matplotlib-users Mailing List for matplotlib

I spent some time trying to cobble one together but it was taking too much
time so I tried the following and it seems to work now:

Instead of using a thick line2D and adding it to the drawing, I added a
patches.Rectangle instead.  The line2D works in 0.88.5 but not in 0.99.  The
newer version was tripping over the Artist.contains() function call when
used with line2D.

On Mon, Aug 31, 2009 at 1:28 PM, Michael Droettboom <md...@st...> wrote:

> Can you provide a standalone example that reproduces this error?
>
> Cheers,
> Mike
>
> Andrew Kelly wrote:
>
>> I recently re-installed matplotlib (0.99) on my vista machine and my code
>> that worked yesterday (no changes) no longer runs because of the following
>> matplotlib error:
>>
>>  File "C:\Python25\Lib\site-packages\matplotlib\lines.py", line 286, in
>> contains
>>    path, affine = self._transformed_path.get_transformed_path_and_affine()
>> AttributeError: 'NoneType' object has no attribute
>> 'get_transformed_path_and_affine'
>>
>> All I am doing is initially drawing a line2D (which works) and then
>> testing if line2D.contains(event)==True if I mouse over the line.  As I said
>> this worked fine yesterday.
>>
>> I tried re-installing with an older version but to no avail.  I am totally
>> clueless as to why this is happening.  Anyone have a guess.
>>
>> -Andrew
>>
>> ------------------------------------------------------------------------
>>
>>
>> ------------------------------------------------------------------------------
>> Let Crystal Reports handle the reporting - Free Crystal Reports 2008
>> 30-Day trial. Simplify your report design, integration and deployment - and
>> focus on what you do best, core application coding. Discover what's new with
>> Crystal Reports now.  http://p.sf.net/sfu/bobj-july
>> ------------------------------------------------------------------------
>>
>> _______________________________________________
>> Matplotlib-users mailing list
>> Mat...@li...
>> https://lists.sourceforge.net/lists/listinfo/matplotlib-users
>>
>>
>
> --
> Michael Droettboom
> Science Software Branch
> Operations and Engineering Division
> Space Telescope Science Institute
> Operated by AURA for NASA
>
>

If you need to stick to eps, another option is to use rasterization
feature of the matplotlib itself. This way you can keep part of plot
in vector format (e.g., texts, lines, etc ) while rasterizing others.
Of course this solution only works if the "quality" of those being
rasterized is not very important.

http://matplotlib.sourceforge.net/examples/misc/rasterization_demo.html

The rasterization in the ps backend works in a way that all the artist
whose zorder is smaller than the specified value are rasterized while
those with larger zorder are not.

Regards,

-JJ

Chuck Pepe-Ranney <cp...@mi...> writes:

>> The postscript format does not support transparency.  This is not a
>> matplotlib limitation but a postscript limitation.  The best solution
>> is to save your matplotlib figures as pdf, [...]
>
> Thanks, I guess my problem is that I am using the latex prosper
> package to make presentation slides but I cannot compile prosper
> documents with pdflatex [...]

You could try creating pdf output from matplotlib and converting it to
ps using ps2ps2, which will flatten the transparent objects into one big
bitmap:

http://ghostscript.com/doc/current/Ps2ps2.htm
http://www.ghostscript.com/pipermail/gs-devel/2009-April/008328.html

You may need to tweak the ps2write parameters, especially the output
resolution, to find the best tradeoff between file size and quality.

-- 
Jouni K. Seppänen
http://www.iki.fi/jks

Can you provide a standalone example that reproduces this error?

Cheers,
Mike

Andrew Kelly wrote:
> I recently re-installed matplotlib (0.99) on my vista machine and my 
> code that worked yesterday (no changes) no longer runs because of the 
> following matplotlib error:
>
>   File "C:\Python25\Lib\site-packages\matplotlib\lines.py", line 286, 
> in contains
>     path, affine = 
> self._transformed_path.get_transformed_path_and_affine()
> AttributeError: 'NoneType' object has no attribute 
> 'get_transformed_path_and_affine'
>
> All I am doing is initially drawing a line2D (which works) and then 
> testing if line2D.contains(event)==True if I mouse over the line.  As 
> I said this worked fine yesterday.
>
> I tried re-installing with an older version but to no avail.  I am 
> totally clueless as to why this is happening.  Anyone have a guess.
>
> -Andrew
>
> ------------------------------------------------------------------------
>
> ------------------------------------------------------------------------------
> Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day 
> trial. Simplify your report design, integration and deployment - and focus on 
> what you do best, core application coding. Discover what's new with 
> Crystal Reports now.  http://p.sf.net/sfu/bobj-july
> ------------------------------------------------------------------------
>
> _______________________________________________
> Matplotlib-users mailing list
> Mat...@li...
> https://lists.sourceforge.net/lists/listinfo/matplotlib-users
>   

-- 
Michael Droettboom
Science Software Branch
Operations and Engineering Division
Space Telescope Science Institute
Operated by AURA for NASA

Thanks, I guess my problem is that I am using the latex prosper package to
make presentation slides but I cannot compile prosper documents with
pdflatex (see here <http://www.nefkom.net/georg.drenkhahn/prosper/>).
Sorry, I see that this is not a matplotlib specific question so I will try
to find a solution somewhere else.

On Mon, Aug 31, 2009 at 12:10 PM, John Hunter <jd...@gm...> wrote:

> On Mon, Aug 31, 2009 at 1:04 PM, Chuck Pepe-Ranney<cp...@mi...>
> wrote:
> > Does anybody have a good method for preserving quality and transparancy
> of
> > eps images when going from matplotlib to latex to pdf?  I can only
> preserve
> > the transparency if I save as png and then convert to eps but then I lose
> > quality.  If I save as eps directly, all my patches are opaque despite
> the
> > alpha values.
>
> The postscript format does not support transparency.  This is not a
> matplotlib limitation but a postscript limitation.  The best solution
> is to save your matplotlib figures as pdf, which is also a vector
> format but which does support transparency, and then use pdflatex
>
> JDH
>



-- 
Chuck Pepe-Ranney
Spear Lab
Colorado School of Mines
Work: (303)384-2477
Cell: (575)313-0993

I recently re-installed matplotlib (0.99) on my vista machine and my code
that worked yesterday (no changes) no longer runs because of the following
matplotlib error:

  File "C:\Python25\Lib\site-packages\matplotlib\lines.py", line 286, in
contains
    path, affine = self._transformed_path.get_transformed_path_and_affine()
AttributeError: 'NoneType' object has no attribute
'get_transformed_path_and_affine'

All I am doing is initially drawing a line2D (which works) and then testing
if line2D.contains(event)==True if I mouse over the line.  As I said this
worked fine yesterday.

I tried re-installing with an older version but to no avail.  I am totally
clueless as to why this is happening.  Anyone have a guess.

-Andrew

Teemu Ikonen wrote:

> 
> I also need error band plotting regularly (also the loglog and semilog
> versions) and currently have some quickly written code to do this in
> matplotlib.
> 
> If I clean up my code and submit a patch to axes and pyplot classes,
> would it have a chance to be accepted?
> 
> Teemu

Also, if you haven't already, see 
http://matplotlib.sourceforge.net/devel/coding_guide.html.

Eric

Teemu Ikonen wrote:


> Hi,
> 
> I also need error band plotting regularly (also the loglog and semilog
> versions) and currently have some quickly written code to do this in
> matplotlib.
> 
> If I clean up my code and submit a patch to axes and pyplot classes,
> would it have a chance to be accepted?
> 

Yes.

Eric

> Teemu

On Mon, Aug 31, 2009 at 1:04 PM, Chuck Pepe-Ranney<cp...@mi...> wrote:
> Does anybody have a good method for preserving quality and transparancy of
> eps images when going from matplotlib to latex to pdf?  I can only preserve
> the transparency if I save as png and then convert to eps but then I lose
> quality.  If I save as eps directly, all my patches are opaque despite the
> alpha values.

The postscript format does not support transparency.  This is not a
matplotlib limitation but a postscript limitation.  The best solution
is to save your matplotlib figures as pdf, which is also a vector
format but which does support transparency, and then use pdflatex

JDH

Does anybody have a good method for preserving quality and transparancy of
eps images when going from matplotlib to latex to pdf?  I can only preserve
the transparency if I save as png and then convert to eps but then I lose
quality.  If I save as eps directly, all my patches are opaque despite the
alpha values.

Thanks,
- Chuck

John Hunter wrote:
>On Wed, Aug 5, 2009 at 9:40 AM, per freem<perfreem@gm...> wrote:
>> hi all,
>>
>> is there a way in matplotlib to plot lines with errorbars, e.g. using
>> errorbar(...) but instead of lines just have shaded, partly transparent
>> regions that represent the error bars? people often use this to show
>> confidence intervals or error bars... an example is here:
>>
>> http://eva.nersc.no/vhost/arctic-roos.org/doc/observations/images/ssmi1_ice_ext.png
>>
>> where the black line has shaded grey bands around. i'd like to plot
>> something like this but have the band be partly transparent.
>
>I use fill_between for this -- draw the line with "plot" and the
>shaded region with "fill_between".

Hi,

I also need error band plotting regularly (also the loglog and semilog
versions) and currently have some quickly written code to do this in
matplotlib.

If I clean up my code and submit a patch to axes and pyplot classes,
would it have a chance to be accepted?

Teemu

Jeff, that's a good point. I remember ...

On Thu, 2009-08-27 at 07:54 -0600, Jeff Whitaker wrote:
> Christian Meesters wrote:
> >> Christian:  That should work, if you created the masked array 
> >> correctly.  Why are you creating the mask with data=='NA'?  I suspect 
> >> that this always evaluates to False, so you don't get a mask.  You 
> >> probably want to check for a numeric value, not a string.  For example:
> >>     
> > Thanks a lot, Jeff!
> >
> > Yes, the non-numerical comparison was indeed causing problems - although
> > I don't understand why.
> >
> > However, I can easily inject numerical non-sense values into the array.
> > As to the 'NA': The data are an R output file. As I don't like R too
> > much, I'm falling back to Python.
> >
> > Christian
> >
> >   
> Christian:  What type of array is that (data.dtype)?  I don't see how a 
> numpy array can have values equal to 'NA', unless it is an array of 
> strings.  In that case, it would not be plottable anyway. 
> 
> -Jeff

I'm trying to change the position of the axis label so that it is a 
certain number of points off the end of the spine representing the 
x-axis.  I'm trying to use the transform attribute and the offset_copy 
function to say "put the label 5 points right of the end of the spine".  
So far, my code looks something like:

xlabel=subplot.xaxis.get_label()
xlabel.set_horizontalalignment('left')
xlabel.set_verticalalignment('baseline')
xtrans=offset_copy(subplot.transData, subplot.figure, x=5, y=0, 
units='points')
subplot.xaxis.set_label_coords(subplot.get_xlim()[1], 0,transform=xtrans)

However, this doesn't handle the case where the spine representing the 
x-axis is not the same as the line y=0.  Any ideas on how to construct 
the transform to say "5 points to the right of the right end of this 
given spine"?

Thanks,

Jason

--
Jason Grout

Thanks for the inputs...  perhaps it will provide the impetus for  
future postings as well...

chris

On Aug 29, 2009, at 11:49 AM, John Hunter wrote:

> On Sat, Aug 29, 2009 at 1:12 PM, Eric Firing<ef...@ha...>  
> wrote:
>
>> This looks interesting.  I successfully ran your program by using  
>> copy
>> and paste to get it into a file, but for the future I certainly
>> recommend that you attach such a file directly--file attachments
>> generally work very well these days, but bad things can happen to  
>> code
>> included as inline text.

I haven't contributed to matplotlib or numpy even though I've used  
them for some years now, so I wasn't sure about the "etiquette" of  
file attachments.

The other thing I recommend is do not use the pylab namespace for any
>
> of the numerics.  pylab is getting all the numerical functions from
> numpy, so if you
>
>  import numpy as np
>
> and then refer to any numerical functions you need as np.somefunc.

Point well taken.  Since pylab exposes most of the numpy calls I use,  
I typically include pylab instead for nump.
>
> Finally, for the functions to be suitable for inclusion in a
> production package like numpy or matplotlib.mlab, you should not use
> any print statements in the function, but rather a combination of
> warnings.warn or exceptions or if it for matplotlib, use the
> verbose.report infrastructure.  That way users can configure how much
> verbosity they want, where the output should be directed, etc.

Point also well taken.  I figured out when there were problems, but  
even after 7 years of writing large Python package, I haven't found  
the best way to handle exceptions.  Usually I purposely cause a  
"crash" so I don't miss the fact that the code had ill formed data.
>
> After a cleanup, you may want to check with numpy or scipy to see of
> it could find a home there.  There was a discussion at scipy on the
> need to improve scipy.interpolate and this seems to go part of the way
> toward that objective.  So I would start there.

I'll send it along to the scipy people.  I figured since I figured out  
a relatively simple solution to a problem that is often encountered,  
it might find use even in its primitive form.  I'll add the URLs to  
the WIkipedia references as well.
>
> JDH

Hi
When I try to import anything from mpl, Python's interpreter exits
without any error.
This happens after the 0.99.0 upgrade. This is example session:

d:\>python
Python 2.6.2 (r262:71605, Apr 14 2009, 22:40:02) [MSC v.1500 32 bit
(Intel)] on win32
Type "help", "copyright", "credits" or "license" for more information.
>>> import matplotlib.pyplot as plt

matplotlib data path D:\msys\opt\python\lib\site-packages\matplotlib\mpl-data
loaded rc file D:\msys\etc\.matplotlib\matplotlibrc
matplotlib version 0.99.0
verbose.level debug-annoying
interactive is False
units is False
platform is win32
loaded modules: ['numpy.lib._iotools', 'xml.sax.urlparse',
'distutils', 'functools', 'matplotlib.matplotlib', 'subprocess', 'gc',
'matplotlib.tempfile', 'distutils.sysconfig', 'ctypes._endian',
'encodings.encodings', 'matplotlib.colors', 'msvcrt',
'numpy.testing.sys', 'numpy.core.info', 'xml', 'numpy.fft.types',
'numpy.ma.operator', 'numpy.ma.cPickle', 'matplotlib.copy',
'numpy.random.info', 'tempfile', 'base64', 'numpy.linalg',
'matplotlib.threading', 'numpy.testing.operator', 'imp',
'numpy.testing', 'collections', 'numpy.core.umath', '_struct',
'distutils.types', 'numpy.lib.numpy', 'numpy.core.scalarmath',
'zipimport', 'string', 'matplotlib.subprocess', 'numpy.testing.os',
'matplotlib.locale', 'numpy.lib.arraysetops',
'numpy.testing.unittest', 'numpy.lib.math', 'textwrap',
'matplotlib.__future__', 'ssl', 'numpy.testing.re', 'itertools',
'numpy.version', 'numpy.lib.re', 'distutils.re',
'ctypes.os', 'numpy.core.os', 'numpy.lib.type_check',
'numpy.lib.__builtin__', 'signal', 'numpy.lib.types',
'numpy.lib._datasource', 'random', 'threading', 'token',
'numpy.fft.fftpack_lite', 'matplotlib.cbook', 'ctypes.ctypes',
'xml.sax.xmlreader', 'numpy.__builtin__', 'dis', 'distutils.version',
'cStringIO', 'numpy.ma.core', 'encodings.cp852',
'matplotlib.StringIO', 'numpy.ma.extras', 'locale',
'numpy.add_newdocs', 'numpy.lib.getlimits', 'xml.sax.saxutils',
'matplotlib.numpy', 'numpy.lib.sys',
'encodings', 'numpy.ma.itertools', 'StringIO', 'numpy.lib.io', 'abc',
'numpy.ctypes', 'numpy.testing.decorators', 'matplotlib.warnings',
'matplotlib.string', '_subprocess', 'urllib', 'matplotlib.sys', 're',
'numpy.lib._compiled_base', 'ntpath', 'numpy.random.mtrand', 'math',
'numpy.fft.helper', 'numpy.ma.warnings', 'inspect',
'numpy.ma.inspect', 'UserDict', 'numpy.lib.function_base',
'distutils.os', 'matplotlib', 'numpy.fft.numpy', 'xml.sax.codecs',
'exceptions', 'numpy.lib.info', 'numpy.core.numerictypes', 'ctypes',
'numpy.lib.warnings', 'ctypes.struct', 'codecs', 'numpy.core._sort',
'numpy.os', 'struct', '_functools', '_locale',
'matplotlib.sre_constants', 'matplotlib.os', 'thread',
'numpy.lib.ufunclike', 'numpy.core.memmap', 'traceback', 'weakref',
'numpy.core._internal', 'numpy.fft.fftpack', 'opcode',
'numpy.linalg.lapack_lite', 'distutils.sys', 'os',
'numpy.lib.itertools', '__future__', '_collections', 'xml.sax.types',
'matplotlib.traceback', '_sre', 'unittest', 'numpy.core.sys',
'numpy.random', 'numpy.linalg.numpy', '__builtin__',
'numpy.lib.twodim_base', 'matplotlib.re', 'numpy.core.cPickle',
'operator', 'numpy.core.arrayprint', 'distutils.string',
'numpy.lib.arrayterator',
'numpy.numpy', 'ctypes.sys', 'matplotlib.errno',
'numpy.lib.financial', 'numpy.core.multiarray', 'errno', '_socket',
'binascii', 'sre_constants', 'datetime', 'numpy.ma',
'xml.sax.handler', 'os.path',
'tokenize', 'numpy.lib.stride_tricks', 'numpy.core.numpy', 'numpy',
'_warnings', 'matplotlib.types', 'numpy.core.defmatrix', 'xml.sax.os',
'cPickle', 'encodings.__builtin__', 'matplotlib.xml', '_codecs',
'numpy.lib.operator', 'numpy.__config__', 'matplotlib.pyparsing',
'nturl2path', 'numpy.ma.numpy', 'copy', 'numpy.core.re', 'socket',
'numpy.core.fromnumeric', 'numpy.ctypeslib', 'keyword',
'numpy.lib.scimath', 'numpy.fft', 'numpy.lib', 'numpy.random.numpy',
'encodings.aliases', 'matplotlib.distutils', 'fnmatch', 'sre_parse',
'numpy.core.ctypes', 'distutils.distutils', 'copy_reg', 'sre_compile',
'xml.sax', '_random', 'numpy.lib.__future__', 'site',
'numpy.lib.polynomial', 'encodings.cp1250',
'numpy._import_tools', '__main__', 'numpy.fft.info',
'numpy.core.records', 'shutil', 'numpy.lib.cPickle', 'numpy.sys',
'matplotlib.weakref', 'xml.sax.urllib', 'numpy.testing.traceback',
'strop', 'numpy.testing.numpytest', 'numpy.core.numeric',
'numpy.linalg.info', 'encodings.codecs', 'ctypes._ctypes', '_abcoll',
'numpy.core', 'matplotlib.rcsetup', 'matplotlib.time', 'nt',
'xml.sax._exceptions', 'genericpath', 'stat', '_ssl',
'numpy.lib.index_tricks', 'warnings', 'numpy.lib.utils',
'numpy.core.defchararray', '_ctypes', 'numpy.lib.shape_base',
'numpy.core.types', 'sys', 'numpy.core.warnings',
'numpy.core.__builtin__', 'xml.sax.sys', 'numpy.lib.format',
'numpy.lib.os', 'numpy.testing.nosetester', 'types',
'numpy.lib.shutil', 'matplotlib.datetime',
'matplotlib.fontconfig_pattern', '_weakref',
'distutils.errors', 'urlparse', 'linecache', 'matplotlib.shutil',
'numpy.lib.cStringIO', 'time', 'numpy.lib.machar',
'numpy.linalg.linalg', 'numpy.testing.utils']

d:\>


Can somebody help me??

On Sat, Aug 29, 2009 at 1:12 PM, Eric Firing<ef...@ha...> wrote:

> This looks interesting.  I successfully ran your program by using copy
> and paste to get it into a file, but for the future I certainly
> recommend that you attach such a file directly--file attachments
> generally work very well these days, but bad things can happen to code
> included as inline text.

The other thing I recommend is do not use the pylab namespace for any
of the numerics.  pylab is getting all the numerical functions from
numpy, so if you

  import numpy as np

and then refer to any numerical functions you need as np.somefunc.

Finally, for the functions to be suitable for inclusion in a
production package like numpy or matplotlib.mlab, you should not use
any print statements in the function, but rather a combination of
warnings.warn or exceptions or if it for matplotlib, use the
verbose.report infrastructure.  That way users can configure how much
verbosity they want, where the output should be directed, etc.

After a cleanup, you may want to check with numpy or scipy to see of
it could find a home there.  There was a discussion at scipy on the
need to improve scipy.interpolate and this seems to go part of the way
toward that objective.  So I would start there.

JDH

Chris Michalski wrote:

> 
> Offered for those who might have the same need – a Python pchip() 
> equivalent ==> pypchip().  Since I'm not sure how attachments work (or 
> if they work at all...), I copied the code I used below, followed by a 
> PNG showing "success":

Chris,

This looks interesting.  I successfully ran your program by using copy 
and paste to get it into a file, but for the future I certainly 
recommend that you attach such a file directly--file attachments 
generally work very well these days, but bad things can happen to code 
included as inline text.

It would also be helpful if you include the actual URLs of the Wikipedia 
articles that you used.  (Or did I miss them?)

Thanks for providing this code, example, and references.

Eric

Is this the only way to plot a sin from x=0 to x=2pi, but shade where x>0.25pi and x<0.75pi?

x0 = np.linspace(0.0, 0.25*np.pi, 100)
x1 = np.linspace(0.25*np.pi, 0.75*np.pi, 100)
x2 = np.linspace(0.75*np.pi,2*np.pi, 100)

y0 = np.sin(x0)
y1 = np.sin(x1)
y2 = np.sin(x2)

fig = figure()
ax1 = fig.add_subplot(111)

ax1.plot(x0,y0)
ax1.fill_between(x1,0,y1)
ax1.plot(x2,y2)
show() 

thanks

Marc Desmarais





      

here's an example

a call to pyplot.axis() after I call pcolor straightens it out though. thx
gokhan, and sorry to all for not picking that up in the docs myself.

and finally, `demolishor' is a great and hilarious song by a band called the
acacia strain, in case there are parties interested in music that might make
your officemates think you're satanic. i am delighted this thread generated
such an enthusiastic response in this regard.

ta ta

On Fri, Aug 28, 2009 at 3:32 PM, John Hunter <jd...@gm...> wrote:

> On Fri, Aug 28, 2009 at 1:59 PM, DEMOLISHOR! the
> Demolishor<des...@gm...> wrote:
> > This is the result from a call to pyplot.pcolor() -- why do the axes
> > automatically expand beyond the range of the data? And how can I set them
> > back? I am not seeing an obvious keyword argument in the pcolor docs to
> do
> > this...
> >
>
> could you post a complete example that replicates the problem?
>
> JDH
>

Here you go.

On Fri, Aug 28, 2009 at 1:59 PM, DEMOLISHOR! the
Demolishor<des...@gm...> wrote:
> This is the result from a call to pyplot.pcolor() -- why do the axes
> automatically expand beyond the range of the data? And how can I set them
> back? I am not seeing an obvious keyword argument in the pcolor docs to do
> this...
>

could you post a complete example that replicates the problem?

JDH

Finetuning with axis(xmin=, xmax= , ymin= , ymax=) should yield a better
looking result.

You might use figure(frameon=False) for a transparent background.

On Fri, Aug 28, 2009 at 1:59 PM, DEMOLISHOR! the Demolishor <
des...@gm...> wrote:

> This is the result from a call to pyplot.pcolor() -- why do the axes
> automatically expand beyond the range of the data? And how can I set them
> back? I am not seeing an obvious keyword argument in the pcolor docs to do
> this...
>
> Thanks in advance
>
> - Craig
>
>
> ------------------------------------------------------------------------------
> Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day
> trial. Simplify your report design, integration and deployment - and focus
> on
> what you do best, core application coding. Discover what's new with
> Crystal Reports now.  http://p.sf.net/sfu/bobj-july
> _______________________________________________
> Matplotlib-users mailing list
> Mat...@li...
> https://lists.sourceforge.net/lists/listinfo/matplotlib-users
>
>


-- 
Gökhan

Recently, I had a need for a monotonic piece-wise cubic Hermite  
interpolator.   Matlab provides the function "pchip" (Piecewise Cubic  
Hermite Interpolator), but when I Googled I didn't find any Python  
equivalent.   I tried "interp1d()" from scipy.interpolate but this was  
a standard cubic spline using all of the data - not a piece-wise cubic  
spline.

I had access to Matlab documentation, so I spent a some time tracing  
through the code to figure out how I might write a Python duplicate.   
This was an massive exercise in frustration and a potent reminder on  
why I love Python and use Matlab only under duress.  I find typical  
Matlab code is poorly documented (if at all) and that apparently  
includes the code included in their official releases.  I also find  
Matlab syntax “dated” and the code very difficult to “read”.

Wikipedia to the rescue.

Not to be deterred, I found a couple of very well written Wikipedia  
entries, which explained in simple language how to compute the  
interpolant.  Hats off to whoever wrote these entries – they are  
excellent.  The result was a surprising small amount of code  
considering the Matlab code was approaching 10 pages of  
incomprehensible code. Again - strong evidence that things are just  
better in Python...

Offered for those who might have the same need – a Python pchip()  
equivalent ==> pypchip().  Since I'm not sure how attachments work (or  
if they work at all...), I copied the code I used below, followed by a  
PNG showing "success":

#
#    pychip.py
#    Michalski
#    20090818
#
#    Piecewise cubic Hermite interpolation (monotonic...) in Python
#
#    References:
#
#        Wikipedia:  Monotone cubic interpolation
#                    Cubic Hermite spline
#
#    A cubic Hermte spline is a third degree spline with each  
polynomial of the spline
#    in Hermite form.  The Hermite form consists of two control points  
and two control
#    tangents for each polynomial.  Each interpolation is performed on  
one sub-interval
#    at a time (piece-wise).  A monotone cubic interpolation is a  
variant of cubic
#    interpolation that preserves monotonicity of the data to be  
interpolated (in other
#    words, it controls overshoot).  Monotonicity is preserved by  
linear interpolation
#    but not by cubic interpolation.
#
#    Use:
#
#    There are two separate calls, the first call, pchip_init(),  
computes the slopes that
#    the interpolator needs.  If there are a large number of points to  
compute,
#    it is more efficient to compute the slopes once, rather than for  
every point
#    being evaluated.  The second call, pchip_eval(), takes the slopes  
computed by
#    pchip_init() along with X, Y, and a vector of desired "xnew"s and  
computes a vector
#    of "ynew"s.  If only a handful of points is needed, pchip() is a  
third function
#    which combines a call to pchip_init() followed by pchip_eval().
#

import pylab as P

#=========================================================
def pchip(x, y, xnew):

     # Compute the slopes used by the piecewise cubic Hermite  
interpolator
     m    = pchip_init(x, y)

     # Use these slopes (along with the Hermite basis function) to  
interpolate
     ynew = pchip_eval(x, y, xnew)

     return ynew

#=========================================================
def x_is_okay(x,xvec):

     # Make sure "x" and "xvec" satisfy the conditions for
     # running the pchip interpolator

     n = len(x)
     m = len(xvec)

     # Make sure "x" is in sorted order (brute force, but works...)
     xx = x.copy()
     xx.sort()
     total_matches = (xx == x).sum()
     if total_matches != n:
         print "*" * 50
         print "x_is_okay()"
         print "x values weren't in sorted order --- aborting"
         return False

     # Make sure 'x' doesn't have any repeated values
     delta = x[1:] - x[:-1]
     if (delta == 0.0).any():
         print "*" * 50
         print "x_is_okay()"
         print "x values weren't monotonic--- aborting"
         return False

     # Check for in-range xvec values (beyond upper edge)
     check = xvec > x[-1]
     if check.any():
         print "*" * 50
         print "x_is_okay()"
         print "Certain 'xvec' values are beyond the upper end of 'x'"
         print "x_max = ", x[-1]
         indices = P.compress(check, range(m))
         print "out-of-range xvec's = ", xvec[indices]
         print "out-of-range xvec indices = ", indices
         return False

     # Second - check for in-range xvec values (beyond lower edge)
     check = xvec< x[0]
     if check.any():
         print "*" * 50
         print "x_is_okay()"
         print "Certain 'xvec' values are beyond the lower end of 'x'"
         print "x_min = ", x[0]
         indices = P.compress(check, range(m))
         print "out-of-range xvec's = ", xvec[indices]
         print "out-of-range xvec indices = ", indices
         return False

     return True

#=========================================================
def pchip_eval(x, y, m, xvec):

     # Evaluate the piecewise cubic Hermite interpolant with  
monoticity preserved
     #
     #    x = array containing the x-data
     #    y = array containing the y-data
     #    m = slopes at each (x,y) point [computed to preserve  
monotonicity]
     #    xnew = new "x" value where the interpolation is desired
     #
     #    x must be sorted low to high... (no repeats)
     #    y can have repeated values
     #
     # This works with either a scalar or vector of "xvec"

     n = len(x)
     mm = len(xvec)

     ############################
     # Make sure there aren't problems with the input data
     ############################
     if not x_is_okay(x, xvec):
         print "pchip_eval2() - ill formed 'x' vector!!!!!!!!!!!!!"

         # Cause a hard crash...
         STOP_pchip_eval2

    # Find the indices "k" such that x[k] < xvec < x[k+1]

     # Create "copies" of "x" as rows in a mxn 2-dimensional vector
     xx = P.resize(x,(mm,n)).transpose()
     xxx = xx > xvec

     # Compute column by column differences
     z = xxx[:-1,:] - xxx[1:,:]

     # Collapse over rows...
     k = z.argmax(axis=0)

     # Create the Hermite coefficients
     h = x[k+1] - x[k]
     t = (xvec - x[k]) / h[k]

     # Hermite basis functions
     h00 = (2 * t**3) - (3 * t**2) + 1
     h10 =      t**3  - (2 * t**2) + t
     h01 = (-2* t**3) + (3 * t**2)
     h11 =      t**3  -      t**2

     # Compute the interpolated value of "y"
     ynew = h00*y[k] + h10*h*m[k] + h01*y[k+1] + h11*h*m[k+1]

     return ynew

#=========================================================
def pchip_init(x,y):

     # Evaluate the piecewise cubic Hermite interpolant with  
monoticity preserved
     #
     #    x = array containing the x-data
     #    y = array containing the y-data
     #
     #    x must be sorted low to high... (no repeats)
     #    y can have repeated values
     #
     #    x input conditioning is assumed but not checked

     n = len(x)

     # Compute the slopes of the secant lines between successive points
     delta = (y[1:] - y[:-1]) / (x[1:] - x[:-1])

     # Initialize the tangents at every points as the average of the  
secants
     m = P.zeros(n, dtype='d')

     # At the endpoints - use one-sided differences
     m[0] = delta[0]
     m[n-1] = delta[-1]

     # In the middle - use the average of the secants
     m[1:-1] = (delta[:-1] + delta[1:]) / 2.0

     # Special case: intervals where y[k] == y[k+1]

     # Setting these slopes to zero guarantees the spline connecting
     # these points will be flat which preserves monotonicity
     indices_to_fix = P.compress((delta == 0.0), range(n))

#    print "zero slope indices to fix = ", indices_to_fix

     for ii in indices_to_fix:
         m[ii]   = 0.0
         m[ii+1] = 0.0

     alpha = m[:-1]/delta
     beta  = m[1:]/delta
     dist  = alpha**2 + beta**2
     tau   = 3.0 / P.sqrt(dist)

     # To prevent overshoot or undershoot, restrict the position vector
     # (alpha, beta) to a circle of radius 3.  If (alpha**2 +  
beta**2)>9,
     # then set m[k] = tau[k]alpha[k]delta[k] and m[k+1] =  
tau[k]beta[b]delta[k]
     # where tau = 3/sqrt(alpha**2 + beta**2).

     # Find the indices that need adjustment
     over = (dist > 9.0)
     indices_to_fix = P.compress(over, range(n))

#    print "overshoot indices to fix... = ", indices_to_fix

     for ii in indices_to_fix:
         m[ii]   = tau[ii] * alpha[ii] * delta[ii]
         m[ii+1] = tau[ii] * beta[ii]  * delta[ii]

     return m

#= 
=======================================================================
def CubicHermiteSpline(x, y, x_new):

     # Piecewise Cubic Hermite Interpolation using Catmull-Rom
     # method for computing the slopes.
     #
     # Note - this only works if delta-x is uniform?

     # Find the two points which "bracket" "x_new"
     found_it = False

     for ii in range(len(x)-1):
         if (x[ii] <= x_new) and (x[ii+1] > x_new):
             found_it = True
             break

     if not found_it:
         print
         print "requested x=<%f> outside X range[%f,%f]" % (x_new,  
x[0], x[-1])
         STOP_CubicHermiteSpline()

     # Starting and ending data points
     x0 = x[ii]
     x1 = x[ii+1]

     y0 = y[ii]
     y1 = y[ii+1]

     # Starting and ending tangents (using Catmull-Rom spline method)

     # Handle special cases (hit one of the endpoints...)
     if ii == 0:

         # Hit lower endpoint
         m0 = (y[1] - y[0])
         m1 = (y[2] - y[0]) / 2.0

     elif ii == (len(x) - 2):

         # Hit upper endpoints
         m0 = (y[ii+1] - y[ii-1]) / 2.0
         m1 = (y[ii+1] - y[ii])

     else:

         # Inside the field...
         m0 = (y[ii+1] - y[ii-1])/ 2.0
         m1 = (y[ii+2] - y[ii])  / 2.0

     # Normalize to x_new to [0,1] interval
     h = (x1 - x0)
     t = (x_new - x0) / h

     # Compute the four Hermite basis functions
     h00 = ( 2.0 * t**3) - (3.0 * t**2) + 1.0
     h10 = ( 1.0 * t**3) - (2.0 * t**2) + t
     h01 = (-2.0 * t**3) + (3.0 * t**2)
     h11 = ( 1.0 * t**3) - (1.0 * t**2)

     h = 1

     y_new = (h00 * y0) + (h10 * h * m0) + (h01 * y1) + (h11 * h * m1)

     return y_new

#==============================================================
def main():

     ############################################################
     # Sine wave test
     ############################################################

     # Create a example vector containing a sine wave.
     x = P.arange(30.0)/10.
     y = P.sin(x)

     # Interpolate the data above to the grid defined by "xvec"
     xvec = P.arange(250.)/100.

     # Initialize the interpolator slopes
     m = pchip_init(x,y)

     # Call the monotonic piece-wise Hermite cubic interpolator
     yvec2 = pchip_eval(x, y, m, xvec)

     P.figure(1)
     P.plot(x,y, 'ro')
     P.title("pchip() Sin test code")

     # Plot the interpolated points
     P.plot(xvec, yvec2, 'b')

      
#########################################################################
     # Step function test...
      
#########################################################################

     P.figure(2)
     P.title("pchip() step function test")

     # Create a step function (will demonstrate monotonicity)
     x = P.arange(7.0) - 3.0
     y = P.array([-1.0, -1,-1,0,1,1,1])

     # Interpolate using monotonic piecewise Hermite cubic spline
     xvec = P.arange(599.)/100. - 3.0

     # Create the pchip slopes slopes
     m    = pchip_init(x,y)

     # Interpolate...
     yvec = pchip_eval(x, y, m, xvec)

     # Call the Scipy cubic spline interpolator
     from scipy.interpolate import interpolate
     function = interpolate.interp1d(x, y, kind='cubic')
     yvec2 = function(xvec)

     # Non-montonic cubic Hermite spline interpolator using
     # Catmul-Rom method for computing slopes...
     yvec3 = []
     for xx in xvec:
         yvec3.append(CubicHermiteSpline(x,y,xx))
     yvec3 = P.array(yvec3)

     # Plot the results
     P.plot(x,    y,     'ro')
     P.plot(xvec, yvec,  'b')
     P.plot(xvec, yvec2, 'k')
     P.plot(xvec, yvec3, 'g')
     P.xlabel("X")
     P.ylabel("Y")
     P.title("Comparing pypchip() vs. Scipy interp1d() vs. non- 
monotonic CHS")
     legends = ["Data", "pypchip()", "interp1d","CHS"]
     P.legend(legends, loc="upper left")

     P.show()

###################################################################
main()

Michael Droettboom wrote:
> It's important to note that we're not, as far as I'm aware, 
> considering anything lighting or ray-tracing-like yet.  I'm not sure 
> that Phong shading actually makes sense without doing that.  The 
> Gouraud shading we're in the process of implementing uses only 
> explicit colors specified in the plot, not any sort of 
> light-source-calculated values.
>
> Add perhaps more importantly, Gouraud is implemented in a number of 
> vector formats, including PDF and PS (though matplotlib doesn't have 
> support for that yet, that's coming...)
>
> Mike

Mike:  I did add some support for crude light-source shading a while 
back.  It works directly on rgb values - see 
pylab_examples/shading_example.py.

-Jeff
>
> On 08/26/2009 07:06 PM, Nicolas Bigaouette wrote:
>> Its great news that the 3D is receiving more polish :)
>>
>> After reading on wikipedia, wouldn't it be nicier to have Phong 
>> reflection[1] instead of Gouraud?
>>
>> Maybe it would be too hard, as the Gouraud seems to be implemented 
>> directly in Agg (from what others just said...)
>>
>> Anyway, just a suggestion ;)
>>
>>
>> [1] http://en.wikipedia.org/wiki/Phong_reflection_model
>>
>>
>> 2009/8/26 Eric Firing <ef...@ha... <mailto:ef...@ha...>>
>>
>>     Reinier Heeres wrote:
>>     > Hi JJ,
>>     >
>>     > Thanks for the examples! I indeed suspected the anti-aliasing
>>     as well.
>>     > I'll include an option for linecolors = facecolors soon
>>     (probably this
>>     > weekend); it seems to be a good solution.
>>
>>     No, not in general.  There are two problems: it distorts the
>>     sizes and
>>     positions of the patches, very slightly, but enough that it might
>>     be a
>>     problem in some applications; and it does not work with alpha != 1.
>>
>>     Eric
>>
>>     ------------------------------------------------------------------------------
>>     Let Crystal Reports handle the reporting - Free Crystal Reports
>>     2008 30-Day
>>     trial. Simplify your report design, integration and deployment -
>>     and focus on
>>     what you do best, core application coding. Discover what's new with
>>     Crystal Reports now.  http://p.sf.net/sfu/bobj-july
>>     _______________________________________________
>>     Matplotlib-users mailing list
>>     Mat...@li...
>>     <mailto:Mat...@li...>
>>     https://lists.sourceforge.net/lists/listinfo/matplotlib-users
>>
>>
>>  
>> ------------------------------------------------------------------------------
>> Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day 
>> trial. Simplify your report design, integration and deployment - and focus on 
>> what you do best, core application coding. Discover what's new with 
>> Crystal Reports now.  http://p.sf.net/sfu/bobj-july
>>  
>> _______________________________________________
>> Matplotlib-users mailing list
>> Mat...@li... <mailto:Mat...@li...>
>> https://lists.sourceforge.net/lists/listinfo/matplotlib-users
>>   
>
> ------------------------------------------------------------------------
>
> ------------------------------------------------------------------------------
> Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day 
> trial. Simplify your report design, integration and deployment - and focus on 
> what you do best, core application coding. Discover what's new with 
> Crystal Reports now.  http://p.sf.net/sfu/bobj-july
> ------------------------------------------------------------------------
>
> _______________________________________________
> Matplotlib-users mailing list
> Mat...@li...
> https://lists.sourceforge.net/lists/listinfo/matplotlib-users
>   

Maybe with the new spines support in 0.99 this could be an 
attribute/method on the spine?

Mike

On 08/27/2009 02:18 PM, Jae-Joon Lee wrote:
> I don't think there is a direct support for this in mpl and I guess
> only way is to adjust the parameters of each ticks.
>
> def set_ticks_both(axis):
>      ticks = list( axis.majorTicks ) # a copy
>      ticks.extend( axis.minorTicks )
>
>      for t in ticks:
>          t.tick1On = True # tick marker on left (or bottom)
>          t.tick2On = True # tick marker on right (or top)
>          t.label1On = True # tick label marker on left (or bottom)
>          t.label2On = True # tick label on right (or top)
>
> set_ticks_both(ax.yaxis)
>
> -JJ
>
>
>
> On Thu, Aug 27, 2009 at 7:22 AM, Peter Saffrey<pz...@dc...>  wrote:
>    
>> Some of my plots are very wide. I'd like Y axis labels on both sides so
>> that it's clearer to read the bars towards the right hand side.
>>
>> I can change the ticks with yaxis.set_ticks_position("both") but there
>> doesn't seem to be a similar call for labels. Any ideas?
>>
>> Thanks,
>>
>> Peter
>>
>> ------------------------------------------------------------------------------
>> Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day
>> trial. Simplify your report design, integration and deployment - and focus on
>> what you do best, core application coding. Discover what's new with
>> Crystal Reports now.  http://p.sf.net/sfu/bobj-july
>> _______________________________________________
>> Matplotlib-users mailing list
>> Mat...@li...
>> https://lists.sourceforge.net/lists/listinfo/matplotlib-users
>>
>>      
> ------------------------------------------------------------------------------
> Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day
> trial. Simplify your report design, integration and deployment - and focus on
> what you do best, core application coding. Discover what's new with
> Crystal Reports now.  http://p.sf.net/sfu/bobj-july
> _______________________________________________
> Matplotlib-users mailing list
> Mat...@li...
> https://lists.sourceforge.net/lists/listinfo/matplotlib-users
>    

It's important to note that we're not, as far as I'm aware, considering 
anything lighting or ray-tracing-like yet.  I'm not sure that Phong 
shading actually makes sense without doing that.  The Gouraud shading 
we're in the process of implementing uses only explicit colors specified 
in the plot, not any sort of light-source-calculated values.

Add perhaps more importantly, Gouraud is implemented in a number of 
vector formats, including PDF and PS (though matplotlib doesn't have 
support for that yet, that's coming...)

Mike

On 08/26/2009 07:06 PM, Nicolas Bigaouette wrote:
> Its great news that the 3D is receiving more polish :)
>
> After reading on wikipedia, wouldn't it be nicier to have Phong 
> reflection[1] instead of Gouraud?
>
> Maybe it would be too hard, as the Gouraud seems to be implemented 
> directly in Agg (from what others just said...)
>
> Anyway, just a suggestion ;)
>
>
> [1] http://en.wikipedia.org/wiki/Phong_reflection_model
>
>
> 2009/8/26 Eric Firing <ef...@ha... <mailto:ef...@ha...>>
>
>     Reinier Heeres wrote:
>     > Hi JJ,
>     >
>     > Thanks for the examples! I indeed suspected the anti-aliasing as
>     well.
>     > I'll include an option for linecolors = facecolors soon
>     (probably this
>     > weekend); it seems to be a good solution.
>
>     No, not in general.  There are two problems: it distorts the sizes and
>     positions of the patches, very slightly, but enough that it might be a
>     problem in some applications; and it does not work with alpha != 1.
>
>     Eric
>
>     ------------------------------------------------------------------------------
>     Let Crystal Reports handle the reporting - Free Crystal Reports
>     2008 30-Day
>     trial. Simplify your report design, integration and deployment -
>     and focus on
>     what you do best, core application coding. Discover what's new with
>     Crystal Reports now. http://p.sf.net/sfu/bobj-july
>     _______________________________________________
>     Matplotlib-users mailing list
>     Mat...@li...
>     <mailto:Mat...@li...>
>     https://lists.sourceforge.net/lists/listinfo/matplotlib-users
>
>
> ------------------------------------------------------------------------
>
> ------------------------------------------------------------------------------
> Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day
> trial. Simplify your report design, integration and deployment - and focus on
> what you do best, core application coding. Discover what's new with
> Crystal Reports now.  http://p.sf.net/sfu/bobj-july
> ------------------------------------------------------------------------
>
> _______________________________________________
> Matplotlib-users mailing list
> Mat...@li...
> https://lists.sourceforge.net/lists/listinfo/matplotlib-users
>