matplotlib-devel Mailing List for matplotlib

Sorry for the delayed response.
I had a discussion thread with Aaron Meurer last year about adding 
pkg-config support to anaconda so that matplotlib would build 
out-of-the-box, but I don't think that's gone anywhere. That would 
allow the extensive patches in the anaconda matplotlib recipe (and 
probably many other recipes for C and Unixy packages) to go away. (Note 
that the "pkgconfig" Python package in Anaconda is just the Python 
wrapper to the underlying Unix tool which is not present in Anaconda).
Mike
On 07/22/2015 07:52 PM, Nathan Goldbaum wrote:
> One way to do this is to build a Conda package using the matplotlib 
> recipe:
>
> https://github.com/conda/conda-recipes/tree/master/matplotlib
>
> Looking at the Conda recipe might give you some hints about how it 
> locates png.h as well, although I haven't checked in detail.
>
> On Wednesday, July 22, 2015, Brian Granger <ell...@gm... 
> <mailto:ell...@gm...>> wrote:
>
> No I am fine linking against the stuff that ships with conda - just
> not clear on how to get the setup.py logic to look in the right place.
>
> On Wed, Jul 22, 2015 at 11:20 AM, Phil Elson <pel...@gm...
> <javascript:;>> wrote:
> > Are you wanting to link against anything other than that
> installed with
> > conda?
> > The output of setup.py is normally pretty helpful at letting you
> know which
> > library it has found to build against.
> >
> > On 20 July 2015 at 01:54, Brian Granger <ell...@gm...
> <javascript:;>> wrote:
> >>
> >> Hi all,
> >>
> >> I am trying to get a dev build of matplotlib working with the
> anaconda
> >> python.
> >>
> >> Any advice on getting matplotlib to detect and use any of the
> >> libpng/freetypes:
> >>
> >> * Those installed with anaconda python.
> >> * Those from homebrew
> >> * Those that ship with OS X
> >>
> >> Cheers,
> >>
> >> Brian
> >>
> >> --
> >> Brian E. Granger
> >> Cal Poly State University, San Luis Obispo
> >> @ellisonbg on Twitter and GitHub
> >> bgr...@ca... <javascript:;> and ell...@gm...
> <javascript:;>
> >>
> >>
> >>
> ------------------------------------------------------------------------------
> >> Don't Limit Your Business. Reach for the Cloud.
> >> GigeNET's Cloud Solutions provide you with the tools and
> support that
> >> you need to offload your IT needs and focus on growing your
> business.
> >> Configured For All Businesses. Start Your Cloud Today.
> >> https://www.gigenetcloud.com/
> >> _______________________________________________
> >> Matplotlib-devel mailing list
> >> Mat...@li... <javascript:;>
> >> https://lists.sourceforge.net/lists/listinfo/matplotlib-devel
> >
> >
>
>
>
> --
> Brian E. Granger
> Cal Poly State University, San Luis Obispo
> @ellisonbg on Twitter and GitHub
> bgr...@ca... <javascript:;> and ell...@gm...
> <javascript:;>
>
> ------------------------------------------------------------------------------
> _______________________________________________
> Matplotlib-devel mailing list
> Mat...@li... <javascript:;>
> https://lists.sourceforge.net/lists/listinfo/matplotlib-devel
>
>
>
> ------------------------------------------------------------------------------
>
>
> _______________________________________________
> Matplotlib-devel mailing list
> Mat...@li...
> https://lists.sourceforge.net/lists/listinfo/matplotlib-devel

nabble is also another fairly commonly used resource for viewing archived
discussions.
On Fri, Jul 31, 2015 at 2:14 PM, Jouni K. Seppänen <jk...@ik...> wrote:
> Neal Becker <ndb...@gm...> writes:
>
> > I read via gmane: I guess this will need to be updated?
>
> I attempted to send a message to gmane.discuss to request this, but it
> seems there is some problem with that mailing list - the latest message
> is from July 17 when viewed via NNTP, and usually there are at several
> messages per week. I have emailed the gmane.org administrator to ask
> about the status.
>
> --
> Jouni K. Seppänen
> http://www.iki.fi/jks
>
>
>
> ------------------------------------------------------------------------------
> _______________________________________________
> Matplotlib-devel mailing list
> Mat...@li...
> https://lists.sourceforge.net/lists/listinfo/matplotlib-devel
>

Neal Becker <ndb...@gm...> writes:
> I read via gmane: I guess this will need to be updated?
I attempted to send a message to gmane.discuss to request this, but it
seems there is some problem with that mailing list - the latest message
is from July 17 when viewed via NNTP, and usually there are at several
messages per week. I have emailed the gmane.org administrator to ask
about the status.
-- 
Jouni K. Seppänen
http://www.iki.fi/jks

I read via gmane: I guess this will need to be updated?

Due to recent technical problems and changes in policy on SourceForge, 
we have decided to move the matplotlib mailing lists to python.org.
To subscribe to the new mailing lists, please visit:
 *
 For user questions and support:
 https://mail.python.org/mailman/listinfo/matplotlib-users
 mat...@py...
 *
 For low-volume announcements about matplotlib releases and related
 events and software:
 https://mail.python.org/mailman/listinfo/matplotlib-announce
 mat...@py...
 *
 For developer discussion:
 https://mail.python.org/mailman/listinfo/matplotlib-devel
 mat...@py...
The old list will remain active in the meantime, but all new posts will 
auto-reply with the location of the new mailing lists.
The old mailing list archives will remain available.
Thanks to Ralf Hildebrandt at python.org for making this possible.
Cheers,
Michael Droettboom
​

One of my co-workers brought this to my attention:
http://savvastjortjoglou.com/nba-shot-sharts.html
Tom

Thomas Caswell <tca...@gm...>
writes:
> The general approach follows R / seaborn / panadas and allows users to pass
> in a `data` kwarg which if present, any data fields which are strings are
> replaced by a call to `data[key]`. In code
>
> ax.plot(labeled_data['a'], labeled_data['b'])
>
> and
>
> ax.plot('a', 'b', data=labeled_data)
>
> are equivalent.
I commented on github briefly, but here's an expanded argument. I'm
proposing that instead of using strings (or only strings) as labels, we
allow arbitrary (hashable) objects to be looked up from the data dict.
I think using strings, or at least restricting to strings only is a
mistake for two reasons. One reason has been touched upon: in
 ax.scatter('a', 'b', c='b', data=data)
should c='b' be interpreted as a constant blue color or a sequence to be
looked up from data['b']?
Another is that since this functionality seems to be modeled after R's
plot functions, people will want to do more than just lookups. A simple
labeled plot in R is
 plot(speed ~ dist, data=cars)
but you can also do expressions, e.g.
 plot(speed^2 ~ dist, data=cars)
if you want to plot the square of speed against dist. This is pretty
neat for trying to find transformations for variables that depend on
each other non-linearly.
If we only allow strings as placeholders for plottable variables,
implementing expressions gets pretty clunky. We'd basically end up
defining a mini-language for parsing expressions from strings. But if
we allow objects for which you can implement methods like __add__,
it's much nicer. There's sample code below.
I'm proposing a small change to the patch. This still allows using
strings but also user-defined objects:
https://github.com/jkseppan/matplotlib/commit/b4709b38426ad5c2905f3ce253ce1bb68d314e7e
Here's a demo of implementing expressions on top of that patch:
https://github.com/jkseppan/matplotlib/blob/label-with-nonstrings/lib/matplotlib/tests/test_labeled.py
Here's how the test case looks, and the (albeit incomplete) expression
classes and evaluator to support this are about 50 lines of pretty simple
code.
 def test_expression_of_labels():
 fig, axes = plt.subplots(2, 2)
 x, y, z = Expr.vars('x y z')
 data = {'x': np.arange(10),
 'y': np.array([3, 1, 4, 1, 5, 9, 2, 6, 5, 3]),
 'z': np.array([2, 7, 1, 8, 2, 8, 1, 8, 2, 8])}
 ev = Evaluator(data)
 axes[0, 0].plot(x, y, data=ev)
 axes[0, 1].plot(x, 2 * y + 1, data=ev)
 axes[1, 0].plot(x, y ** 2, data=ev)
 axes[1, 1].plot(x, 2 * y ** z, data=ev)
The output:
https://raw.githubusercontent.com/jkseppan/matplotlib/label-with-nonstrings/lib/matplotlib/tests/baseline_images/test_labeled/expression_of_labels.png

A couple immediate thoughts: what if the data is spread across a mix of
objects? Also, I think "labeled" might be a better kwarg name. Less likely
to conflict with apis. I'll give this a careful look-see tomorrow.
Ben Root
On Jul 25, 2015 7:03 PM, "Thomas Caswell" <tca...@gm...> wrote:
> Hey all,
>
> Everyone should be aware of
> https://github.com/matplotlib/matplotlib/pull/4787 which is both a very
> simple, but very important change to the mpl API by providing a minimal API
> to pass labeled data (that is anything that `foo[key]` return an array-like
> object) into mpl plotting functions.
>
> This is due to Fernando and Brian's persuasive case to the importance of
> starting to address labeled data in mpl and it is now or in 6-9 months
>
> The general approach follows R / seaborn / panadas and allows users to
> pass in a `data` kwarg which if present, any data fields which are strings
> are replaced by a call to `data[key]`. In code
>
> ax.plot(labeled_data['a'], labeled_data['b'])
>
> and
>
> ax.plot('a', 'b', data=labeled_data)
>
> are equivalent.
>
> This is the minimal change to get quality of life for users who work with
> labeled data at the repl and to put a flag in the sand for the API that
> down stream projects should be targeting.
>
> Major changes to what the plotting functions do (inferring labels,
> inferring what computation to do etc) are out of scope for _this_ PR which
> I want to see included in 1.5. What a higher-level API which can make use
> of the additional meta-data available looks like is a much larger
> discussion which will must have input from all of the stake holders (ex
> IPython, pandas, bokeh, seaborn, xray).
>
> Tom
>
>
> ------------------------------------------------------------------------------
>
> _______________________________________________
> Matplotlib-devel mailing list
> Mat...@li...
> https://lists.sourceforge.net/lists/listinfo/matplotlib-devel
>
>

Hey all,
Everyone should be aware of
https://github.com/matplotlib/matplotlib/pull/4787 which is both a very
simple, but very important change to the mpl API by providing a minimal API
to pass labeled data (that is anything that `foo[key]` return an array-like
object) into mpl plotting functions.
This is due to Fernando and Brian's persuasive case to the importance of
starting to address labeled data in mpl and it is now or in 6-9 months
The general approach follows R / seaborn / panadas and allows users to pass
in a `data` kwarg which if present, any data fields which are strings are
replaced by a call to `data[key]`. In code
 ax.plot(labeled_data['a'], labeled_data['b'])
and
 ax.plot('a', 'b', data=labeled_data)
are equivalent.
This is the minimal change to get quality of life for users who work with
labeled data at the repl and to put a flag in the sand for the API that
down stream projects should be targeting.
Major changes to what the plotting functions do (inferring labels,
inferring what computation to do etc) are out of scope for _this_ PR which
I want to see included in 1.5. What a higher-level API which can make use
of the additional meta-data available looks like is a much larger
discussion which will must have input from all of the stake holders (ex
IPython, pandas, bokeh, seaborn, xray).
Tom

This should be fixed now, Thanks for reporting this!
Tom
On Fri, Jul 24, 2015 at 10:06 AM Thomas Caswell <tca...@gm...> wrote:
> Sorry, i must have broken that when I updated the docs to have the banner,
> i will look into fixing it asap.
>
> Tom
>
> On Fri, Jul 24, 2015, 7:04 AM Pierre Haessig <pie...@cr...>
> wrote:
>
>> Le 24/07/2015 15:10, Jens Nielsen a écrit :
>> > In the mean time you can use the development version of the docs up
>> > here: http://matplotlib.org/devdocs/users/colormaps.html that contains
>> > the plots.
>>
>> thanks !
>>
>> --
>> Pierre
>>
>>
>>
>> ------------------------------------------------------------------------------
>> _______________________________________________
>> Matplotlib-devel mailing list
>> Mat...@li...
>> https://lists.sourceforge.net/lists/listinfo/matplotlib-devel
>>
>

Sorry, i must have broken that when I updated the docs to have the banner,
i will look into fixing it asap.
Tom
On Fri, Jul 24, 2015, 7:04 AM Pierre Haessig <pie...@cr...>
wrote:
> Le 24/07/2015 15:10, Jens Nielsen a écrit :
> > In the mean time you can use the development version of the docs up
> > here: http://matplotlib.org/devdocs/users/colormaps.html that contains
> > the plots.
>
> thanks !
>
> --
> Pierre
>
>
>
> ------------------------------------------------------------------------------
> _______________________________________________
> Matplotlib-devel mailing list
> Mat...@li...
> https://lists.sourceforge.net/lists/listinfo/matplotlib-devel
>

Le 24/07/2015 15:10, Jens Nielsen a écrit :
> In the mean time you can use the development version of the docs up
> here: http://matplotlib.org/devdocs/users/colormaps.html that contains
> the plots.
thanks !
-- 
Pierre

Hi Pierre
I think it is likely that whoever rebuild the documentation did not have
the right dependencies installed causing the graphics to be missing. We
have significantly improved the build process of the documentation in
current master and this should be much less likely to happen when we
release 1.5 and going forward
In the mean time you can use the development version of the docs up here:
http://matplotlib.org/devdocs/users/colormaps.html that contains the plots.
best Jens
fre. 24. jul. 2015 kl. 12.51 skrev Pierre Haessig <pie...@cr...
>:
> Hi,
>
> I'm a bit confused with the colormap doc
> http://matplotlib.org/users/colormaps.html
> I often get to this page to select my colormaps and today I see no more
> images of the colormaps, only text !
>
> And when I look at the raw code
>
> https://raw.githubusercontent.com/matplotlib/matplotlib/master/doc/users/colormaps.rst
> I feel that some plot command are missing.
>
> Was there any error in the build process of the documentation ?
>
>
> best,
> Pierre
>
>
> ------------------------------------------------------------------------------
> _______________________________________________
> Matplotlib-devel mailing list
> Mat...@li...
> https://lists.sourceforge.net/lists/listinfo/matplotlib-devel
>

Hi,
I'm a bit confused with the colormap doc
http://matplotlib.org/users/colormaps.html
I often get to this page to select my colormaps and today I see no more
images of the colormaps, only text !
And when I look at the raw code
https://raw.githubusercontent.com/matplotlib/matplotlib/master/doc/users/colormaps.rst
I feel that some plot command are missing.
Was there any error in the build process of the documentation ?
best,
Pierre

One way to do this is to build a Conda package using the matplotlib recipe:
https://github.com/conda/conda-recipes/tree/master/matplotlib
Looking at the Conda recipe might give you some hints about how it locates
png.h as well, although I haven't checked in detail.
On Wednesday, July 22, 2015, Brian Granger <ell...@gm...> wrote:
> No I am fine linking against the stuff that ships with conda - just
> not clear on how to get the setup.py logic to look in the right place.
>
> On Wed, Jul 22, 2015 at 11:20 AM, Phil Elson <pel...@gm...
> <javascript:;>> wrote:
> > Are you wanting to link against anything other than that installed with
> > conda?
> > The output of setup.py is normally pretty helpful at letting you know
> which
> > library it has found to build against.
> >
> > On 20 July 2015 at 01:54, Brian Granger <ell...@gm...
> <javascript:;>> wrote:
> >>
> >> Hi all,
> >>
> >> I am trying to get a dev build of matplotlib working with the anaconda
> >> python.
> >>
> >> Any advice on getting matplotlib to detect and use any of the
> >> libpng/freetypes:
> >>
> >> * Those installed with anaconda python.
> >> * Those from homebrew
> >> * Those that ship with OS X
> >>
> >> Cheers,
> >>
> >> Brian
> >>
> >> --
> >> Brian E. Granger
> >> Cal Poly State University, San Luis Obispo
> >> @ellisonbg on Twitter and GitHub
> >> bgr...@ca... <javascript:;> and ell...@gm...
> <javascript:;>
> >>
> >>
> >>
> ------------------------------------------------------------------------------
> >> Don't Limit Your Business. Reach for the Cloud.
> >> GigeNET's Cloud Solutions provide you with the tools and support that
> >> you need to offload your IT needs and focus on growing your business.
> >> Configured For All Businesses. Start Your Cloud Today.
> >> https://www.gigenetcloud.com/
> >> _______________________________________________
> >> Matplotlib-devel mailing list
> >> Mat...@li... <javascript:;>
> >> https://lists.sourceforge.net/lists/listinfo/matplotlib-devel
> >
> >
>
>
>
> --
> Brian E. Granger
> Cal Poly State University, San Luis Obispo
> @ellisonbg on Twitter and GitHub
> bgr...@ca... <javascript:;> and ell...@gm... <javascript:;>
>
>
> ------------------------------------------------------------------------------
> _______________________________________________
> Matplotlib-devel mailing list
> Mat...@li... <javascript:;>
> https://lists.sourceforge.net/lists/listinfo/matplotlib-devel
>

No I am fine linking against the stuff that ships with conda - just
not clear on how to get the setup.py logic to look in the right place.
On Wed, Jul 22, 2015 at 11:20 AM, Phil Elson <pel...@gm...> wrote:
> Are you wanting to link against anything other than that installed with
> conda?
> The output of setup.py is normally pretty helpful at letting you know which
> library it has found to build against.
>
> On 20 July 2015 at 01:54, Brian Granger <ell...@gm...> wrote:
>>
>> Hi all,
>>
>> I am trying to get a dev build of matplotlib working with the anaconda
>> python.
>>
>> Any advice on getting matplotlib to detect and use any of the
>> libpng/freetypes:
>>
>> * Those installed with anaconda python.
>> * Those from homebrew
>> * Those that ship with OS X
>>
>> Cheers,
>>
>> Brian
>>
>> --
>> Brian E. Granger
>> Cal Poly State University, San Luis Obispo
>> @ellisonbg on Twitter and GitHub
>> bgr...@ca... and ell...@gm...
>>
>>
>> ------------------------------------------------------------------------------
>> Don't Limit Your Business. Reach for the Cloud.
>> GigeNET's Cloud Solutions provide you with the tools and support that
>> you need to offload your IT needs and focus on growing your business.
>> Configured For All Businesses. Start Your Cloud Today.
>> https://www.gigenetcloud.com/
>> _______________________________________________
>> Matplotlib-devel mailing list
>> Mat...@li...
>> https://lists.sourceforge.net/lists/listinfo/matplotlib-devel
>
>
-- 
Brian E. Granger
Cal Poly State University, San Luis Obispo
@ellisonbg on Twitter and GitHub
bgr...@ca... and ell...@gm...

Are you wanting to link against anything other than that installed with
conda?
The output of setup.py is normally pretty helpful at letting you know which
library it has found to build against.
On 20 July 2015 at 01:54, Brian Granger <ell...@gm...> wrote:
> Hi all,
>
> I am trying to get a dev build of matplotlib working with the anaconda
> python.
>
> Any advice on getting matplotlib to detect and use any of the
> libpng/freetypes:
>
> * Those installed with anaconda python.
> * Those from homebrew
> * Those that ship with OS X
>
> Cheers,
>
> Brian
>
> --
> Brian E. Granger
> Cal Poly State University, San Luis Obispo
> @ellisonbg on Twitter and GitHub
> bgr...@ca... and ell...@gm...
>
>
> ------------------------------------------------------------------------------
> Don't Limit Your Business. Reach for the Cloud.
> GigeNET's Cloud Solutions provide you with the tools and support that
> you need to offload your IT needs and focus on growing your business.
> Configured For All Businesses. Start Your Cloud Today.
> https://www.gigenetcloud.com/
> _______________________________________________
> Matplotlib-devel mailing list
> Mat...@li...
> https://lists.sourceforge.net/lists/listinfo/matplotlib-devel
>

Hi all,
I am trying to get a dev build of matplotlib working with the anaconda python.
Any advice on getting matplotlib to detect and use any of the libpng/freetypes:
* Those installed with anaconda python.
* Those from homebrew
* Those that ship with OS X
Cheers,
Brian
-- 
Brian E. Granger
Cal Poly State University, San Luis Obispo
@ellisonbg on Twitter and GitHub
bgr...@ca... and ell...@gm...

Hi folks,
Is it possible to run the matplotlib test suite without the image comparisons?
Cheers,
Brian
-- 
Brian E. Granger
Cal Poly State University, San Luis Obispo
@ellisonbg on Twitter and GitHub
bgr...@ca... and ell...@gm...

Pretty sure this doesn't impact us, but just thought I'd throw this out
there:
https://web.nvd.nist.gov/view/vuln/detail?vulnId=CVE-2015-2426
Looks like Microsoft's AFM library didn't properly sanitize malicious
OpenType font files and suffered from a buffer underflow, allowing for
arbitrary code execution.
Ben Root

There is a discussion over names happening in a scikit-image PR thread (
https://github.com/scikit-image/scikit-image/pull/1599)
There is a proposal for naming options A-C as {'ignis', 'ortus', 'fyrian'}
in some order.
Commenting here and copying the devel list so it gets some visibility.
On Wed, Jul 15, 2015 at 11:56 AM Nathaniel J. Smith <
not...@gi...> wrote:
> Yep, that's the plan.
>
> On Wed, Jul 15, 2015 at 3:22 AM, Werner Beroux <not...@gi...>
> wrote:
>
> > Would be nice to add also some of the alternatives on
> > http://bids.github.io/colormap/ to matplotlib.
> >
> > —
> > Reply to this email directly or view it on GitHub
> > <
> https://github.com/matplotlib/matplotlib/issues/875#issuecomment-121526766
> >
> > .
> >
>
>
>
> --
> Nathaniel J. Smith -- http://vorpus.org
>
> —
> Reply to this email directly or view it on GitHub
> <https://github.com/matplotlib/matplotlib/issues/875#issuecomment-121660178>
> .
>

On Tue, Jul 14, 2015 at 02:53:13PM -0500, Albert Chin wrote:
> On Tue, Jul 14, 2015 at 09:22:07AM -1000, Eric Firing wrote:
> > On 2015年07月14日 3:59 AM, Albert Chin wrote:
> > > I've built matplotlib-1.4.3 with Python 2.7.5, 3.3.2, 3.4.3. Python
> > > 3.3.2 and 3.4.3 work ok. With Python 2.7.5, I get:
> > > File "/tmp/lines3d_demo.py", line 2, in <module>
> > > from mpl_toolkits.mplot3d import Axes3D
> > > ImportError: No module named mpl_toolkits.mplot3d
> > >
> > > [[ snip snip ]]
> > 
> > I suspect this results from some aspect of your system; we routinely 
> > build and test with python 2.7.x. Maybe your python 2.7 setuptools 
> > needs to be updated?
> 
> Most likely our fault. I rebuilt with setuptools-18.0.1 and that
> didn't help. Looking at a Fedora 22 system,
> matplotlib-1.4.3-py2.7-nspkg.pth is read, but not on our system. This
> is the problem. If I move this file out of the way on the Fedora 22
> system, the failure is the same as above. Just need to figure out how
> Python reads this file.
Ok, found the problem. Python only reads .pth files in the default
Python prefix (sys.prefix and sys.exec_prefix), not from any other
path in PYTHONPATH. I'll need to create sitecustomize.py to fix this.
-- 
albert chin (ch...@th...)

On Tue, Jul 14, 2015 at 09:22:07AM -1000, Eric Firing wrote:
> On 2015年07月14日 3:59 AM, Albert Chin wrote:
> > I've built matplotlib-1.4.3 with Python 2.7.5, 3.3.2, 3.4.3. Python
> > 3.3.2 and 3.4.3 work ok. With Python 2.7.5, I get:
> > File "/tmp/lines3d_demo.py", line 2, in <module>
> > from mpl_toolkits.mplot3d import Axes3D
> > ImportError: No module named mpl_toolkits.mplot3d
> >
> > [[ snip snip ]]
> 
> I suspect this results from some aspect of your system; we routinely 
> build and test with python 2.7.x. Maybe your python 2.7 setuptools 
> needs to be updated?
Most likely our fault. I rebuilt with setuptools-18.0.1 and that
didn't help. Looking at a Fedora 22 system,
matplotlib-1.4.3-py2.7-nspkg.pth is read, but not on our system. This
is the problem. If I move this file out of the way on the Fedora 22
system, the failure is the same as above. Just need to figure out how
Python reads this file.
-- 
albert chin (ch...@th...)

On 2015年07月14日 3:59 AM, Albert Chin wrote:
> I've built matplotlib-1.4.3 with Python 2.7.5, 3.3.2, 3.4.3. Python
> 3.3.2 and 3.4.3 work ok. With Python 2.7.5, I get:
> File "/tmp/lines3d_demo.py", line 2, in <module>
> from mpl_toolkits.mplot3d import Axes3D
> ImportError: No module named mpl_toolkits.mplot3d
>
> There is no __init__.py in the mpl_toolkits directory unlike with
> matplotlib-1.3.1:
> $ ls -l /opt/TWWfsw/matplotlib14/lib/python27/mpl_toolkits
> total 64
> drwxr-xr-x+ 2 root root 53 Jul 14 13:51 axes_grid
> drwxr-xr-x+ 2 root root 32 Jul 14 13:51 axes_grid1
> drwxr-xr-x+ 2 root root 29 Jul 14 13:51 axisartist
> -rw-r--r--+ 1 root root 3830 Dec 23 2014 exceltools.py
> -rw-r--r--+ 1 root root 3817 Jul 14 13:51 exceltools.pyc
> -rw-r--r--+ 1 root root 3817 Jul 14 13:51 exceltools.pyo
> -rw-r--r--+ 1 root root 19346 Dec 23 2014 gtktools.py
> -rw-r--r--+ 1 root root 19745 Jul 14 13:51 gtktools.pyc
> -rw-r--r--+ 1 root root 19706 Jul 14 13:51 gtktools.pyo
> drwxr-xr-x+ 2 root root 17 Jul 14 13:51 mplot3d
> drwxr-xr-x+ 3 root root 12 Jul 14 13:51 tests
>
> If I copy the matplotlib-1.3.1 __init__.py to
> /opt/TWWfsw/matplotlib14/lib/python27/mpl_toolkits, then everything
> works ok.
>
I suspect this results from some aspect of your system; we routinely 
build and test with python 2.7.x. Maybe your python 2.7 setuptools 
needs to be updated?
Eric

I've built matplotlib-1.4.3 with Python 2.7.5, 3.3.2, 3.4.3. Python
3.3.2 and 3.4.3 work ok. With Python 2.7.5, I get:
 File "/tmp/lines3d_demo.py", line 2, in <module>
 from mpl_toolkits.mplot3d import Axes3D
 ImportError: No module named mpl_toolkits.mplot3d
There is no __init__.py in the mpl_toolkits directory unlike with
matplotlib-1.3.1:
 $ ls -l /opt/TWWfsw/matplotlib14/lib/python27/mpl_toolkits 
 total 64
 drwxr-xr-x+ 2 root root 53 Jul 14 13:51 axes_grid
 drwxr-xr-x+ 2 root root 32 Jul 14 13:51 axes_grid1
 drwxr-xr-x+ 2 root root 29 Jul 14 13:51 axisartist
 -rw-r--r--+ 1 root root 3830 Dec 23 2014 exceltools.py
 -rw-r--r--+ 1 root root 3817 Jul 14 13:51 exceltools.pyc
 -rw-r--r--+ 1 root root 3817 Jul 14 13:51 exceltools.pyo
 -rw-r--r--+ 1 root root 19346 Dec 23 2014 gtktools.py
 -rw-r--r--+ 1 root root 19745 Jul 14 13:51 gtktools.pyc
 -rw-r--r--+ 1 root root 19706 Jul 14 13:51 gtktools.pyo
 drwxr-xr-x+ 2 root root 17 Jul 14 13:51 mplot3d
 drwxr-xr-x+ 3 root root 12 Jul 14 13:51 tests
If I copy the matplotlib-1.3.1 __init__.py to
/opt/TWWfsw/matplotlib14/lib/python27/mpl_toolkits, then everything
works ok.
-- 
albert chin (ch...@th...)