matplotlib-devel Mailing List for matplotlib

I just finished a major rewrite of backend agg using the pycxx
extension tools. Last week I presented weave to a CS class that I'm
taking, and encountered cxx (which weave uses). I had seen this long
ago, but hadn't looked closely at it since.
Given the memory leak in agg that was discovered on matplotlib-users
last week, I decided it would be easier and cleaner to use some good
extension building code rather than hunt it down. The result is a
much cleaner and readable backend_agg. There may be some additional
small leaks to clean up (eg in image or ft2font) but if so they are
likely not major. I was able to generate 2000 figures in a test run
with only a small increase in memory (4% -> 8% on my system). As I
get time I hope to do the image module and/or backend_gtkagg and/or
ft2font.
In any case CXX is very nice - it's committed to CVS - you may want to
test build on your platforms of choice.
Cheers,
John Hunter

>>>>> "gary" == gary ruben <gr...@bi...> writes:
 gary> Hi John and Paul, I changed the 'pass' to a 'continue' and a
 gary> different problem has revealed itself. Now, in lieu of any
 gary> plot window, I get a Windows dialog with title 'Windows
 gary> Visual C++ Runtime Library'
 gary> and content
 gary> Runtime Error Program C:\APPS\PYTHON23\PYTHONW.EXE abnormal
 gary> program termination
 gary> Clicking the OK then generates a page fault. This seems a
 gary> bit nasty. When I have time later today, I'll see what
 gary> happens under Win2000 on my laptop. I should probably try
 gary> another backend too.
Hi Gary,
I'm just including your last email to the dev list so others will know
it's taken care of
From: "gary ruben" <gr...@bi...>
Subject: more matplotlib error info
To: jdh...@ac...
Cc: ba...@st...
Date: 2004年4月22日 19:52:55 +1000
Hi John and Paul,
Paul; the font manager problem seems to be fixed by John's suggestion, so
you can ignore my previous offlist email which mentions a page fault. I've
isolated it to a problem with Agg, which I'll post the symptoms of to the
users list.
Gary

>>>>> "David" == David Moore <da...@sj...> writes:
 David> Hi, I'm posting this to the mailing list first because I'm
 David> not sure I've fixed this correctly. This is a patch for a
 David> bug in ticker.py.
Another user told me about this one offlist and the fix is in CVS. My
approach was to revert to what we were doing before (which somehow got
lost in translation) which is to set the view lim to be plus/minus 1
the constant value.
Thanks for the patch, anyhow!
JDH

Hi,
I'm posting this to the mailing list first because I'm not sure I've 
fixed this correctly. This is a patch for a bug in ticker.py.
In Autolocator.get_locator, if the distance passed in is 0, the call to 
log10 will crash with an overflow error. I think I'm hitting this 
problem because I plot a lot of binary data, and the first plot when 
this happened, consisted of all zeros (giving the y axis no distance - I 
could be talking complete nonsense, I haven't really examined the 
code). What I'm not sure about is, in the case that distance is 0, 
should the variable ld be set to 0 (as I have done in the attached 
patch), or is another value more sensible?
thanks,
David Moore

Hi Andrew,
I generally agree; my "fix" is certainly only a hack.
I just sent John privately a slightly less obnoxious
variant that looks at distutils.sysconfig.PREFIX
instead of looking at sys.path.
d
On Apr 21, 2004, at 1:58 PM, Andrew Straw wrote:
> Dear Daishi,
>
> I appreciate that you're trying to make life easier for Fink users, but
> I don't thing the Right Thing on an OS X platform involves
> automatically determining if a user has Fink installed and then using
> it. Fink is, to some degree, a different platform than stock OS X, and
> even if one has Fink installed, it doesn't mean that one wants to build
> one's Python extensions against it. (Even if one has a subdirectory of
> /sw in sys.path.)
>
> I would propose one or both of these solutions: 1) a non-automatic
> "use_fink" flag in the setup(ext) file and/or as a command line switch.
> 2) building (and maintaining!) a python-matplotlib Fink package.
>
> Cheers!
> Andrew
>
> On Apr 21, 2004, at 12:12 PM, Daishi Harada wrote:
>
>> re: nav toolbar
>> Works for me with my fink build of the latest CVS matplotlib.
>>
>> On the other hand, I'm still having problems building with fink
>> out of the box. I'm attaching my latest setupext.py. It includes
>> a hack to distinguish between vanilla darwin vs finked.
>>
>> Finally, I've noticed that under WX the GUI thread doesn't
>> terminate when the close button is pressed.
>>
>> d
>>
>> <setupext.py>
>>
>> On Apr 20, 2004, at 7:11 PM, John Hunter wrote:
>>
>>>
>>> On my OS X machine, the navigation toolbar is not showing up in
>>> backend_wx, though the other widgets (eg axes selector, status bar)
>>> seem to work fine. If anyone has any ideas, please let me know.
>>>
>>> JDH
>>>
>>>
>>> -------------------------------------------------------
>>> This SF.Net email is sponsored by: IBM Linux Tutorials
>>> Free Linux tutorial presented by Daniel Robbins, President and CEO of
>>> GenToo technologies. Learn everything from fundamentals to system
>>> administration.http://ads.osdn.com/?ad_id=1470&alloc_id=3638&op=click
>>> _______________________________________________
>>> Matplotlib-devel mailing list
>>> Mat...@li...
>>> https://lists.sourceforge.net/lists/listinfo/matplotlib-devel
>>>
>

Hi John and Paul,
I changed the 'pass' to a 'continue' and a different problem has revealed
itself. Now, in lieu of any plot window, I get a Windows dialog with title
'Windows Visual C++ Runtime Library'
and content
Runtime Error
Program C:\APPS\PYTHON23\PYTHONW.EXE
abnormal program termination
Clicking the OK then generates a page fault. This seems a bit nasty. When I
have time later today, I'll see what happens under Win2000 on my laptop. I
should probably try another backend too.
Gary
*********** REPLY SEPARATOR ***********
On 21/04/2004 at 10:04 John Hunter wrote:
> >>>>> "gary" == gary ruben <gr...@bi...> writes:
> 
> gary> I just uninstalled the 0.53b release and installed 0.53
> gary> There's still a problem on my Win98 system, but now it seems
> gary> related to the new font manager:
> 
> This is a bug (clearly!). I didn't catch it in my tests because it
> only arises when you get an OSError on a win32 registry call, which I
> never got. I believe Paul lifted this section of the code from
> ttfquery. For a quick fix, in the win32InstalledFonts function, try
> replacing
> 
> try:
> local = _winreg.OpenKey(_winreg.HKEY_LOCAL_MACHINE, fontdir)
> except OSError:
> pass
> 
> with 
> 
> try:
> local = _winreg.OpenKey(_winreg.HKEY_LOCAL_MACHINE, fontdir)
> except OSError:
> continue
> 
> and let me know how things go.
> 
> Paul, is the continue the correct behavior here?
> 
> Thanks for the report,
> JDH
------------------------------------
Gary Ruben gr...@bi...
<http://users.bigpond.net.au/gazzar>

Dear Daishi,
I appreciate that you're trying to make life easier for Fink users, but 
I don't thing the Right Thing on an OS X platform involves 
automatically determining if a user has Fink installed and then using 
it. Fink is, to some degree, a different platform than stock OS X, and 
even if one has Fink installed, it doesn't mean that one wants to build 
one's Python extensions against it. (Even if one has a subdirectory of 
/sw in sys.path.)
I would propose one or both of these solutions: 1) a non-automatic 
"use_fink" flag in the setup(ext) file and/or as a command line switch. 
2) building (and maintaining!) a python-matplotlib Fink package.
Cheers!
Andrew
On Apr 21, 2004, at 12:12 PM, Daishi Harada wrote:
> re: nav toolbar
> Works for me with my fink build of the latest CVS matplotlib.
>
> On the other hand, I'm still having problems building with fink
> out of the box. I'm attaching my latest setupext.py. It includes
> a hack to distinguish between vanilla darwin vs finked.
>
> Finally, I've noticed that under WX the GUI thread doesn't
> terminate when the close button is pressed.
>
> d
>
> <setupext.py>
>
> On Apr 20, 2004, at 7:11 PM, John Hunter wrote:
>
>>
>> On my OS X machine, the navigation toolbar is not showing up in
>> backend_wx, though the other widgets (eg axes selector, status bar)
>> seem to work fine. If anyone has any ideas, please let me know.
>>
>> JDH
>>
>>
>> -------------------------------------------------------
>> This SF.Net email is sponsored by: IBM Linux Tutorials
>> Free Linux tutorial presented by Daniel Robbins, President and CEO of
>> GenToo technologies. Learn everything from fundamentals to system
>> administration.http://ads.osdn.com/?ad_id=1470&alloc_id=3638&op=click
>> _______________________________________________
>> Matplotlib-devel mailing list
>> Mat...@li...
>> https://lists.sourceforge.net/lists/listinfo/matplotlib-devel
>>

re: nav toolbar
Works for me with my fink build of the latest CVS matplotlib.
On the other hand, I'm still having problems building with fink
out of the box. I'm attaching my latest setupext.py. It includes
a hack to distinguish between vanilla darwin vs finked.
Finally, I've noticed that under WX the GUI thread doesn't
terminate when the close button is pressed.
d

John Hunter wrote:
>>>>>>"gary" == gary ruben <gr...@bi...> writes:
> 
> 
> gary> I just uninstalled the 0.53b release and installed 0.53
> gary> There's still a problem on my Win98 system, but now it seems
> gary> related to the new font manager:
> 
> This is a bug (clearly!). I didn't catch it in my tests because it
> only arises when you get an OSError on a win32 registry call, which I
> never got. I believe Paul lifted this section of the code from
> ttfquery. For a quick fix, in the win32InstalledFonts function, try
> replacing
> 
> try:
> local = _winreg.OpenKey(_winreg.HKEY_LOCAL_MACHINE, fontdir)
> except OSError:
> pass
> 
> with 
> 
> try:
> local = _winreg.OpenKey(_winreg.HKEY_LOCAL_MACHINE, fontdir)
> except OSError:
> continue
> 
> and let me know how things go.
> 
> Paul, is the continue the correct behavior here?
Yes. I'll fix it.
-- 
Paul Barrett, PhD Space Telescope Science Institute
Phone: 410-338-4475 ESS/Science Software Branch
FAX: 410-338-4767 Baltimore, MD 21218

>>>>> "gary" == gary ruben <gr...@bi...> writes:
 gary> I just uninstalled the 0.53b release and installed 0.53
 gary> There's still a problem on my Win98 system, but now it seems
 gary> related to the new font manager:
This is a bug (clearly!). I didn't catch it in my tests because it
only arises when you get an OSError on a win32 registry call, which I
never got. I believe Paul lifted this section of the code from
ttfquery. For a quick fix, in the win32InstalledFonts function, try
replacing
 try:
 local = _winreg.OpenKey(_winreg.HKEY_LOCAL_MACHINE, fontdir)
 except OSError:
 pass
with 
 try:
 local = _winreg.OpenKey(_winreg.HKEY_LOCAL_MACHINE, fontdir)
 except OSError:
 continue
and let me know how things go.
Paul, is the continue the correct behavior here?
Thanks for the report,
JDH

I just uninstalled the 0.53b release and installed 0.53
There's still a problem on my Win98 system, but now it seems related to the
new font manager:
>>> from matplotlib.matlab import *
Traceback (most recent call last):
 File "<pyshell#0>", line 1, in -toplevel-
 from matplotlib.matlab import *
 File "C:\APPS\PYTHON23\Lib\site-packages\matplotlib\matlab.py", line 128,
in -toplevel-
 from axes import Axes
 File "C:\APPS\PYTHON23\Lib\site-packages\matplotlib\axes.py", line 10, in
-toplevel-
 from axis import XTick, YTick, XAxis, YAxis
 File "C:\APPS\PYTHON23\Lib\site-packages\matplotlib\axis.py", line 22, in
-toplevel-
 from font_manager import FontProperties
 File "C:\APPS\PYTHON23\Lib\site-packages\matplotlib\font_manager.py",
line 885, in -toplevel-
 fontManager = FontManager()
 File "C:\APPS\PYTHON23\Lib\site-packages\matplotlib\font_manager.py",
line 740, in __init__
 self.ttffiles = findSystemFonts(paths) + findSystemFonts()
 File "C:\APPS\PYTHON23\Lib\site-packages\matplotlib\font_manager.py",
line 159, in findSystemFonts
 for f in win32InstalledFonts(fontdir):
 File "C:\APPS\PYTHON23\Lib\site-packages\matplotlib\font_manager.py",
line 107, in win32InstalledFonts
 if not local:
UnboundLocalError: local variable 'local' referenced before assignment
Gary
*********** REPLY SEPARATOR ***********
On 21/04/2004 at 09:33 John Hunter wrote:
> What's new in matplotlib 0.53
> 
> Improved font manager and support
> 
> Paul Barrett has thoroughly overhauled font support. FontTools and
> ttfquery are no longer required for font finding as matplotlib now
> has a completely freestanding freetype2 implementation and font
> finder. Among other things, this should enable you to specify fonts
> in your scripts and matplotlibrc file and generate consistent
> figures across backends and operating systems. The font finder
> algorithm and implementation are based on the W3C standard
> http://www.w3.org/TR/1999/REC-CSS1-19990111.
> 
> See the font manager module documentation, the fonts documentation
> http://matplotlib.sf.net/fonts.html and the updated .matplotlibrc
> file for more details; please update your .matplotlibrc. Thanks
> Paul!
> 
> Backend WXAgg
> 
> Antigrain rendering to wxpython applications and figure windows. Now
> wx users have access to all the latest matplotlib functionality,
> including mathtext, antialised drawing, alpha blending and image
> support.
> 
> Major and minor ticks
> 
> Full support for major and minor ticks with a bevy of more
> intelligent tick locators supplied in the ticker module. Fully
> customizable and user definable tick locators and formatters. See
> major_minor_demo1.py and major_minor_demo2.py. The default tick
> labeler is much more intelligent in choosing good tick locations.
> See http://matplotlib.sf.net/matplotlib.ticker.html
> 
> Date plots
> 
> A new command a plot_date command for plotting date dependent data;
> see http://matplotlib.sf.net/screenshots.html#date_demo. Converters
> supplied in the dates module allow you to work with a variety of
> datetime instances. Custom date locators and formatters allow you to
> place major and minor ticks by minute, hour, weekday, month, year,
> etc, and use strftime format strings to format the ticks. See
> examples date_demo1.py and date_demo2.py. The dates documentation
> provides an overview and guide to with dates -
> see http://matplotlib.sf.net/matplotlib.dates.html.
> 
> Ported image support to numarray and postscript backend
> 
> The image module now works with Numeric or numarray, and now works
> in the postscript backend as well as GTKAgg, TkAgg, WXAgg, Agg, and
> GTK. Thanks to Todd Miller for the PS work!
> 
> Changes to matplotlibrc
> 
> Many features added to the default config file for font support,
> tkagg windowing in win32, and more. Please use the new file at
> http://matplotlib.sf.net/.matplotlibrc. By default, the installer
> will overwrite the existing file in the install path, so if you want
> to preserve your's, please move it to your HOME dir and set the
> environment variable if necessary.
> 
> load and save commands
> 
> Helper functions for loading and saving ASCII arrays. See load and
> save in the matlab interface.
> 
> Two scales on the same axes
> 
> Added some features to the axis and ticks to allow two plots with
> different scales on the "same" axes with different scales, ticks and
> labels on the left and right side of the x axis. To see why same is
> quoted, see examples/two_scales.py.
> 
> finance module
> 
> The finance module includes a function to fetch quotes from yahoo,
> to draw candlestick plots, and to draw vertical line plots for
> high-low range with open-close ticks to the left and right. I'm
> hoping that user contributions will make up the bulk of this module
> since I'm not a finance guy! See
> http://matplotlib.sf.net/screenshots.html#date_demo.
> 
> 
> http://matplotlib.sourceforge.net
> 
> 
> -------------------------------------------------------
> This SF.Net email is sponsored by: IBM Linux Tutorials
> Free Linux tutorial presented by Daniel Robbins, President and CEO of
> GenToo technologies. Learn everything from fundamentals to system
> administration.http://ads.osdn.com/?ad_id=1470&alloc_id=3638&op=click
> _______________________________________________
> Matplotlib-devel mailing list
> Mat...@li...
> https://lists.sourceforge.net/lists/listinfo/matplotlib-devel

What's new in matplotlib 0.53
Improved font manager and support
 Paul Barrett has thoroughly overhauled font support. FontTools and
 ttfquery are no longer required for font finding as matplotlib now
 has a completely freestanding freetype2 implementation and font
 finder. Among other things, this should enable you to specify fonts
 in your scripts and matplotlibrc file and generate consistent
 figures across backends and operating systems. The font finder
 algorithm and implementation are based on the W3C standard
 http://www.w3.org/TR/1999/REC-CSS1-19990111.
 See the font manager module documentation, the fonts documentation
 http://matplotlib.sf.net/fonts.html and the updated .matplotlibrc
 file for more details; please update your .matplotlibrc. Thanks
 Paul!
Backend WXAgg
 Antigrain rendering to wxpython applications and figure windows. Now
 wx users have access to all the latest matplotlib functionality,
 including mathtext, antialised drawing, alpha blending and image
 support.
Major and minor ticks
 Full support for major and minor ticks with a bevy of more
 intelligent tick locators supplied in the ticker module. Fully
 customizable and user definable tick locators and formatters. See
 major_minor_demo1.py and major_minor_demo2.py. The default tick
 labeler is much more intelligent in choosing good tick locations.
 See http://matplotlib.sf.net/matplotlib.ticker.html
Date plots
 A new command a plot_date command for plotting date dependent data;
 see http://matplotlib.sf.net/screenshots.html#date_demo. Converters
 supplied in the dates module allow you to work with a variety of
 datetime instances. Custom date locators and formatters allow you to
 place major and minor ticks by minute, hour, weekday, month, year,
 etc, and use strftime format strings to format the ticks. See
 examples date_demo1.py and date_demo2.py. The dates documentation
 provides an overview and guide to with dates -
 see http://matplotlib.sf.net/matplotlib.dates.html.
Ported image support to numarray and postscript backend
 The image module now works with Numeric or numarray, and now works
 in the postscript backend as well as GTKAgg, TkAgg, WXAgg, Agg, and
 GTK. Thanks to Todd Miller for the PS work!
Changes to matplotlibrc
 Many features added to the default config file for font support,
 tkagg windowing in win32, and more. Please use the new file at
 http://matplotlib.sf.net/.matplotlibrc. By default, the installer
 will overwrite the existing file in the install path, so if you want
 to preserve your's, please move it to your HOME dir and set the
 environment variable if necessary.
load and save commands
 Helper functions for loading and saving ASCII arrays. See load and
 save in the matlab interface.
Two scales on the same axes
 Added some features to the axis and ticks to allow two plots with
 different scales on the "same" axes with different scales, ticks and
 labels on the left and right side of the x axis. To see why same is
 quoted, see examples/two_scales.py.
finance module
 The finance module includes a function to fetch quotes from yahoo,
 to draw candlestick plots, and to draw vertical line plots for
 high-low range with open-close ticks to the left and right. I'm
 hoping that user contributions will make up the bulk of this module
 since I'm not a finance guy! See
 http://matplotlib.sf.net/screenshots.html#date_demo.
http://matplotlib.sourceforge.net

John,
I'd like to request that the matplotlib/examples are made into
executable scripts by adding 
#!/usr/bin/env python
as first line for all examples. This benefits Unix users and has no
effect (as far as I know) for non-Unix users.
It can be done for all scripts from the command line with a 'sh' style
shell, using something like:
cd examples
echo "#!/usr/bin/env python" >/tmp/header
for file in *.py; do
 mv $file ${file}.old
 cat /tmp/header ${file}.old > $file
done
# check the new files look OK and delete *.old
Regards
Steve

I recommend removing the "Close the Figure" button from the toolbar.
It just destroys the widget / exits the program, which you can already
do by clicking the close button that the window manager provides.
The GNOME Human Interface Guidelines, Chapter 5 Toolbars, says
"Don't add buttons for Help, Close or Quit to your toolbar by default,
as these are rarely used and the space is better used for more useful
controls."
http://developer.gnome.org/projects/gup/hig/1.0/toolbars.html
This reference is for GNOME/GTK+ applications and other windowing
environments will have their own guidelines, but I think they agree that
Close/Quit buttons on toolbars are not needed.
Check your own system - run some applications you use that have toolbars
and check if there is a quit button, and even if there is would it cause
a problem if it wasn't there?
Steve

On my OS X machine, the navigation toolbar is not showing up in
backend_wx, though the other widgets (eg axes selector, status bar)
seem to work fine. If anyone has any ideas, please let me know.
JDH

Here is a beta release of matplotlib-0.53. You'll need to update your
.matplotlibrc files with the one in the included version since a lot
has been added and some things have been changed.
There is still a bug in the date conversion module but I'm working on
it. Everything else looks pretty good from my end. New features you
may want to test
 - font family support - you can flush ttfquery and FontTools; see
 the new matplotlibrc file in the src dist for details
 - wxagg
 - major and minor ticks; see examples/major_minor*; see module docs
 for details
 - date plots (with a small tick location bug) examples/date*; ; see
 module docs for details
 - image module for numarray and PS
 - multiple scales on same axes - examples/two_scales.py
Win32 and OS X testers especially needed....
 http://matplotlib.sf.net/matplotlib-0.53b.tar.gz
 http://matplotlib.sf.net/matplotlib-0.53b.zip
 http://matplotlib.sf.net/matplotlib-0.53b.win32-py2.3.exe
JDH

>>>>> "John" == John Hunter <jdh...@ac...> writes:
 John> http://sourceforge.net/mailarchive/message.php?msg_id=7777103
This is just the usual sourceforge flakiness. This was a link to
Daishi's post "Some notes on building from CVS" which was archived a
few days ago, but now the sf archiver only has messages from April 1st
and before. 
JDH

Just as a bit more info to John's post, I believe that Tk requires the 
X11 headers, even if X11 is not to be used (as in the Aqua case). So 
don't worry that it asks for it. It's just a weird build dependency 
that shouldn't be there...
Also, John, WRT the link your gave, SourceForge says: "Message Not 
Found. This message does not (any longer) exist."
Cheers!
Andrew
On Apr 17, 2004, at 10:28 AM, John Hunter wrote:
>>>>>> "Kenneth" == Kenneth McDonald <kmm...@wi...> writes:
>
> Kenneth> I believe this is the right list, since the question
> Kenneth> deals with compiling, rather than using, matplotlib.
> Kenneth> Hope that's OK.
>
> Fine by me...
>
> Kenneth> I just dl'ed, and after setting the TkAgg flag to 1 in
> Kenneth> setup.py, did a python setup.py install. (Note that I had
> Kenneth> done a previous install without the TkAgg flag set.) On
> Kenneth> my system, this is producing a bunch of errors related to
> Kenneth> the Tk headers:
>
> Generally only the first error message from the compiler is
> informative. Subsequent ones often indicate a downstream component
> that failed after the first failed. In your case, the 1st error is
>
> In file included from src/_tkagg.cpp:18:
> /Library/Frameworks/Tk.framework/Headers/tk.h:96:29: X11/Xlib.h: No
> such file or directory
>
> So it looks to me that either you haven't installed X11 (which is the
> default config on OS X) or it's not in your include path. If you
> haven't installed X11 for OS X, see
> http://www.apple.com/macosx/features/x11.
>
> After you get past this hurdle, you might encounter another road block
> or two, especially if you are a fink user. In that case, you may want
> to edit setupext.py and replace the dictionary entry
>
> 'darwin' : '/usr/local',
>
> with
>
> 'darwin' : '/sw',
>
> and check out this post
>
> http://sourceforge.net/mailarchive/message.php?msg_id=7777103
>
> Good luck!
> JDH
>
>
> -------------------------------------------------------
> This SF.Net email is sponsored by: IBM Linux Tutorials
> Free Linux tutorial presented by Daniel Robbins, President and CEO of
> GenToo technologies. Learn everything from fundamentals to system
> administration.http://ads.osdn.com/?ad_id=1470&alloc_id=3638&op=click
> _______________________________________________
> Matplotlib-devel mailing list
> Mat...@li...
> https://lists.sourceforge.net/lists/listinfo/matplotlib-devel

I overhauled the tick locator and formating in matplotlib CVS. This
provides a lot of nice features, like easily setting major and minor
ticks on multiples of some number, auto locators that pick good ranges
and tick locations dynamically during navigation, a bevy of datetime
tick locators and formatters. If major and minor tick locators are
present, only the major tick locator is used for axis autoranging.
For tick formatting, you can use autoformatting (the default), fixed
tick labels, sprintf format strings, user defined functions, strftime
date format strings or custom classes. This is described in detail in
the matplotlib.dates and matplotlib.ticker module documentation in
CVS.
Perry had mentioned that his group was very interested in some custom
major and minor tick locators and formatters of interest to
astronomers. So if you want to start adding these (or anyone else who
wants to add some useful ones for their application domain), please
check them into CVS or send them to the list.
The new examples major_minor_demo*.py and date_demo*.py show how to
use the new classes.
JDH

I'm using Python/Tkinter under OS X, but with the Aqua version of Tcl/Tk
rather than with the X11 version. Unfortunately, I can't get 
matplotlib's
installation to complete correctly with tkagg turned on; it gets to
a certain point in the compile and then says:
In file included from src/ft2font.c:1:
src/ft2font.h:7:22: ft2build.h: No such file or directory
src/ft2font.h:8:10: #include expects "FILENAME" or <FILENAME>
src/ft2font.h:9:10: #include expects "FILENAME" or <FILENAME>
In file included from src/ft2font.c:1:
src/ft2font.h:28: error: parse error before "FT_Face"
src/ft2font.h:28: warning: no semicolon at end of struct or union
....
The named file is part of the freetype package, which I installed via 
Fink.
Unfortunately, I don't know if this is due to the fact that matplotlib
expects an X11 environment, library/path problems, or something
else.
Anyone else have matplotlib/tkagg working under TckTkAqua? If
so, any hints?
Thanks,
Ken

Hi John,
 My apologies for apparently missing your earlier response.
Interesting enough, last night when i was looking around in the 
backend_wx.py code for what might be causing those debug messages,
i thought that line: 
> wxInitAllImageHandlers() 
in conjunction the "Adding duplicate image handler" looked
suspect. I briefly thought about commenting it out...but i
figured that would probably break the tool bar icons etc. 
As you suggested, i tried commenting that line out, and sure
enough: no debug messages...and my app that does some plotting 
appears to work fine.
Thanks a lot!!
Glad my report led you to a problem you were looking for.
Jim
On 2004年4月17日, John Hunter wrote:
> 
> Hi Jim,
> 
> I responded to this earlier today on matplotlib-users. Here is a copy
> of my post. I'm very interested in getting some feedback if you try
> the suggestion below.
> 
> JDH
> 
> 
> >>>>> "Jim" == Jim Benson <jb...@se...> writes:
> 
> Jim> Hi,
> 
> Jim> I seem to be seeing some extra debug output. I have been
> Jim> using the wx background and the embedding_in_wx.py example as
> Jim> a template for how to make matplotlib plots in an app i'm
> Jim> writing. This part works great!! I am however seeing what
> Jim> appears to be a debug flag left on somewhere?
> 
> No, this is in the wx code. I traced it to the line
> 
> wxInitAllImageHandlers()
> 
> Miraculously, I commented this out, got rid of the annoying messages,
> and fixed the wx exception swallowing bug! Jeremy and I have been
> killing ourselves for about 6 months trying to figure out why backend
> wx was swallowing our exceptions, and I can't really convey how happy
> I am to have found this. We have tried all manner of hacks, google
> searches and plantive pleas to get our exceptions back to no avail.
> It made debugging wx virtually impossible.
> 
> Please test this change on your system. win32 users, I would also be
> much obliged if you could try commenting this out on your system to
> see if you lose functionality. On my linux box, the figures came up
> as expected, I was able to print to jpg and png, so all looks well.
> 
> Note, wxagg is in CVS if you need some of the agg features that native
> wx drawing does not provide. But it looks like all systems are go for
> a release next week so you may as well wait.
> 
> JDH
> 
> 
-- 
Dr. James A. Benson
United States Naval Observatory
Navy Prototype Optical Interferometer
Flagstaff, AZ 86001
Email: jb...@no...
 jb...@se...
Voice: 928-773-4868/928-779-5132
Fax: 928-779-9568

Hi Jim,
I responded to this earlier today on matplotlib-users. Here is a copy
of my post. I'm very interested in getting some feedback if you try
the suggestion below.
JDH
>>>>> "Jim" == Jim Benson <jb...@se...> writes:
 Jim> Hi,
 Jim> I seem to be seeing some extra debug output. I have been
 Jim> using the wx background and the embedding_in_wx.py example as
 Jim> a template for how to make matplotlib plots in an app i'm
 Jim> writing. This part works great!! I am however seeing what
 Jim> appears to be a debug flag left on somewhere?
No, this is in the wx code. I traced it to the line
 wxInitAllImageHandlers()
Miraculously, I commented this out, got rid of the annoying messages,
and fixed the wx exception swallowing bug! Jeremy and I have been
killing ourselves for about 6 months trying to figure out why backend
wx was swallowing our exceptions, and I can't really convey how happy
I am to have found this. We have tried all manner of hacks, google
searches and plantive pleas to get our exceptions back to no avail.
It made debugging wx virtually impossible.
Please test this change on your system. win32 users, I would also be
much obliged if you could try commenting this out on your system to
see if you lose functionality. On my linux box, the figures came up
as expected, I was able to print to jpg and png, so all looks well.
Note, wxagg is in CVS if you need some of the agg features that native
wx drawing does not provide. But it looks like all systems are go for
a release next week so you may as well wait.
JDH

Hi,
 After i posted a version of this question on the 
matplotlib-users list i decided it would probably be better
to post it on the devel list. Here is the problem, if i include 
any one of the following three lines: 
from matplotlib.backends.backend_wx import FigureCanvasWx
#from matplotlib.backends.backend_wx import Toolbar
#from matplotlib.backends.backend_wx import FigureManager
 
(here only the first one is uncommented...same result if i
use either of the other two) in a file (test.py) and then type: 
python test.py
i get:
floyd:/home/jbenson/python>python test.py
[Debug] 15:17:10: Adding duplicate image handler for 'PNG file'
[Debug] 15:17:10: Adding duplicate image handler for 'JPEG file'
[Debug] 15:17:10: Adding duplicate image handler for 'TIFF file'
[Debug] 15:17:10: Adding duplicate image handler for 'GIF file'
[Debug] 15:17:10: Adding duplicate image handler for 'PNM file'
[Debug] 15:17:10: Adding duplicate image handler for 'PCX file'
[Debug] 15:17:10: Adding duplicate image handler for 'XPM file'
[Debug] 15:17:10: Adding duplicate image handler for 'Windows icon file'
[Debug] 15:17:10: Adding duplicate image handler for 'Windows cursor file'
[Debug] 15:17:10: Adding duplicate image handler for 'Windows animated 
cursor file'
floyd:/home/jbenson/python>
If i try the same line from a python prompt line:
floyd:/home/jbenson>python
Python 2.3.3 (#1, Dec 21 2003, 15:13:18)
[GCC 3.3.1] on linux2
Type "help", "copyright", "credits" or "license" for more information.
>>> from matplotlib.backends.backend_wx import FigureCanvasWx
>>>
...no extra debug lines. I'm using matplotlib-0.52. Does anyone recognize
where those [Debug] statements are coming from, and how i could
get them to not show up? 
Thanks,
Jim

>>>>> "Andrew" == Andrew Straw <str...@as...> writes:
 >> The stumper for me is that I have been unable to build
 >> scipy/numpy for the python that is bundled with os x - has
 >> anyone been successful with this? 
I have. Check the scipy mailing list archives (search for OS X).
There are several useful threads. Also, use scipy CVS, as a number of
darwin specific fixes have been added.
JDH

>>>>> "Kenneth" == Kenneth McDonald <kmm...@wi...> writes:
 Kenneth> I believe this is the right list, since the question
 Kenneth> deals with compiling, rather than using, matplotlib.
 Kenneth> Hope that's OK.
Fine by me...
 Kenneth> I just dl'ed, and after setting the TkAgg flag to 1 in
 Kenneth> setup.py, did a python setup.py install. (Note that I had
 Kenneth> done a previous install without the TkAgg flag set.) On
 Kenneth> my system, this is producing a bunch of errors related to
 Kenneth> the Tk headers:
Generally only the first error message from the compiler is
informative. Subsequent ones often indicate a downstream component
that failed after the first failed. In your case, the 1st error is
 In file included from src/_tkagg.cpp:18:
 /Library/Frameworks/Tk.framework/Headers/tk.h:96:29: X11/Xlib.h: No
 such file or directory
So it looks to me that either you haven't installed X11 (which is the
default config on OS X) or it's not in your include path. If you
haven't installed X11 for OS X, see
http://www.apple.com/macosx/features/x11.
After you get past this hurdle, you might encounter another road block
or two, especially if you are a fink user. In that case, you may want
to edit setupext.py and replace the dictionary entry
 'darwin' : '/usr/local',
with
 'darwin' : '/sw',
and check out this post
http://sourceforge.net/mailarchive/message.php?msg_id=7777103
Good luck!
JDH