strategies). This is the default. If you keep the choice auto, you do not have to worry
about the remaining choices. Note that manually provided axis limits will not be modified.
The choice units only will not enlarge axis limits. It will only rescale the units. To this end,
it chooses the scaling factor such that the smaller target dimension is filled as desired. In other
words: if width < height, it will scale to satisfy the width constraint. The height constraint will
be ignored. The case > will be done the other way round. The choice units only typically results
in a square axis as it takes the initial set of unit vectors (which are typically the unit box) and
scales them with a common scaling factor. Consequently, you can choose units only if you want
a boxed axis. You can still change axis limits manually, however.
The choice change vertical limits chooses a common scaling factor for the unit vectors on order
to satisfy the width (!) constraint. This common scaling factor is similar to units only – but
units only can also decide to satisfy the height constraint whereas change vertical limits will
scale unit vectors to satisfy width. In order to satisfy the height constraint, change vertical
limits modifies just the vertical limits. For two–dimensional axes, this is ymin and ymax. For
three–dimensional axes, this is zmin and zmax. Clearly, there is a chance that it will decrease the
displayed range – in this case, parts of the image will be clipped away. This method assumes that
the vertical axis has not been rotated (i.e. that eyx = 0 or ezx = 0, respectively). It refuses to work
and falls back to units only for rotates axes. Choose change vertical limits if you want the
image (i.e. the actual content) as wide as possible. You can modify width and height to improve
its outcome. Note that manually specified axis limits will not be changed, see below for details.
The choice change horizontal limits attempts a similar approach, but for the horizontal limits:
it determines one suitable scaling factor which is applied to all unit vectors and modifies horizontal
axis limits to satisfy the remaining constraints. For two–dimensional axes, this is quite simple
because we typically have exy = 0 (i.e. the x unit vector has vanishing y component) and eyx = 0
such that pgfplots can change axis limits easily. If a two–dimensional axis has an x unit with
exy 6= 0, the method is not applicable and falls back to units only. For three–dimensional axes,
it assumes that the z vector is not rotated, i.e. ezy = 0 and tries to change limits for both x and y.
This choice is much more involved because here, x and y components are coupled. Consequently,
the common unit scaling factor and the two involved axis limit compensation factors for x and y are
tightly coupled as well. pgfplots solves a system of non–linear equations iteratively to arrive at
a suitable solution for all three scalings. Use this method if change vertical limits would clip
away parts of the image (because it reduced the displayed range) and you do not want to change
width and height. The choice change horizontal limits will typically result in more empty
space in the resulting figure. But it will not clip away content. Manually specified axis limits will
not be changed, see below for details.
Manually provided axis limits: Any manually provided arguments for xmin and its variants are
considered to be immutable; pgfplots will not change them. If you assign xmin, pgfplots will only
change xmax and vice–versa. If you assign both xmin and xmax, pgfplots will not change x limits at
all. Note that if you assign both xmin and xmax, pgfplots will simply skip the scaling and will give up
on the constraints. It will not try to compensate the lack of scaling opportunities by changing y limits,
for example. This has the positive e↵ect that assigning limits does not change the complete appearance
of your axis. The allowed set of changes to axis limits can be configured with the following key.
Interaction with enlargelimits: Note that enlargelimits and scale mode are independent of
another: the outcome of enlargelimits is used as “initial axis limits” and these limits may be changed
by scale mode (even if you said enlargelimits=false). See the documentation of enlargelimits for
details on this interaction.
/pgfplots/unit rescale keep size=true|false|unless limits declared (initially unless

limits declared)
In the default configuration unless limits declared, unit rescaling may cause changes to the
axis limits in order to keep the figure’s size intact. However, only those limits which have not been
declared manually are subject to rescaling: if you say xmin=1, only xmax and the limits for y and
z are free to change.
Setting unit rescale keep size=false will disable the modification of axis limits altogether, i.e.
axis limits will not be rescaled to compensate scalings on unit vectors.
4.11. 3D AXIS CONFIGURATION 287
Setting unit rescale keep size=true will always rescale limits, even if they have been declared
manually.
This key mainly a↵ects scale mode=scale uniformly. This, in turn, is used for axis equal and
\addplot3 graphics.
See also the addition documentation for this key and related examples on page 279.
The scale uniformly choice is implicitly used for axis equal and for the \addplot3 graphics fea-
ture, see the documentation in Section 4.3.8 on page 66 for its examples. Note that the common case is
that the initial unit vectors form the unit cube (i.e. those before scaling, see above). In this case, scale
uniformly is the same as axis equal.
4.11 3D Axis Configuration

This section described keys which are used to configure the appearance of three dimensional figures. Some of
them apply for two–dimensional plots as special case as well, and they will also be discussed in the respective
sections of this manual.
4.11.1 View Configuration

/pgfplots/view={hazimuthi}{helevationi} (initially {25}{30})
Changes both view angles of a 3D axis. The azimuth (first argument) is the horizontal angle which is
rotated around the z axis. For a 3D plot, the z axis always points to the top. The elevation (second
argument) is the vertical rotation around the (rotated) x axis. Positive elevation values indicate a view
from above, negative a view from below. All values are measured in degree (but see trig format).
View along the positive y axis \begin{tikzpicture}
\begin{axis}[view={0}{0},
5 xlabel=$x$,
zlabel=$z$,
title=View along the positive $y$ axis]
\addplot3[surf] {x};
\end{axis}
\end{tikzpicture}
0
z
5
4 2 0 2 4
x
View from top \begin{tikzpicture}
\begin{axis}[view={0}{90},
4 xlabel=$x$,
ylabel=$y$,
title=View from top]
2 \addplot3[surf] {x};
\end{axis}
\end{tikzpicture}
0
y
4
4 2 0 2 4
x
\begin{tikzpicture}
\begin{axis}[view={-45}{45},
xlabel=$x$,ylabel=$y$,zlabel=$z$]
5 \addplot3[surf] {x};
\end{axis}
\end{tikzpicture}
0
z
5
5 5
0 0
y 5 5 x
The view is computed as follows. The view is defined by two rotations: the first rotation uses the
hazimuthi angle to rotate around the z axis. Afterwards, the view is rotated helevationi degrees around
the rotated x axis (more precisely, it is rotated helevationi degrees). The resulting transformed x–z–
plane is the viewport, i.e. the view direction is always the transformed positive y axis.
The view argument is compatible with the argument of the Matlab (®) view command, i.e. you can
use
[h,v] = view
in matlab and pack the resulting arguments into pgfplots53 .
If you work with gnuplot, you can convert the view arguments as follows: the gnuplot command
set view v,h
is equivalent to view={h}{90-v}. For example, the default gnuplot configuration set view 60,60 is
equivalent to view={60}{30} in pgfplots.
The view is (currently) always an orthogonal projection, no perspective is possible, yet. You can,
however, specify projection unit vectors for x, y, and z explicitly to get a skewed three–dimensional
axis.
/pgfplots/view/az={hazimuthi}
/pgfplots/view/h={hazimuthi} (initially 25)
Changes only the azimuth view angle, i.e. the horizontal (first) view angle which is rotated around the z
axis.
\begin{tikzpicture}
\begin{axis}[view/h=-30]
1
\addplot3[
surf,
%shader=interp,
0 shader=flat,
samples=50,
domain=-3:3,y domain=-2:2]
1 {sin(deg(x+y^2))};
2 \end{axis}
\end{tikzpicture}
0 2
0
2 2
53 In case it does not work, try h and -v in pgfplots.

\begin{tikzpicture}
\begin{axis}[view/h=10]
\addplot3[
surf,
1 %shader=interp,
shader=flat,
samples=50,
0 {sin(deg(x+y^2))};
2
\end{axis}
\end{tikzpicture}
0
1
2 0 2 2
\begin{tikzpicture}
\begin{axis}[view/h=40,colormap/violet]
\addplot3[
1 surf,
%shader=interp,
shader=flat,
0 samples=50,
{sin(deg(x+y^2))};
1 \end{axis}
2 \end{tikzpicture}
2
0 0
2
2
\begin{tikzpicture}
1 \begin{axis}[view/h=70]
\addplot3[
surf,
0 %shader=interp,
shader=flat,
samples=50,
1 domain=-3:3,y domain=-2:2]
{sin(deg(x+y^2))};
\end{axis}
2 \end{tikzpicture}
0
1 2
2 1 0
2
/pgfplots/view/el={helevationi}
/pgfplots/view/v={helevationi} (initially 30)
Changes only the vertical elevation, i.e. the second argument to view. Positive values view from above,
negative values from below.
4.11.2 Styles Used Only For 3D Axes

/pgfplots/every 3d description (style, no value)
This style allows to change the appearance of descriptions for three dimensional axes. Naturally, a
three dimensional axis will display axis labels for x and y di↵erently than a two dimensional axis (for
example, the y axis label won’t be rotated by 90 degrees). The every 3d description style installs
the necessary display options for three dimensional axis descriptions.
The initial value is:
\pgfkeys{
/pgfplots/every 3d description/.style={
% Only these description styles can be changed here:
every axis x label/.style={at={(ticklabel cs:0.5)},
anchor=near ticklabel},
every axis y label/.style={at={(ticklabel cs:0.5)},
anchor=near ticklabel},
every x tick scale label/.style={
at={(xticklabel cs:0.95,5pt)},
anchor=near xticklabel,inner sep=0pt},
every y tick scale label/.style={
at={(yticklabel cs:0.95,5pt)},
anchor=near yticklabel,inner sep=0pt},
try min ticks=3,
}%
}
As the name suggests, every 3d description can only be used to set styles for axis labels, tick labels
and titles. It has not been designed to reset other styles, you will need to change these options either for
each axis separately or by means of user defined styles. The reason for this limitation is: other options
can (and, in many cases, needs to) be set before the axis is processed. However, the decision whether
we have a two dimensional or a three dimensional axis has to be postponed until the processing is more
or less complete – so only some remaining keys can be set.
/pgfplots/every 3d view {hh i}{hv i} (style, no value)

A style which can be used for fine-tuning of the output for specific views.
This style will be installed right after every 3d description, but before other axis description related
keys are set (in other words: it has higher precedence than every 3d description, but lower precedence
than keys provided to the axis directly).
One example is preconfigured for view={0}{90} (from top):
\pgfplotsset{
/pgfplots/every 3d view {0}{90}/.style={
xlabel near ticks,
ylabel near ticks,
axis on top=true
}
}
4.11.3 Appearance Of The 3D Box

/pgfplots/plot box ratio={hhx stretchi hy stretchi hz stretchii} (initially 1 1 1)
Allows to customize the aspect ratio between the three di↵erent axes in a three dimensional plot.
Note that this key is di↵erent from the related unit vector ratio: the plot box is only useful for
three dimensional axes, and it will usually distort the unit vector ratios. If you want equal unit ratios,
consider using unit vector ratio.
The plot box ratio is applied before any rotations and stretch–to–fill routines have been invoked.
Thus, the initial setting54 1 1 1 makes all axes equally long before the stretch–to–fill routine is applied.
54 Note that you can also use the syntax {1}{1}{1} instead of space-separation.
Initial plot box ratio \begin{tikzpicture}
\begin{axis}[
view/h=60,
plot box ratio=1 1 1,
2 colormap={violet}{[1cm] rgb255(0cm)=(25,25,122)
color(1cm)=(white) rgb255(5cm)=(238,140,238)},
p(x, t)
xlabel=$x$,
1 ylabel=$t$,
zlabel={$p(x,t)$},
shader=faceted,
0 title=Initial \texttt{plot box ratio},
]
5 \addplot3[surf,y domain=0.02:3.5,samples=81]
0 {1/(2*sqrt(pi*y)) * exp(0-x^2/y)};
2 % the ’0’ is a work-around for a bug in PGF 2.00
x 5 \end{axis}
t \end{tikzpicture}
plot box ratio=1 2 1
\begin{tikzpicture}
\begin{axis}[
view/h=60,
plot box ratio=1 2 1,
colormap={violet}{[1cm] rgb255(0cm)=(25,25,122)
2 color(1cm)=(white) rgb255(5cm)=(238,140,238)},
xlabel=$x$,
p(x, t)
ylabel=$t$,
1 zlabel={$p(x,t)$},
shader=flat,
title=\texttt{plot box ratio=1 2 1},
0
]
5 3 \addplot3[surf,y domain=0.02:3.5,samples=81]
0 2 {1/(2*sqrt(pi*y)) * exp(0-x^2/y)};
1 % the ’0’ is a work-around for a bug in PGF 2.00
x 5 t \end{axis}
\end{tikzpicture}
This key applies only to three dimensional axes. After the scaling, the axes will be stretched to fill
the width and height for this plot. Thus, the e↵ects of plot box ratio might be undone by this
stretching for particular views.
/pgfplots/3d box=background|complete|complete* (initially background)

Allows to configure the appearance of boxed three dimensional axes.
Type only 3d box (without value) as alias for 3d box=complete.
The choice background is the initial setting, it does not draw axis lines (and grid lines) which are in
the foreground.
3d box=background \begin{tikzpicture}
\begin{axis}[
3d box=background,
% pretty printing, but irrelevant:
title={3d box=background},
samples=5,
domain=-4:4,
10 xtick=data,
ytick=data,
0 ]
10 4 \end{axis}
2 \end{tikzpicture}
4 0
2 0 2
2 4 4
The choice complete also draws axis lines and tick lines in the foreground, but it doesn’t draw grid
lines in the foreground. The result yields a complete box:
3d box=complete \begin{tikzpicture}
\begin{axis}[
3d box,% same as 3d box=complete
% pretty printing, but irrelevant:
title={3d box=complete},
samples=5,
domain=-4:4,
10 xtick=data,
ytick=data,
0 ]
10 4 \end{axis}
2 \end{tikzpicture}
4 0
2 0 2
2 4 4
Finally, the choice complete* is the same as complete, but it also draws grid lines in the foreground.
3d box=complete 3d box=complete*
10 10
0 0
10 4 10 4
2 2
4 0 4 0
2 0 2 2 0 2
2 4 2 4
4 4
\begin{tikzpicture}
\begin{axis}[
3d box=complete,
grid=major,
title={3d box=complete},
samples=5, domain=-4:4,
xtick=data, ytick=data,
]
\end{axis}
\end{tikzpicture}%
~
\begin{tikzpicture}
\begin{axis}[
3d box=complete*,
grid=major,
title={3d box=complete*},
]
\end{axis}
\end{tikzpicture}
Before any foreground parts are actually processed, the style every 3d box foreground will be in-
stalled. This allows to change the appearance of foreground axis components like tick style or axis
line style separately from the background components.
Note that 3d box=complete is only available for boxed axes, i.e. together with axis lines=box. It is
an error to use a di↵erent combination.
4.12. ERROR BARS 293
4.11.4 Axis Line Variants

Three dimensional axes also benefit from the axis lines=box or axis lines=center styles discussed
in Section 4.9.9. The choice axis lines=box is standard, it draws a box (probably a↵ected by the 3d
box=complete key). The choice axis lines=center draws all three axes such that they pass through the
origin. It might be necessary to combine this key with axis on top as there is no depth information.
% Attention: use compat=1.8 or higher
% to repair label positions
2 \begin{axis}[
4 2 axis lines=center,
2 2 4 axis on top,
4
ztick=\empty, % no z ticks here
]
\end{axis}
\end{tikzpicture}
The remaining choices axis lines*=left and axis lines*=right select di↵erent sets of axes in a way
such that tick labels and axis label won’t disturb the plot’s content. The ‘*’ suppresses the use of special
styles which are mainly adequate for two-dimensional axes, see the documentation of axis lines. Such a
set of axes is always on the boundary of the two-dimensional projection.
The choice axis lines*=left chooses a set of axes which are on the left (or bottom, respectively)
whereas the choice axis lines*=right chooses a set of axes which are on the right (or top, respectively):
\begin{tikzpicture}
\begin{axis}[
10 axis lines*=left,
0 xtick=data, ytick=data,
]
10 4 \addplot3[surf] {x*y};
2 \end{axis}
0 \end{tikzpicture}
4 2 0 2 2
4 4
2 0 2 \begin{axis}[
0 4 axis lines*=right,
2
4 samples=5, domain=-4:4,
10 xtick=data, ytick=data,
]
0 \addplot3[surf] {x*y};
\end{axis}
It is not possible to mix di↵erent styles like axis x line=center,axis z line=top.
4.12 Error Bars

An error bar is used to indicate the reliability of a data point. Typically, a data point is just (x, y). The
reliability would be indicated by additional values, i.e. by means of an error bound ✏x which characterizes the
di↵erence between the coordinate x provided in the plot data and the precise value x̃ (which is unknown).
The reliability can be indicated for both x and y independently (although y might be the typical candidate).
Error bounds can be expressed as absolute errors, i.e. of the form
|x x̃|  ✏x , |y ỹ|  ✏y
where x̃ and ỹ are the (unknown) precise values and x and y are the actual values of the plot. However,
they can also be provided relative to the input values, i.e. of the form
|x x̃| |y ỹ|
 ✏x ,  ✏y .
|x̃| |ỹ|
A relative error of 10% would result in an error value of 0.1 (relative to the precise quantity ỹ). Clearly,
relative errors are only useful if the precise value if not zero, i.e. x̃, ỹ 6= 0.
pgfplots allows to provide the “error values” for each coordinate independently. Thus, it may find
some value ✏x and/or ✏y . Depending on the configuration, it interprets the encountered value as absolute or
relative error. The error value can be the same for every coordinate, for example if you know that each y
coordinate has a fixed error of 10%. The error value can also be di↵erent for every coordinate in which it
is said to be “explicitly provided”. In fact, pgfplots also features asymmetric error values, i.e. the lower
bound on the error can be di↵erent from the upper bound. Thus, a two–dimensional data point (x, y) can
have up to four distinct error values which have to be provided by the end–user.
Thus, the end–user has to provide all needed error values and a configuration to express if these values
are to be interpreted as relative or absolute error and if the values are to be expected explicitly for every
data point or if they are fixed.
\begin{tikzpicture}
1 \begin{axis}
\addplot+[error bars/.cd,
y dir=plus,y explicit]
coordinates {
(0,0) +- (0.5,0.1)
0.5 (0.1,0.1) +- (0.05,0.2)
(0.2,0.2) +- (0,0.05)
(0.5,0.5) +- (0.1,0.2)
(1,1) +- (0.3,0.1)};
\end{axis}
0 \end{tikzpicture}
0 0.2 0.4 0.6 0.8 1
The preceding example has two keys: y dir=plus configures pgfplots to activate error bars for y
coordinates, but only upper bounds. The key y explicit tells pgfplots to expect absolute values in the
input data stream. In our case above, the input data stream is an \addplot coordinates which uses the
special error–value–syntax +- (✏x , ✏y ), see the Section 4.12.1 for details.
It is allowed if the input data contains more error values than needed: our example above has error values
for both x and y and it also contains lower bounds (since +- defines upper- and lower bounds simultaneously).
Consequently, the remaining values can be visualized as well:
\begin{tikzpicture}
1 \begin{axis}
y dir=both,y explicit,
x dir=both,x explicit,
]
0.5
coordinates {
(0,0) +- (0.5,0.1)
(0.1,0.1) +- (0.05,0.2)
(0.2,0.2) +- (0,0.05)
0 (0.5,0.5) +- (0.1,0.2)
(1,1) +- (0.3,0.1)};
\end{axis}
0.5 0 0.5 1 \end{tikzpicture}
Error bars inherit all drawing options of the associated plot, but they use their own error mark and
additional style arguments.
/pgfplots/error bars/x dir=none|plus|minus|both (initially none)

/pgfplots/error bars/y dir=none|plus|minus|both (initially none)
/pgfplots/error bars/z dir=none|plus|minus|both (initially none)

The initial configuration none draws no error bars at all in the provided direction.
The configuration plus draws only upper bounds in the direction of interest.
The configuration minus draws only lower bounds in the direction of interest.
The configuration both draws upper and lower bounds in the direction of interest.
In every case, the actual error value and its character (absolute or relative) is to be determined by other
options (see below). If, for some reason, the error value is missing, the error bar is omitted.
/pgfplots/error bars/x fixed={hvaluei} (initially 0)

/pgfplots/error bars/y fixed={hvaluei} (initially 0)
/pgfplots/error bars/z fixed={hvaluei} (initially 0)
Provides a common, absolute error ✏x = hvaluei for all input coordinates.
\begin{tikzpicture}
1 \begin{axis}
y dir=both,y fixed=0.1,
]
coordinates {
0.5
(0,0)
(0.1,0.1)
(0.2,0.2)
(0.5,0.5)
0 (1,1)
};
\end{axis}
0 0.2 0.4 0.6 0.8 1 \end{tikzpicture}
For linear x axes, the error mark is drawn at x±✏x while for logarithmic x axes, it is drawn at log(x±✏x ).
/pgfplots/error bars/x fixed relative={hpercenti} (initially 0)

/pgfplots/error bars/y fixed relative={hpercenti} (initially 0)
/pgfplots/error bars/z fixed relative={hpercenti} (initially 0)
Provides a common, relative error ✏x = hpercenti · x for all input coordinates. The argument hpercenti
is thus given relatively to input x coordinates such that hpercenti = 1 means 100%.
Error marks are thus placed at x · (1 ± ✏x ) for linear axes and at log(x · (1 ± ✏x )) for logarithmic axes.
Computations are performed in floating point for linear axis and using the identity log(x · (1 ± ✏x )) =
log(x) + log(1 ± ✏x ) for logarithmic scales.
The following example shows that fixed error values ✏x are independent of the input values.
1
\begin{tikzpicture}
\begin{axis}[enlargelimits=false]
0.8 \addplot[red,mark=*]
plot[error bars/.cd,
y dir=minus,y fixed relative=1,
0.6 x dir=minus,x fixed relative=1,
error mark=none,
error bar style={dotted}]
0.4
coordinates
{(0,0) (0.1,0.1) (0.2,0.2)
0.2 (0.5,0.5) (1,1)};
\end{axis}
\end{tikzpicture}
0
0 0.2 0.4 0.6 0.8 1
/pgfplots/error bars/x explicit (no value)

/pgfplots/error bars/y explicit (no value)
/pgfplots/error bars/z explicit (no value)
Configures the error bar algorithm to draw x-error bars at any input coordinate for which user-specified
errors are available. Each error is interpreted as absolute error, see x fixed for details.
The di↵erent input formats of errors are described in Section 4.12.1.
/pgfplots/error bars/x explicit relative (no value)

/pgfplots/error bars/y explicit relative (no value)
/pgfplots/error bars/z explicit relative (no value)
Configures the error bar algorithm to draw x-error bars at any input coordinate for which user-specified
errors are available. Each error is interpreted as relative error, that means error marks are placed at
x(1 ± hvaluei(x)) (works as for error bars/x fixed relative).
\begin{tikzpicture}
105 \begin{loglogaxis}
x dir=both,x fixed relative=0.5,
104 y dir=both,y explicit relative,
]
table[x=x,y=y,y error=error]
103 {
x y error
32 32 0
102 64 64 0
128 128 0.3
1024 1024 0.2
101 102 103 104 105 32068 32068 0.6
64000 64000 0.6
128000 128000 0.6
};
\end{loglogaxis}
\end{tikzpicture}
/pgfplots/error bars/error mark=hmarker i

Sets an error marker for any error bar. {hmarker i} is expected to be a valid plot mark, see Section 4.7.
\begin{tikzpicture}
1 \pgfmathsetseed{42}
\begin{axis}[
ymajorgrids,
]
\addplot+[
0.5 domain=0:180,
samples=11,
error bars/.cd,
y dir=both, y fixed=0.1,
0 error mark=diamond*,
]
{sin(x) + rand*0.1};
0 50 100 150 \end{axis}
\end{tikzpicture}
/pgfplots/error bars/error mark options={hkey-value-listi}

Sets a key-value list of options for any error mark. This option works similarly to the Tik Z ‘mark
options’ key.
/pgfplots/error bars/error bar style={hkey-value-listi}

Appends the argument to ‘/pgfplots/every error bar’ which is installed at the beginning of every
error bar.
/pgfplots/error bars/draw error bar/.code 2 args={h... i}

Allows to change the default drawing commands for error bars. The two arguments are
• the source point, (x, y) and
• the target point, (x̃, ỹ).
Both are determined by pgfplots according to the options described above. The default code is
\pgfplotsset{
/pgfplots/error bars/draw error bar/.code 2 args={%
\pgfkeysgetvalue{/pgfplots/error bars/error mark}%
{\pgfplotserrorbarsmark}%
\pgfkeysgetvalue{/pgfplots/error bars/error mark options}%
{\pgfplotserrorbarsmarkopts}%
\draw #1 -- #2 node[pos=1,sloped,allow upside down] {%
\expandafter\tikz\expandafter[\pgfplotserrorbarsmarkopts]{%
\expandafter\pgfuseplotmark\expandafter{\pgfplotserrorbarsmark}%
\pgfusepath{stroke}}%
};
}
}
4.12.1 Input Formats of Error Coordinates

Error bars with explicit error estimations for single data points require some sort of input format. This
applies to error bars/x explicit and error bars/x explicit relative.
Error Coordinates and Coordinate Lists

Error bar coordinates can be read from ‘\addplot coordinates’ in which they are expected after data point
as such:
(1,2) +- (0.4,0.2)
(2,4) +- (1,0)
(3,5)
(4,6) +- (0.3,0.001)
}
where (1, 2) ± (0.4, 0.2) is the first coordinate, (2, 4) ± (1, 0) the second and so forth. The point (3, 5) has
no error coordinate. The syntax +- defines symmetric error values, i.e. both upper and lower bound receive
the same value.
Alternatively, one can use one of -= and += to define asymmetric values:
\begin{tikzpicture}
4 \begin{axis}
\addplot+[
error bars/.cd,
3 x dir=both, x explicit,
y dir=both, y explicit,
]
2 coordinates {
(1.1,0.9) += (0.4,0.2) -= (0.1,0.1)
(2.7,2) -= (1,0)
1 (3,3)
(3.8,4.2) +- (0.3,0.2)
};
1 2 3 4 \end{axis}
\end{tikzpicture}
If multiple items (like multiple +=) for one coordinate are specified, the last one takes precedence.
Keep in mind that these error values are only displayed as error bars if x dir and y dir are set appro-
priately.
The input type \addplot coordinates also allows point meta=explicit, i.e. values of the form
\addplot coordinates {(0,0) [4]};
This can be combined with error values. However, the point meta value in square brackets needs to be
the last item:
\addplot coordinates {(0,0) +- (0.1,0.2) [4]};
Error Coordinates and Table Input

The ‘\addplot table’ format is
/pgfplots/table/x error={hcolumn namei}
/pgfplots/table/y error={hcolumn namei}
/pgfplots/table/z error={hcolumn namei}
/pgfplots/table/x error index={hcolumn index i}
/pgfplots/table/y error index={hcolumn index i}
/pgfplots/table/z error index={hcolumn index i}
/pgfplots/table/x error expr={hmath expressioni}
/pgfplots/table/y error expr={hmath expressioni}
/pgfplots/table/z error expr={hmath expressioni}
These keys define input sources for error bars with explicit error values.
The x error method provides an input column name (or alias), the x error index method provides
input column indices and x error expr works just as table/x expr: it allows arbitrary mathematical
expressions which may depend on any number of table columns using \thisrow{hcol namei}.
\begin{tikzpicture}
4 \begin{axis}
\addplot+[
error bars/.cd,
x dir=both, x explicit,
]
2 table[y error=error]
{
x y error
1 0.9 0.4
2 2.1 0.2
3 3 0.1
1 2 3 4 4 4.2 0.3
};
\end{axis}
\end{tikzpicture}
In addition, one can provide column indices using
\addplot table[x error index=COLINDEX,y error index=COLINDEX]
These options are used like the ‘x’ and ‘x index’ options.
If you need to specify math expressions, you can use x error expr:
\addplot table[x error expr=\thisrow{errorx}^2]
This is similar to x expr.
/pgfplots/table/x error plus={hcolumn namei}

/pgfplots/table/y error plus={hcolumn namei}
/pgfplots/table/z error plus={hcolumn namei}
/pgfplots/table/x error plus index={hcolumn index i}
/pgfplots/table/y error plus index={hcolumn index i}
/pgfplots/table/z error plus index={hcolumn index i}
/pgfplots/table/x error plus expr={hmath expressioni}
/pgfplots/table/y error plus expr={hmath expressioni}
/pgfplots/table/z error plus expr={hmath expressioni}
/pgfplots/table/x error minus={hcolumn namei}
/pgfplots/table/y error minus={hcolumn namei}
/pgfplots/table/z error minus={hcolumn namei}
/pgfplots/table/x error minus index={hcolumn index i}
/pgfplots/table/y error minus index={hcolumn index i}
/pgfplots/table/z error minus index={hcolumn index i}
/pgfplots/table/x error minus expr={hmath expressioni}
4.13. NUMBER FORMATTING OPTIONS 299
/pgfplots/table/y error minus expr={hmath expressioni}

/pgfplots/table/z error minus expr={hmath expressioni}
These keys define input sources for error bars with asymmetric error values, i.e. di↵erent values for
upper and lower bounds.
They are to be used in the same way as x error. In fact, x error is just a style which sets both x
error plus and x error minus to the same value.
\begin{tikzpicture}
4 \begin{axis}
\addplot+[
error bars/.cd,
3 x dir=both, x explicit,
]
2 table[
x error plus=ex+,
x error minus=ex-,
1 y error plus=ey+,
y error minus=ey-,
] {
1 2 3 4 x y ex+ ey+ ex- ey-
1.1 0.9 0.4 0.2 0.1 0.1
2.7 2 0 0 1 0
3 3 0 0 0 0
3.8 4.2 0.3 0.2 0.3 0.2
};
\end{axis}
\end{tikzpicture}
4.13 Number Formatting Options

pgfplots typesets tick labels rounded to given precision and in configurable number formats. The command
to do so is \pgfmathprintnumber; it uses the current set of number formatting options. In addition,
pgfplots might prepare tick numbers before they are handed over to \pgfmathprintnumber.
The options related to number printing as such are described in all detail in the manual for Pgfplot-
sTable, which comes with pgfplots. This section contains the reference for everything which is specific
to an axis, and only a brief survey over the number formatting options as such.
4.13.1 Frequently Used Number Printing Settings

This section provides a brief survey about the most frequently used aspects of number formatting in pgf-
plots.
1. pgfplots computes common tick scaling factors like ·102 and produces only integers as tick labels.
3
In order to get numbers like 0.001 as tick labels instead of 1 with a separate label ·10 , you can use
scaled ticks=false in your axis. See the description of scaled ticks for details.
2. In order to customize the way numbers are rounded and/or displayed, use something like xticklabel
style={/pgf/number format/.cd,fixed,precision=5}.
Here is a short list of possibilities:
123.46 \pgfmathprintnumber{123.456789}
12,345.68 \pgfmathprintnumber{12345.6789}
12,345.6789 \pgfmathprintnumber
[fixed,precision=5]{12345.6789}
12,345.67890 \pgfmathprintnumber
[fixed,fixed zerofill,precision=5]{12345.6789}
12.345,67890 \pgfmathprintnumber
[fixed,fixed zerofill,precision=5,use comma]
{12345.6789}
1.23 · 104 \pgfmathprintnumber

[sci]{12345.6789}
1.23457 · 104 \pgfmathprintnumber

[sci,sci zerofill,precision=5]{12345.6789}
1.23 ⇥ 101 \pgfmathprintnumber

[sci,sci generic=
{mantissa sep=\times,exponent={10^{#1}}}]
{12.345}
1 1 \pgfmathprintnumber[frac]{0.333333333333333};
3; 2 \pgfmathprintnumber[frac]{0.5}
+2 \pgfmathprintnumber[print sign]{2}
1 000 000.123456 \pgfmathprintnumber

[1000 sep={\,},fixed,precision=6]{1000000.123456}
1 000 000.123 456 \pgfmathprintnumber[

1000 sep={\,},fixed,precision=6,
1000 sep in fractionals]
{1000000.123456}
Each of these keys requires the prefix ‘/pgf/number format/’ when used inside of a pgfplots style
(try /pgf/number format/.cd,hnumber formatting keysi to use the same prefix for many hnumber
formatting keysi).
The number formatting uses \pgfmathprintnumber, a pgf command to typeset numbers. A full
reference of all supported options is shipped with pgfplots: it is documented in the reference manual
for PgfplotsTable, Section ‘Number Formatting Options’. The same reference can be found in the
documentation for pgf.
Note that the number printer knows nothing about pgfplots. In particular, it is not responsible for
logs and their representation.
3. For a logarithmic axis, one may want to modify the number formatting style for the exponent only.
In this case, redefine the style log plot exponent style (its documentation contains a couple of
examples).
4. In order to get fixed point tick labels on a logarithmic axis, you can use log ticks with fixed
point (see below).
4.13.2 PGFPlots-specific Number Formatting

This section contains fine–tuning options to change number formatting aspects – but only things which are
specific to pgfplots like peculiarities of tick labels on logarithmic axes. Consider browsing Section 4.13.1
first to see if you need this section.
\pgfmathprintnumber{hx i}
Generates pretty-printed output for the (real) number hx i. The input number hx i is parsed using
\pgfmathfloatparsenumber which allows arbitrary precision.
Numbers are typeset in math mode using the current set of number printing options, see below. Optional
arguments can also be provided using \pgfmathprintnumber[hoptionsi]{hx i}.
Please refer to the manual of PgfplotsTable (shipped with this package) for details about options
related to number-printing.
/pgfplots/log ticks with fixed point (style, no value)

Reconfigures pgfplots to display tick labels of logarithmic axes using fixed point numbers instead of
the exponential style.
\begin{tikzpicture}
10,000 \begin{semilogyaxis}[log ticks with fixed point]
\addplot+[domain=0:10] {exp(x)};
1,000 \end{semilogyaxis}
\end{tikzpicture}
100
10
1
0 2 4 6 8 10
0.1 Case 1
Case 2
0.01
Error
0.001
0.0001
0.00001
10 100 1,000 10,000 100,000

Cost
\begin{tikzpicture}
\begin{loglogaxis}[
log ticks with fixed point,
xlabel=Cost,ylabel=Error]
(5, 8.31160034e-02)
(17, 2.54685628e-02)
(49, 7.40715288e-03)
(129, 2.10192154e-03)
(321, 5.87352989e-04)
(769, 1.62269942e-04)
(1793, 4.44248889e-05)
(4097, 1.20714122e-05)
(9217, 3.26101452e-06)
};
(7, 8.47178381e-02)
(31, 3.04409349e-02)
(111, 1.02214539e-02)
(351, 3.30346265e-03)
(1023, 1.03886535e-03)
(2815, 3.19646457e-04)
(7423, 9.65789766e-05)
(18943, 2.87339125e-05)
(47103, 8.43749881e-06)
};
\legend{Case 1,Case 2}
\end{loglogaxis}
\end{tikzpicture}
The style replaces log number format basis.
/pgfplots/log plot exponent style={hkey-value-listi}

Allows to configure the number format of log plot exponents. This style is installed just before ‘log
number format basis’ will be invoked. Please note that this style will be installed within the default
code for ‘log number format code’.
ex \pgfplotsset{
e2x samples=15,
width=7cm,
105.0 xlabel=$x$,
ylabel=$f(x)$,
f (x)
extra y ticks={45},
101.7 legend style={at={(0.03,0.97)},
100.0 anchor=north west}}
\begin{tikzpicture}
5.0 log plot exponent style/.style={
10 /pgf/number format/fixed zerofill,
5 0 5 10 /pgf/number format/precision=1},
x domain=-5:10]
\addplot {exp(x)};
\addplot {exp(2*x)};
\legend{$e^x$,$e^{2x}$}
\end{semilogyaxis}
\end{tikzpicture}
ex \pgfplotsset{
e2x samples=15,
width=7cm,
105 xlabel=$x$,
ylabel=$f(x)$,
f (x)
extra y ticks={45},
101,65 legend style={at={(0.03,0.97)},
100 anchor=north west}}
\begin{tikzpicture}
5 log plot exponent style/.style={
10 /pgf/number format/fixed,
5 0 5 10 /pgf/number format/use comma,
x /pgf/number format/precision=2},
domain=-5:10]
\addplot {exp(x)};
\addplot {exp(2*x)};
\legend{$e^x$,$e^{2x}$}
\end{semilogyaxis}
\end{tikzpicture}
/pgfplots/log identify minor tick positions=true|false (initially false)

Set this to true if you want to identify log–plot tick labels at positions
i · 10j
with i 2 {2, 3, 4, 5, 6, 7, 8, 9}, j 2 Z. This may be valuable in conjunction with the ‘extra x ticks’ and
‘extra y ticks’ options.
Standard options \begin{tikzpicture}%
\begin{loglogaxis}
[title=Standard options,
width=6cm]
(1e-2,10)
(3e-2,100)
(6e-2,200)
};
\end{loglogaxis}
\end{tikzpicture}%
101
2 1.5
10 10
with minor tick identification without minor tick identification
102 102
101 101
2 3 · 10 2 6 · 10 2 2 1.52 1.22
10 10 10 10
\pgfplotsset{every axis/.append style={%
width=6cm,
xmin=7e-3,xmax=7e-2,
extra x ticks={3e-2,6e-2},
extra x tick style={major tick length=0pt,font=\footnotesize}
}}%
\begin{loglogaxis}[
xtick={1e-2},
title=with minor tick identification,
extra x tick style={
log identify minor tick positions=true}]
(1e-2,10)
(3e-2,100)
(6e-2,200)
};
\end{loglogaxis}
\end{tikzpicture}%
\begin{loglogaxis}[
xtick={1e-2},
title=without minor tick identification,
log identify minor tick positions=false}]
(1e-2,10)
(3e-2,100)
(6e-2,200)
};
\end{loglogaxis}%
\end{tikzpicture}%
This key is set by the default styles for extra ticks.
/pgfplots/log number format code/.code={h... i}

Provides TEX-code to generate log plot tick labels. Argument ‘#1’ is the (natural) logarithm of the tick
position. The default implementation invokes log base 10 number format code after it changed the
log basis to 10. It also checks the other log plot options.
This key will have a di↵erent meaning when the log basis has been chosen explicitly, see the log basis
x key.
/pgfplots/log base 10 number format code/.code={h... i}

Allows to change the overall appearance of base 10 log plot tick labels. The default implementation
invokes log number format basis={10}{#1}.
Use log plot exponent style if you only want to change number formatting options for the exponent.
/pgfplots/log number format basis/.code={h... i}

Typesets a logarithmic tick. The first supplied argument is the log basis, the second the exponent. The
initial configuration is
\pgfplotsset{
/pgfplots/log number format basis/.code 2 args={$#1^{\pgfmathprintnumber{#2}}$}
}
Use log plot exponent style if you only want to change number formatting options for the exponent.
4.14 Specifying the Plotted Range

4.14.1 Configuration of Limits Ranges
/pgfplots/xmin={hcoord i}
/pgfplots/ymin={hcoord i}
/pgfplots/zmin={hcoord i}
/pgfplots/xmax={hcoord i}
/pgfplots/ymax={hcoord i}
/pgfplots/zmax={hcoord i}
/pgfplots/min={hcoord i}
/pgfplots/max={hcoord i}
These options allow to define the axis limits, i.e. the lower left and the upper right corner. Everything
outside of the axis limits will be clipped away.
Attention: these values (and other related keys in this section) merely a↵ect how pgfplots deter-
mines the axis’ range. The values are unrelated to function sampling. Please refer to domain in order
to define the sampling range for plot expressions. It is valid to sample a larger or smaller range than
the displayed portion of the axis. Use xmin and its friends to specify which portion of the axis is to be
displayed.
Each of these keys is optional, and missing limits will be determined automatically from input data.
Here, the min and max keys set limits for x, y and z to the same hcoord i.
If x-limits have been specified explicitly and y-limits are computed automatically, the automatic com-
putation of y-limits will only considers points which fall into the specified x-range (and vice–versa). The
same holds true if, for example, only xmin has been provided explicitly: in that case, xmax will be up-
dated only for points for which x xmin holds. This feature can be disabled using clip limits=false.
Axis limits can be increased automatically using the enlargelimits option.
4.14. SPECIFYING THE PLOTTED RANGE 305
Auto Limits \begin{tikzpicture}
\begin{axis}[title=Auto Limits]
\addplot {x^2};
\end{axis}
10
0
6 4 2 0 2 4 6
xmin=0
\begin{tikzpicture}
\begin{axis}[title={\texttt{xmin=0}},xmin=0]
\addplot {x^2};
\end{axis}
10
0
0 1 2 3 4 5
ymax=10
\begin{tikzpicture}
10 \begin{axis}[title={\texttt{ymax=10}},ymax=10]
\addplot {x^2};
\end{axis}
\end{tikzpicture}
0
2 0 2
Note that even if you provide ymax=10, data points with y > 10 will still be visualized – producing a
line which leaves the plotted range.
See also the restrict x to domain and restrict x to domain* keys – they allow to discard or clip
input coordinates which are outside of some domain, respectively.
During the visualization phase, i.e. during \end{axis}, these keys will be set to the final axis limits.
You can access the values by means of \pgfkeysvalueof{/pgfplots/xmin}, for example:
Axis limits are [ 6 : 6] ⇥ [ 2.5 : 27.5] \begin{tikzpicture}
\begin{axis}[
% Show (automatically) computed limits:
title={
Axis limits are
20 $
[\pgfmathprintnumber{\pgfkeysvalueof{/pgfplots/xmin}}
:\pgfmathprintnumber{\pgfkeysvalueof{/pgfplots/xmax}}
] \times
10 [\pgfmathprintnumber{\pgfkeysvalueof{/pgfplots/ymin}}
:\pgfmathprintnumber{\pgfkeysvalueof{/pgfplots/ymax}}
]$ },
]
0 \addplot {x^2};
\end{axis}
6 4 2 0 2 4 6 \end{tikzpicture}
This access is possible inside of any axis description (like xlabel, title, legend entries etc.) or
any annotation (i.e. inside of \node, \draw or \path and coordinates in (hx i,hyi)), but not inside of
\addplot (limits may not be complete at this stage).
/pgfplots/xmode=normal|linear|log (initially normal)

/pgfplots/ymode=normal|linear|log (initially normal)
/pgfplots/zmode=normal|linear|log (initially normal)
Allows to choose between linear (=normal) or logarithmic axis scaling or logplots for each
x, y, z-combination.
Logarithmic plots use the current setting of log basis x and its variants to determine the basis (default
is e).
/pgfplots/x dir=normal|reverse (initially normal)

/pgfplots/y dir=normal|reverse (initially normal)
/pgfplots/z dir=normal|reverse (initially normal)
Allows to reverse axis directions such that values are given in decreasing order.
\begin{axis}[
xlabel=$x$ \emph{decreasing} $\to$,
x dir=reverse]
\addplot {x+rand*0.3};
0 \end{axis}
\end{tikzpicture}
5
6 4 2 0 2 4 6
x decreasing !
\begin{axis}[
ylabel=$y$ \emph{decreasing} $\to$,
y decreasing !
y dir=reverse]
10 \addplot {x^2};
\end{axis}
\end{tikzpicture}
20
6 4 2 0 2 4 6
Note that axis descriptions and relative positioning macros will stay at the same place as they would
for non–reversed axes.
reversed axis
·1010 ·1010
2
2
y decreasing !
0 0
2
2
6 4 2 0 2 4 6
x normal
\begin{tikzpicture}
\begin{axis}[
ylabel=$y$ \emph{decreasing} $\to$,
xlabel=$x$ normal,
title=reversed axis,
y dir=reverse,
colorbar,
colorbar style={y dir=reverse}]
\addplot+[mesh,scatter] {x^15};
\end{axis}
\end{tikzpicture}
Note that colorbars won’t be reversed automatically, you will have to reverse the sequence of color
bars manually in case this is required as in the preceding example.
/pgfplots/clip limits=true|false (initially true)

Configures what to do if some, but not all axis limits have been specified explicitly. In case clip
limits=true, the automatic limit computation will only consider points which do not contradict the
explicitly set limits.
This option has nothing to do with path clipping, it only a↵ects how the axis limits are computed.
/pgfplots/enlarge x limits=auto|true|false|upper|lower|hval i|value=hval i|abs value=hval i|

abs=hval i|rel=hval i (initially auto)
/pgfplots/enlarge y limits=auto|true|false|upper|lower|hval i|value=hval i|abs value=hval i|
/pgfplots/enlarge z limits=auto|true|false|upper|lower|hval i|value=hval i|abs value=hval i|
/pgfplots/enlargelimits=hcommon valuei
Enlarges the axis size for one axis (or all of them for enlargelimits) somewhat if enabled.
You can set xmin, xmax and ymin, ymax to the minimum/maximum values of your data and enlarge x
limits will enlarge the canvas such that the axis doesn’t touch the plots.
The value true enlarges the lower and upper limit.
The value false uses tight axis limits as specified by the user (or read from input coordinates).
The value auto will enlarge limits only for axis for which axis limits have been determined automatically.
For three–dimensional figures, the auto mechanism applies only for the z axis. The x and y axis won’t
be enlarged.
\begin{tikzpicture}
500 \begin{axis}[small]
\addplot {5 * x^3 - x^2 + 4*x -2};
\end{axis}
\end{tikzpicture}
0
500
6 4 2 0 2 4 6
Specifying a number value like ‘enlarge x limits=0.2’ will enlarge lower and upper axis limit rela-
tively. The following example adds 20% of the axis limits on both sides:
\begin{tikzpicture}
500 \begin{axis}[small,enlarge x limits=0.2]
\addplot {5 * x^3 - x^2 + 4*x -2};
\end{axis}
\end{tikzpicture}
0
500
6 4 2 0 2 4 6
The choice rel={hvaluei} is the same as true,value={hvaluei}, i.e. it activates relative enlargement
for both upper and lower limit.
The value upper enlarges only the upper axis limit while lower enlarges only the lower axis limit. In
this case, the amount added to the respective limit can be specified using the value={hval i} key. It can
be combined with any of the other possible values. For example,
\pgfplotsset{enlarge x limits={value=0.2,upper}}
will enlarge (only) the upper axis limit by 20% of the axis range. Another example is
\pgfplotsset{enlarge x limits={value=0.2,auto}}
which changes the default threshold of the auto value to 20%.
\begin{tikzpicture}
500 \begin{axis}[small,minor x tick num=1,
enlarge x limits={rel=0.5,upper}
]
\addplot {5 * x^3 - x^2 + 4*x -2};
0
\end{axis}
\end{tikzpicture}
500
4 2 0 2 4 6 8 10
While value uses relative thresholds, abs value accepts absolute values: it adds an absolute value to
the selected axis. The choice abs={hvaluei} is the same as true,abs value={hvaluei}, i.e. it adds an
absolute value to both upper and lower limit:
\begin{tikzpicture}
enlarge x limits={abs=3}
]
\addplot {5 * x^3 - x^2 + 4*x -2};
0
\end{axis}
\end{tikzpicture}
500
8 6 4 2 0 2 4 6 8
Here, we enlarged by 3 units of the x axis. Note that you can also specify dimensions like 1cm:
\begin{tikzpicture}
enlarge x limits={abs=1cm}
]
\addplot {5 * x^3 - x^2 + 4*x -2}
0
coordinate[pos=0] (first)
coordinate[pos=1] (last);
500 \draw[red,->] (first) -- ++(-1cm,0pt);

\draw[red,->] (last) -- ++(1cm,0pt);
\end{axis}
8 6 4 2 0 2 4 6 8 \end{tikzpicture}
Technically, the use of absolute dimensions is a little bit di↵erent. For example, it allows to enlarge by
more than width which is impossible for all other choices. pgfplots will try to fulfill both the provided
width/height and the absolute axis enlargements. If it fails to do so, it will give up on width/height
constraints and print a warning message to your log file. See also the key enlargelimits respects
figure size.
Attention: abs value is applied multiplicatively for logarithmic axes! That means abs value=10 for
a logarithmic axis adds log 10 to upper and/or lower axis limits.
\begin{tikzpicture}
10 1
\begin{loglogaxis}[
small,
enlarge x limits={abs=11},
10 4
]
\addplot+[domain=1:100000] {x^-2};
\end{loglogaxis}
10 7
\end{tikzpicture}
10 10
10 1 100 101 102 103 104 105 106
Note that enlargelimits is applied before any changes to axis limits are considered as part of scale
mode: enlargelimits will always be applied. Afterwards, the choice scale mode=scale uniformly
will enlarge limits once more in order to satisfy all scaling constraints. The two limit enlargements are
independent of each other, i.e. even if you say enlargelimits=false, scale mode will still increase axis
limits if this seems to be necessary. An exception for this rule is enlarge-by-dimension, i.e. something
like abs=1cm (see enlargelimits respects figure size for this case). See scale mode (especially
scale mode=units only) and unit rescale keep size for detail on how to disable limit enlargement
caused by scale mode.
/pgfplots/enlargelimits respects figure size=true|false (initially true)

A key which is only used for something like enlarge x limits={abs=1cm}, i.e. for enlarge-by-
dimension. It controls if pgfplots will try to respect width/height. You should probably always
leave it as its default unless you run into problems.
If pgfplots fails to respect the figure size, it will print a warning message of sorts “enlargelimits
respects figure size=true: could not respect the prescribed width/height” to your log file.
/pgfplots/log origin x=0|infty (initially infty)

/pgfplots/log origin y=0|infty (initially infty)
/pgfplots/log origin z=0|infty (initially infty)
/pgfplots/log origin=0|infty (initially infty)
Allows to choose which coordinate is the logical “origin” of a logarithmic plot (either for a particular
axis or for all of them).
The choice log origin=infty is probably useful for stacked plots: it defines the “origin” in log–
coordinates to be 1. To be compatibly with older versions, this is the default. Note that 1 is
clipped to be the lowest displayed value on the axis in question.
The choice log origin=0 defines the logarithmic origin to be the natural choice log(1) = 0. This is
particularly useful for ycomb plots.
/pgfplots/update limits=true|false (initially true)

Can be used to interrupt updates of the data limits (for example, for single \addplot commands).
This has the same e↵ect as \pgfplotsinterruptdatabb ... \endpgfplotsinterruptdatabb.
\begin{pgfplotsinterruptdatabb}
\end{pgfplotsinterruptdatabb}
Everything in henvironment contentsi will not contribute to the data bounding box.
The same e↵ect can be achieved with update limits=false inside curly braces.
4.14.2 Accessing Computed Limit Ranges

This section is for those who want or need to access the computed limits programmatically. It lists how to
access computed values.
/pgfplots/xmin (no value)

/pgfplots/ymin (no value)
4.15. TICK OPTIONS 311
/pgfplots/zmin (no value)

/pgfplots/xmax (no value)
/pgfplots/ymax (no value)
/pgfplots/zmax (no value)
These values are not only the input: after the survey phase, their values will be overwritten with the
resulting computed values. These may di↵er according to enlargelimits or scale mode.
If you need to access them, you can write \pgfkeysvalueof{/pgfplots/xmin} somewhere in your code.
4.15 Tick Options

4.15.1 Tick Coordinates and Label Texts
/pgfplots/xtick distance={hempty or numerici} (initially empty)
/pgfplots/ytick distance={hempty or numerici} (initially empty)
/pgfplots/ztick distance={hempty or numerici} (initially empty)
Allows to define the distance between generated tick positions.
\begin{axis}[
xtick distance=45,
0.5 ytick distance=0.5,
]
0
\addplot[smooth,domain=0:360] {sin(x)};
\end{axis}
1
0 45 90 135 180 225 270 315 360
10 1 \begin{loglogaxis}[
xtick distance=10^(1.5),
10 2 ytick distance=10^1,
10 3 ]
10 4 \end{loglogaxis}
10 6
10 7
10 8
100 101.5 103
The initial value is empty meaning “generate as needed”.

Any other value is interpreted as constant (evaluated using the math expression parser) and is taken
as-is. Note that such values should be chosen in a way such that not too many and not too few tick
positions are generated.
/pgfplots/xtick=\empty|data|{hcoordinate listi} (initially {hi})

/pgfplots/ytick=\empty|data|{hcoordinate listi} (initially {hi})
/pgfplots/ztick=\empty|data|{hcoordinate listi} (initially {hi})
These options assign a list of positions where ticks shall be placed. The argument is either the empty
string (which is the initial value), the command \empty, the special string ‘data’ or a list of coordinates.
The initial configuration of an empty string means to generate these positions automatically. The choice
\empty will result in no tick at all. The special value ‘data’ will produce tick marks at every coordinate
of the first plot. Otherwise, tick marks will be placed at every coordinate in hcoordinate listi.
The hcoordinate listi will be used inside of a \foreach \x in {hcoordinate listi} statement. The format
is as follows:
• {0,1,2,5,8,1e1,1.5e1} (a series of coordinates),
• {0,...,5} (the same as {0,1,2,3,4,5}),
• {0,2,...,10} (the same as {0,2,4,6,8,10}),
• {9,...,3.5} (the same as {9, 8, 7, 6, 5, 4}),
• See [6, Section 34] for a more detailed definition of the options.
• Please be careful with white spaces inside of hcoordinate listi (at least around the dots).
For logplots, pgfplots will apply log(·) to each element in ‘hcoordinate listi’ (similarly, any custom
transformations are applied to the argument list).
1
\begin{loglogaxis}[xtick={12,9897,1468864}]
2 % see above for this macro:
10
\plotcoords
\end{loglogaxis}
3
4
10
5
10
101.08 104 106.17
\begin{tikzpicture}
4.5 \begin{axis}[
xtick=\empty,
3.7 ytick={-2,0.3,3,3.7,4.5}]
3 \addplot+[smooth] coordinates {
(-2,3) (-1.5,2) (-0.3,-0.2)
(1,1.2) (2,2) (3,5)};
\end{axis}
\end{tikzpicture}
0.3
Attention: You can’t use the ‘...’ syntax if the elements are too large for TEX! For example,
‘xtick=1.5e5,2e7,3e8’ will work (because the elements are interpreted as strings, not as numbers),
but ‘xtick=1.5,3e5,...,1e10’ will fail because it involves real number arithmetics beyond TEX’s
capacities.
The default choice for tick positions in normal plots is to place a tick at each coordinate i · h. The step
size h depends on the axis scaling and the axis limits. It is chosen from a list of “feasible” step sizes
such that neither too much nor too few ticks will be generated. The default for logplots is to place
ticks at positions 10i in the axis’ range. The positions depend on the axis scaling and the dimensions
of the picture. If log plots contain just one (or two) positions 10i in their limits, ticks will be placed at
positions 10i·h with “feasible” step sizes h as in the case of linear axis.
The tick appearance can be (re)configured with
\pgfplotsset{tick style={very thin,gray}}% modifies the style ‘every tick’

\pgfplotsset{minor tick style={black}} % modifies the style ‘every minor tick’
These style commands can be used at any time. The tick line width can be configured with ‘major
tick length’ and ‘minor tick length’.
\begin{tikzpicture}
\begin{axis}[xtick=data,xmajorgrids]
(1,2)
(2,5)
6 (4,6.5)
(6,8)
(10,9)
4 };
\end{axis}
\end{tikzpicture}
2
1 2 4 6 10
A log plot with small axis range \begin{tikzpicture}
\begin{loglogaxis}[
4 title=A log plot with small axis range]
10
(10,1e-4)
(17,8.3176e-05)
(25,7.0794e-05)
(50,5e-5)
4.2 };
10 \end{loglogaxis}
\end{tikzpicture}
101 101.2 101.4 101.6
/pgfplots/minor x tick num={hnumber i} (initially 0)

/pgfplots/minor y tick num={hnumber i} (initially 0)
/pgfplots/minor z tick num={hnumber i} (initially 0)
/pgfplots/minor tick num={hnumber i}
Sets the number of minor tick lines used either for single axes or for all of them.
Minor ticks will be disabled if the major ticks don’t have the same distance and they are currently only
available for linear axes (not for logarithmic ones).
\begin{tikzpicture}
100 \begin{axis}[minor tick num=1]
\addplot {x^3};
\addplot {-20*x};
\end{axis}
\end{tikzpicture}
0
100
6 4 2 0 2 4 6
\begin{tikzpicture}
100 \begin{axis}[minor tick num=3]
\addplot {x^3};
\addplot {-20*x};
\end{axis}
\end{tikzpicture}
0
100
6 4 2 0 2 4 6
\begin{tikzpicture}
100 \begin{axis}[minor x tick num=1,
minor y tick num=3]
\addplot {x^3};
\addplot {-20*x};
\end{axis}
0
\end{tikzpicture}
100
6 4 2 0 2 4 6
/pgfplots/minor xtick=data|{hcoordinate listi} (initially empty)

/pgfplots/minor ytick=data|{hcoordinate listi} (initially empty)
/pgfplots/minor ztick=data|{hcoordinate listi} (initially empty)
/pgfplots/minor tick=data|{hcoordinate listi}
Allows to provide a list of minor tick positions manually. The syntax is almost the same as for xtick
or ytick: simply provide either a comma–separated list of tick positions or the special value data.
An empty argument argument disables the minor tick feature (in contrast to xtick where the special
value \empty clears the list and an empty argument causes pgfplots to compute a default tick list).
In contrast to minor x tick num, this key allows to provide non–uniform minor tick positions.
\begin{tikzpicture}
100 \begin{axis}[minor xtick={-3,1},grid=minor]
\addplot {x^3};
\addplot {-20*x};
\end{axis}
\end{tikzpicture}
0
100
6 4 2 0 2 4 6
\begin{tikzpicture}
\begin{axis}[minor ytick=data]
20 \addplot {x^2};
\end{axis}
\end{tikzpicture}
10
0
6 4 2 0 2 4 6
This key has precedence over minor x tick num and its variants; if both of them are given, minor
xtick is preferred and minor x tick num is ignored.
/pgfplots/extra x ticks={hcoordinate listi}

/pgfplots/extra y ticks={hcoordinate listi}
/pgfplots/extra z ticks={hcoordinate listi}
Adds additional tick positions and tick labels to the x or y axis. ‘Additional’ tick positions do not a↵ect
the normal tick placement algorithms, they are drawn after the normal ticks. This has two benefits:
first, you can add single, important tick positions without disabling the default tick label generation
and second, you can draw tick labels ‘on top’ of others, possibly using di↵erent style flags.
15
e \begin{axis}[
xmin=0,xmax=3,ymin=0,ymax=15,
extra y ticks={2.71828},
10 extra y tick labels={$e$},
extra x ticks={2.2},
extra x tick style={grid=major,
tick label style={
5 rotate=90,anchor=east}},
extra x tick labels={Cut},
e ]
\addplot {exp(x)};
\addlegendentry{$e^x$}
0 \end{axis}
0 1 2 3
Cut
\end{tikzpicture}
Explicitly Provided Limits With Extra Ticks With Extra Ticks; 10e format
5 5 5
10 10 10
6 5.3
5 · 10 10
5.6 5.6
10 10
101 101 2 · 101 4 · 101 101 101.3 101.6
\pgfplotsset{every axis/.append style={width=5.3cm}}
\begin{tikzpicture}
\begin{loglogaxis}[
title=Explicitly Provided Limits,
xtickten={1,2},
ytickten={-5,-6}]
{(10,1e-5) (20,5e-6) (40,2.5e-6)};
\end{loglogaxis}
\end{tikzpicture}
\begin{tikzpicture}
\begin{loglogaxis}[
title=With Extra Ticks,
xtickten={1,2},
ytickten={-5,-6},
extra y ticks={5e-6,2.5e-6}]
{(10,1e-5) (20,5e-6) (40,2.5e-6)};
\end{loglogaxis}
\end{tikzpicture}
\begin{tikzpicture}
\begin{loglogaxis}[
title=With Extra Ticks; $10^e$ format,
extra tick style={log identify minor tick positions=false},
xtickten={1,2},
ytickten={-5,-6},
extra y ticks={5e-6,2.5e-6}]
{(10,1e-5) (20,5e-6) (40,2.5e-6)};
\end{loglogaxis}
\end{tikzpicture}
Remarks:
• Use extra x ticks to highlight special tick positions. The use of extra x ticks does not a↵ect
minor tick/grid line generation, so you can place extra ticks at positions j · 10i in log–plots.
• Extra ticks are always typeset as major ticks.
They are a↵ected by major tick length or options like grid=major.
• Use the style every extra x tick (every extra y tick) to configure the appearance.
• You can also use ‘extra x tick style={h...i}’ which has the same e↵ect.
/pgfplots/xtickten={hexponent base 10 listi}

/pgfplots/ytickten={hexponent base 10 listi}
/pgfplots/ztickten={hexponent base 10 listi}
These options allow to place ticks at selected positions 10k , k 2 {hexponent base 10 listi}. They are
only used for logplots. The syntax for {hexponent base 10 listi} is the same as above for xtick={hlisti}
or ytick={hlisti}.
Using ‘xtickten={1,2,3,4}’ is equivalent to ‘xtick={1e1,1e2,1e3,1e4}’, but it requires fewer com-
putational time and it allows to use the short syntax ‘xtickten={1,...,4}’.
102 2 2x+6
\begin{tikzpicture}
1.5x 3 \begin{semilogyaxis}[
2
samples=8,
100 ytickten={-6,-4,...,4},
domain=0:10]
2
10 \addplot {2^(-2*x + 6)};
\addlegendentry{$2^{-2x + 6}$}
4
10 % or invoke gnuplot to generate coordinates:
\addplot gnuplot[id=pow2]
{2**(-1.5*x -3)};
6
10 \addlegendentry{$2^{-1.5x -3}$}
0 2 4 6 8 10 \end{semilogyaxis}
\end{tikzpicture}
In case log basis x6= 10, the meaning of xtickten changes. In such a case, xtickten will still assign
the exponent, but for the chosen log basis x instead of base 10.
/pgfplots/xticklabels={hlabel listi}
/pgfplots/yticklabels={hlabel listi}
/pgfplots/zticklabels={hlabel listi}
Assigns a list of tick labels to each tick position. Tick positions are assigned using the xtick and
ytick-options.
This is one of two options to assign tick labels directly. The other option is xticklabel={hcommand i}
(or yticklabel={hcommand i}). The option ‘xticklabel’ o↵ers higher flexibility while ‘xticklabels’
is easier to use. See also the variant xticklabels from table.
The argument hlabel listi has the same format as for ticks, that means
xticklabels={$\frac{1}{2}$,$e$}
denotes the two–element–list { 12 , e}. The list indices match the indices of the tick positions. If you need
commas inside of list elements, use
xticklabels={{0,5}, $e$}.
\begin{axis}[
xtick={-1.5,-1,...,1.5},
xticklabels={%
$-1\frac 12$,
$-1$,
0.5
$-\frac 12$,
$0$,
$\frac 12$,
$1$},
% note: \frac can be done automatically:
0 % xticklabel style={/pgf/number format/frac},
]
1 1 0 1 1 \addplot[smooth,blue,mark=*]
2 2
coordinates {
(-1, 1)
(-0.75, 0.5625)
(-0.5, 0.25)
(-0.25, 0.0625)
(0, 0)
(0.25, 0.0625)
(0.5, 0.25)
(0.75, 0.5625)
(1, 1)
};
\end{axis}
\end{tikzpicture}
ytickten={-2,-1,0,1,2},
10 yticklabels={$\frac{1}{100}$,%
$\frac{1}{10}$,%
1,10,100},
1
]
\addplot {exp(x)};
1
10 \end{semilogyaxis}
\end{tikzpicture}
1
100
6 4 2 0 2 4 6
Note that it is also possible to terminate list entries with two backslashes, \\. In that case, the last
entry needs to be terminated by \\ as well (it is the same alternative syntax which is also accepted for
\legend and cycle list).
Please keep in mind that the arguments always refer the a list of tick positions, although it does not
alter or define the list of positions. Consequently, you should also provide the list of positions. Note
that a list of positions might be longer than what is actually displayed (in case the axis limits clip some
of the value away), but the index mapping into hlabel listi still includes the clipped values.
/pgfplots/xticklabel={hcommand i}
/pgfplots/yticklabel={hcommand i}
/pgfplots/zticklabel={hcommand i}
These keys change the TEX-command which creates the tick labels assigned to each tick position (see
options xtick and ytick).
This is one of the two options to assign tick labels directly. The other option is ‘xticklabels={hlabel
listi}’ (or yticklabels={hlabel listi}). The option ‘xticklabel’ o↵ers higher flexibility while
‘xticklabels’ is easier to use.
The argument hcommand i can be any TEX-text. The following commands are valid inside of hcommand i:
\tick The current element of option xtick (or ytick).
\ticknum The current tick number, starting with 0 (it is a macro containing a number).
\nexttick This command is only valid if the x tick label as interval option is set (or the cor-
responding variable for y). It will contain the position of the next tick position, that means the
right boundary of the tick interval.
The default argument is
• \axisdefaultticklabel for normal plots:
\def\axisdefaultticklabel{$\pgfmathprintnumber{\tick}$}
• \axisdefaultticklabellog for logplots:

\def\axisdefaultticklabellog{%
\pgfkeysgetvalue{/pgfplots/log number format code/.@cmd}\pgfplots@log@label@style
\expandafter\pgfplots@log@label@style\tick\pgfeov
}
That means you can configure the appearance of linear axis with the number formatting options de-
scribed in Section 4.13 and logarithmic axis with log number format code, see below.
The key yticklabel is a code fragment which is supposed to handle any incoming \tick value. Con-
sequently, it can be used to append custom suffixes or even units:
log ticks with fixed point,
10 xticklabel={$\pgfmathprintnumber{\tick}_e$},
]
\addplot {exp(x)};
1
\end{semilogyaxis}
\end{tikzpicture}
0.1
0.01
6e 4e 2e 0e 2e 4e 6e
The following example uses explicitly formatted x tick labels and a small TEX script to format y tick
labels as fractions in the form hsignihnumber i/10 (note that the /pgf/number format/frac style can
do similar things automatically, see PgfplotsTable and the documentation therein).
A special Prewavelet
11/10
1/10
1 1 3 1
4 2 4
6/10
12/10
% \usepackage{nicefrace}% required
\begin{tikzpicture}
\begin{axis}[
% x ticks explicitly formatted:
xtick={0,1,0.5,0.25,0.75},
xticklabels={$0$,$1$,$\frac12$,$\frac14$,$\frac34$},
% y ticks automatically by some code fragment:
ytick=data,
yticklabel={%
\scriptsize
\ifdim\tick pt<0pt % a TeX \if -- see TeX Book
\pgfmathparse{-10*\tick}%
$-\nicefrac{\pgfmathprintnumber{\pgfmathresult}}{10}$%
\else
\ifdim\tick pt=0pt
\else
\pgfmathparse{10*\tick}%
$\nicefrac{\pgfmathprintnumber{\pgfmathresult}}{10}$%
\fi
\fi
},
% NOTE: this here does the same:
% yticklabel style={/pgf/number format/.cd,frac,
% frac TeX=\nicefrac,frac whole=false,frac denom=10},
ymajorgrids,
title=A special Prewavelet,
axis x line=center,
axis y line=left,
]
\addplot coordinates {(0,-1.2) (0.25,1.1)
(0.5,-0.6) (0.75,0.1) (1,0)};
\end{axis}
\end{tikzpicture}
The TEX script takes the \tick macro as input and applies some logic. The \ifdim\tick pt<0pt
means “if dimension \tick pt < 0pt”. The \ifdim is TEX’s only way to compare real fixed point
numbers and the author did not want to invoke \pgfmath for this simple task. Since \ifdim expects a
dimension, we have to use the pt suffix which is compatible with \pgfmath. The result is that negative
numbers, zero and positive numbers are typeset di↵erently.
You can change the appearance of tick labels with
\pgfplotsset{tick label style={

font=\tiny,
/pgf/number format/sci}}% this modifies the ‘every tick label’ style
and/or
\pgfplotsset{x tick label style={

above,
/pgf/number format/fixed zerofill}}% this modifies the ‘every x tick label’ style
and
\pgfplotsset{y tick label style={font=\bfseries}}% modifies ‘every y tick label’
/pgfplots/xticklabels from table={h\table or filenamei}{hcolnamei}

/pgfplots/yticklabels from table={h\table or filenamei}{hcolnamei}
/pgfplots/zticklabels from table={h\table or filenamei}{hcolnamei}
A variant of xticklabels={hlisti} which uses each entry in the column named hcolnamei from a table
as tick labels.
The first argument h\table or filenamei can be either a loaded table macro (i.e. the result of
\pgfplotstableread{hfile namei}{h\tablei}) or just a file name.
The second argument can be a column name, a column alias or a create on use specification (see
PgfplotsTable for the latter two). Furthermore, it can be [index]hinteger i in which case hinteger i
is a column index.
The behavior of xticklabels from table is the same as if the column hcolnamei would have been
provided as comma separated list to xticklabels. This means the column can contain text, TEX
macros or even math mode.
If you have white spaces in your cells, enclose the complete cell in curly braces, {example cell}.
The detailed input format for tables is discussed in \addplot table and in the documentation for
PgfplotsTable.
/pgfplots/extra x tick label={hTEX codei}

/pgfplots/extra y tick label={hTEX codei}
/pgfplots/extra z tick label={hTEX codei}
As xticklabel provides code to generate tick labels for each xtick, the key extra x tick label
provides code to generate tick labels for every element in extra x ticks.
/pgfplots/extra x tick labels={hlabel listi}

/pgfplots/extra y tick labels={hlabel listi}
/pgfplots/extra z tick labels={hlabel listi}
As xticklabels provides explicit tick labels for each xtick, the key extra x tick labels provides
explicit tick labels for every element in extra x ticks.
/pgfplots/x tick label as interval=true|false (initially false)

/pgfplots/y tick label as interval=true|false (initially false)
/pgfplots/z tick label as interval=true|false (initially false)
Allows to treat tick labels as intervals; that means the tick positions denote the interval boundaries. If
there are n positions, (n 1) tick labels will be generated, one for each interval.
\begin{tikzpicture}
\begin{axis}[x tick label as interval]
10 \addplot {3*x};
\end{axis}
\end{tikzpicture}
0
10
6 4 2 0 2 4
This mode enables the use of \nexttick inside of xticklabel (or yticklabel). A common application
might be a bar plot.
\begin{tikzpicture}
\begin{axis}[
ybar interval=0.9,
100 x tick label as interval,
xmin=2003,xmax=2030,
ymin=0,ymax=140,
xticklabel={
$\pgfmathprintnumber{\tick}$
50 -- $\pgfmathprintnumber{\nexttick}$},
xtick=data,
rotate=90,anchor=east,
/pgf/number format/1000 sep=}
0 ]
2003 – 2005
2005 – 2006
2006 – 2010
2010 – 2020
2020 – 2030
\addplot[draw=blue,fill=blue!40!white]
coordinates
{(2003,40) (2005,100) (2006,15)
(2010,90) (2020,120) (2030,3)};
\end{axis}
\end{tikzpicture}
/pgfplots/xminorticks=true|false (initially true)

/pgfplots/yminorticks=true|false (initially true)
/pgfplots/zminorticks=true|false (initially true)
/pgfplots/xmajorticks=true|false (initially true)
/pgfplots/ymajorticks=true|false (initially true)
/pgfplots/zmajorticks=true|false (initially true)
/pgfplots/ticks=minor|major|both|none (initially both)
Enables/disables the small tick lines either for single axis or for all of them. Major ticks are those
placed at the tick positions and minor ticks are between tick positions. Please note that minor ticks are
automatically disabled if xtick is not a uniform range55 .
The key minor tick length={hdimeni} configures the tick length for minor ticks while the major
variant applies to major ticks. You can configure the appearance using the following styles:
\pgfplotsset{every tick/.append style={color=black}} % applies to major and minor ticks,

\pgfplotsset{every minor tick/.append style={thin}} % applies only to minor ticks,
\pgfplotsset{every major tick/.append style={thick}} % applies only to major ticks.
There is also the style “every tick” which applies to both, major and minor ticks.
/pgfplots/xtickmin={hcoord i}
/pgfplots/ytickmin={hcoord i}
/pgfplots/ztickmin={hcoord i}
55 A uniform list means the di↵erence between all elements is the same for linear axis or, for logarithmic axes, log(10).
/pgfplots/xtickmax={hcoord i}
/pgfplots/ytickmax={hcoord i}
/pgfplots/ztickmax={hcoord i}
These keys can be used to modify minimum/maximum values before ticks are drawn. Because this
applies to axis discontinuities, it is described on page 258 in Section 4.9.12, “Axis Discontinuities”’.
4.15.2 Tick Alignment: Positions and Shifts

/pgfplots/xtick pos=left|right|both (initially both)
/pgfplots/ytick pos=left|right|both (initially both)
/pgfplots/ztick pos=left|right|both (initially both)
/pgfplots/tick pos=left|right|both
Allows to choose where to place the small tick lines. In the default configuration, this does also a↵ect
tick labels, see below. The tick pos style sets all of them to the same value (aliased by tickpos). This
option is only useful for boxed axes.
For x, the additional choices bottom and top can be used which are equivalent to left and right,
respectively. Both are accepted for y.
Changing tick pos will also a↵ect the placement of tick labels.
Note that it can also a↵ect the axis lines key, although not all combinations make sense. Make sure
the settings are consistent.
/pgfplots/xticklabel pos=left|right|default (initially default)

/pgfplots/yticklabel pos=left|right|default (initially default)
/pgfplots/zticklabel pos=left|right|default (initially default)
/pgfplots/ticklabel pos=left|right|default (initially default)
Allows to choose where to place tick labels. The choices left and right place tick labels either at the
left or at the right side of the complete axis. The choice default uses the same setting as xtick pos
(or ytick pos). This option is only useful for boxed axes – keep it to default for non-boxed figures.
The ticklabel pos style sets all three of them to the same value.
For x, the additional choices bottom and top can be used which are equivalent to left and right,
respectively. Both are accepted for x.
/pgfplots/xtick align=inside|center|outside (initially inside)

/pgfplots/ytick align=inside|center|outside (initially inside)
/pgfplots/ztick align=inside|center|outside (initially inside)
/pgfplots/tick align=inside|center|outside (initially inside)
Allows to change the location of the ticks relative to the axis lines. The tick align sets all of them to
the same value. Default is “inside”.
\begin{axis}[
xtick=data,ytick=data,
xtick align=center]
{(-3,0) (-2,0.1) (-1,-0.6)
0.1 (0,1)
0 (1,-0.6) (2,0.1) (3,0)};
\end{axis}
\end{tikzpicture}
0.6
3 2 1 0 1 2 3
\begin{axis}[
xtick=data,ytick=data,
ytick align=outside]
{(-3,0) (-2,0.1) (-1,-0.6)
0.1 (0,1)
0 (1,-0.6) (2,0.1) (3,0)};
\end{axis}
\end{tikzpicture}
0.6
3 2 1 0 1 2 3
These tick alignment options are set automatically by the axis x line and axis y line methods
(unless one appends an asterisk ‘*’):
1
\begin{tikzpicture}
\begin{axis}[
xtick=data,
axis x line=center,
xticklabels={,,},
ytick={-0.6,0,0.1,1},
1 yticklabels={
10
$-\frac{6}{10}$,,
$\frac{1}{10}$,$1$},
ymajorgrids,
axis y line=left,
6 enlargelimits=0.05]
{(-3,0) (-2,0.1) (-1,-0.6)
(0,1)
(1,-0.6) (2,0.1) (3,0)};
\end{axis}
\end{tikzpicture}
/pgfplots/xticklabel shift={hdimensioni} (initially empty)

/pgfplots/yticklabel shift={hdimensioni} (initially empty)
/pgfplots/zticklabel shift={hdimensioni} (initially empty)
/pgfplots/ticklabel shift={hdimensioni} (initially empty)
Shifts tick labels in direction of the outer unit normal of the axis by an amount of hdimensioni. The
ticklabel shift sets the same value for all axes.
This is usually unnecessary as the anchor of a tick label already yields enough spacing in most cases.
See also axis line shift.
/pgfplots/typeset ticklabels with strut (no value)

/pgfplots/default typeset ticklabel (no value)
Typically, pgfplots typesets ticks “as–is”. This works good in almost all cases, especially if you just
have numbers as tick labels.
But it needs adoptions if one ticklabel has text below the baseline and one does not. It is purpose of
typeset ticklabels with strut to align them.
Default With strut
2
\begin{tikzpicture}
2
\begin{axis}[
1.5
1.5 title=Default,
1
1 tiny,
0.5
0.5 symbolic x coords={a,T,g},
0
0 xticklabel style={name=tick no \ticknum},
a T g a T g xtick=data]
\addplot coordinates {(a,0) (T,1) (g,2)};
\end{axis}
\draw[red,opacity=0.5] (tick no 0.base)
-- (tick no 2.base);
\end{tikzpicture}
%
\begin{tikzpicture}
\begin{axis}[
title=With strut,
tiny,
symbolic x coords={a,T,g},
xtick=data,
xticklabel style={name=tick no \ticknum},
typeset ticklabels with strut]
\addplot coordinates {(a,0) (T,1) (g,2)};
\end{axis}
\draw[red,opacity=0.5] (tick no 0.base)
-- (tick no 2.base);
\end{tikzpicture}
In order to make the normalize the tick labels, pgfplots automatically inserts \strut. A \strut is a
TEX macro which inserts an empty box of “full height”, including anything above or below the text’s
baseline.
For standard axes, this key is basically equivalent to xticklabel=\strut\pgfmathprintnumber{\tick}.
Note that \strut is o↵ by default as enlarges the bounding box unnecessarily for pure numeric nodes.
This is particularly unexpected for polaraxis, but also for normal plots.
4.15.3 Tick Scaling - Common Factors In Ticks

/pgfplots/scaled ticks=true|false|base 10:hei|real:hnumi|manual:{hlabel i}{hcodei} (initially true)
/pgfplots/scaled x ticks=hone of the valuesi (initially true)
/pgfplots/scaled y ticks=hone of the valuesi (initially true)
/pgfplots/scaled z ticks=hone of the valuesi (initially true)
Allows to factor out common exponents in tick labels for linear axes. For example, if you have tick
labels 20000, 40000 and 60000, you may want to save some space and write 2, 4, 6 with a separate factor
‘·104 ’. Use ‘scaled ticks=true’ to enable this feature. In case of true, tick scaling will be triggered
if the data range is either too large or too small (see below).
3
\begin{tikzpicture}
2 \begin{axis}[scaled ticks=true]
(20000,0.0005)
(40000,0.0010)
1.5 (60000,0.0020)
};
\end{axis}
\end{tikzpicture}%
1
0.5
2 3 4 5 6
·104
3
2 · 10 \begin{tikzpicture}
\begin{axis}[scaled ticks=false]
3 (20000,0.0005)
1.5 · 10 (40000,0.0010)
(60000,0.0020)
};
1 · 10 3 \end{axis}
\end{tikzpicture}
4
5 · 10
20,000 30,000 40,000 50,000 60,000
The scaled ticks key is a style which simply sets scaled ticks for both, x and y.
The value base 10:hei allows to adjust the algorithm manually. For example, base 10:3 will divide
every tick label by 103 :
3
\begin{tikzpicture}
415 \begin{axis}[scaled ticks=base 10:3,
/pgf/number format/sci subscript]
{(-0.00001,2e12) (-0.00005,4e12) };
\end{axis}
\end{tikzpicture}
315
215
5 2 4 2 3 2 2 2 1 2
3
·10
Here, the sci subscript option simply saves space. In general, base 10:e will divide every tick by
10e . The e↵ect is not limited by the “too large or too small” decisions mentioned above.
The value real:hnumi allows to divide every tick by a fixed hnumi. For example, the following plot is
physically ranged from 0 to 2⇡, but the tick scaling algorithm is configured to divide every tick label by
⇡.
\begin{axis}[
xtick={0,1.5708,...,10},
domain=0:2*pi,
scaled x ticks={real:3.1415},
xtick scale label code/.code={$\cdot \pi$}]
0
\end{axis}
\end{tikzpicture}
1
0 0.5 1 1.5 2
·⇡
Setting scaled ticks=real:hnumi also changes the tick scale label code to
\pgfkeys{/pgfplots/xtick scale label code/.code=

{$\pgfkeysvalueof{/pgfplots/tick scale binop} \pgfmathprintnumber{#1}$}}.
The key tick scale binop is described below, it is set initially to \cdot.
A further – not very useful – example is shown below. Every x tick label has been divided by 2, every
y tick label by 3.
\begin{tikzpicture}
\begin{axis}[
33.33 scaled x ticks=real:2,
scaled y ticks=real:3]
(3, 9) \addplot {x^3};
\node[pin=135:{$(3,9)$}] at (3,9) {};
\end{axis}
0 \end{tikzpicture}
33.33
3 2 1 0 1 2 3
·2
The last option, scaled ticks=manual:{hlabel i}{hcodei} allows even more customization. It allows
full control over the displayed scaling label and the scaling code: hlabel i is used as-is inside of the tick
scaling label while hcodei is supposed to be a one-argument-macro which scales each tick. Example:
+65 535 % Preamble: \pgfplotsset{width=7cm,compat=1.13}

\begin{tikzpicture}
2 \begin{axis}[
% warning: the ’%’ signs are necessary (?)
scaled y ticks=manual:{$+65\,535$}{%
\pgfmathparse{#1-65535}%
},
yticklabel style={
1 /pgf/number format/fixed,
/pgf/number format/precision=1},
]
(0, 65535)
0 (13, 65535)
(14, 65536)
0 10 20 30 (15, 65537)
(30, 65537)
};
\end{axis}
\end{tikzpicture}
The example uses $+65\,535$ as tick scale label content. Furthermore, it defines the customized tick
label formula y (+6.5535 · 104 ) = y 65535 to generate y tick labels.
The hlabel i can be arbitrary. It is completely in user control. The second argument, hcodei is supposed
to be a one-argument-macro in which #1 is the current tick position in floating point representation.
The macro is expected to assign \pgfmathresult (as a number). The pgf manual [6] contains detailed
documentation about its math engine.
This feature may also be used do transform coordinates in case they can’t be processed with pgfplots:
transform them and supply a proper tick scaling method such that tick labels represent the original
range.
If hlabel i is empty, the tick scale label won’t be drawn (and no space will be occupied).
Tick scaling does not work for logarithmic axes.
/pgfplots/xtick scale label code/.code={h... i}

/pgfplots/ytick scale label code/.code={h... i}
/pgfplots/ztick scale label code/.code={h... i}
Allows to change the default code for scaled tick labels. The default is
\pgfplotsset{
xtick scale label code/.code={$\cdot 10^{#1}$}
}
More precisely, it is
\pgfplotsset{
xtick scale label code/.code={$\pgfkeysvalueof{/pgfplots/tick scale binop} 10^{#1}$}
}
and the initial value of tick scale binop is \cdot, but it can be changed to \times if desired.
If the code is empty, no tick scale label will be drawn (and no space is consumed).
/pgfplots/tick scale label code/.code={h... i}

A style which sets xtick scale label code and those for y and z.
/pgfplots/tick scale binop={hTEX math operator i} (initially \cdot)

Sets the binary operator used to display tick scale labels.
tick scale binop=\cdot
·10
\begin{axis}[
title=\texttt{tick scale
binop=\textbackslash cdot}]
\addplot
4 [mark=none,blue,samples=250,
domain=0:5]
{exp(10*x)};
\end{axis}
2 \end{tikzpicture}
0
0 2 4
tick scale binop=\times
⇥1021 \begin{tikzpicture}
\begin{axis}[
title=\texttt{tick scale
binop=\textbackslash times},
tick scale binop=\times]
4 \addplot
[mark=none,blue,samples=250,
domain=0:5]
{exp(10*x)};
2 \end{axis}
\end{tikzpicture}
0
0 2 4
/pgfplots/scale ticks below exponent={hexponenti} (initially -1)

Allows fine tuning of the ‘scaled ticks’ algorithm: if the axis limits are of magnitude 10e and
e <hexponenti, the common prefactor 10e will be factored out. The default is
/pgfplots/scale ticks above exponent={hexponenti} (initially 3)

Allows fine tuning of the ’scaled ticks’ algorithm: if the axis limits are of magnitude 10e and
e >hexponenti, the common prefactor 10e will be factored out.
4.15.4 Tick Fine-Tuning

The tick placement algorithm depends on a number of parameters which can be tuned to get better results.
/pgfplots/max space between ticks={hnumber i} (initially 35)

Configures the maximum space between adjacent ticks in full points. The suffix “pt” has to be omitted
and fractional numbers are not supported.
/pgfplots/try min ticks={hnumber i} (initially 4)

Configures a loose lower bound on the number of ticks. It should be considered as a suggestion, not a
tight limit. This number will increase the number of ticks if ‘max space between ticks’ produces too
few of them.
The total number of ticks may still vary because not all fractional numbers in the axis’ range are valid
tick positions.
/pgfplots/try min ticks log={hnumber i} (initially 3)

The same as try min ticks, but for logarithmic axis.
/pgfplots/tickwidth={hdimensioni} (initially 0.15cm)

/pgfplots/major tick length={hdimensioni} (initially 0.15cm)
Sets the length of major tick lines.
It can be accessed using \pgfkeysvalueof{/pgfplots/major tick length}.
/pgfplots/subtickwidth={hdimensioni} (initially 0.1cm)

/pgfplots/minor tick length={hdimensioni} (initially 0.1cm)
Sets the length of minor tick lines.
It can be accessed using \pgfkeysvalueof{/pgfplots/minor tick length}.
/pgfplots/xtick placement tolerance (initially 0.05pt)

/pgfplots/ytick placement tolerance (initially 0.05pt)
/pgfplots/ztick placement tolerance (initially 0.05pt)
Tick lines and labels will be placed if they are no more than this tolerance beyond the axis limits. This
threshold should be chosen such that it does not produce visible di↵erences while still providing fault
tolerance.
The threshold is given in paper units of the final figure.
/pgfplots/log basis x={hnumber i} (initially empty)

/pgfplots/log basis y={hnumber i} (initially empty)
/pgfplots/log basis z={hnumber i} (initially empty)
Allows to change the logarithms used for logarithmic axes.
Changing to a di↵erent log basis is nothing but a scale. However, it also changes the way tick labels are
displayed: they will also be shown in the new basis.
101
22
100
1
2
1
10
4
2
4 2 0 2 4 4 2 0 2 4
\begin{tikzpicture}
\begin{semilogyaxis}[log basis y=2,grid=major,samples at={-4,...,4}]
\addplot {2^x};
\end{semilogyaxis}
\end{tikzpicture}
~
\begin{tikzpicture}
\begin{semilogyaxis}[log basis y=10,samples at={-4,...,4}]
\addplot {2^x};
\end{semilogyaxis}
\end{tikzpicture}
The initial setting is ‘log basis x=’ which defaults to: the natural logarithm for any coordinates (basis
exp(1)), and the logarithm base 10 for the display of tick labels.
If the log basis is changed to something di↵erent than the empty string, the chosen logarithm will be
applied to any input coordinate (if the axis scale is log as well) and tick labels will be displayed in this
basis.
In other words: usually, you see log axes base 10 and that’s it. It is only interesting for coordinate
filters: the initial setting (with empty hnumber i) uses coordinate lists basis e although the display will
use basis 10 (i.e. it is rescaled). Any non-empty value hnumber i causes both, coordinate lists and display
to use hnumber i as basis for the logarithm. The javascript code of the clickable library will always
use the display basis (which is usally 10) when it computes slopes.
Technical remarks. When log basis x is used, the style log basis ticks={haxis char i} will be
installed (in this case log basis ticks=x). This style in turn will change log number format code.
Please note that xtickten will be used di↵erently now: it will provide the desired ticks in the new basis!
Despite the misleading name “ten”, xtickten={1,2,3,4} will yield ticks at 21 , 22 , 23 , 24 if log basis
x=2 has been set.
/pgfplots/hide obscured x ticks=true|false (initially true)

/pgfplots/hide obscured y ticks=true|false (initially true)
/pgfplots/hide obscured z ticks=true|false (initially true)
Allows to configure whether pgfplots should hide ticks and their labels if it assumes that they are
obscured by other axis lines.
4
\begin{tikzpicture}
2
\begin{axis}[
4 2 2 4
tiny,
2
axis x line=center,
4 axis y line=center,
%hide obscured x ticks=false,
%hide obscured y ticks=false,
]
\addplot {x};
\end{axis}
\end{tikzpicture}
4
\begin{tikzpicture}
2
\begin{axis}[
0
4 2 0 2 4
tiny,
2
axis x line=center,
4 axis y line=center,
hide obscured x ticks=false,
hide obscured y ticks=false,
]
\addplot {x};
\end{axis}
\end{tikzpicture}
Note that this key applies if only and only if one of axis x line, axis y line, or axis z line is not
box.
4.16 Grid Options

/pgfplots/xminorgrids=true|false (initially false)
/pgfplots/yminorgrids=true|false (initially false)
/pgfplots/zminorgrids=true|false (initially false)
/pgfplots/xmajorgrids=true|false (initially false)
/pgfplots/ymajorgrids=true|false (initially false)
/pgfplots/zmajorgrids=true|false (initially false)
/pgfplots/grid=minor|major|both|none (initially false)
Enables/disables di↵erent grid lines. Major grid lines are placed at the normal tick positions (see
xmajorticks) while minor grid lines are placed at minor ticks (see xminorticks).
This example employs the coordinates defined on page 184.
1
\begin{loglogaxis}[
2 xlabel={\textsc{Dof}},
10
ylabel={$L_2$ Error},
L2 Error
grid=major
3
10 ]
% see above for this macro:
10 4 \plotcoords
\end{loglogaxis}
\end{tikzpicture}
5
10
101 102 103 104 105 106

Dof
4
\begin{loglogaxis}[
grid=both,
tick align=outside,
tickpos=left]
5 {(100,1e-4) (500,1e-5) (1000,3e-6)};
{(100,1e-5) (500,4e-6) (1000,2e-6)};
\end{loglogaxis}
\end{tikzpicture}
102 103
Grid lines will be drawn before tick lines are processed, so ticks will be drawn on top of grid lines. You
can configure the appearance of grid lines with the styles
\pgfplotsset{grid style={help lines}} % modifies the style ‘every axis grid’

\pgfplotsset{minor grid style={color=blue}} % modifies the style ‘every minor grid’
\pgfplotsset{major grid style={thick}} %modifies the style ‘every major grid’
4.17 Custom Annotations

Often, one may want to add custom drawing elements or descriptive texts to an axis. These graphical
elements should be associated to some logical coordinate, grid point, or perhaps they should just be placed
somewhere into the axis.
pgfplots assists with the following ways when it comes to annotations:
1. You can explicitly provide any Tik Z instruction like \draw ... ; into the axis. Here, the axis cs
allows to provide coordinates of pgfplots.
Furthermore, rel axis cs allows to position Tik Z elements relatively (like “50% of the axis’ width).
4.17. CUSTOM ANNOTATIONS 331
2. pgfplots can automatically generate nodes at every coordinate using its nodes near coords feature.
3. pgfplots allows you to place nodes on a plot, using the \addplot ... node[pos=hfractioni] {};
feature.
This section explains all of the approaches, except for the nodes near coords feature which is documented
in its own section.
4.17.1 Accessing Axis Coordinates in Graphical Elements

Coordinate system axis cs
pgfplots provides a new coordinate system for use inside of an axis, the “axis coordinate system”,
axis cs.
Note that this coordinate system is actually the default one whenever one writes graphical elements in
an axis. There is just one place where you really need to provide this coordinate system explicitly: if
you have symbolic x coords.
Consequently, you can refer to this documentation as reference only; most standard use-cases work
directly (as of compat=1.11).
It can be used to draw any Tik Z-graphics at axis coordinates. It is used like
\draw
(axis cs:18943,2.873391e-05)
|- (axis cs:47103,8.437499e-06);
10 \tikzstyle{every pin}=[fill=white,
draw=black,
font=\footnotesize]
2
10 Bad! \begin{tikzpicture}
L2 Error
\begin{loglogaxis}[
xlabel={\textsc{Dof}},
3 ylabel={$L_2$ Error}]
10
4 (11, 6.887e-02)
10 Good! (71, 3.177e-02)
(351, 1.341e-02)
(1471, 5.334e-03)
101 102 103 104 105 106 (5503, 2.027e-03)
(18943, 7.415e-04)
Dof (61183, 2.628e-04)
(187903, 9.063e-05)
(553983, 3.053e-05)
};
\node[coordinate,pin=above:{Bad!}]
at (axis cs:5503,2.027e-03) {};
\node[coordinate,pin=left:{Good!}]
at (axis cs:187903,9.063e-05) {};
\end{loglogaxis}
\end{tikzpicture}
However, since axis cs is the default coordinate system of pgfplots, you can simply omit the prefix
axis cs: in coordinate descriptions56 .
56 As of pgfplots version 1.11 and compat=1.11. All older versions explicitly require the prefix; coordinates without the
prefix will be placed at the wrong position.

1
\begin{loglogaxis}[
2 xlabel=\textsc{Dof},
10
ylabel=$L_2$ Error
L2 Error
]
3
10 \draw
(1793,4.442e-05)
10 4 |- (4097,1.207e-05)
dy node[near start,left]
dx = 1.58 {$\frac{dy}{dx} = -1.58$};
5
10
(5, 8.312e-02)
101 102 103 104 (17, 2.547e-02)
(49, 7.407e-03)
Dof
(129, 2.102e-03)
(321, 5.874e-04)
(769, 1.623e-04)
(1793, 4.442e-05)
(4097, 1.207e-05)
(9217, 3.261e-06)
};
\end{loglogaxis}
\end{tikzpicture}
The e↵ect of axis cs is to apply any custom transformations (including symbolic x coords), loga-
rithms, data scaling transformations or whatever pgfplots usually does and provides a low level pgf
coordinate as result.
In case you need only one component (say, the y component) of such a vector, you can use the
\pgfplotstransformcoordinatey command, see Section 8.4 for details about basic level access.
The result of axis cs is always an absoute position inside of an axis. This means, in particular,
that adding two points has unexpected e↵ects: the expression (0,0) ++ (1,0) is not necessarily the
same as (1,0). The background for such unexpected e↵ects is that pgfplots applies a shifted linear
transformation which moves the origin in order to support its high accuracy and high data range
(compare the documentation of disabledatascaling).
In order to express relative positions (or lengths), you need to use axis direction cs.
Coordinate system axis direction cs

While axis cs allows to supply absolute positions, axis direction cs supplies directions. It allows
to express relative positions, includings lengths and dimensions, by means of axis coordinates.
As noted in the documentation for axis cs, adding two coordinates by means of the Tik Z ++ operator
may have unexpected e↵ects. The correct way for ++ operations is axis direction cs:
\begin{axis}
\draw[red,-stealth]
(1000,0)
-- % = line-to
++ % = calculate a vector sum
0.5
(axis direction cs:1000,0);
\addplot [only marks,mark=*]

coordinates { (1000,0) (2000,1) };
\end{axis}
0 \end{tikzpicture}
1,000 1,200 1,400 1,600 1,800 2,000
Here, the target of the red arrow is the position (2000,0) as expected.
Using relative positions is mainly useful for linear axes. Applying this command to log-axes might still
work, but it requires more care.
One use-case is to supply lengths – for example in order to support circle or ellipse paths. The
correct way to draw an ellipse in pgfplots would be to specify the two involved radii by means of
two (axis direction cs:hx,yi) expressions. In general, this is possible if you use the basic level
macros \pgfpathellipse and \pgfplotspointaxisdirectionxy. Please refer to the documentation
of \pgfplotspointaxisdirectionxy for two examples of drawing arbitrary ellipses by means of this
method.
Since drawing circles and ellipses inside of an axis is a common use-case, pgfplots automatically
communicates its coordinate system transformations to Tik Z: whenever you write \draw ellipse[x
radius=hx i,y radius=hyi], the arguments hx i and hyi are considered to be pgfplots direction vec-
tors and are handed over to axis direction cs. Consequently, ellipses with axis parallel radii are
straight-forward and use the normal Tik Z syntax:
2 % requires \pgfplotsset{compat=1.5.1} !
\begin{tikzpicture}
\begin{axis}[
1 xmin=-2.5, xmax=2.5,
ymin=-2.5, ymax=2.5,
xtick={-2,-1,0,1,2},
0
ytick={-2,-1,0,1,2},
grid=major,
1 ]
% standard tikz syntax:
\draw[black] (0,0)
2 ellipse [
x radius=1, y radius=2];
2 1 0 1 2
\draw[red] (0,0)
ellipse [rotate=90,
x radius=1, y radius=2];
% see \pgfplotspointaxisdirectionxy
% for arbitrary ellipses
\end{axis}
\end{tikzpicture}
Here, the two ellipses are specified as usual in Tik Z. pgfplots ensures that all necessary transformations
are applied to the two radii. Note that pgfplots usually has di↵erent axis scales for x and y. As a
consequence, the rotated red ellipse does not fit into the axis lines; we would need to use axis equal
to allow properly rotated ellipses.
Attention: this modification to circles and ellipses requires \pgfplotsset{compat=1.5.1}.

The same applies to circles: in the standard view, a circle with radius=r will appear as an ellipse due
to the di↵erent axis scales. Supplying axis equal results in true circles:
% requires \pgfplotsset{compat=1.5.1} !
\begin{tikzpicture}
0
\begin{axis}[tiny,enlargelimits,
xmin=-1,xmax=1,
1
ymin=-1,ymax=1,
1 0 1 xtick={-1,0,1},
ytick={-1,0,1},
grid=major,
]
\draw[blue] (0,0) circle[radius=1];
\end{axis}
\end{tikzpicture}
\begin{tikzpicture}
\begin{axis}[tiny,enlargelimits,
0
axis equal,
xmin=-1,xmax=1,
1
ymin=-1,ymax=1,
1 0 1 xtick={-1,0,1},
ytick={-1,0,1},
grid=major,
]
\draw[blue] (0,0) circle[radius=1];
\end{axis}
\end{tikzpicture}
In case you need access to axis direction cs inside of math expressions, you can employ the addi-
tional math function transformdirectionx. It does the same as axis direction cs, but only in x
direction. The result of transformdirectionx is a dimensionless unit which can be interpreted relative
to the current pgf x unit vector ex (see the documentation of \pgfplotstransformdirectionx for
details). There are the math commands transformdirectionx, transformdirectiony, and (if the
axis is three–dimensional) transformdirectionz. Each of them defines \pgfmathresult to contain
the result of \pgfplotstransformdirectionx (or its variants for y and z, respectively).
Coordinate system rel axis cs

The “relative axis coordinate system”, rel axis cs, uses the complete axis vectors as units. That
means ‘x = 0’ denotes the point on the lower x axis range and ‘x = 1’ the point on the upper x axis
range (see the remark below for x dir=reverse).
\begin{tikzpicture}
\begin{axis}
\addplot3[surf] {x^2 - y^2};

20 \draw (rel axis cs:0,0,1)
-- (rel axis cs:1,1,1);
\end{axis}
0 \end{tikzpicture}
20 5
4 0
2 0 2 4 5
\begin{tikzpicture}
\begin{axis}[
xlabel=$x$,
ylabel=$y$,
zlabel=$z$,
100 every axis x label/.style={
at={(rel axis cs:0.5,-0.15,-0.15)}},
z 0 every axis y label/.style={
at={(rel axis cs:1.15,0.5,-0.15)}},
100 5 every axis z label/.style={
at={(rel axis cs:-0.15,-0.15,0.5)}},
4 0 y ]
2 0
x 2 4 5 \addplot3[surf] {x*(1-x)*y};
\end{axis}
\end{tikzpicture}
Points identified by rel axis cs use the syntax

(rel axis cs:hx i,hyi) or
(rel axis cs:hx i,hyi,hz i)
where hx i, hyi and hz i are coordinates or constant mathematical expressions. The second syntax is only
available in three dimensional axes.
There is one specialty: if you reverse an axis (with x dir=reverse), points provided by rel axis cs
will be una↵ected by the axis reversal. This is intended to provide consistent placement even for reversed
axes. Use allow reversal of rel axis cs=false to disable this feature.
There is also a low–level interface to access the transformations and coordinates, see Section 8 on
page 518.
Predefined node current plot begin

This coordinate will be defined for every plot and can be used is htrailing path commandsi or after a
plot. It is the first coordinate of the current plot.
Predefined node current plot end

This coordinate will be defined for every plot. It is the last coordinate of the current plot.
/pgfplots/allow reversal of rel axis cs=true|false (initially true)

A fine-tuning key which specifies how to deal with x dir=reverse and rel axis cs and ticklabel
cs.
The initial configuration true means that points placed with rel axis cs and/or ticklabel cs will
be at the same position inside of the axes even if its ordering has been reversed. The choice false will
disable the special treatment of x dir=reverse.
4.17.2 Placing Nodes on Coordinates of a Plot

The \addplot command is not only used for pgfplots, it can also carry additional drawing instructions
which are handed over to Tik Z after the plot’s path is complete. Among others, this can be used to add
further nodes on the path.
/tikz/pos={hfractioni}
The hfractioni identifies a part of the recently completed plot if it is used before the trailing semicolon:
⇡/2 % Preamble: \pgfplotsset{width=7cm,compat=1.13}

\begin{axis}
\addplot[blue,domain=0:360] {sin(x)}
[yshift=8pt]
node[pos=0] {$0$}
0 ⇡ 2⇡ node[pos=0.25] {$\pi/2$}
0
node[pos=0.5] {$\pi$}
node[pos=0.75] {$3/2\pi$}
node[pos=1] {$2\pi$}
;
3/2⇡ \end{axis}
1 \end{tikzpicture}
0 100 200 300
Here, the [yshift=8pt] tells Tik Z to shift all following nodes upwards. The node[pos=0] {$0$}
instruction tells Tik Z to add a text node at 0% of the recently completed plot. The relative position
0% (pos=0) refers to the first coordinate which has been seen by pgfplots, and 100% (pos=1) refers
to the last coordinate. Any value between 0 and 1 is interpolated in-between. Note that all these nodes
belong to the plot’s visualization (which is terminated by the semicolon). Consequently, all these nodes
inherit the same graphic settings (like color choices).
The position on the plot is computed by pgfplots using logical coordinates. That means: it computes
the overall length of the curve before the curve is projected to screen coordinates and identifies the
desired position57 . Afterwards, it projects the final position to screen coordinates. Thus, the position
identifies a location on the plot which is always the same, even in case of a rotated three-dimensional
axis. pgfplots will linearly interpolate the fraction between successive coordinates.
Valid choices for hfractioni are any numbers in the range [0, 1].
Note that the precise meaning of pos depends on the current plot handler: for most plot handlers, it
defaults to linear interpolation (as in the examples above). For only marks, scatter, ybar, xbar, ybar
57 This can be a time-consuming process. Consider using the external library if you have lots of such figures.
interval, and xbar interval, it snaps to the nearest encountered coordinate. In this context, “snap
to nearest” means that pos=p refers to the coordinate with index i = round(p · N ) where N is the total
number of points:
Snap to nearest for scatter plots
\begin{axis}[title=Snap to nearest for scatter plots]
3 \addplot+[only marks]
coordinates {(0,0) (1,1) (2,2) (3,3)}
0.5 node[pos=0, pin=0 :0 ] {}
node[pos=0.1, pin=90 :0.1 ] {}
2 0.75 node[pos=0.2, pin=200:0.2 ] {}
node[pos=0.3, pin=135:0.3 ] {}
0.3 node[pos=0.4, pin=0 :0.4 ] {}
node[pos=0.5, pin=60 :0.5 ] {}
1 0.4 node[pos=0.75,pin=180:0.75] {}
0.1 0.2 node[pos=1, pin=90 :1 ] {}
;
0 0 \end{axis}
\end{tikzpicture}
0 1 2 3
the previous example shows that pos=p maps to one of the four available coordinates, namely the one
whose index is closest to p · N . Note that in such a case, the distance between coordinates is irrelevant
– only the coordinate index counts.
Note that the fact that pgfplots uses logical coordinates to compute the target positions can produce
unexpected e↵ects if x and y axis operate on a di↵erent scales. Suppose, for example, that x is always of
order 103 whereas y is of order 10 3 . In such a scenario, the y coordinate have no significant contribution
to the curve’s length – although the rescaled axes clearly show “significant” y dynamics. Consider using
axis equal together with pos to produce comparable e↵ects.
/tikz/sloped (initially false)

Providing the Tik Z key sloped to a node identified by pos causes it to be rotated such that it adapts
to the plot’s gradient.
⇡/2 % Preamble: \pgfplotsset{width=7cm,compat=1.13}

\begin{axis}
\addplot[blue,domain=0:360,samples=31] {sin(x)}
[every node/.style={yshift=8pt},sloped]
node[pos=0] {$0$}
node[pos=0.25] {$\pi/2$}
2⇡
⇡
0
0
node[pos=0.75] {$3/2\pi$}
;
3/2⇡ \end{axis}
1 \end{tikzpicture}
0 100 200 300
Note that the sequence in which sloped and shift transformations are applied is important: if shifts
are applied first (as would be the case without the every node/.style construction), the shifts do
not respect the rotation. If sloped is applied first, any subsequent shifts will be applied in the rotated
coordinates. Thus, the case every node/.style={yshift=8pt} shifts every node by 8pt in direction
of its normal vector.
The sloped transformation is based on the gradient between two points (the two points adjacent to
pos). Consequently, it inherits any sampling weaknesses. To see this, consider the example above with
a di↵erent number of samples:
⇡ /2 % Preamble: \pgfplotsset{width=7cm,compat=1.13}
1 % same as above with different number of samples
\begin{tikzpicture}
\begin{axis}
\addplot[blue,domain=0:360,samples=25] {sin(x)}
[every node/.style={yshift=8pt},sloped]
node[pos=0] {$0$}
2⇡
⇡
0 0
node[pos=0.25] {$\pi/2$}
node[pos=0.75] {$3/2\pi$}
1 3/2⇡ ;
\end{axis}
\end{tikzpicture}
0 100 200 300
Here, the two extreme points have small slopes due to the sampling. While this does not seriously a↵ect
the quality of the plot, it has a huge impact on the transformation matrizes. Keep this in mind when
you work with sloped (perhaps it even helps to add a further rotate argument).
/tikz/allow upside down=true|false (initially false)

If /tikz/sloped is enabled and one has some difficult line plot, the transformation may cause nodes to
be drawn upside down. The default configuration allow upside down=false will switch the rotation
matrix, whereas allow upside down allows this case.
/tikz/pos segment={hsegment index i} (initially empty)

Occasionally, one has a single plot which consists of multiple segments (like those generated by empty
line=jump or contour prepared). The individual segments will typically have di↵erent lengths, so it
is tedious to identify a position on one of these segments.
If pos segment=hsegment index i is non-empty, the key pos=hfractioni is interpreted relatively to the
provided segment rather than the whole plot. The argument hsegment index i is an integer, where 0
denotes the first segment.
0.6 \begin{axis}[tiny]
0.4 \addplot coordinates {
0.2
0 0.5 1 (0,0) (1,0)
0
0 0.5 1 1.5 2 (1,1) (2,1)}
[pos segment=0,yshift=7pt,font=\footnotesize]
node[pos=0] {0}
node[pos=0.5] {0.5}
node[pos=1] {1};
\end{axis}
\end{tikzpicture}
Here, the plot has two segments. However, all three annotation nodes are placed with pos segment=0.
\begin{tikzpicture}
\begin{axis}
0. \addplot3[contour gnuplot,domain=0:1] {x*y}
8
[sloped,
allow upside down,
0
0 .6 pos segment=2,
0.
5 every node/.style={yshift=7pt}]
0.5 node[pos=0] {0}
0 .4 1 node[pos=0.5] {0.5}
1 node[pos=1] {1}
0 .2 ;
0.2 0.4 0.5
0.6 0.8 \end{axis}
1 \end{tikzpicture}
This plot has four segments (which are generated automatically by the plot handler). The annotation
nodes are placed on the third segment, where sloped causes them to be rotated, allow upside down
improves the rendering of the ‘0’, and every node/.style install a shift in direction of the normal
vector (see the documentation of sloped for details).
Occasionally, one wants to place a node using pos and one wants to typeset the coordinates of that point
inside of the node. This can be accomplished using \pgfplotspointplotattime:
\pgfplotspointplotattime
\pgfplotspointplotattime{hfractioni}
This command is part of the pos={hfractioni} implementation: it defines the current point of pgf to
hfractioni of the current plot. Without an argument in curly braces, \pgfplotspointplotattime will
take the current argument of the pos key.
Thus, the command computes the basic pgf coordinates – but it also returns the logical coordinates of
the resulting point into the following keys:
/data point/x (no value)
/data point/y (no value)
/data point/z (no value)
After \pgfplotspointplotattime returns, these macros contain the x, y, and z coordinates of the
resulting point. They can be used by means of \pgfkeysvalueof{/data point/x}, for example.
\begin{axis}
\addplot {x}
(2, 2) [left,/pgf/number format/relative=0]
node[pos=0.5] {%
(0, 0) \pgfplotspointplotattime
0
$(\pgfmathprintnumber
{\pgfkeysvalueof{/data point/x}},
( 2.5, 2.5) \pgfmathprintnumber
{\pgfkeysvalueof{/data point/y}})$
}
5 node[pos=0.25] {%
6 4 2 0 2 4 6 $(\pgfmathprintnumber
\pgfmathprintnumber
}
node[pos=0.7,pin=180:{%
\pgfplotspointplotattime{0.7}
\pgfmathprintnumber
}] {}
;
\end{axis}
\end{tikzpicture}
In the example above, three nodes have been placed using di↵erent pos= arguments. Invoking
\pgfplotspointplotattime inside of the associated node’s body checks if pos already has a value
and uses that value. The third node displays the coordinates inside of a pin. Due to inter-
nals of Tik Z, the pin knows nothing about the pos=0.7 argument of its enclosing node, so we
need to replicate the ‘0.7’ argument for \pgfplotspointplotattime{0.7}. The /pgf/number
format/relative=0 style causes the number printer to round relative to 100 (compare against the
same example without this style).
In case you have symbolic x coords (or any other x coord inv tafo which produces non-numeric
results), the output stored in /data point/x will be the symbolic expression:
\begin{axis}[symbolic x coords={A,B,C,D}]
\addplot coordinates {(A,0) (B,1) (C,1) (D,2)}
[left]
(C, 1.19) node[pos=0.3] {%
1
$(\pgfkeysvalueof{/data point/x},
(B, 0.81)
\pgfmathprintnumber
}
node[pos=0.7,pin=180:{%
0 \pgfplotspointplotattime{0.7}
$(\pgfkeysvalueof{/data point/x},
A B C D \pgfmathprintnumber
}] {}
;
\end{axis}
\end{tikzpicture}
In that specific case, you have to avoid \pgfmathprintnumber since the argument is no number.
Note that symbolic x coords cannot return fractions between, say, A and B as you would expect.
However, the point will still be placed at the fractional position (unless you have a scatter or bar
plot).
The computation of coordinates for the pos feature is computationally expensive for plots with
many points. To reduce time, pgfplots will cache computed values: invoking the command
\pgfplotspointplotattime multiple times with the same argument will reuse the computed value.
4.17.3 Placing Decorations on Top of a Plot

Tik Z comes with the powerful decorations library (or better: set of libraries). Decorations allow to replace
or extend an existing path by means of fancy additional graphics. An introduction into the decorations
functionality of Tik Z is beyond the scope of this manual and the interested reader should read the associated
section in [6].
This section shows how to use decorations to enhance plots in pgfplots. Suppose you have some
graphics for which you would like to add “direction pointers”:
Undecorated Graphics \begin{tikzpicture}[]
% An undecorated graphics with a lot of
% pretty-printing styles:
\begin{axis}[
1 axis lines=middle,
title=Undecorated Graphics,
xmin=-2, xmax=2, ymin=-2, ymax=2,
xtick={-1,1}, ytick={-1,1},
% this disables the standard
1 1 % tick label *text* (but not the line)
yticklabel=\ ,
1 % this generates custom y labels to implement
% individual styles for every tick:
\node[below left] at (axis cs:0,-1) {$-1$};
\node[above left] at (axis cs:0,1) {$1$};
},
]
\addplot[blue,samples=100,domain=0:2*pi]
({sin(deg(2*x))}, {sin(deg(x))});
\end{axis}
\end{tikzpicture}
Our aim is to add short pointers indicating the direction of the parameterization.
The solution is to use \usetikzlibrary{decorations.markings} and a decoration inside of \addplot:
Decorated Graphics % requires \usetikzlibrary{decorations.markings}
\begin{tikzpicture}[]
% Same as in previous example, but with decorations:
\begin{axis}[axis lines=middle,
1 title=Decorated Graphics,
xmin=-2, xmax=2, ymin=-2, ymax=2,
xtick={-1,1}, ytick={-1,1},
% this disables the standard
% tick label *text* (but not the line)
1 1 yticklabel=\ ,
% this generates custom y labels to implement
1 % individual styles for every tick:
\node[below left] at (axis cs:0,-1) {$-1$};
\node[above left] at (axis cs:0,1) {$1$};
},
]
\addplot[blue,samples=100,domain=0:2*pi,
postaction={decorate},% ------
decoration={markings, % ------
mark=at position 0.25 with {\arrow{stealth}},
mark=at position 0.5 with {\arrow{stealth}},
mark=at position 0.75 with {\arrow{stealth}}}
]
({sin(deg(2*x))}, {sin(deg(x))});
\end{axis}
\end{tikzpicture}
The only changes are in the option list for \addplot: it contains a postaction={decorate} which activates
the decoration (without replacing the original path) and some specification decoration containing details
about how to decorate the path.
A discussion of details of the decorations libraries is beyond the scope of this manual (see [6] for details),
but the main point is to add the required decorations to \addplot and its option list.
4.18 Style Options

4.18.1 All Supported Styles
pgfplots provides many styles to customize its appearance and behavior. They can be defined and changed
in any place where keys are allowed. Furthermore, own styles are defined easily.
Key handler hkeyi/.style={hkey-value-listi}

Defines or redefines a style hkeyi. A style is a normal key which will set all options in hkey-value-listi
when it is set.
Use \pgfplotsset{hkeyi/.style={hkey-value-listi}} to (re)define a style hkeyi in the namespace
/pgfplots.
Key handler hkeyi/.append style={hkey-value-listi}

Appends hkey-value-listi to an already existing style hkeyi. This is the preferred method to change the
predefined styles: if you only append, you maintain compatibility with future versions.
Use \pgfplotsset{hkeyi/.append style={hkey-value-listi}} to append hkey-value-listi to the style
hkeyi. This will assume the prefix /pgfplots.
Styles installed for linear/logarithmic axis

/pgfplots/every axis (style, initially empty)
Installed at the beginning of every axis. Tik Z options inside of it will be used for anything inside of the
axis rectangle and any axis descriptions.
/pgfplots/every axis post (style, initially empty)

A style which is applied right after arguments provided to an axis are processed.
4.18. STYLE OPTIONS 341
In the following example, such a style is used to override the xmin and xmax options provided as
arguments to \begin{axis}[...]:
\begin{tikzpicture}
\pgfplotsset{
every axis post/.style={
xmin=0,xmax=1,
},
}
\begin{axis}[
xmin=-1,xmax=2,
ymin=0,ymax=1]
...
\end{axis}
\end{tikzpicture}
It is processed right after the arguments of \begin{axis}, but before styles like yticklabel style
etc. are evaluated.
/pgfplots/every semilogx axis (style, initially empty)

Installed at the beginning of every plot with linear x axis and logarithmic y axis, but after ‘every axis’.
/pgfplots/every semilogy axis (style, initially empty)

Likewise, but with interchanged roles for x and y.
/pgfplots/every loglog axis (style, initially empty)

Installed at the beginning of every log–log plot.
/pgfplots/every linear axis (style, initially empty)

Installed at the beginning of every plot with normal axis scaling.
Styles installed for single plots

/pgfplots/every axis plot (style, initially empty)
Installed for each plot. This style may contain options like samples, gnuplot parameters, error bars and
it may contain options which a↵ect the final drawing commands.
/pgfplots/every axis plot post (style, initially empty)

This style is similar to every axis plot in that is applies to any drawing command in \addplot.
However, it is set after any user defined styles or cycle list options.
ex \pgfplotsset{
every axis plot post/.append style=
2 {mark=none}}
\begin{axis}[
legend style={
1 at={(0.03,0.97)},anchor=north west},
domain=0:1]
\addplot {x^2};
\addplot {exp(x)};
0 \legend{$x^2$,$e^x$}
\end{axis}
0 0.2 0.4 0.6 0.8 1 \end{tikzpicture}
/pgfplots/every axis plot no # (style, initially empty)

Used for every #th plot where # = 0, 1, 2, 3, 4, . . . .
/pgfplots/every forget plot (style, initially empty)

Used for every plot which has forget plot activated.
/pgfplots/forget plot style={hkey-value-listi}

An abbreviation for every forget plot/.append style={hkey-value-listi}.
It appends options to the already existing style every forget plot.
Styles for axis descriptions

/pgfplots/every axis label (style, initially empty)
Used for all axis label (like xlabel and ylabel).
/pgfplots/label style={hkey-value-listi}
An abbreviation for every axis label/.append style={hkey-value-listi}.
It appends options to the already existing style every axis label.
/pgfplots/every axis x label (style, no value)
/pgfplots/every axis y label (style, no value)
/pgfplots/every axis z label (style, no value)
Used only for x, y, or z labels, respectively and installed after ‘every axis label’.
The initial settings are set by xlabel absolute and its variants (if the initial configuration compat=pre
1.3 is active) or xlabel near ticks which provides the better spacing as it incorporates the tick label
sizes to compute the position.
Attention: These styles will be overwritten by axis x line and/or axis y line. Please remember
to place your modifications after the axis line variations.
/pgfplots/x label style={hkey-value-listi}
/pgfplots/y label style={hkey-value-listi}
/pgfplots/z label style={hkey-value-listi}
/pgfplots/xlabel style={hkey-value-listi}
/pgfplots/ylabel style={hkey-value-listi}
/pgfplots/zlabel style={hkey-value-listi}
Di↵erent abbreviations for every axis x label/.append style={hkey-value-listi} (or the respec-
tive styles for y, every axis y label/.append style={hkey-value-listi}, and z, every axis z
label/.append style={hkey-value-listi}).
/pgfplots/every axis title (style, no value)
Used for any axis title. The at=(hx,yi) syntax will place the title using axis description cs.
\pgfplotsset{every axis title/.style={at={(0.5,1)},above,yshift=6pt}}
To be more precise, the yshift doesn’t use the hardcoded 6pt: it uses the value of
/pgfplots/every axis title shift={hdefault shifti} (initially 6pt)
which can be reset if needed.
/pgfplots/title style={hkey-value-listi}
An abbreviation for every axis title/.append style={hkey-value-listi}.
It appends options to the already existing style every axis title.
/pgfplots/every axis legend (style, no value)
Installed for each legend. As described for axis description cs, the legend’s position can be placed
using coordinates between 0 and 1 (it employs axis description cs automatically).
\pgfplotsset{every axis legend/.style={
cells={anchor=center},
anchor=north east,
shape=rectangle,
fill=white,draw=black,
at={(0.98,0.98)}}}
/pgfplots/legend style={hkey-value-listi}
An abbreviation for every axis legend/.append style={hkey-value-listi}.
It appends options to the already existing style every axis legend.
/pgfplots/every legend image post (style, no value)

Allows to change the appearance of the small legend images after the options of the plot style have been
applied. Thus, legend formatting can be changed independently of the plot style using every legend
image post.
This key is also documented on page 241.
/pgfplots/legend image post style={hkey-value-listi}

An abbreviation for every legend image post/.append style={hkey-value-listi}.
It appends options to the already existing style every legend image post.
/pgfplots/every legend to name picture (style, no value)

A style for use with legend to image, see the documentation therein.
/pgfplots/every colorbar (style, no value)

A style to change the colorbar. See page 264 for the reference documentation of every colorbar.
/pgfplots/colorbar style={hkey-value-listi}
An abbreviation for every colorbar/.append style={hkey-value-listi}.
It appends options to the already existing style every colorbar.
Styles for axis lines

/pgfplots/every outer x axis line (style, initially empty)
/pgfplots/every outer y axis line (style, initially empty)
/pgfplots/every outer z axis line (style, initially empty)
Installed for every axis line which lies on the outer box.
If you want arrow heads, you may also need to check the separate axis lines boolean key.
/pgfplots/every inner x axis line (style, initially empty)

/pgfplots/every inner y axis line (style, initially empty)
/pgfplots/every inner z axis line (style, initially empty)
Installed for every axis line which is drawn using the center or middle options.
/pgfplots/axis line style={hkey-value-listi}

/pgfplots/inner axis line style={hkey-value-listi}
/pgfplots/outer axis line style={hkey-value-listi}
/pgfplots/x axis line style={hkey-value-listi}
/pgfplots/y axis line style={hkey-value-listi}
/pgfplots/z axis line style={hkey-value-listi}
These options modify parts of the axis line styles. They append options to every inner x axis line
and every outer x axis line and the respective y/z variants.
Please refer to Section 4.9.9 on page 253 for details about styles for axis lines.
/pgfplots/every 3d box foreground (style, no value)

Installed for the parts drawn by 3d box=complete. This a↵ects axis lines, tick lines and grid lines
drawn in the foreground. The background drawing operations have already been done when this style
is evaluated.
/pgfplots/3d box foreground style={hkey-value-listi}

An abbreviation for every 3d box foreground/.append style={hkey-value-listi}.
It appends options to the already existing style every 3d box foreground.
/pgfplots/every colorbar sampled line (style, no value)

To be used in conjunction with colorbar sampled line, see the documentation therein.
/pgfplots/colorbar sampled line style={hkey-value-listi}

An abbreviation for every colorbar sampled line/.append style={hkey-value-listi}.
It appends options to the already existing style every colorbar sampled line.
Styles for ticks

/pgfplots/every tick (style, initially very thin,gray)
Installed for each of the small tick lines.
/pgfplots/tick style={hkey-value-listi}
An abbreviation for every tick/.append style={hkey-value-listi}.
It appends options to the already existing style every tick.
/pgfplots/every minor tick (style, initially empty)

Used for each minor tick line, installed after ‘every tick’.
/pgfplots/minor tick style={hkey-value-listi}

An abbreviation for every minor tick/.append style={hkey-value-listi}.
It appends options to the already existing style every minor tick.
/pgfplots/every major tick (style, initially empty)

Used for each major tick line, installed after ‘every tick’.
/pgfplots/major tick style={hkey-value-listi}

An abbreviation for every major tick/.append style={hkey-value-listi}.
It appends options to the already existing style every major tick.
/pgfplots/every tick label (style, initially empty)

Used for each x and y tick labels.
/pgfplots/tick label style={hkey-value-listi}

/pgfplots/ticklabel style={hkey-value-listi}
Di↵erent abbreviations for every tick label/.append style={hkey-value-listi} (or the respective
styles for y, every tick label/.append style={hkey-value-listi}, and z, every tick label/.append
style={hkey-value-listi}).
/pgfplots/every x tick label (style, initially empty)

/pgfplots/every y tick label (style, initially empty)
/pgfplots/every z tick label (style, initially empty)
Used for each x (or y or z, respectively) tick label, installed after ‘every tick label’.
/pgfplots/x tick label style={hkey-value-listi}

/pgfplots/y tick label style={hkey-value-listi}
/pgfplots/z tick label style={hkey-value-listi}
/pgfplots/xticklabel style={hkey-value-listi}
/pgfplots/yticklabel style={hkey-value-listi}
/pgfplots/zticklabel style={hkey-value-listi}
Di↵erent abbreviations for every x tick label/.append style={hkey-value-listi} (or the respec-
tive styles for y, every y tick label/.append style={hkey-value-listi}, and z, every z tick
label/.append style={hkey-value-listi}).
/pgfplots/every x tick scale label (style, no value)

/pgfplots/every y tick scale label (style, no value)
/pgfplots/every z tick scale label (style, no value)

Configures placement and display of the nodes containing the order of magnitude of tick labels, see
Section 4.15.3 for more information about scaled ticks.
The initial settings with compat=1.8 or higher are
\pgfplotsset{
every x tick scale label/.style={
at={(xticklabel cs:0.9,5pt)},
anchor=near xticklabel,
inner sep=0pt},
every y tick scale label/.style={
at={
(yticklabel* cs:1.03,-0.3em)},
/pgfplots/near ticklabel align=outside,
anchor=near yticklabel opposite,
inner sep=0pt},
every z tick scale label/.style={
at={(zticklabel* cs:1.2,-0.3em)},
anchor=near zticklabel,
inner sep=0pt},
/pgfplots/x tick scale label style={hkey-value-listi}

/pgfplots/y tick scale label style={hkey-value-listi}
/pgfplots/z tick scale label style={hkey-value-listi}
An abbreviation for every x tick scale label/.append style={hkey-value-listi} (or the respective
styles for y, every y tick scale label/.append style={hkey-value-listi}, and the z–axis, every z
tick scale label/.append style={hkey-value-listi}).
It appends options to the already existing style every x tick scale label.
/pgfplots/every x tick (style, initially empty)

/pgfplots/every y tick (style, initially empty)
/pgfplots/every z tick (style, initially empty)
Installed for tick lines on either x or y axis.
/pgfplots/xtick style={hkey-value-listi}
/pgfplots/ytick style={hkey-value-listi}
/pgfplots/ztick style={hkey-value-listi}
An abbreviation for every x tick/.append style={hkey-value-listi} (or the respective styles
for y, every y tick/.append style={hkey-value-listi}, and the z–axis, every z tick/.append
style={hkey-value-listi}).
It appends options to the already existing style every x tick.
/pgfplots/every minor x tick (style, initially empty)

/pgfplots/every minor y tick (style, initially empty)
/pgfplots/every minor z tick (style, initially empty)
Installed for minor tick lines on either x or y axis.
/pgfplots/minor x tick style={hkey-value-listi}

/pgfplots/minor y tick style={hkey-value-listi}
/pgfplots/minor z tick style={hkey-value-listi}
An abbreviation for every minor x tick/.append style={hkey-value-listi} (or the respective styles
for y, every minor y tick/.append style={hkey-value-listi}, and the z–axis, every minor z
tick/.append style={hkey-value-listi}).
It appends options to the already existing style every minor x tick.
/pgfplots/every major x tick (style, initially empty)

/pgfplots/every major y tick (style, initially empty)
/pgfplots/every major z tick (style, initially empty)
Installed for major tick lines on either x or y axis.
/pgfplots/major x tick style={hkey-value-listi}

/pgfplots/major y tick style={hkey-value-listi}
/pgfplots/major z tick style={hkey-value-listi}
An abbreviation for every major x tick/.append style={hkey-value-listi} (or the respective styles
for y, every major y tick/.append style={hkey-value-listi}, and the z–axis, every major z
It appends options to the already existing style every major x tick.
/pgfplots/every extra x tick (style, no value)

/pgfplots/every extra y tick (style, no value)
/pgfplots/every extra z tick (style, no value)
Allows to configure the appearance of ‘extra x ticks’. This style is installed before touching the first
extra x tick. It is possible to set any option which a↵ects tick or grid line generation.
\pgfplotsset{
every extra x tick/.style={/pgfplots/log identify minor tick positions=true},
every extra y tick/.style={/pgfplots/log identify minor tick positions=true}}
Useful examples are shown below.
\pgfplotsset{every extra x tick/.append style={grid=major}}

\pgfplotsset{every extra x tick/.append style={major tick length=0pt}}
\pgfplotsset{every extra x tick/.append style={/pgf/number format=sci subscript}}
\pgfplotsset{extra x tick style={
grid style={
color=red,
},
tickwidth=3mm,
% the initial ’every tick style’ defines a ’line width’.
% this here redefines it:
tick style={
line width=2mm,
},
}
}
/pgfplots/extra x tick style={hkey-value-listi}

/pgfplots/extra y tick style={hkey-value-listi}
/pgfplots/extra z tick style={hkey-value-listi}
An abbreviation for every extra x tick/.append style={hkey-value-listi} (or the respective styles
for y, every extra y tick/.append style={hkey-value-listi}, and the z–axis, every extra z
It appends options to the already existing style every extra x tick.
/pgfplots/extra tick style={hkey-value-listi}

An abbreviation which appends hkey-value-listi to every extra x tick, every extra y tick and
every extra z tick.
Styles for grid lines

/pgfplots/every axis grid (style, initially thin,black!25)
Used for each grid line.
/pgfplots/grid style={hkey-value-listi}
An abbreviation for every axis grid/.append style={hkey-value-listi}.
It appends options to the already existing style every axis grid.
/pgfplots/every minor grid (style, initially empty)

Used for each minor grid line, installed after ‘every axis grid’.
/pgfplots/minor grid style={hkey-value-listi}

An abbreviation for every minor grid/.append style={hkey-value-listi}.
It appends options to the already existing style every minor grid.
/pgfplots/every major grid (style, initially empty)
Likewise, for major grid lines.
/pgfplots/major grid style={hkey-value-listi}
An abbreviation for every major grid/.append style={hkey-value-listi}.
It appends options to the already existing style every major grid.
/pgfplots/every axis x grid (style, initially empty)
/pgfplots/every axis y grid (style, initially empty)
/pgfplots/every axis z grid (style, initially empty)
Used for each grid line in either x or y direction.
/pgfplots/x grid style={hkey-value-listi}
/pgfplots/y grid style={hkey-value-listi}
/pgfplots/z grid style={hkey-value-listi}
An abbreviation for every axis x grid/.append style={hkey-value-listi} (or the respective
styles for y, every axis y grid/.append style={hkey-value-listi}, and the z–axis, every axis z
grid/.append style={hkey-value-listi}).
It appends options to the already existing style every axis x grid.
/pgfplots/every minor x grid (style, initially empty)
/pgfplots/every minor y grid (style, initially empty)
/pgfplots/every minor z grid (style, initially empty)
Used for each minor grid line in either x or y direction.
/pgfplots/minor x grid style={hkey-value-listi}
/pgfplots/minor y grid style={hkey-value-listi}
/pgfplots/minor z grid style={hkey-value-listi}
An abbreviation for every minor x grid/.append style={hkey-value-listi} (or the respective styles
for y, every minor y grid/.append style={hkey-value-listi}, and the z–axis, every minor z
It appends options to the already existing style every minor x grid.
/pgfplots/every major x grid (style, initially empty)
/pgfplots/every major y grid (style, initially empty)
/pgfplots/every major z grid (style, initially empty)
Used for each major grid line in either x or y direction.
/pgfplots/major x grid style={hkey-value-listi}
/pgfplots/major y grid style={hkey-value-listi}
/pgfplots/major z grid style={hkey-value-listi}
An abbreviation for every major x grid/.append style={hkey-value-listi} (or the respective styles
for y, every major y grid/.append style={hkey-value-listi}, and the z–axis, every major z
It appends options to the already existing style every major x grid.
Styles for error bars

/pgfplots/every error bar (style, initially thin)
Installed for every error bar.
/pgfplots/error bars/error bar style={hkey-value-listi}
An abbreviation for every error bar/.append style={hkey-value-listi}.
It appends options to the already existing style every error bar.
4.18.2 (Re)Defining Own Styles

Use \pgfplotsset{hstyle namei/.style={hkey-value-listi}} to create own styles. If hstyle namei exists
already, it will be replaced. Please note that it is not possible to use the Tik Z-command \tikzstyle{hstyle
namei}=[] in this context58 .
1 \pgfplotsset{my personal style/.style=
{grid=major,font=\large}}
\begin{tikzpicture}
\begin{axis}[my personal style]
0.5 \end{axis}
\end{tikzpicture}
0
0 0.2 0.4 0.6 0.8 1
4.19 Alignment Options

4.19.1 Basic Alignment
Alignment works with two main methods: a coordinate where the axis shall be drawn and an “anchor”
inside of the axis which shall be drawn at this particular coordinate. This methodology is common for each
Tik Z node – and an axis is nothing but a (special) Tik Z node. The coordinate can be specified using the at
key, while the anchor can be specified with the anchor key. In most cases, it is sufficient to provide only an
anchor – unless one needs more than one axis in the same picture environment.
/pgfplots/at={hcoordinate expressioni}
Assigns a position for the complete axis image. This option works similarly to the at-option of
\node[at={hcoordinate expressioni}], see [6]. The common syntax is at={(hx,yi)}.
The idea is to provide an hcoordinate expressioni where the axis will be placed. The axis’ anchor will
be placed at hcoordinate expressioni.
/pgfplots/anchor={hnamei} (initially south west)

Chooses one of the di↵erent possible positions inside of an axis which is placed with at. The at key
defines the position where to place the axis inside of the embedding picture, the anchor key defines
which point of the axis shall be positioned by ‘at’. The initial configuration assumes at={(0,0)}. Thus,
anchor=center will place the axis’ center at the logical picture position (0, 0). Similarly, anchor=south
west will position the lower left corner of the axis at (0, 0).
For users who are familiar with Tik Z: an axis is actually a very special node, so anchors work as in [6].
Anchors are useful in conjunction with horizontal or vertical alignment of plots, see the examples below.
There are four sets of anchors available: anchors positioned on the axis bounding box, anchors on the
outer bounding box and anchors which have one coordinate on the outer bounding box and the other
one at a position of the axis rectangle. Finally, one can place anchors near the origin.
In more detail, we have anchors on the axis rectangle (the bounding box around the axis)59 ,
58 This was possible in a previous version and is still supported for backwards compatibility. But in some cases, it may not
work as expected.
59 Versions prior to pgfplots v.1.3 did not use the bounding box of the axis, they used axis coordinates to orient these
anchors. This has been fixed. If you really want to undo the bugfix, see compat/anchors.
4.19. ALIGNMENT OPTIONS 349
(s.north)
(s.north west)
A test plot. (s.north east)
4
·10
2 f (x)
(s.center)
g(x)
(s.west)
0 (s.east)
y
2
4
40 20 0 20 40
(s.south west)
x
(s.south)
(s.south east)
Anchors on the outer bounding box,

(s.outer north)
(s.outer north west) (s.outer north east)
A test plot.
·104
2 f (x)
(s.outer center)
g(x)
0
y
(s.outer west) (s.outer east)
4
40 20 0 20 40
x
(s.outer south west) (s.outer south east)

(s.outer south)
There are anchors which have one coordinate on the outer bounding box, and one on the axis rectangle,
(s.above north)
(s.above north west) (s.above north east)
A test plot.
(s.left of north west)
·104 (s.right of north east)
2 f (x)
g(x)
(s.left of west)
0 (s.right of east)
y
(s.left of south west)

4 (s.right of south east)
40 20 0 20 40
x
(s.below south west) (s.below south east)

(s.below south)
And finally, we have origin anchors which are especially useful when axis lines pass through the origin,
(s.above origin)
100y
50
(s.left of origin) (s.origin) (s.right of origin)
2 2 4 x
(s.below origin)
There is a fifth anchor which is not directly related to the axis: you can provide the anchor of a named
inner node. Thus, you can define your own anchor, by writing \node (hnamei) at (hpoint coordinatei)
{}; as follows (using the baseline option described below):
1
Aligning at .......
0.5
(aninnernode)
0.5
1
6 4 2 0 2 4 6
Aligning at .......
\begin{axis}[small,anchor=aninnernode.center]
\node
[pin=-90:(aninnernode),fill=black,circle,scale=0.3]
(aninnernode) at (-2,0.75) {};
\draw[help lines] (-6,0.75) -- (6,0.75);
\end{axis}
\end{tikzpicture}
What happens is that a node is placed at (-2,0.75). Note that the options [pin=...] are merely to
show the \node (the pin style has been defined by the pgfplots manual). Since a name can also be
assigned using name=hnode’s namei and since any pgfplots description is also a \node, you can align
your plot at selected axis descriptions:
Aligning at ....... The function sin x is very pretty.
0.5
0.5
1
6 4 2 0 2 4 6
Aligning at .......
\begin{axis}[
small,
title={The function $\sin x$ is very pretty.},
title style={name=MyTitleNode},
anchor=MyTitleNode.base,
]
\end{axis}
\end{tikzpicture}
The default value is anchor=south west. You can use anchors in conjunction with the Tik Z baseline
option and/or \begin{pgfinterruptboundingbox} to perform alignment.
Remarks: Each of the anchors on the axis rectangle has an equivalent to a coordinate in the axis
description cs described in Section 4.9.1. That means the first set of anchors actually lives on the
tight bounding box around the axis (without any ticks or descriptions). The south west anchor will
always be the lower left corner of this bounding box, even in case of a rotated or skewed coordinate
system60 . Similar statements hold for the other anchors.
4.19.2 Vertical Alignment with baseline

/tikz/baseline (no value)
The baseline option should be provided as argument to a tikzpicture. It configures Tik Z to
shift the picture position y = 0 to the embedding text’s baseline:
This is a picture, here another one.

This is \tikz[baseline]\fill[red] (0,0) circle(3pt); a picture,
here \tikz[baseline]\fill[red] (0,10pt) circle(3pt); another one.
Consequently, the baseline option allows to align di↵erent tikzpictures. An axis is, by default,
placed with at={(0,0)}, and the anchor key specifies which part of the axis is placed at (0,0).
Consequently, the baseline option, together with anchor, allows to align di↵erent axes with the
embedding text.
The default axis anchor is south west, which means that the picture coordinate (0, 0) is the lower
left corner of the axis. As a consequence, the Tik Z option “baseline” allows vertical alignment of
adjacent plots:
0.5
0.5
0
1 0.5 0 0.5 1
0
N
X
1 0.5 0 0.5 1 ni
A normal sized x label i=0
% 1. Unaligned:
\pgfplotsset{domain=-1:1}
\begin{tikzpicture}
\begin{axis}[xlabel=A normal sized $x$ label]
\addplot[smooth,blue,mark=*] {x^2};
\end{axis}
\end{tikzpicture}%
\hspace{0.15cm}
\begin{tikzpicture}
\begin{axis}[xlabel={$\displaystyle \sum_{i=0}^N n_i $ }]
\end{axis}
\end{tikzpicture}
60 Note that this is only true for versions since 1.3.

1 1
0.5 0.5
0 0
1 0.5 0 0.5 1 1 0.5 0 0.5 1
A normal sized x label N
X
ni
i=0
% 2. Aligned:
\pgfplotsset{domain=-1:1}
\begin{axis}[xlabel=A normal sized $x$ label]
\end{axis}
\end{tikzpicture}%
\hspace{0.15cm}
\begin{axis}[xlabel={$\displaystyle \sum_{i=0}^N n_i $ }]
\end{axis}
\end{tikzpicture}
Note that it is also possible to write baseline=5cm in which case the image o↵set at y =5cm will
be used as baseline.
The baseline key is related to \begin{minipage}[halignmenti] or \begin{tabular}[halignmenti]:

the halignmenti tells LATEX which part of the minipage or tabular shall be positioned on the baseline.
Thus, baseline does the same for pictures (with more freedom for halignmenti).
4.19.3 Horizontal Alignment

Horizontal alignment can be done in two ways:
1. Using separate tikzpicture environments which have reduced bounding boxes or
2. A single tikzpicture environment in which the complete alignment is done.
The first approach requires the use of reduced bounding boxes and is discussed in Section 4.20.1.
The second approach, a single tikzpicture environment, employs the at and anchor keys to align
parts of the images. For example, if you place multiple axes into a single tikzpicture and use the
‘anchor’-option, you can control horizontal alignment:
cycle list={
{red,only marks,mark options={
0.5 fill=red,scale=0.8},mark=*},
{black,only marks,mark options={
fill=black,scale=0.8},mark=square*}}}}
0 \begin{axis}[width=4cm,scale only axis,

name=main plot]
0 0.1 0.2 \addplot file
{plotdata/pgfplots_scatterdata1.dat};
\addplot file
\addplot[blue] coordinates {
0 0.5 1 (0.093947, -0.011481)
(0.101957, 0.494273)
(0.109967, 1.000027)};
\end{axis}
\begin{axis}[
at={(main plot.below south west)},yshift=-0.1cm,
anchor=north west,
width=4cm,scale only axis,height=0.8cm,
ytick=\empty]
\addplot file
{plotdata/pgfplots_scatterdata1_latent.dat};
\addplot file
{plotdata/pgfplots_scatterdata2_latent.dat};
\end{axis}
\end{tikzpicture}
Here, the second axis uses at={(main plot.below south west)} to be placed below the first one.
Furthermore, it has yshift=-0.1cm in order to leave additional space, and it uses anchor=north west
to place the upper left corner at the specified position. Instead of the at={} construction, we could also
have used yshift with larger negative shift.
4.19.4 Alignment In Array Form (Subplots)

Sometimes multiple alignment axes in array form are desired. pgfplots supports this task in several
ways which are described in the following. There are basically three related, yet di↵erent, approaches:
1. Simply place \begin{tikzpicture}...\end{tikzpicture} into a LATEX table. This is straight–
forward; you would do the very same thing with \includegraphics.
In addition to \includegraphics, the baseline feature allows simple yet e↵ective vertical align-
ment. In addition, the trim left and trim right features allow simple yet e↵ective horizontal
alignment (see below).
2. Use a single picture which contains an array of axes, i.e. a pattern like
\begin{tikzpicture} \matrix{ hmultiple axesi}; \end{tikzpicture}.
This allows considerably simpler alignment! Alas, it needs special handling for legend entries
due to a weakness of \matrix. If you use the external library (which is recommended), it takes
more time since the picture gets larger.
3. Use the groupplots library shipped with pgfplots. It is specialized on axes in array form with
particular strength if the axes are closely related (for example if they share axis descriptions like
xlabel or even tick labels). Note, however, that the other approaches are better when it comes to
automatic handling of bounding boxes.
The groupplots library is discussed in all detail in Section 5.8. This section discusses the other two
approaches.
Array Alignment using LATEX Tables The idea is simple: use a LATEX table and provide one
tikzpicture for every cell. You are probably familiar with this sort of alignment, perhaps together
with \includegraphics. It works in the very same way for pgfplots. The approach is the simplest
one since it doesn’t need special knowledge. Its disadvantage, however, is more difficulty to control
positions inside of the image (like di↵erently sized axis descriptions).
Is is strongly recommended to employ the baseline option for each cell picture, which simplifies vertical
alignment considerably. If you want a simple solution to place separate axes in array form, and you
prefer to use one tikzpicture for every axis, the probably most simple and most e↵ective way to get
horizontal alignment are the trim left and trim right features – or styles based on them:
The trim axis left feature can be used to exclude axis descriptions on the left from the bounding
box, and the trim axis right can exclude axis descriptions on the right from the bounding box. Thus,
alignment is done using the vertical axis lines. Since both keys e↵ectively modify the bounding box,
they are documented in Section 4.20.1 “Bounding Box Restrictions”. Here is just a small example for
array alignment by means of tabular, baseline and the trim left/trim right features:
Trimmed bounding boxes Trimmed bounding boxes

6
f (x) = x2
4
20
2
0
10
2
4
0
6
6 4 2 0 2 4 6 6 4 2 0 2 4 6
Trimmed bounding boxes Trimmed bounding boxes
600
100
400
0
200
100
0
6 4 2 0 2 4 6 6 4 2 0 2 4 6
x
\pgfplotsset{
small,
title=Trimmed bounding boxes
}
\begin{center}
\begin{tabular}{rl}
\begin{tikzpicture}[baseline,trim axis left]
\begin{axis}
\addplot {x};
\end{axis}
\end{tikzpicture}
&
\begin{tikzpicture}[baseline,trim axis right]
\begin{axis}[
ylabel={$f(x)=x^2$},
ylabel style={font=\Huge}]
\addplot {x^2};
\end{axis}
\end{tikzpicture}
\\
%
\begin{tikzpicture}[baseline,trim axis left]
\begin{axis}[xlabel=$x$,xlabel style={font=\Huge}]
\addplot {x^3};
\end{axis}
\end{tikzpicture}%
&
\begin{tikzpicture}[baseline,trim axis right]
\begin{axis}[yticklabel pos=right]
\addplot {x^4};
\end{axis}
\end{tikzpicture}%
\\
\end{tabular}%
\end{center}
The example has 2 ⇥ 2 axes. The baseline feature controls the vertical alignment: the lower axis lines
are always on the same height. The trim axis left key is a style which tells Tik Z to trim everything
which is left of the left axis line. Similarly, the trim axis right key does not include picture parts
right of the right axis line. Together with \begin{center} and the yticklabel pos=right key, we
get correct horizontal and vertical alignment together with centering at the left- and right axis lines
(without descriptions).
A strong advantage is that this type of alignment requires almost no changes to your pictures. Thus,
you can copy–paste existing images (TEX code) relatively simple.
Note that the approach is fully compatible with the image externalization library: each picture is
exported separately, and the bounding box restrictions (and the baseline o↵set) are stored in separate
.dpth files. The trim left/trim right approach for horizontal alignment is the only supported way
for reduced bounding boxes and image externalization.
Array Alignment using Tik Z Matrices While it is possible to use (for example) tabular combined
with the vertical and horizontal alignment methods discussed above, it might be better to use a Tik Z
matrix since it automatically handles the size of axis descriptions.
A Tik Z matrix is some sort of “graphical” table. It knows everything about picture alignment and it
has more flexibility than tabular when it comes to graphics. The idea is to pack the complete array
into a single picture.
The complete documentation of a Tik Z matrix is beyond the scope of this manual, please refer to [6]
for details. But we provide an example here:
f (x) = x2
4
20
2
0
10
2
4
0
6
6 4 2 0 2 4 6 6 4 2 0 2 4 6
600
100
400
0
200
100
0
6 4 2 0 2 4 6 6 4 2 0 2 4 6
x
\begin{tikzpicture}
\pgfplotsset{small}
\matrix {
\begin{axis}
\addplot {x};
\end{axis}
&
% differently large labels are aligned automatically:
\begin{axis}[ylabel={$f(x)=x^2$},ylabel style={font=\Huge}]
\addplot {x^2};
\end{axis}
\\
%
\begin{axis}[xlabel=$x$,xlabel style={font=\Huge}]
\addplot {x^3};
\end{axis}
&
\begin{axis}
\addplot {x^4};
\end{axis}
\\
};
\end{tikzpicture}
So, a matrix is a picture element inside of tikzpicture. Its cells are separated by ‘&’ as in tabular (or,
if ‘&’ causes problems, with \pgfmatrixnextcell). Its rows are separated by ‘\\’. Each cell is aligned
using the cells’ anchor. Since, by default, the anchor of an axis is placed at the lower left corner, the
example above is completely aligned, without the need for any bounding box modifications – even the
labels are aligned correctly. If another anchor shall be used, simply place
\pgfplotsset{anchor=....}
\matrix {
...
};
in front of the matrix. This will use the same configuration for every sub-plot.
4.20. THE PICTURE’S SIZE: BOUNDING BOX AND CLIPPING 357
Attention: Unfortunately, the array alignment with \matrix needs special attention with legends.
A legend is also a \matrix and Tik Z matrices can’t be nested. You will need to use the legend to
name feature (or to assemble a legend by means of \label and \ref) to overcome this weakness (see
Section 4.9.6 for details).
4.19.5 Miscellaneous for Alignment

Predefined node current axis
A node which refers to the current axis or the last typeset axis.
You can use this node in axis descriptions, for example to place axis labels or titles.
Remark: If you use current axis inside of axis descriptions, the “current axis” is not yet fin-
ished. That means you can’t use any outer anchor inside of axis descriptions.
It is also possible to use current axis in any drawing or plotting commands inside of an axis (but
no outer anchor as these are not defined when drawing commands are processed). This usage is
similar to the axis description cs.
4.20 The Picture’s Size: Bounding Box and Clipping

This section explains how a picture receives its final dimensions. The picture’s dimension is the bounding
box. It is possible to restrict the bounding box, but display graphical elements outside of the bounding
box. This is called subject of Section 4.20.1. Another use-case is to restrict both the bounding box and
the clip the graphical elements to some outer path which is subject of Section 4.20.2.
4.20.1 Bounding Box Restrictions

Bounding box restrictions are a useful and often necessary tool if multiple pictures need to be aligned
properly. Consequently, it is often applied together with the Alignments methods of Section 4.19.
Bounding box restrictions can be archieved with several methods of pgf:
1. The overlay option,
2. The pgfinterruptboundingbox environment,
3. The \pgfresetboundingbox command,
4. The \useasboundingbox path,
5. The trim left and trim right feature (which is the only supported way of restricted bounding
boxes and image externalization; at least for pdf output).
An additional item is a specific use-case of pgfplots:
6. The hide axis feature will exclude any axis–specific stu↵ from the bounding box. See the reference
for hide axis for details.
Note that image externalization (the external library) is more or less incompatible with methods
(1.)–(4.). The problem is that pdflatex crops everything outside of the bounding box away. There
are only two safe ways to “restrict” bounding boxes of external .pdf images: the first is the mentioned
trim left/trim right feature and the second is to use negative \hspace or \vspace commands (or
options to \includegraphics).
/tikz/overlay (no value)

A special key of pgf which disables bounding box updates for (parts of) the image. The e↵ect is
that those parts are an “overlay” over the document.
For pgfplots, overlay can be useful to position legends or other axis descriptions outside of the
axis – without a↵ecting its size (and without a↵ecting alignment).
For example, one may want to include only certain parts of the axis into the final bounding box.
This would allow horizontal alignment (centering):
A title \begin{tikzpicture}%
\begin{axis}[
x2 x3 x4 title=A title,
ylabel style={overlay},
yticklabel style={overlay},
10 xlabel={$x$},
ylabel={$y$},
y
0 domain=-2:2
]
\addplot {x^2};
\addplot {x^3};
\addplot {x^4};
10 \legend{$x^2$,$x^3$,$x^4$}
2 1 0 1 2 \end{axis}
x \end{tikzpicture}%
Now, the left axis descriptions (y label and y ticks) stick out of the bounding box.
The following example places a legend somewhere without a↵ecting the bounding box.
1
\begin{tikzpicture}
\begin{axis}[
domain=0:6.2832,samples=200,
0.5 legend style={
overlay,
Signal 1 at={(-0.5,0.5)},
0 anchor=center},
Signal 2
every axis plot post/.append style={mark=none},
enlargelimits=false]
0.5
\addplot {sin(deg(x)+3)+rand*0.05};
\addplot {cos(deg(x)+2)+rand*0.05};
1 \legend{Signal 1,Signal 2}
0 2 4 6 \end{axis}
\end{tikzpicture}
More information about the overlay option can be found in the pgf manual [6].
\pgfresetboundingbox
This command of pgf resets the bounding box of the current picture. The computation starts from
scratch afterwards, allowing to compute a user–defined bounding box.
A title
x2 x3 x4 \setlength{\fboxsep}{0pt}%
\fbox{%
10 \begin{axis}[
title=A title,
xlabel={$x$},
y
ylabel={$y$},
0 legend style={at={(0.5,0.97)},
domain=-2:2
]
\addplot {x^2};
10 \addplot {x^3};
2 1 0 1 2 \addplot {x^4};
x \legend{$x^2$,$x^3$,$x^4$}
\end{axis}
\pgfresetboundingbox
\path
(current axis.south west)
rectangle (current axis.north east);
\end{tikzpicture}%
}%
4.20. THE PICTURE’S SIZE: BOUNDING BOX AND CLIPPING 359
The example draws a normal picture, containing an axis. Afterwards, it throws the bounding box
away and creates a new one based on the current axis node and its anchors.
\begin{pgfinterruptboundingbox}
\end{pgfinterruptboundingbox}
Yet another approach with the same e↵ect is shown below: the bounding box is interrupted man-
ually, and resumed afterwards.
A title \setlength{\fboxsep}{0pt}%
\fbox{%
x2 x3 x4 \begin{tikzpicture}%
\begin{pgfinterruptboundingbox}
\begin{axis}[
10 title=A title,
xlabel={$x$},
ylabel={$y$},
y
domain=-2:2
]
\addplot {x^2};
\addplot {x^3};
10 \addplot {x^4};
2 1 0 1 2 \legend{$x^2$,$x^3$,$x^4$}
x \end{axis}
\end{pgfinterruptboundingbox}
\useasboundingbox
(current axis.below south west)
rectangle (current axis.above north east);
\end{tikzpicture}%
}%
The pgfinterruptboundingbox environment does not include its content into the image’s bounding
box, and \useasboundingbox sets the pictures bounding box to the following argument (see [6]).
/tikz/trim left={hx coordinate or pointi} (default 0pt)

/tikz/trim right={hx coordinate or pointi}
These two keys allow to reduce the size of the bounding box.
The trim left key expects either a single x coordinate like 1cm or a point like (current
axis.west). If a point is provided, is uses only the x coordinate of that point. Then, the left
end of the bounding box is set to the resulting x coordinate and everything left of it is outside of
the bounding box.
The trim right key has the same e↵ect, only for the right end of the bounding box.
More detailed documentation can be found in the Tik Z manual.
/tikz/trim axis left (style, no value)

A style with value trim left=(current axis.south west).
The style needs to be provided as argument to \begin{tikzpicture}[trim axis left]. It ex-
pects (at least) one pgfplots environment in the picture. The e↵ect is to trim everything which
is left of the last axis’ anchor south west (i.e. everything left of the left axis boundary).
/tikz/trim axis right (style, no value)

A style with value trim right=(current axis.south east).
It works similarly to trim axis left: the e↵ect is that everything right of the right axis line of
the last axis environment is truncated from the bounding box.
/tikz/trim axis group left (style, no value)

A style which has the same e↵ect as trim axis left, but is tailored for the groupplots library.
It has the value trim left=(group c1r1.south west).
The style needs to be provided as argument to \begin{tikzpicture}[trim axis group left]. It

expects (at least) one groupplot environment in the picture. The e↵ect is to trim everything which
is left of the first group axis’ anchor south west (i.e. everything left of the left axis boundary).
/tikz/trim axis group right (style, no value)

A style which has the same e↵ect as trim axis right, but is tailored for the groupplots library.
It works similarly to trim axis group left: the e↵ect is that everything right of the rightmost
axis in a group plot (the last element of the groupplot environment) is truncated from the bounding
box.
4.20.2 Clipping
Clipping influences both the picture size and the visible output in contrast to bounding box restrictions
which reduce the picture’s final size while keeping the same graphical output.
Typically, pgfplots uses the path for a boxed axis as clip path. However, clipping has some special
features and fine–tuning keys which are explained in this section.
/pgfplots/clip=true|false (initially true)

Controls whether any paths inside of an axis shall be clipped.
This is in e↵ect even if hide axis=true.
Note that a clip path can contribute to the picture’s bounding box. Starting with compat=1.8, pgf-
plots applies intelligence to separate the responsabilities clipping and bounding box control, see clip
bounding box and its choices. As of compat=1.8, a clip path can be in e↵ect although the bounding
box is considerably smaller than the clip path. This is typically what one expects if the clip path is
invisible.
The clip path is generated using \pgfplotspathaxisoutline, i.e. it is the path induced by boxed axis
lines. For a three–dimensional plot, only the outer axis lines are used. A plot with centered axis lines
uses the outer axis lines as well.
/pgfplots/clip marker paths=true|false (initially false)

The initial choice clip marker paths=false causes markers to be drawn after the clipped region. Only
their positions will be clipped.
As a consequence, markers will be drawn completely, or not at all. The value clip marker paths=true
is here for backwards compatibility: it does not introduce special marker treatment, so markers may be
drawn partially if they are close to the clipping boundary61 .
This key has no e↵ect if clip=false.
Note that clip marker paths also a↵ects the sequence in which plots and their markers are drawn on top
of each other. See also the related key clip mode.
/pgfplots/clip bounding box=default tikz|upper bound (initially controlled by compat key)

Controls how the path generated by clip=true contributes to the bounding box. This has a consequence
for axis lines6=box, in particular, for hide axis: if the value is default tikz, hiding (parts of) the
axis will not reduce the bounding box because the clip path is as large as before. The value upper
bound allows to reduce the bounding box also in case of hide axis.
More precisely, the choice default tikz installs the clip path induced by the axis as ordinary Tik Z
path (see \pgfplotspathaxisoutline). That means its bounding box essentially contributes to the
picture’s bounding box, irrespective of the size of contained paths.
The choice upper bound allows to reduce the picture’s bounding box to what is actually shown:
if the picture only contains graphical elements which are completely within the bounding box of
\pgfplotspathaxisoutline, the bounding box is made up of those contained elements. If the con-
tained elements are actually larger than the bounding box of \pgfplotspathaxisoutline, they are
clipped to the outline’s path (“upper bound”). The latter case ensures that parts of the graphics which
are excluded by clip are not counted for the bounding box.
Keep in mind that hide axis is independent of clip=true: the clip path might still be in e↵ect even
though the axis outline is invisible.
61 Please note that clipped marker paths may be slightly faster during TEX compilation.
4.21. SYMBOLIC COORDINATES AND USER TRANSFORMATIONS 361
This key is irrelevant if clip=false. In addition, it has no e↵ect for axis lines=box since the box
path is made up from \pgfplotspathaxisoutline. It has an e↵ect for hide axis=true or for choices
of axis lines in which parts of the axis are empty.
This key is controlled by the compat level. Its default is default tikz. Since compat=1.8, it is set to
upper bound.
The key has no e↵ect if clip=false.
/pgfplots/clip mode=global|individual (initially global)

This key controls how pgfplots implements the clip=true feature (which is on by default). Its primary
motivation is control where markers are placed: are markers on top of everything else (choice global)
or are they overdrawn by following plots (choice individual)?
The choice global tells pgfplots to install one single clip path for the complete picture62 . In order
to avoid clipped marker paths, any markers are processed after the clip path has been closed, i.e. on a
separate layer (see clip marker paths). An unexpected side–e↵ect is that marks are on top of plots,
even if the plots have been added after the markers.
The choice individual instructs pgfplots to install a separate clip path for every \addplot command.
Consequently, the plot will be clipped. But most importantly, its markers will be drawn immediately
after the clip path has been deactivated.
An unexpected side–e↵ect of clip mode=individual is that
1. the resulting pdf will be slightly larger due to the repeated paths,
2. custom drawing instruction like \node or \draw need to be clipped manually: use
\begin{tikzpicture}
\begin{axis}[
20 clip mode=individual,
]
\addplot+[samples=3] {x^2};
10
\begin{scope}
\clip \pgfextra{\pgfplotspathaxisoutline};
0 \draw (-20,15) -- (20,15);
\draw (-20,20) -- (20,20);

\end{scope}
6 4 2 0 2 4 6
\end{axis}
\end{tikzpicture}
to install a custom clip path around your \draw instructions for such a use–case. Here, the path
instruction \pgfplotspathaxisoutline results in a path of the axis outline, i.e. the path which
is used for the background paths or for clipping. Since it is a basic level macro, it needs to be
encapsulated by \pgfextra.
Note that clip marker paths can lead to the same result as clip mode=individual if the plot does
not reach the boundaries.
4.21 Symbolic Coordinates and User Transformations

pgfplots supports user transformations which can be applied to input and output coordinates. Suppose
the plot shall display days versus account statements over time. Then, one wants to visualize date versus
credit balance. But: dates need to be transformed to numbers before doing so! Furthermore, tick labels
shall be displayed as dates as well. This, and more general transformations, can be implemented using the
x coord trafo and y coord trafo keys.
62 The choice clip mode=global was the only supported clipping mechanism up to and including version 1.5.
Remark: This section applies to users who want to have non-standard input coordinates. If you have
normal numbers which don’t need to be transformed and you like to have special symbols as tick labels, you
should consider using the xticklabels (yticklabels) key described on page 317.
See also Section 4.23.1 for di↵erent types of transformations and their interaction.
/pgfplots/x coord trafo/.code={h... i}

/pgfplots/y coord trafo/.code={h... i}
/pgfplots/z coord trafo/.code={h... i}
/pgfplots/x coord inv trafo/.code={h... i}
/pgfplots/y coord inv trafo/.code={h... i}
/pgfplots/z coord inv trafo/.code={h... i}
These code keys allow arbitrary coordinate transformations which are applied to input coordinates and
output tick labels.
The x coord trafo and y coord trafo command keys take one argument which is the input coordi-
nate. They are expected to set \pgfmathresult to the final value.
At this level, the input coordinate is provided as it is found in the \addplot statement. For example,
if x coordinates are actually of the form hyear i-hmonthi-hdayi, for example 2008-01-05, then a use-
ful coordinate transformation would transform this string into a number (see below for a predefined
realization).
In short, no numerics has been applied to input coordinates when this transformation is applied63 .
The input coordinate transformation is applied to
• any input coordinates (specified with \addplot or axis cs),
• any user-specified xtick or ytick options,
• any user-specified extra x ticks and extra y ticks options,
• any user-specified axis limits like xmin and xmax.
The output coordinate transformation x coord inv trafo is applied to tick positions just before eval-
uating the xticklabel and yticklabel keys. The argument to x coord inv trafo is a fixed point
number (which may have trailing zeros after the period). The tick label code may use additional macros
defined by the inverse transformation.
Remark: pgfplots will continue to produce tick positions as usual, no extra magic is applied. It
may be necessary to provide tick positions explicitly if the default doesn’t respect the coordinate space
properly.
The initial value of these keys is
\pgfplotsset{
x coord trafo/.code={},
x coord inv trafo/.code={}}
which simply disables the transformation (the same for y, of course).
Remark: It might be necessary to set
\pgfplotsset{
xticklabel={\tick},
scaled x ticks=false,
plot coordinates/math parser=false,
}
in order to avoid number formatting routines on \tick or numerics for tick scale methods. This is done
automatically by the predefined symbolic coordinate styles (see below).
4.21.1 String Symbols as Input Coordinates

It is possible to provide a string dictionary to pgfplots. An input coordinate can then use any symbol
provided in that dictionary.
63 Of course, if coordinates have been generated by gnuplot or pgf, this does no longer hold.
/pgfplots/symbolic x coords={hdictionaryi}
/pgfplots/symbolic y coords={hdictionaryi}
/pgfplots/symbolic z coords={hdictionaryi}
A style which sets x coord trafo and x coord inv trafo (or the respective y or z variants) such that
any element in hdictionaryi is a valid input coordinate. The hdictionaryi can be a comma separated list
or a list terminated with ‘\\’. In both cases, white space is considered to be part of the names (use ‘%’
at end of lines).
The dictionary will assign integer numbers to every element (starting with 0). These integers are used
internally for arithmetics. Finally, the inverse transformation takes a fixed point number and maps it
to the nearest integer, and that integer is mapped into the dictionary.
\begin{tikzpicture}
\begin{axis}[symbolic x coords={a,b,c,d,e,f,g,h,i}]
80 \addplot+[smooth] coordinates {
(a,42)
(b,50)
(c,80)
(f,60)
60 (g,62)
(i,90)};
\end{axis}
\end{tikzpicture}
40
a c e g i
The e↵ect of the transformation is simply that input coordinates can be elements of the dictionary and
tick labels will be chosen out of this dictionary as well.
Note that symbolic x coords is more-or-less equivalent to explicitly provided xtick positions and
xticklabels:
\begin{tikzpicture}
\begin{axis}[
80 xtick={0,1,2,...,20},
xticklabels={a,b,c,d,e,f,g,h,i},
xticklabel style={
anchor=base,
yshift=-\baselineskip
60 },
]
\addplot+[smooth] coordinates {
(0,42)
(1,50)
40 (2,80)
a b c d e f g h i (5,60)
(6,62)
(8,90)};
\end{axis}
\end{tikzpicture}
The di↵erence is that the approach with symbolic x coords is simpler to read whereas the xtick
approach is simpler with respect to coordinate arithmetics (for example to increase limits using
enlargelimits). The xticklabel style here is an attempt to align all tick labels at their base
line (which would be useful for symbolic x coords as well as soon as labels have characters which
exceed the baseline).
The key symbolic x coords and its variants accepts a comma–separated list of strings. These can be
arbitrary and can even contain unexpandable material64 :
64 As of pgfplots version 1.11.
\begin{tikzpicture}
\begin{axis}[
80 symbolic x coords={a,$1 \pm 2$,c,d,1\&1,f,g,h,i},
xtick=data,
]
(a,42)
60 ($1 \pm 2$,50)
(c,80)
(1\&1,60)
(g,62)
(i,90)};
40 \end{axis}
a1±2c 1&1 g i \end{tikzpicture}
In any case, the arguments of hdictionaryi must be met precisely, including any white spaces.
Symbolic coords are useful since they reduce the burden to map strings to indices and vice–versa.
However, they have a caveat: what if you want to set xmin to something to the left of the first symbolic
x coord? This is impossible since any input coordinate is expected to be contained in symbolic x
coord. To allow such modifications nevertheless, pgfplots checks for the magic prefix [normalized]:
\begin{tikzpicture}
\begin{axis}[
80 symbolic x coords={a,b,c,d,e,f,g,h,i},
xmin={[normalized]-1.7},
minor tick num=1,
1.5 ]
60 (a,42)
(b,50)
(c,80)
(f,60)
(g,62)
40 (i,90)};
a c e g i
\fill (axis cs:{[normalized]1.5},60)
circle(2pt) coordinate[pin={$1.5$}];
\end{axis}
\end{tikzpicture}
Whenever pgfplots finds a symbolic coordinate which starts with the precise string ‘[normalized]’,
it will interprete everything after the prefix to be a normalized number and no symbol. In our example,
xmin will be set to 1.7, i.e. something to the left of a (which would be 0 according to the definition
above). Furthermore, the custom \node is placed at (1.5, 60) despite the fact the 1.5 is no defined
symbol – but we know that it is between b and c which are dictionary entry 1 and 2, respectively.
See also the option to add tick and/or grid lines at every encountered coordinate using xtick=data
(or minor xtick=data).
4.21.2 Dates as Input Coordinates

The already mentioned application of using dates as input coordinates has been predefined, together with
support for hours and minutes. It relies on the pgf calendar library which converts dates to numbers in the
Julian calendar. Then, one coordinate unit is one day.
\usepgfplotslibrary{dateplot} % LATEX and plain TEX

\usepgfplotslibrary[dateplot] % ConTEXt
\usetikzlibrary{pgfplots.dateplot} % LATEX and plain TEX
\usetikzlibrary[pgfplots.dateplot] % ConTEXt
Loads the coordinate transformation code.
/pgfplots/date coordinates in=hcoordinatei

Installs x coord trafo and x coord inv trafo (or the respective variant for hcoordinatei) such that
ISO dates of the form hyear i-hmonthi-hdayi are accepted. Here, hcoordinatei is usually one of x, y, or
z, but it can also contain stu↵ like hist/data.
After installing this style, input values like 2006-02-28 will be converted to an “appropriate” integer
using the Julian calender. Input coordinates may be of the form
hyear i-hmonthi-hdayi
or they may contain times as
hyear i-hmonthi-hdayi hhour i:hminutei.
The result of the transformation are numbers where one unit is one day and times are fractional numbers.
The transformation is implemented using the pgf-calendar module, see [6, Calendar Library]. This
reference also contains more information about extended syntax options for dates.
The inverse transformation provides the following macros which are available during tick label evaluation
(i.e. when used inside of xticklabel or yticklabel):
• \year expands to the year component,
• \month expands to the month component,
• \day expands to the day component,
• \hour expands to the hour component (using two digits),
• \Hour expands to the hour component (but omits leading zeros),
• \minute expands to the minute component (two digits),
• \Minute expands to the minute component (omits leadings zeros),
• \lowlevel expands to the low level number representing the tick,
• \second will always be 00.
This allows to use \day.\month.\year or \day. \hour:\minute inside of xticklabel, for example.
A complete example (with fictional data) is shown below.
3,000e
date account1 account2 account3
2008-01-03 60 1200 400
Total credit
2008-02-06 120 1600 410 2,000e

2008-03-15 -10 1600 410
2008-04-01 1800 500 410
2008-05-20 2300 500 410 1,000e
2008-06-15 800 1920 410
0e
03.01. 22.02. 12.04. 01.06.
2008
Giro Tagesgeld Sparbuch

% Preamble: \pgfplotsset{width=7cm,compat=1.13}\usepgfplotslibrary{dateplot}
% requires \usepgfplotslibrary{dateplot} !
\pgfplotstabletypeset[string type]{plotdata/accounts.dat}
\begin{tikzpicture}
\begin{axis}[
date coordinates in=x,
xticklabel={\day.\month.},
xlabel={2008},
stack plots=y,
yticklabel={\pgfmathprintnumber{\tick}\EUR{}}, % <- requires \usepackage{eurosym}
ylabel=Total credit,
ylabel style={yshift=10pt},
legend style={
at={(0.5,-0.3)},anchor=north,legend columns=-1}]
\addplot table[x=date,y=account1] {plotdata/accounts.dat};

\legend{Giro,Tagesgeld,Sparbuch}
\end{axis}
\end{tikzpicture}
% Preamble: \pgfplotsset{width=7cm,compat=1.13}\usepgfplotslibrary{dateplot}
100 % requires \usepgfplotslibrary{dateplot} !
\begin{tikzpicture}
\begin{axis}[
date coordinates in=x,
xtick=data,
xticklabel style=
50
{rotate=90,anchor=near xticklabel},
xticklabel=\day. \hour:\minute,
]
(2009-08-18 09:00, 050)
0 (2009-08-18 12:00, 100)
(2009-08-18 15:00, 100)
18. 09:00
18. 12:00
18. 15:00
18. 18:35
18. 21:30
19. 00:00
19. 03:00
19. 06:00
(2009-08-18 18:35, 100)

(2009-08-18 21:30, 040)
(2009-08-19, 020)
(2009-08-19 3:00, 000)
(2009-08-19 6:0, 035)
};
\end{axis}
\end{tikzpicture}
/pgfplots/date ZERO=hyear i-hmonthi-hdayi

A technical key which defines the 0 coordinate of date coordinates in.
pgfplots up to and including version 1.12 required this value. As of 1.13, the first encountered
coordinate is used as date ZERO, i.e. it is assigned automatically without user intervention.
Users will never see the resulting numbers, so one probably never needs to change this value. This key
allows to set the result as needed.
4.22 Skipping Or Changing Coordinates – Filters

pgfplots o↵ers filters. A filter expects a (numeric) input coordinate and is allowed to modify the coordinate
or throw it away. Filters can either operate on individual coordinates or on all simultaneously. One main
application is to skip coordinates, i.e. to install a filter which causes the coordinate to be skipped. This works
by assigning special “this coordinate does not exist” values instead of the original values. Valid choices for
“this coordinate does not exist” are defined using unbounded coords as described in Section 4.5.14 about
Interrupted Plots, namely empty line or unbounded coords.
See also Section 4.23.1 for di↵erent types of transformations and their interaction.
/pgfplots/x filter/.expression={hmath expression i}

/pgfplots/y filter/.expression={hmath expression i}
4.22. SKIPPING OR CHANGING COORDINATES – FILTERS 367
/pgfplots/z filter/.expression={hmath expression i}

Installs a coordinate filter which allows to modify the current value of a single coordinate.
The argument math expression is a math expression which contains x, y, or z.
\addplot+[
x filter/.expression={x+1},
y filter/.expression={y==3 ? nan : y},
3
]
table {
x y
1 2
1.5 2.5
2 2 3
2.5 3.5
2 2.5 3 3.5 4 3 4
};
\end{axis}
\end{tikzpicture}
The x filter is evaluated first. It can depend on x, y, and z whose values are the “prepared” coordi-
nates: values which have been found after applying x coord trafo and any logarithms (for logarithmic
axes).
The y filter is evaluated as next. It can depend on x which is the result of x filter. It can also
depend on y and z which have the same value as discussed in the previous paragraph.
The z filter is evaluated as last. It can depend on x and y which are result of their respective filters.
It can also depend on z which is the plain z coordinate (as discussed for x filter).
Defining filters by math expression is actually a special case of x filter, see below.
/pgfplots/x filter/.code={h... i}
/pgfplots/y filter/.code={h... i}
/pgfplots/z filter/.code={h... i}
/pgfplots/filter point/.code={h... i}
The code keys x filter and y filter allow coordinate filtering which are based on a single coordinate.
A coordinate filter gets an input coordinate as #1 (on input, the same value is stored in \pgfmathresult),
applies some operation and writes the result into the macro \pgfmathresult. If \pgfmathresult is
empty afterwards, the coordinate is discarded. You can also set \pgfmathresult to nan or inf in which
case the coordinate can be either discarded (if unbounded coords=discard is set) or the plot can be
interrupted (the case unbounded coords=jump).
The filter point/.code filter allows filtering depending on all components forming a complete point
(x, y and z); it is described below.
It is allowed that filters do not change \pgfmathresult. In this case, the unfiltered coordinate will be
used.
Coordinate filters are useful in automatic processing system, where pgfplots is used to display au-
tomatically generated plots. You may not want to filter your coordinates by hand, so these options
provide a tool to do this automatically.
The following filter adds 0.5 to every x coordinate.
% equivalent to ’x filter/.expression={x+0.5}’:
\begin{axis}[x filter/.code=
{\pgfmathadd{#1}{0.5}}]
(4,0)
0.5
(6,1)
};
\end{axis}
\end{tikzpicture}
0
4.5 5 5.5 6 6.5
Please refer to [6, pgfmath manual] for details about the math engine of pgf. Please keep in mind that
the math engine works with limited TEX precision.
During evaluation of the filter, the macro \coordindex contains the number of the current coordinate
(starting with 0). Thus, the following filter discards all coordinates after the 5th and before the 10th.
\begin{tikzpicture}
\begin{axis}[
20 samples=20,
x filter/.code={
\ifnum\coordindex>4
\ifnum\coordindex<11
\def\pgfmathresult{}
10 \fi
\fi
}]
\addplot {x^2};
0 \end{axis}
\end{tikzpicture}
6 4 2 0 2 4 6
There is also a style key which simplifies selection by index, see below.
pgfplots invokes the filter with argument #1 set to the input coordinate. For x-filters, this is the
x-coordinate as it is specified to \addplot, for y-filters it is the y-coordinate.
If the corresponding axis is logarithmic, #1 is the logarithm (see log basis x and its variants) of the
coordinate as a real number, for example #1=4.2341. In case the logarithm was undefined, the argument
will be empty.
The arguments to coordinate filters are minimally preprocessed: first, for logarithmic axes, the log of
the argument is supplied. Second, any high level coordinate maps like x coord trafo (which may be
used to map dates to numbers or string to numbers or so) are applied. In consequence, the #1 argument
is supposed to be a number. No further transformation has been applied.
Occasionally, it might be handy to get the “raw”, completely unprocessed input coordinate as it has
been reported by the coordinate input routine. This unprocessed data is available in the three math
parser constants rawx, rawy and rawz. All these values are ready for use in filters (and some other
methods influence plots as well). Note that rawy is to be used like a function without arguments, i.e.
filters can employ it where math parsing is done.
An application could be to filter log values based on the normal scale:
4.22. SKIPPING OR CHANGING COORDINATES – FILTERS 369
\begin{tikzpicture}
1
10 \begin{semilogyaxis}
\addplot+[
restrict expr to domain={rawy}{1e0:1.5e1},
]{exp(x)};
\end{semilogyaxis}
\end{tikzpicture}
100
0 1 2
The preceding example uses rawy to throw all samples outside of the range [1, 15] away.
If key filters are invoked for plot table, access to the current row’s data can be achieved using
\thisrow{hcolumn namei} (and its variants). This includes all columns of the table.
The filter point key is more technical. It doesn’t take an argument: its arguments are given in
terms of the pgfkeys variables /data point x, /data point y and /data point z. It may change
its coordinates using \pgfkeyssetvalue{/data point x}{hnew valuei}; access to variables can be
accessed with \pgfkeysvalueof{/data point/x} or, if the argument shall be written into a macro,
with \pgfkeysgetvalue. This filter is evaluated after the other ones.
Note that you can provide di↵erent x filter/y filter arguments to each \addplot command. It
seems there are only problems with the ‘#1’ argument, and I haven’t yet found out why. Please use
\pgfmathresult in place of #1 if you provide \addplot[x filter/.code={...}].
Note that coordinate filtering is also available for mesh, surf, and patch plots. In this context, a
patch type is drawn if and only if all its vertices have bounded coordinates. In other words: if one
vertex of, say, a rectangle has been filtered away, the entire rectangle will be omitted. Coordinate
filtering for mesh and surface plots has a further special requirement: the default for such plots is
mesh input=lattice. If a coordinate filter silently discards a coordinate, the lattice will break and
pgfplots will become confused. Consequently, coordinate filtering for mesh and surface plots always
needs unbounded coords=jump, and any point which is filtered away should receive the value nan
instead of an empty string (since empty strings will always be discarded even in presence of unbounded
coords=jump). Please see the reference documentation of unbounded coords and the example therein
on page 118 for details about coordinate filtering and three dimensional plots.
/pgfplots/pre filter/.code={h... i}
Applied before x filter, y filter, and z filter.
/pgfplots/skip coords between index={hbegini}{hend i}

A style which appends an x filter which discards selected coordinates. The selection is done by index
where indexing starts with 0, see \coordindex. Every coordinate with index hbegini  i < hend i will
be skipped.
\begin{tikzpicture}
\begin{axis}[
20 samples=20,
skip coords between index={5}{11},
skip coords between index={15}{18}]
\addplot {x^2};
10 \end{axis}
\end{tikzpicture}
0
6 4 2 0 2 4 6
Technical note : this style usually applies to x coordinates (i.e. it counts x coordinates). In case you
want to apply it to something like hist/data or quiver/u, you can
1. append an asterisk ‘*’ to the style’s name and
2. provide the target coordinate’s name as first argument.
For example, skip coords between index*={hist/data}{2} applies to hist/data.
/pgfplots/each nth point={hinteger i}

A style which appends an x filter which discards all but each nth input coordinate.
This downsampling works fairly well. It can be used to reduce a huge amount of coordinates from
an input file. In this case, you should also set filter discard warning=false to avoid repeated
notifications about skipped coordinates and unbounded coords=discard such that pgfplots should
silently forget any discarded points (rather than generated interrupted plots).
Note that there is also a mark repeat style which applies the same operation to plot marks only.
Technical note : this style usually applies to x coordinates (i.e. it counts x coordinates). In case you
want to apply it to something like hist/data or quiver/u, you can
1. append an asterisk ‘*’ to the style’s name and
2. provide the target coordinate’s name as first argument.
For example, each nth point*={hist/data}{2} applies to hist/data.
/pgfplots/restrict x to domain=hmini:hmax i
/pgfplots/restrict y to domain=hmini:hmax i
/pgfplots/restrict z to domain=hmini:hmax i
/pgfplots/restrict x to domain*=hmini:hmax i
/pgfplots/restrict y to domain*=hmini:hmax i
/pgfplots/restrict z to domain*=hmini:hmax i
These keys append x (or y or z) coordinate filters to restrict the respective coordinate to a domain.
The versions without star (like restrict x to domain) will assign the value -inf if the coordinate
is below hmini and +inf if the coordinate is above hmax i. The starred versions (like restrict x to
domain*) will truncate coordinates to [hmini, hmax i], i.e. they assign the value hmini if the coordinate
falls outside of the lower limit and hmax i if the value falls outside of the upper limit.
For logarithmic axes, hmini and hmax i are logs of the respective values. A variant which uses the
non-logarithmic number might be to use restrict expr to domain={\pgfmathrawx}{hmini}{hmax i}.
The non-starred versions also set unbounded coords=jump which leads to interrupted plots.
tan(x)
3 ⇡/2 ⇡/2 3
2⇡ 2⇡
5
4.23. TRANSFORMING COORDINATE SYSTEMS 371
\begin{tikzpicture}
\begin{axis}[
restrict y to domain=-10:10,
samples=1000,
% some fine-tuning for the display:
width=10cm, height=210pt,
xmin=-4.7124, xmax=4.7124,
xtick={-4.7124,-1.5708,...,10},
xticklabels={$-\frac32 \pi$,$-\pi/2$,$\pi/2$,$\frac32 \pi$},
axis x line=center,
axis y line=center]
\addplot[blue] gnuplot[id=tangens,domain=-1.5*pi:1.5*pi] {tan(x)};

\legend{$\tan(x)$}
\end{axis}
\end{tikzpicture}
/pgfplots/restrict expr to domain={hexpressioni}{hhmini:hmax ii}

/pgfplots/restrict expr to domain*={hexpressioni}{hhmini:hmax ii}
Appends an x coordinate filter which sets the x coordinate to -inf if the hexpressioni evaluates to
something less than hmini and to inf if hexpressioni evaluates to something larger than hmax i.
The starred variant, restrict to domain* assigns hmini if hexpressioni is less then the lower limit and
hmax i if it is larger than the upper limit.
The non-starred version also sets unbounded coords=jump which leads to interrupted plots.
In contrast to restrict x to domain, hexpressioni can depend on anything which is valid during
\addplot, in particular \coordindex or table columns (\thisrow{hcolumn namei} and friends). The
expression doesn’t need to depend on x at all.
/pgfplots/@restrict to domain={hfilter namei}{hexpressioni}{hhmini:hmax ii}0|1

A low–level (technical) key which allows to apply the restrict * to ... features also to something
like hist/data.
For example, @restrict to domain={hist/data}{}{0:1}{0} applies the domain-restriction to the
histogram-input hist/data. The final ‘0’ means that it works in a similar way as the key restrict
x to domain=0:1, i.e. it skips everything which is outside of [0, 1]. In a similar way, @restrict
to domain={hist/data}{}{0:1}{1} applies the functionality of restrict x to domain*=0:1 to
hist/data: it truncates values outside of [0, 1] to the domain’s end-points.
The hfilter namei is expected to be a coordinate name like x, y, z (or hist/data).
The hexpressioni configures an expression which will be used rather than the value of hfilter namei. It
can be empty.
The hmini:hmax i are as described above.
If the last argument is 1, any coordinate outside of the allowed domain will take the domain boundary
as value. If it is 0, such a coordinate will get either inf or -inf.
/pgfplots/filter discard warning=true|false (initially true)

Issues a notification in your logfile whenever coordinate filters discard coordinates.
You can find somewhat more on coordinate filtering in Section 4.5.14: “Interrupted Plots”.
4.23 Transforming Coordinate Systems

Usually, pgfplots works with cartesian coordinates. However, one may want to provide coordinates in a
di↵erent coordinate system.
In this case, the data cs key can be used to identify the input coordinate system:
/pgfplots/data cs=cart|polar|polarrad (initially cart)

Defines the coordinate system (‘cs’) of the input coordinates. pgfplots will apply transformations if
the argument does not match the expected coordinate system.
Use data cs if your input has a di↵erent coordinate system than the axis. More precisely, every axis
type has its own coordinate system. For example, a normal axis has the cart coordinate system,
whereas a polaraxis has a polar coordinate system. The use of data cs with a di↵erent argument
than the default of your axis instructs pgfplots to apply transformations.
At the time of this writing, pgfplots supports the following values for data cs:
The data cs=cart denotes the cartesian coordinate system. It is the coordinate system of the usual
axis (or its logarithmic variants). It can have three components, x, y, and z. Specifying it is only
necessary if you have a non-cartesian axis:
90
120 60 % requires \usepgfplotslibrary{polar}
\begin{tikzpicture}
\begin{polaraxis}
150 30 \addplot+[data cs=cart]
coordinates {(1,0) (0.5,0.5)};
\end{polaraxis}
\end{tikzpicture}
0 0.5 1
180 0
210 330
240 300
270
The data cs=polar is the (two–dimensional) coordinate system with (angle, radius), i.e. the first com-
ponent “x” is the angle and the second component “y” is the radius. The angle is a number in the
periodic range [0, 360); the radius is any number. If a polar coordinate has a z component, it is taken
as-is (the transformations ignore it).
\begin{axis}
\addplot+[data cs=polar,domain=0:360] (\x,1);
\end{axis}
\end{tikzpicture}
0
1
1 0.5 0 0.5 1
The data cs=polarrad is similar to polar, but it expects the angle in radians, i.e. in the periodic
range [0, 2⇡).
4.23. TRANSFORMING COORDINATE SYSTEMS 373
\begin{axis}
\addplot+[data cs=polarrad,domain=0:2*pi] (\x,1);
\end{axis}
\end{tikzpicture}
0
1
1 0.5 0 0.5 1
Note that the math function deg(hrad i) transforms hrad i into degrees and rad(hdegreei) transforms
hdegreei into radians. Consequently, polar and polarrad are more-or-less equivalent for plot expression.
\begin{axis}[
axis equal,
20 minor tick num=1,
]
\def\FREQUENCY{3}
0 \addplot[red,domain=0:360,samples=200,
smooth,data cs=polar]
(x,{30-8*sin(\FREQUENCY*x)});
20
\addplot[samples=40,domain=0:2*pi,dashed,
data cs=polar] (deg(x),30);
40 20 0 20 40 \addplot[mark=oplus,only marks] coordinates {(0,0)};

\end{axis}
\end{tikzpicture}
At the point of this writing, the data cs method will work for most plot handlers. But for complicated
plot handlers, further logic may be needed which is not yet available (for example, the quiver plot
handler might not be able to convert its direction vectors correctly)65 .
\pgfplotsaxistransformcs{hfromnamei}{htonamei}
Expects the current point in a set of keys, provided in the coordinate system hfromnamei and replaces
them by the same coordinates represented in htonamei.
On input, the coordinates are stored in /data point/x, /data point/y, and /data point/z (the latter
may be empty). The macro will test if there is a declared coordinate transformation from hfromnamei
to htonamei and invoke it. If there is none, it will attempt to convert to cart first and then from cart
to htonamei. If that does not exist either, the operation fails.
\pgfplotsdefinecstransform{hfromnamei}{htonamei}{hcodei}
Defines a new coordinate system transformation. The hcodei is expected to get input and write output
as described for \pgfplotsaxistransformcs.
Implementing a new coordinate system immediately raises the question in which math mode the oper-
ations shall be applied. pgfplots supports di↵erent so–called “coordinate math systems” for generic
operations, and for each individual coordinate as well. These coordinate math systems can either use
basic pgf math arithmetics, the fpu, or perhaps there will come a LuaTEX library.
The documentation of this system is beyond the scope of this manual66 . Please consider reading the
source-code comments and the source of existing transformations if you intend to write own transfor-
mations.
65 In case you run into problems, consider writing a bug report or ask others in TEX online discussion forums.
66 Which is quite comprehensive even without API documentation, as you will certainly agree...
4.23.1 Interaction of Transformations

There are a couple of coordinate mappings in pgfplots. For each encountered coordinate in a coordinate
stream (\addplot), it applies the following steps:
1. Remember the “raw” coordinates in math constants rawx, rawy, rawz,
2. Apply pre filter,
3. Apply x coord trafo, logarithm if necessary, and x filter (in this order),
4. Apply y coord trafo, logarithm if necessary, and y filter,

5. Apply z coord trafo, logarithm if necessary, and z filter,
6. Apply filter point,
7. Transfrom from data cs to the coordinate system of axis type,
8. Handle coordinate stacking (stack plots=x and/or stack plots=y).

Here, pre filter takes no arguments; it simply prepares the following filters. Consequently, the first item
which actually accepts the input argument is x coord trafo. This method is part of the parsing; it accepts
the “raw” coordinate which may be in symbolic form. The output of x coord trafo and its variants is a
number. This number can be filtered or transformed by means of x filter and its variants. Note that x
filter simply takes one argument: the result of x coord trafo.
The intented meaning of x coord trafo is to define how the “raw” string form as provided by the user
makes its way into pgfplots. It is also the only transformation which has an inverse, the key x coord
inv trafo. The inverse can be used to transform back from internal numeric form to some string for the
end user. The key x coord trafo can only be defined as option to an axis, i.e. it it cannot be provided to
\addplot.
The meaning of x filter and its variants is to transform numbers or to conditionally throw away
numbers (and thus the entire coordinate). Each \addplot can receive its own (set of) filters.
The key filter point accepts all arguments (more precisely: the result of the preceding steps). It can
rely on x, y, and z, and it can change any of them. The way it accepts the coordinates is by means of keys
/data point/x and its variants for y and z. In principle, filter point is the most general filter. However,
you can define both x filter and filter point, and both will be applied. Each \addplot can receive its
own filter point.
The implementation for the key data cs is similar to filter point in that it accepts the result of all
previous mapping steps. However, it maps from a well–defined input coordinate system to some well–defined
internal coordinate system (which is inherent to the axis). Each \addplot can receive its own data cs.
Finally, stack plots takes measures to add/subtract the results. Stacking of plots is a property of the
axis; it is not intented to be provided as argument to \addplot67 .
4.24 Fitting Lines – Regression

This section documents the attempts of pgfplots to fit lines to input coordinates. pgfplots currently
supports create col/linear regression applied to columns of input tables. The feature relies on Pgf-
plotsTable, it is actually implemented as a table postprocessing method.
/pgfplots/table/create col/linear regression={hkey-value-configi}

A style for use in \addplot table which computes a linear (least squares) regression y(x) = a · x + b
using the sample data (xi , yi ) which has to be specified inside of hkey-value-configi (see below).
It creates a new column on-the-fly which contains the values y(xi ) = a · xi + b. The values a and b will
be stored (globally) into \pgfplotstableregressiona and \pgfplotstableregressionb.
67 Also it appears to do something which is not entirely useless.
4.24. FITTING LINES – REGRESSION 375
y(x)
7 · x 9.33
20
2 4 6
\begin{tikzpicture}
\begin{axis}[legend pos=outer north east]
\addplot table {% plot X versus Y. This is original data.
X Y
1 1
2 4
3 9
4 16
5 25
6 36
};
\addplot table[
y={create col/linear regression={y=Y}}] % compute a linear regression from the input table
{
X Y
1 1
2 4
3 9
4 16
5 25
6 36
};
%\xdef\slope{\pgfplotstableregressiona} %<-- might be handy occasionally
\addlegendentry{$y(x)$}
\addlegendentry{%
$\pgfmathprintnumber{\pgfplotstableregressiona} \cdot x
\pgfmathprintnumber[print sign]{\pgfplotstableregressionb}$}
\end{axis}
\end{tikzpicture}
The example above has two plots: one showing the data and one containing the linear regression
line. We use y={create col/linear regression={}} here, which means to create a new column68
containing the regression values automatically. As arguments, we need to provide the y column name ex-
plicitly69 . The x value is determined from context: linear regression is evaluated inside of \addplot
table, so it uses the same x as \addplot table (i.e. if you write \addplot table[x={hcol namei}],
the regression will also use hcol namei as its x input). Furthermore, it shows the line parameters a and
b in the legend.
Note that the uncommented line with \xdef\slope{\pgfplotstableregressiona} is useful if you have
more than one regression line: it copies the value of \pgfplotstableregressiona (in this case) into a
new global variable called ‘\slope’. This allows to use ‘\slope’ instead of \pgfplotstableregressiona
– even after \pgfplotstableregressiona has been overwritten.
The following hkey-value-configi keys are accepted as comma–separated list:
/pgfplots/table/create col/linear regression/table={h\macro or file namei} (initially empty)

Provides the table from where to load the x and y columns. It defaults to the currently processed
one, i.e. to the value of \pgfplotstablename.
68 The y={create col/ feature is available for any other PgfplotsTable postprocessing style, see the create on use docu-
mentation in the PgfplotsTable manual.

69 In fact, pgfplots sees that there are only two columns and uses the second by default. But you need to provide it if there
are at least 3 columns.

/pgfplots/table/create col/linear regression/x={hcolumni} (initially empty)

/pgfplots/table/create col/linear regression/y={hcolumni} (initially empty)
Provides the source of xi and yi data, respectively. The argument hcolumni is usually a column
name of the input table, yet it can also contain [index]hinteger i to designate column indices
(starting with 0), create on use specifications or aliases (see the PgfplotsTable manual for
details on create on use and alias).
The initial configuration (an empty value) checks the context where the linear regression is
evaluated. If it is evaluated inside of \pgfplotstabletypeset, it uses the first and second table
columns. If it is evaluated inside of \addplot table, it uses the same x input as the \addplot
table statement. The y key needs to be provided explicitly (unless the table has only two columns).
/pgfplots/table/create col/linear regression/xmode=auto|linear|log (initially auto)

/pgfplots/table/create col/linear regression/ymode=auto|linear|log (initially auto)
Enables or disables processing of logarithmic coordinates. Logarithmic processing means to apply
ln before computing the regression line and exp afterwards.
The choice auto checks if the column is evaluated inside of a pgfplots axis. If so, it uses the axis
scaling of the embedding axis. Otherwise, it uses linear.
In case of logarithmic coordinates, the log basis x and log basis y keys determine the basis.
100 y(x) \begin{tikzpicture}

slope 0.57 \begin{loglogaxis}
\addplot table[x=dof,y=error2]
1 {pgfplotstable.example1.dat};
10 \addlegendentry{$y(x)$}
\addplot table[
2 x=dof,
10
y={create col/linear regression={y=error2}}]
{pgfplotstable.example1.dat};
3
10 % might be handy occasionally:
%\xdef\slope{\pgfplotstableregressiona}
101 102 103 104 105 106 \addlegendentry{slope
$\pgfmathprintnumber{\pgfplotstableregressiona}$}
\end{loglogaxis}
\end{tikzpicture}
The (commented) line containing \slope is explained above; it allows to remember di↵erent re-
gression slopes in our example.
/pgfplots/table/create col/linear regression/variance list={hlisti} (initially empty)

/pgfplots/table/create col/linear regression/variance={hcolumn namei} (initially empty)
Both keys allow to provide uncertainties (variances) to single data points. A high (relative) variance
indicates an unreliable data point, a value of 1 is standard.
The variance list key allows to provide variances directly as comma–separated list, for example
variance list={1000,1000,500,200,1,1}.
The variance key allows to load values from a table hcolumn namei. Such a column name is
(initially, see below) loaded from the same table where data points have been found. The hcolumn
namei may also be a create on use name.
4.25. MISCELLANEOUS OPTIONS 377
100 y(x) \begin{tikzpicture}
slope 0.52 \begin{loglogaxis}
\addplot table[x=dof,y=error2]
1 {pgfplotstable.example1.dat};
10
\addlegendentry{$y(x)$}
\addplot table[
2 x=dof,
10
y={create col/linear regression={
y=error2,
variance list={1000,800,600,500,400}}
3
10 }
]
101 102 103 104 105 106 {pgfplotstable.example1.dat};
\addlegendentry{slope
$\pgfmathprintnumber{\pgfplotstableregressiona}$}
\end{loglogaxis}
\end{tikzpicture}
If both, variance list and variance are given, the first one will be preferred. Note that it is not
necessary to provide variances for every data point.
/pgfplots/table/create col/linear regression/variance src={h\table or file namei} (initially

empty)
Allows to load the variance from another table. The initial setting is empty. It is acceptable if
the variance column in the external table has fewer entries than expected, in this case, only the
first ones will be used.
Limitations: Currently, pgfplots supports only linear regression, and it only supports regression together
with \addplot table. Furthermore, long input tables might need quite some time.
4.25 Miscellaneous Options

/pgfplots/disablelogfilter=true|false (initially false, default true)
Disables numerical evaluation of log(x) in TEX. If you specify this option, any plot coordinates and tick
positions must be provided as log(x) instead of x. This may be faster and – possibly – more accurate
than the numerical log. The current implementation of log(x) normalizes x to m · 10e and computes
log(x) = log(m) + e log(10)
where y = log(m) is computed with a Newton method applied to exp(y) m. The normalization involves
string parsing without TEX-registers. You can safely evaluate log(1 · 10 7 ) although TEX-registers would
produce an underflow for such small numbers.
/pgfplots/disabledatascaling=true|false (initially false, default true)

Disables internal re-scaling of input data. Normally, every input data like plot coordinates, tick positions
or whatever, are parsed without using TEX’s limited number precision. Then, a transformation like
T (x) = 10q m
·x a
is applied to every input coordinate/position where m is “the order of x” base 10. Example: x = 1234 =
1.234 · 103 has order m = 4 while x = 0.001234 = 1.234 · 10 3 has order m = 2. The parameter q is
the order of the axis’ width/height.
The e↵ect of the transformation is that your plot coordinates can be of arbitrary magnitude like
0.0000001 and 0.0000004. For these two coordinates, pgfplots will use 100pt and 400pt internally.
The transformation is quite fast since it relies only on period shifts. This scaling allows precision beyond
TEX’s capabilities.
The option “disabledatascaling” disables this data transformation. This has two consequences: first,
coordinate expressions like (haxis cs:x,yi) have the same e↵ect as (hx,yi), no re-scaling is applied.
Second, coordinates are restricted to what TEX can handle70 .
So far, the data scale transformation applies only to normal axes (logarithmic scales do not need it).
/pgfplots/execute at begin plot={hcommandsi}

This axis option allows to invoke hcommandsi at the beginning of each \addplot command. The
argument hcommandsi can be any TEX content.
You may use this in conjunction with x filter=... to reset any counters or whatever. An example
would be to change every 4th coordinate.
/pgfplots/execute at end plot={hcommandsi}

This axis option allows to invoke hcommandsi after each \addplot command. The argument
hcommandsi can be any TEX content.
/pgfplots/execute at begin axis={hcommandsi}

Allows to invoke hcommandsi at the end of \begin{axis} (or the other “begin axis” statements).
The statement is execute as (almost) last statement before the preparation has been completed.
/pgfplots/execute at end axis={hcommandsi}

The counterpart for execute at begin axis. The hook is actually superfluos, it is executed immedi-
ately after before end axis. It is executed in the same TEX group as execute at begin axis.
/pgfplots/execute at begin plot visualization={hcommandsi}

Allows to add customized code which is executed at the beginning of each plot visualization. In con-
trast to execute at begin plot, this happens not immediately during \addplot, but late during the
postprocessing of \end{axis} when actual drawing commands are generated.
One possible application is shown below71 : suppose you want to use \usepackage{ocg} in order to
switch layers dynamically, for example in a beamer package. This can be implemented as follows:
Dynamic PDF Layer Support (see Acrobat Layers)
10
5 0
0 0.5
0 0.2 0.4
0.6 0.8
11
70 Please note that the axis’ scaling requires to compute 1/(xmax xmin ). The option disabledatascaling may lead to
overflow or underflow in this context, so use it with care! Normally, the data scale transformation avoids this problem.
71 This example from the game theory was provided by Pavel Střı́ž.
% requires \usepackage[pdftex]{ocg}
\begin{tikzpicture}
\begin{axis}[
title=Dynamic PDF Layer Support (see Acrobat Layers),
view={110}{35}]
\addplot3+[
execute at begin plot visualization=\begin{ocg}{First Layer}{FirstLayer}{0},
execute at end plot visualization=\end{ocg},
]
coordinates {(0,0,12) (0,1,2) (1,0,6) (0,0,12)};
\addplot3+[
execute at begin plot visualization=\begin{ocg}{Second Layer}{SecondLayer}{0},
]
coordinates {(0,0,9) (0,1,8) (1,0,4) (0,0,9)};
\addplot3+[
execute at begin plot visualization=\begin{ocg}{Third Layer}{ThirdLayer}{0},
]
coordinates {(0,0,1) (0,1,7) (1,0,3) (0,0,1)};
\end{axis}
\end{tikzpicture}
The execute * hooks insert the ocg-statements at the correct positions, and the single plot commands
are added to di↵erent dynamic layers. Use the Acrobat Reader and its “Layers” Tab to switch each
of them on or o↵. Note that it would not be enough to add the \begin{ocg}... statements right
into the text since pgfplots postpones drawing commands until \end{axis} (splitting of survey and
visualization phase).
See http://www.texample.net/weblog/2008/nov/02/creating-pdf-layers for more details on ocg
and how to obtain it.
Technical note: these hooks are also inserted for \pgfplotsextra commands.
/pgfplots/execute at end plot visualization={hcommandsi}

This is the counter–part of execute at begin plot visualization.
/pgfplots/forget plot={htrue,falsei} (initially false)

Allows to include plots which are not remembered for legend entries, which do not increase the number
of plots and which are not considered for cycle lists.
A forgotten plot can be some sort of decoration which has a separate style and does not influence the
axis state, although it is processed as any other plot. Provide this option to \addplot as in the following
example.
New Experiments (old in gray) \begin{tikzpicture}
\begin{loglogaxis}[
e1 % some descriptions:
e2 table/x=Basis,
1 e3
10 table/y={L2/r},
relative Error
xlabel=Degrees of Freedom,
ylabel=relative Error,
10 3 title=New Experiments (old in gray),
legend entries={$e_1$,$e_2$,$e_3$}
]
5 \addplot[black!15,forget plot]
10 table {plotdata/oldexperiment1.dat};
\addplot[black!15,forget plot]
table {plotdata/oldexperiment2.dat};
7
10 \addplot[black!15,forget plot]
100 101 102 103 104 105 106 107 table {plotdata/oldexperiment3.dat};
\addplot table {plotdata/newexperiment1.dat};
Degrees of Freedom \addplot table {plotdata/newexperiment2.dat};
\addplot table {plotdata/newexperiment3.dat};
\end{loglogaxis}
\end{tikzpicture}
Since forgotten plots won’t increase the plot index, they will use the same cycle list entry as following
plots.
The style every forget plot can be used to configure styles for each such plot:
New Experiments (old in transparent) \begin{tikzpicture}
\begin{loglogaxis}[
e1 forget plot style={opacity=0.2},
e2 % same as above:
1 e3
10 table/x=Basis,
relative Error
table/y={L2/r},
xlabel=Degrees of Freedom,
10 3 ylabel=relative Error,
title=New Experiments (old in transparent),
legend entries={$e_1$,$e_2$,$e_3$},
5 ]
10 \foreach \exp in {1,2,3} {
\addplot+[forget plot]
table {plotdata/oldexperiment\exp.dat};
7
10 \addplot table {plotdata/newexperiment\exp.dat};
100 101 102 103 104 105 106 107 }
\end{loglogaxis}
Degrees of Freedom \end{tikzpicture}
Here, the \addplot+ command means we are using the same cycle list as the following plot and
forget plot style modifies every forget style and yields transparency of the “old experiments”.
Please note that every plot no hindex i styles are not applicable here.
A forgotten plot will be stacked normally if stack plots is enabled!
/pgfplots/before end axis/.code={h... i}

Allows to insert hcommandsi just before the axis is ended (see also execute at end axis). This option
takes e↵ect inside of the clipped area.
before end axis/.code={
20 \fill[red] (1,10) circle(5pt);
\node at (-4,10)
{\large This text has been inserted
using \texttt{before end axis}.};
}}}
This text has been10
inserted using before end axis. \begin{tikzpicture}
\begin{axis}
\addplot {x^2};
\end{axis}
0 \end{tikzpicture}
6 4 2 0 2 4 6
/pgfplots/after end axis/.code={h... i}

Allows to insert hcommandsi right after the end of the clipped drawing commands. While before end
axis has the same e↵ect as if hcommandsi had been placed inside of your axis, after end axis allows
to access axis coordinates without being clipped.
20
10
This text has been inserted

0 using after end axis.
6 4 2 0 2 4 6
after end axis/.code={
\fill[red] (1,10) circle(5pt);
\node at (-4,10)
{\large This text has been inserted using \texttt{after end axis}.};
}}}
\begin{tikzpicture}
\begin{axis}
\addplot {x^2};
\end{axis}
\end{tikzpicture}
/pgfplots/axis on top=true|false (initially false)

If set to true, axis lines, ticks, tick labels and grid lines will be drawn on top of plot graphics.
\begin{axis}[
axis on top=true,
axis x line=middle,
axis y line=middle]
\addplot+[fill] {x^3} \closedcycle;
\end{axis}
4 2 2 4 \end{tikzpicture}
100
\begin{axis}[
axis on top=false,
axis x line=middle,
axis y line=middle]
\addplot+[fill] {x^3} \closedcycle;
\end{axis}
4 2 2 4 \end{tikzpicture}
100
Please note that this feature does not a↵ect plot marks. I think it looks unfamiliar if plot marks are
crossed by axis descriptions.
/pgfplots/visualization depends on=h\macroi (initially empty)

/pgfplots/visualization depends on=hexpressioni\ash\macroi (initially empty)
/pgfplots/visualization depends on=value hcontenti\ash\macroi (initially empty)
Allows to communicate data to pgfplots which is essential to perform the visualization although
pgfplots isn’t aware of it.
Suppose you want a scatter plot, which depends on the (x, y) coordinates, the point meta data to draw
individual colors and furthermore data which influences the mark size. Thus, you need a total of 4
coordinates for every data point, although pgfplots supports only 3 in its initial configuration.
Before we actually come to the main point of the problem, we’ll talk about how to get a scatter plot
which has individual colors and individual sizes. It is not sufficient to set mark size alone, since mark
size is evaluated only once, before markers are processed (the same holds for every mark). Thus, we
can use scatter combined with
scatter/@pre marker code/.append style={/tikz/mark size=\perpointmarksize}.
The @pre marker code is installed for every marker of a scatter plot individually. Now, we come to the
problem as such: where can we get the value for mark size, in our case called \perpointmarksize?
A solution is visualization depends on (using the second input syntax at this point):
\begin{axis}
\addplot+[
scatter,
scatter src=y,
samples=40,
0
visualization depends on=
{5*cos(deg(x)) \as \perpointmarksize},
scatter/@pre marker code/.append style=
{/tikz/mark size=\perpointmarksize}
]
1 {sin(deg(x))};
\end{axis}
Here, we define \perpointmarksize as 5*cos(deg(x)). The expression will be evaluated together with
all other coordinates. Thus, everything which is available during the survey phase can be used here.
This includes the final coordinates x, y, z; the constant meta expands to the current per point meta
data. Furthermore, \thisrow{hcolnamei} expands to the value of a table column.
The command visualization depends on evaluates and remembers every value in internal data
structures. The remembered value is then available as h\macroi during the visualization phase. In
our example, the @pre marker code is evaluated during the visualization phase and applies mark
size=5*cos(deg(x)).
The first syntax, visualization depends on=h\macroi, tells pgfplots to use an already defined
h\macroi. The second syntax with hcontenti\ash\macroi provides also the value.
There can be more than one visualization depends on phrase.
In case the stored value is not of numerical type72 , you can use the prefix ‘value’ before the argument,
i.e.
visualization depends on=value h\macroi or
visualization depends on=value hcontenti\as h\macroi.
Such a value will be expanded and stored, but not parsed as number (at least not by pgfplots).
/pgf/fpu={htrue,falsei} (initially true)

This key activates or deactivates the floating point unit. If it is disabled (false), the core pgf math
engine written by Mark Wibrow and Till Tantau will be used for plot expression. However, this
engine has been written to produce graphics and is not suitable for scientific computing. It is limited
to fixed point numbers in the range ±16384.00000.
If the fpu is enabled (true, the initial configuration) the high-precision floating point library of pgf
written by Christian Feuersänger will be used. It o↵ers the full range of IEEE double precision computing
in TEX. This FPU is also part of PgfplotsTable, and it is activated by default for create col/expr
and all other predefined mathematical methods.
72 Or if it is just a constant and you’d like to improve speed.
4.26. TI KZ INTEROPERABILITY 383
Use
\pgfkeys{/pgf/fpu=false}
in order to de-activate the extended precision. If you prefer using the fp (fixed point) package, possibly
combined with Mark Wibrows corresponding pgf library, the fpu will be deactivated automatically.
Please note, however, that fp has a smaller data range (about ±1017 ) and may be slower.
4.26 Tik Z Interoperability

pgfplots uses Tik Z/pgf as its “backend layer”. This implies that it inherits most of Tik Z’s graphical
features and adds a lot of own stu↵ on top of it. However, the coordinate systems of Tik Z and pgfplots do
not match up – for good reason: pgfplots operates on logical (data) coordinates whereas Tik Z operates
on image coordinates.
Occasionally, one may want to synchronize both in order to generate a graphic. In this context, “synchro-
nize both” means that they really use the same coordinate systems. This is far beyond the simple use-case
of “enrich a pgfplots picture by means of Tik Z annotations”, compare Section 4.17.
Consequently, this section addresses the question how to match the coordinates from Tik Z to pgfplots
and vice–versa. It explains how to match coordinates and it discusses the necessary configuration.
There are a couple of keys in pgfplots which control the mapping of coordinates. The purpose of these
keys is to implement visualization techniques, but they do things di↵erent than Tik Z (and they should). To
match coordinates with Tik Z, one needs the following aspects:
1. Restrict your visualization type: a logarithmic axis simply may not fit into Tik Z (to be more precise:
it may fit, but a Tik Z unit will correspond to a log–unit in pgfplots).
2. Configure matching unit vectors by means of the x and y keys. The default configuration of Tik Z is to
use x=1cm,y=1cm,z={(0,0)}. Note that these settings are usually overridden by pgfplots in order
to respect width and height (and view for three–dimensional axes).
3. Disable the data scaling by means of disabledatascaling: pgfplots will internally apply linear
coordinate transformations in order to provide the data range required for floating point arithmetics
(using approximately floating point precision). Disabling the data scaling means to restrict yourself to
the (small) data range supported by Tik Z—but that’s probably what you want in that case.
4. Define anchor and position of the axis, probably using anchor=origin,at={(0,0)}. The at={(0,0)}
configures pgfplots to place the axis at the Tik Z position (0,0) whereas anchor=origin means that
pgfplots will place its data origin (0, 0, 0) at the place designated by at (see Section 4.19 for details).
5. Make sure that the pgfplots axis contains the data origin (0, 0, 0) in the displayed data range (i.e.
configure xmin, xmax, ymin, and ymax appropriately).
Without this, the anchor=origin key required in the previous item will be truncated to the next
coordinate which is part of the displayed range.
Here is a simple example, first with Tik Z:

\begin{tikzpicture}
\coordinate (Point) at (1,2);
(1,2) \draw [gray] (-1,-1) grid (3,3);

\draw [blue,fill] (Point) circle (2pt)
node [right] {(1,2)};
\end{tikzpicture}
it displays a grid with x, y 2 [ 1, 3] and shows a node inside of it. Now, we apply the keys discussed above
to match this setting in pgfplots:
3
\begin{tikzpicture}
\coordinate (Point) at (1,2);
2 (1,2)
\begin{axis}[
% tell pgfplots to "grab" the axis at its
% internal (0,0) coord:
1
anchor=origin,
% tell pgfplots to place its anchor at (0,0):
% (This is actually the default and can
0 (-1,0) (2,0)
% be omitted)
at={(0pt,0pt)},
1 % tell pgfplots to use the "natural" dimensions:
1 0 1 2 3 disabledatascaling,
% tell pgfplots to use the same unit vectors
% as tikz:
x=1cm,y=1cm,
%
% this is just as usual in pgfplots. I guess
% it is only useful if (0,0) is part of the
% range... try it out.
xmin=-1,xmax=3, ymin=-1,ymax=3,grid=both]
% this uses the point defined OUTSIDE of the axis
\draw [blue,fill] (Point) circle (2pt)
% this uses a TIKZ coordinate (2,0) in the axis:

\draw [blue,fill] (2,0) circle (2pt)
% this here will always work inside of an axis:

\draw [blue,fill] (-1,0) circle (2pt)
node [right] {(-1,0)};
\end{axis}
\end{tikzpicture}
The example demonstrates several things: first, it defines a coordinate in the enclosing tikzpicture and
uses it inside of the axis (at the correct position). Second, it uses the standard Tik Z coordinate (2,0) inside
of the axis, and it is placed at the expected position. Third, it uses the approach provided by pgfplots
by using the axis cs to designate a coordinate (this last approach does also work without the coordinate
matching).
Here is an example which inserts a pgfplots graphics correctly into a tikzpicture:
second
third
first
fourth
4.27. LAYERS 385
% requires \usepgfplotslibrary{patchplots}
\begin{tikzpicture}
\begin{axis}[
% tell pgfplots to "grab" the axis at its internal (0,0) coord:
anchor=origin,
% tell pgfplots to place its anchor at (0,0):
% (This is actually the default and can be omitted)
at={(0pt,0pt)},
% tell pgfplots to use the "natural" dimensions:
disabledatascaling,
% tell pgfplots to use the same unit vectors as tikz:
x=1cm,y=1cm,
%
hide axis,
]
\addplot[patch,patch type=coons,
shader=interp,point meta=explicit]
coordinates {
(0,0) [0] % first corner
(1,-1) [0] % bezier control point between (0) and (3)
(4,0.7) [0] % bezier control point between (0) and (3)
%
(3,2) [1] % second corner
(7,2) [1] % bezier control point between (3) and (6)
%
(7,1) [2] % third corner
(4.5,-0.5) [2] % bezier control point between (6) and (9)
%
(5,-2) [3] % fourth corner
(4,-2.5) [3] % bezier control point between (9) and (0)
(-1,-2) [3] % bezier control point between (9) and (0)
};
\end{axis}
% this requires pgf 2.10

\begin{scope}[every node/.style={circle,inner sep=2pt,fill=black}]
\node[pin=140:first] at (0,0) {};
\node[pin=second] at (3,2) {};
\node[pin=45:third] at (7,1) {};
\node[pin=0:fourth] at (5,-2) {};
\end{scope}
\end{tikzpicture}
The example employs one of the patch plots of the patchplots library. Since these graphical elements
typically require depth information (z buffering) and color data (point meta), they are only available
inside of pgfplots. However, the configuration above ensures that coordinates match one-to-one between
pgfplots and Tik Z. The hide axis flag disables anything of pgfplots, so only the visualized patch plot
remains73 .
4.27 Layers
It is important that several parts of an axis are drawn “on top” of others. Usually, pgfplots ensures this
by drawing them in a suitable sequence (usually background followed by grid lines, followed by tick lines
and tick labels, followed by plots and finally axis descriptions). While this works reasonable in most cases,
there are cases where more control is desired. One common use-case is if multiple axes shall be drawn into
the same picture: here, the sequence from above should be applied to all involved axes simultaneously.
4.27.1 Summary
This section is the technical reference for using and customizing layered graphics in pgfplots. As such, it
is hard reading.
For most purposes, the following is completely sufficient for you: If you want to enable layered graphics,
put the following statement into the tikzpicture which is supposed to have layered graphics:
73 Note that the (0, 0, 0) coordinate of pgfplots is part of the data range here.
\begin{tikzpicture}
\begin{axis}
...
\end{axis}
% perhaps a second axis which should use the same layers?

\begin{axis}
...
\end{axis}
\end{tikzpicture}
This enables layered graphics for that specific tikzpicture.

You may want layered graphics if you have multiple axes in the same picture, of if you have specific needs
for your plot.
Consider reading on layer if you want to move particular elements of your axis to a di↵erent layer.
4.27.2 Using Predefined Layers

The main key to control layered graphics with pgfplots is set layers:
/pgfplots/set layers=none|hlayer configuration namei (initially none)

This key enables layered graphics for either the current axis or for all following axes.
Enabling layered graphics has the e↵ect that the order in which graphical elements are given is unrelated
to the ordering in which they will be drawn. The main benefit is if you have multiple axes in the same
figure: the axes can share the same layers.
The invocation set layers=none disables layered graphics.
The invocation set layers (without equal sign and without arguments) is the same as if you would
write set layers=default.
In all other cases, set layers expects a hlayer configuration namei. There are two predefined configu-
rations available (the prefix /pgfplots/layers/ is optional):
/pgfplots/layers/standard (no value)

A layer configuration which defines the layers axis background, axis grid, axis ticks, axis
lines, axis tick labels, main, axis descriptions, axis foreground. They are drawn in the
order of appearance.
/pgfplots/layers/axis on top (no value)

A layer configuration which uses the same layer names as layers/standard, but with a di↵er-
ent sequence: axis background, main, axis grid, axis ticks, axis lines, axis tick labels,
axis descriptions, axis foreground.
This layer is automatically used if the key axis on top is used together with set layers=hany
layer configuration namei.
As soon as the key set layers=hlayer configuration namei is encountered, pgfplots starts the pgf
command \pgfsetlayers{hlayer namesi} with the layer names of the respective configuration. Usually,
this replaces the current layer configuration of the embedding tikzpicture. Furthermore, set layers
stores the name of hlayer configuration namei such that every following axis knows how to map graphical
elements to layer names.
There is one huge di↵erence to any other key which tunes pgfplots: layer configurations are properties
of a complete tikzpicture whereas any other option a↵ects only axis objects and their contents.
Layers, however, a↵ect every graphical element of the embedding picture. Due to this property, layer
configurations need to be given at one of several supported positions:
1. Directly within the picture:
\begin{tikzpicture}
\pgfplotsset{set layers=default}
\begin{axis}
...
\end{axis}
\end{tikzpicture}
4.27. LAYERS 387
This option explicitly tells the reader of your source code that a significant portion of your picture
has been changed: the complete picture has and uses a hlayer configuration namei (in this case
default).
2. As option for one or more axes which is/are directly within the picture:
\begin{tikzpicture}
\begin{axis}[set layers]
...
\end{axis}
\end{tikzpicture}
Here, pgfplots implicitly communicates its layer configuration to the enclosing tikzpicture.
Thus, the e↵ect of set layers is not local to an axis; it survives until \end{tikzpicture}. Any
other option only survives until \end{axis}.
In this case, only the last activated layer configuration will apply to the picture.
Limitation: no environments or local TEX groups allowed. Standard usages as within the
examples of this manual will always work. But since the layer name configuration is essentially
part of a pgf picture (at a low level), one cannot arbitrarily set them; pgf will complain if they
are changed within some nested TEX groups or LATEX environments. Typically, you will never need
to worry about this.
In short, the following examples are forbidden because the axis is within locally nested groups.
\begin{tikzpicture}
{% FORBIDDEN! Consider using case (1) above!
...
\end{axis}
}
\end{tikzpicture}
\begin{tikzpicture}
\begin{scope} % FORBIDDEN! Consider using case (1) above!
...
\end{axis}
\end{scope}
\end{tikzpicture}
These examples are forbidden because the layer configuration will be cleared by the ‘}’ of the first
forbidden example and by the ‘\end{scope}’ of the second example. A solution would be one of
the di↵erent placement options (i.e. choice (1.) or (3.)).
3. outside of any picture:
\pgfplotsset{set layers=default}
\begin{tikzpicture}
\begin{axis}
...
\end{axis}
\end{tikzpicture}
This choice configures the layer configuration for every following tikzpicture.
Limitation: axis alignment restricted to inner anchors. This applies only if you changed the
default value of anchor (which is anchor=south west). Any axis which uses layered graphics should
use one of the following values of anchor: north, north west, west, south west, south, south east,
east, north east, north, center, origin, above origin, left of origin, right of origin, below
origin. In case you really need another anchor, pgfplots requires the use of cell picture=true,
causing the layers to be local for that specific axis.
The technical background for this limitation is a hen-and-egg problem: outer anchors (like outer
south west) are only available after the complete axis has been generated – and layers can only be
drawn after each drawing instruction has been issued. The technical keys for further reading are cell
picture=false or cell picture=if necessary (one of them is active for layered graphics).
\pgfplotssetlayers
An alias for \pgfplotsset{set layers}. It activates the layers/default layer configuration.
\pgfplotssetlayers{hlayer configuration namei}

An alias for \pgfplotsset{set layers={hlayer configuration namei}}.
Key handler hkeyi/.define layer set={hordered layer namesi}{hstyle definitionsi}

Allows to define a new layer set configuration named hkeyi. Afterwards, hkeyi is a valid argument for
set layers=hkeyi.
The first argument hordered layer namesi is a comma-separated list of layer names. The names are
arbitrary, and \pgfdeclarelayer will be called for every encountered argument74 . There is just one
“magic” name: the layer main should be part of every hordered layer namesi as it will contain every
graphical element which is not associated with a specific layer.
The second argument hstyle definitionsi contains options – just as if you would have written
hkeyi/.style={hstyle definitionsi}. The hstyle definitionsi are supposed to contain pgfplots style
redefinitions which make use of each encountered element of hordered layer namesi. This is probably
best explained by an example: the layers/standard layer configuration is defined by
\pgfplotsset{
layers/standard/.define layer set=
{axis background,axis grid,axis ticks,axis lines,axis tick labels,main,%
axis descriptions,axis foreground}
{
grid style= {/pgfplots/on layer=axis grid},
tick style= {/pgfplots/on layer=axis ticks},
axis line style= {/pgfplots/on layer=axis lines},
label style= {/pgfplots/on layer=axis descriptions},
legend style= {/pgfplots/on layer=axis descriptions},
title style= {/pgfplots/on layer=axis descriptions},
colorbar style= {/pgfplots/on layer=axis descriptions},
ticklabel style= {/pgfplots/on layer=axis tick labels},
axis background@ style={/pgfplots/on layer=axis background},
3d box foreground style={/pgfplots/on layer=axis foreground},
},
}
This definition declares a couple of layers, and it adjusts pgfplots styles by adding on layer com-
mands. The arguments for on layer are the elements of hordered layer namesi.
Note that if you have an element in hordered layer namesi which is never referenced inside of hstyle
definitionsi, this layer will always be empty. In other words: the only reference to the names in hordered
layer namesi is hstyle definitionsi, pgfplots has no hard-coded magic layer names (except for main as
explained above).
Since the second argument hstyle definitionsi defines hkeyi to be a normal style key, one can simply use
hkeyi in order to set hstyle definitionsi. This allows to inherit them. For example, the layers/axis on
top layer configuration is defined by means of
\pgfplotsset{
/pgfplots/layers/axis on top/.define layer set=
{axis background,main,axis grid,axis ticks,axis lines,axis tick labels,%
axis descriptions,axis foreground}
{/pgfplots/layers/standard}
}
i.e. it only redefines the sequence of the layers and re-uses the style definitions of layers/standard.
Any number of layer configurations can be defined.
4.27.3 Changing the Layer of Graphical Elements

There are a couple of keys which change the layer of a graphical element.
/pgfplots/on layer={hlayer namei}

74 To be more precise: set layers calls \pgfdeclarelayer when it uses hordered layer namesi.
4.28. TECHNICAL INTERNALS 389
Providing this key somewhere in a pgfplots style or inside of a pgfplots axis will change the layer
for all graphical elements for which the style applies.
For example,
...
\begin{axis}[set layers,grid style={/pgfplots/on layer=axis foreground}]
...
will change the layer for any grid lines to axis foreground.
The argument hlayer namei is expected to be part of the current layer configuration, i.e. the argument
of set layers should contain it.
Note that if you have two plots with di↵erent values of on layer, you may also want to enable clip
mode=clip individual or to deactivate clipping altogether using clip=false. Clipping options need
to be provided as option to the axis, not to the plot. The technical background is that clip paths needs
to be replicated for the layer on which the drawing is supposed to happen – otherwise they will be
applied to the wrong layer.
/pgfplots/mark layer=auto|like plot|hlayer name i (initially auto)

An advanced key which defines the layer for plot marks. It is typically the best choice to leave it at
auto.
If you write \addplot[on layer=hlayer namei], the layer will be used for the complete plot. Plot
marks are treated with special care, so you can define an own layer for plot marks.
The initial choice auto will automatically define a “suitable” choice, leaving the responsability with
pgfplots. Here, “suitable” means to respect clip mode and clip marker paths in a way such that
plot marks will not be clipped even though the default layer for your plot will be clipped.
The choice like plot will pack the marks onto the same layer as the plot they belong to. This might
cause clipped markers, i.e. markers which are only displayed partially if they are close to the boundary
of the axis.
Finally, one can provide any hlayer namei, just as for on layer – but the layer can be di↵erent from
the layer used for the plot.
4.28 Technical Internals

This section describes keys which are usually set by internal routines – it is typically unnecessary to use
them. However, they may impose limitations or influence performance. Such cases are documented clearly
in other sections of this manual. This here is the reference on the involved internals.
/pgfplots/cell picture=true|false|if necessary (initially true)

This key is set automatically by pgfplots if necessary (for example by set layers).
Typically, pgfplots creates a so-called “cell picture”. A cell picture is a separate picture which is
typeset into a node. Finally, the node is shifted to fulfill special anchor requirements. The necessity for
a cell picture is given if the anchor of an axis is only known after the complete axis has been drawn.
The initial choice true means that pgfplots will create a cell picture for every axis. This allows all
anchors, but it is unsuited if multiple graphics layers are desired or if one wants SVG export. In order
to create a cell picture, pgfplots interrupts the embedding tikzpicture, draws a new tikzpicture,
and finally typesets the result into a node.
The choice false tells pgfplots to draw its paths directly into the embedding tikzpicture. Such an
approach is necessary if the axis shall use layers of the embedding tikzpicture. This is possible if and
only if the anchor can be determined without actually drawing the complete axis. If so, pgfplots will
modify the transformation matrix in advance. Note that axes with cell picture=false will contain
all the usual anchors – the only di↵erence is that the axis itsself can only use one of the following
anchors for its alignment: north, north west, west, south west, south, south east, east, north
east, north, center, origin, above origin, left of origin, right of origin, below origin.
The choice if necessary will check if the chosen anchor is one of the list above. If so, it will use cell
picture=false. Otherwise, it will use cell picture=true.
/pgfplots/compat/show suggested version=true|false (initially true)

If enabled, pgfplots will show you which value for compat=hversioni results in the largest active feature
set and highest quality.
This key will generate a warning if the current version is so old that the quality degrades seriously.
The notification will be printed to your .log file (during \end{document}).
Chapter 5
Related Libraries
This section describes some libraries which come with pgfplots, but they are more or less special and need
to be activated separately.
5.1 Clickable Plots

\usepgfplotslibrary{clickable} % LATEX and plain TEX
\usepgfplotslibrary[clickable] % ConTEXt
\usetikzlibrary{pgfplots.clickable} % LATEX and plain TEX
\usetikzlibrary[pgfplots.clickable] % ConTEXt
A library which generates small popups whenever one clicks into a plot. The popup displays the
coordinate under the mouse pointer, supporting the optional snap–to–nearest clickable coords feature
with customizable displayed information. Furthermore, the library allows to display slopes if one holds
the mouse pressed and drags it to another point in the plot.
The library has two purposes: to compute slopes in a simple way1 and to provide related, optional
information to single data points which are not important enough to be listed in the main text (like
prototype parameters or other technical things).
5.1.1 Overview
It is completely sufficient to write
\usepgfplotslibrary{clickable}
in the document preamble. This will automatically prepare every plot.

The library works with Acrobat Javascript and pdf forms: every plot becomes a push–button.
These screenshots show the result of clicking into the axis range (left column) and of dragging from one
point to another (right column). The second case shows the result of Drag-and-Drop: it displays start- and
1 The author is applied mathematician...
391
392 CHAPTER 5. RELATED LIBRARIES
end points and the equation for the line segment between between the first point of the drag- and drop and
the second point where the mouse has been released. The line segment is
l(x; x0 , y0 , x1 , y1 ) = m · x + n
where m = (y1 y0 )/(x1 x0 ) is the slope and n the o↵set chosen such that l(x0 ; . . . ) = y0 . For logarithmic
plots, logarithms will be applied before computing slopes.
These screen shots show the result of drag- and drop for logarithmic axes: the end points show, again, the
coordinates (without logs) and the form field in the middle shows the slope and o↵set of the linear equation
in log coordinates.
The log basis for any logarithmic axes is usually 10, but it respects the current setting of log basis
x and log basis y. The applied log will always use the same logarithm which is also used for the axis
descriptions (this is not necessarily the same as used by PgfplotsTable!).
This document has been produced with the clickable library, so it is possible to load it into Acrobat
Reader and simply click into a plot.
/pgfplots/clickable coords={hdisplayed texti}
Activates a snap–to–nearest feature when clicking onto plot coordinates. The hdisplayed texti is the
coordinate’s x and y value by default (i.e. you write just clickable coords without an equal sign).
% Preamble: \pgfplotsset{width=7cm,compat=1.13}\usepgfplotslibrary{clickable}
\begin{tikzpicture}
\begin{loglogaxis}[clickable coords=
{Level \thisrow{level} (q=\thisrow{q})}]
2
10 \addplot table[x=dof,y=error] {
level dof error q
1 4 2.50000000e-01 48
2 16 6.25000000e-02 25
4
10 3 64 1.56250000e-02 41
4 256 3.90625000e-03 8
5 1024 9.76562500e-04 22
6 4096 2.44140625e-04 46
6
10 7 16384 6.10351562e-05 40
8 65536 1.52587891e-05 3
101 102 103 104 105 106 9 262144 3.81469727e-06 1
10 1048576 9.53674316e-07 9
};
\end{loglogaxis}
\end{tikzpicture}
Now, clicking onto a data point yields ‘Level 7 (q=40)’ whereas clicking besides a data point results in
the click coordinates as before,
.
5.1. CLICKABLE PLOTS 393
Note that logarithmic slopes work as before.

If you want the (x, y) values to be displayed, use the special placeholder string ‘(xy)’ inside of hdisplayed
texti. As an example, we consider again the scatter/classes example of page 112:
\begin{tikzpicture}
\begin{axis}[%
clickable coords={(xy): \thisrow{label}},%
0.4 scatter/classes={%
c={mark=o,draw=black}}]
\addplot[scatter,only marks,%
0.2
scatter src=explicit symbolic]%
table[meta=label] {
x y label
0.1 0.15 a
0.45 0.27 c
0 0.2 0.4 0.6 0.8 0.02 0.17 a
0.06 0.1 a
0.9 0.5 b
0.5 0.3 c
0.85 0.52 b
0.12 0.05 a
0.73 0.45 b
0.53 0.25 c
0.76 0.5 b
0.55 0.32 c
};
\end{axis}
\end{tikzpicture}
Here, we find popups like
.
The hdisplayed texti is a richtext string displayed with Javascript. For most purposes, it is used like an
unformatted C string: it contains characters, perhaps line breaks with ‘\n’ or tabulators with ‘\t’, but it
should not contain TEX formatting instructions, especially no math mode (the ‘(xy)’ replacement text is
formatted with sprintf, see below). Consider clickable coords code in case you’d like to preprocess
data before displaying it. If you experience problems with special characters, try prepending a backslash
to them. If that doesn’t work either, try to prefix the word with ‘\\’ and/or with ‘\string’. Consider
using clickable coords size if you intend to work with multiline fields and the size allocation needs
improvements.
In fact, hdisplayed texti can even contain richtext (=XHTML) formatting instructions like ‘<br/>’ (note
the final slash) or ‘<span style="color:\#7E0000;">text</span>’ (note the backslash before ‘#’)
which changes the color for text. The <span style=""> arguments are CSS fields, consider an HTML
reference for a list of CSS attributes.
It is possible to use clickable coords together with three dimensional axes. Note that dynamic
(clickable) features of a three dimensional axis without clickable coords will be disabled (they ap-
pear to be useless). Furthermore, three dimensional axes do not support slope calculations; only the
snap–to–nearest feature is available.
Consider using annot/snap dist=6 to increase the snap–to–nearest distance.
The clickable coords can be specified for all plots in an axis (as in the examples above), but also once
for every single \addplot commands for which the snap–to–nearest feature is desired (with di↵erent
hdisplayed texti).
If multiple clickable coords are on the same position, each click chooses the next one (in the order
of appearance).
/pgfplots/clickable coords code={hTEX code which defines \pgfplotsretvali}
A variant of clickable coords which allows to prepare the displayed information before it is handed
over to Javascript.
The value should be TEX code which defines \pgfplotsretval somehow. The result is used as simple,
unformatted string which is associated to coordinates.
Consider using
\pgfmathprintnumberto[verbatim]{hnumber i}\macroname
\edef\pgfplotsretval{Number=\macroname}
to provide number printing. The \pgfmathprintnumberto[verbatim] doesn’t use math mode to format
a number2 , and it writes its result into \macroname. The name ‘\macroname’ is arbitrary, use anything
like ‘\eps’ or ‘\info’. The \edef means “expanded definition” and has the e↵ect of expanding all
macros to determine the value, in our case “Number= hthe valuei”. The following example uses it twice
to pretty–print the data:
\begin{tikzpicture}
\begin{loglogaxis}[clickable coords code={%
\pgfmathprintnumberto[verbatim,precision=1]%
2
10 {\thisrow{error}}%
\error%
\pgfmathprintnumberto[verbatim,frac]%
{\thisrow{frac}}%
4
10 \fraccomp%
\edef\pgfplotsretval{error \error, R=\fraccomp}%
}]%
\addplot table[x=dof,y=error] {
6
10 level dof error frac
1 4 2.50000000e-01 0.5
101 102 103 104 105 106 2 16 6.25000000e-02 0.75
3 64 1.56250000e-02 0.1
4 256 3.90625000e-03 0.2
5 1024 9.76562500e-04 0.5
6 4096 2.44140625e-04 0.8
7 16384 6.10351562e-05 0.125
8 65536 1.52587891e-05 0.725
9 262144 3.81469727e-06 0.625
10 1048576 9.53674316e-07 1
};
\end{loglogaxis}
\end{tikzpicture}
resulting in
.
The hTEX codei is evaluated inside of a local scope, all locally declared variables are freed afterwards
(that’s why you can use any names you want).
/pgfplots/clickable coords size=auto or {hmax charsi} or {hmax chars x,max chars yi} (initially
auto)
This is actually just another name for annot/popup size snap, see its documentation below.
5.1.2 Requirements for the Library

• The library relies on the LATEX packages insdljs (“Insert document level Javascript”) and eforms
which are both part of the freely available AcroTeX education bundle [5]3 . The insdljs package creates
a temporary file with extension .djs.
• At the time of this writing, only Adobe Acrobat Reader interpretes Javascript and Forms properly.
The library doesn’t have any e↵ect if the resulting document is used in other viewers (as far as I know).
Note that although this library has been written for pgfplots, it can be used independently of a pgfplots
environment.
2 See the PgfplotsTable manual for details about number printing.
3 These packages rely on LATEX, so the library is only available for LATEX, not for plain TEX or ConTEXt.
5.1. CLICKABLE PLOTS 395
Compatibility issues: There a several restrictions when using this library. Most of them will vanish in
future versions – but up to now, I can’t do magic.
• The library does not yet support rotated axes. Use clickable=false for those axes.
• The library works only with pdflatex; dvips or dvipdfm are not supported4 .
• Up to now, it is not possible to use this library together with the external library and other image
externalization methods of Section 7.
To be more precise, you can (with two extra preamble lines, see below) get correctly annotated, exported
pdf documents, but the \includegraphics command does not import the dynamic features.
In case you decide to use this work–around, you need to insert
% \maxdeadcycles=10000 % in case you get the error ‘Output loop---<N> consecutive dead cycles.’
\usepackage[pdftex]{eforms}
before loading pgf, Tik Z or pgfplots. The \maxdeadcycles appears to be necessary for large docu-
ments, try it out.
As long as you are working on a draft version of your document, you might want to use
\pgfkeys{/pgf/images/include external/.code={\href{file:#1}{\pgfimage{#1}}}}
in your preamble. This will generate hyperlinks around the graphics files which link to the exported
figures. Clicking on the hyperlinks opens the exported figure which, in turn, has been generated with
the clickable library and allows dynamic features5 .
• The library automatically calls \begin{Form} at \begin{document} and \end{Form} at the end of
the document. This environment of hyperref is necessary for dynamic user interaction and should be
kept in mind if the document contains other form elements.
Acknowledgements:
• I have used a Javascript sprintf implementation of Kevin van Zonneveld [7] (the Javascript API has
only a limited set of conversions).
5.1.3 Customization
It is possible to customize the library with several options.
/pgfplots/clickable=true|false (initially true)

Allows to disable the library for single plots.
/pgfplots/annot/js fillColor={hJavascript color i} (initially ["RGB",1,1,.855])

Sets the background (fill) color of the short popup annotations.
Possible choices are transparent, gray, RGB or CMYK color specified as four–element–arrays of the
form ["RGB", hred i,hgreeni,hbluei]. Each color component is between 0 and 1.
Again: this option is for Javascript. It is not possible to use colors as in pgf.
/pgfplots/annot/point format={hsprintf-formati} (initially (%.1f,%.1f))

/pgfplots/annot/point format 3d={hsprintf-formati} (initially (%.1f,%.1f,%.1f))
Allows to provide an sprintf format string which is used to fill the annotations with text. The first
argument to sprintf is the x-coordinate and the second argument is the y-coordinate.
The point format 3d variant is used for any three–dimensional axis whereas the point format is used
(only) for two–dimensional ones.
The every semilogx axis, every semilogy axis and every loglog axis styles have been updated
to
4 In fact, they should be. I don’t really know why they don’t . . . any hint is welcome.
5 This special treatment needs the external files in the same base directory as the main document, so this approach is most
certainly not suitable for a final document.
\pgfplotsset{
every semilogy axis/.append style={/pgfplots/annot/point format={(\%.1f,\%.1e)}},
every semilogx axis/.append style={/pgfplots/annot/point format={(\%.1e,\%.1f)}},
every loglog axis/.append style={/pgfplots/annot/point format={(\%.1e,\%.1e)}}
}
such that every logarithmic coordinate is displayed in scientific format.

/pgfplots/annot/slope format={hsprintf-formati} (initially %.1f*x %+.1f)
Allows to provide an sprintf format string which is used to fill the slope–annotation with text. The
first argument is the slope and the second the line o↵set.
/pgfplots/annot/printable=true|false (initially false)
Allows to configure whether the small annotations will be printed. Otherwise, they are only available
on screen.
/pgfplots/annot/font={hJavascript font namei} (initially font.Times)
Allows to choose a Javascript font for the annotations. Possible choices are limited to what Javascript
accepts (which is not the same as LATEX). The default fonts and its names are shown below.
Font Name Name in Javascript
Times-Roman font.Times
Times-Bold font.TimesB
Times-Italic font.TimesI
Times-BoldItalic font.TimesBI
Helvetica font.Helv
Helvetica-Bold font.HelvB
Helvetica-Oblique font.HelvI
Helvetica-BoldOblique font.HelvBI
Courier font.Cour
Courier-Bold font.CourB
Courier-Oblique font.CourI
Courier-BoldOblique font.CourBI
Symbol font.Symbol
ZapfDingbats font.ZapfD
/pgfplots/annot/textSize={hSize in Pointi} (initially 11)

Sets the text size of annotations in points.
/pgfplots/annot/popup size generic=auto or {hx i} or {hx,yi} (initially auto)
/pgfplots/annot/popup size snap=auto or {hx i} or {hx,yi} (initially auto)
/pgfplots/annot/popup size={hvaluei}
The first key defines the size of popups if you just click into an axis. The second key defines the size of
popups for the snap–to–nearest feature (i.e. those prepared by clickable coords). The third key sets
both to the same hvaluei.
The argument can be auto in which case pgfplots tries to be smart and counts characters. This may fail
for multiline texts. The choice hx i provides the horizontal size only, in units of annot/textSize. Thus,
annot/popup size generic=6 makes the popup 6 · 11 points wide. In this case, only one line will be
allocated. Finally, hx,yi allows to provide horizontal and vertical size, both in units of annot/textSize.
See also clickable coords size which is an alias for annot/popup size snap.
/pgfplots/annot/snap dist={hSize in Pointi} (initially 4)
Defines the size within two mouse clicks are considered to be equivalent, meased in points (Euclidean
distance).
/pgfplots/annot/richtext=true|false (initially true)
Enables or disables richtext formatting in clickable coords arguments. Richtext is kind of XHTML
and allows CSS styles like colors, font changes and other CSS attributes, see the documentation for
clickable coords for details.
The case annot/richtext=false is probably more robust.
5.2. COLORBREWER 397
5.1.4 Using the Clickable Library in Other Contexts

This library provides essentially one command, \pgfplotsclickablecreate which creates a clickable area
of predefined size, combined with Javascript interaction code. It can be used independently of pgfplots.
\pgfplotsclickablecreate[hrequired key-value-optionsi]
Creates an area which is clickable. A click produces a popup which contains information about the
point under the cursor.
The complete (!) context needs to be provided using key-value-pairs, either set before calling this
method of inside of [hrequired key-value-optionsi].
This command actually creates an AcroForm which invokes Javascript whenever it is clicked. A
Javascript Object is created which represents the context (axis limits and options). This Javascript
object is available at runtime.
This method is public and it is not restricted to pgfplots. The pgfplots hook simply initializes the
required key-value-pairs.
This method does not draw anything. It initializes only a clickable area and Javascript code.
The required key-value-pairs are documented below.
Attention: Complete key-value validation is not performed here. It can happen that invalid options
will produce Javascript bugs when opened with Acrobat Reader. Use the Javascript console to find
them.
All options described in the following are only interesting for users who intend to use this library without
pgfplots.
/pgfplots/annot/width={hdimensioni} (initially -)
This required key communicates the area’s width to \pgfplotsclickablecreate. It must be a TEX
dimension like 5cm.
/pgfplots/annot/height={hdimensioni} (initially -)
This required key communicates the area’s height to \pgfplotsclickablecreate. It must be a TEX
dimension like 5cm.
/pgfplots/annot/jsname={hstringi} (initially -)
This required key communicates a unique identifier to \pgfplotsclickablecreate. This identifier is
used to identify the object in Javascript, so there can’t be more than one of them. If it is empty, a
default identifier will be generated.
/pgfplots/annot/xmin={hnumber i}
/pgfplots/annot/xmax={hnumber i}
/pgfplots/annot/ymin={hnumber i}
/pgfplots/annot/ymax={hnumber i} (initially empty)
These required keys communicate the axis limits to \pgfplotsclickablecreate. They should be set to
numbers which can be assigned to a Javascript floating point number (standard IEEE double precision).
/pgfplots/annot/collected plots={hnested arraysi} (initially empty)

The low level interface to implement a snap–to–nearest feature. The value is an array of plots, where
each plot is again an array of coordinates and each coordinate is an array of three elements, x, y and
text. Please consult the code comments for details and examples.
5.2 ColorBrewer
An extension by Vincent A. Traag and Stefan Pinnow
\usepgfplotslibrary{colorbrewer} % LATEX and plain TEX

\usepgfplotslibrary[colorbrewer] % ConTEXt
\usetikzlibrary{pgfplots.colorbrewer} % LATEX and plain TEX
\usetikzlibrary[pgfplots.colorbrewer] % ConTEXt
This library brings the color schemes created by Cynthia Brewer published at http://colorbrewer2.
org/ to PGFPlots. These where originally designed for cartography needs, but are also used in other
kind of data visualization in the mean time. The ColorBrewer schemes are divided into the types
sequential for ordered data progressing from low to high,
diverging to highlight changes from a mean value, and
qualitative where colors have no special order.
They can consist of up to 12 di↵erent data classes, i.e. colors, per scheme and are provided as color
maps as well as as cycle lists.
Diverging colormap/PiYG-11 Sequential colormap/PuBu-9
0
0.2
0.5
0.4
0
0.6
0.5
0.8
1
Diverging colormap/RdGy-11 Sequential colormap/YlOrBr
1 1
0.8
0.5
0.6
0
0.4
0.5
0.2
1 0
Qualitative cycle list/Set1-5 Diverging cycle list/RdYlBu-4
d=2
d=3
d=4
d=5
d=6
5.2.1 Usage
In the following the available schemes are presented in graphical form as “swatches”.
Color Schemes as “Swatches”
A swatch is a matrix showing all available colors for a specific scheme, and the available color compila-
tions.
(short)
scheme
name
AB CD E F GH I J K LM letters to create scheme color name
GnBu all colors of scheme
3
4
5
6
7 colors of scheme GnBu-7
8
9 colormap based on
previous row colors;
CM
(also) accessible by the
short scheme name
numbers to create
(full) scheme name
(number of data classes)
Such swatches are read as follows:

1. At the top left of the block you find the (short) name of the scheme.
2. The following color blocks are the colors the scheme consists of.
3. To get the (full) color name, combine the (short) scheme name with a hyphen and a letter of the
first row, e.g.
\tikz \fill[color=GnBu-H] (0,0) rectangle (1em,2ex);
which results in .
4. To get the full scheme name, which can be used as colormap or cycle list, combine the (short)
scheme name with a hyphen and a number of the first column, e.g. cycle list name=GnBu-7.
The special case of using the (short) scheme name only is also provided and is an alias for the full
scheme name with the highest number, e.g. scheme name GnBu equals GnBu-9.
5. The last row shows a continuous color map based on the previous row, that is in the example
GnBu-9.
6. The rest of the matrix shows the colors used in the corresponding scheme.
Activating Color Schemes
In order to activate a colorbrewer colormap, say, BuGn-5, you have to use the key
colormap/BuGn-5.
This will initialize and select the associated colormap. It will also initialize the associated cycle list
(but will not select it). In order to initialize and select the cycle list of name BuGn-5, you have to
use the key
cycle list/BuGn-5.
This will initialize and select the associated cycle list. It will also initialize (but not select) the
associated colormap.
Note that cycle lists shipped with colorbrewer merely consist of colors. However, a good cycle
list typically also comes with markers and perhaps line style variations. In order to combine a pure
color-based cycle list with markers, you should make use of the features cycle multi list, cycle
multiindex list, and cycle multiindex* list, for example using
\pgfplotsset{
% initialize Set1-5:
cycle list/Set1-5,
% combine it with ’mark list*’:
mark list*\nextlist
Set1-5\nextlist
},
}
Please refer to the reference manual for cycle multiindex* list for details.
5.2.2 Sequential Schemes

Sequential schemes are useful for ordered data progressing low to high.
ABCDE FGH I J K LM ABCDE FGH I J K LM ABCDE FGH I J K LM
BuGn PuRd Blues
3 3 3
4 4 4
5 5 5
6 6 6
7 7 7
8 8 8
9 9 9
CM CM CM
BuPu RdPu Greens
3 3 3
4 4 4
5 5 5
6 6 6
7 7 7
8 8 8
9 9 9
CM CM CM
ABCDE FGH I J K LM ABCDE FGH I J K LM A BCDE FGH I J K LM

GnBu YlGn Greys
3 3 3
4 4 4
5 5 5
6 6 6
7 7 7
8 8 8
9 9 9
CM CM CM
OrRd YlGnBu Oranges
3 3 3
4 4 4
5 5 5
6 6 6
7 7 7
8 8 8
9 9 9
CM CM CM

PuBu YlOrBr Purples
3 3 3
4 4 4
5 5 5
6 6 6
7 7 7
8 8 8
9 9 9
CM CM CM
PuBuGn YlOrRd Reds
3 3 3
4 4 4
5 5 5
6 6 6
7 7 7
8 8 8
9 9 9
CM CM CM
5.2.3 Diverging Schemes

Diverging schemes highlight changes from some mean value.
A B C D E F G H I J K LMN O A B C D E F G H I J K LMN O A B C D E F G H I J K LMN O
BrBG RdGy RdYlBu
3 3 3
4 4 4
5 5 5
6 6 6
7 7 7
8 8 8
9 9 9
10 10 10
11 11 11
CM CM CM
PiYG PuOr RdYlGn
3 3 3
4 4 4
5 5 5
6 6 6
7 7 7
8 8 8
9 9 9
10 10 10
11 11 11
CM CM CM

PRGn RdBu Spectral
3 3 3
4 4 4
5 5 5
6 6 6
7 7 7
8 8 8
9 9 9
10 10 10
11 11 11
CM CM CM
Note that you can adopt point meta min and point meta max such that the colormap’s mean value
fits the data (for example by forcing point meta min=-2 and point meta max=+2).
5.2.4 Qualitative Schemes

Qualitative schemes are useful if colors have no special order, but should be distinguishable.
ABCDE FGH ABCDE FGH I ABCDE FGH
Accent Pastel1 Pastel2
3 3 3
4 4 4
5 5 5
6 6 6
7 7 7
8 8 8
9
ABCDE FGH ABCDE FGH I ABCDE FGH

Dark2 Set1 Set2
3 3 3
4 4 4
5 5 5
6 6 6
7 7 7
8 8 8
9
ABCDE FGH I J K L ABCDE FGH I J K L

Paired Set3
3 3
4 4
5 5
6 6
7 7
8 8
9 9
10 10
11 11
12 12
5.2.5 Interaction with the ColorBrewer website

To find a scheme, e.g. the above chosen GnBu-7 on http://colorbrewer2.org/
• first select the “Nature of your data” type; in this case “sequential”,
• then select the “Number of data classes”, which is 7,
• and last select the corresponding scheme in the “Pick a color scheme” section. (This is a bit tricky,
because you can only see the name at the top of the lower right corner, where all the colors are
listed. When you have selected the right one you should read there “7-class GnBu”. But perhaps
it helps when you know that “Gn” stands for green and “Bu” for blue, so you are searching for a
scheme going from green to blue, which in this case is the third color scheme in the first row of
“Multi-hue”.)
• The color names, e.g. GnBu-H, cannot be extracted from the page directly though, because you
cannot simply count to the number and replace it with the corresponding letter. This intentionally
was avoided, because then one would define the same color multiple times with di↵erent names,
e.g. GnBu-H 3 times.
• If you should need let’s say the 5th color of the scheme GnBu-7 you don’t have to trial and error
or have a look at the manual. You can simply extract this color from the corresponding color map
via the index of colormap feature, e.g.
\tikz \fill[index of colormap={4 of GnBu-7}] (0,0) rectangle (1em,2ex);
which results in . Note that the index is 4 instead of 5 because the index starts with 0. Note
furthermore that you have to invoke \pgfplotsset{colormap/GnBu-7} before using this key.
5.2.6 External Examples

If you want to see some examples, where Brewer has used her schemes, have a look at https://www.
census.gov/population/www/cen2000/atlas/index.html.
\usetikzlibrary{colorbrewer} % LATEX and plain TEX

\usetikzlibrary[colorbrewer] % ConTEXt
A library which contains just the color definitions like GnBu-B. Please refer to Section 5.2.1 for a list of
available colors.
5.3 Colormaps
An extension by Patrick Häcker
\usepgfplotslibrary{colormaps} % LATEX and plain TEX
\usepgfplotslibrary[colormaps] % ConTEXt
\usetikzlibrary{pgfplots.colormaps} % LATEX and plain TEX
\usetikzlibrary[pgfplots.colormaps] % ConTEXt
A small library providing a number of additional colormaps. Many of these colormaps originate from
the free Matlab package “SC — powerful image rendering” of Oliver Woodford.
The purpose of this library is to provide further colormaps to all users and to provide some of them
which are similar to those used by Matlab (®).
/pgfplots/colormap/autumn (style, no value)

\pgfplotsset{
/pgfplots/colormap={autumn}{rgb255=(255,0,0) rgb255=(255,255,0)}
}
/pgfplots/colormap/bled (style, no value)

\pgfplotsset{
/pgfplots/colormap={bled}{rgb255=(0,0,0) rgb255=(43,43,0) rgb255=(0,85,0)
rgb255=(0,128,128) rgb255=(0,0,170) rgb255=(213,0,213) rgb255=(255,0,0)}
}
/pgfplots/colormap/bright (style, no value)

\pgfplotsset{
/pgfplots/colormap={bright}{rgb255=(0,0,0) rgb255=(78,3,100) rgb255=(2,74,255)
rgb255=(255,21,181) rgb255=(255,113,26) rgb255=(147,213,114) rgb255=(230,255,0)
rgb255=(255,255,255)}
}
/pgfplots/colormap/bone (style, no value)

\pgfplotsset{
/pgfplots/colormap={bone}{[1cm]rgb255(0cm)=(0,0,0) rgb255(3cm)=(84,84,116)
rgb255(6cm)=(167,199,199) rgb255(8cm)=(255,255,255)}
}
5.3. COLORMAPS 403
/pgfplots/colormap/cold (style, no value)

\pgfplotsset{
/pgfplots/colormap={cold}{rgb255=(0,0,0) rgb255=(0,0,255) rgb255=(0,255,255)
rgb255=(255,255,255)}
}
/pgfplots/colormap/copper (style, no value)

\pgfplotsset{
/pgfplots/colormap={copper}{[1cm]rgb255(0cm)=(0,0,0) rgb255(4cm)=(255,159,101)
rgb255(5cm)=(255,199,127)}
}
/pgfplots/colormap/copper2 (style, no value)

\pgfplotsset{
/pgfplots/colormap={copper2}{rgb255=(0,0,0) rgb255=(68,62,63) rgb255=(170,112,95)
rgb255=(207,194,138) rgb255=(255,255,255)}
}
/pgfplots/colormap/earth (style, no value)

\pgfplotsset{
/pgfplots/colormap={earth}{rgb255=(0,0,0) rgb255=(0,28,15) rgb255=(42,39,6)
rgb255=(28,73,33) rgb255=(67,85,24) rgb255=(68,112,46) rgb255=(81,129,83)
rgb255=(124,137,87) rgb255=(153,147,122) rgb255=(145,173,164) rgb255=(144,202,180)
rgb255=(171,220,177) rgb255=(218,229,168) rgb255=(255,235,199) rgb255=(255,255,255)}
}
/pgfplots/colormap/gray (style, no value)

\pgfplotsset{
/pgfplots/colormap={gray}{rgb255=(0,0,0) rgb255=(255,255,255)}
}
This colormap is an alias for the standard colormap/blackwhite.

/pgfplots/colormap/hot2 (style, no value)

\pgfplotsset{
/pgfplots/colormap={hot2}{[1cm]rgb255(0cm)=(0,0,0) rgb255(3cm)=(255,0,0)
rgb255(6cm)=(255,255,0) rgb255(8cm)=(255,255,255)}
}
Note that this particular choice ships directly with pgfplots, you do not need to load the
colormaps library for this value.
/pgfplots/colormap/hsv (style, no value)

\pgfplotsset{
/pgfplots/colormap={hsv}{rgb255=(255,0,0) rgb255=(255,255,0) rgb255=(0,255,0)
rgb255=(0,255,255) rgb255=(0,0,255) rgb255=(255,0,255) rgb255=(255,0,0)}
}
/pgfplots/colormap/hsv2 (style, no value)

\pgfplotsset{
/pgfplots/colormap={hsv2}{rgb255=(0,0,0) rgb255=(128,0,128) rgb255=(0,0,230)
rgb255=(0,255,255) rgb255=(0,255,0) rgb255=(255,255,0) rgb255=(255,0,0)}
}
/pgfplots/colormap/jet (style, no value)

\pgfplotsset{
/pgfplots/colormap={jet}{rgb255(0cm)=(0,0,128) rgb255(1cm)=(0,0,255)
rgb255(3cm)=(0,255,255) rgb255(5cm)=(255,255,0) rgb255(7cm)=(255,0,0)
rgb255(8cm)=(128,0,0)}
}
Note that this particular choice ships directly with pgfplots, you do not need to load the
colormaps library for this value.
5.3. COLORMAPS 405
/pgfplots/colormap/pastel (style, no value)

\pgfplotsset{
/pgfplots/colormap={pastel}{rgb255=(0,0,0) rgb255=(120,0,5) rgb255=(0,91,172)
rgb255=(215,35,217) rgb255=(120,172,78) rgb255=(255,176,24) rgb255=(230,255,0)
rgb255=(255,255,255)}
}
/pgfplots/colormap/pink (style, no value)

\pgfplotsset{
/pgfplots/colormap={pink}{rgb255=(0,0,0) rgb255=(12,16,46) rgb255=(62,22,43)
rgb255=(53,53,65) rgb255=(79,72,58) rgb255=(122,80,67) rgb255=(147,91,102)
rgb255=(147,115,140) rgb255=(144,145,154) rgb255=(173,163,146) rgb255=(216,171,149)
rgb255=(250,179,179) rgb255=(255,198,227) rgb255=(246,229,255) rgb255=(255,255,255)}
}
/pgfplots/colormap/sepia (style, no value)

\pgfplotsset{
/pgfplots/colormap={sepia}{rgb255(0cm)=(0,0,0) rgb255(1cm)=(26,13,0)
rgb255(18cm)=(255,230,204) rgb255(20cm)=(255,255,255)}
}
/pgfplots/colormap/spring (style, no value)

\pgfplotsset{
/pgfplots/colormap={spring}{rgb255=(255,0,255) rgb255=(255,255,0)}
}
/pgfplots/colormap/summer (style, no value)

\pgfplotsset{
/pgfplots/colormap={summer}{rgb255=(0,128,102) rgb255=(255,255,102)}
}
/pgfplots/colormap/temp (style, no value)

\pgfplotsset{
/pgfplots/colormap={temp}{rgb255=(36,0,217) rgb255=(25,29,247) rgb255=(41,87,255)
rgb255=(61,135,255) rgb255=(87,176,255) rgb255=(117,211,255) rgb255=(153,235,255)
rgb255=(189,249,255) rgb255=(235,255,255) rgb255=(255,255,235) rgb255=(255,242,189)
rgb255=(255,214,153) rgb255=(255,172,117) rgb255=(255,120,87) rgb255=(255,61,61)
rgb255=(247,40,54) rgb255=(217,22,48) rgb255=(166,0,33)}
}
/pgfplots/colormap/thermal (style, no value)

\pgfplotsset{
/pgfplots/colormap={thermal}{rgb255=(0,0,0) rgb255=(77,0,179) rgb255=(255,51,0)
rgb255=(255,255,0) rgb255=(255,255,255)}
}
/pgfplots/colormap/winter (style, no value)

\pgfplotsset{
/pgfplots/colormap={winter}{rgb255=(0,0,255) rgb255=(0,255,128)}
}
/pgfplots/colormap/viridis high res (style, no value)

A style which installs the colormap “viridis” which has been defined by Stéfan van der Walt and
Nathaniel Smith for Matplotlib. It is designated to be the default colormap for Matplotlib
starting with version 2.0 and is released under the CC06 . This is the original suggestion as released
by the authors.
Please refer to http://bids.github.io/colormap/ for details.
\pgfplotsset{
colormap name=viridis,
}
\pgfplotsset{
colormap/viridis high res,
}
6 https://creativecommons.org/about/cc0
5.4. DATES AS INPUT COORDINATES 407
5.4 Dates as Input Coordinates

\usepgfplotslibrary{dateplot} % LATEX and plain TEX
\usepgfplotslibrary[dateplot] % ConTEXt
\usetikzlibrary{pgfplots.dateplot} % LATEX and plain TEX
\usetikzlibrary[pgfplots.dateplot] % ConTEXt
A library which allows to use dates like 2008-01-01 or dates with time like 2008-01-01 11:35 as input
coordinates in plots. The library converts dates to numbers and tick labels will be pretty-printed dates
(or times).
This library is documented in Section 4.21.2 on page 364.
5.5 Decoration: Soft Clipping

\usepgfplotslibrary{decorations.softclip} % LATEX and plain TEX
\usepgfplotslibrary[decorations.softclip] % ConTEXt
\usetikzlibrary{pgfplots.decorations.softclip} % LATEX and plain TEX
\usetikzlibrary[pgfplots.decorations.softclip] % ConTEXt
Activates decoration=softclip.
A “soft clip” is a part of an input path, namely that part which is inside of the “clip path”. This is
typically known as clipping: you set \clip hpathi; and all following paths are clipped against hpathi.
Soft–clipping is similar, but instead of installing a low–level clip path, it modifies the input path in
a way such that only parts inside of hpathi remain. This makes a di↵erence if decorations are to be
applied. It also makes a di↵erence for fill between/soft clip.
Note that this library is loaded implicitly by the fillbetween library in order to address its fill
between/soft clip key.
Attention: this library is considered to be experimental. It will work for paths which are similar to a
plot, i.e. paths which do not intersect themselves and which have a clear direction. The library might
fail, in general.
An application could be to draw a path twice, but the second time should only a↵ect portions of the
path:
\begin{tikzpicture}
\draw[
postaction={decorate,draw,ultra thick},
decoration={soft clip,soft clip path={
(1.5,-1) rectangle (4,2)
},
},
]
(0,0) -- (1,1) -- (2,1) -- (3,0);
\end{tikzpicture}%
The soft clip feature is tailored for use with fill between. Please refer to the documentation of
fill between/soft clip for more examples and explanation on soft–clipping.
/pgf/decoration/soft clip path=hcorner1 i rectangle hcorner2 i

Assigns the path which is to be used for the soft clip decoration. This argument is mandatory in
order to apply a soft clip decoration.
Please refer to the documentation of fill between/soft clip for details; it has the same syntax and
a similar motivation.
/pgf/decoration/every soft clipped path (style, no value)

A style which is applied just before the reduced path is generated.
5.6 Image Externalization

\usepgfplotslibrary{external} % LATEX and plain TEX
\usepgfplotslibrary[external] % ConTEXt
\usetikzlibrary{pgfplots.external} % LATEX and plain TEX

\usetikzlibrary[pgfplots.external] % ConTEXt
The external library o↵ers a convenient method to export every single tikzpicture into a sepa-
rate .pdf (or .eps). Later runs of LATEX will simply include these graphics, thereby reducing typesetting
time considerably.
This library is documented in more detail in Section 7.1 “Export to pdf/eps”.
The external library has been written by Christian Feuersänger (author of pgfplots). It has been
contributed to Tik Z as general purpose library, so the reference documentation along with all tweaks can
be found in [6, Section “Externalization Library”]. The command \usepgfplotslibrary{external} is
actually just a wrapper which loads \usetikzlibrary{external} or, if this library does not yet exist
because the installed pgf has at most version 2.00, it will load a copy which is shipped with pgfplots.
5.7 Fill between

\usepgfplotslibrary{fillbetween} % LATEX and plain TEX
\usepgfplotslibrary[fillbetween] % ConTEXt
\usetikzlibrary{pgfplots.fillbetween} % LATEX and plain TEX
\usetikzlibrary[pgfplots.fillbetween] % ConTEXt
The fillbetween library allows to fill the area between two arbitrary named plots. It can also identify
segments of the intersections and fill the segments individually.
5.7.1 Filling an Area

\addplot fill between [{hoptions defined with prefix /tikz/fill betweeni}];
\addplot[hoptionsi] fill between [{hoptions defined with prefix /tikz/fill betweeni}] htrailing path
commandsi;
\addplot3 . . .
A special plotting operation which takes two named paths on input and generates one or more paths
resembling the filled area between the input paths.
% Preamble: \pgfplotsset{width=7cm,compat=1.13}\usepgfplotslibrary{fillbetween}
\begin{tikzpicture}
\begin{axis}
\addplot+[name path=A,domain=0:1,samples=2] {x};
2
\addplot+[name path=B] table {
x y
0 2
0.5 -1
0 1 3
};
\addplot fill between[of=A and B];

\end{axis}
0 0.2 0.4 0.6 0.8 1 \end{tikzpicture}
The operation fill between requires at least one input key within hoptions defined with prefix
/tikz/fill betweeni: the two involved paths in the form of=hfirsti and hsecond i. Here, both hfirsti
and hsecond i need to be defined using name path (or name path global). The arguments can be
exchanged7 , i.e. we would achieve the same e↵ect for of=B and A.
The argument hoptions defined with prefix /tikz/fill betweeni can contain any number of options
which have the prefix /tikz/fill between. Note that the prefix refers to the reference manual, you
do not need to type the prefix. This excludes drawing options like fill=orange; these options should
be given as hoptionsi (or inside of styles like every segment). Allowed options include of, split, soft
clip, and style definitions of every segment and its friends, i.e. those which define which paths are to
be acquired and how they should be processed before they can be visualized.
7 Note that some options refer explicitly to either the first or the second input path. These options are documented accord-
ingly.
5.7. FILL BETWEEN 409
A fill between operation takes the two input paths, analyzes their orientation (i.e. are its coordinates
given increasing in x direction?), connects them, and generates a fill path.
As mentioned above, the input paths need to be defined in advance (forward references are unsupported).
If you would generate the filled path manually, you would draw it before the other ones such that it does
not overlap. This is done implicitly by pgfplots: as soon as pgfplots encounters a fill between
plot, it will activate layered graphics. The filled path will be placed on layer pre main which is between
the main layer and the background layer.
A fill between operation is just like a usual plot: it makes use of the cycle list, i.e. it receives
default plot styles. Our first example above uses the default cycle list which has a brown color. We
can easily redefine the appearance just as for any other plot by adding options in square braces:
\begin{axis}
\addplot[blue,name path=A,domain=0:1] {sqrt(x)};
\addplot[red, name path=B,domain=0:1] {sqrt(x/2)};

0.5
\addplot[gray] fill between[of=A and B];
\end{axis}
\end{tikzpicture}
0
0 0.2 0.4 0.6 0.8 1
Note that the number of data points does not restrict fill between. In particular, you can combine
di↵erent arguments easily.
\begin{axis}
\addplot[blue,name path=A,domain=0:1]
{sin(360*x)};
\addplot[red, name path=B,domain=0:1,samples=2]

0
{0.5};
\addplot[orange] fill between[of=A and B];

\end{axis}
\end{tikzpicture}
1
0 0.2 0.4 0.6 0.8 1
The combination of input plots is also possible if one or both of the plots make use of smooth interpo-
lation:
\begin{axis}
\addplot+[name path=A,samples=7,smooth,domain=0:1]
{sin(360*x)};
\addplot+[name path=B,samples=15,domain=0:1]
0
{cos(360*x)};

\end{axis}
\end{tikzpicture}
1
0 0.2 0.4 0.6 0.8 1
Actually, a fill between path operates directly on the low–level input path segments. As such, it is
much closer to, say, a Tik Z decoration than to a plot; only its use-cases (legends, styles, layering) are
tailored to the use as a plot. However, the input paths can be paths and/or plots. The example below
combines one \addplot and one \path.
\begin{axis}[ymin=-0.2,enlargelimits]
\addplot+[name path=A,smooth]
coordinates {(0,0) (1,1) (2,0)};
0.5 \path[name path=B]
(0.5,-0.2) -- (1.8,-0.2);

0 \end{axis}
\end{tikzpicture}
0 0.5 1 1.5 2
As mentioned above, fill between takes the two input paths as such and combines them to a filled
segment. To this end, it connects the end–points of both paths. This can be seen in the example above:
the path named ‘B’ has di↵erent x coordinates than ‘A’ and results in a trapezoidal output.
Here is another example in which a plot and a normal path are combined using fill between. Note
that the \draw path is generated using nodes of path ‘A’. In such a scenario, we may want to fill only
the second segment which is also possible, see split below.
\begin{axis}
\addplot+[name path=A,samples=15,domain=0:1]
{cos(360*x)}
coordinate[pos=0.25] (nodeA0) {}
coordinate[pos=0.75] (nodeA1) {};
0
\draw[name path=B]
(nodeA0) -- (nodeA1);

1 \end{axis}
\end{tikzpicture}
0 0.2 0.4 0.6 0.8 1
A fill between plot is di↵erent from other plotting operations with respect to the following items:
1. It has no own markers and no nodes near coords. However, its input paths can have both.
2. It supports no pos nodes. However, its input paths can have any annotations as usual.
3. It supports no error bars. Again, its input paths support what pgfplots o↵ers for plots.
4. It cannot be stacked (its input plots can be, of course).
Note that more examples can also be found in Section 4.5.10 on page 103 which covers Area Plots and
has a lot of examples on fill between.
5.7.2 Filling Di↵erent Segments of the Area

/tikz/fill between/split=true|false (initially false)
Activates the generation of more than one output segment.
The initial choice split=false is quite fast and robust, it simply concatenates the input paths and
generates exactly one output segment.
The choice split=true results in a computation of every intersection of the two curves (by means of the
tikz library intersections). Then, each resulting segment results in a separate drawing instruction.
The choice split=false is the default and has been illustrated with various examples above.
The choice split=true is very useful in conjunction with the various styles. For example, we could use
every odd segment to choose a di↵erent color for every odd segment:
\begin{axis}
{sin(360*x)};
\addplot+[name path=B,samples=15,domain=0:1]
0
{cos(360*x)};

split,
every odd segment/.style={yellow},
1 ];
\end{axis}
0 0.2 0.4 0.6 0.8 1 \end{tikzpicture}
Similarly, we could style the regions individually using every segment no:
\begin{axis}
{sin(720*x)};
\path[name path=B]
0
(\pgfkeysvalueof{/pgfplots/xmin},0)
-- (\pgfkeysvalueof{/pgfplots/xmax},0)
;
\addplot fill between[of=A and B,

1 split,
0 0.2 0.4 0.6 0.8 1 {orange},
{pattern=north east lines},
{pattern=north west lines},
{top color=orange, bottom color=blue},
];
\end{axis}
\end{tikzpicture}
The split option allows us to revisit our earlier example in which we wanted to draw only one of the
segments:
\begin{axis}
\addplot+[name path=A,samples=15,domain=0:1]
{cos(360*x)}
coordinate[pos=0.25] (nodeA0)
coordinate[pos=0.75] (nodeA1);
0
\draw[name path=B]
(nodeA0) -- (nodeA1);
\addplot[fill=none] % default: fill none

1 fill between[of=A and B,
split,
0 0.2 0.4 0.6 0.8 1 % draw only selected ones:
% every segment no 0/.style: invisible
every segment no 1/.style={fill,orange},
% every segment no 2/.style: invisible
];
\end{axis}
\end{tikzpicture}
Each segment results in an individual \fill instruction, i.e. each segment is its own, independent, path.
This allows to use all possible Tik Z path operations, including pattern, shade, or decorate.
% requires
% \usetikzlibrary{decorations.markings,shapes.arrows}
\begin{tikzpicture}
2 \begin{axis}
\addplot+[name path=A,domain=0:1,samples=2] {x};
\addplot+[name path=B] table {

x y
0 0 2
0.5 -1
1 3
};
0 0.2 0.4 0.6 0.8 1 \addplot fill between[of=A and B,

split,
every even segment/.style={
postaction={decorate},
decoration={
markings,
mark=between positions 0 and 1
step 1cm with {
\node [single arrow,fill=orange,
single arrow head extend=3pt,
transform shape]
{};
}
},
},
];
\end{axis}
\end{tikzpicture}
5.7.3 Filling only Parts Under a Plot (Clipping)

/tikz/fill between/soft clip=hargumenti fill between/soft clip first=hargumenti
/tikz/fill between/soft clip second=hargumenti
Installs “soft–clips” on both or just one of the involved paths. Soft–clipping means to modify the input
paths such that they respect a clipping region.
In its default configuration, fill between connects the start/end points of the two involved paths.
This is often what you want, but there are use–cases where only parts between the input parts should
be filled: suppose we have f (x) = x2 and we want to fill below the interval [3, 5]. The case “fill below”
means to fill between our function and the x axis:
\begin{tikzpicture}
\begin{axis}
20 \addplot+[name path=A,domain=0:5] {x^2};
\path[name path=B]
10
\end{axis}
\end{tikzpicture}
0
0 2 4
Clearly, we have filled too much. A solution might be to shorten the path B — but that would still
connect the left and right end points of f (x) with the shortened line.
This is where soft clip has its uses: we can select the area of interest by installing a soft clip path:
\begin{tikzpicture}
\begin{axis}
\path[name path=B]
10
];
0 \end{axis}
\end{tikzpicture}
0 2 4
Soft–clipping is similar to clipping. In fact, we could have installed a clip path to achieve the same
e↵ect8 . However, soft–clipping results in a new path which is aware of the boundaries. Consequently,
decorations will be correct, without su↵ering from missing image parts due to the clipping:
\begin{tikzpicture}
\begin{axis}
\path[name path=B]
10
\addplot[
orange,
decorate,decoration={
0 footprints,
foot of=bird,
0 2 4 stride length=15pt,
foot sep=2pt,
foot length=6pt},
]
fill between[of=A and B,
];
\end{axis}
\end{tikzpicture}
The feature soft clip is a part of fill between in the sense that it determines the fill path.
The hargumenti can be one of the following items:
• It can be of the form domain=hxmini:hxmax i. This choice is equivalent to
(hxmini,\pgfkeysvalueof{/pgfplots/ymin}) rectangle
(hxmax i,\pgfkeysvalueof{/pgfplots/ymax}).
• It can be of the form domain y=hymini:hymax i. This choice is equivalent to
(\pgfkeysvalueof{/pgfplots/xmin},hymini) rectangle
(\pgfkeysvalueof{/pgfplots/xmax},hymax i).
• It can be (hx i,hyi) rectangle (hX i,hY i). In this case, it is the rectangle defined by the given
points.
• It can be the name of a named path, i.e. soft clip=A if there exists a path with name path=A.
In this case, the named path has to be “reasonable simple”. In particular, it should be convex,
that is like a rectangle, a circle, or some cloud. It should also be closed.
In any case, the soft clip path should be larger than the paths it applies to. Please avoid infinitely many
intersections points.
8 Installing a clip path might need to adopt layers: fill between is on layer pre main and the clip path would need to be
on the same layer.

\begin{axis}
\draw[help lines,name path=clippath]
(0,-5) -- (-5,0)
-- (2,4) -- (4,-4) --cycle;
0
\addplot+[name path=A] {x};
\draw[red,name path=B]
(0,\pgfkeysvalueof{/pgfplots/ymin})
-- (0,\pgfkeysvalueof{/pgfplots/ymax});
5
6 4 2 0 2 4 6 soft clip={clippath}];
\end{axis}
\end{tikzpicture}
The previous example defines three named paths: the path A is f (x) = x. The named path clippath
serves as clip path, it is some rotated rectangular form. Finally, the path named B is a straight line –
and we fill between A and B with the given clippath.
The choice soft clip first applies the clip path only to the first input path (“A” in our case).
The choice soft clip second applies the clip path only to the second input path (“B” in our case).
Finally, soft clip applies the clip path to both input paths.
\begin{tikzpicture}
20 \begin{axis}
\addplot+[name path=A,domain=0:5] {x^2};
\path[name path=B]
10 (\pgfkeysvalueof{/pgfplots/xmax},0);

soft clip={
0 (3,-1)
rectangle
0 2 4 (4,100)
},
];

soft clip={
(2,-1)
rectangle
(3,100)
},
];
\end{axis}
\end{tikzpicture}
Note that there is also a separate module which allows to apply soft–clipping to individual paths. To
this end, a decoration=soft clip is available. A use–case could be to highlight parts of the input
path:
\begin{tikzpicture}
\begin{axis}
20 \addplot[name path=A,domain=0:5,
postaction={decorate,red,thick},
decoration={
soft clip,
soft clip path={domain=3:5},
10 },
] {x^2};
\path[name path=B]
0 (\pgfkeysvalueof{/pgfplots/xmin},0) --
0 2 4
soft clip={
(3,-1)
rectangle
(5,100)},
];
\end{axis}
\end{tikzpicture}
Note that more examples can also be found in Section 4.5.10 on page 103 which covers Area Plots and
has a lot of examples on fill between.
5.7.4 Styles Around Fill Between

/tikz/fill between/every segment (style, no value)
A style which installed for every segment generated by fill between.
The sequence of styles which are being installed is
every segment, any hoptionsi provided after \addplot[hoptionsi] fill between, then the appropri-
ate every segment no hindex i, then one of every odd segment or every even segment, then every
last segment.
/tikz/fill between/every odd segment (style, no value)

A style which is installed for every odd segment generated by fill between.
Attention: this style makes only sense if split is active.
See every segment for the sequence in which the styles will be invoked.
/tikz/fill between/every even segment (style, no value)

A style which is installed for every even segment generated by fill between.
/tikz/fill between/every segment no hindex i (style, no value)

A style which is installed for every segment with the designated index hindex i generated by fill
between.
An index is a number starting with 0 (which is used for the first segment).
/tikz/fill between/every last segment (style, no value)

A style which installed for every last segment generated by fill between.
\tikzsegmentindex
A command which is valid while the paths of \addplot fill between are generated.
It expands to the segment hindex i, i.e. the same value which is used by every segment no hindex i.
An index is a number starting with 0 (which is used for the first segment).
/pgfplots/every fill between plot (style, no value)

A style which is installed for every \addplot fill between. Its default is
\pgfkeys{
/pgfplots/every fill between plot/.style={
/pgfplots/area legend,/tikz/fill},
}
5.7.5 Key Reference

/tikz/fill between/of=hfirsti and hsecond i
This key is mandatory. It defines which paths should be combined.
The arguments are names which have been assigned to paths or plots in advance using name path.
Paths with these two names are expected in the same tikzpicture.
The fillbetween library supports a variety of input paths, namely
• plots of functions, i.e. each x coordinate has at most one y coordinate,
• plots with interruptions,
• Tik Z paths which meet the same restrictions and are labelled by name path,
• smooth curves or curveto paths,
• mixed smooth / non-smooth parts.
However, it has at most restricted support (or none at all) for paths which
• have self-intersections (i.e. parametric plots might pose a problem),
• have coordinates which are given in a strange input sequence,
• consist of lots of individually separated sub-paths (like mesh or surf plots).
Note that the input paths do not necessarily need to be given in the same sequence, see fill
between/reverse.
/tikz/name path={hnamei}
A Tik Z instruction which assigns a name to a path or plot.
This is mandatory to define input arguments for fill between/of.
/tikz/fill between/reverse=auto|true|false (initially auto)

Configures whether the input paths specified by of need to be reversed in order to arrive at a suitable
path concatenation.
The initial choice auto will handle this automatically. To this end, it applies the following heuristics:
it compares the two first coordinates of each plot: if both plots have their x coordinates in ascending
order, one of them will be reversed (same if both are in descending order). If one is in ascending and
on in descending, they will not be reversed. If the x coordinates of the first two points are equal, the y
coordinates are being compared.
The choice true will always reverse one of the involved paths. This is suitable if both paths have the
same direction (for example, both are specified in increasing x order).
The choice false will not reverse the involved paths. This is suitable if one path has, for example,
coordinates in increasing x order whereas the other path has coordinates in decreasing x order.
Manual reversal is necessary if pgfplots chose the wrong one.
\begin{axis}
\addplot[blue,-stealth,name path=A] {x};
\draw[-stealth,name path=B]
0
(0,-5)
-- (0,5);

reverse=true];
5 \end{axis}
\end{tikzpicture}
6 4 2 0 2 4 6
In this case, we chose reverse=true. This is essentially equivalent to the instruction

• use path A
• append the reversed path B
• fill the result.
In other words, it is equivalent to the following path construction based on intersection segments:
0
(0,-5)
-- (0,5);
\pgfonlayer{pre main}
\fill[gray!50,intersection segments={of=A and B,
5 sequence={A* -- B*[reverse]}}];
\endpgfonlayer
6 4 2 0 2 4 6 \end{axis}
\end{tikzpicture}
Here, we filled the intersection segments of A and B by taking all (indicated by *) segments of A and
all of B in reversed order.
\begin{axis}
0
(0,-5)
-- (0,5);

reverse=false];
5 \end{axis}
\end{tikzpicture}
6 4 2 0 2 4 6
In this case, we chose reverse=false. This, in turn, can be expressed as

• use path A,
• connect with path B,
• fill the result.
In other words, it resembles the follower intersections segments construction:
0
(0,-5)
-- (0,5);
\fill[gray!50,intersection segments={of=A and B,
5 sequence={A* -- B*}}];
\endpgfonlayer
6 4 2 0 2 4 6 \end{axis}
\end{tikzpicture}
/tikz/fill between/on layer={hlayer namei} (initially pre main)

Defines the layer on which \addplot fill between will be drawn. pgfplots defines the layer pre
main to be right before the main layer and pre main is also the initial configuration for any fill
between path.
As soon as you type \addplot fill between, pgfplots will activate layered graphics using set
layers such that this works automatically. pgfplots will also install a clip path the first time it
encounters this layer.
Set hlayer namei to the empty string to place deactivate special layer support for fill between.
Note that this auto-activation of set layers and the installation of a clip path is done for \addplot
fill between, not for the lower–level drawing instructions like \tikzfillbetween or intersection
segments. If you need them, you have to install a layer list manually using either set layers (if inside
of an axis) or \pgfsetlayers. The clip path for an axis can be installed manually using
\pgfplotsextra{
\pgfplotspathaxisoutline
\pgfusepath{clip}
\endpgfonlayer
}
(should this ever be necessary).

/tikz/fill between/inner moveto=connect|keep (initially connect)
Sometimes input paths contain the leading moveto operation and some inner movetos. This key con-
figures how to deal with them.
The initial choice connect replaces them by lineto operations (and connects them)9 .
The choice keep keeps them.
Typically, fill between requires the initial choice connect as it allows to deal with interrupted paths:
\begin{axis}
\addplot[blue,ultra thick,name path=A,

jump mark left] {x};
0
\path[name path=B]
(-5,0)
-- (5,0);
\addplot[top color=red, bottom color=white]

5 fill between[of=A and B];
\end{axis}
9 Note that the actual implementation of inner moveto=connect is more complicated: it also deduplicates multiple adjacent
movetos and it eliminates “empty” moveto operations at the end of a path (i.e. moveto operations which are not followed by
any path operation).
5.7.6 Intersection Segment Recombination

The implementation of fill between relies on path recombination internally: all intersection segments
are computed and concatenated in a suitable order (possibly reversed).
This method can also be applied to plain Tik Z paths to achieve interesting e↵ects:
1 N (0)
N ( 52 )
0.8 fill
lower
0.6
0.4
0.2
0
6 4 2 0 2 4 6 8
\begin{tikzpicture}
\def\verticalbar{2}
\begin{axis}[domain=-5:8,samples=25,smooth,width=12cm,height=7cm]
% Draw curves
\addplot[name path=g0,thin] {exp(-x^2/4)};
\addlegendentry{$\mathcal{N}(0)$}
\addplot[name path=g2.5,thin] {exp(-(x-2.5)^2/4)};
\addlegendentry{$\mathcal{N}(\frac52)$}
% Draw vertical bar:

\draw [name path=red,red,thick]
(\verticalbar,-1) -- (\verticalbar,2);
\path[name path=axis] (-10,0) -- (16,0);
\addplot[black!10] fill between[of=g0 and g2.5,

soft clip={domain=-6:\verticalbar},
split,
every segment no 0/.style={left color=blue,right color=white},
every segment no 1/.style={pattern=grid,pattern color=gray},
];
\addlegendentry{fill}
% compute + label the lower segment (but do not draw it):

\path[name path=lower,
%draw=red,ultra thick,
intersection segments={of=g0 and g2.5,sequence=R1 -- L2}
];
\addplot[pattern=north west lines, pattern color=orange]

fill between[
of=axis and lower,
soft clip={domain=-6:\verticalbar},
];
\addlegendentry{lower}
\end{axis}
\end{tikzpicture}
This example has two plots, one with a Gauss peak at x = 0 and one with a Gauss peak at x = 52 . Both
have standard legend entries. Then we have a red line drawn at x =\verticalbar which is defined as x = 2.
The third plot is a fill between with splitted segments where the left segment has a shading and the right
one has a pattern – and both are clipped to the part which is left of \verticalbar. The option list which
comes directly after \addplot, i.e. the [black!10] will be remembered for the legend entry of this plot. The
next \path... instruction has no visible e↵ect (and does not increase the size of the document10 ). However,
it contains the key intersection segments which computes a path consistent of intersection segments of
the two functions. In our case, we connect the first (1th) segment of the path named g2.5 (which is referred
to as R in the context of sequence) and the second (2nd) segment of the path named g0 (which is referred
to as L in the context of sequence). The result receives name path=lower. Finally, the last \addplot is
a fill between which fills everything between the axis and this lower path segment, again clipped to the
parts left of \verticalbar. Note that axis is no magic name; it has been defined in our example as well.
This is explained in more detail in the following paragraphs.
/tikz/intersection segments={hoptions with prefix /tikz/segmentsi}

Evaluates hoptionsi and appends intersection segments to the current path.
This key is actually more a command: it acquires the two input paths which are argument of the
mandatory of key and it parses the hseries specificationi which is argument of sequence. Afterwards,
it computes the intersection segments and concatenates them according to hseries specificationi. Each
resulting segment is appended to the current path as–is.
The key intersection segments can occur more than once in the current path. Since the key generates
path elements, it is sufficient to terminate the path right after options have been processed (i.e. to add
‘;’ right after the option list).
Note that this method operates on the transformed paths and does not apply coordinate transformations
on its own. It is more like a decoration than a standard path command, although it does not have an
input path.
There is also a related key name intersections in the intersections library of Tik Z which assigns
names to all intersections, see the manual of Tik Z for details.
/tikz/segments/of={hname1 i} and {hname2 i}

Defines the two input paths.
/tikz/segments/sequence={hseries specificationi} (initially L1 -- R2)

Selects the intersection segments and their sequence.
The hseries specificationi consists of a sequence of entries of the form Lhindex i -- Rhindex i -- Lhindex i.
It is probably best shown before we delve into the details:
\begin{tikzpicture}
\begin{axis}[domain=0:10]
4
\addplot[name path=first,blue]
{sin(deg(x)) + 2/5*x};
\addplot[name path=second,red,
2
samples=16,smooth]
{cos(deg(1.2*x)) + 2/5*x};
\draw[black,-stealth,
0 decorate,decoration={
saw,
0 2 4 6 8 10 post=lineto,
post length=10pt},
intersection segments={
of=first and second,
sequence={L1 -- R2 -- R3
-- R4 -- L4[reverse]}
},
];
\end{axis}
\end{tikzpicture}
The preceding example defines two input plots of di↵erent sampling density, one is a sharp plot and
one is smooth. Afterwards, it draws a third path with a saw decoration – and that path concatenates
intersection segments.
10 Well, perhaps 8 byte for scopes.
The entry L1 means to take the first intersection segment of the first input path. In this context, the
first input path is always called ‘L’, regardless of its actual name. Note that L1 refers to an entire
segment (not just a point).
The second item is -- which means to connect the previous segment with the next one. Without this
item, it would have moved to the next one, leaving a gap. Note that -- is normally a connection between
two points. In this context, it is a connection between two segments.
The third item is R2 which means to use the second intersection segment of the second input path.
Again, the second input path is always called ‘B’, regardless of its actual name. The other items are
straightforward until we arrive at L4[reverse]: it is possible to append an reversed path segment this
way.
The general syntax is to add an (arbitrary) sequence of [--] [L|R]{hindex i}[hoptionsi] where
hoptionsi is optional. There is one special case: if hindex i is *, the entire path will be used. If all
encountered indices are *, the intersection will not be computed at all (in this case, intersection
segments degenerates to “path concatenation”). Consequently, the following example is a degenerate
case in which we did “path concatenation” rather than intersection concatenation:
\pgfdeclarelayer{pre main}
\begin{tikzpicture}
\pgfsetlayers{pre main,main}
\draw[-stealth,thick,name path=A,red]
(0,0) -- (1,1);
\draw[-stealth,thick,name path=B,blue]
(0,-1) -- (1,0);
\fill[
orange,
of=A and B,
sequence={L* -- R*[reverse]}}];
\endpgfonlayer
\end{tikzpicture}
Note that segment indices start at 1. They will be processed by means of pgf’s math parser and may
depend on \pgfintersectionsolutions (which is the number of intersections).
Note that there is actually another syntax: you can use Ah0-based-index i for the first input path and
Bh0-based-index i for the second input path. These identifiers were introduced in pgfplots 1.10, but
have been deprecated in favor of 1-based indices. The old syntax with 0-based indices will still remain
available. It is advised to use L and R instead of A and B.
Note that curly braces around {hindex i} can be omitted if the index has just one digit: L1, L2, or L-2
are all valid.
It is also possible to use negative indices to count from the last:
\pgfdeclarelayer{pre main}
\begin{tikzpicture}
\pgfsetlayers{pre main,main}
\draw[-stealth,thick,name path=A,red]
(0,0) -- (1,1);
\draw[-stealth,thick,name path=B,blue]
(0,1) -- (1,0);
\draw[
green,
line width=2pt,
of=A and B,
sequence={L-1}}];
\endpgfonlayer
\end{tikzpicture}
In this case, 1 denotes the first segment in reverse ordering.

Note that \pgfintersectionsegments is the number of segments in this context.
/tikz/segments/reverse=true|false (false)
Allows to reverse a↵ected path segments.
\begin{tikzpicture}
\begin{axis}[xmin=0,xmax=5,ymin=0,ymax=5]
4
\addplot [name path=f,domain=0:5,samples=100]
{(x - 2)^2/(7*x) + 1};
\addplot [name path=border,

color=blue, ultra thick, dashed]
2 coordinates {(2,0) (2,5) } ;
\fill[
of=f and border,
0 sequence={L1 -- R2}},
0 1 2 3 4 5 pattern=north east lines,
]
-- cycle;
\fill[
of=f and border,
sequence={R2[reverse] -- L2}},
pattern=north west lines,
]
-- (rel axis cs:1,1) -- cycle;
\end{axis}
\end{tikzpicture}
The preceding example defines two input plots: the plot named f which is the plot of the curve and the
blue border line.
It then computes fills paths relying on two intersection segments, one which resembles the part
above the curve which is left of the border and one which resembles the part on the right of the border.
It works by concatenating intersection segments in a suitable sequence (add further \draw state-
ments to visualize the individual segments). Note that R2 is used twice: once in normal direction and
once reversed. The intersection segments merely constitute the start of the paths; they are ex-
tended by --cycle and -- (rel axis cs:1,1) -- cycle, respectively. Keep in mind that rel axis
cs is a relative coordinate system in which 1 means 100% of the respective axis – in this case, we have
the upper right corner which is 100% of x and 100% of y.
5.7.7 Basic Level Reference

There are a couple of basic level functions which allow to control fillbetween on a lower level of abstraction.
The author of this package lists them here for highly experienced power-users (only). It might be suitable
to study the source code to get details about these methods and how and where they are used.
\tikzfillbetween[hoptionsi]{hdraw stylei}
This is the low-level interface of fill between; it generates one or more paths.
This command can be used inside of a plain Tik Z picture, it is largely independent of pgfplots:
\begin{tikzpicture}
\draw[name path=first] (0,0) -- (1,1) -- (2,0);
\draw[name path=second] (0,0.5) -- (2,0.5);
\tikzfillbetween[of=first and second,

split,
every even segment/.style={orange}]
{red}
\end{tikzpicture}
The first argument hoptionsi describes how to compute the filled regions like of or split. It corresponds
to those items which are in \addplot fill between[hoptionsi].
The second argument hdraw stylei is the default draw style which is installed for every generated path
segment.
Note that \tikzfillbetween is no typically \path statement: it generates one or more of Tik Z \path
statements (each with their own, individual draw style).
Inside of \tikzfillbetween, the macro \tikzsegmentindex will expand to the current segment index.
It can be used inside of styles.
The key on layer is respected here: if on layer has a valid layer name (pre main by default), the
generated paths will be on that layer. However, unlike \addplot fill between, this command does
not ensure that layered graphics is active. As soon as you write, say, \pgfsetlayers{pre main,main},
it will automatically use these layers. If not, you will see a warning message in your .log file.
\pgfcomputeintersectionsegments{h1 or 2 i}
Given that some intersections have been computed already (and are in the current scope), this command
computes the intersection segments for one of the input arguments.
On output, \pgfretval contains the number of computed segments. The segments as such can be
accessed via \pgfgetintersectionsegmentpath.
The argument h1 or 2 i should be 1 if intersection segments of the first argument of
\pgfintersectionofpaths are to be computed and 2 if the second argument should be used as
input.
This macro is part of fillbetween.
Let us illustrate the e↵ects of some of these methods on the following example.
\begin{tikzpicture}
\draw[red] (0,0) -- (1,1) -- (2,0);
\draw[blue] (0,0.5) -- (2,0.5);
\end{tikzpicture}
We have two lines, both start on the left-hand-side. Our goal is to get a new path consisting of the
intersections segments on the lower part of the picture, i.e. we would like to see
\begin{tikzpicture}
% this is our goal - but computed automatically.
\draw[black] (0,0) -- (0.5,0.5) --
(1.5,0.5) -- (2,0);
\end{tikzpicture}
In order to let fillbetween compute the target path, we assign names to the input paths, compute the
intersections – and recombined them using \pgfcomputeintersectionsegments.
Attention: Before you want to replicate this example, you may want to read about intersection
segments which is a much simpler way to get the same e↵ect!
\begin{tikzpicture}[line join=round,x=3cm,y=3cm]
\draw[name path=first,red] (0,0) -- (1,1) -- (2,0);
\draw[name path=second,blue] (0,0.5) -- (2,0.5);
% from ’name path’ to softpaths...

\tikzgetnamedpath{first}
\let\A=\pgfretval
\tikzgetnamedpath{second}
\let\B=\pgfretval
% compute intersections using the PGF intersection lib...

\pgfintersectionofpaths{\pgfsetpath\A}{\pgfsetpath\B}%
% ... and compute the intersection *segments* for both input

% paths...
\pgfcomputeintersectionsegments1
% ... recombine the intersection segment paths!

\pgfgetintersectionsegmentpath{1}{0}% path 1, segment 0
\pgfsetpathandBB\pgfretval% this starts a new path
% connect, not move. Try to eliminate this line to see the effect
\pgfpathreplacefirstmoveto\pgfretval%
\pgfaddpathandBB\pgfretval% append
\pgfgetintersectionsegmentpath{1}{2}%
\pgfpathreplacefirstmoveto\pgfretval
\pgfaddpathandBB\pgfretval
\pgfsetlinewidth{3}
\pgfsetcolor{black}
\pgfusepath{stroke}
\end{tikzpicture}
Note that this operates on a relatively low level. However, you can easily insert these statements into a
\pgfextra in order to embed it into Tik Z. This allows access to any Tik Z options, including decorate:
\begin{tikzpicture}[line join=round,x=3cm,y=3cm]
\draw[name path=first,red] (0,0) -- (1,1) -- (2,0);
\draw[name path=second,blue] (0,0.5) -- (2,0.5);
\draw[orange,
decorate,decoration={
footprints,
foot of=bird,
stride length=15pt,
foot sep=2pt,
foot length=6pt},
]
\pgfextra
% from ’name path’ to softpaths...
\tikzgetnamedpath{first}
\let\A=\pgfretval
\tikzgetnamedpath{second}
\let\B=\pgfretval
%
% compute intersections using the PGF intersection lib...
\pgfintersectionofpaths{\pgfsetpath\A}{\pgfsetpath\B}%
%
% ... and compute the intersection *segments* for both input
% paths...
%
% ... recombine the intersection segment paths!
\pgfsetpathandBB\pgfretval% this starts a new path
\pgfpathreplacefirstmoveto\pgfretval% connect, not move
\pgfaddpathandBB\pgfretval% append
\pgfgetintersectionsegmentpath{1}{2}%
\pgfpathreplacefirstmoveto\pgfretval
\pgfaddpathandBB\pgfretval
\endpgfextra
;
\end{tikzpicture}
Attention: Before you want to replicate this example, you may want to read about intersection
segments which is a much simpler way to get the same e↵ect!
\pgfgetintersectionsegmentpath{h1 or 2 i}{hindex i}
Defines \pgfretval to contain the desired path segment as softpath.
The result has the same quality as a path returned by \pgfgetpath and can be used by means of
\pgfsetpath, \pgfsetpathandBB, or \pgfaddpathandBB.
The value h1 or 2 i resembles the argument of a preceding call to \pgfcomputeintersectionsegments:
it identifies which of the two paths for which intersections have been computed is to be selected.
The second argument hindex i is a number 0  i < N where N is the total number of computed segments.
The total number of computed segments is returned by \pgfcomputeintersectionsegments.
\tikzgetnamedpath{hstring namei}
Defines \pgfretval to contain the softpath associated with hstring namei. The hstring namei is sup-
posed to be the value of name path or name path global.
The resulting value is a softpath, i.e. it has the same quality as those returned by \pgfgetpath.
\tikznamecurrentpath{hstring namei}
Takes the current softpath (the one assembled by previous moveto, lineto, or whatever operations), and
assigns the name hstring namei to it.
\pgfcomputereversepath{h\softpathmacroi}
Takes a softpath h\softpathmacroi and computes its reversed path.
It stores the resulting softpath into \pgfretval.
\pgfgetpath{h\softpathmacroi}
Stores the current softpath into the macro h\softpathmacroi.
See also \tikzgetnamedpath.
This macro is part of pgf.
\pgfsetpath{h\softpathmacroi}
Replaces the current softpath from the macro h\softpathmacroi.
This does not update any bounding boxes. Note that this takes a soft–path as it is, no transformation
will be applied. The only way to modify the path and its coordinates is a decoration or a canvas
transformation.
\pgfaddpath{h\softpathmacroi}
Appends the softpath from the macro h\softpathmacroi to the current softpath.
This does not update any bounding boxes. Note that this takes a soft–path as it is, no transformation
will be applied. The only way to modify the path and its coordinates is a decoration or a canvas
transformation.
\pgfsetpathandBB{h\softpathmacroi}
Replaces the current softpath from the macro h\softpathmacroi.
This updates the picture’s bounding box by the coordinates found inside of h\softpathmacroi. Aside
from that, the same restrictions as for \pgfsetpath hold here as well.
\pgfaddpathandBB{h\softpathmacroi}
Appends the softpath of macro h\softpathmacroi to the current softpath.
This updates the picture’s bounding box by the coordinates found inside of h\softpathmacroi. Aside
from that, the same restrictions as for \pgfsetpath hold here as well.
\pgfpathreplacefirstmoveto{h\softpathmacroi}
Takes a macro containing a softpath on input, replaces its first moveto operation by a lineto operation
and returns it as \pgfretval.
The argument h\softpathmacroi is one which can be retrieved by \pgfgetpath.
\pgfintersectionofpaths{hfirsti}{hsecond i}
The pgf basic layer command to compute intersections of two softpaths. In contrast to the name path
method provided by Tik Z, this command accepts di↵erent argument: hfirsti and hsecond i are supposed
to set paths, i.e. they should contain something like \pgfsetpath{\somesoftpath}.
Results are stored into variables of the current scope.
\tikzpathintersectionsegments[hoptions with prefix /tikz/segmentsi]

An alias for \tikzset{intersection segments={hoptionsi}}.
\pgfpathcomputesoftclippath{h\inputsoftpathi}{h\softclippathi}
Does the work for soft clip: it computes the soft–clip–path of h\inputsoftpathi when it is clipped
against h\softclippathi.
5.8. GROUPING PLOTS 427
The algorithm has been tested and written for rectangular soft clip paths. It will accept complicated
clip paths, and might succeed with some of them. Nevertheless, rectangular soft clip paths are the ones
which are supported officially.
See soft clip for details.
5.7.8 Pitfalls and Limitations

The fillbetween backend is quite powerful and successfully computes many use–cases around “filling path
segments of arbitrary plots” out of the box.
However, it has a couple of limitations which may or may not be overcome by future releases:
1. The first limitation is scalability. The underlying algorithms are relatively inefficient and scale badly
if the number of samples is large. Please apply it to “reasonable sample sizes” and plots with a
“reasonable number of intersections”. That means: if it takes too long, you may need to reduce the
sampling density.
2. The second limitation is accuracy. The fillbetween functionality relies on the intersections library
of pgf which, in turn, may fail to find all intersections (although its accuracy and reliability has been
improved considerably as part of the work on fillbetween, thanks to Mark Wibrow of the pgf team
for his assistance).
The work-around for this limitation might be to reduce the sampling density – and to file bug reports.
3. Another limitation is generality. The fillbetween library allows to combine smooth and sharp plots,
even const plots with jumps, and all out of the box. It will do so successfully as long as you have
“plot–like” structures. But it may fail if
• plots intersect themselves and you try to compute individual segments using split or
intersection segments,
• the plot has circles,
• the two involved plots have infinitely many intersections (i.e. are on top of each other).
Many of these limitations are present in pgf as well, especially when decorateing paths or when using name
intersections.
5.8 Grouping plots

by Nick Papior Andersen
\usepgfplotslibrary{groupplots} % LATEX and plain TEX

\usepgfplotslibrary[groupplots] % ConTEXt
\usetikzlibrary{pgfplots.groupplots} % LATEX and plain TEX
\usetikzlibrary[pgfplots.groupplots] % ConTEXt
A library which allows the user to typeset several plots in a matrix like structure. Often one has to
compare two plots to one another, or you simply need to display two plots in conjunction with each
other. Either way the following section describes this library which makes matrix structure easier than
alternative methods discussed in Section 4.19.4.
\begin{groupplot}[hoptionsi]
\end{groupplot}
Once you have loaded the groupplots library you will gain access to this environment. This environment
is limited to the same restrictions as the axis environment. It actually utilizes this environment so
consider it as an extension of this. What is important to note is that [hoptionsi] are applied to all
plots in the entire environment. This can be really handy when you need the same xmin, xmax, ymin
and ymax.
With such an environment one can typeset plots in matrix like styles
2 2 2 2
1 1 1 1
0 0 0 0
0 1 2 0 1 2 0 1 2 0 1 2
2 2 2 2
1.5 1 1.5 1
1 0 1 0
0 1 2 0 0.5 1 0 1 2 0 0.5 1
% Preamble: \pgfplotsset{width=7cm,compat=1.13}\usepgfplotslibrary{groupplots}
% Example using groupplots library
\begin{tikzpicture}
\begin{groupplot}[group style={group size=2 by 2},height=3cm,width=3cm]
\nextgroupplot
\addplot coordinates {(0,0) (1,1) (2,2)};
\nextgroupplot
\nextgroupplot
\nextgroupplot
\end{groupplot}
\end{tikzpicture}
% Same example created as done without the library
\begin{tikzpicture}
\begin{axis}[name=plot1,height=3cm,width=3cm]
\end{axis}
\begin{axis}[name=plot2,at={($(plot1.east)+(1cm,0)$)},anchor=west,height=3cm,width=3cm]
\end{axis}
\begin{axis}[name=plot3,at={($(plot1.south)-(0,1cm)$)},anchor=north,height=3cm,width=3cm]
\end{axis}
\begin{axis}[name=plot4,at={($(plot2.south)-(0,1cm)$)},anchor=north,height=3cm,width=3cm]
\end{axis}
\end{tikzpicture}
The equivalent code is seen as the second example and it is clear that you have to type a lot less. So how
do you use it? First of all you need to utilize the new environment groupplot. Within this environment the
following command works.
\nextgroupplot[haxis optionsi] hnormal plot commandsi

This command shifts the placement of the plot. Therefore one should always start the environment
groupplot with the command \nextgroupplot in order to create the first plot. The [haxis optionsi]
are the options that are supplied to the following axes until the next \nextgroupplot command is seen
by TEX. The order in which figures are typeset are as seen in the next example.
1 1
\begin{tikzpicture}[shorten >=4pt,shorten <=4pt]
\begin{groupplot}[group style={group size=2 by 2},
0.5 1. 0.5 2. height=3.5cm,width=3.5cm,/tikz/font=\small]
\nextgroupplot%1
0 0 \addplot coordinates {(0,1) (1,0)};
0 0.5 1 0 0.5 1 \nextgroupplot%2
\nextgroupplot%3
1 1 \addplot coordinates {(0,1) (1,0)};
\nextgroupplot%4
0.5 3. 0.5 4. \end{groupplot}
\draw[thick,>=latex,->,red]
0 0 (group c1r1.center) node {1.} --
0 0.5 1 0 0.5 1 (group c2r1.center) node {2.};
(group c2r1.center) --
(group c1r2.center) node {3.};
(group c1r2.center) --
(group c2r2.center) node {4.};
\end{tikzpicture}
The plot first fills the first row, then the next row and so on. Just like a table, thus the names group
chcolumnirhrow i. The power of the groupplot is to quickly create an aligned structure of plots. But
you can also utilize it to structure data more creatively. Consider the next example.
100
\begin{tikzpicture}
\begin{groupplot}[group style={group size=2 by 2,
80 horizontal sep=0pt,vertical sep=0pt,
xticklabels at=edge bottom},
xmin=0,ymin=0,
60 height=3.7cm,width=4cm,no markers]
\nextgroupplot[group/empty plot]
40 \nextgroupplot[xmin=5,xmax=10,ymin=50,ymax=100]
\addplot[very thick] file {plotdata/group-1.dat};
\nextgroupplot[xmax=5,ymax=50]
20 \addplot[very thick] file {plotdata/group-1.dat};
\nextgroupplot[xmin=5,xmax=10,ymax=50,
0 yticklabels={}]
0 2 4 6 8 10 \addplot[very thick] file {plotdata/group-1.dat};
\end{groupplot}
\end{tikzpicture}
Or for instance zooming in on data as in the next example.
30 25
80
60 20 20
40
10 15
20
0 0 10
0 5 10 0 2 4 3 3.5 4 4.5 5
\begin{tikzpicture}
\begin{groupplot}[group style={group size=3 by 1},xmin=0,ymin=0,height=4cm,width=5cm,no markers]
\nextgroupplot
\draw[red,dashed,thick] (0,0) rectangle (5,30);
\nextgroupplot[xmax=5,ymax=30]
\draw[red,dashed,thick] (3,10) rectangle (5,25);
\nextgroupplot[xmin=3,xmax=5,ymin=10,ymax=25]
\end{groupplot}
\draw[thick,blue,->,shorten >=2pt,shorten <=2pt]
(group c1r1.east) -- (group c2r1.west);
\draw[thick,blue,->,shorten >=2pt,shorten <=2pt]
(group c2r1.east) -- (group c3r1.west);
\end{tikzpicture}
5.8.1 Grouping options

/pgfplots/group style={hoptions with group/ prefix i}
This key sets all hoptionsi using the /pgfplots/group/ prefix.
Note that the distinction between group/ and normal options is important as some of them are quite
similar.
For example, the following statements are all equivalent:
\pgfplotsset{group style={a=2,b=3}}
\pgfplotsset{group/a=2,group/b=3}
\pgfplotsset{group/.cd,a=2,b=3}
All the following keys are in the subdirectory group.
/pgfplots/group/group size=hcolumnsi by hrowsi (initially 1 by 1)

/pgfplots/group/columns=hcolumnsi (initially 1)
/pgfplots/group/rows=hrowsi (initially 1)
These keys determine the total number of plots that can be in one environment groupplot. It is thus
important not to add more \nextgroupplot in the environment than hcolumnsi⇥hrowsi. This is critical
to set if one uses more than 1 more plot. As the key group size uses columns and rows you should
stick to either group size or both columns and rows.
/pgfplots/group/horizontal sep=hdimensioni (initially 1cm)

/pgfplots/group/vertical sep=hdimensioni (initially 1cm)
The spacing between the plots in the horizontal and vertical direction, respectively. If you thus want
them to be glued together you should set them both to a length of 0pt.
/pgfplots/group/every plot/.style={hstylei} (initially empty)

This style is used on every plot as the first style. It is thus equivalent as hoptionsi in the groupplot
environment.
/pgfplots/group/xlabels at=all|edge bottom|edge top (initially all)

/pgfplots/group/ylabels at=all|edge left|edge right (initially all)
In order to determine which plots get labels typeset one can use these keys. By default all axes get
typeset normally and thus have both x and y axis labels.
2 2
\begin{tikzpicture}
c / mol/L 1.5 1.5 \begin{groupplot}[
1 1 group style={
group name=my plots,
0.5 0.5 group size=2 by 2,
0 0 xlabels at=edge bottom,
ylabels at=edge left,
0 0.5 1 1.5 2 0 0.5 1 1.5 2 },
footnotesize,
width=4cm,
2 2
height=4cm,
c / mol/L
1.5 1.5 %
xlabel=time $t$ / h,
1 1
ylabel=$c$ / mol/L,
0.5 0.5 ]
\nextgroupplot
0 0 \addplot coordinates{(0,0) (1,2) (2,1)};
0 0.5 1 1.5 2 0 0.5 1 1.5 2 \nextgroupplot
time t / h time t / h \addplot coordinates{(0,0) (1,2) (2,1)};
\nextgroupplot
\addplot coordinates{(0,0) (1,2) (2,1)};
\nextgroupplot
\end{groupplot}
\end{tikzpicture}
In the example above, only the bottom row gets the label defined in the beginning
groupplot-environment on the x axis and only the first column of plots gets labels on the y axis
on their left side. These keys are especially handy when using glued plots.
/pgfplots/group/xticklabels at=all|edge top|edge bottom (initially all)

/pgfplots/group/yticklabels at=all|edge left|edge right (initially all)
In order to determine which plots get tick labels typeset one can use these keys. By default all axes
gets typeset normally and thus have both x and y axis tick labels. If one sets
\pgfplotsset{group/xticklabels at=edge bottom,group/yticklabels at=edge right}
only the bottom row gets tick labels on the x axis and only the last column gets tick labels on the y
axis on their right side. These keys are specially handy when using glued plots.
Keep in mind that this is implies the same ticks for all plots.
/pgfplots/group/x descriptions at=all|edge top|edge bottom (initially all)

/pgfplots/group/y descriptions at=all|edge left|edge right (initially all)
These are simply a short hand for using both xticklabels at and xlabels at simultaneously:
2
\begin{tikzpicture}
c / mol/L
1.5 \begin{groupplot}[
1 group style={
group name=my plots,
0.5 group size=2 by 2,
0 %
x descriptions at=edge bottom,
y descriptions at=edge right,
2 horizontal sep=0.5cm,
vertical sep=0.5cm,
c / mol/L
1.5
},
1 footnotesize,
width=4cm,
0.5 height=4cm,
0 %
0 0.5 1 1.5 2 0 0.5 1 1.5 2 xlabel=time $t$ / h,
ylabel=$c$ / mol/L,
time t / h time t / h ]
\nextgroupplot
\nextgroupplot
\nextgroupplot
\nextgroupplot
\end{groupplot}
\end{tikzpicture}
Here, x descriptions at=edge bottom yields that x descriptions (xlabel and xticklabel) are only
used for the lowest row. Furthermore, y descriptions at=edge right places y descriptions only for
the rightmost column. Consider modifying the horizontal sep and vertical sep for your needs.
As for xticklabels at, usage of this key implies the same ticks for all plots.
This might require compat=1.3 (or newer).
/pgfplots/group/group name={hnamei} (initially group)

This sets what you can refer the plots to after typesetting. Thus you can use their anchors later. See
the following example
north
\begin{groupplot}[group style={
group name=my plots,group size=2 by 2},
1 East 1 width=4cm,height=4cm]
\nextgroupplot
0 0 \nextgroupplot
0 1 2 0 1 2
\nextgroupplot
North west \nextgroupplot
2 2
\end{groupplot}
1 center 1 \draw (my plots c1r1.east)
circle (3pt) node {East};
\draw (my plots c2r1.north)
0 0 circle (3pt) node {north};
\draw (my plots c1r2.center)
0 1 2 0 1 2 circle (3pt) node {center};
\draw (my plots c2r2.north west)
circle (3pt) node {North west};
\end{tikzpicture}
/pgfplots/group/empty plot/.style={hstylei} (initially /pgfplots/hide axis)

This key can be used as an option to the command \nextgroupplot. This makes the next plot invisible
(only the axes) but maintains it anchors and name. If you want it to behave in another style then you
can redefine it. Consider the same example as before.
5.9. PATCHPLOTS LIBRARY 433
north
\begin{groupplot}[group style={
group name=my plots,group size=2 by 2},
East 1 width=4cm,height=4cm]
\nextgroupplot[group/empty plot]
\nextgroupplot
0 \addplot coordinates{(0,0) (1,2) (2,1)};
\nextgroupplot
0 1 2
\nextgroupplot
North west \addplot coordinates{(0,0) (1,2) (2,1)};
2 2
\end{groupplot}
\draw (my plots c1r1.east)
1 center 1 circle (3pt) node {East};
\draw (my plots c2r1.north)
circle (3pt) node {north};
0 0 \draw (my plots c1r2.center)
circle (3pt) node {center};
0 1 2 0 1 2 \draw (my plots c2r2.north west)
circle (3pt) node {North west};
\end{tikzpicture}
Notice that you need to call a \nextgroupplot againwards to jump to the next plot.
5.9 Patchplots Library

\usepgfplotslibrary{patchplots} % LATEX and plain TEX
\usepgfplotslibrary[patchplots] % ConTEXt
\usetikzlibrary{pgfplots.patchplots} % LATEX and plain TEX
\usetikzlibrary[pgfplots.patchplots] % ConTEXt
A library for advanced patch plots. Its strength is the creation of patches with smooth boundaries and
smoothly shaded colors.
A patch plot is a plot in which each individual patch is available. Here, “available” means that the user
provided each individual patch manually. This can be achieved by means of a long series of patches which
have been concatenated in a suitable way (compare the description of patch plots in section 4.6.13)
or by means of a mathematical expression which is sampled (compare the key patch type sampling).
Most patch types expect a series of point evaluations in a specific sequence.
Note that even though each individual patch might have a smooth boundary, the patchplots library
does not interpolate smoothly between adjacent patches. Consequently, it is task of the one who creates
the patches (which means: evaluated some function at its vertices) to ensure that patches can be
glued together in an adequate way. This allows a lot of freedom, including both jumps and smoothly
concatenated edges.
The patchplots library comes with a couple of inherently two–dimensional patch types (including
second order triangles/rectangular patches and cubic tensor product patches known for finite elements).
Typically, these patches live in a three–dimensional axis. Often, they are used to visualize the surface
of function values f (x, y). The patchplots library ensures that such patches are drawn in a way which
respects the current view. In particular, if a patch folds over itsself (which is possible), it is drawn such
that foreground areas are in the foreground and background areas are in the background.
The patchplots library comes with smoothly shaded patches. More precisely, both the boundary of
patches and their color shading are smooth. Note, however, that the patch boundary typically has much
more smoothness than the color shading.
The patchplots library also allows automatic conversion from a higher–order patch to triangles (tri-
angulation) by means of the key patch to triangles. Furthermore, it features automatic patch
refines.
Use the patchplots library if you want to have smooth boundaries for your patches, or if you need
advanced shadings, or if you want polygon plots, or if you want more freedom in one–dimensional
patches.
5.9.1 Additional Patch Types

/pgfplots/patch type=default|rectangle|triangle|line|quadratic spline|cubic spline|
bilinear|triangle quadr|biquadratic|bicubic|polygon|coons|tensor bezier (initially
default)
The patchplots library supports several new patch types in addition to the initially available
choices (which are rectangle,triangle and line). The documentation of the two–dimensional
choices from page 174 is repeated here.
The new patch types are discussed in detail on the following pages.
One–Dimensional Patch Types
There are two new one–dimensional patch types, namely quadratic spline and cubic spline. Here,
patch type=quadratic spline consists of quadratic patches of n = 3 vertices each. The vertices are
interpolated exactly:
% Preamble: \pgfplotsset{width=7cm,compat=1.13}\usepgfplotslibrary{patchplots}
patch type=quadratic spline
\begin{tikzpicture}
(5) \begin{axis}[nodes near coords={(\coordindex)},
2 title={\texttt{patch type=quadratic spline}}]
\addplot[
mark=*,
patch,
patch type=quadratic spline]
(1)(3) (4) coordinates {
1 % left, right, middle-> first segment
(0,0) (1,1) (0.5,0.5^2)
% left, right, middle-> second segment
(2) (1.2,1) (2.2,1) (1.7,2)
(0) };
0 \end{axis}
\end{tikzpicture}
0 0.5 1 1.5 2
In our example, the first segment interpolates f (x) = x2 at the points {0, 1/2, 1}. The quadratic spline
is actually nothing but piecewise Lagrangian interpolation with quadratic polynomials: it expects three
points in the sequence ‘(left end), (right end), (middle)’ and interpolates these three points with a
quadratic polynomial. Unlike the default 1d mesh visualization (which uses patch type=line implic-
itly), you have to use the special syntax above (or the equivalent approach by means of patch table).
Note that patch type=quadratic spline results in correct shapes, but uses just constant color for
each segment; high–order color shading is only supported approximately using patch refines.
The patch type=cubic spline is very similar: it expects patches of n = 4 vertices and interpolates
them with a cubic polynomial:
patch type=cubic spline
\begin{tikzpicture}
1 title={\texttt{patch type=cubic spline}}]
\addplot[
mark=*,
patch,
patch type=cubic spline]
(3) coordinates {
0 % left, right, left middle, right middle
(2) (-1,-1)
(1,1)
(-1/3,{(-1/3)^3})
(1/3,{(1/3)^3})
1 };
(0) \end{axis}
1 0.5 0 0.5 1 \end{tikzpicture}
Here, we interpolated f (x) = x3 at the four equidistant points { 1, 1/3, 1/3, 1} with a cubic polynomial
(which is x3 ). The cubic spline expects a sequence of patches, each with four coordinates, given in
the sequence ‘(left end), (right end), (interpolation point at 1/3), (interpolation point at 2/3)’. It has
limitations and features like quadratic spline, see above.
Providing Patches by means of Mathematical Expressions
Most patch types expect a specific number of vertices in a specific sequence. This is part of what the
patchplots library is. But is is still tedious to provide this sort of data.
For simple patch types like line,rectangle and bilinear, you can provide the input coordinates
with any of the input methods which are available for all other plot handlers. In particular, line is
just a sharp plot (with individually colored segments) and rectangle is nothing but a surf plot.
Note that both rectangle and bilinear also accept the standard matrix input (with scanlines, see
mesh/ordering and its documentation). In summary: simple patch types accept a simple input format.
/pgfplots/patch type sampling=true|false (initially false)

There are some complicated patch types. In particular, all patch types of higher order (i.e.
quadratic spline, cubic spline, triangle quadr, biquadratic, bicubic) need more points
than just their corners. For such patch types, you need to resort to mesh input=patches. That
means you need to provide extra vertices and their function evaluation values in a specific sequence.
The patch type sampling method allows to simplify the procedure for such complicated patch
types11 : it works together with \addplot expression and evaluates the mathematical expression
at each of the required vertices (in the correct sequence):
\begin{axis}
\addplot[
samples=5,domain=-3:3,
mesh,patch type=cubic spline,
0.5 patch type sampling,
% avoid individual colors per segment:
blue,point meta=none,
]
{exp(-x^2)};
0
% a second plot which shows the
% generated x positions:
2 0 2 \addplot[
mark=*,only marks,scatter,
patch type=cubic spline,
patch type sampling,
point meta={exp(-x^2)},
]
{-0.1};
% a third plot which shows the marks

% without patch type sampling:
\addplot[
mark=*,only marks,scatter,
point meta={exp(-x^2)},
]
{-0.15};
\end{axis}
\end{tikzpicture}
The first plot above is almost a normal plot by expression. The samples and domain key controls
the sampling procedure, and blue,point meta=none defines the global color to use. Note that
the special choice point meta=none simply disables individual colors per mesh segment (which is
the default for mesh plots). However, the patch type sampling key here makes a huge di↵erence:
it tells pgfplots to check the current value of patch type and to sample a coordinate sequence
which is suitable as input for that patch type. We see that the outcome is a partially smooth
function (more about that below).
11 Note that patch type sampling is more or less useless for simple patch types.
The method patch type sampling samples x just as usual. The result is a sequence [x0 , x1 , . . . , xk ].
For each interval [xi , xi+1 ], a patch type is sampled inside of the interval. To this end, the
current patch type is used to generate a standardized vertex pattern in the unit cube. For
patch type=cubic spline, this generates four points 0, 1/3, 2/3, 1. These standardized numbers
are mapped into [xi , xi+1 ]. Then, any mathematical expressions (in our case exp(-x^2)) are eval-
uated at the resulting positions.
The second plot in our example above shows the markers resulting from patch type sampling.
Note that we see 13 markers even though we have said samples=5. These 5 samples are shown
in the third plot. This is because patch type=cubic spline needs 4 points for each patch (i.e. 4
points in each sampled interval).
Note that even though the result in our example above is partially smooth, it is not globally
smooth. In other words: each resulting mesh segment is a polynomial of third order. But: the five
cubic polynomials are determined independently; and they are simple glued together without any
intelligence. In particular, they are unsmooth at the five initial sampling points! This key cannot
apply global smoothing. It is really just a convenient method which simplifies sampling of such
patch types.
The method patch type sampling can also be used for surf plots, i.e. for matrix sampling. It
works in the same way:
\begin{tikzpicture}
\begin{axis}
patch type=bicubic,
1 patch type sampling,
samples=5,domain=-3:3]
{exp(-x^2-y^2)};
0.5
% show the generated grid on top:
\addplot3[
0 2 mark=*,mark size=1pt,only marks,scatter,
0 samples=5,domain=-3:3,
2
0 2 patch type=bicubic,
2 patch type sampling,
point meta={exp(-x^2-y^2)},
]
{1.1};
\end{axis}
\end{tikzpicture}
The example is similar to our one–dimensional example above: it uses the same 1d function as
product. We see that it has 132 samples instead of just 52 , and we see that the geometry is
partially smooth (see above for “partially”). Note, however, that the color interpolation is only
applied once per patch. The following example shows a bilinear patch with unsmooth geometry,
but higher resolution for the color data, on a 13 ⇥ 13 mesh:
\begin{tikzpicture}
\begin{axis}
patch type=bilinear,
1 samples=13,domain=-3:3]
{exp(-x^2-y^2)};
0.5 % show the generated grid on top:

\addplot3[
mark=*,mark size=1pt,only marks,scatter,
0 2 samples=13,domain=-3:3,
0 point meta={exp(-x^2-y^2)},
2
0 2 ]
2 {1.1};
\end{axis}
\end{tikzpicture}
Note that you may want to view the preceding examples in Acrobat Reader. Many free pdf viewers
cannot display these shadings properly.
Global One–Dimensional Curves with Smooth Splines
Typically, pgfplots assumes that you want individually colored patch segments whenever you use one
of the plot handlers mesh, surf, or patch. The individual colors are determined by the current colormap
and the value of point meta (compare section 4.8).
Technically, individually colored path segments are one unit. If you fill them, you fill only one segment.
You cannot fill them against the axis. In particular, you cannot use \closedcycle for individually
colored mesh or patch plots.
The patchplots library comes with one–dimensional patch types like quadratic spline or cubic
spline. It would be useful to draw a global path, that is: one which has a single color such that
\closedcycle works. This is supported if you write point meta=none:
Global path with cubic spline \begin{tikzpicture}
\begin{axis}[
axis lines=middle,
axis on top,
1 enlargelimits,
title={Global path with
\texttt{cubic spline}}]
\addplot[
mark=*,
patch,
1 1 2 patch type=cubic spline,
point meta=none,% allow \closedcycle
blue,
1 fill=blue!60!black,
]
table {
% left, right, left middle, right middle
-1 -1
1 1
-0.333333 -0.037037
0.333333 +0.037037
1 1
2 -0.5
1.333333 1.5
1.666666 1
}
\closedcycle;
\end{axis}
\end{tikzpicture}
The use of point meta=none activates a special processing: the outcome is precisely one path.
Two–Dimensional Patch Types
The patchplots library is especially strong for shader=interp, so this is our main focus in the remaining
documentation here.
Attention: At the time of this writing, many free pdf viewers do not fully support the following
shadings12 . The preferred viewer is Adobe Acrobat Reader.
The choice rectangle expects one or more rectangular patches with n = 4 vertices each. These vertices
are either encoded as a matrix or as individual patches (using mesh input=patches), in the sequence
in which you would connect the vertices:
12 The author of this package has submitted bugfixes to Linux viewers based on xpdf/libpoppler, so the problem will (hopefully)
vanish in future versions.

Rectangle from matrix input \begin{tikzpicture}
\begin{axis}[nodes near coords={(\coordindex)},
title=Rectangle from matrix input]
(3)
(0) \addplot3[surf,shader=interp,samples=2,
patch type=rectangle]
20 {x*y};
\end{axis}
0 (2) \end{tikzpicture}
20 5
4 0
2 0 2 4 (1) 5
Rectangle from patch input \begin{tikzpicture}
title=Rectangle from patch input]
\addplot3[patch,shader=interp,patch
(0) type=rectangle] coordinates {
1 (0,0,1) (1,0,0) (1,1,0) (0,1,0)
(3) };
\end{axis}
0.5 (2) \end{tikzpicture}
1
0
0 (1)
0.2 0.4 0.5
0.6 0.8
1 0
As already documented on page 174, the shader=interp implementation for rectangle uses two tri-
angles and interpolates them linearly. The di↵erences between the two examples above arise due to z
bu↵ering approaches: the matrix input reorders the matrix in linear time, whereas the second example
would sort complete rectangles. In our case, this yields to the di↵erent corner sequence.
The choice bilinear is essentially the same as rectangular with respect to its input formats and
stroke paths, but it uses correct bilinear shading for shader=interp. Moreover, the geometry is also
interpolated bilinearly instead of just two triangles. The two examples from above now become
Bilinear from 2 ⇥ 2 matrix input \begin{tikzpicture}
title=Bilinear from $2\times 2$ matrix input]
(3)
(0) \addplot3[surf,shader=interp,samples=2,
patch type=bilinear]
20 {x*y};
\end{axis}
0 (2) \end{tikzpicture}
20 5
4 0
2 0 2 4 (1) 5
Bilinear from 4–point patch input \begin{tikzpicture}
title=Bilinear from $4$--point patch input]
\addplot3[patch,shader=interp,patch type=bilinear]
(0) coordinates {
1 (0,0,1) (1,0,0) (1,1,0) (0,1,0)
(3) };
\end{axis}
0.5 (2) \end{tikzpicture}
1
0
0 (1)
0.2 0.4 0.5
0.6 0.8
1 0
Use patch type=bilinear if you want to improve the shape of individual patches and the quality of
the color interpolation. In contrast to the simpler patch type=rectangle, it might result in a huger
output document.
The choice triangle expects a sequence of linear triangles, each encoded using n = 3 vertices:
Single Triangle patch \begin{tikzpicture}
\begin{axis}[enlargelimits,
nodes near coords={(\coordindex)},
title=Single Triangle patch]
(0) \addplot3[patch,shader=interp] coordinates {
1 (0,0,1)
(1,0,0)
(1,1,0)
0.5 (2) };
\end{axis}
\end{tikzpicture}
1
0 (1)
0 0.5
0.5 0
1
The choice triangle quadr expects a sequence of isoparametric quadratic triangles, each defined by
n = 6 vertices:
Quadratic Triangle \begin{tikzpicture}
title=Quadratic Triangle]
(4)
6 \addplot[patch,patch type=triangle quadr,
(5) (1) coordinates {
4 (0,0) [1] (5,4) [2] (0,7) [3]
(3) (2,3) [1] (3,6) [2] (-1,4) [3]
};
2 \end{axis}
\end{tikzpicture}
(0)
0
0 2 4
Quadratic Triangle \begin{tikzpicture}
title=Quadratic Triangle]
\addplot3[patch,patch type=triangle quadr,

(0) shader=interp]
1 coordinates {
(2) (0,0,1) (5,4,0) (0,7,0)
(2,3,0) (3,6,0) (-1,4,0)
0.5 (5) (4) };
(3) (1)
\end{axis}
\end{tikzpicture}
0 5
0
2
4 0
Here, the edges have the correct quadratic shape. However, the color interpolation is just bilinear ;
using the color values of the corners and ignoring the rest (consider using patch refines to improve
the color interpolation). For three dimensions, pgfplots checks the depth of corners to determine
foreground/background. For two dimensions, strongly distorted elements may fold over each other in
unexpected ways.
The choice biquadratic expects a sequence of isoparametric biquadratic quadrilaterals each defined by
n = 9 vertices. Their main use is to get “rectangles” with smooth boundaries:
Single Biquadratic Quadrilateral \begin{tikzpicture}
6 title=Single Biquadratic Quadrilateral]
(3) (2)
\addplot[patch,patch type=biquadratic,
(8) shader=interp,point meta=explicit]
4 coordinates {
(7) (5) (0,0) [1] (6,1) [2] (5,5) [3] (-1,5) [4]
(3,1) [1] (6,3) [2] (2,6) [3] (0,3) [4]
(3,3.75) [4]
2 };
(4) (1)
\end{axis}
(0) \end{tikzpicture}
0
0 2 4 6
Single Biquadratic Quadrilateral \begin{tikzpicture}
title=Single Biquadratic Quadrilateral]
\addplot3[patch,patch type=biquadratic,shader=interp]
(0) coordinates {
1 (0,0,1) (6,1,0) (5,5,0) (-1,5,0)
(3,1,0) (6,3,0) (2,6,0) (0,3,0)
(3) (6) (3,3.75,0)
0.5 (7) (8) (2) };
\end{axis}
(5) 6 \end{tikzpicture}
(4)
0 (1) 4
0 2 2
4 6 0
Similar to triangle quadr, the edges have the correct quadratic shape – but the color interpolation
is just bilinear ; using the color values of the corners and ignoring the rest. Again, ensure that the
mesh width is small enough in order to improve the quality of the color interpolation (see also patch
refines).
Note that a function of (x, y) is biquadratic if it is quadratic w.r.t. x if y = const and also quadratic
w.r.t. y if x = const (note that this is not an “if and only if”). For example, f (x, y) = x2 y 2 is
biquadratic. Consequently, we can represent a surface plot of f with just one biquadratic patch – only
the color interpolation is just bilinear. We do so using \addplot table[z expr=hexpressioni]:
\begin{tikzpicture}
\begin{axis}
\addplot3[patch,patch refines=3,
shader=faceted interp,
4 patch type=biquadratic]
2 table[z expr=x^2-y^2]
{
0 x y
2 -2 -2
2 2 -2
4 2 2
2 0 -2 2
1 0 0 -2
1 2 2 2 0
0 2
-2 0
0 0
};
\end{axis}
\end{tikzpicture}
We see that the shape’s boundary is reconstructed exactly using the biquadratic patch. In addi-
tion, patch refines improves the (first order) color interpolation. Details for patch refines are
discussed in Section 5.9.2 and details and limitations regarding superimposed grid lines are discussed in
Section 5.9.4.
Note that biquadratic can easily be combined with patch type sampling in order to sample an
arbitrary surface plot with smooth boundaries.
A patch with type biquadratic and shader=interp has a bounding box which is determined from the
input vertices. Due to the high order of the patch, parts of the patch can be outside of that bounding
box. This holds for all advanced patch types.
The choice bicubic is similar to biquadratic: it allows to defines two–dimensional patches whose
boundary is defined by four cubic polynomials. Consequently, it allows very smooth boundaries – espe-
cially since the viewer constructs these boundaries at every zoom level. A bicubic patch is constructed
from 16 points which are arranged in a 4 ⇥ 4 matrix. Each consecutive 16 points make up a single
bicubic patch. The 17th point starts the next bicubic patch (just as for any other patch type).
Single Bicubic Quadrilateral \begin{tikzpicture}
title=Single Bicubic Quadrilateral]
\addplot3[patch,patch type=bicubic,shader=interp]
coordinates {
(0) (0,0,1) (1,0,0) (2,0,0) (3,0,0)
1 (0,1,0) (1,1,0) (2,1,0) (3,1,0)
(12) (0,2,0) (1,2,0) (2,2,0) (3,2,0)
(13) (0,3,0) (1,3,0) (2,3,0) (3,3,0)
(8) (14)
0.5 (9) (15) };
(4) (10)
(5) (11) \end{axis}
(6) \end{tikzpicture}
(1) (7)
0 (2) 2
0 (3)
1
2
3 0
Just as for biquadratic, the color interpolation of bicubic is (just) bilinear, even though the geometry
is of higher order. The color interpolation uses the point meta values determined at the four corners
of each patch; all other values of point meta are ignored by the shader (although their values are used
to compute point meta min and point meta max).
Two Bicubic Patches \begin{tikzpicture}
\begin{axis}[
title=Two Bicubic Patches]
\addplot3[patch,patch type=bicubic,
coordinates {
1 (0,0,1)[1] (1,0,0)[0] (2,0,0)[0] (3,0,0)[0]
(0,1,0)[0] (1,1,0)[0] (2,1,0)[0] (3,1,0)[0]
(0,2,0)[0] (1,2,0)[0] (2,2,0)[0] (3,2,0)[0]
0.5 (0,3,0)[0] (1,3,0)[0] (2,3,0)[0] (3,3,0)[0]
(3,0,0)[0] (4,0,0)[0] (5,0,0)[0] (6,0,0)[0.7]

0 2 (3,1,0)[0] (4,1,.5)[1](5,1,0)[0] (6,1,0)[0]
0 (3,2,0)[0] (4,2,0)[0] (5,2,0)[0] (6,2,0)[0]
2 (3,3,0)[0] (4,3,0)[0] (5,3,0)[0] (6,3,0)[0.1]
4
6 0 };
\end{axis}
\end{tikzpicture}
The previous example uses two patches of type bicubic. Note that the color data (point meta) has
been provided explicitly – and its values are only used at the corners (the [1] value after the point
(4,1,.5) is ignored). Color interpolation of bicubic patches uses only the color data at the patch’s
corners. The remaining color data values are ignored. Note that if you leave the default (which is point
meta=f(x) instead of point meta=explicit), the second patch will be blue. This is because the four
corner vertices of the second patch define the color shading – and their z value is 0.
Note that bicubic can easily be combined with patch type sampling in order to sample an arbitrary
surface plot with smooth boundaries.
Just as described for biquadratic, a patch with type bicubic and shader=interp can have a bounding
box which is slightly smaller than the region which is actually drawn (because the bounding box is
computed from the input points).
The choice coons expects a sequence of one or more Coons patches, made up of n = 12 points each.
A Coons patch is delimited by four cubic Bézier curves, with the end points attached to each other
– and the n points provide the required control points for these curves in a specific ordering which is
illustrated in the following example:
A Coons Patch
4
(4)
(3) (5)
2
(6)
1 (2) (7)
(0)
0
(8)
(1)
1
(11) (9)
2
(10)
3
1 0 1 2 3 4 5 6 7
\begin{tikzpicture}
width=12cm,
title=A Coons Patch]
\addplot[mark=*,patch,patch type=coons,
coordinates {
(0,0) [0] % first corner
(1,-1) [0] % Bezier control point between (0) and (3)
(4,0.7) [0] % Bezier control point between (0) and (3)
%
(3,2) [1] % second corner
(7,2) [1] % Bezier control point between (3) and (6)
%
(7,1) [2] % third corner
(4.5,-0.5) [2] % Bezier control point between (6) and (9)
%
(5,-2) [3] % fourth corner
(4,-2.5) [3] % Bezier control point between (9) and (0)
(-1,-2) [3] % Bezier control point between (9) and (0)
};
\end{axis}
\end{tikzpicture}
The four cubic Bézier curves are equivalent to curveto paths of pgf, i.e. to a sequence of the form
(hcorner 1 i).. controls(hcontrol point Ai) and (hcontrol point B i) .. (hcorner 2 i). The inter-
polated shading is bilinear. More precisely, a bilinear shading in the unit cube [0, 1]2 is initialised which
is then mapped into the Coons patch such that the corners match. The color interpolation uses only
the color data of the four corners, color values of intermediate control points are ignored for the shading
(although their value will be respected for the upper and lower limit of color data). In contrast to the
finite element patches, a Coons patch is inherently two–dimensional. While you can still use three–
dimensional coordinates, pgfplots will draw the shading as you provide it, without checking for the
depth information (as it does for the other patch types). In other words: depending on the current
view angle, the shading might fold over itself in unexpected ways.
Even for two dimensions, Coons patches may fold over themselves. To determine which part is foreground
and which part is background, the following rule applies: the four corner points (0), (3), (6), (9) are
associated to the unit cube points (u, v) = (0, 0), (0, 1), (1, 1) and (1, 0), respectively. The edge between
corner (3) and (6) (i.e. the one with v = 1) is foreground, the edge between (1) and (9) is background.
Thus, large values of v are drawn on top of small values of v. If v is constant, large values of u are
drawn on top of small values of u. Thus, reordering the patch vertices (choosing a di↵erent first vertex
and/or reversing the sequence) allows to get di↵erent foreground/background configurations13 .
Note that patch type sampling is unavailable for patch type=coons because the control points are
no point evaluation of the same function.
The choice tensor bezier is similar to patch type=coons: it allows to define a bezier patch. However,
it allows more freedom: it has 16 control points instead of the 12 of a coons patch. The four additional
control points are situated in the center of each patch. This patch type generates .pdf shadings of
type 7 (whereas coons patches are shadings of type 6). It has been added for reasons of completeness,
although it has not been tested properly. Please refer to the specification of the .pdf format for de-
tails14 . The choice tensor bezier is actually the same as patch type=bicubic – except that bicubic
automatically respects the view depth (foreground/background) and is given in a di↵erent by means of
function evaluations rather than control points.
Note that patch type sampling is unavailable for patch type=tensor bezier because the control
points are no point evaluation of the same function.
The choice polygon expects polygons with a fixed number of vertices. This patch type requires the
number of vertices as argument:
/pgfplots/vertex count=hcounti
13 Internally, pgfplots employs such mechanisms to map the higher order isoparametric patch types to Coons patches, sorting
according their corner’s depth information.

14 If someone is willing to test it and document it, feel free to email me!
The number of vertices to be used for patch type=polygon. The number can be arbitrary. All
input patches are expected to have this many vertices – but it is acceptable if a patch uses the
same vertex multiple times. This means that patch type=polygon accepts polygons with di↵erent
numbers of vertices, but you need to apply some sort of “manual padding”.
This parameter is (currently) mandatory.
A patch plot with patch type=polygon simply connects the n=vertex count vertices in their order
of appearance and closes the resulting path:
3
\begin{tikzpicture}
2 \begin{axis}[view/h=120,xlabel=$x$,ylabel=$y$]
4
\addplot3[
opacity=0.5,
table/row sep=\\,
7
5 patch,
2 patch type=polygon,
vertex count=5,
0 patch table with point meta={%
6 % pt1 pt2 pt3 pt4 pt5 cdata
0 0 0 1 7 2 2 0\\
1 6 5 5 5 1\\
0 1 1 1 5 4 2 7 2\\
1 x 2 4 3 3 3 3\\
y 2 2 }]
table {
x y z\\
0 2 0\\%0
2 2 0\\%1
0 1 3\\%2
0 0 3\\%3
1 0 3\\%4
2 0 2\\%5
2 0 0\\%6
1 1 2\\%7
};
% replicate the vertex list to show \coordindex:
\addplot3[only marks,nodes near coords=\coordindex]
table[row sep=\\] {
0 2 0\\ 2 2 0\\ 0 1 3\\ 0 0 3\\
1 0 3\\ 2 0 2\\ 2 0 0\\ 1 1 2\\
};
\end{axis}
\end{tikzpicture}
The example above defines the patch by means of a connectivity table (patch table with point
meta) and a vertex list (the normal input coordinates of the plot): there are 8 vertices and 4 polygons.
Note that 2 of these polygons are triangles, one has 4 corners and only of them actually has all 5
allocated corners. This e↵ect can be achieved by replicating one of the corners. The connectivity table
in our example defines a unique color for each polygon: 0 for the first patch, 1 for the second, 2 for the
third, and 3 for the last. These numbers map into the current colormap.
The patch type=polygon supports neither triangulation nor shading nor refinement. The order of
appearance of the input points is supposed to be the order in which the line–to operations of the
resulting path are generated.
5.9.2 Automatic Patch Refinement and Triangulation

pgfplots supports automatic patch refinement for most of its patch types. There are mainly two
purposes for patch refinement: to increase the quality of z buffer=sort and/or to improve color
interpolation for high–order patches.
/pgfplots/patch refines={hlevelsi} (initially 0)

This key controls patch refinement. The initial choice patch refines=0 disables refinement and
visualizes elements as they have been found in input files.
A positive hlevelsi enables (recursive) patch refinement: each patch is refined individually.
The following example illustrates the patch refines feature for a triangle quadr shape function
on an edge. Note that since pgfplots uses only first order shading which is based on the corner
points (0), (1) and (2), the specified shape function of patch refines=0 has constant color. Higher
hlevelsi approximate the patch with increasing quality:
patch refines=0 patch refines=1 patch refines=2

(4) (4) (4)
1 1 1
(2) (2) (2)
(5) (5) (5)
0.5 (3) 0.5 (3) 0.5 (3)
(0) (1) (0) (1) (0) (1)
0 5 0 5 0 5
0 2 0 2 0 2
4 0 4 0 4 0
\foreach \level in {0,1,2} {%
\begin{tikzpicture}
\begin{axis}[
footnotesize,
title={patch refines=\level}]
\addplot3[patch,patch type=triangle quadr,

shader=faceted interp,patch refines=\level]
coordinates {
(0,0,0) (5,4,0) (0,7,0)
(2,3,0) (3,6,1) (-1,4,0)
};
\end{axis}
\end{tikzpicture}
}
In this example, patch refinement makes a huge di↵erence since it is just one element with huge
displacements. For practical examples, you probably won’t need many refinement levels.
The refined patches reproduce the geometry’s shape exactly. In addition, they improve color inter-
polation. Note that its purpose is just visualization, therefor hanging nodes are allowed (and will
be generated by patch refine for most patch types).
Patch refinement is implemented for all supported patches except for patch type=coons, tensor
bezier, bicubic (might follow eventually) and polygon.
/pgfplots/patch to triangles=true|false (initially false)

Occasionally, one has a complicated patch type on input and would like to visualize it as a
triangle mesh. pgfplots supports automatic triangulation of patches. Triangulation means
to replace each individual input patch by one or more triangles. Triangulation can be combined
with patch refines in which case patch refines is applied first and the resulting refined patches
are then triangulated.
Triangulation + 0 refines Triangulation + 1 refines Triangulation + 2 refines

(8) (8) (8)
1 1 1
(3) (6) (3) (6) (3) (6)
0.5 (7) (2) 0.5 (7) (2) 0.5 (7) (2)
(0) (4) (5) (0) (4) (5) (0) (4) (5)
0 (1) 6 0 (1) 6 0 (1) 6
4 4 4
0 2 2 0 2 2 0 2 2
4 6 0 4 6 0 4 6 0
\begin{tikzpicture}
\begin{axis}[
footnotesize,
title={Triangulation + \level\ refines}]
\addplot3[patch,patch type=biquadratic,shader=faceted interp,

patch to triangles,patch refines=\level]
coordinates {
(0,0,0) (6,1,0) (5,5,0) (-1,5,0)
(3,1,0) (6,3,0) (2,6,0) (0,3,0)
(3,3.75,1)
};
\end{axis}
\end{tikzpicture}%
}
For one–dimensional patch types like quadratic spline, patch to triangles results in approx-
imation by means of patch type=line instead of triangle.
The patch to triangles feature is implemented for all supported patches except for patch
type=coons, tensor bezier, and polygon.
5.9.3 Peculiarities of Flat Shading and High Order Patches

The patchplots library has been optimized for use with interpolated shadings, i.e. for shader=interp:
it allows the filled area to fold over itself or to be outside of the patch boundaries.
pgfplots also supports shader=flat and shader=faceted by simply stroking and/or filling the patch
boundaries. Naturally, such an approach works only if the enclosed patch boundary and the filled area
are essentially the same! Consider using shader=flat or shader=faceted only if the mesh width is
small enough such that patches do not fold over themselves.
The following example illustrates the e↵ect: the coarse single element on the left folds over itsself,
resulting in strange fill patterns. Refining the mesh reduces the e↵ect.
Faceted + 0 refines Faceted + 1 refines Faceted + 2 refines
1 1 1
0.5 0.5 0.5
0 6 0 6 0 6
4 4 4
0 2 2 0 2 2 0 2 2
4 6 0 4 6 0 4 6 0
\begin{tikzpicture}
\begin{axis}[
footnotesize,
title={Faceted + \level\ refines}]
\addplot3[patch,patch type=biquadratic,shader=faceted,
patch refines=\level]
coordinates {
(0,0,1) (6,1,0) (5,5,0) (-1,5,0)
(3,1,0) (6,3,0) (2,6,0) (0,3,0)
(3,3.75,0)
};
\end{axis}
\end{tikzpicture}
}
5.9.4 Drawing Grids

The patchplots library supports grid (mesh) visualization in the same way as for two/three–dimensional
mesh- and surf plots. This includes four di↵erent approaches: the first is shader=faceted, which uses
constant fill color and faceted color for stroke paths (as we already saw in Section 5.9.3). The second
approach is to use shader=faceted interp which uses interpolated shadings for filling and issues stroke
paths on top of each interpolated element. The third approach is to issue two \addplot commands,
one with the filled patch plot, and one with a patch,mesh style which only draws (colored) grid lines
on top of the previous plot. The three approaches are shown below.
Grids with shader=faceted \begin{tikzpicture}
\begin{axis}[
title={Grids with shader=faceted}]
\addplot3[patch,patch type=biquadratic,
shader=faceted,patch refines=3]
coordinates {
(0,0,1) (6,1,1.6) (5,5,1.3) (-1,5,0)
1 (3,1,0) (6,3,0.4) (2,6,1.1) (0,3,0.9)
(3,3.75,0.5)
};
6 \end{axis}
0 2 2
4 6 0
As already discussed in Section 5.9.3, the approach with shader=faceted works well if the mesh width is
small enough (such that single patches do not overlap and their fill area is within the patch boundaries).
Grids with shader=faceted interp \begin{tikzpicture}
\begin{axis}[
title={Grids with shader=faceted interp}]
\addplot3[patch,patch type=biquadratic,
shader=faceted interp,patch refines=3]
coordinates {
(0,0,1) (6,1,1.6) (5,5,1.3) (-1,5,0)
1 (3,1,0) (6,3,0.4) (2,6,1.1) (0,3,0.9)
(3,3.75,0.5)
};
6 \end{axis}
0 2 2
4 6 0
Here, grid lines are defined to be the patch boundary, so it may occasionally happen for coarse patches
that grid lines cross the filled area. If you experience problems, consider using the patch refines key.
The shader=faceted interp supports z buffer – at the cost of generating one shading for each patch
element (the stroke path is drawn immediately after the patch element is shaded). This can become
quite expensive15 at display time and may lead to huge pdf files. However, shader=faceted interp
provides smooth shadings and, at the same time, good grid lines which are drawn in the correct order.
15 I would really like to hear any well–founded ideas how to improve this issue. In case you have an idea– let me know!
Mesh on top of patches (i): obscured \begin{tikzpicture}
\begin{axis}[
title={Mesh on top of patches (i): obscured}]
\addplot3[patch,patch type=biquadratic,shader=interp,
patch refines=3]
coordinates {
(0,0,1) (6,1,1.6) (5,5,1.3) (-1,5,0)
1 (3,1,0) (6,3,0.4) (2,6,1.1) (0,3,0.9)
(3,3.75,0.5)
};
6 \addplot3[patch,patch type=biquadratic,mesh,black,
0 4
patch refines=3]
coordinates {
0 2 2 (0,0,1) (6,1,1.6) (5,5,1.3) (-1,5,0)
4 6 0 (3,1,0) (6,3,0.4) (2,6,1.1) (0,3,0.9)
(3,3.75,0.5)
};
\end{axis}
\end{tikzpicture}
Mesh on top of patches (ii): unobscured \begin{tikzpicture}
Geometry provided by Prof. Chernov, Bonn \begin{axis}[
title={Mesh on top of patches (ii): unobscured\\
\tiny Geometry provided by Prof. Chernov, Bonn},
title style={align=center},
view={156}{28}]
1 \addplot3[patch,patch type=bilinear,
shader=interp,
patch table=plotdata/patchexample_conn.dat]
file {plotdata/patchexample_verts.dat};
0 \addplot3[patch,patch type=bilinear,
mesh,black,
1 patch table=plotdata/patchexample_conn.dat]
file {plotdata/patchexample_verts.dat};
0 1
0.5 \end{axis}
1 1
The approach to draw grids separately is done by means of two \addplot statements; the first using
patch as before, the second using patch,mesh. This configures pgfplots to visualize just the mesh.
Make sure you provide ‘mesh’ after ‘patch’ since the latter activates filled surf visualization. The
approach of meshes on top of patches implies to draw grid lines simply over any previous drawing
operations. Thus, depth information is lost (as displayed in the first example above). Overlaying grid
lines on top of the surface works in special cases (see bottom picture). An approach which always works
is to provide the mesh at a fixed z position as displayed in the following example:
5.10. POLAR AXES 449
Separate Grids (iii) \begin{tikzpicture}
\begin{axis}[
title={Separate Grids (iii)}]
\addplot3[patch,patch type=biquadratic,shader=interp,
patch refines=3]
coordinates {
(0,0,1) (6,1,1.6) (5,5,1.3) (-1,5,0)
(3,1,0) (6,3,0.4) (2,6,1.1) (0,3,0.9)
1 (3,3.75,0.5)
};
6 \addplot3[patch,patch type=biquadratic,
0 4
mesh,black,
z filter/.code={\def\pgfmathresult{1.8}},
0 2 2 patch refines=3]
4 6 0 coordinates {
(0,0,1) (6,1,1.6) (5,5,1.3) (-1,5,0)
(3,1,0) (6,3,0.4) (2,6,1.1) (0,3,0.9)
(3,3.75,0.5)
};
\end{axis}
\end{tikzpicture}
Here, the first \addplot3 command is the same as above, just with shader=interp. The second
reproduces the same geometry, but uses a z filter to fix the z coordinate (in this case to z = 1.8).
This e↵ectively overrules all z coordinates.
Thus, grid lines can be drawn either by means of flat fill color with shader=faceted (efficient), by
means of interpolated fill colors with shader=faceted interp (inefficient, see above) or, for special
applications, using a separate patch,mesh plot which is drawn on top of the patches (efficient). In any
case, the mesh visualization considers the faceted color which can depend on mapped color.
5.10 Polar Axes

\usepgfplotslibrary{polar} % LATEX and plain TEX
\usepgfplotslibrary[polar] % ConTEXt
\usetikzlibrary{pgfplots.polar} % LATEX and plain TEX
\usetikzlibrary[pgfplots.polar] % ConTEXt
A library to draw polar axes and plot types relying on polar coordinates, represented by angle (in
degrees or, optionally, in radians) and radius.
5.10.1 Polar Axes

\begin{polaraxis}
\end{polaraxis}
The polar library provides the polaraxis environment. Inside of such an environment, all coordinates
are expected to be given in polar representation of the form (hanglei, hradiusi), i.e. the x coordinate is
always the angle and the y coordinate the radius:
% Preamble: \pgfplotsset{width=7cm,compat=1.13}\usepgfplotslibrary{polar}
90
\begin{polaraxis}
\addplot coordinates {(0,1) (90,1)
(180,1) (270,1)};
150 30 \end{polaraxis}
\end{tikzpicture}
0 0.5 1
180 0
210 330
240 300
270
90
\begin{polaraxis}
\addplot+[domain=0:3] (360*x,x); % (angle,radius)
\end{polaraxis}
0 1 2 3
180 0
210 330
240 300
270
90
\begin{polaraxis}
\addplot+[mark=none,domain=0:720,samples=600]
{sin(2*x)*cos(2*x)};
150 30 % equivalent to (x,{sin(..)cos(..)}), i.e.
% the expression is the RADIUS
\end{polaraxis}
\end{tikzpicture}
0 0.2 0.4
180 0
210 330
240 300
270
Polar axes support most of the pgfplots user interface, i.e. legend entries, any axis descriptions,
xtick/ytick and so on:
A polar axis \begin{tikzpicture}
90 \begin{polaraxis}[
First xtick={0,90,180,270},
Second title=A polar axis]

\addlegendentry{First}
\addplot coordinates {(180,0.5) (0,0)};

0 0.5 1 \addlegendentry{Second}
\end{tikzpicture}
270
Furthermore, you can use all of the supported input coordinate methods (like \addplot coordinates,
\addplot table, \addplot expression). The only di↵erence is that polar axes interpret the (first two)
input coordinates as polar coordinates of the form (hangle in degreesi, hradiusi).
It is also possible to provide \addplot3; in this case, the third coordinate will be ignored (although it
can be used as color data using point meta=z). An example can be found below in Section 5.10.3.
5.10.2 Using Radians instead of Degrees

The initial configuration uses degrees for the angle (x component of every input coordinate). pgfplots also
supports to provide the angle in radians using the data cs=polarrad switch:
Degrees and/or Radians \begin{tikzpicture}
90 \begin{polaraxis}[title={Degrees and/or Radians}]
120 Deg \addplot
60
Rad coordinates {(0,1) (90,1) (180,1) (270,1)};
\addlegendentry{Deg}
150 30 \addplot+[data cs=polarrad]
coordinates {(0,1.5) (pi/2,1.5)
(pi,1.5) (pi*3/2,1.5)};
\addlegendentry{Rad}
0 0.5 1 1.5
\end{tikzpicture}
210 330
240 300
270
The data cs key is described in all detail on page 371; it tells pgfplots the coordinate system of input
data. pgfplots will then take steps to automatically transform each coordinate into the required coordinate
system (in our case, this is data cs=polar).
5.10.3 Mixing With Cartesian Coordinates

Similarly to the procedure described above, you can also provide Cartesian coordinates inside of a polar
axis: simply tell pgfplots that it should automatically transform them to polar representation by means
of data cs=cart:
Cartesian Input \begin{tikzpicture}
90 \begin{polaraxis}[title=Cartesian Input]
120 60 \addplot+[data cs=cart]
coordinates {(1,0) (0,1) (-1,0) (0,-1)};
\end{polaraxis}
0 0.5 1
180 0
210 330
240 300
270
More details about the data cs key can be found on page 371.
This does also allow more involved visualization techniques which may operate on Cartesian coordinates.
The following example uses \addplot3 to sample a function f : R2 ! R, computes contour lines (with the
help of gnuplot) and displays the result in a polaraxis:
90
\begin{polaraxis}
0 .2
\addplot3[contour gnuplot,domain=-3:3,
data cs=cart]
2
150 30
0.
{exp(-x^2-y^2)};
0.
0 .6
4
\end{polaraxis}
\end{tikzpicture}
0.
8
0 0.5 1
0.2
180 0
0.4
0 .2
0 .4
0 .6
0 .6
0.8
210 330
2
0.4
0.
0 .2
240 300
270
What happens is that z = exp( x2 y 2 ) is sampled for x, y 2 [ 3, 3], then contour lines are computed on
(x, y, z), then the resulting triples (x, y, z) are transformed to polar coordinates (↵, r, z) (leaving z intact).
Finally, the z coordinate is used as point meta to determine the color.
Note that \addplot3 allows to process three–dimensional input types, but the result will always be
two–dimensional (the z coordinate is ignored for point placement in polaraxis). However, the z coordinate
can be used to determine point colors (using point meta=z).
5.10.4 Polar Descriptions

Polar axes support axis labels just as other plot types.16 In this case, xticklabel cs refers to a label
position for the angles (x axis) and yticklabel cs refers to a label position for the radius (y axis).
The default axis label for the angles is placed at 45 degrees,
16 Attention: support for polar axis labels is broken in versions before 1.13. Please upgrade to pgfplots 1.13 and use
compat=1.13 or higher in order to use polar axis descriptions.

x
120 60
ax
\begin{polaraxis}[
i s
xlabel=x axis,
ylabel=y axis,
150 30 ]
y axis (180,1) (270,1)};
\end{polaraxis}
180 0
210 330
240 300
270
and is configured as a sloped label using the default configuration

every axis x label/.style={
at={(xticklabel cs:0.125)},
sloped={at position=45},
anchor=near ticklabel,
near ticklabel at=45,
},
every axis y label/.style={
at={(yticklabel cs:0.5)},
anchor=near ticklabel,
},
This also allows sloped x labels as in the following example.17

90
120
x
ax
\begin{polaraxis}[
is
xlabel=x axis,
ylabel=y axis,
150
30
xticklabel style={sloped like x axis},

]
y axis \addplot coordinates {(0,1) (90,1)
(180,1) (270,1)};
0 0.5 1
180
\end{polaraxis}
0
\end{tikzpicture}
210
330
240
300
270
5.10.5 Special Polar Plot Types

/tikz/polar comb (no value)
\addplot+[polar comb]
The polar comb plot handler is provided by Tik Z; it draws paths from the origin to the designated
position and places marks at the positions (similar to the comb plot handler). Since the paths always
start at the origin, it is particularly suited for polaraxis:
17 Note that sloped polar labels also requires pgfplots 1.13 and compat=1.13 or newer.
90
\begin{polaraxis}
\addplot+[polar comb]
coordinates {(300,1) (20,0.3) (40,0.5)
150 30 (120,1) (200,0.4)};
\end{polaraxis}
\end{tikzpicture}
0 0.5 1
180 0
210 330
240 300
270
5.10.6 Partial Polar Axes

The polar library also supports partial axes. If you provide xmin/xmax, you can restrict the angles used for
the axis:
90
\begin{polaraxis}[xmin=45,xmax=360]
1 \addplot coordinates {(0,1) (90,1) (180,1) (270,1)};
\end{polaraxis}
0.5
0
180 360
210 330
240 300
270
Currently, the first angle must be lower than the second one. But you can employ the periodicity to get
pies as follows:
90
1
135 45
0.5
180 0 0 360
0.5
225 315
1
270 270
5.11. SMITH CHARTS 455
\begin{tikzpicture}
\addplot coordinates {(0,1) (90,1) (180,1) (270,1)};
\end{polaraxis}
\end{tikzpicture}~%
\begin{tikzpicture}
\addplot coordinates {(0,1) (90,1) (180,1) (270,1)};
\end{polaraxis}
\end{tikzpicture}
Similarly, an explicitly provided value for ymin allows to reduce the displayed range away from 0:
90
\begin{polaraxis}[ymin=0.3]
(180,1) (270,1)};
\end{tikzpicture}
0.4 0.6 0.8 1

180 0
210 330
240 300
270
Modifying xmin and xmax manually can also be used to move the y axis line (the line with ytick and
yticklabels):
90
1 \addplot coordinates {(0,1) (90,1) (180,1) (270,1)};
\end{polaraxis}
0.5
0
180 360
210 330
240 300
270
5.11 Smith Charts

\usepgfplotslibrary{smithchart} % LATEX and plain TEX
\usepgfplotslibrary[smithchart] % ConTEXt
\usetikzlibrary{pgfplots.smithchart} % LATEX and plain TEX
\usetikzlibrary[pgfplots.smithchart] % ConTEXt
A library to draw Smith Charts.
A Smith Chart maps the complex half plane with positive real parts to the unit circle. The smithchart
library allows pgfplots to visualize Smith Charts: it visualizes two–dimensional input coordinates
z 2 C of the form z = x + jy 2 C (j being the imaginary unit, j 2 = 1) with x 0 using the map
z 1
r : [0, 1] ⇥ [ 1, 1] ! {a + jb | a2 + b2 = 1}, r(z) =
z+1
using complex number division. The result is always in the unit circle.
The main application for Smith Charts is in the area of electrical and electronics engineers specializing
in radio frequency: to show the reflection coefficient r(z) for normalised impedance z. It is beyond the
scope of this manual to delve into the radio frequency techniques; for us, it is important to note that
the smithchart library supports
• the data map r(z) shown above,
• an axis class which interprets x as the real components and y as the imaginary components,
• a visualization of grid lines as arcs,
• the possibility to stop grid lines to allow uniform spacing in Smith Charts,
• a large set of the pgfplots axis fine tuning parameters,
• input of already mapped coordinates r(z) (i.e. Cartesian coordinates in the unit circle),
• many of the pgfplots plot handlers.
5.11.1 Smith Chart Axes

\begin{smithchart}[hoptionsi]
\end{smithchart}
The \begin{smithchart} environment draws Smith Charts. It accepts the same hoptionsi as
\begin{axis}. In fact, it is equivalent to \begin{axis}[hoptionsi,axis type=smithchart].
% Preamble: \pgfplotsset{width=7cm,compat=1.13}\usepgfplotslibrary{smithchart}
Impedance Smith Chart \begin{tikzpicture}
1 \begin{smithchart}[title=Impedance Smith Chart]
\addplot coordinates {(0.5,0.2) (1,0.8) (2,2)};
0.5 2 \end{smithchart}
\end{tikzpicture}
0.2 5
0.2 0.5 1 2 5
0
0.2 5
0.5 2
1
The example above visualizes three data points using the initial configuration of Smith Charts; the data
points are interpreted as complex numbers z = x + jy and are mapped using r(z).
Here, the x coordinate refers to the cycles described by the horizontal line whereas the y coordinate
refers to the cycles described by the tick labels on the outside.
/pgfplots/smithchart mirrored=true|false (initially false)

pgfplots also supports Admittance Smith Charts. Here, the origin is on the right side of the
circle:
Admittance Smith Chart \begin{tikzpicture}
1 \begin{smithchart}[
2 0.5 smithchart mirrored,
title=Admittance Smith Chart]
\end{smithchart}
5 0.2 \end{tikzpicture}
5 2 1 0.5 0.2
0
5 0.2
2 0.5
1
Since pgfplots can draw two axes on top of each other, a combined Impedance/Admittance Smith
Chart is also possible.
Impedance and Admittance Smith Chart

1
1
0. 5
0.5
2
(3)
0.2
0. 2
5 (2) 5
(1)
0.2 0.5 1 2 5
0
0 5 2 1 0.5 0.2
5
5
0.2
0.2
2
2
0. 5
0.5
1
1
\begin{tikzpicture}[]
% First, the Admittance chart:
\begin{smithchart}[
title=Impedance and Admittance Smith Chart,
smithchart mirrored,
xticklabel shift=-19pt,
grid style={blue},
ticklabel style={blue},
yticklabel around circle,
]
\end{smithchart}
% Second, overlay the impedance chart:

\begin{smithchart}[
show origin,
grid style={red},
ticklabel style={red},
yticklabel around circle*,
]
\addplot+[black,mark=o,only marks,point meta=explicit symbolic,nodes near coords]
coordinates {
(0.2,0.2) [(2)]
(1,0.2) [(1)]
};
\addplot+[black,no marks,domain=0.2:1] {0.2};
\addplot+[black,mark=o,only marks,point meta=explicit symbolic,nodes near coords]

coordinates{
(0.2,0.5) [(3)]
};
\end{smithchart}
\end{tikzpicture}
Since such a chart easily becomes crowded, it should be tuned manually by means of “suitable” appear-
ance options (if you feel that there is a “suitable default”, let me know).
Details for show origin, yticklabel around circle, and yticklabel around circle* can be found
later in this section.
5.11.2 Size Control

A Smith Chart can be resized by providing either width or height as argument to the axis. If you provide
both, the Chart is drawn as an ellipsis.
The tick and grid positions for smithchart axes are realized by means of three manually tuned sets of grid
lines: one for small-sized plots, one for medium-sized plots and one for huge plots. The actual parameters
for width or height are considered to select one of the following sets:
/pgfplots/few smithchart ticks (style, no value)

This produces the output of the example above – it constitutes the initial configuration for Smith Chart
which has a width of less than 14cm.
The few smithchart ticks style is defined by:
\pgfplotsset{
few smithchart ticks/.style={
default smithchart xtick/.style={
xtick={0.2,0.5,1,2,5},
},
default smithchart ytick/.style={
ytick={%
0,%
0.2, 0.5, 1, 2, 5,%
-0.2,-0.5,-1,-2,-5},
},
default smithchart xytick/.style={
xgrid each nth passes y={2},
ygrid each nth passes x={2},
},
},
}
Note that few smithchart ticks contains syntactical overhead to distinguish between “default ticks”
and final tick positions: it does not assign xtick and ytick directly. Instead, it provides them as
separate default xtick style arguments. The purpose of this distinction is to mark them as “default”
arguments – the underlying styles smithchart/every default xtick is used if and only if there is no
xtick value given.
In case you want to override this default, you can either
• copy–paste the definition above and adjust it or
• omit all the default smithchart xtick/.style stu↵ and write xtick={hyour listi} directly.
As mentioned, the only purpose of the default smithchart xtick/.style overhead is to distinguish
between \begin{smithchart}[xtick={huser defined i}] and default arguments (see the documentation
of default smithchart xtick/.style for more about this technical detail).
For fine tuning of the scaling decisions, see the smith chart ticks by size key.
/pgfplots/many smithchart ticks (style, no value)

The many smithchart ticks style is used for every Smith Chart whose width exceeds 14cm although
it is less than 20cm:
Medium–Sized Smith Chart
0.9
1
0.8
0.7
0. 6
1.5
0.5
2
4
0.
0.3 3
0. 2 4
5
0.1
10
0 0.1 0.2 0.3 0.4 0.5 1 1.5 2 3 4 5 10 2020
20
0. 1
10
0. 2
5
0. 3
4
4
0.
3
0.5
2
0.6
0.7
1.5
0.8
0.9
1
\begin{tikzpicture}
\begin{smithchart}[
title=Medium--Sized Smith Chart,
width=14cm]
\end{smithchart}
\end{tikzpicture}
We see that many smithchart ticks has di↵erent placement and alignment options than few
smithchart ticks: it uses sloped tick labels inside of the unit circle for the y descriptions (imagi-
nary axis).
The initial configuration is realized by means of two separate styles: one which defines only the tick
positions (the many smithchart ticks* style) and one which also changes placement and alignment
options. The initial configuration can be changed individually (see the end of this section for examples).
The initial configuration is:
\pgfplotsset{
many smithchart ticks*/.style={
xtick={
0.1,0.2,0.3,0.4,0.5,1,1.5,2,3,4,5,10,20%
},
minor xtick={0.6,0.7,0.8,0.9,1.1,1.2,1.3,1.4,1.6,1.7,1.8,1.9,
2.2,2.4,2.6,2.8,3.2,3.4,3.6,3.8,4.5,6,7,8,9,50},
},
ytick={%
0,%
0.1,0.2,...,1,1.5,2,3,4,5,10,20,%
-0.1,-0.2,...,-1,-1.5,-2,-3,-4,-5,-10,-20%
},
minor ytick={%
1.1,1.2,1.3,1.4,1.6,1.7,1.8,1.9,2.2,2.4,2.6,2.8,3.2,3.4,3.6,3.8,
4.5,6,7,8,9,50,%
-1.1,-1.2,-1.3,-1.4,-1.6,-1.7,-1.8,-1.9,-2.2,-2.4,-2.6,-2.8,
-3.2,-3.4,-3.6,-3.8,-4.5,-6,-7,-8,-9,-50%
},
},
xgrid each nth passes y={1,2,4,5,10,20},
ygrid each nth passes x={1,2,3,5,10:3,20:3},
},
},
/pgfplots/many smithchart ticks/.style={
many smithchart ticks*,
yticklabel in circle,
show origin=true,
},
}
See the documentation for few smithchart ticks for an explanation of the default smithchart
xtick/.style overhead.
/pgfplots/dense smithchart ticks (style, no value)

The dense smithchart ticks style assigns the set of tick positions for every Smith Chart whose width
is at least 20cm:
Huge Smith Chart (rescaled)
1
0.9
0.8
1.2
0.7
1.4
0.6
1. 6
1.8
0. 5
2
4
0.
0.3 3
4
0.2
5
0.1
10
20
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.2 1.4 1.6 1.8 2 3 4 5 10 20
20
0.1
10
0. 2
5
4
0. 3
3
4
0.
0.5
2
1.8
0.6
1.6
0.7
1.4
0.8
1.2
0.9
\begin{tikzpicture}[scale=0.75]
\begin{smithchart}[
title=Huge Smith Chart (rescaled),
width=20cm]
\end{smithchart}
\end{tikzpicture}
Attention: This style might change in future versions!

Similarly to many smithchart ticks (see above), the initial configuration is realized by means of two
separate styles: one which defines only the tick positions (the many smithchart ticks* style) and one
which also changes placement- and alignment options:
\pgfplotsset{
dense smithchart ticks*/.style={
xtick={
0.1,0.2,0.3,0.4,0.5,0.6,0.7,0.8,0.9,1,1.2,1.4,1.6,1.8,2,3,4,5,10,20%
},
minor xtick={%
0.01,0.02,0.03,0.04,0.05,0.06,0.07,0.08,0.09,0.11,0.12,0.13,0.14,0.15,0.16,0.17,
0.18,0.19,0.22,0.24,0.26,0.28,0.32,0.34,0.36,0.38,0.42,0.44,0.46,0.48,%
0.52,% This is sub-optimal and will (hopefully) be improved in the future.
0.55,0.65,0.75,0.85,0.95,%
%0.6,0.7,0.8,0.9,%
1.1,1.3,1.5,1.7,1.9,%
2.2,2.4,2.6,2.8,3.2,3.4,3.6,3.8,4.5,6,7,8,9,50},
},
ytick={%
0,%
0.1,0.2,...,1,1.2,1.4,1.6,1.8,2,3,4,5,10,20,%
-0.1,-0.2,...,-1,-1.2,-1.4,-1.6,-1.8,-2,-3,-4,-5,-10,-20%
},
minor ytick={%
0.01,0.02,0.03,0.04,0.05,0.06,0.07,0.08,0.09,0.11,0.12,0.13,0.14,0.15,0.16,0.17,
0.18,0.19,0.22,0.24,0.26,0.28,0.32,0.34,0.36,0.38,0.42,0.44,0.46,0.48,%
0.55,0.65,0.75,0.85,0.95,%
1.1,1.3,1.5,1.7,1.9,2.2,2.4,2.6,2.8,3.2,3.4,3.6,3.8,4.5,6,7,8,9,50,%
-0.01,-0.02,-0.03,-0.04,-0.05,-0.06,-0.07,-0.08,-0.09,-0.11,-0.12,-0.13,-0.14,
-0.15,-0.16,-0.17,-0.18,-0.19,-0.22,-0.24,-0.26,-0.28,-0.32,-0.34,-0.36,-0.38,
-0.42,-0.44,-0.46,-0.48,-0.55,-0.65,-0.75,-0.85,-0.95,%
-1.1,-1.3,-1.5,-1.7,-1.9,-2.2,-2.4,-2.6,-2.8,-3.2,-3.4,-3.6,-3.8,-4.5,-6,-7,-8,
-9,-50%
},
},
xgrid each nth passes y={0.2 if < 0.2001,0.5 if < 0.50001,1 if < 1.001,2,4,5,10,20},
ygrid each nth passes x={0.2 if < 0.2001,0.52 if < 0.52001,1 if < 1.001,2,3,5,10:3,20:3},
},
},
dense smithchart ticks/.style={
yticklabel in circle,
dense smithchart ticks*,
show origin=true,
every major grid/.style={black!60},
},
}
See the documentation for few smithchart ticks for an explanation of the default smithchart
xtick/.style overhead.
/pgfplots/default smithchart xtick (no value)

/pgfplots/default smithchart ytick (no value)
/pgfplots/default smithchart xytick (no value)
The default smithchart xtick style is installed if and only if you do not provide xtick manually.
Similarly, the default smithchart ytick style is installed if and only if you do not provide ytick
manually.
Finally, the default smithchart xytick style is installed if and only if you provide neither xtick nor
ytick.
These styles are usually defined in few smithchart ticks and its variants, see above.
5.11.3 Working with Prepared Data

/pgfplots/is smithchart cs=true|false (initially false)
Occasionally, you may already have input data transformed into unit–circle Cartesian coordinate r(z) =
(x, y).
You can provide them to pgfplots with the is smithchart cs key:
1
\begin{tikzpicture}
0.5 2 \begin{smithchart}
% smithchart_data.dat contains
% 0.78395 -0.40845
% 0.78165 -0.41147
0.2 5 % 0.77934 -0.41466
% 0.77774 -0.41869
% ...
0.2 0.5 1 2 5 \addplot[blue,is smithchart cs]
0 file {plotdata/smithchart_data.dat};
\end{smithchart}
\end{tikzpicture}
0.2 5
0.5 2
1
Using is smithchart cs tells pgfplots to skip the transformation r(z).
5.11.4 Appearance Control and Styles

/pgfplots/show origin=true|false (initially false)
Allows to place an extra description at the point (0, 0) to mark the origin.
1
\begin{tikzpicture}
0.5 2 \begin{smithchart}[show origin]
\end{smithchart}
\end{tikzpicture}
0.2 5
0.2 0.5 1 2 5
0
0.2 5
0.5 2
1
/pgfplots/show origin code/.code={h... i}

Allows to redefine the code to draw the origin marker. The initial configuration is
\pgfplotsset{
show origin code/.code={%
\path[draw=black,fill=white] (0pt,0pt) circle (2.5pt);
\path[fill=black] (0pt,0pt) circle (0.5pt);
}
}
/pgfplots/yticklabel in circle (style, no value)

This style draws Smith Chart tick labels for imaginary components (the ytick arguments) inside of the
circle.
It installs transformations to rotate and shift tick labels. See the many smithchart ticks style for an
example.
The initial configuration for this style is
\pgfplotsset{
yticklabel in circle/.style={
ytick align=inside,
yticklabel style={
rotate=90,
sloped like y axis={%
execute for upside down={\tikzset{anchor=north east}},
%allow upside down,
reset nontranslations=false},
anchor=south west,
%font=\tiny,
}
}
}
/pgfplots/yticklabel around circle (style, no value)

This style draws Smith Chart tick labels for imaginary components (the ytick arguments) outside of
the circle, but rotated to fit the slope of the circle.
1
\begin{tikzpicture}
0.5
\begin{smithchart}[yticklabel around circle]

\end{smithchart}
\end{tikzpicture}
5
0.2
0 0.2 0.5 1 2 5
0.2
5
0. 5
2
1
\pgfplotsset{
yticklabel around circle/.style={
ytick align=center,
yticklabel style={
rotate=90,
execute for upside down={\tikzset{anchor=south west}},
%allow upside down,
anchor=south east,
}
},
}
/pgfplots/yticklabel around circle* (style, no value)

A variant of yticklabel around circle which exchanges the anchors:
1
0. 5
\begin{tikzpicture}
\begin{smithchart}[yticklabel around circle*]
2
\end{smithchart}
\end{tikzpicture}
0.2
5
0.2 0.5 1 2 5
0
5
0.2
2
0.5
If you have two Smith Charts in the same figure, you can overlay them if the first uses yticklabel
around circle and the second uses yticklabel around circle* (see the beginning of this section for
an example).
\pgfplotsset{
yticklabel around circle*/.style={
ytick align=center,
yticklabel style={
rotate=90,
execute for upside down={\tikzset{anchor=north west}},
%allow upside down,
anchor=north east,
}
}
}
/pgfplots/every smithchart axis (style, no value)

This style is installed for every Smith Chart. It is defined as
\pgfplotsset{
every smithchart axis/.style={
grid=both,
xmin=0,
scaled ticks=false, % never draw the \cdot 10^4 labels
major tick style={draw=black},
xtick align=center,
ytick align=center,
},
}
5.11.5 Controlling Arcs and Their Stop Points

This section allows advanced control over Smith Chart arcs (grid lines). The two features xgrid each nth
passes y and xgrid stop at y (and their counterparts for y) allow to draw only partial arcs in order to
get a more uniform appearance.
/pgfplots/xgrid each nth passes y={hlist of stop entriesi} (initially empty)

This key constitutes the main idea to draw only partial arcs: you provide a couple of y tick coordinates
which constitute “boundaries”. Then, only each (say) second x grid line is allowed to pass these
boundaries:
1
\begin{tikzpicture}
0.5 2 \begin{smithchart}[
xtick={0.2,0.5,1,2,5},
ytick={
0,
0.2 5 0.2, 0.5, 1, 2, 5,
-0.2,-0.5,-1,-2,-5},
xgrid each nth passes y={1,2},
0.2 0.5 1 2 5 ]
0 \end{smithchart}
\end{tikzpicture}
0.2 5
0.5 2
1
The example overwrites the default smithchart ticks to define a new layout: now, every ytick uses
the complete arc, but some of the grid lines for xtick stop at y = 1 and, if they pass, they may stop at
y = 2.
The argument hlist of stop entriesi is a comma–separated list of entries. Each entry is, in the simplest
case, a y coordinate (it should be a coordinate which appears in the ytick list). This simplest case
means “only each second x grid line may pass the grid line for this y”. The second syntax allows to
provide a natural number, using hy coord i:hnumber i. This means to let only each hnumber i’s x grid
line pass the designated y grid line. The third syntax also allows to write if < hx valuei. It means
the entry is considered only for x grid lines which are less than hx valuei. To summarize: there are the
three possible forms of entries
1. single y coordinates, for example xgrid each nth passes y={1,2} or
2. the same as above, followed by an integer, for example xgrid each nth passes y={1:3,2:2} or
3. an additional restriction clause like xgrid each nth passes y={0.2 if <0.3}.
In this case, the all x grid lines which fulfill x  0.3 will be checked if they are allowed to pass
y = 0.2. All x grid lines with x > 0.3 are not a↵ected by the constraint. See the dense smithchart
ticks style for an application example.
Note that xgrid each nth passes y always employs symmetry; you do not need to provide y and y
(if you want to, you may use the xgrid stop at y key to overrule the “each nth”-strategy).
In order to check if a given xtick argument is the “nth” grid line, pgfplots collects all xtick and
minor xtick arguments into one large array and sorts it. Then, it uses the resulting sequence to assign
the indices. Consequently, you can freely intermix minor and major ticks; it will still work. The only
way to a↵ect the counting is the xgrid each nth passes y start key, see below.
/pgfplots/ygrid each nth passes x={hlist of stop entriesi} (initially empty)

As you may already have guessed, this is the y counterpart of xgrid each nth passes y. It restricts
the arcs for y grid lines by provided x ticks:
1
\begin{tikzpicture}
xtick={0.2,0.5,1,2,5},
ytick={
0,
0.2 5 0.2, 0.5, 1, 2, 5,
-0.2,-0.5,-1,-2,-5},
ygrid each nth passes x={0.2,1:2},
0.2 0.5 1 2 5 ]
0 \end{smithchart}
\end{tikzpicture}
0.2 5
0.5 2
1
The syntax is exactly the same as explained for xgrid each nth passes y. The only di↵erence is that
the if < syntax uses absolute values y (to maintain symmetry).
Now, we know how to use xgrid each nth passes y and the corresponding ygrid each nth passes
x separately. Can we use both keys at the same time? Yes – but it may happen that lines end in white space!
pgfplots applies some logic to avoid arcs ending in white space by extending them to the next feasible
stopping point. The result of mixing both of these keys is thus corrected automatically.
/pgfplots/xgrid each nth passes y start={hinteger i} (initially 0)
/pgfplots/ygrid each nth passes x start={hinteger i} (initially 0)
Allows to modify where the “each nth” counting starts. The argument can be considered as a shift. I
consider this key to be more or less experimental – in the hope it may be useful. Try it out.
/pgfplots/xgrid stop at y={hlisti} (initially empty)
/pgfplots/ygrid stop at x={hlisti} (initially empty)
These keys allow to provide individual stop points for explicitly chosen tick positions. These explicit
stop points have higher precedence over the each nth features described above.
The ygrid stop at x key accepts a comma–separated list of entries hy coord i:hx stop pointi:
1
\begin{tikzpicture}
ygrid stop at x={0.5:0.5,-0.2:0.2}
]
\end{smithchart}
0.2 0.5 1 2 5
0
0.2 5
0.5 2
1
In this example, the y = 0.5 arc stops at the x = 0.5 arc whereas the y = 0.2 arc stops at x = 0.2.
The ygrid stop at x key allows unsymmetric layouts (di↵erent stop points for y and y).
5.12 Statistics
\usepgfplotslibrary{statistics} % LATEX and plain TEX
5.12. STATISTICS 469
\usepgfplotslibrary[statistics] % ConTEXt
\usetikzlibrary{pgfplots.statistics} % LATEX and plain TEX
\usetikzlibrary[pgfplots.statistics] % ConTEXt
A library which provides plot handlers for statistics.
5.12.1 Box Plots

Box plots are visualizations for one–dimensional distributions. They provide a fast overview over charac-
teristics of the distribution. Box plots are inherently one–dimensional; they only use a second axis to place
multiple box plots next to each other.
pgfplots supports two related plot handlers: boxplot and boxplot prepared. The boxplot handler
takes a one–dimensional sample as input, computes the median, the lower quartile, the upper quartile,
the lower whisker and the upper whisker, and visualizes the result using the boxplot prepared handler.
The boxplot prepared handler expects all required values on input and visualizes them.
Prepared Box Plots and Common Options

The boxplot prepared handler is discussed first; all its customizations apply to boxplot as well.
/pgfplots/boxplot prepared={hoptions with boxplot/ prefix i}
\addplot+[boxplot prepared={hoptions with boxplot/ prefix i}]

A boxplot prepared takes a couple of key–value pairs which describe the required statistics and a
coordinate stream of outliers on input and draws a box plot.
% Preamble: \pgfplotsset{width=7cm,compat=1.13}\usepgfplotslibrary{statistics}
\begin{axis}
\addplot+[
1.2 boxplot prepared={
lower whisker=5,
lower quartile=7,
1
median=8.5,
upper quartile=9.5,
0.8 upper whisker=10,
},
]
0.6 table[row sep=\\,y index=0] {
data\\ 1\\ 3\\
2 4 6 8 10 };
\end{axis}
\end{tikzpicture}
The previous example shows the main idea of boxplot prepared: each required quantity has to be pro-
vided explicitly using a key-value syntax. The following coordinate stream can be empty; all coordinates
inside of it are considered to be outliers. They are drawn as scatter plot.
A box plot produces two coordinates: one which belongs to the input data (for example median) and
one which is only used for drawing purposes. Limits will be updated for both of them. While this is
clear for the axis which shows the input data (x in this example, see also draw direction), it should
be noted that limits and scaling parameters for the other axis will be chosen just as for any other plot.
The box’s extend is a little bit less than one unit by default (compare box extend). As a consequence,
would might need to adjust either the limits or the scaling parameters for the remaining axis:
1.4
1 \begin{axis}[
0.8 y=1.5cm,
0.6 ]
2 4 6 8 10 \addplot+[
boxplot prepared={
lower whisker=5,
lower quartile=7,
median=8.5,
upper quartile=9.5,
upper whisker=10,
},
]
table[row sep=\\,y index=0] {
data\\ 1\\ 3\\
};
\end{axis}
\end{tikzpicture}
If you place multiple blots with handler boxplot prepared into the same axis, they will automatically
be placed next to each other by means of the default value of draw position:
/pgfplots/boxplot/draw position={haxis unit to place box i} (initially 1+\plotnumofactualtype)

The draw position key determines how to choose the position of the box plot, i.e. the axis unit which
is free to choose.
The initial configuration places the first box plot at y = 1 and all following ones at the next integer
numbers.
\begin{tikzpicture}
\begin{axis}
3 \addplot+[
boxplot prepared={
lower whisker=42, lower quartile=45,
median=47,
2
upper quartile=47.5, upper whisker=48,
},
]
table[row sep=\\,y index=0] { 40\\ 34\\ 56\\ };
1
\addplot+[
boxplot prepared={
35 40 45 50 55 lower whisker=36, lower quartile=39,
median=40,
upper quartile=41, upper whisker=43,
},
]
% no outliers:
coordinates {};
\addplot+[
boxplot prepared={
median=45,
},
]
coordinates {(0,35) (0,55)};
\end{axis}
\end{tikzpicture}
The preceding example shows three box plots in the same axis. Note that they have been aligned using
the default setting.
The fact that the first N boxplot (or the equivalent boxplot prepared) are placed at the coordinates
1, 2, 3, . . . , N makes it simple to assign tick labels: either use ytick=data or ytick=1,2,3 combined
with yticklabels:
\begin{tikzpicture}
Group C \begin{axis}[
ytick={1,2,3},
yticklabels={Group A, Group B, Group C},
]
Group B \addplot+[
boxplot prepared={
median=47,
Group A upper quartile=47.5, upper whisker=48,
},
]
35 40 45 50 55
\addplot+[
boxplot prepared={
median=40,
},
]
coordinates {};
\addplot+[
boxplot prepared={
median=45,
},
]
coordinates {(0,35) (0,55)};
\end{axis}
\end{tikzpicture}
The draw position may be read from some input table using draw position=\thisrowhcolnamei. In
this case, the last encountered data row will be used (this remark is, of course, only useful if a data
stream is present).
While the default choice of draw position hopefully covers the most common use-cases, one can also
assign a custom value to it if specific box plots should be placed individually:
\begin{axis}[
boxplot/draw direction=y,
]
10 \addplot+[boxplot prepared={
lower whisker=2.5, lower quartile=4,
median=8.5, upper quartile=12,
upper whisker=15}]
coordinates {};
5 \addplot+[boxplot prepared={
upper whisker=15}]
1 2 3 4 5 coordinates {};
\addplot+[boxplot prepared={draw position=5,
upper whisker=15}]
coordinates {};
\end{axis}
\end{tikzpicture}
The example shows the first two plots using the default draw position. As this uses the plot index
+1, they are placed at x = 1 and x = 2, respectively (using x due to draw direction=y). The third
plot has draw position=5 and is drawn at x = 5.
Note that if you assign draw position for one plot, you may also need to adopt all followings ones as
pgfplots does not automatically detect collisions.
The preceeding examples read their outlier data streams from the y coordinate of the input streams: for
\addplot table, we have explicitly said y index=0 and for \addplot coordinates, we have used (0,35)
(0,55) where the x components are ignored. This default can be changed using the boxplot/data key.
/pgfplots/boxplot/data={hexpressioni} (initially y)
Tells boxplot how to get its data. The common idea is to provide a mathematical hexpressioni which de-
pends on data supplied by the \addplot statement. For example, if you have \addplot expression, the
hexpressioni may depend upon x, y or z. In case of an \addplot table input routine, the hexpressioni
can employ \thisrow{hcolnamei} to access the currently active table row in the designated column.
It is also possible to avoid invocations of the math parser. Use boxplot/data value={hvaluei} instead
to do so. Here, hvaluei should be of a numeric constant.
The initial configuration employs what would usually become the final y coordinate as input (to be
more precise, the initial value is data value=\pgfkeysvalueof{/data point/y}).
/pgfplots/boxplot/lower whisker={hvaluei} (initially auto)

/pgfplots/boxplot/lower quartile={hvaluei} (initially auto)
/pgfplots/boxplot/median={hvaluei} (initially auto)
/pgfplots/boxplot/upper quartile={hvaluei} (initially auto)
/pgfplots/boxplot/upper whisker={hvaluei} (initially auto)
/pgfplots/boxplot/average={hvaluei} (initially empty)
These keys constitute the supported statistics. Typically, a box plot uses each of them except for
average.
Any numeric value for hvaluei will be used as-is. This holds for both boxplot prepared and boxplot.
An empty hvaluei disables the respective key: its associated visualization will be omitted. This is the
default for average.
The value auto tells pgfplots to include the statistics in the automatic computation applied by
boxplot. It is irrelevant for boxplot prepared (where it is essentially the same as an empty hvaluei).
The definition of the values is as follows. Assume that we have a given sample of a distribution, say
x1 , . . . , xN , and assume that the values are sorted, x1 < · · · < xN (which is not a requirement for
boxplot, by the way). For any real number p with 0  p  1, the “p–quantile” (or p–percentage) is
defined as (
xN ·p if N · p is an integer number
xp := 1
2 (x bN pc + x dN ·pe ) if N · p is not an integer.
median is the 0.5–quantile of the input data: half of the points are less and half of the points are larger
than the median.
lower quartile is the 0.25–quantile of the input data.
upper quartile is the 0.75–quartile of the input data.
lower whisker is the smallest data value which is larger than lower quartile 1.5 · IQR where IQR
is the “inter–quartile–range”, i.e. the di↵erence between upper quartile and lower quartile.
upper whisker is the largest data value which is smaller than upper quartile+1.5 · IQR.
average is the sample average. It is omitted by boxplot in its default configuration. Set it to auto to
enable its auto-computation.
/pgfplots/boxplot/draw direction=x|y (initially x)

Since boxplot is inherently one–dimensional, it can be visualized along the x or the y axis.
The default configuration uses the x direction as seen above.
The alternative choice y lets the boxes and their whiskers extend along the y axis and stacks multiple
box plots along the x axis:
\begin{tikzpicture}
\begin{axis}[
boxplot/draw direction=y,
50 x axis line style={opacity=0},
axis x line*=bottom,
axis y line=left,
enlarge y limits,
ymajorgrids,
40 xtick={1,2,3},
xticklabels={Group A, Group B, Group C},
]
\addplot+[
boxplot prepared={
Group A Group B Group C lower whisker=42, lower quartile=45,
median=47,
},
]
\addplot+[
boxplot prepared={
median=40,
},
]
coordinates {};
\addplot+[
boxplot prepared={
median=45,
},
]
coordinates {(0,35) (0,55)};
\end{axis}
\end{tikzpicture}
/pgfplots/boxplot/variable width=true|false (initially false)

If enabled, the box extend will be scaled according to the sample size relative to all other boxplot
or boxplot prepared within the same axis.
The key variable width only has an e↵ect if sample size is available. For boxplot prepared, one
needs to provide sample size explicitly.
\begin{tikzpicture}
Group C \begin{axis}[
ytick={1,2,3},
yticklabels={Group A, Group B, Group C},
boxplot/variable width,
Group B ]
\addplot+[% Group A:
boxplot prepared={
Group A median=47,
sample size=1000,
},
35 40 45 50 55 ]
\addplot+[% Group B:
boxplot prepared={
median=40,
sample size=100000,
},
]
coordinates {};
\addplot+[% Group C:
boxplot prepared={
median=45,
sample size=50000,
},
]
coordinates {(0,35) (0,55)};
\end{axis}
\end{tikzpicture}
The variable width computation computes the largest and smallest value of sample size, chosen
among all box plots in the same axis. If a single plot has no sample size, it will be omitted from the
computation (and it will not be scaled). If a single box plot has variable width=false, its sample
size will not contribute either. The box plot with largest value of sample size will be drawn with
100% of box extend. The box plot with smallest value will be drawn with variable width min target
times box extend as size (i.e. it will receive the smallest configured size). All box plots in-between are
scaled linearly.
Note that the previous paragraph is not entirely true: sample size is only indirectly related to the
scaling factor. Instead, the variable width expr is evaluated with the sample size as argument (see
below for details).
/pgfplots/boxplot/sample size={hnumber i} (initially auto)

The number of samples used to derive the statistics. This number is used if variable width=true.
The value auto means to “use it whenever it can be acquired somewhere”. For a boxplot, it means
that the size of the input sample is taken as-is. For a boxplot prepared, it means that the data is
unavailable.
The empty string means that the value is unavailable.
Otherwise, a number is expected.
/pgfplots/boxplot/variable width expr={hmath expressioni} (initially sqrt(#1))

A math expression which is used to evaluate the scaling factors of variable width. The argument is
the current value of sample size. This key is used to implement common (nonlinear) transformations
which are to be applied to the sample size before the result is used to scale down box sizes.
Typically, the argument should be a monotonically increasing function.
/pgfplots/boxplot/sample size min={hmin sample size of groupi} (initially empty)

/pgfplots/boxplot/sample size max={hmax sample size of groupi} (initially empty)

This is part of the variable width scaling: it is used to determine the box extend relative to all other
box plots of the same group. It fixes the range.
/pgfplots/boxplot/variable width min target={hfactor for the box width minimal sizei} (initially
0.2)
Used for the variable width feature to determine the size for the box plot with smallest value of
sample size. The argument is interpreted to be a scaling factor in the range [0, 1].
It is to be understood as percentage of box extend: a value of 1 means 100% of box extend. The
initial configuration is 0.2, meaning 20% of box extend.
The box plot with largest value of sample size has 100% of box extend.
/pgfplots/boxplot/box extend={haxis unit for box extensioni} (initially 0.8)

A parameter which controls the size of the boxes with respect to the axis orthogonal to the data axis.
The box extend is used as follows: if a box is centered at say, y = 1 with box extend=1, the box will
start at 1/2 and will end at 11/2. In other words: the box size is 1 y unit.
It is interpreted as coordinate in the axis, and it a↵ects the automatic computation of ymin and ymax
values for a boxplot.
\begin{tikzpicture}
\begin{axis}[minor y tick num=1]
2 \addplot+[
boxplot prepared={
1.5 median=47,
box extend=1, }, ]
1 table[row sep=\\,y index=0] { 40\\ 34\\ 56\\ };
\addplot+[
0.5 boxplot prepared={
35 40 45 50 55 median=47,
box extend=0.5, }, ]
\end{axis}
\end{tikzpicture}
The box extend controls the size of the box and the length of the median line. It also controls the size
of whiskers, although they have a separate parameter whisker extend.
It is supposed to be the low–level size of a box plot, although it can be interpreted and used as a
“low–level–variant” of variable width.
/pgfplots/boxplot/whisker extend={haxis unit for whisker extensioni} (initially

\pgfkeysvalueof{/pgfplots/boxplot/box extend}*0.8)
A parameter which configures how large whisker lines are with respect to the non–data axis.
It is used in the same way as box extend, and it also a↵ects axis limits.
The initial configuration couples its value to box extend (it is 80% of box extend, to be more precise).
/pgfplots/boxplot/draw relative anchor={hnumber between 0 and 1i} (initially 0.5)

A key which customizes the anchor inside of the box where the whiskers are attached.
1.8
1.4 \begin{axis}[y=1cm]
1.2
1 \addplot+[
35 40 45 50 55 boxplot prepared={
median=47,
draw relative anchor=0, }, ]
\end{axis}
\end{tikzpicture}
1.4
1 \begin{axis}[y=1cm]
0.8
0.6 \addplot+[
median=47,
draw relative anchor=0.5, }, ]
\end{axis}
\end{tikzpicture}
1
0.6 \begin{axis}[y=1cm]
0.4
0.2 \addplot+[
median=47,
draw relative anchor=1, }, ]
\end{axis}
\end{tikzpicture}
The value 0 means that whisker lines are attached to the bottom of the box (0% of box extend). The
value 1 means that whisker lines are attached to the top edge of the box. Any value in-between is scaled
linearly. The initial configuration is 0.5 which means that whiskers are attached to the middle of the
box.
Analyzing Samples Automatically

/pgfplots/boxplot={hoptions with boxplot/ prefix i}
\addplot+[boxplot={hoptions with boxplot/ prefix i}]
The boxplot handler takes a one–dimensional sample as input, computes the median, the lower
quartile, the upper quartile, the lower whisker and the upper whisker, and visualizes the re-
sult using the boxplot prepared handler.
Attention: Computing the statistics automatically is considerably faster if you use compat=1.12
combined with lualatex: this library has a special lua backend which allows scalability, speed, and
accuracy beyond TeX’s capabilities.
1.4
1 \begin{axis}[y=1cm]
0.8
0.6 \addplot+[boxplot]
2 4 6 8 10 table[row sep=\\,y index=0] {
data\\
1\\ 2\\ 1\\ 5\\ 4\\ 10\\
7\\ 10\\ 9\\ 8\\ 9\\ 9\\
};
\end{axis}
\end{tikzpicture}
The values do not need to be sorted. However, if they are sorted in ascending order, pgfplots might
need less time to analyze them.
Data points can be given by means of any supported input stream, although the most useful ones
are probably \addplot table and \addplot coordinates. In any case, boxplot acquires only one–
dimensional data. To this end, it uses the current value of the boxplot/data key to see which input
coordinate is to be used. In the default configuration, this is the y coordinate of the input stream. All
other input items are ignored (except for point meta, which is handed down to the outlier stream).
/pgfplots/boxplot/estimator=value (initially Excel)

Selects one of 10 available boxplot value estimators.
The default estimator is R7 alias Excel if pgfplots is configured to use compat=1.12 or higher. For
all older compatibility levels, it is legacy.
The choice R1 resembles the estimator type 1 used by R. It has aliases SAS3 and Maple1. This choice is
currently limited to the lua backend.
The choice R2 resembles the estimator type 2 used by R. It has aliases SAS5 and Maple2. This choice is
currently limited to the lua backend.
The choice R3 resembles the estimator type 3 used by R. It has aliases SAS2.
The choice R4 resembles the estimator type 4 used by R. It has aliases SAS1, SciPy0-1, and Maple3.
The choice R5 resembles the estimator type 5 used by R. It has aliases SciPy12-12 and Maple4.
The choice R6 resembles the estimator type 6 used by R. It has aliases SAS4, SciPy0-0, and Maple5.
The choice R7 resembles the estimator type 7 used by R. It has aliases Excel, SciPy1-1 and Maple6.
The choice R8 resembles the estimator type 8 used by R. It has aliases ScuPy13-13 and Maple7.
The choice R9 resembles the estimator type 9 used by R. It has aliases SciPy38-38 and Maple8.
The choice legacy is a minimally repaired variant of the estimator which was shipped with the first
version of the statistics library. It is merely kept for reasons of backwards compatibility18 .
/pgfplots/boxplot/whisker range={hnumber i} (initially 1.5)

Defines how to determine lower whisker and upper whisker. In the default configuration, the lower
whisker is placed at the smallest data point which is larger than lower quartile 1.5·IQR. The upper
whisker is placed at the largest data point which is smaller than upper quartile +1.5 · IQR. Here,
IQR is the inter–quartile–range, defined as
IQR := upper quartile lower quartile.
Everything outside of the whisker range is supposed to be an outlier.
Styles
/pgfplots/boxplot/every boxplot (style, no value)
A style which is immediately installed whenever boxplot or boxplot prepared are set.
The initial value is empty.
/pgfplots/boxplot/every whisker (style, no value)

A style which is installed whenever a whisker is drawn. It is empty initially.
/pgfplots/boxplot/every box (style, no value)

A style which is installed whenever a box is drawn. It is empty initially. Note that this does not apply
to the path for the median.
/pgfplots/boxplot/every median (style, no value)

A style which is installed whenever a median is drawn. It is empty initially.
/pgfplots/boxplot/every average (style, no value)

A style which is installed whenever an average is drawn. The initial configuration is
18 There is also an estimator called legacy*. This is the original one shipped with the first version of this library. It is
discouraged but kept in case someone really needs it.
\pgfplotsset{
boxplot/every average/.style={%
/tikz/mark=diamond*,
},
}
Placing Annotations
\pgfplotsboxplotvalue{hkey namei}
Same as
\pgfkeysvalueof{/pgfplots/boxplot/hkey namei}.
\boxplotvalue{hkey namei}
Same as \pgfplotsboxplotvalue{hkey namei} (just shorted).
Coordinate system boxplot box

The boxplot box cs accepts two arguments specified a tuple of the form boxplot box cs=(hdata
coordinate, box–relative o↵seti) where the first is a value of the box plot’s data (it is expressed in the
same space as median or upper whisker).
The second argument is an o↵set expressed as signed multiple of box extend. An o↵set of 0 means to
place the point exactly on the bottom line of the box. An o↵set of 1 places the point on the top line of
the box. An o↵set of 0.5 places the point in the middle.
2
\begin{tikzpicture}
\begin{axis}[y=1.5cm, ymax=2]
1.5 1 3.5 9 10
\addplot+[boxplot]
1 7.5 data\\
1\\ 2\\ 1\\ 5\\ 4\\ 10\\
0.5 7\\ 10\\ 9\\ 8\\ 9\\ 9\\
2 4 6 8 10 }
[above]
node at
(boxplot box cs: \boxplotvalue{lower whisker},1)
{\pgfmathprintnumber{\boxplotvalue{lower whisker}}}
node at
(boxplot box cs: \boxplotvalue{lower quartile},1)
{\pgfmathprintnumber{\boxplotvalue{lower quartile}}}
node[left] at
(boxplot box cs: \boxplotvalue{median},0.5)
{\pgfmathprintnumber{\boxplotvalue{median}}}
node at
(boxplot box cs: \boxplotvalue{upper quartile},1)
{\pgfmathprintnumber{\boxplotvalue{upper quartile}}}
node at
(boxplot box cs: \boxplotvalue{upper whisker},1)
{\pgfmathprintnumber{\boxplotvalue{upper whisker}}}
;
\end{axis}
\end{tikzpicture}
Coordinate system boxplot whisker

A coordinate system which is almost the same as boxplot box cs, except that it aligns at whisker
extend instead of box extend.
The boxplot whisker cs accepts two arguments of the form boxplot whisker cs=(hdata coordinate,
whisker–relative o↵seti) where the first is a value of the box plot’s data (it is expressed in the same
space as median or upper whisker).
The second argument is an o↵set expressed as signed multiple of whisker extend. An o↵set of 0 means
to place the point exactly on the lower end of the whisker line. An o↵set of 1 places the point on the
upper end of the whisker line. An o↵set of 0.5 places the point in the middle of the whisker line.
2
\begin{tikzpicture}
\begin{axis}[y=1.5cm, ymax=2]
1.5 1 7.5 10 \addplot+[boxplot]
1 data\\
1\\ 2\\ 1\\ 5\\ 4\\ 10\\
0.5 7\\ 10\\ 9\\ 8\\ 9\\ 9\\
2 4 6 8 10 }
[above]
node at
(boxplot whisker cs:\boxplotvalue{lower whisker},1)
{\pgfmathprintnumber{\boxplotvalue{lower whisker}}}
node at
(boxplot box cs: \boxplotvalue{median},1)
{\pgfmathprintnumber{\boxplotvalue{median}}}
node at
(boxplot whisker cs:\boxplotvalue{upper whisker},1)
{\pgfmathprintnumber{\boxplotvalue{upper whisker}}}
;
\end{axis}
\end{tikzpicture}
Customizing Visualization Paths

The following keys are of interest if you want to redefine the shape of a box, of a median, or of the whiskers.
Note that you should customize styles like boxplot/every box if you merely wish to change fill colors.
/pgfplots/boxplot/draw/lower whisker/.code={h... i}
/pgfplots/boxplot/draw/upper whisker/.code={h... i}
/pgfplots/boxplot/draw/whisker/.code={h... i}
A couple of code keys which customize the stroke paths for whiskers.
\pgfplotsset{
boxplot/draw/lower whisker/.style={%
/pgfplots/boxplot/draw/whisker=%
{\pgfplotsboxplotvalue{lower quartile}}
{\pgfplotsboxplotvalue{lower whisker}}
},
boxplot/draw/upper whisker/.style={%
/pgfplots/boxplot/draw/whisker=%
{\pgfplotsboxplotvalue{upper quartile}}
{\pgfplotsboxplotvalue{upper whisker}}%
},
boxplot/draw/whisker/.code 2 args={%
\draw[/pgfplots/boxplot/every whisker/.try]
(boxplot cs:#1) -- (boxplot cs:#2)
(boxplot whisker cs:#2,0)
--
(boxplot whisker cs:#2,1)
;
},%
}
The key draw/lower whisker key is used if and only if lower whisker has a numeric value. The key
draw/upper whisker is used if and only if upper whisker has a value.
If one of lower quartile or upper quartile is empty, both are replaced by the following values:
lower quartile := upper whisker and
upper quartile := lower whisker.
Thus, if the box cannot be drawn but you only have whiskers, the two whiskers will be connected with
each other.
/pgfplots/boxplot/draw/box/.code={h... i}
A path which is used for every box.
\pgfplotsset{
boxplot/draw/box/.code={%
\draw[/pgfplots/boxplot/every box/.try]
(boxplot box cs:\pgfplotsboxplotvalue{lower quartile},0)
rectangle
(boxplot box cs:\pgfplotsboxplotvalue{upper quartile},1)
;
},%
}
It either lower quartile or upper quartile is empty, this key will not be invoked.
Note that draw/median will be invoked after this key.
/pgfplots/boxplot/draw/median/.code={h... i}
A path which is used for every median. Its initial configuration is
\pgfplotsset{
boxplot/draw/median/.code={%
\draw[/pgfplots/boxplot/every median/.try]
(boxplot box cs:\pgfplotsboxplotvalue{median},0)
--
(boxplot box cs:\pgfplotsboxplotvalue{median},1)
;
},%
}
This key will be omitted if median is empty.
/pgfplots/boxplot/draw/average/.code={h... i}
The path which is used to visualize an average. The initial configuration is
\pgfplotsset{
boxplot/draw/average/.code={%
\draw[/pgfplots/boxplot/every average/.try]
\pgfextra
% do NOT use \draw[mark=*] plot coordinates because
% boxplots uses the same plot handler to draw its
% outliers.
\pgftransformshift{%
% basic level access to ’boxplot box cs’:
\pgfplotsboxplotpointabbox
{\pgfplotsboxplotvalue{average}}
{0.5}%
}%
\pgfuseplotmark{\tikz@plot@mark}%
\endpgfextra
;
},
}
This key will be omitted if average has an empty value (the default).
The key draw/average will be evaluated after draw/median and after draw/box.
5.12.2 Histograms
/pgfplots/hist={hoptions with hist/ prefix i}
\addplot+[hist={hoptions with hist/ prefix i}]

A histogram plot takes one-dimensional input data and counts the occurrence of values: it determines
the data range [m, m] and subdivides it into N equally sized bins with (N + 1) end–points. Then,
it counts the number of points falling into each bin. More precisely, it computes the N + 1 points
m =: x0 < x1 < · · · < xN := m using xi := m + i · (m m)/N . Then, it creates the N + 1 coordinates
(xi , yi ), i = 0, . . . , N 1 by means of
( ⌘
bincount [xi , xi+1 ) i<N
yi :=
yN 1 i = N,
i.e. the value of the last coordinate is replicated. This set of (N +1) interval boundaries is then visualized
by an ybar interval plot handler.
\begin{tikzpicture}
\begin{axis}[
ybar interval,
6 xticklabel=
\pgfmathprintnumber\tick--\pgfmathprintnumber\nexttick
]
\addplot+[hist={bins=3}]
4 table[row sep=\\,y index=0] {
data\\
1\\ 2\\ 1\\ 5\\ 4\\ 10\\
7\\ 10\\ 9\\ 8\\ 9\\ 9\\
};
2
\end{axis}
\end{tikzpicture}
1–4 4–7 7–10
We see that hist={bins=3} takes a table with one column as input. The data values fall into the range
[1, 10] which is partitioned into 3 intervals (of equal lengths). Finally, the number of points falling into
each of the three bins is plotted. The xticklabel key shows the range (note that it works only in
conjunction with x tick label as interval which has been enabled by ybar interval before). We
see that there are 3 elements in the range [1, 4), 2 elements in the range [4, 7) and finally 7 elements in
the range [7, 10].
The bins are half–open intervals, i.e. the end–point does not belong to the bin. Only the last bin contains
its right end point.
\begin{tikzpicture}
\begin{axis}[
2,000 ybar interval,
xtick=,% reset from ybar interval
xticklabel=
{$[\pgfmathprintnumber\tick,%
\pgfmathprintnumber\nexttick)$}
1,000 ]]
% a data file containing 8000 normally distributed
% random numbers of mean 0 and variance 1
\addplot+[hist={data=x}]
file {plotdata/pgfplots.randn.dat};
0
\end{axis}
[ 4, 2)[ 2, 0) [0, 2) [2, 4) \end{tikzpicture}
The hist plot type can be combined with plot expression as well: provide the usual hexpressioni as
you would for a line plot. Then, configure the value for data=hexpressioni in dependence of x, y, or z:
25
20
15
10
0
0 [0, ·) [0.1, ·) [0.2, ·) [0.3, ·) [0.4, ·) [0.5, ·) [0.6, ·) [0.7, ·) [0.8, ·) [0.9, ·) 1
\begin{tikzpicture}
\begin{axis}[
tiny,
height=4cm,width=12cm,
ybar interval,
ymin=0,
xmin=0,xmax=1,
axis on top,
grid=none,
x tick label as interval=false,
xticklabel=$\pgfmathprintnumber\tick$
},
xticklabel={$[\pgfmathprintnumber[fixed]\tick,\cdot)$}
]
\addplot+[samples=200,hist] {rnd};
\end{axis}
\end{tikzpicture}
The example uses the rnd method of pgf which defines y to contain uniform random numbers in the
range [0, 1]. Then, it configures hist. Note that hist has the default data=y such that it uses the y
coordinate as input. Note furthermore that the x value is e↵ectively ignored here. The options after
\begin{axis}[...] are mainly to scale the graphics and to insert the right limits. The extra x ticks
method is inserted to demonstrate how to add further tick marks without a↵ecting the overall layout.
Note that the extra x tick style sets x tick label as interval=false to disable the special tick
handling which is active for the rest of the plot.
The following keys configure hist. If they are provided inside of hoptionsi, the common key prefix
hist/ can be omitted.
/pgfplots/hist/data={hexpressioni} (initially y)
Tells hist how to get its data. The common idea is to provide a mathematical hexpressioni
which depends on data supplied by the \addplot statement. For example, if you have \addplot
expression, the hexpressioni may depend upon x, y or z. In case of an \addplot table input
routine, the hexpressioni can employ \thisrow{hcolnamei} to access the currently active table row
in the designated column.
It is also possible to avoid invocations of the math parser. Use hist/data value={hvaluei} instead
to do so. Here, hvaluei should be of a numeric constant.
The initial configuration employs what would usually become the final y coordinate as input (to be
more precise, the initial value is data value=\pgfkeysvalueof{/data point/y}).
/pgfplots/hist/data min={hmin valuei} (initially /pgfplots/xmin)

/pgfplots/hist/data max={hmax valuei} (initially /pgfplots/xmax)
Allows to provide the min/max values (the m and m) values manually.
If empty, these values will be deduced from the input data range.
The resulting interval will be splitted into hist/bins intervals.
The initial configuration uses any provided data limits, i.e. the (natural) choices hist/data
min=xmin and hist/data max=xmax.
/pgfplots/hist/bins={hnumber of intervalsi} (initially 10)

Specifies the number of intervals to use.
/pgfplots/hist/intervals={htrue,falsei} (initially true)

If intervals=true (the initial configuration), hist will generate N + 1 coordinates, with
m = x0 < x 1 < · · · < x N = m
where [m, m] is the data range. In this case, the data points for xN 1 and xN will get the same
value, namely the number of elements in the last bin. This is (only) useful in conjunction with
const plot or ybar interval.
If intervals=false, the last data point will be omitted and exactly N coordinates will be gener-
ated. In this case, the right end point is not returned explicitly.
/pgfplots/hist/cumulative={htrue,falsei} (initially false)

Allows to compute a cumulative histogram.
A cumulative histogram uses the sum of all previous bins and the current one as final value.
Here is the example from above, this time with hist/cumulative:
8,000 \begin{tikzpicture}
\begin{axis}[
ybar interval,
6,000 xtick=,% reset from ybar interval
xticklabel=
{$[\pgfmathprintnumber\tick,
4,000 \pgfmathprintnumber\nexttick)$}
]]
% a data file containing 8000 normally distributed
2,000 % random numbers of mean 0 and variance 1
\addplot+[hist={
data=x,
0
cumulative}]
file {plotdata/pgfplots.randn.dat};
[ 4, 2)[ 2, 0) [0, 2) [2, 4)
\end{axis}
\end{tikzpicture}
/pgfplots/hist/density={htrue,falsei} (initially false)

An extension by Jürnjakob Dugge
Enables density estimation mode. If hist/density is active, the resulting data points will be
renormalized such that the overall “mass” equals 1.
hist hist=density
2,500 0.4
2,000
0.3
1,500
0.2
1,000
0.1
500
0 0
4 2 0 2 4 4 2 0 2 4
\begin{tikzpicture}
\begin{axis}[small,ymin=0,title=\texttt{hist}]
\addplot [
hist,
fill=orange!75,
draw=orange!50!black]
table [y index=0] {plotdata/pgfplots.randn.dat};
\end{axis}
\end{tikzpicture}
%
\pgfplotsexpensiveexample
\begin{tikzpicture}
\begin{axis}[small,ymin=0, title=\texttt{hist=density}]
\addplot [
hist=density,
fill=orange!75,
\end{axis}
\end{tikzpicture}
The keys hist/density and hist/cumulative can be combined as well:
hist=cumulative hist={cumulative,density}
8,000 1
0.8
6,000
0.6
4,000
0.4
2,000
0.2
0 0
4 2 0 2 4 4 2 0 2 4
\begin{tikzpicture}
\begin{axis}[small,ymin=0, title=\texttt{hist=cumulative}]
\addplot [
hist=cumulative,
fill=orange!75,
\end{axis}
\end{tikzpicture}
%
\pgfplotsexpensiveexample
\begin{tikzpicture}
\begin{axis}[small,ymin=0, title=\texttt{hist=\{cumulative,density\}}]
\addplot [
hist={cumulative,density},
fill=orange!75,
\end{axis}
\end{tikzpicture}
/pgfplots/hist/handler (style, initially ybar interval)

Allows to change the way the generated coordinates are visualized. The hist/handler key is a
style, so use hist/handler/.style={const plot} to change it.
\begin{tikzpicture}
\begin{axis}
6
data\\
1\\ 2\\ 1\\ 5\\ 4\\ 10\\
4 7\\ 10\\ 9\\ 8\\ 9\\ 9\\
};
\end{axis}
\end{tikzpicture}
2
2 4 6 8 10
\begin{tikzpicture}
\begin{axis}
6
\addplot+[hist={bins=3,
handler/.style={const plot}}
]
4 data\\
1\\ 2\\ 1\\ 5\\ 4\\ 10\\
7\\ 10\\ 9\\ 8\\ 9\\ 9\\
};
2 \end{axis}
\end{tikzpicture}
2 4 6 8 10
\begin{tikzpicture}
\begin{axis}
6
handler/.style={sharp plot}}
]
4 data\\
1\\ 2\\ 1\\ 5\\ 4\\ 10\\
7\\ 10\\ 9\\ 8\\ 9\\ 9\\
};
2 \end{axis}
\end{tikzpicture}
2 4 6 8 10
Note that sharp plot might benefit from intervals=false:
\begin{tikzpicture}
\begin{axis}
6
handler/.style={sharp plot},
intervals=false}
]
data\\
1\\ 2\\ 1\\ 5\\ 4\\ 10\\
7\\ 10\\ 9\\ 8\\ 9\\ 9\\
2 };
\end{axis}
/pgfplots/hist/data filter/.code={h... i}
Allows to define coordinate filters, similar to the coordinate filter key x filter described in Sec-
tion 4.22. The argument #1 is the coordinate as it has been found after processing hist/data. The
code is supposed to assign \pgfmathresult to contain the result. If \pgfmathresult is empty
afterwards, it will be skipped. Otherwise, it is supposed to contain a number.
This filter is applied before the histogram is computed. Note that x filter and y filter are
applied after the histogram is computed.
Note that predefined styles like each nth point can also be applied to hist/data if
1. an asterisk ‘*’ is appended to the predefined style’s name and
2. the first argument to the style is hist/data.
For example, each nth point*={hist/data}{2} will skip each second input value of hist/data
(try it out).
/pgfplots/hist/data coord trafo/.code={h... i}

/pgfplots/hist/data coord inv trafo/.code={h... i}
These keys work in the same way as for x coord trafo and x coord inv trafo. They are applied
to the hist/data value before the histogram is evaluated and after the result value is assigned,
respectively.
Note that hist will apply the hist/data coord inv trafo before it visualizes its results. Conse-
quently, it may be necessary to assign a similar transformation to x coord trafo as well.
See the documentation of x coord trafo for more information about custom transformations.
/pgfplots/hist/symbolic coords={hlisti}
A style which enables symbolic x coords for an axis containing hist plots:
\begin{tikzpicture}
\begin{axis}[
ybar interval,
6 hist/symbolic coords={A,B,C,D,E,F,G,H,I,J},
xticklabel={[\tick--\nexttick[}],
]
data\\
A\\ B\\ A\\ D\\ F\\ J\\
G\\ J\\ I\\ H\\ I\\ I\\
};
2
\end{axis}
\end{tikzpicture}
[A–D[ [D–G[ [G–J[
The style does two things: first, it defines hist/data coord trafo and hist/data coord inv
trafo, then, it calls symbolic x coords with the same argument.
Attention : do not use hist/data=x or other symbolic values as input when you have
symbolic coords. Rather than symbolic values, you need to provide expandable values like
\pgfkeysvalueof{/data point/x} (which has the same e↵ect, but directly expands to the correct
value).
Please refer to the documentation of symbolic x coords for further details about symbolic coor-
dinates.
5.13 Ternary Diagrams

\usepgfplotslibrary{ternary} % LATEX and plain TEX
\usepgfplotslibrary[ternary] % ConTEXt
\usetikzlibrary{pgfplots.ternary} % LATEX and plain TEX
\usetikzlibrary[pgfplots.ternary] % ConTEXt
A library to draw ternary diagrams.
5.13. TERNARY DIAGRAMS 487
A ternary diagram visualizes three–component systems such that the sum of them yields 100%. Ternary
diagrams are triangular axes.
5.13.1 Ternary Axis

\begin{ternaryaxis}[hoptionsi]
\end{ternaryaxis}
The axis environment for ternary axes.
0
100 10°
20 20°
80
40
60
60
40
80
20
100
0
0 20 40 60 80 100
% Preamble: \pgfplotsset{width=7cm,compat=1.13}\usepgfplotslibrary{ternary}
\begin{tikzpicture}
\begin{ternaryaxis}
\addplot3 coordinates {
(0.81, 0.19, 0.00)
(0.76, 0.17, 0.07)
(0.66, 0.16, 0.16)
(0.76, 0.07, 0.17)
(0.81, 0.00, 0.19)
};
(0.85, 0.15, 0.00)
(0.82, 0.13, 0.05)
(0.73, 0.14, 0.13)
(0.82, 0.06, 0.13)
(0.84, 0.00, 0.16)
};
\legend{$10$\textdegree, $20$\textdegree}
\end{ternaryaxis}
\end{tikzpicture}
A ternaryaxis works with relative coordinates: each data point consists of three components x, y, z.
Their sum forms a compound entity which has 100% (of whatever). In the standard configuration, we
have x, y, z 2 [0, 1]. The unit interval is not necessary: you can as well choose absolute data ranges
x 2 [xmin , xmax ], y 2 [ymin , ymax ] and z 2 [zmin , zmax ]. The important thing is that the relative values
x xmin y ymin z zmin
x̃ := , ỹ := , z̃ :=
xmax xmin ymax ymin zmax zmin
sum up to 100%, i.e. x̃ + ỹ + z̃ = 1. Thus, pgfplots computes x̃, ỹ and z̃ and interpretes them as
barycentric (triangular) coordinates.
For this to work, it is crucial to provide xmin, xmax, ymin, ymax and zmin, zmax precisely! The
initial configuration fixes them to the unit interval.
What happens behind the scenes is that a data point (x, y, z) is placed at X, Y determined by
 1
X(x, y, z) x̃ + 2z̃
= x̃A + ỹB + z̃C = 2 p3
Y (x, y, z) 2 x̃
p
where A = (1/2, 3/2) is top corner of the triangle, B = (0, 0) the lower left and C = (1, 0) the lower
right one. The ỹ component is not really necessary due to the linear dependency x̃ + ỹ + z̃ = 1.
The input coordinate (100%, 0%, 0%) is mapped to A, the input coordinate (0%, 100%, 0%) to B and
(0%, 0%, 100%) to C (Acrobat Reader: click into the axis to verify it).
0
100 10°
20 20°
80
B 40 A
60
60
40
Deduced z
80
20
100
0
0 20 40 60 80 100
C
\begin{tikzpicture}
\begin{ternaryaxis}[xlabel=A,ylabel=B,zlabel=C]
(0.81, 0.19, 0.00)
(0.76, 0.17, 0.07)
(0.66, 0.16, 0.16)
(0.76, 0.07, 0.17)
(0.81, 0.00, 0.19)
};
(0.85, 0.15, 0.00)
(0.82, 0.13, 0.05)
(0.73, 0.14, 0.13)
(0.82, 0.06, 0.13)
(0.84, 0.00, 0.16)
};
\node[pin=130:Deduced $z$,draw=black] at (0.2,0.2) {};
\legend{$10$\textdegree, $20$\textdegree}
\end{ternaryaxis}
\end{tikzpicture}
A ternaryaxis can contain zero, one or more \addplot3 commands, just as a usual axis. In case you
provide only two–dimensional coordinates (for example using \addplot or axis cs), the third component
is deduced automatically such that components sum to 100%. The \addplot3 command can use any of the
accepted input formats, for example using coordinates, table, expression or whatever – but the input is
always interpreted as barycentric coordinates (three components summing up to 100%).
Sloped labels and minor ticks \begin{tikzpicture}
0 \begin{ternaryaxis}[
100 title=Sloped labels and minor ticks,
xlabel=Water,
20 ylabel=D--Threonine,
e 80 zlabel=L--Threonine,
ni n
label style={sloped},
40
Wa
r eo
minor tick num=2,

60
t er
Th
]
D–
60 (0.82, 0.18, 0.00)
40
(0.75, 0.17, 0.08)
80 (0.77, 0.12, 0.11)
20 (0.75, 0.08, 0.17)
(0.81, 0.00, 0.19)
100 };
0 \addplot3 coordinates {
0 20 40 60 80 100 (0.75, 0.25, 0.00)
(0.69, 0.25, 0.06)
L–Threonine (0.64, 0.24, 0.12)
(0.655, 0.23, 0.115)
(0.67, 0.17, 0.16)
(0.66, 0.12, 0.22)
(0.64, 0.11, 0.25)
(0.69, 0.05, 0.26)
(0.76, 0.01, 0.23)
};
\end{ternaryaxis}
\end{tikzpicture}
Sloped labels and minor grids \begin{tikzpicture}
100 title=Sloped labels and minor grids,
xlabel=Water,
20 ylabel=D--Threonine,
80
e
zlabel=L--Threonine,
ni n
label style={sloped},
40
Wa
r eo
minor tick num=2,

60
te
Th
grid=both,
r
D–
]
40
(0.82, 0.18, 0.00)
80 (0.75, 0.17, 0.08)
20 (0.77, 0.12, 0.11)
(0.75, 0.08, 0.17)
100 (0.81, 0.00, 0.19)
0 };
0 20 40 60 80 100 \addplot3 coordinates {
(0.75, 0.25, 0.00)
L–Threonine (0.69, 0.25, 0.06)
(0.64, 0.24, 0.12)
(0.655, 0.23, 0.115)
(0.67, 0.17, 0.16)
(0.66, 0.12, 0.22)
(0.64, 0.11, 0.25)
(0.69, 0.05, 0.26)
(0.76, 0.01, 0.23)
};
\end{ternaryaxis}
\end{tikzpicture}
A ternaryaxis supports (most of) the pgfplots axis interface, among them the grid option, the
xtick={hpositionsi} way to provide ticks, including extra x ticks and its variants. Of course, it can
also contain any of the mark, color and cycle list options of a normal axis.
The following example is a (crude) copy of an example of
http://www.sv.vt.edu/classes/MSE2094_NoteBook/96ClassProj/experimental/ternary2.html
and uses area style to change cycle list and the legend appearance.
Want–be–Stainless Steel
0
100 Cr
We
n Cr+ FeNi
ig
20
Ir o
ht
80 FeNi
ent
Pe
Cr+ FeNi
rc
40
rc
ent
Pe
60 + FeNi
ht
Ch
60
ig
rom
We
40
i um
80
20
Stainless Steel
100
0
0 20 40 60 80 100
Weight Percent Nickel
\begin{tikzpicture}
\begin{ternaryaxis}[
title=Want--be--Stainless Steel,
xlabel=Weight Percent Chromium,
ylabel=Weight Percent Iron,
zlabel=Weight Percent Nickel,
label style=sloped,
area style,
]
\addplot3 table {
A B C
1 0 0
0.5 0.4 0.1
0.45 0.52 0.03
0.36 0.6 0.04
0.1 0.9 0
};
\addlegendentry{Cr}
\addplot3 table {
A B C
1 0 0
0.5 0.4 0.1
0.28 0.35 0.37
0.4 0 0.6
};
\addlegendentry{Cr+$\gamma$FeNi}
\addplot3 table {
0.4 0 0.6
0.28 0.35 0.37
0.25 0.6 0.15
0.1 0.9 0
0 1 0
0 0 1
};
\addlegendentry{$\gamma$FeNi}
\addplot3 table {
0.1 0.9 0
0.36 0.6 0.04
0.25 0.6 0.15
};
\addlegendentry{Cr+$\gamma$FeNi}
\addplot3 table {
0.5 0.4 0.1
0.45 0.52 0.03
0.36 0.6 0.04
0.25 0.6 0.15
0.28 0.35 0.37
};
\addlegendentry{$\sigma$+$\gamma$FeNi}
\node[inner sep=0.5pt,circle,draw,fill=white,pin=-15:\footnotesize Stainless Steel]

at (0.18,0.74,0.08) {};
\end{ternaryaxis}
\end{tikzpicture}
Ternary plots can also use contour prepared to plot contour lines. The following example is a (crude)
copy of an example of
http://www.sv.vt.edu/classes/MSE2094_NoteBook/96ClassProj/experimental/ternary2.html:
Want–be–Stainless Steel \begin{tikzpicture}
100 title=Want--be--Stainless Steel,
We
0 xlabel=Weight Percent Chromium,
1 ,7 0
n
ig
Ir o 20 ylabel=Weight Percent Iron,
ht
80
00
zlabel=Weight Percent Nickel,
ent
Pe
1,6
label style=sloped,
rc
40
rc
ent
]
Pe
60
0
50 1 ,5 0 0
50
50
ht
Ch
1 ,6 0 0
1,3
1,
60 % plotdata/pgfplotsternary.example1.dat:
00
ig
1,
1
rom
0 ,4
35
We
0 40 %
% Chromium Iron Nickel Temperature
i um
00
1,
80 6
1, 1 ,3 6 0
% 0.90 0.0 0.10 1700
1,370 20 % 0.85 0.14 0.00 1700
00
%
1 ,4
1,400
1 ,5
100
1,500
% 0.85 0.00 0.15 1600

90
1,420
0 % 0.78 0.22 0.00 1600
0 20 40 60 80 100 % 0.71 0.29 0.00 1600
% ....
Weight Percent Nickel
\addplot3[contour prepared={labels over line},
point meta=\thisrow{Temperature}]
table[x=Chromium,y=Iron,z=Nickel]
{plotdata/pgfplotsternary.example1.dat};
\end{ternaryaxis}
\end{tikzpicture}
The contour prepared={labels over line} installs the display style contour/labels over line and
expects precomputed contour lines from the input stream. Here, the input stream is a table, consisting
of the three relative components for Chromium, Iron and Nickel – and the point meta is set to be the
Temperature column. The contour prepared style uses the (x, y, z) coordinate to plot the data point and
the point meta to determine contour labels (the initial configuration of contour prepared is to use point
meta=z). The output thus allows to use both barycentric coordinates (ternary components) and contour
labels.
/pgfplots/ternary limits relative=true|false (initially true)

/pgfplots/ternary relative limits=true|false (initially true)
Allows to switch tick labels between relative numbers in the range [0, 100] or absolute numbers.
The choice ternary limits relative=true accepts data in any input number range, for example
(x, y, z) 2 [0, 1]3 , or (x, y, z) 2 [0, 100]3 or in any absolute scala of the form xi 2 [xi , xi ] for xi 2 {x, y, z}
(remember that it is crucial to communicate these limits to pgfplots explicitly using xmin, xmax,
ymin, ymax and zmin, zmax such that relative coordinates can be computed, see the description above
for details). In every case, relative tick labels are drawn, i.e. tick labels in the range [0, 100].
Data range [0, 1], limits relative \begin{tikzpicture}
100 ternary limits relative,
title={Data range $[0,1]$, limits relative},
20 area style]
F 43 (0.2,0.8,0)
40 (0.31,0.4,0.29)
60 (0.34,0.2,0.46)
F 42
(0.4,0,0.6)
60 (1,0,0)
40 };
20 (0.4,0,0.6)
(0.34,0.2,0.46)
100 (0.31,0.4,0.29)
0 (0.14,0.46,0.4)
0 20 40 60 80 100 (0,0.37,0.63)
(0,0,1)
};
\node[fill=white]
at (axis cs:0.56,0.28,0.16) {$F 42$};
\node[fill=white]
at (0.7,0.2) {$F 43$};
\end{ternaryaxis}
\end{tikzpicture}
Data range x 2 [0, 500], y 2 [1, 2], z 2 [0, 1] limits relative

0
100
20
80
40
60
F 42
60
40
80
20
100
0 F 43
0 20 40 60 80 100
\begin{tikzpicture}
xmax=500,ymin=1,ymax=2,
ternary limits relative,
title={Data range $x\in[0,500]$,
$y\in[1,2]$, $z\in[0,1]$ limits relative},
area style]
(100,1.8,0)
(155,1.4,0.29)
(170,1.2,0.46)
(200,1,0.6)
(500,1,0)
};
(200,1,0.6)
(170,1.2,0.46)
(155,1.4,0.29)
(70,1.46,0.4)
(0,1.37,0.63)
(0,1,1)
};
\node[fill=white]
at (axis cs:280,1.28,0.16) {$F 42$};
\node[fill=white]
at (0.7,0.2) {$F 43$};
\end{ternaryaxis}
\end{tikzpicture}
The choice ternary limits relative=false accepts the same data ranges, but it draws tick labels in
the very same data ranges.
Data range x 2 [0, 500], y 2 [1, 2], z 2 [0, 1] limits absolute

1
500
1.2
400
1.4
300
1.6
F 42
200
1.8
100
2
0 F 43
0 0.2 0.4 0.6 0.8 1
\begin{tikzpicture}
ternary limits relative=false,
xmax=500,ymin=1,ymax=2,
title={Data range $x\in[0,500]$,
$y\in[1,2]$, $z\in[0,1]$ limits absolute},
footnotesize, % just for the sake of demonstration...
area style]
(100,1.8,0)
(155,1.4,0.29)
(170,1.2,0.46)
(200,1,0.6)
(500,1,0)
};
(200,1,0.6)
(170,1.2,0.46)
(155,1.4,0.29)
(70,1.46,0.4)
(0,1.37,0.63)
(0,1,1)
};
\node[fill=white]
at (axis cs:280,1.28,0.16) {$F 42$};
\node[fill=white]
at (0.7,0.2) {$F 43$};
\end{ternaryaxis}
\end{tikzpicture}
Coordinate system cartesian cs

A coordinate system which allows Cartesian coordinates. The lower left pointp
has coordinate (0, 0), the
lower right point has (1, 0) and the upper point of the triangle is at (1/2, 3/2).
If you use the standard point syntax (x, y) in path commands inside of the axis, you’ll get Cartesian
coordinates. If you want to use it for axis descriptions (like xlabel), you’ll have to write cartesian
cs:0,0 explicitly (axis labels have the default coordinate system axis description cs).
Cartesian Annotations \begin{tikzpicture}
z(1, 0)
100 title=Cartesian Annotations,
clip=false]
20
(0.1,0.5,0.4)
40
60 (0.2,0.5,0.3)
p (0.3,0.6,0.1)
x( 12 , 23 ) 60 };
40
\node[fill=white,draw] at (0,0) {$y (0,0)$};
80 \node[fill=white,draw] at (1,0) {$z (1,0)$};
20 \node[fill=white,draw] at (0.5,{sqrt(3)/2})
{$x (\frac12,\frac{\sqrt3}{2})$};
100
y(0, 0)
0 \draw[red,-stealth] (0.5,0) -- (0.5,0.7);
0 20 40 60 80 100 \end{ternaryaxis}
\end{tikzpicture}
/pgfplots/every ternary axis (style, no value)

A style which is installed at the beginning of every ternary axis. It is used to adjust some of the
pgfplots keys to fit the triangular shape.
\pgfplotsset{
every ternary axis/.style={
tick align=outside,
grid=major,
xticklabel style={anchor=west},
every 3d description/.style={},
every axis x label/.style={at={(ticklabel cs:0.5)},anchor=near ticklabel},
every axis y label/.style={at={(ticklabel cs:0.5)},anchor=near ticklabel},
every axis z label/.style={at={(ticklabel cs:0.5)},anchor=near ticklabel},
every x tick scale label/.style=
{at={(xticklabel cs:0.95,5pt)},anchor=near xticklabel,inner sep=0pt},
every y tick scale label/.style=
{at={(yticklabel cs:0.95,5pt)},anchor=near yticklabel,inner sep=0pt},
every z tick scale label/.style=
{at={(yticklabel cs:0.95,5pt)},anchor=near yticklabel,inner sep=0pt},
every axis legend/.style={
cells={anchor=center},
shape=rectangle,
fill=white,
draw=black,
at={(1.03,1.03)},
anchor=north west,
},
annot/point format 3d/.initial={(\%.2f, \%.2f, \%.2f)},
},
}
5.13.2 Tieline Plots

/tikz/tieline={hoptions with tieline/ prefix i}
\addplot+[tieline={hoptions with tieline/ prefix i}]

A plot handler for use in ternary diagrams which plots tie lines and binodal curves.
(i) (i) (i) (1) (2) (3)
On input, it accepts pairs of coordinates, A(i) = (Ax , Ay , Az ) and B (i) = (Bx , By , Bz ), for
i = 1, . . . , N (i.e. it requires a total of six coordinates, perhaps plus additional color data).
On output, it connects the pairs, i.e. for every fixed i = 1, . . . , N , it connects A(i) — B (i) (the so-called
“tie lines”). In addition, it also draws the binodal curve, which is made up by connecting all A(i) and
then, in reverse ordering, all B (i) : A(1) — A(2) — · · · — A(N ) — B (N ) — B (N 1) — · · · — B (1) :
0
20 xlabel=x (IPA),
80 ylabel=y (water),
y (water) zlabel=z (propene),
40 x (IPA) axis on top,
60 ]
% plotdata/ternary_data.txt is a table of the form
60 %A_propene A_water A_IPA B_propene B_water B_IPA
40 % 0.0009 0.9990 0 0.9333 0.0667 0
% 0.0009 0.9988 0.0002 0.9303 0.0665 0.0032
80 % 0.0011 0.9975 0.0013 0.9135 0.0673 0.0191
20 % 0.0013 0.9962 0.0024 0.8956 0.0693 0.0351
100 %...
0 \addplot3[tieline,fill=blue!10]
table [x=A_IPA,y=A_water,z=A_propene]
0 20 40 60 80 100
{plotdata/ternary_data.txt};
z (propene) \end{ternaryaxis}
\end{tikzpicture}
We see that each input line has six columns, and each six columns are taken into account (this is di↵erent
from other plot handlers!). The six columns make up the three components of the A and B points,
respectively. In the example above, we used explicit column names and provided Ax using x=A_IPA, Ay
using y=A_water and Az using z=A_propene. Note that these keys are the common input method for
\addplot table; they are nothing special (that means we could also use x index instead). The three
columns for B can be provided manually (see below), or deduced automatically: in our case, the value
for Bx has been found in the third column after x=A_IPA (which is B_IPA); the value for By has been
found in the third column after y=A_water and Bz is made up from the third column after z=A_propene.
In other words, the B value is searched (by default) by adding 3 to the column index of the respective
A coordinate.
You do not need to provide any column names; in this case, the first three columns make up A (in the
order of appearance) and the following three make up B.
The only supported input type for tieline plots is table input. It is optimized to use \addplot3 table
(as described above). To use the two–dimensional variant \addplot table, you need to tell pgfplots
explicitly which columns make up Ax , Ay , Bx , By ; the z coordinates are deduced automatically such
that the result sums to 100%.
/pgfplots/table/tie end x={hcolnamei} (initially empty)
/pgfplots/table/tie end y={hcolnamei} (initially empty)
/pgfplots/table/tie end z={hcolnamei} (initially empty)
/pgfplots/table/tie end x index={hcol index i} (initially empty)
/pgfplots/table/tie end y index={hcol index i} (initially empty)
/pgfplots/table/tie end z index={hcol index i} (initially empty)
These keys can be used to provide column names or column indices for Bx , By and Bz , respectively.
They can be provided like
\addplot3[tieline] table[tie end y=B_water] ....
Note that the tie end x keys are only available if the tieline option has been used before.
The values for A are provided with table/x, table/x index and its variants as for any other plot
type.
The tieline plot handler accepts several options to customize the appearance. You can provide them as
argument after tieline, using tieline={hoptionsi}. In this case, the tieline/ prefix can be omitted.
The keys are described in the following:
/pgfplots/tieline/each nth tie={hnumber i} (initially empty)
Allows to draw only each nth tie line, even though the binodal curve uses all provided coordinates:
0
20 xlabel=x (IPA),
60 ]
40 % 0.0009 0.9990 0 0.9333 0.0667 0
% 0.0009 0.9988 0.0002 0.9303 0.0665 0.0032
80 % 0.0011 0.9975 0.0013 0.9135 0.0673 0.0191
20 % 0.0013 0.9962 0.0024 0.8956 0.0693 0.0351
100 %...
0 \addplot3[
tieline={each nth tie=5},
0 20 40 60 80 100
fill=blue!10,
z (propene) ]
\end{ternaryaxis}
\end{tikzpicture}
Note that plot marks (if any) are drawn on every input position, use the mark repeat option to
change that.
/pgfplots/tieline/tieline style={hoptionsi}
Appends hoptionsi to the style tieline/every tieline.
Useful hoptionsi are, for example, other plot handlers to adjust the appearance of tie lines. Suppose
that you have additional color data for every tie line (which might have been provided as further
input column). In our case, we provide random color data using point meta=rand, and visualize
the single tielines as with contour prepared:
0
20 xlabel=x (IPA),
60 ]
40 % 0.0009 0.9990 0 0.9333 0.0667 0
% 0.0009 0.9988 0.0002 0.9303 0.0665 0.0032
80 % 0.0011 0.9975 0.0013 0.9135 0.0673 0.0191
20 % 0.0013 0.9962 0.0024 0.8956 0.0693 0.0351
100 %...
0 \addplot3[
point meta=rand,
0 20 40 60 80 100
tieline={
z (propene) each nth tie=8,
tieline style={contour prepared}
},
fill=blue!10,
]
\end{ternaryaxis}
\end{tikzpicture}
The e↵ect here is that contour labels and line colors are chosen for every tie line, where the actual
color is determined using point meta and colormap. Other choices for plot handlers in tieline
style might be the mesh.
/pgfplots/tieline/curve style={hoptionsi}
Appends hoptionsi to the style tieline/every curve.
The curve style allows to customize the plot handler for the curve. A possible choice might be
curve style={smooth} or a separate fill/draw color.
5.14 Units in Labels

by Nick Papior Andersen
\usepgfplotslibrary{units} % LATEX and plain TEX

\usepgfplotslibrary[units] % ConTEXt
\usetikzlibrary{pgfplots.units} % LATEX and plain TEX
\usetikzlibrary[pgfplots.units] % ConTEXt
A library which allows to use automatic typesetting of units in labels. The library utilizes di↵erent keys
to typeset the final output in a consistent way. Calling one of the commands automatically sets the key
‘use units=true’ so one does not have to worry about this.
pgfplots has the capability of supporting units. This provides quick customization of the plot as well as
the addition of units in labels.
Loading the library automatically enables the typesetting of units in labels. Currently it only supports
predefined SI units but a per-user customization is also implemented such that it can be used in any way
you like.
First the key which enables you to switch on/o↵ the unit system.
/pgfplots/use units={hbooleani} (initially true)
This key simply enables pgfplots to use what is described next. This key will be set to true if you
load the library. You can use this to temporarily determine whether the unit library should be used in
plots.
5.14. UNITS IN LABELS 499
/pgfplots/x unit={huniti} (initially empty)

/pgfplots/y unit={huniti} (initially empty)
/pgfplots/z unit={huniti} (initially empty)
These keys set the unit in their respective axis. In SI units you could for instance set the x unit in
Newton as x unit=N.
/pgfplots/x unit prefix={hprefix i} (initially empty)

/pgfplots/y unit prefix={hprefix i} (initially empty)
/pgfplots/z unit prefix={hprefix i} (initially empty)
These keys set the prefix of the unit. If a value on the y axis is in kilo you would set the y unit
prefix=k. Prefix will be typeset in front of the unit.
This command will not intervene with the basis of the axis system. I.e. a prefix as just mentioned will
not divide every y axis number by 1000. In order to do this, see key haxisi SI prefix, see Section
5.14.1.
Notice that if the haxisi unit isn’t set the entire unit will not be typeset.
Remarks: Remember that all typesetting of labels occur within math mode (i.e. within $$ delimiters).
Therefore one can use \frac and other mathematics commands.
Often one just has to utilize the above mentioned keys. It is the basis of the unit typesetting system provided
by pgfplots.
% Preamble: \pgfplotsset{width=7cm,compat=1.13}\usepgfplotslibrary{units}
\begin{tikzpicture}
2.5 \begin{axis}[use units,
x unit=m,x unit prefix=k,
y unit=N,y unit prefix=m,
Force [mN]
xlabel=Distance,ylabel=Force]
(1,2.3)
(2,2.7)
1.5 (3,2.1)
(4,1.8)
(5,1.5)
(6,1.1)
1 };
2 4 6 \end{axis}
Distance [km] \end{tikzpicture}
Below is an example of what would be obtained according to the styles

% x label becomes ‘‘Temperature [T]’’, y label becomes ‘‘Nothing’’
\pgfplotsset{use units,x unit=T,xlabel=Temperature,ylabel=Nothing}
% x label becomes ‘‘Temperature’’, y label becomes ‘‘Nothing’’
\pgfplotsset{use units,x unit prefix=m,xlabel=Temperature,ylabel=Nothing}
Notice the second example. Only setting the prefix will not activate the unit typesetting. Therefore one
should ensure to use the x unit key if the typesetting of the labels should be done.
For typesetting the units one can also change the appearance. For instance one might not like the square
brackets which surround the unit. These can luckily be changed using the below keys.
/pgfplots/unit marking pre={hprei} (initially \left[)
/pgfplots/unit marking post={hposti} (initially \right])
/pgfplots/unit markings=parenthesis|square brackets|slash space (initially square brackets)
⇥1⇤
These keys set the surroundings of the unit. The initial yields 2 such that you can typeset fractions
in units. Be aware that you can only obtain large fractions if you use \dfrac. These can easily be set
using the option key unit markings where the options typesets as the following
\pgfplotsset{x unit=T,unit markings=parenthesis} % x unit becomes ‘‘ \left(T\right)’’

\pgfplotsset{x unit=T,unit markings=square brackets} % x unit becomes ‘‘ \left[T\right]’’
\pgfplotsset{x unit=T,unit markings=slash space} % x unit becomes ‘‘ / T’’
Notice that all typesetting of units first inserts a space and then the unit marking pre code.
Of course you can just manually set each of them with the unit marking pre and unit marking post
keys. Just remember that they are typeset within a $$.
One will typically typeset the unit with a specific font. To do so an option of changing the typesetting
command is supplied.
/pgfplots/unit code/.code 2 args={h... i}
This can be utilized to great extent. By default, units are typeset as \mathrm{hunit prefix ihuniti}. But
if one for instance wishes to utilize the package siunitx, which has great capabilities in typesetting
both units, numbers and angles, one can just set the key as
\pgfplotsset{unit code/.code 2 args={\si{#1#2}}}
which would yield the unit as \si{hunit prefix ihuniti}.

The first argument is typeset as hunit prefix i and the second argument is huniti.
The most important thing is that the command needs exactly two arguments. So if you would like a
command that typesets the prefix in bold face and the unit in normal roman font you should call
\pgfplotsset{unit code/.code 2 args={\mathbf{#1}\mathrm{#2}}
5.14.1 Preset SI prefixes

To support the SI system a number of preset keys are defined. This should yield a more intuitive way
of supplying the prefix as well as add some more functionality. For instance it provides an easy scaling
mechanism.
/pgfplots/x SI prefix=yocto|...|milli|centi|deci|deca|hecto|kilo|...|yotta (initially none)
/pgfplots/y SI prefix=yocto|...|milli|centi|deci|deca|hecto|kilo|...|yotta (initially none)
/pgfplots/z SI prefix=yocto|...|milli|centi|deci|deca|hecto|kilo|...|yotta (initially none)
/pgfplots/change x base=true|false (initially false)
/pgfplots/change y base=true|false (initially false)
/pgfplots/change z base=true|false (initially false)
These keys sets the prefix of the unit. The allowed prefixes are:
Prefix Power Prefix Power

yocto 24 deca 1
zepto 21 hecto 2
atto 18 kilo 3
femto 15 mega 6
pico 12 giga 9
nano 9 tera 12
micro 6 peta 15
milli 3 exa 18
centi 2 zetta 21
deci 1 yotta 24
As well as resetting the base of the axis if the key change haxisi base=true. Just remember to set
the change haxisi base before using the haxisi SI prefix key.
See the utilization as in the example below.
5.14. UNITS IN LABELS 501
% Preamble: \pgfplotsset{width=7cm,compat=1.13}\usepgfplotslibrary{units}
\begin{axis}[change x base,
x SI prefix=kilo,x unit=m,
y SI prefix=milli,y unit=N,
Force [mN]
1.2 xlabel=Distance,ylabel=Force]
(1000,1)
1.1 (2000,1.1)
(3000,1.2)
(4000,1.3)
};
1 \end{axis}
\end{tikzpicture}
1 2 3 4
Distance [km]
Notice that the x axis has changed base without displaying the ·103 . This is done by using the key
change x base. Even though you have used the key y SI prefix=milli the base isn’t changed on the
y axis. Try adding change y base just after change x base and see the result!
The above keys are the easy implementation of the base change. Below is a further customization of the
base change. It makes it easy to implement a prefix with a custom base change.
/pgfplots/axis base prefix=axis {haxisi} base {hbasei} prefix {hprefix i} (initially empty)
One can utilize this key to customize further of the base and setting the prefix.
\pgfplotsset{change x base,axis base prefix={axis x base -3 prefix k}}

\pgfplotsset{change x base,x SI prefix=kilo}
The above two commands are thus equivalent. Remember that the base should operate in opposite of
prefix!
Chapter 6
Memory and Speed considerations
6.1 Memory Limits of TEX

pgfplots can typeset plots with several thousand points if memory limits of TEX are configured properly.
Its runtime is roughly proportional to the number of input points1 .
Scatter plot with 2250 points \begin{tikzpicture}
·104 \begin{axis}[
3 enlargelimits=0.01,
title style={yshift=5pt},
title=Scatter plot with $2250$ points]
2 \addplot[blue,
mark=*,only marks,mark options={scale=0.3}]
file[skip first]
1 \end{axis}
\end{tikzpicture}
0
0 1 2 3 4 5
6
·10
Ornstein-Uhlenbeck sample (13000 time steps) \begin{tikzpicture}
\begin{axis}[
enlarge x limits=0.03,
title=Ornstein-Uhlenbeck sample
1.5 ($13000$ time steps),
xlabel=$t$]
1 \addplot[blue] file {plotdata/ou.dat};

\end{axis}
\end{tikzpicture}
0.5
0
0 1 2 3 4 5
t
1 In fact, the runtime is pseudo–linear: starting with about 100,000 points, it will become quadratic. This limitation applies
to the path length of pgf paths as well. Furthermore, the linear runtime is not possible yet for stacked plots.
502
6.2. MEMORY LIMITATIONS 503
80 ⇥ 80 Smooth Surface % huger graphs are possible; consider lualatex
\begin{tikzpicture}
\begin{axis}[
title=$80 \times 80$ Smooth Surface,
xlabel=$x$,
2 ylabel=$y$]
\addplot3[surf,samples=80,shader=interp,domain=0:1]
{sin(deg(8*pi*x))* exp(-20*(y-0.5)^2)
+ exp(-(x-0.5)^2*30
- (y-0.25)^2 - (x-0.5)*(y-0.25))};
0 \end{axis}
1 \end{tikzpicture}
0 0.2 0.4 0.5

0.6 0.8 y
1 0
x
pgfplots relies completely on TEX to do all typesetting. It uses the front-end-layer and basic layer of
pgf to perform all drawing operations. For complicated plots, this may take some time, and you may want
to read Section 7 for how to write single figures to external graphics files. Externalization is the best way to
reduce typesetting time.
However, for large scale plots with a lot of points, limitations of TEX’s capacities are reached easily.
6.2 Memory Limitations

The default settings of most TEX-distributions are quite restrictive, so it may be necessary to adjust them.
Usually, the log–file or the final error message contains a summary about the used resources, giving a
hint which parameter needs to be increased.
6.2.1 LuaLaTEX
One solution which works quite well is to switch the LaTEX executable: if you have a decent TEX distribution,
you will have the lualatex executable as well. This, in turn, uses dynamic memory allocation such that it
usually has enough memory for any pgfplots axis.
The LuaLaTEX executable lualatex is supposed to be almost compatible with pdflatex.
This approach works for any platform.
6.2.2 MikTEX
If you are running MikTEX and you do not want to (or cannot switch) to lualatex, you can proceed as
follows.
For MikTEX, memory limits can be increased in two ways. The first is to use command line switches:
pdflatex
--stack-size=n --save-size=n
--main-memory=n --extra-mem-top=n --extra-mem-bot=n
--pool-size=n --max-strings=n
Experiment with these settings if MikTEX runs out of memory. Usually, one doesn’t invoke pdflatex
manually: there is a development aid which does all the invocations, so this one needs to be adjusted.
Sometimes it might be better to adjust the MikTEX configuration file permanently, for example to avoid
reconfiguring the TEX development program. This can be implemented using the command
initexmf --edit-config-file=pdflatex
which can be typed either on a command prompt in Windows or using Start Execute. As a result, an
editor will be opened with the correct config file. A sample config file could be
main_memory=90000000
save_size=80000
or any of the config file entries which are listed below can be entered. Thanks to “LeSpocky” for his
documentation in
http://blog.antiblau.de/2009/04/21/speicherlimits-von-miktex-erhoehen.
504 CHAPTER 6. MEMORY AND SPEED CONSIDERATIONS
6.2.3 TEXLive or similar installations

In addition to the option to switch to lualatex, you can proceed as follows to keep existing dvips or
pdflatex workflows.
For Unix installations, one needs to adjust config files. This can be done as follows:
1. Locate texmf.cnf on your system. On my Ubuntu installation, it is in
/usr/share/texmf/web2c/texmf.cnf.
2. Either change texmf.cnf directly, or copy it to some convenient place. If you copy it, here is how to
proceed:
• keep only the changed entries in your local copy to reduce conflicts. TEX will always read all
config files found in its search path.
• Adjust the search path to find your local copy. This can be done using the environment variable
TEXMFCNF. Assuming your local copy is in ~/texmf/mytexcnf/texmf.cnf, you can write
export TEXMFCNF=~/texmf/mytexcnf:
to search first in your directory, then in all other system directories.

3. You should change the entries
main_memory = n
extra_mem_top = n
extra_mem_bot = n
max_strings = n
param_size = n
save_size = n
stack_size = n
The log–file usually contains information about the parameter which needs to be enlarged.
An example of this config file thing is shown below. It changes memory limits.
1. Create the file ~/texmf/mytexcnf/texmf.cnf (and possibly the paths as well).
% newly created file ~/texmf/mytexcnf/texmf.cnf:

% If you want to change some of these sizes only for a certain TeX
% variant, the usual dot notation works, e.g.,
% main_memory.hugetex = 20000000
main_memory = 230000000 % words of inimemory available; also applies to inimf&mp
extra_mem_top = 10000000 % extra high memory for chars, tokens, etc.
extra_mem_bot = 10000000 % extra low memory for boxes, glue, breakpoints, etc.
save_size = 150000 % for saving values outside current group
stack_size = 150000 % simultaneous input sources
% Max number of characters in all strings, including all error messages,

% help texts, font names, control sequences. These values apply to TeX and MP.
%pool_size = 1250000
% Minimum pool space after TeX/MP’s own strings; must be at least
% 25000 less than pool_size, but doesn’t need to be nearly that large.
%string_vacancies = 90000
% Maximum number of strings.
%max_strings = 100000
% min pool space left after loading .fmt
%pool_free = 47500
2. Run texhash such that TEX updates its ~/texmf/ls-R database.

3. Create the environment variable TEXMFCNF and assign the value ‘~/texmf/mytexcnf:’ (including the
trailing ‘:’ !). For my linux system, this can be done using by adding
export TEXMFCNF=~/texmf/mytexcnf:
to ~/.bashrc.
Unfortunately, TEX does not allow arbitrary memory limits, there is an upper bound hard coded in the
executables.
6.3. REDUCING TYPESETTING TIME 505
6.3 Reducing Typesetting Time

pgfplots does a lot of computations ranging from abstract coordinate computations to low level .pdf
drawing commands (implemented by pgf). For complex plots, this may take a considerable time – especially
for 3D plots.
6.3.1 LUA
If you use compat=1.12 (or newer) and compile your documents by means of lualatex, pgfplots activates
its lua backend. This switch reduces the time to generate output files, especially for 3D plots.
/pgfplots/lua backend=true|false (initially true)

If lua backend is active and the document is processed by means of lualatex, pgfplots activates
scalability and performance improvements which result in the same output as without it, but with less
time and with a smarter memory management.
The feature relies on a partial reimplementation of pgfplots in the fast scripting language lua. In
order to benefit from it, you need to
1. write compat=1.12 (or newer) into your preamble and
2. use lualatex to translate your .tex files (or at least those which contain pgfplots figures).
The time to generate surface plots with this feature can be reduced to 25%–50% of the time required
by the pure TEX implementation (i.e. pdflatex). Other plot types will also benefit from the feature.
The lua backend works in two ways: first, it substitutes isolated routines by faster lua pendants.
This is relatively generic. Second, it replaces entire processing steps by an equivalent lua implemen-
tation.2 pgfplots 1.12 comes with lua implementations for important and long–running operations.
This implementation will be used whenever possible. However, it only covers parts of the feature set:
some features of pgfplots are unsupported by the lua backend. In this case, it will be switched o↵
automatically and the TEX implementation will be used as fallback.
The lua backend will be extended in order to gain more performance improvements and in order to
cover more features. Please take a look at lua debug if you want to see if lua backend has been
(de)activated for your figure.
Note that the lua backend uses a di↵erent math engine with a higher accuracy. As a consequence,
the appearance of a plot might have insignificant pixel di↵erences compared to the output generated by
pdflatex. It is generally recommended to stick with one way to generate a document, i.e. to use either
lualatex or pdflatex. Migrating from one to another might change the appearance of your document
due to lua backend.
Eventually, pgfplots might add computationally expensive features which can be implemented as part
of lua backend and which will be unavailable without it.
Note that lua backend requires a TEX distribution which supports at least lua 5.2 (like TEX Live
2014). It will be deactivated automatically if your version of lualatex is shipped with an older lua
interpreter.
Here are some guidelines how to benefit from lua backend:
• Math expressions which involve macro definitions are unavailable in lua backend. Best-practise:
prefer declare function over macro constants:
2 An example for the first way is: every time pgfplots makes a lookup in its colormap it transfers control over to the
luabackend and the result is immediately communicated back to TEX. The main control flow resides in the slow TEX imple-
mentation. An example for the second way is \addplot expression: pgfplots will copy the math expression and any related
input arguments (including samples and domain) over to the lua backend. Then, the lua backend will apply all loops and
collect coordinates which are finally handed over to the (TEX) implementation of pgf. Thus, the control flow resides in the fast
lua interpreter.
506 CHAPTER 6. MEMORY AND SPEED CONSIDERATIONS
Do \begin{tikzpicture}
\begin{axis}[small,title=Do,samples=2]
20 \addplot+[declare function={C=4;}] {C*x};
\end{axis}
10
20
6 4 2 0 2 4 6
Don’t \begin{tikzpicture}
\begin{axis}[small,title=Don’t,samples=2]
20 \def\constantC{4}
\addplot {\constantC*x};
10 \end{axis}
\end{tikzpicture}
10
20
6 4 2 0 2 4 6
Both are semantically equivalent, but since lua cannot interprete TEX macros, it refuses to process
the “Don’t” case. The “Don’t” case results in a log message
Package pgfplots info on input line 16: Deactivating LUA version of plot expression
for plot 0 (type ’pgfplothandlerlineto’): y expression ’\constantC *x’ contains
a TeX macro.
• Any operation which requires “native” TEX code is unavailable in lua backend. This in-
cludes x filter/.code={...} since /.code cannot be mapped to lua. Best-practise: prefer
x filter/.expression over x filter/.code:
Do \begin{tikzpicture}
\begin{axis}[small,grid=major,title=Do]
4 \addplot+[
3.5 y filter/.expression={y==3 ? nan : y},
]
table {
3 x y
1 2
2.5 1.5 2.5
2 3
2.5 3.5
2 3 4
};
1 1.5 2 2.5 3
\end{axis}
\end{tikzpicture}
• Plot expression can be processed entirely in lua, all other input coordinate types use TEX to read
the value and hand-over to lua.
• There are a couple of operations for which the lua backend is planned for future releases.
/pgfplots/lua debug (no value)

6.3. REDUCING TYPESETTING TIME 507
/pgfplots/lua debug=false|off|off and silent|verbose|compileerror

Typically, the lua backend works silently: pgfplots decides if it is active and if it supports the current
operation and proceeds accordingly. If both conditions are satisfied, it will transfer control to the lua
backend. If one of them is not met, it will automatically fall back to the old TEX implementation. As
a consequence, it cannot be sensed right-away, only the time to compile pictures will vary.
This key controls how pgfplots handles the case where the lua backend is active, but cannot be used
because some encountered feature is unsupported.
The choices false and off deactivate lua debugging. In this case, pgfplots will only write log
messages if it used the TEX implementation instead of the lua backend. In addition, it will only write
the messages into the .log file (but not into the console output).
The choice off and silent writes neither information messages nor debug messages: it is the same
as off but without the log message if the TEX implementation has been used. It is the least verbose
choice.
The choice verbose will also write success messages into the .log file (not into the console).
The choice compileerror will abort compilation if the lua backend is active but does not support the
current figure or plot.
Note that pgfplots might come with limited expensive operations which are only available in the lua
backend. These items will result in compile errors if the lua backend cannot be used for some reason.
6.3.2 Compiling Images Just Once

One possibility to reduce typesetting time is to tell pgf to generate single, temporary .pdf (or .eps)
documents for a subset (or all) graphics in one run and re-use these temporary images in successive runs.
For pgfplots, this is the most e↵ective way to reduce typesetting time for larger documents. It can be
accomplished using the external library described in Section 7.1.
Chapter 7
Import/Export From Other Formats
This section contains information of how to single pictures into separate pdf graphics files (or eps graphics
files). Furthermore, it explains a matlab (®) script which allows to convert from matlab to pgfplots.
7.1 Export to pdf/eps

It is possible to export images to single pdf-documents using routines of pgf and/or Tik Z.
7.1.1 Using the Automatic Externalization Framework of Tik Z

\usepgfplotslibrary{external} % LATEX and plain TEX
\usepgfplotslibrary[external] % ConTEXt
\usetikzlibrary{pgfplots.external} % LATEX and plain TEX
\usetikzlibrary[pgfplots.external] % ConTEXt
The external library o↵ers a convenient method to export every single tikzpicture into a sepa-
rate .pdf (or .eps). Later runs of LATEX will simply include these graphics, thereby reducing typesetting
time considerably.
The library can also be used to submit documents to authors who do not even have pgfplots or Tik Z
installed.
Technical foreword: The external library has been written by Christian Feuersänger (author of
pgfplots). It has been contributed to Tik Z as general purpose library, so the reference documen-
tation along with all tweaks can be found in [6, Section “Externalization Library”]. The command
\usepgfplotslibrary{external} is actually just a wrapper which loads \usetikzlibrary{external}
or, if this library does not yet exist because the installed pgf has at most version 2.00, it will load a
copy which is shipped with pgfplots.
The external library has been designed such that no changes to the document as such are necessary.
The idea is as follows:
1. Every \begin{tikzpicture} . . . \end{tikzpicture} gets a file name. The file name can be
assigned manually with \tikzsetnextfilename{houtput file namei} or automatically, in which
case htex file namei-figurehnumber i is used with an increasing hnumber i.
2. The library writes the resulting images using system calls of the form pdflatex --jobname houtput
file namei automatically, using the write18 system call of TEX. It is the same framework which
can be used to call gnuplot.
The only steps which are necessary is to use
\tikzexternalize
somewhere in your document’s preamble (see below for system-dependent configuration options). No
further modification to the document is necessary. Suppose we have a file called test.tex:
508
7.1. EXPORT TO PDF/EPS 509
\tikzexternalize% activate externalization!
\begin{document}
\begin{figure}
\begin{tikzpicture}
\begin{axis}
\addplot {x^2};
\end{axis}
\end{tikzpicture}
\caption{Our first external graphics example}
\end{figure}
\begin{figure}
\begin{tikzpicture}
\begin{axis}
\addplot {x^3};
\end{axis}
\end{tikzpicture}
\caption{A second graphics}
\end{figure}
\end{document}
To enable the system calls, we type
pdflatex -shell-escape test
and LATEX will now generate the required graphics files test-figure0.pdf and test-figure1.pdf
automatically. Any further call to pdflatex will simply use \includegraphics and the tikzpictures
as such are no longer considered (you need a di↵erent command line switch for MikTEX, see the shell
escape option).
If a figure shall be remade, one can simply delete all or selected graphics files and regenerate them. Al-
ternatively, one can use the command \tikzset{external/force remake} somewhere in the document
to remake every following picture automatically.
There are three ways to modify the file names of externalized figures:
• Changing the overall file name using a prefix,
• Changing the file name for a single figure using \tikzsetnextfilename,
• Changing the file name for a restricted set of figures using figure name.
/tikz/external/prefix={hfile name prefix i} (initially empty)

A shortcut for \tikzsetexternalprefix{hfile name prefix i}, see below.
\tikzsetexternalprefix{hfile name prefix i}

Assigns a common prefix used by all file names. For example,
\tikzsetexternalprefix{figures/}
will prepend figures/ to every external graphics file name.
\tikzsetnextfilename{hfile namei}
Sets the file name for the next Tik Z picture or \tikz short command. It will only be used for the
next picture.
Pictures for which no explicit file name has been set will get automatically generated file names.
Please note that prefix will still be prepended to hfile namei.
510 CHAPTER 7. IMPORT/EXPORT FROM OTHER FORMATS
% main document, called main.tex
\usepackage{tikz}
\tikzexternalize[prefix=figures/]% activate with a name prefix
\begin{document}
\tikzsetnextfilename{firstplot}
\begin{tikzpicture} % will be written to ’figures/firstplot.pdf’
\begin{axis}
\addplot {x};
\end{axis}
\end{tikzpicture}
\begin{tikzpicture} % will be written to ’figures/main-figure0.pdf’

\draw[help lines] (0,0) grid (5,5);
\end{tikzpicture}
\end{document}
pdflatex -shell-escape main
/tikz/external/figure name={hnamei}
Same as \tikzsetfigurename{hnamei}.
\tikzsetfigurename{hnamei}
Changes the names of all following figures. It is possible to change figure name during the doc-
ument using \tikzset{external/figure name={hnamei}}. A unique counter1 will be used for
each di↵erent hnamei, and each counter will start at 0.
The value of prefix will be applied after figure name has been evaluated.
% main document, called main.tex
\usepackage{tikz}
\tikzexternalize% activate externalization!
\begin{document}
% will be written to ’main-figure0.pdf’

\begin{tikzpicture}
\begin{semilogyaxis}
\addplot {exp(x)};
\end{semilogyaxis}
\end{tikzpicture}
{
\tikzset{external/figure name={subset_}}
A simple image is \tikz \fill (0,0) circle(5pt);. % will be written to ’subset_0.pdf’
\begin{tikzpicture} % will be written to ’subset_1.pdf’

\begin{axis}
\addplot {x^2};
\end{axis}
\end{tikzpicture}
}% here, the old file name will be restored:
\begin{tikzpicture} % will be written to ’main-figure1.pdf’

\begin{axis}
\addplot[domain=1e-3:100] {1/x};
\end{axis}
\end{tikzpicture}
\end{document}
The scope of figure name ends with the next closing brace (as all values set by \tikzset do).
1 These counters are stored into di↵erent macros. In other words: no TEX register will be needed.
Remark: Use \tikzset{external/figure name/.add={hprefix i}{hsuffix i}} to prepend a hprefix i

and append a hsuffix i to the actual value of figure name. Might be useful for something like
\tikzset{external/figure name=main}
% uses main_0.pdf, main_1.pdf, ...
\chapter{The first section}

{\tikzset{external/figure name/.add={}{_firstsection}}
...
% uses main_firstsection_0.pdf, main_firstsection_1.pdf, ...
}
\chapter{The second section}

{\tikzset{external/figure name/.add={}{secondsection_}}
...
% uses main_secondsection_0.pdf, main_secondsection_1.pdf, ...
\section{Second subsection}
{\tikzset{external/figure name/.add={}{sub_}}
...
% uses main_secondsection_sub_0.pdf, main_secondsection_sub_1.pdf, ...
}
% uses main_secondsection_2.pdf, main_secondsection_3.pdf, ...
}
\tikzappendtofigurename{hsuffix i}
Appends hsuffix i to the actual value of figure name.
It is a shortcut for \tikzset{external/figure name/.add={}{hsuffix i}} (a shortcut which is also
supported if Tik Z is not installed, see below).
Configuration option for eps output or MikTEX: Since the external lib works by means of
system calls, it has to be modified to fit the local system. This is necessary for MikTEX since it uses
a di↵erent option to enable these system calls. It is also necessary for eps output since this involves a
di↵erent set of utilities.
Note that the most important part is to enable system calls. This is typically done by typesetting your
document with pdflatex -shell-escape or pdflatex -enable-write18 (MikTEX). These options
need to be configured in your TEX editor. Besides this step, one may want to configure the system call:
/tikz/external/system call={htemplatei}
A template string used to generate system calls. Inside of htemplatei, the macro \image can be
used as placeholder for the image which is about to be generated while \texsource contains the
main file name (in truth, it contains \input{hmain file namei}, but that doesn’t matter).
The default is
\tikzset{external/system call={pdflatex \tikzexternalcheckshellescape -halt-on-error

-interaction=batchmode -jobname "\image" "\texsource"}
where \tikzexternalcheckshellescape inserts the value of the configuration key shell escape
if and only if the current document has been typeset with -shell-escape2 .
For eps output, you can (and need to) use
\tikzset{external/system call={latex \tikzexternalcheckshellescape -halt-on-error

-interaction=batchmode -jobname "\image" "\texsource" &&
dvips -o "\image".ps "\image".dvi}}
The argument htemplatei will be expanded using \edef, so any control sequences will be expanded.
During this evaluation, ‘\\’ will result in a normal backslash, ‘\’. Furthermore, double quotes ‘"’,
single quotes ‘’’, semicolons and dashes ‘-’ will be made to normal characters if any package uses
them as macros. This ensures compatibility with the german package, for example.
2 Note that this is always true for the default configuration. This security consideration applies mainly for mode=list and
make which will also work without shell escapes.

/tikz/external/shell escape={hcommand-line argi} (initially -shell-escape)

Contains the command line option for latex which enables the \write18 feature.
For TEX-Live, this is -shell-escape. For MikTEX, you should use \tikzexternalize[shell
escape=-enable-write18].
Support for Labels and References In External Files The external library comes with extra
support for \label and \ref (and other commands which usually store information in the .aux file)
inside of external files.
There are, however, some points which need your attention when you try to use
a) \ref to something in the main document inside of an externalized graphics or
b) \label in the externalized graphics which is referenced in the main document.
For point a), a \ref inside of an externalized graphics works only if you issue the required system call
manually or by make. The initial configuration mode=convert with system call does not support
\ref. But you can copy–paste the system call generated by mode=convert with system call and
issue it manually. The reason is that \ref information is stored in the main .aux file – but this
auxiliary file is not completely written when mode=convert with system call is invoked (there is a
race condition). Note that \pageref is not supported (sorry). Thus: if you have \ref inside of external
graphics, consider using mode=list and make or copy–paste the system call for the image(s) and issue
it manually.
Point b) is realized automatically by the external library. In detail, a \label inside of an externalized
graphics causes the external library to generate separate auxiliary files for every external image. These
files are called himagenamei.dpth. The extension .dpth indicates that the file also contains the image’s
depth (the baseline key of Tik Z). Furthermore, anything which would have been written to an .aux file
will be redirected to the .dpth file – but only things which occur inside of the externalized tikzpicture
environment. When the main document loads the image, it will copy the .dpth file into the main .aux
file. Then, successive compilations of the main document contain the external \label information. In
other words, a \label in an external graphics needs the following work flow:
1. The external graphics needs to be generated together with its .dpth (usually automatically by
Tik Z).
2. The main document includes the external graphics and copies the .dpth content into its main .aux
file.
3. The main document needs to be translated once again to re-read its .aux file3 .
There is just a special case if a \label/\ref drawn as a tikzpicture. This is, for example, the case
for the legend \ref images or for the \pgfplotslegendfromname feature. In such cases, you need to
proceed as for case a) since mode=convert with system call can’t handle that stu↵ on its own.
In other words: a \label in an external document works automatically, just translate the main document
often enough. A \ref might need manual adjustments as described for case a) above.
Operation Modes
/tikz/external/mode=convert with system call|list and make|. . . (initially convert with
system call)
This allows to change the default operation mode. There are a handful of choices possible, all of
them are described in detail in [6, section “Externalization Library”]. The most useful ones are
probably the initial configuration convert with system call and the specialized choice list and
make.
The choice list and make configures the library to check if there are already external graphics and
uses them. If there are no graphics, the library will skip the figure. However, it will also generate
a makefile to generate the graphics, and a list of all required graphics files.
It is not required to use make: the library expects you to generate the images somehow and it doesn’t
care about the “how”. Using make -f hname-of-tex-filei.makefile -j 2 allows parallel execution
3 Note that it is not possible to activate the content of an auxiliary file after \begin{document} in LATEX.
which might, indeed, be an option. Furthermore, the makefile also supports file dependencies: if
one of your data tables has been updated, the external graphics will be remade automatically.
pgfplots tells the external library about any file dependencies (input files and tables).
The two modes have the following characteristics:
1. convert with system call is automatic and does everything on–the–fly. However, it can’t
work with \ref and/or \label information in external pictures.
2. list and make requires either manual (by issuing the system calls manually) or semi–
automatic conversion (using the generated hmaini.makefile), and multiple runs of pdflatex.
The generated Makefile can be processed in parallel. Furthermore, list and make provides
full support for \ref and \label: any \label defined inside of an externalized graphics is still
available for the main document.
If you have legends with legend to name or \label/\ref, you need to generate the graphics
defining the \label (or legend to name), then run pdflatex twice on the main document.
Afterwards, you can externalize the legend graphics.
The complete reference documentation and remaining options are documented in [6, “Externalization
Library”]. This reference also contains information about
• how to use \tikzset{external/force remake} and \tikzset{external/remake next} to re-
make selected figures,
• how to disable the externalization partially with \tikzset{external/export=false} or com-
pletely with \tikzexternaldisable,
• how to optimize the speed of the conversion process using \tikzset{external/optimize command
away=\myExpensiveMacro},
• how to add further remake-dependencies with \tikzpicturedependsonfile{hnamei} and/or
\tikzexternalfiledependsonfile{hexternal filei}{hnamei},
• examples how to enable png export,
• how to typeset such a document without pgf installed or
• how to provide work-arounds with .pdf images and bounding box restrictions.
Using the Library Without pgf or pgfplots Installed There is a small replacement package
tikzexternal.sty which can be used once every figure has been exported. The idea is to uncomment
\usepackage{tikz} and \usepackage{pgfplots} and write \usepackage{tikzexternal} instead:
% \usepackage{tikz}
% \usepackage{pgfplots}
\usepackage{tikzexternal}
\tikzexternalize% activate externalization
\begin{document}
\begin{tikzpicture}
...
\end{tikzpicture}
...
\end{document}
You do not need pgf, Tik Z or pgfplots installed. What you need is tikzexternal.sty and all
generated figures (consisting of the image files, ‘.pdf’ and the ‘.dpth’ files containing information of
the baseline option). The file tikzexternal.sty is shipped with pgf in the directory
latex/pgf/utilities/tikzexternal.sty
and a copy is shipped with pgfplots in
tex/generic/pgfplots/oldpgfcompatib/pgfplotsoldpgfsupp_tikzexternal.sty
Just copy the file into your directory and rename it to tikzexternal.sty.
Attention: The small replacement package doesn’t support key–value interfaces. Thus, it is necessary
to use \tikzsetexternalprefix instead of the prefix option and \tikzsetfigurename instead of the
figure name option since \tikzset is not available in such a context. Also, you may want to define a
dummy–macro \pgfplotsset if you have used \pgfplotsset.
7.1.2 Using the Externalization Framework of pgf “By Hand”

Another way to export TEX-pictures to single graphics files is to use the externalization framework of
pgf, which requires more work but works more generally than the external library. The basic idea is to
encapsulate the desired parts with
\beginpgfgraphicnamed{houtput file namei}
hpicture contentsi
\endpgfgraphicnamed.
Furthermore, one needs to tell pgf the name of the main document using
\pgfrealjobname{hthe real job’s namei}
in the preamble. This enables two di↵erent modes:
1. The first is the normal typesetting mode. LATEX checks whether a file named houtput file namei
with one of the accepted file extensions exists – if that is the case, the graphics file is included with
\pgfimage and the hpicture contentsi is skipped. If no such file exists, the hpicture contentsi is typeset
normally. This mode is applied if \jobname equals hthe real job’s namei.
2. The second mode applies if \jobname equals houtput file namei, it initiates the “conversion mode”
which is used to write the graphics file houtput file namei. In this case, only hpicture contentsi is
written to \jobname, the complete rest of the LATEX is processed as normal, but it is silently discarded.
This mode needs to be started manually with pdflatex --jobname houtput file namei for every ex-
ternalized graphics file.
A complete example may look as follows.

\pgfrealjobname{test}
\begin{document}
\begin{figure}
\beginpgfgraphicnamed{testfigure}
\begin{tikzpicture}
\begin{axis}
\addplot {x^2};
\end{axis}
\end{tikzpicture}
\endpgfgraphicnamed
\caption{Our first external graphics example}
\end{figure}
\begin{figure}
\beginpgfgraphicnamed{testfigure2}
\begin{tikzpicture}
\begin{axis}
\addplot {x^3};
\end{axis}
\end{tikzpicture}
\endpgfgraphicnamed
\caption{A second graphics}
\end{figure}
\end{document}
The file is named test.tex, and it is processed (for example) with

pdflatex test
Now, we type
pdflatex --jobname testfigure test
pdflatex --jobname testfigure2 test
7.2. IMPORTING FROM MATLAB 515
to enter conversion mode. These last calls will only write the contents of our named graphics environ-
ments, one for htestfigurei and one for htestfigure2 i into the respective output files testfigure.pdf and
testfigure2.pdf.
In summary, one needs \pgfrealjobname and calls pdflatex --jobname hgraphics filei for every ex-
ternalized graphics environment. Please note that it is absolutely necessary to use the syntax above, not
\begin{pgfgraphicnamed}.
These steps are explained in much more detail in Section“Externalizing Graphics” of [6].
Attention: Do not forget a correct \pgfrealjobname statement! If it is missing, externalization simply

won’t work. If it is wrong, any call to LATEX will produce empty output files.
It should be noted that this approach of image externalization is not limited to Tik Z picture environments.
In fact, it collects everything between the begin and end statements into the external file. It is implicitly
assumed that the encapsulated stu↵ is one box, but you can also encapsulate complete paragraphs using
something like the LATEX minipage (or a \vbox which is not as powerful but does not a↵ect the remaining
document that much).
/pgf/images/aux in dpth=true|false (initially false)

If this boolean is set to true, any \label information generated inside of the external image is stored
into the already mentioned .dpth file. The main document can thus reference label information of
externalized parts of the document (although you may need to run latex several times).
Label support is provided for \ref, and probably \cite. The \pageref command is only partially
supported.
Using the Library Without pgf Installed Simply uncomment the packages \usepackage{tikz} and
\usepackage{pgfplots} and use
\long\def\beginpgfgraphicnamed#1#2\endpgfgraphicnamed{%
\begingroup
\setbox1=\hbox{\includegraphics{#1}}%
\openin1=#1.dpth
\ifeof1 \box1
\else
\read1 to\pgfincludeexternalgraphicsdp \closein1
\dimen0=\pgfincludeexternalgraphicsdp\relax
\hbox{\lower\dimen0 \box1 }%
\fi
\endgroup
}
instead. This will include the generated graphics files (and it will respect the baseline information
stored in .dpth files). Consequently, you won’t need pgf or pgfplots installed. See Section“Externalizing
Graphics” of [6] for details.
7.2 Importing From Matlab

7.2.1 Importing Mesh Data From Matlab To PGFPlots
While it is easy to write Matlab vectors to files (using save P.dat data -ASCII), it is more involved to
export mesh data.
The main problem is to communicate the mesh structure to pgfplots.
Here is an example how to realize this task: in Matlab, we have mesh data X, Y and Z which are matrices
of the same size. For example, suppose we have
[X,Y] = meshgrid( linspace(-1,1,5), linspace(4,5,10) );
Z = X + Y;
surf(X,Y,Z)
as data. Then, we can generate an N ⇥ 3 table containing all single elements in column–wise ordering with
data = [ X(:) Y(:) Z(:) ]
save P.dat data -ASCII
where the second command stores the N ⇥ 3 table into P.dat. Finally, we can use
\addplot3[surf,mesh/rows=10,mesh/ordering=colwise,shader=interp] file {P.dat};
in pgfplots to read this data. We need to provide either the number of rows (10 here) or the number
of columns – and the ordering (which is colwise for Matlab matrices).
An alternative which is faster in pgfplots would be to transpose the matrices in Matlab and tell
pgfplots they are in rowwise ordering. So, the last step becomes
XX=X’; YY=Y’; ZZ=Z’;
data = [ XX(:) YY(:) ZZ(:) ]
save P.dat data -ASCII
with pgfplots command
\addplot3[surf,mesh/cols=10,mesh/ordering=rowwise,shader=interp] file {P.dat};.
7.2.2 matlab2pgfplots.m
This is a Matlab (®) script which attempts to convert a Matlab figure to pgfplots. It requires Matlab
version 7.4 (or higher).
Attention: This script is largely outdated and supports only a very small subset of pgfplots. You may
want to look at matlab2tikz, a conversion script of Nico Schlömer available at
http://www.mathworks.com/matlabcentral/fileexchange/22022-matlab2tikz
which also uses pgfplots for the LATEX conversion.
The idea of matlab2pgfplots.m is to
• use a complete matlab figure as input,
• acquire axis labels, axis scaling (log or normal) and legend entries,
• acquire all plot coordinates
and write an equivalent .pgf file which typesets the plot with pgfplots.
The intention is not to simulate matlab. It is a first step for a conversion. Type
> help matlab 2 pgfplots
on your matlab prompt for more information about its features and its limitations.
This script is experimental.
7.2.3 matlab2pgfplots.sh
A bash-script which simply starts matlab and runs
f = hgload ( ’ somefigure . fig ’ ) ;
matlab 2 pgfplots ( ’ outputfile . pgf ’ , ’fig ’ , f ) ;
See matlab2pgfplots.m above.
7.2.4 Importing Colormaps From Matlab

Occasionally, you may want to reuse your matlab colormap in pgfplots. Here is a small Matlab script
which converts it to pgfplots:
C = colormap; % gets data of the current colormap.
% C = colormap(jet) % gets data of "jet"
eachnth = 1;
I = 1:eachnth:size(C,1); % this is nonsense for eachnth=1 -- but perhaps you don’t want each color.
CC = C(I,:);
TeXstring = [ ...
sprintf(’\\pgfplotsset{\n\tcolormap={matlab}{\n’) ...
sprintf(’\t\trgb=(%f,%f,%f)\n’,CC’) ...
sprintf(’\t}\n}\n’) ]
7.3 SVG Output

It is possible to write every single Tik Z picture into a scalable vector graphics (.svg) file. This has nothing
to do with pgfplots, it is a separate driver of pgf. Please refer to [6, Section “Producing HTML / SVG
Output”].
7.4. GENERATE PGFPLOTS GRAPHICS WITHIN PYTHON 517
7.4 Generate pgfplots Graphics Within Python

Mario Orne DÍAZ ANADÓN contributed a small python script pgfplots.py which provides a simple in-
terface to generate pgfplots figures from within python. It can be found in the pgfplots installation
directory, in pgfplots/scripts/pgfplots/pgfplots.py; documentation can be found in the file.
Chapter 8
Utilities and Basic Level Commands
This section documents commands which provide access to more basic elements of pgfplots. Most of them
are closely related to the basic level of pgf, especially various point commands which are specific to an axis.
Some of them are general purpose utilities like loops.
However, most elements in this section are only interesting for advanced users – and perhaps only for
special cases.
8.1 Utility Commands

\foreachhvariablesi in hlisti {hcommandsi}
A powerful loop command provided by Tik Z, see [6, Section Utilities].
Iterating 1. Iterating 2. Iterating 3. Iterating 4. \foreach \x in {1,2,...,4} {Iterating \x. }%
A pgfplots related example could be
\foreach \i in {1,2,...,10} {\addplot table {datafile\i}; }%
\pgfplotsforeachungroupedhvariablei in hlisti {hcommand i}

A specialised variant of \foreach which can do two things: it does not introduce extra groups while
executing hcommand i and it allows to invoke the math parser for (simple!) hx0 i,hx1 i,...,hxn i expres-
sions.
Iterating 1. Iterating 2. Iterating 3. Iterating 4. All collected = , 1, 2, 3, 4.

\def\allcollected{}
\pgfplotsforeachungrouped \x in {1,2,...,4} {Iterating \x. \edef\allcollected{\allcollected, \x}}%
All collected = \allcollected.
A more useful example might be to work with tables. The following example is taken from Pgfplot-
sTable:
\pgfplotsforeachungrouped \i in {1,2,...,10} {%
\pgfplotstablevertcat{\output}{datafile\i} % appends ‘datafile\i’ -> ‘\output’
}%
% since it was ungrouped, \output is still defined (would not work
% with \foreach)
Remark: The special syntax hlisti=hx0 i,hx1 i,...,hxn i, i.e. with two leading elements, followed by
dots and a final element, invokes the math parser for the loop. Thus, it allows larger number ranges
than any other syntax if /pgf/fpu is active. In all other cases, \pgfplotsforeachungrouped invokes
\foreach and provides the results without TEX groups.
Keep in mind that inside of an axis environment, all loop constructions (including custom loops,
\foreach and \pgfplotsforeachungrouped) need to be handled with care: loop arguments can only
be used in places where they are immediately evaluated; but pgfplots postpones the evaluation of
518
8.1. UTILITY COMMANDS 519
many macros. For example, to loop over something and to generate axis descriptions of the form \node
at (axis cs:\i,0.5)...., the loop macro \i will be evaluated in \end{axis} – but at that time, the
loop is over and its value is lost. The correct way to handle such an application is to expand the loop
variable explicitly. For example:
\pgfplotsforeachungrouped \i/\j in {
1 / a,
2 / b,
3 / c
}{
\edef\temp{\noexpand\node at (axis cs: \i,0.5) {\j};}
% \show\temp % lets TeX show you what \temp contains
\temp
}
The example generates three loop iterations: \i=1, \j=a; then \i=2, j=b; then \i=3, \j=c. Inside of the
loop body, it expands them and assigns the result to a macro using an “expanded definition”, \edef.
The result no longer contains either \i or \j (since these have been expanded). Then, it invokes the
resulting macro. Details about the TEX command \edef and expansion control can be found in the
document TeX-programming-notes.pdf which comes with pgfplots.
\pgfplotsinvokeforeach{hlisti} {hcommand i}
A variant of \pgfplotsforeachungrouped (and such also of \foreach) which replaces any occurrence
of #1 inside of hcommand i once for every element in hlisti. Thus, it actually assumes that {hcommand i}
is like a \newcommand body.
In other words, hcommand i is invoked for every element of hlisti. The actual element of hlisti is available
as #1.
As \pgfplotsforeachungrouped, this command does not introduce extra scopes (i.e. it is ungrouped
as well).
The di↵erence to \foreach \x in hlisti{hcommand i} is subtle: the \x would not be expanded whereas
#1 is.
Invoke them: [a] [b] [c] [d] \pgfkeys{

otherstyle a/.code={[a]},
otherstyle b/.code={[b]},
otherstyle c/.code={[c]},
otherstyle d/.code={[d]}}
\pgfplotsinvokeforeach{a,b,c,d}
{\pgfkeys{key #1/.style={otherstyle #1}}}
Invoke them:
\pgfkeys{key a} \pgfkeys{key b}
\pgfkeys{key c} \pgfkeys{key d}
The counter example would use a macro (here \x) as loop argument:
Invoke them: [d] [d] [d] [d] \pgfkeys{

otherstyle a/.code={[a]},
otherstyle b/.code={[b]},
otherstyle c/.code={[c]},
otherstyle d/.code={[d]}}
\pgfplotsforeachungrouped \x in {a,b,c,d}
{\pgfkeys{key \x/.style={otherstyle \x}}}
Invoke them:
\pgfkeys{key a} \pgfkeys{key b}
\pgfkeys{key c} \pgfkeys{key d}
Restrictions: you can’t nest this command yet (since it does not introduce protection by scopes).
\pgfmathparse{hexpressioni}
Invokes the pgf math parser for hexpressioni and defines \pgfmathresult to be the result.
The result is ‘42.0’. \pgfmathparse{1+41}
The result is ‘\pgfmathresult’.

520 CHAPTER 8. UTILITIES AND BASIC LEVEL COMMANDS
The math engine in pgf typically uses TEX’s internal arithmetics. That means: it is well suited for
numbers in the range [ 16384, 16384] and has a precision of 5 digits.
The number range is typically too small for plotting applications. pgfplots improves the number range
by means of \pgfkeys{/pgf/fpu}\pgfmathparse{1+41} to activate the “floating point unit” (fpu) and
to apply all following operations in floating point.
In pgfplots, the key /pgfplots/use fpu is typically on, which means that any coordinate arithmetics
are carried out with the fpu. However, all pgf related drawing operations still use the standard math
engine.
In case you ever need to process numbers of extended precision, you may want to use
The result is ‘1 · 106 ’. \pgfkeys{/pgf/fpu}%

\pgfmathparse{1000*1000}
The result is ‘\pgfmathprintnumber{\pgfmathresult}’.
Note that results of the fpu are typically not in human-readable format, so \pgfmathprintnumber is
the preferred way to typeset such numbers.
Please refer to [6] for more details.
/pgf/declare function=hfunction definitionsi

Allows to define one or more functions.
The argument hfunction definitionsi can cantain one or more definitions, and each must be terminated
by a semicolon:
\begin{tikzpicture}
\begin{axis}[
20 declare function={
C=4;
square(\t)=(\t)^2 + C;
},
]
0 \addplot+[samples=2] {C*x};
\addplot {square(x)};
\end{axis}
6 4 2 0 2 4 6
The definitions as such have the form hfunctionihargument listi = hdefinitioni where the hargument
listi contains a comma-separated-list of arguments like \t or \t,\a,\b. The hdefinitioni is a math
expression which makes use of these arguments.
Please refer to [6] for more details.
\pgfplotstableread{hfilei}
Please refer to the manual of PgfplotsTable, pgfplotstable.pdf, which is part of the pgf-
plots-bundle.
\pgfplotstabletypeset{h\macroi}
Please refer to the manual of PgfplotsTable, pgfplotstable.pdf, which is part of the pgf-
plots-bundle.
\pgfplotsiffileexists{hfilenamei}{htrue codei}{hfalse codei}

Invokes htrue codei if hfilenamei exists and hfalse codei if not. Can be used in looping macros, for
example to plot every data file until there are no more of them.
\pgfplotsutilifstringequal{hfirsti}{hsecond i}{htrue codei}{hfalse codei}

A simple “strcmp” tool which invokes htrue codei if hfirsti =hsecond i and hfalse codei otherwise. This
does not expand macros.
8.2. COMMANDS INSIDE OF PGFPLOTS AXES 521
\pgfkeys
\pgfeov
\pgfkeysvalueof
\pgfkeysgetvalue
These commands are part of the Tik Z way of specifying options, its sub-package pgfkeys. The
\pgfplotsset command is actually nothing but a wrapper around \pgfkeys.
A short introduction into \pgfkeys can be found in [8] whereas the complete reference is, of course, the
Tik Z manual [6].
The key \pgfkeysvalueof{hkey namei} expands to the value of a key; \pgfkeysgetvalue{hkey
namei}{h\macroi} stores the value of hkey namei into h\macroi. The \pgfeov macro is used to delimit
arguments for code keys in \pgfkeys, please refer to the references mentioned above.
8.2 Commands Inside Of pgfplots Axes

\autoplotspeclist
This command should no longer be used, although it will be kept as technical implementation detail.
Please use the ‘cycle list’ option, Section 4.7.7.
\logten
Expands to the constant log(10). Useful for logplots because log(10i ) = i log(10). This command is
only available inside of a Tik Z-picture.
\pgfmathprintnumber{hnumber i}
Generates pretty–printed output1 for hnumber i. This method is used for every tick label.
The number is printed using the current number printing options, see the manual of PgfplotsTable
which comes with this package for the di↵erent number styles, rounding precision and rounding methods.
\numplots
Inside of any of the axis environments, associated style, option or command, \numplots expands to the
total number of plots.
\numplotsofactualtype
Like \numplots, this macro returns the total number of plots which have the same plot handler. Thus,
if you have sharp plot active, it returns the number of all sharp plots. If you have ybar active, it
returns the number of ybar plots and so on.
\plotnum
Inside of \addplot or any associated style, option or command, \plotnum expands to the current plot’s
number, starting with 0.
\plotnumofactualtype
Like \plotnum, but it returns the number among all plots of the same type. The number of all such
plots is available using \numplotsofactualtype.
\coordindex
Inside of an \addplot command, this macro expands to the number of the actual coordinate (starting
with 0).
It is useful together with x filter or y filter to (de)select coordinates.
8.3 Path Operations

\path
\draw
1 This method was previously \prettyprintnumber. Its functionality has been included into pgf and the old command is
now deprecated.
\fill
\node
\matrix
These commands are Tik Z drawing commands all of which are documented in [6]. They are used to
draw or fill paths, generate text nodes or aligned text matrices. They are equivalent to \path[draw],
\path[fill], \path[node], \path[matrix], respectively.
\path . . . --hcoordinatei . . . ;
A Tik Z path operation which connects the current point (the last one before --) and hcoordinatei with
a straight line.
\path . . . |-hcoordinatei . . . ;
A Tik Z path operation which connects the current point and hcoordinatei with two straight lines: first
vertical, then horizontal.
\path . . . -|hcoordinatei . . . ;
A Tik Z path operation which connects the current point and hcoordinatei with two straight lines: first
horizontal, then vertical.
/tikz/xshift={hdimensioni}
/tikz/yshift={hdimensioni}
These Tik Z keys allow to shift something by hdimensioni which is any TEX size (or expression).
\pgfplotsextra{hlow-level path commandsi}

A command to execute hlow-level path commandsi in a pgfplots axis. Since any drawing commands
inside of an axis need to be postponed until the axis is complete and the scaling has been initialised,
it is not possible to simply draw any paths. Instead, it is necessary to draw them as soon as the axis
is finished. This is done automatically for every Tik Z path – and it is also done manually if you write
\pgfplotsextra{hcommandsi}.
\begin{tikzpicture}
4
\pgfplotsextra{%
\pgfpathmoveto{\pgfplotspointaxisxy{1}{2}}%
\pgfpathlineto{\pgfplotspointaxisxy{2}{4}}%
\pgfusepath{stroke}%
}
2 \end{axis}
\end{tikzpicture}
0
0 1 2 3
The example above initializes an axis and executes the basic level path commands as soon as the axis
is ready. The execution of multiple \path, \addplot and \pgfplotsextra commands is in the same
sequence as they occur in the environment2 .
\pgfplotspathaxisoutline
Generates a path which resembles the outline of the current axis. This path is used for clip paths and
the background paths (if any).
8.4 Specifying Basic Coordinates

\pgfplotspointaxisxy{hx coordinatei}{hy coordinatei}
2 Except for stacked plots where the sequence may be reverse, see the key reverse stack plots.
8.4. SPECIFYING BASIC COORDINATES 523
\pgfplotspointaxisxyz{hx coordinatei}{hy coordinatei}{hz coordinatei}

Point commands like \pgfpointxy which take logical, absolute coordinates and return a low–level point.
Every transformation from user transformations to logarithms is applied.
Since the transformations are initialized after the axis is complete, this command needs to be postponed
(see \pgfplotsextra).
This command is the basic–level variant of axis cs:hx coordinatei,hy coordinatei,hz coordinatei.
Note that this is also the default coordinate system during the visualization phase; in other words: if you
write \draw (1,2) -- (1,4), pgfplots will automatically use (axis cs:1,2) -- (axis cs:1,4).
\pgfplotspointaxisdirectionxy{hx coordinatei}{hy coordinatei}
\pgfplotspointaxisdirectionxyz{hx coordinatei}{hy coordinatei}{hz coordinatei}
Point commands like \pgfpointxy which take logical, relative coordinates and return a low–level
point. Every transformation from user transformations to logarithms is applied. The di↵erence
to \pgfplotspointaxisxy is that the shift of the linear transformation is skipped here (compare
disabledatascaling).
This command is the basic–level variant of axis direction cs:hx coordinatei,hy coordinatei,hz
coordinatei. Please refer to the documentation of axis direction cs for more details.
Use this command whenever something of relative character like directions or lengths need to be supplied.
One use-case is to draw ellipses:
\begin{tikzpicture}
2 \begin{axis}[
xmin=-3, xmax=3,
ymin=-3, ymax=3,
extra x ticks={-1,1},
extra y ticks={-2,2},
0
extra tick style={grid=major},
]
\draw[red] \pgfextra{
\pgfpathellipse{\pgfplotspointaxisxy{0}{0}}
2 {\pgfplotspointaxisdirectionxy{1}{0}}
{\pgfplotspointaxisdirectionxy{0}{2}}
% see also the documentation of
2 1 0 1 2 % ’axis direction cs’ which
% allows a simpler way to draw this ellipse
};
\draw[blue] \pgfextra{
\pgfpathellipse{\pgfplotspointaxisxy{0}{0}}
};
\addplot [only marks,mark=*] coordinates
{ (0,0) };
\end{axis}
\end{tikzpicture}
Since the transformations are initialized after the axis is complete, this command needs to be provided
either inside of a Tik Z \path command (like \draw in the example above) or inside of \pgfplotsextra.
\pgfplotspointrelaxisxy{hrel x coordinatei}{hrel y coordinatei}
\pgfplotspointrelaxisxyz{hrel x coordinatei}{hrel y coordinatei}{hrel z coordinatei}
Point commands which take relative coordinates such that x = 0 is the lower x axis limit and x = 1
the upper x axis limit.
These commands are used for rel axis cs.
Please note that the transformations are only initialised if the axis is complete! This means you need
to provide \pgfplotsextra.
\pgfplotspointdescriptionxy{hx fractioni}{hy fractioni}
\pgfplotsqpointdescriptionxy{hx fractioni}{hy fractioni}
Point commands such that {0}{0} is the lower left corner of the axis’ bounding box and {1}{1} the
upper right one; everything else is in between. The ‘q’ variant is quicker as it doesn’t invoke the math
parser on its arguments.
They are used for axis description cs, see Section 4.9.1.
\pgfplotspointaxisorigin
A point coordinate at the origin, (0, 0, 0). If the origin is not part of the axis limits, the nearest point
on the boundary is returned instead.
This is the same coordinate as returned by the origin anchor.
\pgfplotstransformcoordinatex{hx coordinate of an axisi}

\pgfplotstransformcoordinatey{hy coordinate of an axisi}
\pgfplotstransformcoordinatez{hz coordinate of an axisi}
Defines \pgfmathresult to be the low-level pgf coordinate corresponding to the input argument.
The command applies any [xyz] coord trafo keys, data scalings and/or logarithms or whatever pgf-
plots does to map input coordinates to internal coordinates.
The result can be used inside of a \pgfpointxy statement (i.e. it still needs to be scaled with the
respective pgf unit vector).
\begin{tikzpicture}
4
\pgfplotsextra{%
\pgfplotstransformcoordinatex{1}%
\let\xcoord=\pgfmathresult
\pgfplotstransformcoordinatey{1}%
\let\ycoord=\pgfmathresult
2 \pgfpathcircle
{\pgfqpointxy{\xcoord}{\ycoord}}
{5pt}%
\pgfusepath{fill}%
}%
0 \end{axis}
0 0.5 1 1.5 2 \end{tikzpicture}
The result of this command is also available as math method transformcoordinatex (see the docu-
mentation for axis cs).
Please note that the transformations are only initialised if the axis is complete. This means you need
to provide \pgfplotsextra as is shown in the example above.
\pgfplotstransformdirectionx{hx direction of an axisi}

\pgfplotstransformdirectiony{hy direction of an axisi}
\pgfplotstransformdirectionz{hz direction of an axisi}
Defines \pgfmathresult to be a low-level pgf direction vector component.
A direction vector needs to be added to some coordinate in order to get a coordinate, compare the
documentation for \pgfplotspointaxisdirectionxy and axis direction cs.
The argument hx direction of an axisi is processed in (almost) the same way as for the macro which
operates on absolute positions, \pgfplotstransformcoordinatex. The only di↵erence is that directions
need no shifting transformation.
The result of this command is also available as math method transformdirectionx (see the documen-
tation for axis direction cs).
See axis direction cs for details and examples about this command.
\pgfplotspointunitx
\pgfplotspointunity
\pgfplotspointunitz
Low–level point commands which return the canvas x, y or z unit vectors.
The \pgfplotspointunitx is the pgf unit vector in x direction.
These vectors are essentially the same as \pgfqpointxyz{1}{0}{0}, \pgfqpointxyz{0}{1}{0}, and
\pgfqpointxyz{0}{0}{1}, respectively.
The unit z vector is only defined for three dimensional axes.
8.4. SPECIFYING BASIC COORDINATES 525
\pgfplotsunitxlength
\pgfplotsunitylength
\pgfplotsunitzlength
\pgfplotsunitxinvlength
\pgfplotsunityinvlength
\pgfplotsunitzinvlength
Macros which expand to the vector length kxi k of the respective unit vector xi or the inverse vector
length, 1/kxi k. These macros can be used inside of \pgfmathparse, for example.
The xi are the \pgfplotspointunitx variants.
\pgfplotsqpointoutsideofaxis{hthree-char-stringi}{hcoordinatei}{hnormal distancei}
Provides a point coordinate on one of the available four axes in case of a two dimensional figure or on
one of the available twelve axes in case of a three dimensional figure.
The desired axis is uniquely identified by a three character string, provided as first argument to the
command. The first of the three characters is ‘0’ if the x coordinate of the specified axis passes through
the lower axis limit. It is ‘1’, if the x coordinate of the specified axis passes through the upper axis
limit. Furthermore, it is ‘2’ if it passes through the origin. The second character is also either 0, 1 or
2 and it characterizes the position on the y axis. The third character is for the third dimension, the z
axis. It should be left at ‘0’ for two dimensional plots. However, one of the three characters should be
‘v’, meaning the axis varies. For example, v01 denotes {(x, ymin , zmax )|x 2 R}.
The second argument, hcoordinatei is the logical coordinate on that axis. Since two coordinates of the
axis are fixed, hcoordinatei refers to the varying component of the axis. It must be a number without
unit; no math expressions are supported here.
The third argument hnormal distancei is a dimension like 10pt. It shifts the coordinate away from the
designated axis in direction of the outer normal vector. The outer normal vector always points away
from the axis. It is computed using \pgfplotspointouternormalvectorofaxis.
There are several variants of this command which are documented in the source code. One of them is
particularly useful:
\pgfplotsqpointoutsideofaxisrel{hthree-char-stringi}{haxis fractioni}{hnormal distancei}

This point coordinate is a variant of \pgfplotsqpointoutsideofaxis which allows to provide an haxis
fractioni instead of an absolute coordinate. The fraction is a number between 0 (lower axis limit) and
1 (upper axis limit), i.e. it is given in percent of the total axis. It is possible to provide negative values
or values larger than one.
The \pgfplotsqpointoutsideofaxisrel command is similar in spirit to rel axis cs.
There is one speciality in conjunction with reversed axes: if the axis has been reversed by x dir=reverse
and, in addition, allow reversal of rel axis cs is true, the value 0 denotes the upper limit while
1 denotes the lower limit. The e↵ect is that coordinates won’t change just because of axis reversal.
\pgfplotspointouternormalvectorofaxis{hthree-char-stringi}
A point command which yields the outer normal vector of the respective axis. The normal vector
has length 1 (computed with \pgfpointnormalised). It is the same normal vector used inside of
\pgfplotsqpointoutsideofaxis and its variants.
The output of this command will be cached and re-used during the lifetime of an axis.
\pgfplotsticklabelaxisspec{hx, y or z i}
Expands to the three-character-identification for the axis containing tick labels for the chosen axis,
either hx i, hyi or hz i.
\pgfplotsvalueoflargesttickdimen{hx, y or z i}
Expands to the largest distance of a tick position to its tick label bounding box in direction of the outer
unit normal vector. It does also include the value of the ticklabel shift key.
This value is used for ticklabel cs.
\pgfplotsmathfloatviewdepthxyz{hx i}{hyi}{hz i}
\pgfplotsmathviewdepthxyz{hx i}{hyi}{hz i}
Both macros define \pgfmathresult to be the “depth” of a three dimensional point x̄ = (x, y, z). The
~ the view direction of the current axis.
depth is defined to be the scalar product of x̄ with d,
For \pgfplotsmathfloatviewdepthxyz, the arguments are parsed as floating point numbers and
the result is encoded in floating point. A fixed point representation can be generated with
\pgfmathfloattofixed{\pgfmathresult}.
For \pgfplotsmathviewdepthxyz, TEX arithmetics is employed for the inner product and the result is
assigned in fixed point. This is slightly faster, but has considerably smaller data range.
Both commands can only be used inside of a three dimensional pgfplots axis (as soon as the axis is
initialised, see \pgfplotsextra).
\ifpgfplotsthreedimhtrue codei\elsehelse codei\fi

A TEX \if which evaluates the htrue codei if the axis is three dimensional and the helse codei if not.
8.5 Accessing Axis Limits

It is also possible to access axis limits during the visualization phase, i.e. during \end{axis}. Please refer
to the reference documentation for xmin on page 306.
8.6 Accessing Point Coordinate Values

During the visualization phase, pgfplots provides access to the currently processed coordinate and its
values.
This access requires a call to specific macros. These macros write the coordinate values to some publicy
available key–value pairs. Then, the current point’s x, y, z, and color data can be accessed.
\pgfplotspointgetcoordinates
A macro which is supposed to be called during the visualization phase. Afterwards, the followings keys
will be set:
/data point/x will contain the current point’s x coordinate.
/data point/y will contain the current point’s y coordinate.
/data point/z will contain the current point’s z coordinate (if applicable).
/data point/meta will contain the current point’s point meta value (if applicable).
/data point/index will contain the current point’s index in the coordinate stream. This is actually
the same as \coordindex.
The returned value are the same as they can be read on the axes, they are also the same as you would
write them into axis cs.
This means that any x coord inv trafo has been applied on the value. It also means that the
exponential function has been called even though the internal coordinate was present in log format.
This function is implicitly called for any scatter plot (including nodes near coords).
This allows to access all coordinate values at once:
8.6. ACCESSING POINT COORDINATE VALUES 527
(90, 1)
(45, 0.71)
(135, 0.71) \begin{axis}
\addplot+[
domain=0:360,
samples=9,
(0, 0) (180, 0) only marks,
0 nodes near coords={%
(360, 0)
\footnotesize
\pgfmathprintnumber
(225, 0.71)
(315, 0.71)
1 {\pgfkeysvalueof{/data point/y}})$%
(270, 1) },
0 100 200 300 ]
{sin(x)};
\end{axis}
\end{tikzpicture}
Note that the preceding example would have been simpler if we would have printed just one value:
nodes near coords resorts to the point meta. And that, in turn, contains the y coordinate anyway
by default.
A more advanced example would be a ybar plot in which nodes shall be placed at the lower end of the
axis, together with some dotted lines to the respective bars:
\begin{axis}[
ybar,
nodes near coords,
%
% we want to provide absolute ’at’ values
0 % for the nodes:
scatter/position=absolute,
every node near coord/.style={
at={(\pgfkeysvalueof{/data point/x},-1)},
% pretty printing:
1 anchor=north,
0 0.7 1 0.7 0 0.7 1 0.7 0 /pgf/number format/fixed,
0 45 90 135 180 225 270 315 360 % assign a name which can be referenced below:
name=NNC\pgfkeysvalueof{/data point/index},
},
% ... draw a dotted line between
% the marker and the bar:
/pgfplots/scatter/@post marker code/.add code={}{
\draw[dotted,help lines]
(NNC\pgfkeysvalueof{/data point/index})
--
(\pgfkeysvalueof{/data point/x},
{min(0,\pgfkeysvalueof{/data point/y})});
},
% assign suitable tick labels:
xtick=data,
]
% some dummy data:
\addplot+[
domain=0:360,
bar width=360/9,
samples=9]
{sin(x)};
\end{axis}
\end{tikzpicture}
\pgfplotspointgetnormalizedcoordinates
A macro which is very similar to \pgfplotspointgetcoordinates. Consequently, it is supposed to be
called during the visualization phase.
It assigns the very same output macros, but the values are di↵erent. More precisely, it defines the
macros
/data point/x will contain the current point’s normalized x coordinate.

/data point/y will contain the current point’s normalized y coordinate.
/data point/z will contain the current point’s normalized z coordinate (if applicable).
/data point/meta will contain the current point’s point meta value (if applicable).
/data point/index will contain the current point’s index in the coordinate stream. This is actually
the same as \coordindex.
The keyword normalized means that the values are in a suitable numerical form which can be consumed
by the axis. To be more specific: any user x coord inv trafo is ignored. An important example
would be symbolic x coords: the normalized coordinates would be some associated numbers, not
the symbols. The results returned by \pgfplotspointgetcoordinates would be the symbols. For
logarithm axes, the normalized values are the logs.
Typically, normalized values are much more useful when you want to apply some math operation like
averaging or subtraction.
This function needs to be called explicitly. It is currently used by ybar stacked to align nodes near
coords.
8.7 Layer Access

\pgfplotsonlayer{hlayer namei}
A low-level command which will check if the current axis has layer support activated and, if so, calls
\pgfonlayer{hlayer namei}.
There must be a \endpgfplotsonlayer to delimit the environment.
\endpgfplotsonlayer
The end of \pgfplotsonlayer.
\pgfonlayer{hlayer namei}
A low-level command of pgf which will collect everything until the matching \endpgfonlayer into
layer hlayer namei.
The hlayer namei must be active, i.e. it must be part of the layer names of set layers.
The only special case is if you call \pgfdeclarelayer{discard} somewhere: this special layer has a
“magical name” which serves as /dev/null if it is enabled using \pgfonlayer{discard}: it does not
need to be active and everything assigned to this layer will be thrown away if it is not part of the layer
name configuration.
There must be a \endpgfonlayer to delimit the environment.
\endpgfonlayer
The end of \pgfonlayer.
\pgfsetlayers{hlayer listi}
This is a low-level command of pgf. At the time of this writing, it is the only way to tell pgf which
layers it shall use for the current / next picture. It is used implicitly by set layers.
8.7. LAYER ACCESS 529
Index
— Symbols — width . . . . . . . . . . . . . . . . . . . . . . . . . . . . 397
plot (hx expressioni,hy expressioni) . . . . . . . . . 55 xmax . . . . . . . . . . . . . . . . . . . . . . . . . . . . 397
(hx expressioni,hy expressioni,hz expressioni) . 127 xmin . . . . . . . . . . . . . . . . . . . . . . . . . . . . 397
-- path operation . . . . . . . . . . . . . . . . . . . . . . 522 ymax . . . . . . . . . . . . . . . . . . . . . . . . . . . . 397
.define key . . . . . . . . . . . . . . . . . . . . . . . . . . . 203 ymin . . . . . . . . . . . . . . . . . . . . . . . . . . . . 397
.style key . . . . . . . . . . . . . . . . . . . . . . . 430, 432 .append style handler . . . . . . . . . . . . . . . . . . 340
/data point/x . . . . . . . . . . . . . . . . . . . . . . . . . 374 area cycle list key . . . . . . . . . . . . . . . . . . . . 100
plot {hmath expressioni} . . . . . . . . . . . . . . . . . 53 area legend key . . . . . . . . . . . . . . . . . . . . . . . 242
{hmath expressioni} . . . . . . . . . . . . . . . . . . . . . 126 area style key . . . . . . . . . . . . . . . . . . . . . . . . 100
3d box key . . . . . . . . . . . . . . . . . . . . . . . . . . . 291 array
3d box foreground style key . . . . . . . . . . . . . 343 Array Alignment . . . . . . . . . . . . . . . . . . . . 353
3d view at key . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 348
Assign rotation angles . . . . . . . . . . . . . . . . 287 at position key . . . . . . . . . . . . . . . . . . . . . . . 231
Assign unit vectors . . . . . . . . . . . . . . . . . . 272 \autoplotspeclist . . . . . . . . . . . . . . . . . . . . . 521
3d: line or mesh . . . . . . . . . . . . . . . . . . . . . . . . 123 autumn key . . . . . . . . . . . . . . . . . . . . . . . . . . . 402
aux in dpth key . . . . . . . . . . . . . . . . . . . . . . . 515
—A— average key . . . . . . . . . . . . . . . . . . . . . . . 472, 480
Accuracy axis environment . . . . . . . . . . . . . . . . . . . . . . . . 40
Data Transformation . . . . . . . . . . . . . . . . . 377 axis background key . . . . . . . . . . . . . . . . . . . . 210
Floating Point in pgfplots . . . . . . . . . . . . . 44 axis base prefix key . . . . . . . . . . . . . . . . . . . 501
High Precision for Plot Expression . . . . . . . . 54 axis cs coordinate system . . . . . . . . . . . . . . . . 331
\addlegendentry . . . . . . . . . . . . . . . . . . . . . . . 234 axis description cs coordinate system . . . . . . 220
\addlegendentryexpanded . . . . . . . . . . . . . . . . 235 axis direction cs coordinate system . . . . . . . 332
\addlegendimage . . . . . . . . . . . . . . . . . . . . . . . 249 axis equal key . . . . . . . . . . . . . . . . . . . . . . . . 276
\addplot . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43 axis equal image key . . . . . . . . . . . . . . . . . . . 277
\addplot3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123 axis line shift key . . . . . . . . . . . . . . . . . . . . 255
after arrow key . . . . . . . . . . . . . . . . . . . . . . . . 93 axis line style key . . . . . . . . . . . . . . . . 253, 343
after end axis key . . . . . . . . . . . . . . . . . . . . . 380 axis lines key . . . . . . . . . . . . . . . . . . . . . . . . 250
Alignment axis lines* key . . . . . . . . . . . . . . . . . . . . . . . 250
Array . . . . . . . . . . . . . . . . . . . . . . . . .... 353 axis on top key . . . . . . . . . . . . . . . . . . . 381, 386
Subplots . . . . . . . . . . . . . . . . . . . . . . .... 353 axis x discontinuity key . . . . . . . . . . . . . . . 258
allow reversal of rel axis cs . . . . . . . . . . . . .... 525 axis x line key . . . . . . . . . . . . . . . . . . . . . . . 250
allow reversal of rel axis cs key . . . . .... 335 axis x line shift key . . . . . . . . . . . . . . . . . . 255
allow upside down key . . . . . . . . . . . . . . 231, 337 axis x line* key . . . . . . . . . . . . . . . . . . . . . . 250
anchor key . . . . . . . . . . . . . . . . . . . . . . . .... 348 axis y discontinuity key . . . . . . . . . . . . . . . 258
Anchors axis y line key . . . . . . . . . . . . . . . . . . . . . . . 250
near ticklabel . . . . . . . . . . . . . . . . . . . . 226 axis y line shift key . . . . . . . . . . . . . . . . . . 255
near ticklabel opposite . . . . . . . . . . . . 228 axis y line* key . . . . . . . . . . . . . . . . . . . . . . 250
near xticklabel . . . . . . . . . . . . . . . . . . . 226 axis z discontinuity key . . . . . . . . . . . . . . . 258
near xticklabel opposite . . . . . . . . . . . 228 axis z line shift key . . . . . . . . . . . . . . . . . . 255
near yticklabel . . . . . . . . . . . . . . . . . . . 226 \axisdefaultheight . . . . . . . . . . . . . . . . . . . . 272
near yticklabel opposite . . . . . . . . . . . 228 \axisdefaultwidth . . . . . . . . . . . . . . . . . . . . . 272
near zticklabel . . . . . . . . . . . . . . . . . . . 226 az key . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 288
near zticklabel opposite . . . . . . . . . . . 228
annot/ —B—
collected plots . . . . . . . . . . . . . . . . . . . 397 bar cycle list key . . . . . . . . . . . . . . . . . . . . . . 85
font . . . . . . . . . . . . . . . . . . . . . . . . . . . . 396 bar direction key . . . . . . . . . . . . . . . . . . . . . . 87
height . . . . . . . . . . . . . . . . . . . . . . . . . . . 397 bar nodes key . . . . . . . . . . . . . . . . . . . . . . . . . . 10
js fillColor . . . . . . . . . . . . . . . . . . . . . . 395 Bar Plots
jsname . . . . . . . . . . . . . . . . . . . . . . . . . . . 397 Skewed axes problems . . . . . . . . . . . . . . . . 275
point format . . . . . . . . . . . . . . . . . . . . . . 395 bar shift key . . . . . . . . . . . . . . . . . . . . . . . . . . 86
point format 3d . . . . . . . . . . . . . . . . . . . 395 bar shift auto key . . . . . . . . . . . . . . . . . . . . . . 83
popup size . . . . . . . . . . . . . . . . . . . . . . . 396 bar width key . . . . . . . . . . . . . . . . . . . . . . . . . . 86
popup size generic . . . . . . . . . . . . . . . . . 396 bar width by units key . . . . . . . . . . . . . . . . . . 10
popup size snap . . . . . . . . . . . . . . . . . . . 396 baseline key . . . . . . . . . . . . . . . . . . . . . . . . . . 351
printable . . . . . . . . . . . . . . . . . . . . . . . . 396 BB key . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
richtext . . . . . . . . . . . . . . . . . . . . . . . . . 396 before arrow key . . . . . . . . . . . . . . . . . . . . . . . 93
slope format . . . . . . . . . . . . . . . . . . . . . . 396 before end axis key . . . . . . . . . . . . . . . . . . . . 380
snap dist . . . . . . . . . . . . . . . . . . . . . . . . 396 Behavior Options . . . . . . . . . . . . . . . . . . . . . . . . 43
textSize . . . . . . . . . . . . . . . . . . . . . . . . . 396 bins key . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 482
blackwhite key . . . . . . . . . . . . . . . . . . . . . . . . 190 clickable coords key . . . . . . . . . . . . . . . . . . . 392

bled key . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 402 clickable coords code key . . . . . . . . . . . . . . . 393
bluered key . . . . . . . . . . . . . . . . . . . . . . . . . . . 190 clickable coords size key . . . . . . . . . . . . . . . 394
bone key . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 402 clip key . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 360
Bounding Box Control . . . . . . . . . . . . . . . . . . . 359 clip bounding box key . . . . . . . . . . . . . . . . . . 360
clip . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 360 clip limits key . . . . . . . . . . . . . . . . . . . . . . . 307
Disable data bounding box modifications . . 310 clip marker paths key . . . . . . . . . . . . . . . . . . 360
Excluding Image Parts . . . . . . . . . . . . . . . 357 clip mode key . . . . . . . . . . . . . . . . . . . . . . . . . 361
hide axis . . . . . . . . . . . . . . . . . . . . . 259, 357 closedcycle
Image Externalization Problems . . . . . . . . 513 Mesh or patch plots . . . . . . . . . . . . . . . . . 437
pgfinterruptboundingbox . . . . . . . . . . . . . . 359 \closedcycle . . . . . . . . . . . . . . . . . . . . . . . . . 105
box key . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 479 cmd key . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
box extend key . . . . . . . . . . . . . . . . . . . . . . . . 475 .code handler . . . . . . . . . . . . . . . . . . . . . . . . . . 76
boxplot/ .code 2 args handler . . . . . . . . . . . . . . . . . . . . 76
average . . . . . . . . . . . . . . . . . . . . . . . . . . 472 col sep key . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
box extend . . . . . . . . . . . . . . . . . . . . . . . 475 cold key . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 403
data . . . . . . . . . . . . . . . . . . . . . . . . . . . . 472 collected plots key . . . . . . . . . . . . . . . . . . . . 397
draw/ color key . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
average . . . . . . . . . . . . . . . . . . . . . . . . 480 color burning . . . . . . . . . . . . . . . . . . . . . . . . . . 149
box . . . . . . . . . . . . . . . . . . . . . . . . . . . . 479 color input key . . . . . . . . . . . . . . . . . . . . . . . 142
lower whisker . . . . . . . . . . . . . . . . . . . 479 color of colormap
median . . . . . . . . . . . . . . . . . . . . . . . . . 480 In colormap specification . . . . . . . . . . . . . . 192
upper whisker . . . . . . . . . . . . . . . . . . . 479 color of colormap key . . . . . . . . . . . . . . . . . . 193
whisker . . . . . . . . . . . . . . . . . . . . . . . . 479 colorbar/
draw direction . . . . . . . . . . . . . . . . . . . . 472 draw . . . . . . . . . . . . . . . . . . . . . . . . . . . . 267
draw position . . . . . . . . . . . . . . . . . . . . . 470 width . . . . . . . . . . . . . . . . . . . . . . . . . . . . 266
draw relative anchor . . . . . . . . . . . . . . . 475 colorbar key . . . . . . . . . . . . . . . . . . . . . . . . . . 260
estimator . . . . . . . . . . . . . . . . . . . . . . . . 477 colorbar as palette key . . . . . . . . . . . . . . . . 216
every average . . . . . . . . . . . . . . . . . . . . . 477 colorbar horizontal key . . . . . . . . . . . . . . . . 263
every box . . . . . . . . . . . . . . . . . . . . . . . . 477 colorbar left key . . . . . . . . . . . . . . . . . . . . . 263
every boxplot . . . . . . . . . . . . . . . . . . . . . 477 colorbar right key . . . . . . . . . . . . . . . . . . . . . 262
every median . . . . . . . . . . . . . . . . . . . . . . 477 colorbar sampled key . . . . . . . . . . . . . . . . . . . 267
every whisker . . . . . . . . . . . . . . . . . . . . . 477 colorbar sampled line key . . . . . . . . . . . . . . . 268
lower quartile . . . . . . . . . . . . . . . . . . . . 472 colorbar sampled line style key . . . . . . . . . . 344
lower whisker . . . . . . . . . . . . . . . . . . . . . 472 colorbar shift key . . . . . . . . . . . . . . . . . . . . . 267
median . . . . . . . . . . . . . . . . . . . . . . . . . . . 472 colorbar source key . . . . . . . . . . . . . . . . . . . . 265
sample size . . . . . . . . . . . . . . . . . . . . . . . 474 colorbar style key . . . . . . . . . . . . . . . . . 266, 343
sample size max . . . . . . . . . . . . . . . . . . . 475 colorbar to name key . . . . . . . . . . . . . . . . . . . 269
sample size min . . . . . . . . . . . . . . . . . . . 474 colorbrewer library . . . . . . . . . . . . . . . . . 397, 401
upper quartile . . . . . . . . . . . . . . . . . . . . 472 colored key . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
upper whisker . . . . . . . . . . . . . . . . . . . . . 472 Colormap
variable width . . . . . . . . . . . . . . . . . . . . 473 Convert into Cycle list . . . . . . . . . . . . . . . 200
variable width expr . . . . . . . . . . . . . . . . 474 colormap/
variable width min target . . . . . . . . . . . 475 autumn . . . . . . . . . . . . . . . . . . . . . . . .... 402
whisker extend . . . . . . . . . . . . . . . . . . . . 475 blackwhite . . . . . . . . . . . . . . . . . . . .... 190
whisker range . . . . . . . . . . . . . . . . . . . . . 477 bled . . . . . . . . . . . . . . . . . . . . . . . . .... 402
boxplot key . . . . . . . . . . . . . . . . . . . . . . . . . . . 476 bluered . . . . . . . . . . . . . . . . . . . . . . .... 190
boxplot box coordinate system . . . . . . . . . . . . 478 bone . . . . . . . . . . . . . . . . . . . . . . . . .... 402
boxplot prepared key . . . . . . . . . . . . . . . . . . . 469 bright . . . . . . . . . . . . . . . . . . . . . . . .... 402
boxplot whisker coordinate system . . . . . . . . . 478 cold . . . . . . . . . . . . . . . . . . . . . . . . .... 403
\boxplotvalue . . . . . . . . . . . . . . . . . . . . . . . . 478 cool . . . . . . . . . . . . . . . . . . . . . . . . .... 190
bright key . . . . . . . . . . . . . . . . . . . . . . . . . . . 402 copper . . . . . . . . . . . . . . . . . . . . . . . .... 403
copper2 . . . . . . . . . . . . . . . . . . . . . . .... 403
—C— earth . . . . . . . . . . . . . . . . . . . . . . . . .... 403
cartesian cs coordinate system . . . . . . . . . . . . 495 gray . . . . . . . . . . . . . . . . . . . . . . . . .... 403
.cd handler . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77 greenyellow . . . . . . . . . . . . . . . . . . . .... 191
cell picture key . . . . . . . . . . . . . . . . . . . . . . 389 hot . . . . . . . . . . . . . . . . . . . . . . . . . .... 189
change x base key . . . . . . . . . . . . . . . . . . . . . 500 hot2 . . . . . . . . . . . . . . . . . . . . . . . . 189, 404
change y base key . . . . . . . . . . . . . . . . . . . . . 500 hsv . . . . . . . . . . . . . . . . . . . . . . . . . .... 404
change z base key . . . . . . . . . . . . . . . . . . . . . 500 hsv2 . . . . . . . . . . . . . . . . . . . . . . . . .... 404
check key . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134 jet . . . . . . . . . . . . . . . . . . . . . . . . . 190, 404
classes key . . . . . . . . . . . . . . . . . . . . . . . . . . . 112 pastel . . . . . . . . . . . . . . . . . . . . . . . .... 405
clickable key . . . . . . . . . . . . . . . . . . . . . . . . . 395 pink . . . . . . . . . . . . . . . . . . . . . . . . .... 405
clickable library . . . . . . . . . . . . . . . . . . . . . . 391
redyellow . . . . . . . . . . . . . . . . . . . . . . . . 191 contour label style key . . . . . . . . . . . . . . . . 159

sepia . . . . . . . . . . . . . . . . . . . . . . . . . . . . 405 contour prepared key . . . . . . . . . . . . . . . . . . . 156
spring . . . . . . . . . . . . . . . . . . . . . . . . . . . 405 contour prepared format key . . . . . . . . . . . . . 156
summer . . . . . . . . . . . . . . . . . . . . . . . . . . . 405 cool key . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190
temp . . . . . . . . . . . . . . . . . . . . . . . . . . . . 406 Coordinate systems
thermal . . . . . . . . . . . . . . . . . . . . . . . . . . 406 axis cs . . . . . . . . . . . . . . . . . . . . . . . . . . 331
violet . . . . . . . . . . . . . . . . . . . . . . . . . . . 191 axis description cs . . . . . . . . . . . . . . . . 220
viridis . . . . . . . . . . . . . . . . . . . . . . . . . . 189 axis direction cs . . . . . . . . . . . . . . . . . 332
viridis high res . . . . . . . . . . . . . . . . . . 406 boxplot box . . . . . . . . . . . . . . . . . . . . . . . 478
winter . . . . . . . . . . . . . . . . . . . . . . . . . . . 406 boxplot whisker . . . . . . . . . . . . . . . . . . . 478
colormap key . . . . . . . . . . . . . . . . . . . . . . . . . . 186 cartesian cs . . . . . . . . . . . . . . . . . . . . . . 495
colormap access key . . . . . . . . . . . . . . . . . . . . 216 rel axis cs . . . . . . . . . . . . . . . . . . . . . . . 334
colormap default colorspace key . . . . . . . . . . 188 ticklabel cs . . . . . . . . . . . . . . . . . . . . . . 222
colormap name key . . . . . . . . . . . . . . . . . . . . . 186 ticklabel* cs . . . . . . . . . . . . . . . . . . . . . 222
colormaps library . . . . . . . . . . . . . . . . . . . . . . 402 xticklabel cs . . . . . . . . . . . . . . . . . . . . . 222
colors of colormap xticklabel* cs . . . . . . . . . . . . . . . . . . . . 222
In colormap specification . . . . . . . . . . . . . . 193 yticklabel cs . . . . . . . . . . . . . . . . . . . . . 222
colorspace explicit color input key . . . . . . 146 yticklabel* cs . . . . . . . . . . . . . . . . . . . . 222
colorspace explicit color output key . . . . . 146 zticklabel cs . . . . . . . . . . . . . . . . . . . . . 222
cols key . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125 zticklabel* cs . . . . . . . . . . . . . . . . . . . . 222
columns key . . . . . . . . . . . . . . . . . . . . . . . . . . . 430 plot coordinates . . . . . . . . . . . . . . . . . . . . . . . 46
comment chars key . . . . . . . . . . . . . . . . . . . . . . 52 coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . 124
compat/ \coordindex . . . . . . . . . . . . . . . . . . . . . . . 59, 521
bar nodes . . . . . . . . . . . . . . . . . . . . . . . . . 10 copper key . . . . . . . . . . . . . . . . . . . . . . . . . . . 403
bar width by units . . . . . . . . . . . . . . . . . . 10 copper2 key . . . . . . . . . . . . . . . . . . . . . . . . . . . 403
BB . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 crossref file suffix . . . . . . . . . . . . . . . . . . . . . . . 247
empty line . . . . . . . . . . . . . . . . . . . . . . . . 10 cube/
general . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 size x . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
labels . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 size y . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
path replacement . . . . . . . . . . . . . . . . . . . 10 size z . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
pgfpoint substitution . . . . . . . . . . . . . . . 10 cumulative key . . . . . . . . . . . . . . . . . . . . . . . . 483
plot3graphics . . . . . . . . . . . . . . . . . . . . . . 10 current axis node . . . . . . . . . . . . . . . . . . . . . 357
scale mode . . . . . . . . . . . . . . . . . . . . . . . . 10 current colorbar axis node . . . . . . . . . . . . . . 267
scaling . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 current plot begin node . . . . . . . . . . . . . . . . 335
show suggested version . . . . . . . . . . . . . 390 current plot end node . . . . . . . . . . . . . . . . . . 335
compat key . . . . . . . . . . . . . . . . . . . . . . . 10, 232f. curve style key . . . . . . . . . . . . . . . . . . . . . . . 498
const plot key . . . . . . . . . . . . . . . . . . . . . . . . . 78 cycle from colormap manual style key . . . . . 202
const plot mark left key . . . . . . . . . . . . . . . . 79 cycle list/
const plot mark mid key . . . . . . . . . . . . . . . . . 79 .define . . . . . . . . . . . . . . . . . . . . . . . . . . 203
const plot mark right key . . . . . . . . . . . . . . . . 79 cycle list key . . . . . . . . . . . . . . . . . . . . . . . . 194
contour/ cycle list name key . . . . . . . . . . . . . . . . . . . . 194
contour dir . . . . . . . . . . . . . . . . . . . . . . . 153 cycle list shift key . . . . . . . . . . . . . . . . . . . 200
contour label style . . . . . . . . . . . . . . . . 159 cycle multi list key . . . . . . . . . . . . . . . . . . . 203
draw color . . . . . . . . . . . . . . . . . . . . . . . 158 cycle multiindex list key . . . . . . . . . . . . . . . 206
every contour label . . . . . . . . . . . . . . . . 158 cycle multiindex* list key . . . . . . . . . . . . . . 208
every contour plot . . . . . . . . . . . . . . . . . 158
handler . . . . . . . . . . . . . . . . . . . . . . . . . . 159 —D—
label distance . . . . . . . . . . . . . . . . . . . . 158 dashdotdotted key . . . . . . . . . . . . . . . . . .... 182
label node code . . . . . . . . . . . . . . . . . . . 160 dashdotted key . . . . . . . . . . . . . . . . . . . . .... 181
labels . . . . . . . . . . . . . . . . . . . . . . . . . . . 158 dashed key . . . . . . . . . . . . . . . . . . . . . . . .... 181
labels over line . . . . . . . . . . . . . . . . . . 159 data key . . . . . . . . . . . . . . . . . . . . . . . . . 472, 482
levels . . . . . . . . . . . . . . . . . . . . . . . . . . . 152 data coord inv trafo key . . . . . . . . . . . .... 486
number . . . . . . . . . . . . . . . . . . . . . . . . . . . 152 data coord trafo key . . . . . . . . . . . . . . . .... 486
contour dir key . . . . . . . . . . . . . . . . . . . . . . . 153 data cs . . . . . . . . . . . . . . . . . . . . . . . . . . .... 374
contour external/ data cs key . . . . . . . . . . . . . . . . . . . . . . . .... 371
cmd . . . . . . . . . . . . . . . . . . . . . . . . . . ... 161 data filter key . . . . . . . . . . . . . . . . . . . .... 485
file . . . . . . . . . . . . . . . . . . . . . . . . . ... 160 data max key . . . . . . . . . . . . . . . . . . . . . . .... 482
output point meta . . . . . . . . . . . . . . ... 161 data min key . . . . . . . . . . . . . . . . . . . . . . .... 482
scanline marks . . . . . . . . . . . . . . . . . ... 160 /data point/
script . . . . . . . . . . . . . . . . . . . . . . . . ... 160 x ................ . . . . . . . . . . . . . . . 338
script extension . . . . . . . . . . . . . . . ... 161 y ................ . . . . . . . . . . . . . . . 338
contour external key . . . . . . . . . . . . . . . . ... 160 z ................ . . . . . . . . . . . . . . . 338
contour gnuplot key . . . . . . . . . . . . . . . . 150, 161 date coordinates in key . . . . . . . . . . . . . . . . 364
date ZERO key . . . . . . . . . . . . . . . . . . . . . . . . . 366
dateplot library . . . . . . . . . . . . . . . . . . . 364, 407 x explicit relative . . . . . . . . . . . . . . . . 296

debug key . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73 x fixed . . . . . . . . . . . . . . . . . . . . . . . . . . 295
declare function key . . . . . . . . . . . . . . . . . . . 520 x fixed relative . . . . . . . . . . . . . . . . . . 295
decorations.softclip library . . . . . . . . . . . . . 407 y dir . . . . . . . . . . . . . . . . . . . . . . . . . . . . 294
default key . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96 y explicit . . . . . . . . . . . . . . . . . . . . . . . 295
default smithchart xtick key . . . . . . . . . . . . 463 y explicit relative . . . . . . . . . . . . . . . . 296
default smithchart xytick key . . . . . . . . . . . 463 y fixed . . . . . . . . . . . . . . . . . . . . . . . . . . 295
default smithchart ytick key . . . . . . . . . . . . 463 y fixed relative . . . . . . . . . . . . . . . . . . 295
default typeset ticklabel key . . . . . . . . . . . 323 z dir . . . . . . . . . . . . . . . . . . . . . . . . . . . . 295
.define layer set handler . . . . . . . . . . . . . . . 388 z explicit . . . . . . . . . . . . . . . . . . . . . . . 295
dense smithchart ticks key . . . . . . . . . . . . . . 461 z explicit relative . . . . . . . . . . . . . . . . 296
densely dashdotdotted key . . . . . . . . . . . . . . . 182 z fixed . . . . . . . . . . . . . . . . . . . . . . . . . . 295
densely dashdotted key . . . . . . . . . . . . . . . . . 182 z fixed relative . . . . . . . . . . . . . . . . . . 295
densely dashed key . . . . . . . . . . . . . . . . . . . . . 181 error mark key . . . . . . . . . . . . . . . . . . . . . . . . 296
densely dotted key . . . . . . . . . . . . . . . . . . . . . 181 error mark options key . . . . . . . . . . . . . . . . . 296
density key . . . . . . . . . . . . . . . . . . . . . . . . . . . 483 Error Messages
disabledatascaling key . . . . . . . . . . . . . . . . . 377 No room for a new dimen . . . . . . . . . . . . . . 13
disablelogfilter key . . . . . . . . . . . . . . . . . . . 377 errorbars key . . . . . . . . . . . . . . . . . . . . . . . . . . 96
domain key . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55 Errors
domain y key . . . . . . . . . . . . . . . . . . . . . . . . . . . 55 dimension too large . . . . . . . . . . . . . . . 54, 377
dotted key . . . . . . . . . . . . . . . . . . . . . . . . . . . 181 enlargelimits respects figure size=true: could not
Downsampling . . . . . . . . . . . . . . . . . . . . . . . . . 370 respect the prescribed width/height . . . 309f.
\draw . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 521 Patch Input and Memory Problems . . . . . . 172
draw key . . . . . . . . . . . . . . . . . . . . . . . . . 185, 267 Skewed axes and bar plots . . . . . . . . . . . . . 275
draw color key . . . . . . . . . . . . . . . . . . . . . . . . 158 estimator key . . . . . . . . . . . . . . . . . . . . . . . . . 477
draw direction key . . . . . . . . . . . . . . . . . . . . . 472 every 3d box foreground key . . . . . . . . . . . . . 343
draw error bar key . . . . . . . . . . . . . . . . . . . . . 296 every 3d description key . . . . . . . . . . . . . . . 289
draw position key . . . . . . . . . . . . . . . . . . . . . 470 every 3d view {hh i}{hv i} key . . . . . . . . . . . . 290
draw relative anchor key . . . . . . . . . . . . . . . 475 every arrow key . . . . . . . . . . . . . . . . . . . . . . . . 93
every average key . . . . . . . . . . . . . . . . . . . . . 477
—E— every axis key . . . . . . . . . . . . . . . . . . . . . . . . 340
each nth point key . . . . . . . . . . ........ . . . 370 every axis grid key . . . . . . . . . . . . . . . . . . . . 346
each nth tie key . . . . . . . . . . . ........ . . . 497 every axis label key . . . . . . . . . . . . . . . . . . . 342
earth key . . . . . . . . . . . . . . . . . ........ . . . 403 every axis legend key . . . . . . . . . . . . . . 237, 342
el key . . . . . . . . . . . . . . . . . . . . ........ . . . 289 every axis plot key . . . . . . . . . . . . . . . . . . . . 341
empty legend key . . . . . . . . . . . ........ . . . 242 every axis plot no # key . . . . . . . . . . . . . . . 341
empty line key . . . . . . . . . . . . . ........ . 10, 45 every axis plot post key . . . . . . . . . . . 181, 341
\endpgfonlayer . . . . . . . . . . . . . ........ . . . 528 every axis post key . . . . . . . . . . . . . . . . . . . . 340
\endpgfplotsonlayer . . . . . . . . ........ . . . 528 every axis title key . . . . . . . . . . . . . . . . . . . 342
enlarge x limits key . . . . . . . . ........ . . . 307 every axis title shift key . . . . . . . . . . . . . . 342
enlarge y limits key . . . . . . . . ........ . . . 307 every axis x grid key . . . . . . . . . . . . . . . . . . 347
enlarge z limits key . . . . . . . . ........ . . . 307 every axis x label key . . . . . . . . . . . . . . . . . 342
enlargelimits key . . . . . . . . . . ........ . . . 307 every axis y grid key . . . . . . . . . . . . . . . . . . 347
enlargelimits respects figure size key . . . . 310 every axis y label key . . . . . . . . . . . . . . . . . 342
Environments every axis z grid key . . . . . . . . . . . . . . . . . . 347
axis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40 every axis z label key . . . . . . . . . . . . . . . . . 342
groupplot . . . . . . . . . . . . . . . . . . . . . . . . 427 every box key . . . . . . . . . . . . . . . . . . . . . . . . . 477
loglogaxis . . . . . . . . . . . . . . . . . . . . . . . . 41 every boxed x axis key . . . . . . . . . . . . . . . . . 253
pgfinterruptboundingbox . . . . . . . . . . . . 359 every boxed y axis key . . . . . . . . . . . . . . . . . 253
pgfplotsinterruptdatabb . . . . . . . . . . . . 310 every boxed z axis key . . . . . . . . . . . . . . . . . 254
polaraxis . . . . . . . . . . . . . . . . . . . . . . . . 449 every boxplot key . . . . . . . . . . . . . . . . . . . . . 477
semilogxaxis . . . . . . . . . . . . . . . . . . . . . . . 41 every colorbar key . . . . . . . . . . . . . . . . . 264, 343
semilogyaxis . . . . . . . . . . . . . . . . . . . . . . . 41 every colorbar sampled line key . . . . . . 269, 344
smithchart . . . . . . . . . . . . . . . . . . . . . . . 456 every colorbar to name picture key . . . . . . . 270
ternaryaxis . . . . . . . . . . . . . . . . . . . . . . . 487 every contour label key . . . . . . . . . . . . . . . . 158
tikzpicture . . . . . . . . . . . . . . . . . . . . . . . . 40 every contour plot key . . . . . . . . . . . . . . . . . 158
error bar style key . . . . . . . . . . . . . . . . 296, 347 every crossref picture key . . . . . . . . . . . . . . 247
error bars/ every error bar key . . . . . . . . . . . . . . . . . . . . 347
draw error bar . . . . . . . . . . . . . . . . .... 296 every even segment key . . . . . . . . . . . . . . . . . 415
error bar style . . . . . . . . . . . . . . . 296, 347 every extra x tick key . . . . . . . . . . . . . . . . . 346
error mark . . . . . . . . . . . . . . . . . . . .... 296 every extra y tick key . . . . . . . . . . . . . . . . . 346
error mark options . . . . . . . . . . . . . .... 296 every extra z tick key . . . . . . . . . . . . . . . . . 346
x dir . . . . . . . . . . . . . . . . . . . . . . . . .... 294 every fill between plot key . . . . . . . . . . . . . 416
x explicit . . . . . . . . . . . . . . . . . . . .... 295
every forget plot key . . . . . . . . . . . . . . . . . . 341 expression . . . . . . . . . . . . . . . . . . . . . . . . . . . 127

every inner x axis line key . . . . . . . . . 253, 343 external library . . . . . . . . . . . . . . . . . . . 407, 508
every inner y axis line key . . . . . . . . . 253, 343 External Graphics
every inner z axis line key . . . . . . . . . 253, 343 Bounding Box Issues . . . . . . . . . . . . . . . . . 513
every last segment key . . . . . . . . . . . . . . . . . 415 extra description key . . . . . . . . . . . . . . . . . . 234
every legend image post key . . . . . . . . . 241, 343 extra tick style key . . . . . . . . . . . . . . . . . . . 346
every legend to name picture key . . . . . 249, 343 extra x tick label key . . . . . . . . . . . . . . . . . 320
every linear axis key . . . . . . . . . . . . . . . . . . 341 extra x tick labels key . . . . . . . . . . . . . . . . 320
every loglog axis key . . . . . . . . . . . . . . . . . . 341 extra x tick style key . . . . . . . . . . . . . . . . . 346
every major grid key . . . . . . . . . . . . . . . . . . . 347 extra x ticks key . . . . . . . . . . . . . . . . . . . . . 315
every major tick key . . . . . . . . . . . . . . . . . . . 344 extra y tick label key . . . . . . . . . . . . . . . . . 320
every major x grid key . . . . . . . . . . . . . . . . . 347 extra y tick labels key . . . . . . . . . . . . . . . . 320
every major x tick key . . . . . . . . . . . . . . . . . 345 extra y tick style key . . . . . . . . . . . . . . . . . 346
every major y grid key . . . . . . . . . . . . . . . . . 347 extra y ticks key . . . . . . . . . . . . . . . . . . . . . 315
every major y tick key . . . . . . . . . . . . . . . . . 345 extra z tick label key . . . . . . . . . . . . . . . . . 320
every major z grid key . . . . . . . . . . . . . . . . . 347 extra z tick labels key . . . . . . . . . . . . . . . . 320
every major z tick key . . . . . . . . . . . . . . . . . 345 extra z tick style key . . . . . . . . . . . . . . . . . 346
every mark key . . . . . . . . . . . . . . . . . . . . . . . . 178 extra z ticks key . . . . . . . . . . . . . . . . . . . . . 315
every median key . . . . . . . . . . . . . . . . . . . . . . 477
every minor grid key . . . . . . . . . . . . . . . . . . . 346 —F—
every minor tick key . . . . . . . . . . . . . . . . . . . 344 faceted color key . . . . . . . . . . . . . . . . . . . . . 140
every minor x grid key . . . . . . . . . . . . . . . . . 347 few smithchart ticks key . . . . . . . . . . . . . . . 458
every minor x tick key . . . . . . . . . . . . . . . . . 345 figure name key . . . . . . . . . . . . . . . . . . . . . . . 510
every minor y grid key . . . . . . . . . . . . . . . . . 347 plot file . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
every minor y tick key . . . . . . . . . . . . . . . . . 345 file . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
every minor z grid key . . . . . . . . . . . . . . . . . 347 file key . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160
every minor z tick key . . . . . . . . . . . . . . . . . 345 \fill . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 522
every node near coord key . . . . . . . . . . . . . . . 116 fill key . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
every non boxed x axis key . . . . . . . . . . . . . . 254 plot fill between . . . . . . . . . . . . . . . . . . . . . 408
every non boxed y axis key . . . . . . . . . . . . . . 254 fillbetween library . . . . . . . . . . . . . . . . . . . . . 408
every non boxed z axis key . . . . . . . . . . . . . . 254 filter discard warning key . . . . . . . . . . . . . . 371
every odd segment key . . . . . . . . . . . . . . . . . . 415 filter point . . . . . . . . . . . . . . . . . . . . . . . . . . . . 374
every outer x axis line key . . . . . . . . . 253, 343 filter point key . . . . . . . . . . . . . . . . . . . . . . 367
every outer y axis line key . . . . . . . . . 253, 343 Floating Point Unit . . . . . . . . . . . . . . . . . . . . . . 54
every outer z axis line key . . . . . . . . . 253, 343 font key . . . . . . . . . . . . . . . . . . . . . . 183, 282, 396
every patch key . . . . . . . . . . . . . . . . . . . . . . . 176 footnotesize key . . . . . . . . . . . . . . . . . . . . . . 283
every segment key . . . . . . . . . . . . . . . . . . . . . 415 \foreach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 518
every segment no hindex i key . . . . . . . . . . . . 415 forget plot key . . . . . . . . . . . . . . . . . . . . . . . 379
every semilogx axis key . . . . . . . . . . . . . . . . 341 forget plot style key . . . . . . . . . . . . . . . . . . 342
every semilogy axis key . . . . . . . . . . . . . . . . 341 fpu key . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 382
every smithchart axis key . . . . . . . . . . . . . . . 466 plot function . . . . . . . . . . . . . . . . . . . . . . . . . 60
every soft clipped path key . . . . . . . . . . . . . 407
every ternary axis key . . . . . . . . . . . . . . . . . 495 —G—
every tick key . . . . . . . . . . . . . . . . . . . . . . . . 344 general key . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
every tick label key . . . . . . . . . . . . . . . . . . . 344 plot gnuplot . . . . . . . . . . . . . . . . . . . . . . . . . . 59
every whisker key . . . . . . . . . . . . . . . . . . . . . 477 plot graphics . . . . . . . . . . . . . . . . . . . . . . . . . 62
every x tick key . . . . . . . . . . . . . . . . . . . . . . 345 gray key . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 403
every x tick label key . . . . . . . . . . . . . . . . . 344 greenyellow key . . . . . . . . . . . . . . . . . . . . . . . 191
every x tick scale label key . . . . . . . . . . . . 344 grid key . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 330
every y tick key . . . . . . . . . . . . . . . . . . . . . . 345 grid style key . . . . . . . . . . . . . . . . . . . . . . . . 346
every y tick label key . . . . . . . . . . . . . . . . . 344 group/
every y tick scale label key . . . . . . . . . . . . 344 columns . . . . . . . . . . . . . . . . . . . . . . . . . . 430
every z tick key . . . . . . . . . . . . . . . . . . . . . . 345 empty plot/
every z tick label key . . . . . . . . . . . . . . . . . 344 .style . . . . . . . . . . . . . . . . . . . . . . . . . 432
every z tick scale label key . . . . . . . . . . . . 345 every plot/
execute at begin axis key . . . . . . . . . . . . . . . 378 .style . . . . . ... . . . . . . . . . . . . . . . . . 430
execute at begin plot key . . . . . . . . . . . . . . . 378 group name . . . ... . . . . . . . . . . . . . . . . . 432
execute at begin plot visualization key . . . 378 group size . . . ... . . . . . . . . . . . . . . . . . 430
execute at end axis key . . . . . . . . . . . . . . . . 378 horizontal sep ... . . . . . . . . . . . . . . . . . 430
execute at end plot key . . . . . . . . . . . . . . . . 378 rows . . . . . . . . ... . . . . . . . . . . . . . . . . . 430
execute at end plot visualization key . . . . 379 vertical sep . . ... . . . . . . . . . . . . . . . . . 430
execute for upside down key . . . . . . . . . . . . . 231 x descriptions at . . . . . . . . . . . . . . . . . 431
plot expression . . . . . . . . . . . . . . . . . . . . . . . . 55 xlabels at . . . ... . . . . . . . . . . . . . . . . . 430
xticklabels at ... . . . . . . . . . . . . . . . . . 431
y descriptions at . . . . . . . . . . . . . . . . . 431 intersection segments key . . . . . . . . . . . . . . . 420

ylabels at . . . . . . . . . . . . . . . . . . . . . . . 430 intervals key . . . . . . . . . . . . . . . . . . . . . . . . . 482
yticklabels at . . . . . . . . . . . . . . . . . . . . 431 invoke after crossref tikzpicture key . . . . 247
Group library invoke before crossref tikzpicture key . . . . 247
Subplots . . . . . . . . . . . . . . . . . . . . . . . . . . 427 is smithchart cs key . . . . . . . . . . . . . . . . . . . 463
group name key . . . . . . . . . . . . . . . . . . . . . . . . 432
group size key . . . . . . . . . . . . . . . . . . . . . . . . 430 —J—
group style key . . . . . . . . . . . . . . . . . . . . . . . 430 jet key . . . . . . . . . . . . . . . . . . . . . . . . . . 190, 404
groupplot environment . . . . . . . . . . . . . . . . . . 427 js fillColor key . . . . . . . . . . . . . . . . . . . . . . 395
groupplots library . . . . . . . . . . . . . . . . . . . . . . 427 jsname key . . . . . . . . . . . . . . . . . . . . . . . . . . . 397
jump mark left key . . . . . . . . . . . . . . . . . . . . . . 80
—H— jump mark mid key . . . . . . . . . . . . . . . . . . . . . . 80
h key . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 288 jump mark right key . . . . . . . . . . . . . . . . . . . . . 80
handler key . . . . . . . . . . . . . . . . . . . . . . . 159, 484
header key . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50 —K—
height key . . . . . . . . . . . . . . . . . . . . . . . 272, 397 Key handlers
hide axis key . . . . . . . . . . . . . . . . . . . . . . . . . 259 .append style .... . . . . . . . . . . . . . . . . . 340
hide obscured x ticks key . . . . . . . . . . . . . . . 329 .cd . . . ..... .... . . . . . . . . . . . . . . . . . . 77
hide obscured y ticks key . . . . . . . . . . . . . . . 329 .code . . ..... .... . . . . . . . . . . . . . . . . . . 76
hide obscured z ticks key . . . . . . . . . . . . . . . 329 .code 2 args . .... . . . . . . . . . . . . . . . . . . 76
hide x axis key . . . . . . . . . . . . . . . . . . . . . . . 259 .define layer set . . . . . . . . . . . . . . . . . 388
hide y axis key . . . . . . . . . . . . . . . . . . . . . . . 259 .style . ..... .... . . . . . . . . . . . . . . . . . 340
hide z axis key . . . . . . . . . . . . . . . . . . . . . . . 259
hist/ —L—
bins . . . . . . . . . . . . . . . . . . . . . . . . .... 482 \label . . . . . . . . . . . . . . . . . . . . . . . . . . . ... 246
cumulative . . . . . . . . . . . . . . . . . . . .... 483 label distance key . . . . . . . . . . . . . . . . . . ... 158
data . . . . . . . . . . . . . . . . . . . . . . . . .... 482 label node code key . . . . . . . . . . . . . . . . . ... 160
data coord inv trafo . . . . . . . . . . . .... 486 label shift key . . . . . . . . . . . . . . . . . . . . ... 232
data coord trafo . . . . . . . . . . . . . . .... 486 label style key . . . . . . . . . . . . . . . . . . . . ... 342
data filter . . . . . . . . . . . . . . . . . . . .... 485 labels key . . . . . . . . . . . . . . . . . . . . . . . . 10, 158
data max . . . . . . . . . . . . . . . . . . . . . .... 482 labels over line key . . . . . . . . . . . . . . . . ... 159
data min . . . . . . . . . . . . . . . . . . . . . .... 482 layers/
density . . . . . . . . . . . . . . . . . . . . . . .... 483 axis on top . . . . . . . . . . . . . . . . . . . .... 386
handler . . . . . . . . . . . . . . . . . . . . . . .... 484 standard . . . . . . . . . . . . . . . . . . . . . .... 386
intervals . . . . . . . . . . . . . . . . . . . . .... 482 \legend . . . . . . . . . . . . . . . . . . . . . . . . . . .... 235
symbolic coords . . . . . . . . . . . . . . . .... 486 legend cell align key . . . . . . . . . . . . . . .... 240
hist key . . . . . . . . . . . . . . . . . . . . . . . . . .... 480 legend columns key . . . . . . . . . . . . . . . . . .... 241
horizontal sep key . . . . . . . . . . . . . . . . . .... 430 legend entries key . . . . . . . . . . . . . . . . . .... 236
hot key . . . . . . . . . . . . . . . . . . . . . . . . . . .... 189 legend image code key . . . . . . . . . . . . . . .... 241
hot2 key . . . . . . . . . . . . . . . . . . . . . . . . . 189, 404 legend image post style key . . . . . . . . . 241, 343
hsv key . . . . . . . . . . . . . . . . . . . . . . . . . . .... 404 legend plot pos key . . . . . . . . . . . . . . . . .... 241
hsv2 key . . . . . . . . . . . . . . . . . . . . . . . . . .... 404 legend pos key . . . . . . . . . . . . . . . . . . . . .... 239
legend reversed key . . . . . . . . . . . . . . . . .... 244
—I— legend style key . . . . . . . . . . . . . . . . . . 239, 343
id key . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61f. legend to name key . . . . . . . . . . . . . . . . . .... 247
ifpgfplotsthreedim . . . . . . . . . . . . . . . . . . . . . . 526 legend transposed key . . . . . . . . . . . . . . .... 244
ignore chars key . . . . . . . . . . . . . . . . . . . . . . . 52 levels key . . . . . . . . . . . . . . . . . . . . . . . .... 152
ignore first key . . . . . . . . . . . . . . . . . . . . . . . 74 Libraries
imagesc key . . . . . . . . . . . . . . . . . . . . . . . . . . . 169 clickable . . . . . . . . . . . . . . . . . . . . .... 391
imagesc* key . . . . . . . . . . . . . . . . . . . . . . . . . . 169 colorbrewer . . . . . . . . . . . . . . . . . . . 397, 401
implicit . key . . . . . . . . . . . . . . . . . . . . . . . . . 52 colormaps . . . . . . . . . . . . . . . . . . . . .... 402
includegraphics key . . . . . . . . . . . . . . . . . . . . . 66 dateplot . . . . . . . . . . . . . . . . . . . . . 364, 407
includegraphics cmd key . . . . . . . . . . . . . . . . . 66 decorations.softclip . . . . . . . . . . . .... 407
index of colormap external . . . . . . . . . . . . . . . . . . . . . 407, 508
In colormap specification . . . . . . . . . . . . . . 192 fillbetween . . . . . . . . . . . . . . . . . . . .... 408
index of colormap key . . . . . . . . . . . . . . . . . . 194 groupplots . . . . . . . . . . . . . . . . . . . .... 427
indices of colormap patchplots . . . . . . . . . . . . . . . . . . . .... 433
In colormap specification . . . . . . . . . . .... 193 polar . . . . . . . . . . . . . . . . . . . . . . . . .... 449
inner axis line style key . . . . . . . . . . . 253, 343 smithchart . . . . . . . . . . . . . . . . . . . .... 455
inner moveto key . . . . . . . . . . . . . . . . . . .... 418 statistics . . . . . . . . . . . . . . . . . . . .... 468
interior colormap key . . . . . . . . . . . . . . .... 140 ternary . . . . . . . . . . . . . . . . . . . . . . .... 486
interior colormap name key . . . . . . . . . . .... 140 units . . . . . . . . . . . . . . . . . . . . . . . . .... 498
interior colormap thresh key . . . . . . . . .... 142 line cap key . . . . . . . . . . . . . . . . . . . . . . .... 182
Interrupted Plots . . . . . . . . . . . . . . . . . . . .... 118 line join key . . . . . . . . . . . . . . . . . . . . . .... 182
line legend key . . . . . . . . . . . . . . . . . . . . . . . 242 scanlinewise . . . . . . . . . . . . . . . . . . . . . . . 154

line width key . . . . . . . . . . . . . . . . . . . . . . . . 183 mesh/
linear regression key . . . . . . . . . . . . . . . . . . 374 check . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
\lineno . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59 color input . . . . . . . . . . . . . . . . . . . . . . . 142
log base 10 number format code key . . . . . . . 304 colorspace explicit color input . . . . . . 146
log basis x key . . . . . . . . . . . . . . . . . . . . . . . 328 colorspace explicit color output . . . . . 146
log basis y key . . . . . . . . . . . . . . . . . . . . . . . 328 cols . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
log basis z key . . . . . . . . . . . . . . . . . . . . . . . 328 interior colormap . . . . . . . . . . . . . . . . . 140
log identify minor tick positions key . . . . 302 interior colormap name . . . . . . . . . . . . . 140
log number format basis key . . . . . . . . . . . . . 304 interior colormap thresh . . . . . . . . . . . 142
log number format code key . . . . . . . . . . . . . . 303 ordering . . . . . . . . . . . . . . . . . . . . . . . . . 126
log origin key . . . . . . . . . . . . . . . . . . . . . . . . 310 rows . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
log origin x key . . . . . . . . . . . . . . . . . . . . . . 310 scanline verbose . . . . . . . . . . . . . . . . . . 126
log origin y key . . . . . . . . . . . . . . . . . . . . . . 310 mesh key . . . . . . . . . . . . . . . . . . . . . . . . . 117, 133
log origin z key . . . . . . . . . . . . . . . . . . . . . . 310 mesh input . . . . . . . . . . . . . . . . . . . . . . . . . . . 169
log plot exponent style key . . . . . . . . . . . . . 301 mesh input key . . . . . . . . . . . . . . . . . . . . . . . . 174
log ticks with fixed point key . . . . . . . . . . 301 mesh legend key . . . . . . . . . . . . . . . . . . . . . . . 243
loglogaxis environment . . . . . . . . . . . . . . . . . . 41 meta key . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
\logten . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 521 meta expr key . . . . . . . . . . . . . . . . . . . . . . . . . . 51
loosely dashdotdotted key . . . . . . . . . . . . . . . 182 meta index key . . . . . . . . . . . . . . . . . . . . . . . . . 51
loosely dashdotted key . . . . . . . . . . . . . . . . . 182 min key . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 304
loosely dashed key . . . . . . . . . . . . . . . . . . . . . 181 minor grid style key . . . . . . . . . . . . . . . . . . . 347
loosely dotted key . . . . . . . . . . . . . . . . . . . . . 181 minor tick key . . . . . . . . . . . . . . . . . . . . . . . . 314
lower quartile key . . . . . . . . . . . . . . . . . . . . . 472 minor tick length key . . . . . . . . . . . . . . . . . . 328
lower whisker key . . . . . . . . . . . . . . . . . 472, 479 minor tick num key . . . . . . . . . . . . . . . . . . . . . 313
lowlevel draw key . . . . . . . . . . . . . . . . . . . . . . 66 minor tick style key . . . . . . . . . . . . . . . . . . . 344
lua backend key . . . . . . . . . . . . . . . . . . . . . . . 505 minor x grid style key . . . . . . . . . . . . . . . . . 347
lua backend required . . . . . . . . . . . . . . . . . . . . 477 minor x tick num key . . . . . . . . . . . . . . . . . . . 313
lua debug key . . . . . . . . . . . . . . . . . . . . . . . . 506f. minor x tick style key . . . . . . . . . . . . . . . . . 345
minor xtick key . . . . . . . . . . . . . . . . . . . . . . . 314
—M— minor y grid style key . . . . . . . . . . . . . . . . . 347
major grid style key . . . . . . . . . . . . . . . . . . . 347 minor y tick num key . . . . . . . . . . . . . . . . . . . 313
major tick length key . . . . . . . . . . . . . . . . . . 328 minor y tick style key . . . . . . . . . . . . . . . . . 345
major tick style key . . . . . . . . . . . . . . . . . . . 344 minor ytick key . . . . . . . . . . . . . . . . . . . . . . . 314
major x grid style key . . . . . . . . . . . . . . . . . 347 minor z grid style key . . . . . . . . . . . . . . . . . 347
major x tick style key . . . . . . . . . . . . . . . . . 346 minor z tick num key . . . . . . . . . . . . . . . . . . . 313
major y grid style key . . . . . . . . . . . . . . . . . 347 minor z tick style key . . . . . . . . . . . . . . . . . 345
major y tick style key . . . . . . . . . . . . . . . . . 346 minor ztick key . . . . . . . . . . . . . . . . . . . . . . . 314
major z grid style key . . . . . . . . . . . . . . . . . 347 miter limit key . . . . . . . . . . . . . . . . . . . . . . . 182
major z tick style key . . . . . . . . . . . . . . . . . 346 mode key . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 512
many smithchart ticks key . . . . . . . . . . . . . . . 459
mark color key . . . . . . . . . . . . . . . . . . . . . . . . 180 —N—
mark indices key . . . . . . . . . . . . . . . . . . . . . . 180 name path key . . . . . . . . . . . . . . . . . . . . . . . . . 416
mark layer key . . . . . . . . . . . . . . . . . . . . . . . . 389 near ticklabel anchor . . . . . . . . . . . . . . . . . . 226
mark list fill key . . . . . . . . . . . . . . . . . . . . . 198 near ticklabel align key . . . . . . . . . . . . . . . 229
mark options key . . . . . . . . . . . . . . . . . . . . . . 180 near ticklabel at key . . . . . . . . . . . . . . . . . . 230
mark phase key . . . . . . . . . . . . . . . . . . . . . . . . 179 near ticklabel opposite anchor . . . . . . . . . . 228
mark repeat key . . . . . . . . . . . . . . . . . . . . . . . 179 near xticklabel anchor . . . . . . . . . . . . . . . . . 226
mark size key . . . . . . . . . . . . . . . . . . . . . 178, 282 near xticklabel opposite anchor . . . . . . . . . . 228
markers key . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96 near yticklabel anchor . . . . . . . . . . . . . . . . . 226
math parser key . . . . . . . . . . . . . . . . . . . . . . . . 47 near yticklabel opposite anchor . . . . . . . . . . 228
\matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 522 near zticklabel anchor . . . . . . . . . . . . . . . . . 226
matrix plot near zticklabel opposite anchor . . . . . . . . . . 228
Limits inaccurate warning . . . . . . . . . . ... 166 \nextgroupplot . . . . . . . . . . . . . . . . . . . . . . . . 428
matrix plot key . . . . . . . . . . . . . . . . . . . . ... 164 no markers key . . . . . . . . . . . . . . . . . . . . . . . . 179
matrix plot* key . . . . . . . . . . . . . . . . . . . ... 168 \node . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 522
max key . . . . . . . . . . . . . . . . . . . . . . . . . . . ... 304 node key . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
max space between ticks key . . . . . . . . . 282, 328 node near coord style key . . . . . . . . . . . . . . . 116
median key . . . . . . . . . . . . . . . . . . . . . . . 472, 480 node near coords style key . . . . . . . . . . . . . . 116
mesh nodes near coords key . . . . . . . . . . . . . . . . . . 114
closedcycle . . . . . . . . . . . . . . . . . . . . .... 437 nodes near coords align key . . . . . . . . . . . . . 116
Coordinate filtering . . . . . . . . . . . . . . 120, 369 nodes near coords style key . . . . . . . . . . . . . 116
patch type=line . . . . . . . . . . . . . . . . .... 154 nodes near coords* key . . . . . . . . . . . . . . . . . 114
point meta=none and global paths . . . .... 437 normalsize key . . . . . . . . . . . . . . . . . . . . . . . . 282
number key . . . . . . . . . . . . . . . . . . . . . . . . . . . 152 \pgfeov . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 521

\numplots . . . . . . . . . . . . . . . . . . . . . . . . . . . . 521 \pgfgetintersectionsegmentpath . . . . . . . . . . 425
\numplotsofactualtype . . . . . . . . . . . . . . . . . . 521 \pgfgetpath . . . . . . . . . . . . . . . . . . . . . . . . . . 426
pgfinterruptboundingbox environment . . . . . . 359
—O— \pgfintersectionofpaths . . . . . . . . . . . . . . . . 426
of key . . . . . . . . . . . . . . . . . . . . . . . . . . . 416, 420 \pgfkeys . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 521
on layer key . . . . . . . . . . . . . . . . . . . . . . 388, 418 \pgfkeysgetvalue . . . . . . . . . . . . . . . . . . . . . . 521
only marks key . . . . . . . . . . . . . . . . . . . . . . . . 107 \pgfkeysvalueof . . . . . . . . . . . . . . . . . . . . . . . 521
Options \pgfmathparse . . . . . . . . . . . . . . . . . . . . . . . . 519
Distinction Behavior, Style Options . . . . . . . 43 \pgfmathprintnumber . . . . . . . . . . . . . . . 300, 521
ordering key . . . . . . . . . . . . . . . . . . . . . . . . . . 126 \pgfonlayer . . . . . . . . . . . . . . . . . . . . . . . . . . 528
outer axis line style key . . . . . . . . . . . 253, 343 \pgfpathcomputesoftclippath . . . . . . . . . . . . . 426
output each nth key . . . . . . . . . . . . . . . . . . . . 192 \pgfpathreplacefirstmoveto . . . . . . . . . . . . . 426
output format key . . . . . . . . . . . . . . . . . . . . . 192 \pgfplotbarshift . . . . . . . . . . . . . . . . . . . . . . . 87
output point meta key . . . . . . . . . . . . . . . . . . 161 \pgfplotbarwidth . . . . . . . . . . . . . . . . . . . . . . . 86
overlay key . . . . . . . . . . . . . . . . . . . . . . . . . . . 357 \pgfplotsaxistransformcs . . . . . . . . . . . . . . . 373
\pgfplotsboxplotvalue . . . . . . . . . . . . . . . . . . 478
—P— \pgfplotsclickablecreate . . . . . . . . . . . . . . . 397
parametric/ \pgfplotscolorbarfromname . . . . . . . . . . . . . . 270
var 1d . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61 \pgfplotscolormaplastindexof . . . . . . . . . . . . 194
var 2d . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61 \pgfplotscolormapsizeof . . . . . . . . . . . . . . . . 194
parametric key . . . . . . . . . . . . . . . . . . . . . . . . . 60 \pgfplotscolormaptodatafile . . . . . . . . . . . . . 192
parent axis node . . . . . . . . . . . . . . . . . . . . . . 265 \pgfplotscolormaptoshadingspec . . . . . . . . . . 191
parent axis height key . . . . . . . . . . . . . . . . . 265 \pgfplotscreateplotcyclelist . . . . . . . . . . . . 199
parent axis width key . . . . . . . . . . . . . . . . . . 265 \pgfplotsdefinecstransform . . . . . . . . . . . . . 373
pastel key . . . . . . . . . . . . . . . . . . . . . . . . . . . 405 \pgfplotsextra . . . . . . . . . . . . . . . . . . . . . . . . 522
patch \pgfplotsforeachungrouped . . . . . . . . . . . . . . 518
closedcycle . . . . . . . . . . . . . . . . . . . . . . . . 437 \pgfplotsiffileexists . . . . . . . . . . . . . . . . . . 520
point meta=none and global paths . . . . . . . 437 pgfplotsinterruptdatabb environment . . . . . . 310
patch key . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169 \pgfplotsinvokeforeach . . . . . . . . . . . . . . . . . 519
Patch plots \pgfplotslegendfromname . . . . . . . . . . . . . . . . 248
Bounding Box . . . . . . . . . . . . . . . . . . . . . 441 \pgfplotsmarklistfill . . . . . . . . . . . . . . . . . . 198
patch refines key . . . . . . . . . . . . . . . . . . . . . 444 \pgfplotsmathfloatviewdepthxyz . . . . . . . . . . 525
patch table key . . . . . . . . . . . . . . . . . . . . . . . 171 \pgfplotsmathviewdepthxyz . . . . . . . . . . . . . . 526
patch table with individual point meta key 171 \pgfplotsonlayer . . . . . . . . . . . . . . . . . . . . . . 528
patch table with point meta key . . . . . . . . . . 171 \pgfplotspathaxisoutline . . . . . . . . . . . . . . . 522
patch to triangles key . . . . . . . . . . . . . . . . . 445 \pgfplotspointaxisdirectionxy . . . . . . . . . . . 523
patch type key . . . . . . . . . . . . . . . . . . . . 174, 434 \pgfplotspointaxisdirectionxyz . . . . . . . . . . 523
patch type sampling key . . . . . . . . . . . . . . . . 435 \pgfplotspointaxisorigin . . . . . . . . . . . . . . . 524
patchplots library . . . . . . . . . . . . . . . . . . . . . . 433 \pgfplotspointaxisxy . . . . . . . . . . . . . . . . . . 522
\path . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 521 \pgfplotspointaxisxyz . . . . . . . . . . . . . . . . . . 523
Path operations \pgfplotspointdescriptionxy . . . . . . . . . . . . . 523
-- . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 522 \pgfplotspointgetcoordinates . . . . . . . . . . . . 526
path replacement key . . . . . . . . . . . . . . . . . . . . 10 \pgfplotspointgetnormalizedcoordinates . . . 527
/pgf/ \pgfplotspointouternormalvectorofaxis . . . . 525
bar shift . . . . . . . . . . . . . . . . . . . . . . . . . 86 \pgfplotspointplotattime . . . . . . . . . . . . . . . 338
bar width . . . . . . . . . . . . . . . . . . . . . . . . . 86 \pgfplotspointrelaxisxy . . . . . . . . . . . . . . . . 523
declare function . . . . . . . . . . . . . . . . . . 520 \pgfplotspointrelaxisxyz . . . . . . . . . . . . . . . 523
decoration/ \pgfplotspointunitx . . . . . . . . . . . . . . . . . . . 524
every soft clipped path . . . . . . . . . . . 407 \pgfplotspointunity . . . . . . . . . . . . . . . . . . . 524
soft clip path . . . . . . . . . . . . . . . . . . 407 \pgfplotspointunitz . . . . . . . . . . . . . . . . . . . 524
fpu . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 382 \pgfplotsqpointdescriptionxy . . . . . . . . . . . . 523
images/ \pgfplotsqpointoutsideofaxis . . . . . . . . . . . . 525
aux in dpth . . . . . . . . . . . . . . . . . . . . . 515 \pgfplotsqpointoutsideofaxisrel . . . . . . . . . 525
mark color . . . . . . . . . . . . . . . . . . . . . . . 180 \pgfplotsset . . . . . . . . . . . . . . . . . . . . . . . . . . 76
text mark . . . . . . . . . . . . . . . . . . . . . . . . 180 \pgfplotssetlayers . . . . . . . . . . . . . . . . . . . . 388
text mark as node . . . . . . . . . . . . . . . . . 181 \pgfplotstableread . . . . . . . . . . . . . . . . . . . . 520
text mark style . . . . . . . . . . . . . . . . . . . 180 \pgfplotstabletypeset . . . . . . . . . . . . . . . . . . 520
trig format . . . . . . . . . . . . . . . . . . . . . . . . 57 \pgfplotsticklabelaxisspec . . . . . . . . . . . . . 525
\pgfaddpath . . . . . . . . . . . . . . . . . . . . . . . . . . 426 \pgfplotstransformcoordinatex . . . . . . . . . . . 524
\pgfaddpathandBB . . . . . . . . . . . . . . . . . . . . . . 426 \pgfplotstransformcoordinatey . . . . . . . . . . . 524
\pgfcomputeintersectionsegments . . . . . . . . . 423 \pgfplotstransformcoordinatez . . . . . . . . . . . 524
\pgfcomputereversepath . . . . . . . . . . . . . . . . . 426 \pgfplotstransformdirectionx . . . . . . . . . . . . 524
\pgfdeclareplotmark . . . . . . . . . . . . . . . . . . . 181
\pgfplotstransformdirectiony . . . . . . . . . . . . 524 point meta max key . . . . . . . . . . . . . . . . . 215, 265

\pgfplotstransformdirectionz . . . . . . . . . . . . 524 point meta min key . . . . . . . . . . . . . . . . . 215, 265
\pgfplotsunitxinvlength . . . . . . . . . . . . . . . . 525 point meta rel key . . . . . . . . . . . . . . . . . . . . . 215
\pgfplotsunitxlength . . . . . . . . . . . . . . . . . . 525 points key . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
\pgfplotsunityinvlength . . . . . . . . . . . . . . . . 525 polar library . . . . . . . . . . . . . . . . . . . . . . . . . . 449
\pgfplotsunitylength . . . . . . . . . . . . . . . . . . 525 polar comb key . . . . . . . . . . . . . . . . . . . . . . . . 453
\pgfplotsunitzinvlength . . . . . . . . . . . . . . . . 525 polaraxis environment . . . . . . . . . . . . . . . . . . 449
\pgfplotsunitzlength . . . . . . . . . . . . . . . . . . 525 popup size key . . . . . . . . . . . . . . . . . . . . . . . . 396
\pgfplotsutilifstringequal . . . . . . . . . . . . . 520 popup size generic key . . . . . . . . . . . . . . . . . 396
\pgfplotsvalueoflargesttickdimen . . . . . . . . 525 popup size snap key . . . . . . . . . . . . . . . . . . . . 396
pgfpoint substitution key . . . . . . . . . . . . . . . . 10 pos key . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 335
\pgfresetboundingbox . . . . . . . . . . . . . . . . . . 358 pos segment key . . . . . . . . . . . . . . . . . . . . . . . 337
\pgfsetlayers . . . . . . . . . . . . . . . . . . . . . . . . 528 position key . . . . . . . . . . . . . . . . . . . . . . . . . . 116
\pgfsetpath . . . . . . . . . . . . . . . . . . . . . . . . . . 426 pre filter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 374
\pgfsetpathandBB . . . . . . . . . . . . . . . . . . . . . . 426 pre filter key . . . . . . . . . . . . . . . . . . . . . . . . 369
pink key . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 405 Precision . . . . . . . . . . . . . . . . . . . . . . . . . . 54, 382
plot box ratio key . . . . . . . . . . . . . . . . . . . . . 290 precision key . . . . . . . . . . . . . . . . . . . . . . . . . 142
plot coordinates/ Predefined node
math parser . . . . . . . . . . . . . . . . . . . . . . . . 47 current axis . . . . . . . . . . . . . . . . . . . ... 357
plot file/ current colorbar axis . . . . . . . . . . . ... 267
ignore first . . . . . . . . . . . . . . . . . . . . . . . 74 current plot begin . . . . . . . . . . . . . . ... 335
skip first . . . . . . . . . . . . . . . . . . . . . . . . 74 current plot end . . . . . . . . . . . . . . . ... 335
plot graphics/ parent axis . . . . . . . . . . . . . . . . . . . . ... 265
debug . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73 prefix key . . . . . . . . . . . . . . . . . . . . . . . 61f., 509
includegraphics . . . . . . . . . . . . . . . . . . . . 66 prefixes
includegraphics cmd . . . . . . . . . . . . . . . . . 66 unit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 500
lowlevel draw . . . . . . . . . . . . . . . . . . . . . . 66 printable key . . . . . . . . . . . . . . . . . . . . . . . . . 396
node . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
points . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65 —Q—
xmax . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65 quiver/
xmin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65 after arrow . . . . . . . . . . . . . . . . . . . . . . . . 93
ymax . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65 before arrow . . . . . . . . . . . . . . . . . . . . . . . 93
ymin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65 colored . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
zmax . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65 every arrow . . . . . . . . . . . . . . . . . . . . . . . . 93
zmin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65 quiver legend . . . . . . . . . . . . . . . . . . . . . . 93
plot graphics key . . . . . . . . . . . . . . . . . . . . . . 66 scale arrows . . . . . . . . . . . . . . . . . . . . . . . 93
Plot operations u .......... . . . . . . . . . . . . . . . . . . . . . . 91
(hx expressioni,hy expressioni,hz expressioni) u value . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
127 update limits . . . . . . . . . . . . . . . . . . . . . . 93
coordinates . . . . . . . . . . . . . . . . . . . . . . . 124 v .......... . . . . . . . . . . . . . . . . . . . . . . 91
expression . . . . . . . . . . . . . . . . . . . . . . . 127 v value . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
file . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124 w .......... . . . . . . . . . . . . . . . . . . . . . . 91
table . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125 w value . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
{hmath expressioni} . . . . . . . . . . . . . . . . . 126 quiver key . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
plot (hx expressioni,hy expressioni) . . . . . 55 quiver legend key . . . . . . . . . . . . . . . . . . . . . . 93
plot coordinates . . . . . . . . . . . . . . . . . . . 46
plot expression . . . . . . . . . . . . . . . . . . . . 55 —R—
plot file . . . . . . . . . . . . . . . . . . . . . . . . . 73 raw gnuplot key . . . . . . . . . . . . . . . . . . . . . . . . 61
plot fill between . . . . . . . . . . . . . . . . . 408 rawx . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 374
plot function . . . . . . . . . . . . . . . . . . . . . . 60 rawy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 374
plot gnuplot . . . . . . . . . . . . . . . . . . . . . . . 59 rawz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 374
plot graphics . . . . . . . . . . . . . . . . . . . . . . 62 read completely key . . . . . . . . . . . . . . . . . . . . . 52
plot shell . . . . . . . . . . . . . . . . . . . . . . . . 61 redyellow key . . . . . . . . . . . . . . . . . . . . . . . . . 191
plot table . . . . . . . . . . . . . . . . . . . . . 47, 57 \ref . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 246
plot {hmath expressioni} . . . . . . . . . . . . . . 53 refstyle key . . . . . . . . . . . . . . . . . . . . . . . . . . 246
plot3graphics key . . . . . . . . . . . . . . . . . . . . . . 10 rel axis cs coordinate system . . . . . . . . . . . . 334
\plotnum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 521 reset nontranslations key . . . . . . . . . . . . . . . 231
\plotnumofactualtype . . . . . . . . . . . . . . . . . . 521 restrict expr to domain key . . . . . . . . . . . . . 371
point format key . . . . . . . . . . . . . . . . . . . . . . 395 restrict expr to domain* key . . . . . . . . . . . . 371
point format 3d key . . . . . . . . . . . . . . . . . . . . 395 restrict x to domain key . . . . . . . . . . . . . . . 370
point meta restrict x to domain* key . . . . . . . . . . . . . . . 370
point meta=none for smooth patch plots . . 437 restrict y to domain key . . . . . . . . . . . . . . . 370
point meta key . . . . . . . . . . . . . . . . . . . . . . . . 212 restrict y to domain* key . . . . . . . . . . . . . . . 370
restrict z to domain key . . . . . . . . . . . . . . . 370
restrict z to domain* key . . . . . . . . . . . . . . . 370 sloped/

reverse key . . . . . . . . . . . . . . . . . . . . . . . 416, 421 allow upside down . . . . . . . . . . . . . . . . . 231
reverse legend key . . . . . . . . . . . . . . . . . . . . . 244 at position . . . . . . . . . . . . . . . . . . . . . . . 231
reverse stacked plots key . . . . . . . . . . . . . . . . 96 execute for upside down . . . . . . . . . . . . 231
richtext key . . . . . . . . . . . . . . . . . . . . . . . . . . 396 reset nontranslations . . . . . . . . . . . . . . 231
row sep key . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51 sloped key . . . . . . . . . . . . . . . . . . . . . . . . . . . 336
rows key . . . . . . . . . . . . . . . . . . . . . . . . . 125, 430 sloped like x axis key . . . . . . . . . . . . . . . . . 231
sloped like y axis key . . . . . . . . . . . . . . . . . 231
—S— sloped like z axis key . . . . . . . . . . . . . . . . . 231
sample size key . . . . . . . . . . . . . . . . . . . . . . . 474 sloped Polar labels . . . . . . . . . . . . . . . . . . . . . . 453
sample size max key . . . . . . . . . . . . . . . . . . . . 475 small key . . . . . . . . . . . . . . . . . . . . . . . . . . . . 283
sample size min key . . . . . . . . . . . . . . . . . . . . 474 smithchart environment . . . . . . . . . . . . . . . . . 456
samples key . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55 smithchart library . . . . . . . . . . . . . . . . . . . . . . 455
samples at key . . . . . . . . . . . . . . . . . . . . . . . . . 56 smithchart mirrored key . . . . . . . . . . . . . . . . 456
samples of colormap smooth key . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
In colormap specification . . . . . . . . . . . . . . 193 snap dist key . . . . . . . . . . . . . . . . . . . . . . . . . 396
samples y key . . . . . . . . . . . . . . . . . . . . . . . . . . 56 soft clip key . . . . . . . . . . . . . . . . . . . . . . . . . 412
scale key . . . . . . . . . . . . . . . . . . . . . . . . . . . . 280 soft clip path key . . . . . . . . . . . . . . . . . . . . . 407
scale arrows key . . . . . . . . . . . . . . . . . . . . . . . 93 soft clip second key . . . . . . . . . . . . . . . . . . . 412
scale mode key . . . . . . . . . . . . . . . . . . . . . 10, 285 solid key . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
scale only axis key . . . . . . . . . . . . . . . . . . . . 272 split key . . . . . . . . . . . . . . . . . . . . . . . . . . . . 410
scale ticks above exponent key . . . . . . . . . . 327 spring key . . . . . . . . . . . . . . . . . . . . . . . . . . . 405
scale ticks below exponent key . . . . . . . . . . 327 stack dir key . . . . . . . . . . . . . . . . . . . . . . . . . . 96
scale uniformly strategy key . . . . . . . . . . . . 285 stack negative key . . . . . . . . . . . . . . . . . . . . . . 98
scaled ticks key . . . . . . . . . . . . . . . . . . . . . . 324 stack plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . 374
scaled x ticks key . . . . . . . . . . . . . . . . . . . . . 324 stack plots key . . . . . . . . . . . . . . . . . . . . . . . . 94
scaled y ticks key . . . . . . . . . . . . . . . . . . . . . 324 stacked ignores zero/
scaled z ticks key . . . . . . . . . . . . . . . . . . . . . 324 default . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
scaling key . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 errorbars . . . . . . . . . . . . . . . . . . . . . . . . . 96
scanline marks key . . . . . . . . . . . . . . . . . . . . . 160 markers . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
scanline verbose key . . . . . . . . . . . . . . . . . . . 126 stacked ignores zero key . . . . . . . . . . . . . . . . 96
scatter/ standard key . . . . . . . . . . . . . . . . . . . . . . . . . . 386
classes . . . . . . . . . . . . . . . . . . . . . . . . . . 112 statistics library . . . . . . . . . . . . . . . . . . . . . . 468
position . . . . . . . . . . . . . . . . . . . . . . . . . 116 .style handler . . . . . . . . . . . . . . . . . . . . . . . . 340
use mapped color . . . . . . . . . . . . . . . . . . 110 Subplots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 353
scatter key . . . . . . . . . . . . . . . . . . . . . . . . . . . 108 subtickwidth key . . . . . . . . . . . . . . . . . . . . . . 328
scatter src key . . . . . . . . . . . . . . . . . . . . . . . 109 summer key . . . . . . . . . . . . . . . . . . . . . . . . . . . 405
script key . . . . . . . . . . . . . . . . . . . . . . . . . . . 160 surf
script extension key . . . . . . . . . . . . . . . . . . . 161 Coordinate filtering . . . . . . . . . . . . . . 120, 369
search path key . . . . . . . . . . . . . . . . . . . . . . . . 52 point meta=none and global paths . . . . . . . 437
semilogxaxis environment . . . . . . . . . . . . . . . . . 41 surf key . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
semilogyaxis environment . . . . . . . . . . . . . . . . . 41 surf shading/
semithick key . . . . . . . . . . . . . . . . . . . . . . . . . 183 precision . . . . . . . . . . . . . . . . . . . . . . . . 142
separate axis lines Symbolic coords
Combined with axis line shift . . . . . . . . . . 256 [normalized] . . . . . . . . . . . . . . . . . . . . . . . 364
separate axis lines key . . . . . . . . . . . . . . . . 254 symbolic coords key . . . . . . . . . . . . . . . . . . . . 486
sepia key . . . . . . . . . . . . . . . . . . . . . . . . . . . . 405 symbolic x coords key . . . . . . . . . . . . . . . . . . 363
sequence key . . . . . . . . . . . . . . . . . . . . . . . . . . 420 symbolic y coords key . . . . . . . . . . . . . . . . . . 363
set layers key . . . . . . . . . . . . . . . . . . . . . . . . 386 symbolic z coords key . . . . . . . . . . . . . . . . . . 363
set point meta if empty key . . . . . . . . . . . . . 214 system call key . . . . . . . . . . . . . . . . . . . . . . . 511
shader key . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
sharp plot key . . . . . . . . . . . . . . . . . . . . . . . . . 77 —T—
plot shell . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61 plot table . . . . . . . . . . . . . . . . . . . . . . . . . 47, 57
shell escape key . . . . . . . . . . . . . . . . . . . . . . 512 Unbalanced Columns . . . . . . . . . . . . . . . . . 49
show origin key . . . . . . . . . . . . . . . . . . . . . . . 464 table/
show origin code key . . . . . . . . . . . . . . . . . . . 464 col sep . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
show suggested version key . . . . . . . . . . . . . . 390 comment chars . . . . . . . . . . . . . . . . . . . . . . 52
size x key . . . . . . . . . . . . . . . . . . . . . . . . . . . 178 create col/
size y key . . . . . . . . . . . . . . . . . . . . . . . . . . . 178 linear regression . . . . . . . . . . . . . . . . 374
size z key . . . . . . . . . . . . . . . . . . . . . . . . . . . 178 header . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
skip coords between index key . . . . . . . . . . . 369 ignore chars . . . . . . . . . . . . . . . . . . . . . . . 52
skip first key . . . . . . . . . . . . . . . . . . . . . . . . . 74 meta . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
skip first n key . . . . . . . . . . . . . . . . . . . . . . . 52 meta expr . . . . . . . . . . . . . . . . . . . . . . . . . 51
slope format key . . . . . . . . . . . . . . . . . . . . . . 396 meta index . . . . . . . . . . . . . . . . . . . . . . . . 51
read completely . . . . . . . . . . . . . . . . . . . . 52 thin key . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183

row sep . . . . . . . . . . . . . . . . . . . . . . . . . . . 51 \thisrow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
search path/ \thisrowno . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
implicit . . . . . . . . . . . . . . . . . . . . . . . . 52 tick align key . . . . . . . . . . . . . . . . . . . . . . . . 322
search path . . . . . . . . . . . . . . . . . . . . . . . . 52 tick label style key . . . . . . . . . . . . . . . . . . . 344
skip first n . . . . . . . . . . . . . . . . . . . . . . . 52 Tick Labels
tie end x . . . . . . . . . . . . . . . . . . . . . . . . 497 Configure texts by command: xticklabel . . . 318
tie end x index . . . . . . . . . . . . . . . . . . . 497 Di↵erent Depths . . . . . . . . . . . . . . . . . . . . 323
tie end y . . . . . . . . . . . . . . . . . . . . . . . . 497 Distances: xtick with foreach . . . . . . . . . . . 311
tie end y index . . . . . . . . . . . . . . . . . . . 497 Placement of log tick labels: xtickten . . . . . 316
tie end z . . . . . . . . . . . . . . . . . . . . . . . . 497 Positions: xtick . . . . . . . . . . . . . . . . . . . . . 311
tie end z index . . . . . . . . . . . . . . . . . . . 497 Read from table . . . . . . . . . . . . . . . . . . . . 320
white space chars . . . . . . . . . . . . . . . . . . 52 Shift along normal vector: ticklabel shift . . 323
x .................. . . . . . . . . . . . . . . 50 strut . . . . . . . . . . . . . . . . . . . . . . . . . . . . 323
x error . . . . . . . . . . . . . . . . . . . . . . . 50, 298 Texts as list: xticklabels . . . . . . . . . . . . . . 317
x error expr . . . . . . . . . . . . . . . . . . . 50, 298 Units (km, m, etc) . . . . . . . . . . . . . . . . . . 498
x error index . . . . . . . . . . . . . . . . . . 50, 298 tick pos key . . . . . . . . . . . . . . . . . . . . . . . . . . 322
x error minus . . . . . . . . . . . . . . . . . . 51, 298 tick scale binop key . . . . . . . . . . . . . . . . . . . 327
x error minus expr . . . . . . . . . . . . . . 51, 298 tick scale label code key . . . . . . . . . . . . . . . 327
x error minus index . . . . . . . . . . . . . 51, 298 tick style key . . . . . . . . . . . . . . . . . . . . . . . . 344
x error plus . . . . . . . . . . . . . . . . . . . 50, 298 ticklabel cs coordinate system . . . . . . . . . . . . 222
x error plus expr . . . . . . . . . . . . . . 51, 298 ticklabel pos key . . . . . . . . . . . . . . . . . . . . . 322
x error plus index . . . . . . . . . . . . . . 50, 298 ticklabel shift key . . . . . . . . . . . . . . . . . . . . 323
x expr . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50 ticklabel style key . . . . . . . . . . . . . . . . . . . . 344
x index . . . . . . . . . . . . . . . . . . . . . . . . . . . 50 ticklabel* cs coordinate system . . . . . . . . . . . 222
y .................. . . . . . . . . . . . . . . 50 ticks key . . . . . . . . . . . . . . . . . . . . . . . . . . . . 321
y error . . . . . . . . . . . . . . . . . . . . . . . 50, 298 tickwidth key . . . . . . . . . . . . . . . . . . . . . . . . . 328
y error expr . . . . . . . . . . . . . . . . . . . 50, 298 tie end x key . . . . . . . . . . . . . . . . . . . . . . . . . 497
y error index . . . . . . . . . . . . . . . . . . 50, 298 tie end x index key . . . . . . . . . . . . . . . . . . . . 497
y error minus . . . . . . . . . . . . . . . . . . 51, 298 tie end y key . . . . . . . . . . . . . . . . . . . . . . . . . 497
y error minus expr . . . . . . . . . . . . . . 51, 299 tie end y index key . . . . . . . . . . . . . . . . . . . . 497
y error minus index . . . . . . . . . . . . . 51, 298 tie end z key . . . . . . . . . . . . . . . . . . . . . . . . . 497
y error plus . . . . . . . . . . . . . . . . . . . 50, 298 tie end z index key . . . . . . . . . . . . . . . . . . . . 497
y error plus expr . . . . . . . . . . . . . . 51, 298 tieline/
y error plus index . . . . . . . . . . . . . . 50, 298 curve style . . . . . . . . . . . . . . . . . . . . . . . 498
y expr . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50 each nth tie . . . . . . . . . . . . . . . . . . . . . . 497
y index . . . . . . . . . . . . . . . . . . . . . . . . . . . 50 tieline style . . . . . . . . . . . . . . . . . . . . . 497
z .................. . . . . . . . . . . . . . . 50 tieline key . . . . . . . . . . . . . . . . . . . . . . . . . . . 496
z error . . . . . . . . . . . . . . . . . . . . . . . 50, 298 tieline style key . . . . . . . . . . . . . . . . . . . . . 497
z error expr . . . . . . . . . . . . . . . . . . . 50, 298 /tikz/
z error index . . . . . . . . . . . . . . . . . . 50, 298 allow upside down . . . . . . . . . . . . . . . . . 337
z error minus . . . . . . . . . . . . . . . . . . 51, 298 baseline . . . . . . . . . . . . . . . . . . . . . . . . . 351
z error minus expr . . . . . . . . . . . . . . 51, 299 color . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
z error minus index . . . . . . . . . . . . . 51, 298 const plot . . . . . . . . . . . . . . . . . . . . . . . . 78
z error plus . . . . . . . . . . . . . . . . . . . 50, 298 const plot mark left . . . . . . . . . . . . . . . . 79
z error plus expr . . . . . . . . . . . . . . 51, 298 const plot mark mid . . . . . . . . . . . . . . . . . 79
z error plus index . . . . . . . . . . . . . . 50, 298 const plot mark right . . . . . . . . . . . . . . . 79
z expr . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50 dashdotdotted . . . . . . . . . . . . . . . . . . . . . 182
z index . . . . . . . . . . . . . . . . . . . . . . . . . . . 50 dashdotted . . . . . . . . . . . . . . . . . . . . . . . 181
table . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125 dashed . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
table key . . . . . . . . . . . . . . . . . . . . . . . . . . . . 375 densely dashdotdotted . . . . . . . . . . . . . . 182
temp key . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 406 densely dashdotted . . . . . . . . . . . . . . . . . 182
tension key . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78 densely dashed . . . . . . . . . . . . . . . . . . . . 181
ternary library . . . . . . . . . . . . . . . . . . . . . . . . 486 densely dotted . . . . . . . . . . . . . . . . . . . . 181
ternary limits relative key . . . . . . . . . . . . . 492 dotted . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
ternary relative limits key . . . . . . . . . . . . . 492 draw . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
ternaryaxis environment . . . . . . . . . . . . . . . . . 487 every mark . . . . . . . . . . . . . . . . . . . . . . . 178
text mark key . . . . . . . . . . . . . . . . . . . . . . . . . 180 external/
text mark as node key . . . . . . . . . . . . . . . . . . 181 figure name . . . . . . . . . . . . . . . . . . . . . 510
text mark style key . . . . . . . . . . . . . . . . . . . . 180 mode . . . . . . . . . . . . . . . . . . . . . . . . . . . 512
textSize key . . . . . . . . . . . . . . . . . . . . . . . . . . 396 prefix . . . . . . . . . . . . . . . . . . . . . . . . . 509
thermal key . . . . . . . . . . . . . . . . . . . . . . . . . . . 406 shell escape . . . . . . . . . . . . . . . . . . . . 512
thick key . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183 system call . . . . . . . . . . . . . . . . . . . . . 511
fill . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185 trim axis right . . . . . . . . ... . . . . . . . . 359

fill between/ trim left . . . . . . . . . . . . . ... . . . . . . . . 359
every even segment . . . . . . . . . . . . . . . 415 trim right . . . . . . . . . . . . ... . . . . . . . . 359
every last segment . . . . . . . . . . . . . . . 415 ultra thick . . . . . . . . . . . . ... . . . . . . . . 183
every odd segment . . . . . . . . . . . . . . . . 415 ultra thin . . . . . . . . . . . . ... . . . . . . . . 183
every segment . . . . . . . . . . . . . . . . . . . 415 very thick . . . . . . . . . . . . ... . . . . . . . . 183
every segment no hindex i . . . . . . . . . . 415 very thin . . . . . . . . . . . . . ... . . . . . . . . 183
inner moveto . . . . . . . . . . . . . . . . . . . . 418 xbar . . . . . . . . . . . . . . . . . ... . . . . . . . . . 81
of . . . . . . . . . . . . . . . . . . . . . . . . . . . . 416 xbar interval . . . . . . . . . . ... . . . . . . . . . 88
on layer . . . . . . . . . . . . . . . . . . . . . . . 418 xcomb . . . . . . . . . . . . . . . . . ... . . . . . . . . . 89
reverse . . . . . . . . . . . . . . . . . . . . . . . . 416 xshift . . . . . . . . . . . . . . . . ... . . . . . . . . 522
soft clip . . . . . . . . . . . . . . . . . . . . . . 412 ybar . . . . . . . . . . . . . . . . . ... . . . . . . . . . 83
soft clip second . . . . . . . . . . . . . . . . 412 ybar interval . . . . . . . . . . ... . . . . . . . . . 87
split . . . . . . . . . . . . . . . . . . . . . . . . . . 410 ycomb . . . . . . . . . . . . . . . . . ... . . . . . . . . . 89
font . . . . . . . . . . . . . . . . . . . . . . . . 183, 282 yshift . . . . . . . . . . . . . . . . ... . . . . . . . . 522
id . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61f. \tikzappendtofigurename . . . . . ... . . . . . . . . 511
imagesc . . . . . . . . . . . . . . . . . . . . . . . . . . 169 \tikzfillbetween . . . . . . . . . . . ... . . . . . . . . 422
imagesc* . . . . . . . . . . . . . . . . . . . . . . . . . 169 \tikzgetnamedpath . . . . . . . . . . ... . . . . . . . . 425
intersection segments . . . . . . . . . . . . . . 420 \tikznamecurrentpath . . . . . . . ... . . . . . . . . 425
jump mark left . . . . . . . . . . . . . . . . . . . . . 80 \tikzpathintersectionsegments ... . . . . . . . . 426
jump mark mid . . . . . . . . . . . . . . . . . . . . . . 80 tikzpicture environment . . . . . . ... . . . . . . . . . 40
jump mark right . . . . . . . . . . . . . . . . . . . . 80 \tikzsegmentindex . . . . . . . . . . ... . . . . . . . . 415
line cap . . . . . . . . . . . . . . . . . . . . . . . . . 182 \tikzsetexternalprefix . . . . . . ... . . . . . . . . 509
line join . . . . . . . . . . . . . . . . . . . . . . . . 182 \tikzsetfigurename . . . . . . . . . ... . . . . . . . . 510
line width . . . . . . . . . . . . . . . . . . . . . . . 183 \tikzsetnextfilename . . . . . . . ... . . . . . . . . 509
loosely dashdotdotted . . . . . . . . . . . . . . 182 tiny key . . . . . . . . . . . . . . . . . . ... . . . . . . . . 284
loosely dashdotted . . . . . . . . . . . . . . . . . 182 title key . . . . . . . . . . . . . . . . . ... . . . . . . . . 233
loosely dashed . . . . . . . . . . . . . . . . . . . . 181 title style key . . . . . . . . . . . . ... . . . . . . . . 342
loosely dotted . . . . . . . . . . . . . . . . . . . . 181 translate gnuplot key . . . . . . . ... . . . . . . . . . 60
mark indices . . . . . . . . . . . . . . . . . . . . . . 180 transpose legend key . . . . . . . . ... . . . . . . . . 244
mark options . . . . . . . . . . . . . . . . . . . . . . 180 trig format key . . . . . . . . . . . . ... . . . . . . . . . 57
mark phase . . . . . . . . . . . . . . . . . . . . . . . 179 trig format plots key . . . . . . . ... . . . . . . . . . 56
mark repeat . . . . . . . . . . . . . . . . . . . . . . . 179 trim axis group left key . . . . ... . . . . . . . . 359
mark size . . . . . . . . . . . . . . . . . . . . 178, 282 trim axis group right key . . . . ... . . . . . . . . 360
matrix plot* . . . . . . . . . . . . . . . . . . . . . . 168 trim axis left key . . . . . . . . . . ... . . . . . . . . 359
miter limit . . . . . . . . . . . . . . . . . . . . . . . 182 trim axis right key . . . . . . . . . ... . . . . . . . . 359
name path . . . . . . . . . . . . . . . . . . . . . . . . 416 trim left key . . . . . . . . . . . . . . ... . . . . . . . . 359
only marks . . . . . . . . . . . . . . . . . . . . . . . 107 trim right key . . . . . . . . . . . . . ... . . . . . . . . 359
overlay . . . . . . . . . . . . . . . . . . . . . . . . . . 357 try min ticks key . . . . . . . . . . ... . . . . 282, 328
polar comb . . . . . . . . . . . . . . . . . . . . . . . 453 try min ticks log key . . . . . . . ... . . . . . . . . 328
pos . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 335 typeset ticklabels with strut key . . . . . . . . 323
pos segment . . . . . . . . . . . . . . . . . . . . . . . 337
prefix . . . . . . . . . . . . . . . . . . . . . . . . . . . 61f. —U—
raw gnuplot . . . . . . . . . . . . . . . . . . . . . . . . 61 u key . . . . . . . . . . . . . . . ... . . . . . . . . . . . . . . . 91
segments/ u value key . . . . . . . . . . ... . . . . . . . . . . . . . . . 92
of . . . . . . . . . . . . . . . . . . . . . . . . . . . . 420 ultra thick key . . . . . . ... . . . . . . . . . . . . . . 183
reverse . . . . . . . . . . . . . . . . . . . . . . . . 421 ultra thin key . . . . . . . ... . . . . . . . . . . . . . . 183
sequence . . . . . . . . . . . . . . . . . . . . . . . 420 Unbalanced Columns . . . ... . . . . . . . . . . . . . . . 49
semithick . . . . . . . . . . . . . . . . . . . . . . . . 183 unbounded coords key . . ... . . . . . . . . . . . . . . 119
sharp plot . . . . . . . . . . . . . . . . . . . . . . . . 77 unit code key . . . . . . . . ... . . . . . . . . . . . . . . 500
sloped . . . . . . . . . . . . . . . . . . . . . . . . . . . 336 unit marking post key . ... . . . . . . . . . . . . . . 499
sloped like x axis . . . . . . . . . . . . . . . . . 231 unit marking pre key . . ... . . . . . . . . . . . . . . 499
sloped like y axis . . . . . . . . . . . . . . . . . 231 unit markings key . . . . ... . . . . . . . . . . . . . . 499
sloped like z axis . . . . . . . . . . . . . . . . . 231 unit rescale keep size key . . . . . . . . . . 279, 286
smooth . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78 Unit vector
solid . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181 Assign . . . . . . . . . . . . . . . . . . . . . . . . ... 272
tension . . . . . . . . . . . . . . . . . . . . . . . . . . . 78 unit vector ratio key . . . . . . . . . . . . . . . ... 278
thick . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183 unit vector ratio* key . . . . . . . . . . . . . . ... 279
thin . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183 units library . . . . . . . . . . . . . . . . . . . . . . . ... 498
tieline . . . . . . . . . . . . . . . . . . . . . . . . . . 496 update limits key . . . . . . . . . . . . . . . . . . 93, 310
trim axis group left . . . . . . . . . . . . . . . 359 upper quartile key . . . . . . . . . . . . . . . . . .... 472
trim axis group right . . . . . . . . . . . . . . 360 upper whisker key . . . . . . . . . . . . . . . . . 472, 479
trim axis left . . . . . . . . . . . . . . . . . . . . 359 use mapped color key . . . . . . . . . . . . . . . .... 110
use units key . . . . . . . . . . . . . . . . . . . . . .... 498
—V— x expr key . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50

v key . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91, 289 x filter key . . . . . . . . . . . . . . . . . . . . . . . . . 366f.
v value key . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92 x fixed key . . . . . . . . . . . . . . . . . . . . . . . . . . . 295
var 1d key . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61 x fixed relative key . . . . . . . . . . . . . . . . . . . 295
var 2d key . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61 x grid style key . . . . . . . . . . . . . . . . . . . . . . 347
variable key . . . . . . . . . . . . . . . . . . . . . . . . . . . 56 x index key . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
variable width key . . . . . . . . . . . . . . . . . . . . . 473 x label style key . . . . . . . . . . . . . . . . . . . . . 342
variable width expr key . . . . . . . . . . . . . . . . 474 x post scale key . . . . . . . . . . . . . . . . . . . . . . 280
variable width min target key . . . . . . . . . . . 475 x SI prefix key . . . . . . . . . . . . . . . . . . . . . . . 500
variable y key . . . . . . . . . . . . . . . . . . . . . . . . . 56 x tick label as interval key . . . . . . . . . . . . 320
variance key . . . . . . . . . . . . . . . . . . . . . . . . . . 376 x tick label style key . . . . . . . . . . . . . . . . . 344
variance list key . . . . . . . . . . . . . . . . . . . . . 376 x tick scale label style key . . . . . . . . . . . . 345
variance src key . . . . . . . . . . . . . . . . . . . . . . 377 x unit key . . . . . . . . . . . . . . . . . . . . . . . . . . . 499
vertex count key . . . . . . . . . . . . . . . . . . . . . . 443 x unit prefix key . . . . . . . . . . . . . . . . . . . . . 499
vertical sep key . . . . . . . . . . . . . . . . . . . . . . 430 xbar key . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81f.
very thick key . . . . . . . . . . . . . . . . . . . . . . . . 183 xbar interval key . . . . . . . . . . . . . . . . . . . . . 88f.
very thin key . . . . . . . . . . . . . . . . . . . . . . . . . 183 xbar interval legend key . . . . . . . . . . . . . . . 242
View xbar interval stacked key . . . . . . . . . . . . . . . . 96
Assign rotation angles . . . . . . . . . . . . . . . . 287 xbar legend key . . . . . . . . . . . . . . . . . . . . . . . 242
Assign unit vectors . . . . . . . . . . . . . . . . . . 272 xbar stacked key . . . . . . . . . . . . . . . . . . . . . . . 95
view/ xcomb key . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
az . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 288 xgrid each nth passes y key . . . . . . . . . . . . . 466
el . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 289 xgrid each nth passes y start key . . . . . . . . 468
h .................... . . . . . . . . . . . 288 xgrid stop at y key . . . . . . . . . . . . . . . . . . . . 468
v .................... . . . . . . . . . . . 289 xlabel
view key . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 287 Line break . . . . . . . . . . . . . . . . . . . . . . . . 232
violet key . . . . . . . . . . . . . . . . . . . . . . . . . . . 191 Multiline . . . . . . . . . . . . . . . . . . . . . . . . . 232
viridis key . . . . . . . . . . . . . . . . . . . . . . . . . . . 189 xlabel key . . . . . . . . . . . . . . . . . . . . . . . . . . . 231
viridis high res key . . . . . . . . . . . . . . . . . . . 406 xlabel absolute key . . . . . . . . . . . . . . . . . . . . 233
visualization depends on key . . . . . . . . . . . . 381 xlabel near ticks key . . . . . . . . . . . . . . . . . . 232
xlabel shift key . . . . . . . . . . . . . . . . . . . . . . 232
—W— xlabel style key . . . . . . . . . . . . . . . . . . . . . . 342
w key . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91 xlabels at key . . . . . . . . . . . . . . . . . . . . . . . . 430
w value key . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92 xmajorgrids key . . . . . . . . . . . . . . . . . . . . . . . 330
waterfall chart . . . . . . . . . . . . . . . . . . . . . . . . . . 98 xmajorticks key . . . . . . . . . . . . . . . . . . . . . . . 321
whisker key . . . . . . . . . . . . . . . . . . . . . . . . . . . 479 xmax key . . . . . . . . . . . . . . . . . . . 65, 304, 311, 397
whisker extend key . . . . . . . . . . . . . . . . . . . . . 475 xmin key . . . . . . . . . . . . . . . . . . . 65, 304, 310, 397
whisker range key . . . . . . . . . . . . . . . . . . . . . 477 xminorgrids key . . . . . . . . . . . . . . . . . . . . . . . 330
white space chars key . . . . . . . . . . . . . . . . . . . 52 xminorticks key . . . . . . . . . . . . . . . . . . . . . . . 321
width key . . . . . . . . . . . . . . . . . . . . . 266, 271, 397 xmode key . . . . . . . . . . . . . . . . . . . . . 275, 306, 376
winter key . . . . . . . . . . . . . . . . . . . . . . . . . . . 406 xshift key . . . . . . . . . . . . . . . . . . . . . . . . . . . 522
xtick key . . . . . . . . . . . . . . . . . . . . . . . . . . . . 311
—X— xtick align key . . . . . . . . . . . . . . . . . . . . . . . 322
x xtick distance key . . . . . . . . . . . . . . . . . . . . . 311
Assign unit vector . . . . . . . . . . . . . . . . . . . 272 xtick placement tolerance key . . . . . . . . . . . 328
x key . . . . . . . . . . . . . . . . . . . . . . 50, 272, 338, 376 xtick pos key . . . . . . . . . . . . . . . . . . . . . . . . . 322
\x In Coordinate Expressions . . . . . . . . . . . . . . . 54 xtick scale label code key . . . . . . . . . . . . . . 326
x axis line style key . . . . . . . . . . . . . . 253, 343 xtick style key . . . . . . . . . . . . . . . . . . . . . . . 345
x coord inv trafo key . . . . . . . . . . . . . . . . . . 362 xticklabel key . . . . . . . . . . . . . . . . . . . . . . . . 318
x coord trafo . . . . . . . . . . . . . . . . . . . . . . . . . . 374 xticklabel cs coordinate system . . . . . . . . . . . 222
x coord trafo key . . . . . . . . . . . . . . . . . . . . . 362 xticklabel interval boundaries key . . . . . . . . 89
x descriptions at key . . . . . . . . . . . . . . . . . . 431 xticklabel pos key . . . . . . . . . . . . . . . . . . . . . 322
x dir key . . . . . . . . . . . . . . . . . . . . . 276, 294, 306 xticklabel shift key . . . . . . . . . . . . . . . . . . . 323
x error key . . . . . . . . . . . . . . . . . . . . . . . . 50, 298 xticklabel style key . . . . . . . . . . . . . . . . . . . 344
x error expr key . . . . . . . . . . . . . . . . . . . 50, 298 xticklabel* cs coordinate system . . . . . . . . . . 222
x error index key . . . . . . . . . . . . . . . . . . 50, 298 xticklabels key . . . . . . . . . . . . . . . . . . . . . . . 317
x error minus key . . . . . . . . . . . . . . . . . . 51, 298 xticklabels at key . . . . . . . . . . . . . . . . . . . . . 431
x error minus expr key . . . . . . . . . . . . . . 51, 298 xticklabels from table key . . . . . . . . . . . . . . 320
x error minus index key . . . . . . . . . . . . . 51, 298 xtickmax key . . . . . . . . . . . . . . . . . . . . . . 258, 322
x error plus key . . . . . . . . . . . . . . . . . . . 50, 298 xtickmin key . . . . . . . . . . . . . . . . . . . . . . 258, 321
x error plus expr key . . . . . . . . . . . . . . . 51, 298 xtickten key . . . . . . . . . . . . . . . . . . . . . . . . . . 316
x error plus index key . . . . . . . . . . . . . . 50, 298
x explicit key . . . . . . . . . . . . . . . . . . . . . . . . 295 —Y—
x explicit relative key . . . . . . . . . . . . . . . . 296 y
Assign unit vector . . . . . . . . . . . . . . . . . . . 272 ytick scale label code key . . . . . . . . . . . . . . 326
y key . . . . . . . . . . . . . . . . . . . . . . 50, 272, 338, 376 ytick style key . . . . . . . . . . . . . . . . . . . . . . . 345
y axis line style key . . . . . . . . . . . . . . 253, 343 yticklabel key . . . . . . . . . . . . . . . . . . . . . . . . 318
y coord inv trafo key . . . . . . . . . . . . . . . . . . 362 yticklabel around circle key . . . . . . . . . . . . 465
y coord trafo . . . . . . . . . . . . . . . . . . . . . . . . . . 374 yticklabel around circle* key . . . . . . . . . . . 465
y coord trafo key . . . . . . . . . . . . . . . . . . . . . 362 yticklabel cs coordinate system . . . . . . . . . . . 222
y descriptions at key . . . . . . . . . . . . . . . . . . 431 yticklabel in circle key . . . . . . . . . . . . . . . 464
y dir key . . . . . . . . . . . . . . . . . . . . . 276, 294, 306 yticklabel interval boundaries key . . . . . . . . 89
y domain key . . . . . . . . . . . . . . . . . . . . . . . . . . . 55 yticklabel pos key . . . . . . . . . . . . . . . . . . . . . 322
y error key . . . . . . . . . . . . . . . . . . . . . . . . 50, 298 yticklabel shift key . . . . . . . . . . . . . . . . . . . 323
y error expr key . . . . . . . . . . . . . . . . . . . 50, 298 yticklabel style key . . . . . . . . . . . . . . . . . . . 344
y error index key . . . . . . . . . . . . . . . . . . 50, 298 yticklabel* cs coordinate system . . . . . . . . . . 222
y error minus key . . . . . . . . . . . . . . . . . . 51, 298 yticklabels key . . . . . . . . . . . . . . . . . . . . . . . 317
y error minus expr key . . . . . . . . . . . . . . 51, 299 yticklabels at key . . . . . . . . . . . . . . . . . . . . . 431
y error minus index key . . . . . . . . . . . . . 51, 298 yticklabels from table key . . . . . . . . . . . . . . 320
y error plus key . . . . . . . . . . . . . . . . . . . 50, 298 ytickmax key . . . . . . . . . . . . . . . . . . . . . . 258, 322
y error plus expr key . . . . . . . . . . . . . . . 51, 298 ytickmin key . . . . . . . . . . . . . . . . . . . . . . 258, 321
y error plus index key . . . . . . . . . . . . . . 50, 298 ytickten key . . . . . . . . . . . . . . . . . . . . . . . . . . 316
y explicit key . . . . . . . . . . . . . . . . . . . . . . . . 295
y explicit relative key . . . . . . . . . . . . . . . . 296 —Z—
y expr key . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50 z
y filter key . . . . . . . . . . . . . . . . . . . . . . . . . 366f. Assign unit vector . . . . . . . . . . . . . . . . . . . 272
y fixed key . . . . . . . . . . . . . . . . . . . . . . . . . . . 295 z key . . . . . . . . . . . . . . . . . . . . . . . . . 50, 272, 338
y fixed relative key . . . . . . . . . . . . . . . . . . . 295 z axis line style key . . . . . . . . . . . . . . 253, 343
y grid style key . . . . . . . . . . . . . . . . . . . . . . 347 z buffer key . . . . . . . . . . . . . . . . . . . . . . . . . . 135
y index key . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50 z coord inv trafo key . . . . . . . . . . . . . . . . . . 362
y label style key . . . . . . . . . . . . . . . . . . . . . 342 z coord trafo . . . . . . . . . . . . . . . . . . . . . . . . . . 374
y post scale key . . . . . . . . . . . . . . . . . . . . . . 280 z coord trafo key . . . . . . . . . . . . . . . . . . . . . 362
y SI prefix key . . . . . . . . . . . . . . . . . . . . . . . 500 z dir key . . . . . . . . . . . . . . . . . . . . . 276, 295, 306
y tick label as interval key . . . . . . . . . . . . 320 z error key . . . . . . . . . . . . . . . . . . . . . . . . 50, 298
y tick label style key . . . . . . . . . . . . . . . . . 344 z error expr key . . . . . . . . . . . . . . . . . . . 50, 298
y tick scale label style key . . . . . . . . . . . . 345 z error index key . . . . . . . . . . . . . . . . . . 50, 298
y unit key . . . . . . . . . . . . . . . . . . . . . . . . . . . 499 z error minus key . . . . . . . . . . . . . . . . . . 51, 298
y unit prefix key . . . . . . . . . . . . . . . . . . . . . 499 z error minus expr key . . . . . . . . . . . . . . 51, 299
ybar key . . . . . . . . . . . . . . . . . . . . . . . . . . . 83, 85 z error minus index key . . . . . . . . . . . . . 51, 298
ybar interval key . . . . . . . . . . . . . . . . . . . . . 87f. z error plus key . . . . . . . . . . . . . . . . . . . 50, 298
ybar interval legend key . . . . . . . . . . . . . . . 242 z error plus expr key . . . . . . . . . . . . . . . 51, 298
ybar interval stacked key . . . . . . . . . . . . . . . . 96 z error plus index key . . . . . . . . . . . . . . 50, 298
ybar legend key . . . . . . . . . . . . . . . . . . . . . . . 242 z explicit key . . . . . . . . . . . . . . . . . . . . . . . . 295
ybar stacked key . . . . . . . . . . . . . . . . . . . . . . . 94 z explicit relative key . . . . . . . . . . . . . . . . 296
ycomb key . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89 z expr key . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
ygrid each nth passes x key . . . . . . . . . . . . . 467 z filter key . . . . . . . . . . . . . . . . . . . . . . . . . . 367
ygrid each nth passes x start key . . . . . . . . 468 z fixed key . . . . . . . . . . . . . . . . . . . . . . . . . . . 295
ygrid stop at x key . . . . . . . . . . . . . . . . . . . . 468 z fixed relative key . . . . . . . . . . . . . . . . . . . 295
ylabel key . . . . . . . . . . . . . . . . . . . . . . . . . . . 231 z grid style key . . . . . . . . . . . . . . . . . . . . . . 347
ylabel absolute key . . . . . . . . . . . . . . . . . . . . 233 z index key . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
ylabel near ticks key . . . . . . . . . . . . . . . . . . 232 z label style key . . . . . . . . . . . . . . . . . . . . . 342
ylabel shift key . . . . . . . . . . . . . . . . . . . . . . 232 z post scale key . . . . . . . . . . . . . . . . . . . . . . 280
ylabel style key . . . . . . . . . . . . . . . . . . . . . . 342 z SI prefix key . . . . . . . . . . . . . . . . . . . . . . . 500
ylabels at key . . . . . . . . . . . . . . . . . . . . . . . . 430 z tick label as interval key . . . . . . . . . . . . 320
ymajorgrids key . . . . . . . . . . . . . . . . . . . . . . . 330 z tick label style key . . . . . . . . . . . . . . . . . 344
ymajorticks key . . . . . . . . . . . . . . . . . . . . . . . 321 z tick scale label style key . . . . . . . . . . . . 345
ymax key . . . . . . . . . . . . . . . . . . . 65, 304, 311, 397 z unit key . . . . . . . . . . . . . . . . . . . . . . . . . . . 499
ymin key . . . . . . . . . . . . . . . . . . . 65, 304, 310, 397 z unit prefix key . . . . . . . . . . . . . . . . . . . . . 499
yminorgrids key . . . . . . . . . . . . . . . . . . . . . . . 330 zbar interval legend key . . . . . . . . . . . . . . . 242
yminorticks key . . . . . . . . . . . . . . . . . . . . . . . 321 zbar legend key . . . . . . . . . . . . . . . . . . . . . . . 242
ymode key . . . . . . . . . . . . . . . . . . . . . 275, 306, 376 zlabel key . . . . . . . . . . . . . . . . . . . . . . . . . . . 231
yshift key . . . . . . . . . . . . . . . . . . . . . . . . . . . 522 zlabel absolute key . . . . . . . . . . . . . . . . . . . . 233
ytick key . . . . . . . . . . . . . . . . . . . . . . . . . . . . 311 zlabel near ticks key . . . . . . . . . . . . . . . . . . 232
ytick align key . . . . . . . . . . . . . . . . . . . . . . . 322 zlabel shift key . . . . . . . . . . . . . . . . . . . . . . 232
ytick distance key . . . . . . . . . . . . . . . . . . . . . 311 zlabel style key . . . . . . . . . . . . . . . . . . . . . . 342
ytick placement tolerance key . . . . . . . . . . . 328 zmajorgrids key . . . . . . . . . . . . . . . . . . . . . . . 330
ytick pos key . . . . . . . . . . . . . . . . . . . . . . . . . 322 zmajorticks key . . . . . . . . . . . . . . . . . . . . . . . 321
zmax key . . . . . . . . . . . . . . . . . . . . . . 65, 304, 311
zmin key . . . . . . . . . . . . . . . . . . . . . . 65, 304, 311

zminorgrids key . . . . . . . . . . . . . . . . . . . . . . . 330
zminorticks key . . . . . . . . . . . . . . . . . . . . . . . 321
zmode key . . . . . . . . . . . . . . . . . . . . . . . . 276, 306
ztick key . . . . . . . . . . . . . . . . . . . . . . . . . . . . 311
ztick align key . . . . . . . . . . . . . . . . . . . . . . . 322
ztick distance key . . . . . . . . . . . . . . . . . . . . . 311
ztick placement tolerance key . . . . . . . . . . . 328
ztick pos key . . . . . . . . . . . . . . . . . . . . . . . . . 322
ztick scale label code key . . . . . . . . . . . . . . 326
ztick style key . . . . . . . . . . . . . . . . . . . . . . . 345
zticklabel key . . . . . . . . . . . . . . . . . . . . . . . . 318
zticklabel cs coordinate system . . . . . . . . . . . 222
zticklabel interval boundaries key . . . . . . . . 89
zticklabel pos key . . . . . . . . . . . . . . . . . . . . . 322
zticklabel shift key . . . . . . . . . . . . . . . . . . . 323
zticklabel style key . . . . . . . . . . . . . . . . . . . 344
zticklabel* cs coordinate system . . . . . . . . . . 222
zticklabels key . . . . . . . . . . . . . . . . . . . . . . . 317
zticklabels from table key . . . . . . . . . . . . . . 320
ztickmax key . . . . . . . . . . . . . . . . . . . . . . 258, 322
ztickmin key . . . . . . . . . . . . . . . . . . . . . . 258, 321
ztickten key . . . . . . . . . . . . . . . . . . . . . . . . . . 316
Bibliography
[1] C. Feuersänger. PgfplotsTable package – Loading, rounding and formatting tables in LaTeX. Available
as separate package \usepackage{pgfplotstable}, as part of pgfplots.
[2] C. Feuersänger. Programming in TEX and Library Functions from pgf and pgfplots. Available as part
of pgfplots, TeX-programming-notes.pdf, January 6, 2016.
[3] U. Kern. Extending LATEX’s color facilities: the xcolor package.

[4] Signal and I. P. Institute. “Lena” or “Lenna”. http://sipi.usc.edu/database/database.php?volume=
misc&image=12.
[5] D. P. Story. The AcroTEX eDucation Bundle. http://www.ctan.org/tex-archive/macros/latex/
contrib/acrotex. Sub packages insdljs and eforms are required for the clickable library.
[6] T. Tantau. Tik Z and pgf manual. http://sourceforge.net/projects/pgf. v. 2.00.

[7] K. van Zonneveld. PhP to javascript conversion project (GPL). http://kevin.vanzonneveld.net/
techblog/article/phpjs_licensing.
[8] J. Wright and C. Feuersänger. Implementing keyval input: an introduction. http://pgfplots.
sourceforge.net as .pdf, 2008.
544

PGFplots PDF

Uploaded by

Copyright:

Available Formats

PGFplots PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

PGFplots PDF

Uploaded by

Copyright:

Available Formats

Manual for Package pgfplots

2D/3D Plots in LATEX, Version 1.13

Dr. Christian Feuersänger

Revision 1.13 (2016/01/06)

pgfplots draws high–quality function plots in normal or logarithmic scaling with

2 About PGFPlots: Preliminaries 8

3.5 Solving a Real Use–Case: Functions of Two Variables . . . . . . . . . . . . . . . . . . . . . . 33

4.7 Markers, Linestyles, (Background-) Colors and Colormaps . . . . . . . . . . . . . . . . . . . . 176

4.19 Alignment Options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 348

5 Related Libraries 391

5.10.1 Polar Axes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 449

6 Memory and Speed considerations 502

7 Import/Export From Other Formats 508

8 Utilities and Basic Level Commands 518

• consistency of data and figures (no third party tool necessary),

About pgfplots: Preliminaries

2.2 Upgrade remarks

2.2.1 New Optional Features

in your preamble. This will configure the compatibility layer.

4. pgfplots 1.10 has no di↵erences to 1.9 with respect to compatibility.

2.2.2 Old Features Which May Need Attention

I apologize for any inconvenience caused by these changes.

/pgfplots/compat=1.13|1.12|1.11|1.10|1.9|1.8|1.7|1.6|1.5.1|1.5|1.4|1.3|pre 1.3|default (initially

allows to choose between backwards compatibility and most recent features.

The compat/plot3graphics controls new features for \addplot3 graphics.

2.3 The Team

2.5 Installation and Prerequisites

See the GNU General Public License for more details.

2.5.3 Installation in Windows

2.5.4 Installation of Linux Packages

2.5.5 Installation in Any Directory - the TEXINPUTS Variable

2.5.6 Installation Into a Local TDS Compliant texmf-Directory

2.5.7 Installation If Everything Else Fails...

2.6 Troubleshooting – Error Messages

2.6.1 Problems with available Dimen-registers

2.6.2 Dimension Too Large Errors

2.6.3 Restrictions for DVI-Viewers and dvipdfm

The uncommented commands could be used to set other drivers explicitly.

2.6.4 Problems with TEX’s Memory Capacities

2.6.5 Problems with Language Settings and Active Characters

2.6.6 Other Problems

3.2 Solving a Real Use–Case: Function Visualization

3.2.1 Getting the Data Into TEX

title=Inv. cum. normal,

The code listing already shows a couple of important aspects:

3.2.2 Fine–Tuning of the First Picture

\addplot[blue] table {invcum.dat};

0 0.2 0.4 0.6 0.8 1

3.2.3 Adding the Second Picture with a Di↵erent Plot

Inv. cum. normal

Inv. cum. normal

3.2.5 Satisfying Di↵erent Tastes

3.2.6 Finishing Touches: Automatic Generation of Individual Pdf Graphics

But now, we have to provide a command-line switch to pdflatex :

3.3 Solving a Real Use–Case: Scientific Data Analysis

Finally, the last table is data_d4.dat

3.3.1 Getting the Data into TeX

3.3.2 Adding the Remaining Data Files of Our Example.

\addplot table {data_d2.dat};