matplotlib-users Mailing List for matplotlib

I didn't get a response to the post below from 20 Dec 2007, and also
didn't get a response to posts I made earlier in 2007, so I thought
I'd check whether posts made via gmane make it to the list.

On a probably unrelated note, I checked the list archives at

  http://sourceforge.net/mailarchive/forum.php?forum_name=matplotlib-users

and don't see any messages after 17 Dec.

Dan

Dan Christensen <jd...@uw...> writes:

> I regularly make log-log errorbar plots where the errors are larger than
> the y values.  This means that the error bar goes a finite distance
> upwards, but should go infinitely far downwards.  In Debian's 0.87.7-0.3
> package, just the top portion is drawn, which seems like a reasonable
> solution.  But in Ubuntu's 0.90.1-2ubuntu1 package, trying to do a
> plot like this causes an error.  Any ideas?
>
> Dan
>
> $ python
> Python 2.5.1 (r251:54863, Oct  5 2007, 13:36:32) 
> [GCC 4.1.3 20070929 (prerelease) (Ubuntu 4.1.2-16ubuntu2)] on linux2
> Type "help", "copyright", "credits" or "license" for more information.
>>>> from pylab import *
>>>> errorbar([1],[1],[2])
> (<matplotlib.lines.Line2D instance at 0xb5ac024c>, [<matplotlib.lines.Line2D instance at 0xb5ac00cc>, <matplotlib.lines.Line2D instance at 0xb5ac00ec>], [<matplotlib.collections.LineCollection instance at 0xb5abdfac>])
>>>> gca().set_yscale('log')
>>>> show()
> Exception in Tkinter callback
> Traceback (most recent call last):
>   File "lib-tk/Tkinter.py", line 1406, in __call__
>     return self.func(*args)
>   File "/usr/lib/python2.5/site-packages/matplotlib/backends/backend_tkagg.py", line 151, in resize
>     self.show()
>   File "/usr/lib/python2.5/site-packages/matplotlib/backends/backend_tkagg.py", line 154, in draw
>     FigureCanvasAgg.draw(self)
>   File "/usr/lib/python2.5/site-packages/matplotlib/backends/backend_agg.py", line 392, in draw
>     self.figure.draw(renderer)
>   File "/usr/lib/python2.5/site-packages/matplotlib/figure.py", line 601, in draw
>     for a in self.axes: a.draw(renderer)
>   File "/usr/lib/python2.5/site-packages/matplotlib/axes.py", line 1286, in draw
>     a.draw(renderer)
>   File "/usr/lib/python2.5/site-packages/matplotlib/collections.py", line 700, in draw
>     transoffset)
> ValueError: Domain error on transOffset->operator in draw_line_collection
>
>
> -------------------------------------------------------------------------
> This SF.net email is sponsored by: Microsoft
> Defy all challenges. Microsoft(R) Visual Studio 2005.
> http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/

In case anyone else runs into this issue, let me offer a few suggestions.

First, and most importantly, make sure that you are using python from the
directory /System/Library/Python.frameworks and not a third-party python from
/Library/Python.frameworks.  This was the key issue that for me fixed the
problem described in previous post with ppc architecture. 

Second, you may have problems with which gcc you are using especially if you
have inadvertently installed another gcc when gfortran was installed.  It is
best to install gfortran from 

  http://r.research.att.com/tools/

Hope that helps.

Travis

On Mon, 7 Jan 2008, John Hunter wrote:

> The function is document
> http://matplotlib.sourceforge.net/matplotlib.pyplot.html#-hold and the
> usage in the "Simple Plots" Section 3.1 of the User's Guide at
> http://matplotlib.sourceforge.net/users_guide_0.91.2svn.pdf .  The hold
> functionality is part of the state-machine interface inherited from
> matlab, where plotting commands are targets to the current axes in the
> current figure, and overbplotting is controlled by the "hold" state.  Se
> also, "ishold"

   Thank you John and Darren.

Rich

-- 
Richard B. Shepard, Ph.D.               |  Integrity            Credibility
Applied Ecosystem Services, Inc.        |            Innovation
<http://www.appl-ecosys.com>     Voice: 503-667-4517      Fax: 503-667-8863

John Hunter wrote:
> On Jan 7, 2008 2:15 PM, Jeff Whitaker <js...@fa...> wrote:
>
>   
>> I think namespace packages were being used before, but were removed.  I
>> don't remember why.
>>     
>
> We removed the namespace packages support because we were using it
> improperly.  To do it correctly would require moving all of the
> functionality out of matplotlib/__init__.py.  Since there is currently
> a fair amount of functionality there, adding namespace support would
> require moving it and breaking the current API. In addition, there are
> some ongoing issues with performance and namespace packages.   Because
> of this, we had some ambivalence about namespace packages and decided
> to remove the broken support until we could add proper support, if we
> decide to go that route.  A good time to do that, if indeed we want
> to, would be when we release Michael's transforms branch, so we can
> break as much as possible at once.
>
> JDH
>   

John:  OK, then I can think of two options for basemap in the interim:

1) remove setuptools support, since basemap cannot be installed as an 
egg without namespace packages.  This means that python2.3 support would 
have to be dropped. 

2) move the toolkits to basemap_toolkits, as Philip Eby suggested, and 
make matplotlib_toolkits a namespace package.  Unfortunately, this may 
break lots of existing code.

-Jeff

-- 
Jeffrey S. Whitaker         Phone  : (303)497-6313
Meteorologist               FAX    : (303)497-6449
NOAA/OAR/PSD  R/PSD1        Email  : Jef...@no...
325 Broadway                Office : Skaggs Research Cntr 1D-124
Boulder, CO, USA 80303-3328 Web    : http://tinyurl.com/5telg

On Jan 7, 2008 2:15 PM, Jeff Whitaker <js...@fa...> wrote:

> I think namespace packages were being used before, but were removed.  I
> don't remember why.

We removed the namespace packages support because we were using it
improperly.  To do it correctly would require moving all of the
functionality out of matplotlib/__init__.py.  Since there is currently
a fair amount of functionality there, adding namespace support would
require moving it and breaking the current API. In addition, there are
some ongoing issues with performance and namespace packages.   Because
of this, we had some ambivalence about namespace packages and decided
to remove the broken support until we could add proper support, if we
decide to go that route.  A good time to do that, if indeed we want
to, would be when we release Michael's transforms branch, so we can
break as much as possible at once.

JDH

Fought, Richard wrote:
>>> Hi Jeff,
>>>
>>> Thanks for your response.  I might be able to upgrade my prototype 
>>> machine to Python 2.4, but I'm not sure about my customer's box.
>>>
>>> Here is the listing from
>>>
>>>       
>> python2.3/site-packages/matplotlib-0.91.1_r0-py2.3-linux-i686.egg/matp
>>     
>>> lo
>>> tlib/toolkits:
>>>
>>> __init__.py
>>> __init__.pyc
>>>
>>> Regards,
>>> Rich Fought
>>>   
>>>       
>> Rich:  That's it?  I expected to see a basemap directory 
>> installed there.  Can you send me a listing of the basemap 
>> directory in site-packages?
>>
>> -Jeff
>>
>>     
>
> Directory listing of
> python2.3/site-packages/basemap-0.9.9-py2.3-linux-i686.egg :
>
> drwxr-xr-x  4 root root  4096 Jan  4 17:33 dap
> drwxr-xr-x  2 root root  4096 Jan  4 17:33 dbflib
> -rwxr-xr-x  1 root root   258 Jan  4 17:33 dbflibc.py
> -rwxr-xr-x  1 root root   614 Jan  4 17:33 dbflibc.pyc
> -rwxr-xr-x  1 root root 74526 Jan  4 17:33 dbflibc.so
> drwxr-xr-x  2 root root  4096 Jan  4 17:33 EGG-INFO
> -rwxr-xr-x  1 root root   256 Jan  4 17:33 _geos.py
> -rwxr-xr-x  1 root root   608 Jan  4 17:33 _geos.pyc
> -rwxr-xr-x  1 root root 88557 Jan  4 17:33 _geos.so
> drwxr-xr-x  2 root root  4096 Jan  4 17:33 httplib2
> drwxr-xr-x  3 root root  4096 Jan  4 17:33 matplotlib
> drwxr-xr-x  2 root root  4096 Jan  4 17:33 shapelib
> -rwxr-xr-x  1 root root   260 Jan  4 17:33 shapelibc.py
> -rwxr-xr-x  1 root root   620 Jan  4 17:33 shapelibc.pyc
> -rwxr-xr-x  1 root root 92671 Jan  4 17:33 shapelibc.so
> -rwxr-xr-x  1 root root   258 Jan  4 17:33 shptree.py
> -rwxr-xr-x  1 root root   614 Jan  4 17:33 shptree.pyc
> -rwxr-xr-x  1 root root 19958 Jan  4 17:33 shptree.so
>
> As you can see, there is a matplotlib folder here with the following
> structure:
>
> matplotlib
>  |_toolkits
>     |_basemap
>
> I created a symbolic link to this basemap directory in the
> matplotlib/toolkits directory and simpletest.py works now, though I
> still get the warning message
>
> /usr/lib/python2.3/site-packages/matplotlib-0.91.1_r0-py2.3-linux-i686.e
> gg/pytz/ __init__.py:29: UserWarning: Module matplotlib was already
> imported from /usr/li
> b/python2.3/site-packages/matplotlib-0.91.1_r0-py2.3-linux-i686.egg/matp
> lotlib/_ _init__.pyc, but
> /usr/lib/python2.3/site-packages/basemap-0.9.9-py2.3-linux-i686 .egg is
> being added to sys.path
>   from pkg_resources import resource_stream
>
> I guess the basemap installer just installed the toolkit in the wrong
> place.
>
> Regards,
> Rich
>   

Rich:  I think this is happening because matplotlib (and the basemap 
toolkit) needs to use namespace packages 
(http://peak.telecommunity.com/DevCenter/setuptools#namespace-packages) 
in order for the basemap stuff to be picked up from the 
matplotlib.toolkits namespace when basemap is installed as an egg.  If 
it's installed using distutils (instead of setuptools), the basemap 
stuff just gets written in the matplotlib toolkits directory directly, 
and everything works fine.

I think namespace packages were being used before, but were removed.  I 
don't remember why.

Until we get this fixed, I think your symlink trick is the only way to 
get it to work with python2.3.  The warning message appears to be harmless.

-Jeff


-- 
Jeffrey S. Whitaker         Phone  : (303)497-6313
Meteorologist               FAX    : (303)497-6449
NOAA/OAR/PSD  R/PSD1        Email  : Jef...@no...
325 Broadway                Office : Skaggs Research Cntr 1D-124
Boulder, CO, USA 80303-3328 Web    : http://tinyurl.com/5telg

Fought, Richard wrote:
>  
>
>   
>> -----Original Message-----
>> From: Jeff Whitaker [mailto:js...@fa...] 
>> Sent: Monday, January 07, 2008 1:26 PM
>> To: Fought, Richard
>> Cc: mat...@li...
>> Subject: Re: [Matplotlib-users] Basemap installation question
>>
>> Fought, Richard wrote:
>>     
>>> Hi all,
>>>
>>> I'm trying to set up matplotlib with basemap on CentOS 4.6 
>>>       
>> with python
>>     
>>> 2.3.4
>>>
>>> I installed setuptools 0.6c7, then numpy 1.0.4, then matplotlib 
>>> 0.91.1_r0, then basemap 0.9.9 (building and installing the GEOS 
>>> library from source).  When I try to run the example simpletest.py 
>>> script, I get the following message:
>>>
>>>
>>>       
>> /usr/lib/python2.3/site-packages/matplotlib-0.91.1_r0-py2.3-linux-i686
>>     
>>> .e
>>> gg/pytz/__init__.py:29: UserWarning: Module matplotlib was already 
>>> imported from 
>>>
>>>       
>> /usr/lib/python2.3/site-packages/matplotlib-0.91.1_r0-py2.3-linux-i686
>>     
>>> .e
>>> gg/matplotlib/__init__.pyc, but
>>>
>>>       
>> /usr/lib/python2.3/site-packages/basemap-0.9.9-py2.3-linux-i68
>> 6.egg is 
>>     
>>> being added to sys.path
>>>   from pkg_resources import resource_stream Traceback (most recent 
>>> call last):
>>>   File "simpletest.py", line 1, in ?
>>>     from matplotlib.toolkits.basemap import Basemap
>>> ImportError: No module named basemap
>>>
>>> Any ideas what might have gone wrong?  I was expecting basemap to 
>>> install under /site-packages/matplotlib/toolkits but it installed 
>>> under a site-package of it's own.
>>>
>>> Thanks,
>>> Rich
>>>   
>>>       
>> Rich:  Unfortunately, I've no idea what's wrong and don't 
>> have 2.3 to test on.  The install process for 2.3 is a bit 
>> different in that it uses setuptools to substitute for 
>> features in distutils that didn't appear until python 2.4.  
>> Can you easily upgrade to 2.4? If not, can you send a listing 
>> of  site-packages/matplotlib/toolkits to me, off-list?
>>
>> -Jeff
>>     
>
> Hi Jeff,
>
> Thanks for your response.  I might be able to upgrade my prototype
> machine to Python 2.4, but I'm not sure about my customer's box.
>
> Here is the listing from
> python2.3/site-packages/matplotlib-0.91.1_r0-py2.3-linux-i686.egg/matplo
> tlib/toolkits:
>
> __init__.py
> __init__.pyc
>
> Regards,
> Rich Fought
>   
Richard:  I've confirmed your problem - in fact, it occurs whenever 
setuptools is used (and the package is installed as an egg),  This 
happens by default with python2.3, but only if you use setupegg.py for 
python2.4 and python2.5.  I haven't figured out why yet - perhaps 
someone on the list who knows more about eggs will chime in.

-Jeff

-- 
Jeffrey S. Whitaker         Phone  : (303)497-6313
Meteorologist               FAX    : (303)497-6449
NOAA/OAR/PSD  R/PSD1        Email  : Jef...@no...
325 Broadway                Office : Skaggs Research Cntr 1D-124
Boulder, CO, USA 80303-3328 Web    : http://tinyurl.com/5telg

On Jan 7, 2008 1:43 PM, Rich Shepard <rsh...@ap...> wrote:
> On Mon, 7 Jan 2008, Darren Dale wrote:
>
> > you can call hold(True) so each call to plot() adds a new curve to the axes.
>
> Darren,
>
>    Excellent! Where is this documented, please? I did not see it when I
> looked in the docs.

The function is document
http://matplotlib.sourceforge.net/matplotlib.pyplot.html#-hold and the
usage in the "Simple Plots" Section 3.1 of the User's Guide at
http://matplotlib.sourceforge.net/users_guide_0.91.2svn.pdf .  The
hold functionality is part of the state-machine interface inherited
from matlab, where plotting commands are targets to the current axes
in the current figure, and overbplotting is controlled by the "hold"
state.  Se also, "ishold"

JDH

On Monday 07 January 2008 04:43:26 pm Rich Shepard wrote:
> On Mon, 7 Jan 2008, Darren Dale wrote:
> > you can call hold(True) so each call to plot() adds a new curve to the
> > axes.
>
> Darren,
>
>    Excellent! Where is this documented, please? I did not see it when I
> looked in the docs.

See section 3.1 in the users guide: 
http://matplotlib.sourceforge.net/users_guide_0.91.2svn.pdf

On Mon, 7 Jan 2008, Darren Dale wrote:

> you can call hold(True) so each call to plot() adds a new curve to the axes.

Darren,

   Excellent! Where is this documented, please? I did not see it when I
looked in the docs.

> Agg does not produce jpg. Can you live with a png? png are not lossy and
> so they yield much nicer line art than jpgs. Just give your filename a
> .png extension and save will recognize the format.

   Yes, a .png image can be incorporated into a .pdf file as easily as can a
.jpg impage.

Thank you very much,

Rich

-- 
Richard B. Shepard, Ph.D.               |  Integrity            Credibility
Applied Ecosystem Services, Inc.        |            Innovation
<http://www.appl-ecosys.com>     Voice: 503-667-4517      Fax: 503-667-8863

Eric Firing wrote:
> thread is a standard python module, part of the basic python
distribution.  > I don't know why it is not being found.  If you start
python on a command > line, can you import thread?

No, I can't.  Apparently whoever installed this version of python (it's on
a cluster I've got time on) didn't compile the thread module.  If you
don't use the thread module, there's a drop-in replacement called
dummy_thread instead--I went into the install files and changed every
instance of

import thread

to

try:
    import thread
except:
    import dummy_thread

Then installed and it worked fine.  At least, everything I've tried to do
with it so far works fine.

Jordan

On Monday 07 January 2008 03:05:30 pm Rich Shepard wrote:
>    I've looked at all the docs I can find on the matplotlib web site
> without finding the answers to two questions. Pointers to references are
> greatly appreciated.
>
>    1)  I want to plot a series of curves on the same set of axes. For
> example, shoulder- and trapezoidal curves:
>
>  	 _______          __________          _______
>
>  	|       \        /          \        /
>  	|        \      /            \      /
>  	|         \    /              \    /
>  	|          \  /                \  /
>  	|           \/                  \/
>  	|           /\                  /\
>  	|          /  \                /  \
>  	|         /    \              /    \
>  	|        /      \            /      \
>  	|_______/________\__________/________\_______
>
> With labels and text, of course. Each curve represents values in a separate
> row of the database table.
>
>    How do I specify that subsequent curves are to be plotted on the same
> axes as the first one?

you can call hold(True) so each call to plot() adds a new curve to the axes.

>
>    2)  When each plot is created I want to save it as a .jpg file so it can
> be included in a ReportLab report. Do I specify the filename and extension
> in the save() command, or is there a different way?

Agg does not produce jpg. Can you live with a png? png are not lossy and so 
they yield much nicer line art than jpgs. Just give your filename a .png 
extension and save will recognize the format.

Darren

   I've looked at all the docs I can find on the matplotlib web site without
finding the answers to two questions. Pointers to references are greatly
appreciated.

   1)  I want to plot a series of curves on the same set of axes. For
example, shoulder- and trapezoidal curves:

 	 _______          __________          _______
 	|       \        /          \        /
 	|        \      /            \      /
 	|         \    /              \    /
 	|          \  /                \  /
 	|           \/                  \/
 	|           /\                  /\
 	|          /  \                /  \
 	|         /    \              /    \
 	|        /      \            /      \
 	|_______/________\__________/________\_______

With labels and text, of course. Each curve represents values in a separate
row of the database table.

   How do I specify that subsequent curves are to be plotted on the same axes
as the first one?

   2)  When each plot is created I want to save it as a .jpg file so it can
be included in a ReportLab report. Do I specify the filename and extension
in the save() command, or is there a different way?

Rich

-- 
Richard B. Shepard, Ph.D.               |  Integrity            Credibility
Applied Ecosystem Services, Inc.        |            Innovation
<http://www.appl-ecosys.com>     Voice: 503-667-4517      Fax: 503-667-8863

Fought, Richard wrote:
> Hi all,
>
> I'm trying to set up matplotlib with basemap on CentOS 4.6 with python
> 2.3.4
>
> I installed setuptools 0.6c7, then numpy 1.0.4, then matplotlib
> 0.91.1_r0, then basemap 0.9.9 (building and installing the GEOS library
> from source).  When I try to run the example simpletest.py script, I get
> the following message:
>
> /usr/lib/python2.3/site-packages/matplotlib-0.91.1_r0-py2.3-linux-i686.e
> gg/pytz/__init__.py:29: UserWarning: Module matplotlib was already
> imported from
> /usr/lib/python2.3/site-packages/matplotlib-0.91.1_r0-py2.3-linux-i686.e
> gg/matplotlib/__init__.pyc, but
> /usr/lib/python2.3/site-packages/basemap-0.9.9-py2.3-linux-i686.egg is
> being added to sys.path
>   from pkg_resources import resource_stream
> Traceback (most recent call last):
>   File "simpletest.py", line 1, in ?
>     from matplotlib.toolkits.basemap import Basemap
> ImportError: No module named basemap
>
> Any ideas what might have gone wrong?  I was expecting basemap to
> install under /site-packages/matplotlib/toolkits but it installed under
> a site-package of it's own.
>
> Thanks,
> Rich
>   

Rich:  Unfortunately, I've no idea what's wrong and don't have 2.3 to 
test on.  The install process for 2.3 is a bit different in that it uses 
setuptools to substitute for features in distutils that didn't appear 
until python 2.4.  Can you easily upgrade to 2.4? If not, can you send a 
listing of  site-packages/matplotlib/toolkits to me, off-list?

-Jeff

-- 
Jeffrey S. Whitaker         Phone  : (303)497-6313
Meteorologist               FAX    : (303)497-6449
NOAA/OAR/PSD  R/PSD1        Email  : Jef...@no...
325 Broadway                Office : Skaggs Research Cntr 1D-124
Boulder, CO, USA 80303-3328 Web    : http://tinyurl.com/5telg

We have uploaded source and binary releases of matplotlib-0.91.2 to
http://sourceforge.net/project/showfiles.php?group_id=80706&package_id=82474&release_id=566411.
Thanks to Charlie Moad for doing the release.

This is a bugfix release and includes several important fixes listed
below.

2008-01-06 Released 0.91.2 at revision 4802

2007-12-26 Reduce too-late use of matplotlib.use() to a warning
           instead of an exception, for backwards compatibility - EF

2007-12-25 Fix bug in errorbar, identified by Noriko Minakawa - EF

2007-12-25 Changed masked array importing to work with the upcoming
           numpy 1.05 (now the maskedarray branch) as well as with
           earlier versions. - EF

2007-12-16 rec2csv saves doubles without losing precision. Also, it
           does not close filehandles passed in open. - JDH,ADS

2007-12-13 Moved rec2gtk to matplotlib.toolkits.gtktools and rec2excel
           to matplotlib.toolkits.exceltools - JDH

2007-12-12 Support alpha-blended text in the Agg and Svg backends -
           MGD

2007-12-10 Fix SVG text rendering bug. - MGD

2007-12-10 Increase accuracy of circle and ellipse drawing by using an
           8-piece bezier approximation, rather than a 4-piece one.
           Fix PDF, SVG and Cairo backends so they can draw paths
           (meaning ellipses as well). - MGD

2007-12-07 Issue a warning when drawing an image on a non-linear axis. - MGD

2007-12-06 let widgets.Cursor initialize to the lower x and y bounds
           rather than 0,0, which can cause havoc for dates and other
           transforms - DSD

2007-12-06 updated references to mpl data directories for py2exe - DSD

2007-12-06 fixed a bug in rcsetup, see bug 1845057 - DSD

2007-12-05 Fix how fonts are cached to avoid loading the same one
multiple times.
           (This was a regression since 0.90 caused by the refactoring of
           font_manager.py) - MGD

2007-12-05 Support arbitrary rotation of usetex text in Agg backend. - MGD

2007-12-04 Support '|' as a character in mathtext - MGD

Hi all,

I'm trying to set up matplotlib with basemap on CentOS 4.6 with python
2.3.4

I installed setuptools 0.6c7, then numpy 1.0.4, then matplotlib
0.91.1_r0, then basemap 0.9.9 (building and installing the GEOS library
from source).  When I try to run the example simpletest.py script, I get
the following message:

/usr/lib/python2.3/site-packages/matplotlib-0.91.1_r0-py2.3-linux-i686.e
gg/pytz/__init__.py:29: UserWarning: Module matplotlib was already
imported from
/usr/lib/python2.3/site-packages/matplotlib-0.91.1_r0-py2.3-linux-i686.e
gg/matplotlib/__init__.pyc, but
/usr/lib/python2.3/site-packages/basemap-0.9.9-py2.3-linux-i686.egg is
being added to sys.path
  from pkg_resources import resource_stream
Traceback (most recent call last):
  File "simpletest.py", line 1, in ?
    from matplotlib.toolkits.basemap import Basemap
ImportError: No module named basemap

Any ideas what might have gone wrong?  I was expecting basemap to
install under /site-packages/matplotlib/toolkits but it installed under
a site-package of it's own.

Thanks,
Rich