0% found this document useful (0 votes)
106 viewsMatplotlib
matplotlib with python
Uploaded by
Vijay DubeyCopyright
© © All Rights Reserved
Available Formats
Download as PDF or read online on Scribd
0% found this document useful (0 votes)
106 viewsMatplotlib
matplotlib with python
Uploaded by
Vijay DubeyCopyright
© © All Rights Reserved
Available Formats
Download as PDF or read online on Scribd
/ 20
Python Plotting With Matplotlib (Guide)
byBtadSolomon ®30Comments > [Gaas)(Gmcsiase)
Mark asCompleted
Table of Contents
Why Can Matpltlib Be Confusing?
‘Pylab: What sit, and Should 1Use I?
“The Matootib Object Hierarchy.
Slatful Versus Stateless
+ The figures” Behind 1
‘Appendix: Configuration and Styling
ixB: Interactive Mode
Opancizads
[EATEN This tutorial has a related video course created by the Real Python team. Watch it together with the
written tutorial to deepen your understanding: Python Plotting With Matplot
“Apicture is worth a thousand words, and with Python's matploti library it fortunately take far less than a
thousand words of code to create a production-qualty graphic.
Improve Your Python However, atplotiis also a massive library, and getting a lotto lookjustrightis often achieved through tial and
error. Using one-tinersto generate basic plots in matplotii is fay simple, bt skilfully commanding the remaining
98% ofthe ibrary can be daunting.
Thisarticleisa beginnerto-intermediate-evel walkthrough on matplotib that mixes theory with examples. While
learning by example canbe tremendously insightful it helps to have even just a surface-level understanding the
brary’ inner workings and layout as wel
Here's what well cover:
+ Pylab and pyplot: whichis which?
+ Key concepts of matpletlb’s design Improve Your Python
+ Understanding ott. subplots()
3} Python Trick
code snippet every couple of days:
+ Visualizing arrays with matplotip sith afresh
+ Plotting with the pandas + matplot
Email Address
Free Bonus: Click hereto download:
2 a bass for making your own plots 2
‘Thisarticle assumes the user knows tin, -
ata, drawing samples from different statistical distribution
instalee, see here fora walkthrough before proceeding.
Learning matpltlb canbe a frustrating process at times. The problem is not that matpltib’s documentation is
lacking: the documentation is actualy extensive, ut the following issues can cause some challenges:
+ The library itself is huge, at something lke 70,000 total lnes of code,
+ Matplotib is home to several diferent interfaces {ways of constructing figure) and capable of interacting with
handful ofeitferent backends. (2ackends deal witn the process of how charts are actully rendered, not just
structured internally.)
+ White itis comprehensive, some of matpoti’s own public documentation is seriously out-of-date, The brary
still evolving, and many older examples floating around ontne may take 70% fewer ines of code in their modern
So, before we getto-any ality examples, its useful to grasp the care concepts of matpotlibs design.
‘Master Real-world Python Skills
‘ith a Community of Experts
Level Up With Unlimited Access fo Our Vast Library
‘of ython Tutorials and Video Lessons
Watch Now »
Pylab: What Is It, and Should I Use It?
Lets start with abit of history: John D. Hunter, a neurobiologst began developing matplotib around 2003 originally
inspired o emulate commands from Mathworks’ MATLAS software, Jahn passed away tragically young at age 4, in
2012, ane matpotib is now afullledged community effort, developed and maintained by ahostof others. (John
gave a talk about the evolution of matpotib atthe 2012 SeiPy conference, which is worth a watch
(One relevant featur of MATLAB\isits global style. The Python concept of importing isnot heavily used in MATLAB, and
‘most of MATLABYS functions are readily available to the user at the top level,
Knowing that matpltlb as its roots in MATLAB helps to explain why pylab exists. pylab isa module within the
ratploti library that was bull to mimic MATLAB' global syle. elsts only to bring 2 numberof functions and
classes from both NumPy and matplotib into the namespace, making for an easy transition fr former MATLAB Users
who were not used to needing ingort statements, [ExMATLAB converts (who areal fine people | promise! ked this functionality, because with from pylab snport
they could simply eal plot) orarray() directly, as they would in MATLAB.
‘Thessue here may be apparent to some Python users using ron pyleb import “Ina session or scripts generally
bad practice. Matpotib now directly advises against this in its own tutorials:
“Ipylab) still exists for historical reasons, butts highly advised not to use tpollutes namespaces with
functions that will shadow Python builtins and can lead to hard-to-track bugs. To get IPython integration
withoutimports the use ofthe anatptot1sb magic is preferred." [Source
Internally there area ton of potentially ©
‘oython ~-pyab from the terminalfcom
ort * under the hood.
Improve Your Python
sth afresh
> Python Tick 2
The bottom nes that matplatib has al ‘cove snippet ever couple of days:
against using pylab, bringing things mo cmaaadvess
Without the need fr pylab, we can usual
Python
While we're ait let's aso import NumPy, which we'l use for generating data later on, and callnp. randon. seed) £0
‘make examples with (pseudorandom data reproducible:
thon
29> prandon seed)
The Matplotlib Object Hierarchy
Cone important bg-picture matplotb concept ists abject hierarchy.
I you've worked through any introductory matplotliptutoral, you've probably called something lke ptt. plot(2, 2,
2). This one-liner hides the fact that a plots really a hierarchy of nested Python objects A“hierarchy" here means
that there isa tree-ke structure of matplotlib objects underlying each plot.
/Arigure objects the outermost container for a matplotlb graphic, which can contain multiple wees objects. One
source af confusions thename: an axes actually translates into what we think of35 an individual pos or graph rather
than te plural of “axis,” as we might expect)
‘You can tink of the Figure object as aboxcke container holding ne or more Axes (actual pts). Below the nee in
the hierarchy are smaller objects suchas tick marks, individual Ines, legends, and textboxes. Almost every “element”
ofachartisits own manipulable Python object, all the way down tothe ticks and labels:
Figure Heres an illustration ofthis hierarchy in action. Don't worry if you'e nt completely familiar with this notation, which
well cover lateron
Python
D> ig, _ = pt smpletscy
>>> sypotiey
‘Above, we created two variables with it.suplots()-The isis atoplevel igure object. The second isa
“throwavay?vatablethatwe dont neediest anand uith an nareron iin at tating He nnn
traverse down te igure hierarchy and st
Improve Your Python
) Python Trick 3
code snippet every couple of days:
sth afresh
stick = fig.anes{0] yards. get,
>>> typetone te)
Email Address
Above, fig a Figure class instance) has
yanss and xaxs, each of which have a col
Matplotib presents this as a igure anato
(totrue matploti style the igure aboveis created inthe matplotlib docs here)
Tes
Mi the »
raha Online Python Training for Teams
Oemeseads
Stateful Versus Stateless Approaches
Aight, we need one more chunk of theory before we can get around tothe shiny visualizations: the cifference
between the stateful state-based, state-machine) and stateless (object-oriented, 00) interfaces.
Above, we used inport eatplotiib.pyplot as pit toimportthe pyplot madule from matplatlb and nameit pit
‘Almost al functions from pyplot, such as pie p2ot(}, are implicitly ether referring toan existing current Figure and
current Axes, or creating them anew if none exist Hidden inthe matplotib docs this helpful snippet: {With pyplot, simple functions are used to add plot elements nes, images, text, etc) tothe current axes in
‘the current figure. [emphasis added]
Hardcore ex-MATLAB users may choose to word ths by saying somethinglike,*pls.plor() is state-machine
interface that implicitly tracks the current figure!” In English, tis means that
+ The stateful interface makes ts alls with pit. ple) and other top-level pyplot functions. There is only ever one
Figure or Axes that you're manipulating at agven time, and you don't need to explicitly refer toi
‘+ Modifying the underying objects directly isthe object-oriented approach. We usually do this by calling methods
af anaxes object, whichis te objec
Improve Your Python
The flow ofthis process, ata high evel, le
sth afresh
3} Python Trick
[eee code snippet every couple of days:
Email Address
‘Tying these together, most ofthe functio:
‘Thisiseasierto see by pecking under the
>>> dat plot args
fe gea0)
ecurn ae plos(-ares, *oares)
5)
Doo dof geahrgs)
turn pltgc#()-2caCorenaegs)
Calling 1¢.plot() is just a convenient way to get the current Axes of the curent Figure and then callitsplot()
‘method. Thisis what is meant by the assertion thatthe stateful interface always “implicit tracks" the plot thatit
wants toreference,
pyplot shome toa batch af functions that are really just wrappers around matpotib’s object-oriented interface. For
example, with pit.title(), there are corresponding setter and getter methods within the 00 approach,
10() (Use of getters and setters tends to be more popularin languages such as Java
but is 2 key feature of matpotibs 00 approach)
au set sitet) and ax. gett
-titlet) gets translated into this one line: gca()-set_titie(s, *args, *shwargs). Here's what thats
e() is setter method that sets the title for that Axes abject. The “convenience” heres that we didn't
need to specify any Axes object explicitly with pie. tstiet)
Similarly, fyou take few moments to look atthe source for topevel functions ike ls.
plt.yabe1s(, youll natice that all of them follow the same structure of delegating
then calling some method ofthe current Axes. (Thisis the underiying object-oriented approach!
gend(), and
cat) and
the euren
Understanding p1t.subplots() Notation
Aight, enough theory, Now, we're ready to tie everything together and do some plotting, From here on out, we'll
‘mostly rely on the stateless (object-oriented) approach, which is more customizable and comesin handy as graphs
become more complex.
The prescribed way to create a Figure witha single Axes unde the 00 approach s (not too intuitively with
pit.subplots(. This is really the only time thatthe 00 approach uses pyplot to create a Figure and Aes: >>> fits ax = ple. suplots(>
‘Above, we took advantage ofterable unpacking to assign a separate variable to each of the two results of
plt.subplots(). Notice that we didn't pass arguments to subplots() here. The default cals subplots(arows-2,
reolse1). Consequently, axisa single txessvoplot object:
thon
>>> ‘ypetand
Improve Your Python
We can callitsinstance methods to man
a stacked area graph of three time series:
5} Python Trick
code snippet every couple of days:
sth afresh
Python
o> me = apsarmen(se) Emailaddress
>>> me = ap randon.randint(®, 18, 51
>>> 4p, ax = pt. svplors(Festze-(5
de> persteexplor(yrs, tag = ond, labessy canconie
doo mecnet title( combined cent growth over tine!)
>>> ae lagend(Lo=' ope” Leet")
Doo pecset ylabel("Torat Seb")
23s peret inGaninyts(@], smaeyesl a1)
>>> Fag tant Layout()
Here's what's going on above:
+ After creating three random time series, we defined one Figure (rig containing one Axes (a plot, 2)
+ We call methods of ax directly to create a stacked area chart and to adda legend, tte, anéy-axs label. Under
the object-oriented approach, it's clear that allo these ae attributes ofa.
+ eighe_tayout() applies tothe Figure object as a whole to clean up whitespace padding,
Combined debt growth over time
PO) mm Oceania
B34
2
Lets look at an example with multiple subplots (Axes) within one Figure, plotting two correlated arrays that re drawn
from the ciscrete uniform distribution >>> X= mpinandon rancine (low, Nah, size)
doo y = nm.randan.randet(t, 5, s8ze%size)
29> data ~ apicolunnstaex((x, 99)
b> fig, (2, 92) = ple aplota(rraues, eaten,
fasize-(s, ©)
oo mscatter(nen, yoy, aakers'o", ce'e', edpecolorsb)
vos pen get title( scoters inh veneer 86°)
>>> ae Set eabel ss")
>>> we set ylabel 35°)
>> weznist(data, ninsong-arangetdet Improve Your Python
Tabel-'s 9)
>>> ax. Legend(loe-(0.65, 8.8)? .with a fresh 3 Pythen Trick
>>> a. set tte Frequencies of $x code snippet every couple of days
oss pe yards tek rent)
Email Address
Scatter: x vi
vee +e
een ] 2
‘There's a tle bit more goingon inthis example
+ Because we'te creating a“1x2" Figure, the returned esult of pit subplots(2, 2) isnow a Figure abject and a
NumPy aray of Axes objects. (You caninspect this with ig, axs = ple.sueplots(2, 2) and takinga look at 2x5)
*+ We deal with axa and axa individually, which would be cificult todo withthe stateful approach. The final ine isa
00d illustration ofthe abject hierarchy, where we are madifying the yexis belonging ta the second Axes,
placing its ticks and tiekiabels tothe right.
+ Text inside dollar signs utilizes eX markup to pu variables in italics.
Remember that multiple Axes can be enclosed inor "belong to” agen figure. In the case above, Fig owes gets usa
lista the Axes objects
oo» (fag sees[0] 45 ae, Fe acest] 4s a2)
(ve, Tove)
(sx. axes is lowercase, not uppercase There's no denying te terminology is abit confusing)
‘Taking this one step further, we could altematively create a figure that holds a 2x2 grid of tees objects:
Python >
D> Fag, Ok = pe ubplote(mrone 2, meets 2, Figsize-(, 79)
Now, what
2x2 i's no longer single wees, but 2 two-dimensional NumPy arta of them >>> eypeton)
srray{[[enatplot1ib.otes._sutplots.AxesSubplot object at &422064a638>,
“matplosLsb.sxas, splat. AxesSubplot object a @a130117495)]
ecipenonsect)
oo Improve Your Python
This is reaffirmed by the docstring sith afresh
5} Python Trick
code snippet every couple of days:
“pecan be ethera single natpiot1i
created” Emailaddress
Wenowneed
plotting methods on
case). common way to address thisist
Doo Hig, ax = I seplos(orons2, neols-2, #3E852e-(7, 7)
D> web, 0, OG, as ae Flatten?) een 2 24 NPY array co ad
We could've also dane this with ((axt, 242), (ax2, ax4)) = ax, but the ist approach tends to be more flexible,
Toilustrate some more advanced subplot features, let's pull some macroeconomic California housing data extracted
from a compressed tar archive, using 4, xarféle, and unl from Python's Standard Library.
Python >
Sytest0fertopen(urt) read}
D> with tanfte openfnoges's, eabj:b) as archive:
housing ~ na loaétnt{archive,extractfile(fpath), deliniter-',")
‘The “response” variable y below, to use the statistical term, isan area's average home value. pep and agearethe area's
population and average house age, respectively
D> y= heusingtts 2)
25> pop, age = housings, [4 711-7
Neat let's define a “helper function” that places a text box inside ofa plot and acts as an “n-pot title’
D> df aed ttehebor(an, text)
Se Fent(.55, 8, tent,
Ssse', aloha.)
fontsize-22-5)
We're ready todo some plotting. Matplolb's gsassac module allows for more subplot customization. pyplots
supplotzgrie() interacts with this module nicely Lets say we want to create a layout lke this Improve Your Python
.with a fresh 3 Pythen Trick
code snippet every couple of days:
Email Address
ad
oe.
axt
oes
204 ‘
0s ax
Above, what we actualy haveis 030g
columns and two rows
oT
1
joe. 1
Vo6 uy
1s eS RE ee
ios : 1
\ 1
oes
1 ' 1
1oo+ + ' 1
Peon sw Ly
Toy ho '
Vos. a tos !
1 1
1°%.5 92 es oo 0 104°%G0 aa !
Loe et ee oe 1 OP or ok ge Be 38 1
‘The second argument to subplotagrid) isthe (ow, column) Location ofthe Axes within the gi
o> enidsize = (3, 2)
>> ig = alt. Figure rigsize-(22, 89)
Doo aah pt supletagridgrisize, (0, 0), colspan2, rawspan2)
23s we = pt seplotzeria(eriersze, (2, 9)
>>> aa = pit sumpletamiaceresize, (2, 0)
"Now, we can proceed 2s normal, modifying each Axes individually:
boo meh seb ite Hone
font size-28)
aut. seatter(seage, ¥-po
pit colorvar(s
bes Set_yscalet
Sos ez nese age, ines
eB nistteop, bins 200
nario")
ste")
doo adé_tilebox(ae,
aee_tittenox(a03,
istoa (doe set.) Home value asa function ofhome age (X-axis) & area population (Y-axis)
Improve Your Python
. sith esh 3) Python Trek 9s
Tt code sippetever couple of day:
we. os Email Address
‘Above, coLervar() (diferent from ColorMap earlier) gets called on the Figure directly, rather than the Ate. sf
argument uses Matpltlb’s .scatter()andisthe result of ax.scatter(), which functions asa mapping ofy-values
a Colona.
Visually, there isn’t much differentiation in color (they-variable) as we move up and down they-axls, indicating that
home age seems to bea stronger determinant of house value,
Real Python for Teams »
The “Figures” Behind The Scenes
Each time you clit. subplots) or the less frequently used pic. +igure() (which creates a Figure, with no Axes), you
‘are creating 2 new Figure object that matplotlesneakly keeps around in memory Earle, we alluded tothe concept
‘of current Figure and current Axes By default, these are the most recently created Figure and Axes, which we can
show with the builtin function sat) to cisplay the address ofthe objectin memory
>> igh, aa = ple sumpletsey
>> ss¢Hien)
doo td(oltigetOo) # “Fer As the current Are
>> fie, a2 = ple. subplots.
doo Ld(Higa) = id(adt.gefO) 4 The current Figure has changed bo “Fig
(ve could also use the builtin ix operator here.)
After the above routine, the current figures 42, the most recently created figure. However, both figures are till
hanging around in memory, each witha corresponding I number (indexed, in MATLAB styl}: >>> pt get_Fignans
ta
‘useful way to get all of the Figures themselves is with a mapping of ple. rigure() to each ofthese integers:
>>> de get_ald_tigures(y
retson [plt-figure(s) for £ Sn plt-get_FSgnans()]
>>> get at flesres
Centploiitigre igure 9 daar Improve Your Python
sth afresh
3 Python Trick
Be cognizant of this eunning a script wh code snippet every couple of days
them ater use to avoid aeroryErrer. By
rpumbernun, and pt. close('312") close: Email Address
Python
bop pt etase(“al")
>> got a iaores)
i
A Burst of Color: imshow() and matshow()
While ax.pior() is ane ofthe mast common plating methods an an Axes, there area whole host of others, aswell
(iveused ax. stackplot() above, You can find the compete list her)
Methods that get heavy use ae inshou() and matshox(), with the latter being a wrapper around the former. These are
Useful anytime thata raw numerical array can be visualized asa colored grid
First, e's create two distinct grids wi
some fancy NumPy indexing:
Python
bos x = mpataglop.aranget, 22))022-1,
D> a etag_andices fren(x(=!-1))] = p.arangel2, 22)
oo 2 np aranga(e size) reshape shape)
Next, we can map these to their image representations, In this specific case we toggle “oll” all axis labels and ticks by
using dictionary comprehension and passing the result to ax. tick parans()
bop sides = (Lett, “rapes “tops “botten'y
oso molabels = (5: false for's in sides)
doo molavels.opeate(("1sbelks) X81 False for sin siees))
ne(nonabels)
(eft; False, ‘night's False, “taps False, “bo
False}
‘Then, we an use a context manager to cisable the grid, and callnatshout) on each Axes. Lastly, we need to put the
colorbarin whats technically anew Ates within rig. For this, we can use abit ofan esoteric function from deep
within matploti: ooo with plt.re_eontoxe(re-(sen gia: False)
fig, (art, 2:2) = ple sibplotstt, 2, figsize-(8, 4))
sua natstowts)
fing? = wx2-natshon(, enape'2eVion.')
for ax in (@xd, 292)
bt tlck_parane(atise “both”, whieh "both’, “tnalabels)
for 8, 4 an riot x-ranters)
eastenttJ, 4, m(Ss J}, cOl0"—"ahite!, man‘center", vac‘center")
oc cuvter append nes( Improve Your Python
fi scolrber (ing eaeany#
Far oriiet nentane wih with afresh
) Python Trick 3
code snippet every couple of days:
Hea
Email Address
6
20
URC
Plotting in Pandas
‘The pandas library has become popular fr not jus for enabling powerful dats analysis, but also forts handy pre
canned platting methods. Interestingly though, pandas plotting methods ae realy ust convenient wrappers around
existing matplotib cals
‘Thatis, the plott) method on pandas’ Series and Datafvame's 3 wrapper around pit.plat()-One convenience
provided, for example, is that ifthe DataFrame's Index consist of dates, ecf().autornt_xdata() is calledintemally by
pandas to get the current Figure and nicely auto-format the xos
In tur, remember that pis. pio() (the state-based approach) is implicitly aware ofthe current Figure and current
‘nes, So pandasis following the state-based approach by extension
We can prove this chain” of function calls with bit of introspection, First let's construct a plain-vanilla pandas
Series, assuming were starting outin afresh interpreter session: pandas a5 oa
>> 5 = pe Sertes(ng. oranges), de
os aes pot
see(9860'9)
>>> sypetan)
‘enatgloe1i9.0%65,_subplots.AsesSubplot at @xI21085eb6>
bo» soit aca) ~ S000)
Thisinterma architectures helpful tok Improve Your Python
cals, whichis done below in plating the
Series fr which we ca alla. 510) th
by thiscall pga) for matplti 01
sith afresh Python Trick
code snippet every couple of days:
Python Email Address
>>> Import pandas 25 pd
>>> import matplotib-eransforns 25
Send Python Tricks »
b> unk = Inttps://feedsthoutsfed.or
>>> WIK = pd.rebd_esv(url, Index col-6, parse dates-rrue, na_values-'.",
squeeze-Truc)-aropna()
995 ma = vbs rottang( 900") reon()
ooo ates pleats, bine-np. in, 8, 8, 28, e9.tne)
labels ranee(4))
29> emap = pe get_ena( Vion =")
>os aa. plot(color-Toisek', Lineieth.s, wankers, Hastze-(8, Dy
abel= "V2" 964 9")
333 an = placa) # Get the currant Ants that ma.plot() refarances
>>> ax set abel)
95) ax set_ylabel(‘200 soving average: CAOE VDC")
23> pecgettésle( voasity Regine See")
oo axentacrnse
555 ax Tegend(toe="upee” center")
>> pecset xlinGamin-na index, xaneea.inéex-10)
boo trans « mtnanafonss blondes transfora factory(ae tranbata, ax: transhxes)
>>> for 4, color in ennerace(cnap([®.2, 8:4, 8.6, @.8)))
a Fill between(na index, @ 1, where-staten-s,
aceccloe=ctler, transfore-trans)
ecavhline(vie.nean(), Linestyle=cishes", color cede 9"
‘Blphacb.6, 2bbel="Full-persed sean", markers")
Volatility Regime State
Soo ta
=~ Fulkperiod mean
8 8
90 moving average: CBOE VIX
8 8
Pog gh PP ptt
There's alothappening above: + nals 90-day moving average ofthe VIX Index, a measure of market expectations of near-term stock volatity
states binning ofthe moving average into itferent regime states, high Vk is seen as signaling a heightened
levelof earn the markerplace
+ enapis a ColorMap—a matplatlb object thats essentially 2 mapping of floats to RGBA colors. Any colormap can
be reversed by appending "_r',50 Ravicn_r" isthe reversed Red Yellow. Green colormap. Matpotib maintains
‘handy visual reference guide to ColorMaps in its docs,
+ The only real pandas call we're making hee isra. plot) Tis cllsplt.piet() internally, soto integrate the
abject oriented approach, we nced to get an explicit reference to the current Axes with ax = pit.geat)
+ The second chunk of ce creates ©
e.8) 5395,
ColorMaps' spectrum.” ensneratet) sith afresh
Improve Your Python
et us an RGBA sequen
5} Python Trick
code snippet every couple of days:
Pandas also comes bultout with a smatt pe "
theirown). However, all ofthese, ke the Email Address
Wrapping Up
‘As shown by some ofthe examples abou.
syntascheavy lbrary Creating a production-ready chart sometimes requires aha hour of Googling and combining a
hodgepodge of lines in order to fine-tune a plot
However, understanding how matpltlib’s interfaces interacts an investment that can pay off down the road. As Real
Python's own Dan Bader has advise, aking the time to dissect code rather than resorting tothe Stack Overflow.
“copy pasta” solution tends tobe a smarter long-term solution ticking to the object-oriented approach can save
hours of rustration when you want to take a plot fom plan to a warkof at
More Resources
From the matplotib documentation:
+ Anindex of matplotlib examples
+ The usage FAQ
+ The tutorials page, which s broken up into beginner, intermediate, and advanced sections
* Lifecycle ofa plot which touches on the object-oriented versus stateful approaches
Free Bonus: Click hereto download 5 Python + Mat
asa basis for making your own plots and graphics
examples with full source code shat you can use
Third-party esources:
+ atacamp's matpltlb cheat shoes
+ PLOS Computational Biology: Ten Simole Rules for Better Figures
+ Chapters (Plotting & Visualization) of Wes McKinney's Bython for Data Analysis, 2d ed.
* Chaper 11 (visualization with Matplotlib, Pandas, and Seaborn) of Ted Petrou's Pandas Cookbook
+ Section 14 (Matplotlb: Potting) ofthe Sey Notes
+ Thesked color palette
+ The matplotib external resources page
from queiroztcom
+ The sualization page_in the pandas documentation
Other plotting libraries:
‘+ The seaborn library, bult ontop of matplotib and designed for advanced statistical graphics, which could take
pan entire tutorial all on its own
+ atashader,a graphics library geared specifically towards large datasets
+ Alistof other third-party packages from the matplotit Fi Become a Python Expert »
Appendix A: Configuration and Styling
you've been following along with tis
stylistically than the ones shown here
ren lookcitferent
torial, it's Ukely thatthe ph
9s popping up on your se
Notpltibofers two ways to configures Improve Your Python
1 Bycustomizing a matplatire fle sith afresh Python Trick
2. By changing your configuration pars code snippet every couple of days
‘Amatpltibrcfile (Option #L above) isbs Emailaddress
between Python sessions. On Mac OSX,
‘Quick Tip: GitHub isa great place tok
contain personaly identifiable or private wiormaton, suchas passwords or 20m private nays
‘Atematively, you can change your configuration parameters interactively (Option #2 above). When you Seport
ratplotlib.pyplot as plt, yOu get access to an rcParans abject that resembles a Python dictionary of settings Allof
the module objects starting with rc" are a means tointeract with your plot styles and settings:
>>> atte for atte in dirgolt) i at
Ure
startewith(re')]
‘earans', ‘"cParansbefaule”, “re_context", “redefaults*)
ofthese:
+ put.redefeuite() restores there parameters from matplatibs internal defauits, which are listed at
pit.reParanabefault This will vert (overurite) whatever you've already customizeé ina matplotibre file.
+ pit.ret) is used for setting parameters interactively
+ pit.reParans sa (mutable) dictionary.ike object that lets you manipulate settings directly. you have
customized settings in a matplotibe file, these willbe reflected inthis dictionary
With pit.re() andpls.reparan, these two syntaxes are equivalent for adjusting settings:
Python =
boo pltce( Mines’, Mnewen2, ot
oy # satan
do pt.echaran{'Lines.Linewidth'] = 2.4 Syntax 2
>> pltsneparanstLines-color')
Notably, the Figure class then uses some of these as its default arguments.
Relatedy, a style isjust a predefined
of custom settings. To vew available styles, use:
Python
[eeatoredari'y“eeatorcdarkgria,‘eaborectiske', “hvethirgeeht
dark bockgraund’, "seabor-poster", "sesborn-deep | .
Toseta style, make this call >>> pt. style uset!svetnioreiahe')
Your plots will now take on anew look:
“fivethirtyeight’ style sheet
» .
on Improve Your Python
sith afesh Bs Python Trek
ao code snippetevery couple of days:
20 mall address
This full example is available here.
For inspiration, matplotlb keeps some style sheet displays for reference as well,
A Learn Python »
en)
emo
Appendix B: Interactive Mode
Behind the scenes, matpltib also interacts wih diffrent backends. A backend’ the workhorse behind actually
rendering a char. (On the popular Anaconda distribution, for instance, the default backend is QtSAgg) Some
backenes are interactive, meaning they are dynamically updated and “pop up" tothe ser when changed
While interactive mode i off by default, you can chec
plt.ssincaractive(), and toggleiton and off with pls. sont) and pit. to##(), respectively
ts status with p1t-reParans interactive") oF
bo» pltneparans{"interactive"] a ort plt.sstnteractivet
Python a
bo» pt. 40FF0)
doo pt ecbarna[ interactive")
In some code examples, you may notice the presence of pit-stou() atthe end ofa chunk of code. The main purpose
of pit.show(), asthe name implies, isto actualy “show” (open) the figure when you're running with interactive mode
turned of n other wards
+ tfinterative mode is on, you don't need pit.shaw(),and images will automatically pop-up and be updated as
youreferencethem,
+ Hfinterative modes off you'll need pit. show() to dlsplay a figure and pis.areu() to update a plot
Below, we make sure that interactive mode is off, which requires that we call pit-shout) after building the plot ise: bee pit. 40f€0
>> pe plete, 7h e's Mnestyle='solie")
>> ae HAL between, JL, 2, wheeryDyt, interpolates,
olor prcsn'y alpha-6.3)
29> land = axlegend(C'yt', "72", loe~‘uppar center’, shadow-True)
>>> gn. get_tave()-set_Facecolor( 8109")
b> pt aha)
Improve Your Python
Notaby interactive oe has nothing to
) Python Trick 3
code snippet every couple of days:
something ike juoyter notebook --naty sith afresh
Email Address
(CTD Tis tutoriat has a eater
written tutorial to deepen your unders
B python Tricks
Geta short & sweet Python Trick delvere to your inbox every couple of
ays. No spam ever. Unsubscribe anytime. Curated by the Real Python
Email Address
About Brad Solomon
Bradisa software engineer and a member of the Real Python Tutorial Team
>More about Brad
ach tutorial at Real Python is created by a team of developers so that it meets our igh quality stondards. The team members who
worked on tis tutorial are: E
Master Real-World Python Skills
With unl’ ~
Improve Your Python
sith afesh Bs Python Trick
ove snippet every couple of ay:
dress
Join us and get access to hundreds of tutorials, hands
‘on video courses, and a community of expert
Pythonistas:
What Do You Think?
(ieee) (Fear) (tem)
Real Python Comment Policy: The mast useful comments are those written with the goal of learning
{rom or helping out other readers~after reading the whole article andall he earler comments.
Complaints and insults generally won't make the cut here.
Whats your#1 takeaway or favorite thing you learned? How are you going to put your newfound sills touse?
Leave a comment below and letus know.
Ins stpoystap ued, nme gud, your the nis stpy-stop ce 30 Comments Real Python Disqus’ Privacy Policy @ Lovin
O Racamenend 24 sory Best
Join the siscussion
seam enstonsewmcncus @
Keep Learning
Improve Your Python
Related Tutorial Categories: (Baia)
with a fresh 3 Python Trick
code snippet every couple of days:
Recommended Video Cours:
Email Address
mall
(No spar. Unsubscribe anytime
All Tutorial Topics
