matplotlib-users Mailing List for matplotlib (Page 4)

Chris <listservs@...> writes:
> > 
> > Based on recent emails, this looks like a problem attributed to the gcc 
> > version, not the python version. Suggested solutions are compile with 
> > the -Os flag or use gcc 4.2.
> > 
> I think universal builds may have to 
> wait for another day, when gcc 4.2 is ready for prime time.
> 
Actually, the -0s flag with gcc 4.0 seemed to work just fine. Thanks.

Eric Firing <efiring@...> writes:
> 
> Chris,
> 
> Based on recent emails, this looks like a problem attributed to the gcc 
> version, not the python version. Suggested solutions are compile with 
> the -Os flag or use gcc 4.2.
> 
Hmm. The problem with using gcc 4.2 is that numpy does not seem to
like it. I am also leery about using a developer preview of *anything* to
build packages for distribution. I think universal builds may have to 
wait for another day, when gcc 4.2 is ready for prime time.

Chris,
Based on recent emails, this looks like a problem attributed to the gcc 
version, not the python version. Suggested solutions are compile with 
the -Os flag or use gcc 4.2.
Eric
Chris wrote:
> I'm trying to get a built of Matplotlib built under Python.org Python 2.5.2, 
> but get
> the following build error, which did not occur under Leopard's 
> python:
> 
> src/_image.cpp: In member function ‘Py::Object _image_module::
> from_images (const Py::Tuple&)’:
> src/_image.cpp:848: error: insn does not satisfy its constraints:
> (insn 2573 1070 2574 126 agg24/include/agg_color_rgba.h:268 (set 
> (mem:QI (plus:SI (reg/f:SI 6 bp)
> (const_int -280 [0xfffffffffffffee8])) [0 SR.2970+0 S1 A8])
> (reg:QI 5 di)) 56 {*movqi_1} (nil)
> (nil))
> src/_image.cpp:848: internal compiler error: in 
> reload_cse_simplify_operands, at postreload.c:391
> Please submit a full bug report,
> with preprocessed source if appropriate.
> See <URL:http://developer.apple.com/bugreporter> for instructions.
> error: command 'gcc' failed with exit status 1
> 
> 
> Any idea as to what the problem may be?
> 
> Thanks in advance.
> 
> 
> -------------------------------------------------------------------------
> This SF.net email is sponsored by the 2008 JavaOne(SM) Conference 
> Don't miss this year's exciting event. There's still time to save 100ドル. 
> Use priority code J8TL2D2. 
> http://ad.doubleclick.net/clk;198757673;13503038;p?http://java.sun.com/javaone
> _______________________________________________
> Matplotlib-users mailing list
> Mat...@li...
> https://lists.sourceforge.net/lists/listinfo/matplotlib-users

Michael Droettboom <mdroe@...> writes:
> 
> Paul Smith wrote:
> > Hi Michael,
> >
> > I put in the rc line you suggested below into fonts_demo.py but didn't see 
it 
> > print any extra info (but did confirm in ipython that rcParams showed 
> > verbose.level had changed to "annoying"). It just quietly finished 
otherwise. 
> > Did I miss something here? 
> > 
> Does it work if you put this into your matplotlibrc?
> 
> verbose.level: debug-annoying
> 
> (Note it's debug-annoying, not simply annoying)
> > I've linked the output of fonts_demo.py to;
> > https://www.box.net/shared/static/o693hq3soo.png
> > 
> Hmm. That font is definitely not Vera Sans. Something really odd is 
> going on here.
> I can appreciate that goal -- and AFAIK it does work for other users on 
Ubuntu 7.10 so there is probably just
> some configuration problem here that we can hopefully get to the bottom of.
> 
> Do you have any customizations in your matplotlibrc?
> 
> Cheers,
> Mike
> 
Mike,
I've not made any other changes to matplotlibrc. In fact I only just now have 
copied the one in /etc to my $HOME/.matplotlib directory. It still seemed to 
pick up the /etc version as you'll see below, I guess coz I was sudo'd as 
root? :) anyway...
Interesting the font search path seems to only include mpl-data fonts, but it 
does seem to know about the ones in /usr/share/fonts/truetype/freefont. Then 
fails to use one, so we're always back to cmr10. The output below was for my 
test plot but it's pretty much the same (longer) story for fonts_demo.py I 
tried copying all the freetype fonts into mpl-data/font and everything came 
out FreeSansOblique. Still failed to find the correct font, but used something 
a little closer. What do you make of it?
I'll be away for a while but I'll check the list if I can.
Cheers,
Paul
debug-annoying dump from my test plot program
----------------------------------------
loaded rc file /etc/matplotlibrc
matplotlib version 0.90.1
verbose.level debug-annoying
interactive is False
units is True
platform is linux2
loaded modules: 
['_bisect', 'distutils', 'random', 'datetime', 'matplotlib.tempfile', 'distutil
s.sysconfig', 'encodings.encodings', 'pytz.cStringIO', 'struct', 'tempfile', 'p
ytz.os', 'zipimport', 'string', 'encodings.utf_8', 'matplotlib.__future__', 'py
tz.tzinfo', 'pytz.datetime', 'distutils.re', 'bisect', 'signal', 'matplotlib.py
tz', 'pytz.tzfile', 'cStringIO', 'locale', 'encodings', 'dateutil', 'matplotlib
.warnings', 'pytz.pytz', 'matplotlib.sys', 're', 'math', 'fcntl', 'UserDict', '
distutils.os', 'matplotlib', 'codecs', 'md5', '_locale', 'matplotlib.os', 'thre
ad', 'itertools', 'distutils.sys', 'os', '__future__', '_sre', '__builtin__', '
matplotlib.re', 'operator', 'distutils.string', 'matplotlib.datetime', 'posixpa
th', 'errno', 'binascii', 'sre_constants', 'matplotlib.md5', 'types', 'pytz.sys
', '_codecs', 'pytz', 'copy', '_struct', '_types', 'matplotlib.dateutil', 'hash
lib', 'posix', 'encodings.aliases', 'exceptions', 'sre_parse', 'pytz.bisect', '
distutils.distutils', 'copy_reg', 'sre_compile', '_hashlib', '_random', 'pytz.s
truct', 'site', '__main__', 'shutil', 'strop', 'encodings.codecs', 'gettext', '
pytz.sets', 'stat', 'warnings', 'encodings.types', 'sys', 'os.path', 'pytz.gett
ext', 'matplotlib.distutils', 'distutils.errors', 'linecache', 'matplotlib.shut
il', 'sets']
numerix numpy 1.0.3
font search path ['/usr/share/matplotlib/mpl-
data/fonts/ttf', '/usr/share/matplotlib/mpl-data/fonts/afm']
trying fontname /usr/share/matplotlib/mpl-data/fonts/ttf/cmr10.ttf
trying fontname /usr/share/matplotlib/mpl-data/fonts/ttf/cmtt10.ttf
trying fontname /usr/share/matplotlib/mpl-data/fonts/ttf/cmmi10.ttf
trying fontname /usr/share/matplotlib/mpl-data/fonts/ttf/cmex10.ttf
trying fontname /usr/share/matplotlib/mpl-data/fonts/ttf/cmsy10.ttf
trying fontname /usr/share/fonts/truetype/ttf-dejavu/DejaVuSansMono-Bold.ttf
trying fontname /usr/share/fonts/truetype/ttf-dejavu/DejaVuSerifCondensed-
Bold.ttf
trying fontname /usr/share/fonts/truetype/ttf-dejavu/DejaVuSansMono-Oblique.ttf
trying fontname /usr/share/fonts/truetype/ttf-dejavu/DejaVuSansCondensed-
Oblique.ttf
trying fontname /usr/share/fonts/truetype/ttf-dejavu/DejaVuSansCondensed.ttf
trying fontname /usr/share/fonts/truetype/ttf-dejavu/DejaVuSerif-Italic.ttf
trying fontname /usr/share/fonts/truetype/freefont/FreeSerifBoldItalic.ttf
trying fontname /usr/share/fonts/truetype/freefont/FreeMonoOblique.ttf
trying fontname /usr/share/fonts/truetype/ttf-dejavu/DejaVuSerif-BoldItalic.ttf
trying fontname /usr/share/fonts/truetype/ttf-dejavu/DejaVuSans.ttf
trying fontname /usr/share/fonts/truetype/ttf-dejavu/DejaVuSansMono.ttf
trying fontname /usr/share/fonts/truetype/freefont/FreeSerif.ttf
trying fontname /usr/share/fonts/truetype/ttf-dejavu/DejaVuSerif-Bold.ttf
trying fontname /usr/share/fonts/truetype/freefont/FreeSerifBold.ttf
trying fontname /usr/share/fonts/truetype/ttf-dejavu/DejaVuSerif.ttf
trying fontname /usr/share/fonts/truetype/ttf-dejavu/DejaVuSerifCondensed-
Italic.ttf
trying fontname /usr/share/fonts/truetype/freefont/FreeSansOblique.ttf
trying fontname /usr/share/fonts/truetype/ttf-dejavu/DejaVuSerifCondensed.ttf
trying fontname /usr/share/fonts/truetype/ttf-dejavu/DejaVuSansCondensed-
Bold.ttf
trying fontname /usr/share/fonts/truetype/freefont/FreeSansBoldOblique.ttf
trying fontname /usr/share/fonts/truetype/ttf-dejavu/DejaVuSans-Oblique.ttf
trying fontname /usr/share/fonts/truetype/ttf-dejavu/DejaVuSerifCondensed-
BoldItalic.ttf
trying fontname /usr/share/fonts/truetype/freefont/FreeMonoBoldOblique.ttf
trying fontname /usr/share/fonts/truetype/freefont/FreeSansBold.ttf
trying fontname /usr/share/fonts/truetype/freefont/FreeSans.ttf
trying fontname /usr/share/fonts/truetype/ttf-dejavu/DejaVuSans-BoldOblique.ttf
trying fontname /usr/share/fonts/truetype/freefont/FreeMono.ttf
trying fontname /usr/share/fonts/truetype/ttf-dejavu/DejaVuSansMono-
BoldOblique.ttf
trying fontname /usr/share/fonts/truetype/freefont/FreeSerifItalic.ttf
trying fontname /usr/share/fonts/truetype/ttf-dejavu/DejaVuSansCondensed-
BoldOblique.ttf
trying fontname /usr/share/fonts/truetype/freefont/FreeMonoBold.ttf
trying fontname /usr/share/fonts/truetype/ttf-dejavu/DejaVuSans-Bold.ttf
trying fontname /usr/share/fonts/truetype/ttf-dejavu/DejaVuSans-ExtraLight.ttf
$HOME=/root
CONFIGDIR=/root/.matplotlib
loaded ttfcache file /root/.matplotlib/ttffont.cache
matplotlib data path /usr/share/matplotlib/mpl-data
backend Agg version v2.2
backend_agg.new_figure_manager
FigureCanvasAgg.draw
RendererAgg.__init__
RendererAgg.__init__ width=640.0, height=480.0
RendererAgg.__init__ _RendererAgg done
RendererAgg.__init__ done
RendererAgg._get_agg_font
	findfont failed Arial
	findfont failed FreeSans
Could not match FreeSans, normal, normal. Returning /usr/share/matplotlib/mpl-
data/fonts/ttf/cmr10.ttf
RendererAgg._get_agg_font
RendererAgg.draw_text
RendererAgg._get_agg_font
----cut bunch of rendering stuff------

I'm trying to get a built of Matplotlib built under Python.org Python 2.5.2, 
but get
the following build error, which did not occur under Leopard's 
python:
src/_image.cpp: In member function ‘Py::Object _image_module::
from_images (const Py::Tuple&)’:
src/_image.cpp:848: error: insn does not satisfy its constraints:
(insn 2573 1070 2574 126 agg24/include/agg_color_rgba.h:268 (set 
(mem:QI (plus:SI (reg/f:SI 6 bp)
 (const_int -280 [0xfffffffffffffee8])) [0 SR.2970+0 S1 A8])
 (reg:QI 5 di)) 56 {*movqi_1} (nil)
 (nil))
src/_image.cpp:848: internal compiler error: in 
reload_cse_simplify_operands, at postreload.c:391
Please submit a full bug report,
with preprocessed source if appropriate.
See <URL:http://developer.apple.com/bugreporter> for instructions.
error: command 'gcc' failed with exit status 1
Any idea as to what the problem may be?
Thanks in advance.

Hi,
 I want to inset one axes within a larger one... as in
 http://matplotlib.sourceforge.net/screenshots/axes_demo.py and
 http://matplotlib.sourceforge.net/screenshots/axes_demo_small.png
 but I was wondering if there is an option similar to "best" location
 in the legend function. So, instead of hardwiring, like:
 a = axes([.65, .6, .2, .2], axisbg='y'), I like it automatically place
 the inset in a "best" location!
 Thanks,
 Nino

One more comment:
Chris wrote:
> maptlotlib is the only missing piece of the "superpack" of
> modules that I distribute for OSX.
The Superpack is great, but please, please, please built it Universal -- 
if you do that, then we'll solve a lot of distribution issues for OS-X 
python users everywhere. Yes, it's more pain for you, but the Python 
packages for OS-X situation is just way too ugly -- we really need to 
make it cleaner.
-Chris
-- 
Christopher Barker, Ph.D.
Oceanographer
Emergency Response Division
NOAA/NOS/OR&R (206) 526-6959 voice
7600 Sand Point Way NE (206) 526-6329 fax
Seattle, WA 98115 (206) 526-6317 main reception
Chr...@no...

Chris wrote:
> export CFLAGS="-arch i386 -I/Developer/src/libpng 
> -I/Developer/src/freetype/include"
> export LDFLAGS="-arch i386 -L/Developer/src/libpng 
> -L/Developer/src/freetype"
> rm -rf build
> python setupegg.py bdist_egg
> 
> The build of freetype in /Developer/src/freetype does not even have 
> dynamic libs built. The basedir dict in setupext.py contains the following 
> entry 
> for OSX:
> 
> 'darwin' : []
even with that, I think setup.py looks in "standard" places.
> ImportError: dlopen(/Library/Python/2.5/site-packages/matplotlib-
> 0.98pre-py2.5-macosx-10.5-i386.egg/matplotlib/ft2font.so, 2): 
> Library not loaded: /usr/local/lib/libfreetype.6.dylib
Do you have /usr/local/lib/libfreetype.6.dylib on your build system? I 
assume so, or it wouldn't work at all.
The easiest thing to do is remove it. Not a good solution if you need it 
for other things, but if nothing else you could temporarily re-name it 
in your build script, then name it back at end.
Otherwise, poke more into setup.py and setupext.py, and remove any 
references to /usr/local/lib. Indeed, in an older version of MPL that I 
have handy, it's added in setupext.py.
An Alternative:
I posted a note about this yesterday, with no replies, so I'll try again:
Instead of all of us going through the pain of figuring out how to build 
and link static libs for MPL, and PIL, and GDAL, and ???, why don't we 
just build against the nice Frameworks here:
http://www.kyngchaos.com/wiki/software:frameworks
yes, it's an extra download and install, but it's easy, they can be 
provided by package distributors, and they can be shared by a bunch of 
python packages (and other *nix-y software). See my message yesterday 
for more detail.
I'm doing some testing with PIL -- it's very easy to build, and works 
fine. It looks like py2app picks up the libs fine, too, though I need a 
bit more testing.
-Chris
 Referenced from:
> /Library/Python/2.5/site-packages/matplotlib-0.98pre-py2.5-macosx-10.5-
> i386.egg/matplotlib/ft2font.so Reason: image not found
> 
> I thought I had my bases covered -- if anyone has some insight here, please
> let me know. maptlotlib is the only missing piece of the "superpack" of 
> modules that I distribute for OSX.
> 
> 
> -------------------------------------------------------------------------
> This SF.net email is sponsored by the 2008 JavaOne(SM) Conference 
> Don't miss this year's exciting event. There's still time to save 100ドル. 
> Use priority code J8TL2D2. 
> http://ad.doubleclick.net/clk;198757673;13503038;p?http://java.sun.com/javaone
> _______________________________________________
> Matplotlib-users mailing list
> Mat...@li...
> https://lists.sourceforge.net/lists/listinfo/matplotlib-users
-- 
Christopher Barker, Ph.D.
Oceanographer
Emergency Response Division
NOAA/NOS/OR&R (206) 526-6959 voice
7600 Sand Point Way NE (206) 526-6329 fax
Seattle, WA 98115 (206) 526-6317 main reception
Chr...@no...

On Thu, Apr 24, 2008 at 08:59:32AM -0500, John Hunter wrote:
> On Wed, Apr 23, 2008 at 1:49 PM, Glen W. Mabey <Gle...@sw...> wrote:
> > Hello,
> >
> > I'm using today's svn source and I'm surprised that the following loop
> > does not get redrawn 10 times.
> >
> > for it in range( 10 ):
> > plot( arange( it ) )
> > draw()
> > raw_input();
> >
> > That is, within a 'ipython -pylab' session. Is this really a question
> > for the ipython folks? Or am I missing something here?
> 
> Can you explain what is happening in your code? First off, in pylab
> mode this should be drawn 20 times, because in interactive mode plot
> forces a draw and draw does too, so you should get 20 draws.
> Secondly, ipython in pylab mode for some backends (Qt, GTK and WX) run
> the GUI mainloop in a thread and I am not sure what will happen with
> the "raw_input" which you expect to be blocking (it is not supported).
Hi John,
I guess it was for that reason (that the GUI mainloop runs in a
different thread) that I was expecting the draw to succeed.
> Folks have had various successes with getting blocking input for
> matplotlib in interactive sessions, but Gael has made the most
> progress. See the ginput function in svn
> 
> -- "ginput'' in pylab or matplotlib.pyplot
> 
> -- matplotlib.figure.Figure.ginput in the API
> 
> from pylab import arange, plot, sin, ginput, show
> t = arange(10)
> plot(t, sin(t))
> print "Please click"
> x = ginput(3, verbose=True)
> show()
After looking at that code, I don't immediately see how a similar
approach could be used for retrieving keyboard input from the user while
still allowing MPL drawing in the same loop. But thank you for pointing
out that example.
Best Regards,
Glen Mabey

Paul Smith wrote:
> Hi Michael,
>
> I put in the rc line you suggested below into fonts_demo.py but didn't see it 
> print any extra info (but did confirm in ipython that rcParams showed 
> verbose.level had changed to "annoying"). It just quietly finished otherwise. 
> Did I miss something here? 
> 
Does it work if you put this into your matplotlibrc?
 verbose.level: debug-annoying
(Note it's debug-annoying, not simply annoying)
> I've linked the output of fonts_demo.py to;
> https://www.box.net/shared/static/o693hq3soo.png
> 
Hmm. That font is definitely not Vera Sans. Something really odd is 
going on here.
> If I need to upgrade matplotlib above 0.90.1 I guess I'll have to build it 
> separately unfortunately. We're trying to stick to the standard ubuntu 
> packages that get pulled in for that release.
> 
I can appreciate that goal -- and AFAIK it does work for other users on Ubuntu 7.10 so there is probably just some configuration problem here that we can hopefully get to the bottom of.
Do you have any customizations in your matplotlibrc?
Cheers,
Mike
-- 
Michael Droettboom
Science Software Branch
Operations and Engineering Division
Space Telescope Science Institute
Operated by AURA for NASA

Darren Dale <dar...@co...> writes:
> pdf output uses dviread.py to parse the dvi files created by latex,
> get the font layout information, and place the glyphs. [...] there are
> some subtle and difficult to resolve limitations of dviread (like
> rendering greek letters in math mode) that have prevented us from
> using it everywhere. Jouni, would you care to comment?
The details are a little complicated, but yes, dviread.py lacks some
features of the font support in dvips. The PS backend calls dvips and so
gets to use all the font machinery in a modern TeX system, and dviread
doesn't (yet) implement all the features. Another potential problem is
that dviread makes more assumptions about the TeX system than the PS
backend, and I have only tested it on TeX Live on a Mac. Both TeX Live
and teTeX on a Unix-ish system should be fine, but I have no idea about
TeX implementations on Windows, or any commercial implementations.
For various reasons, I haven't had much time recently to hack on
matplotlib, but improving dviread is high on my todo list when I do find
the time.
-- 
Jouni K. Seppänen
http://www.iki.fi/jks

Michael Droettboom <mdroe@...> writes:
> 
> The font lookup mechanism has been much improved in 0.91.2 -- you may 
> want to try using that. In 0.90.x, often if you don't get a perfectly 
> exact match for a font, it reverts back to the default "Vera Sans". 
> Vera Sans, however, is not a fixed-width font. Can you provide the png 
> file of fonts_demo.py so I can be sure of what is happening? One way to 
> diagnose this is to do
> 
> rc("verbose", level="debug-annoying")
> 
> which will print out a bunch of stuff related to font lookup. Attach 
> the output here and I'll have a look at it to try to figure out what may 
> be going wrong.
> 
> Cheers,
> Mike
> 
> Paul Smith wrote:
> > Hi all,
> >
> > I've been trying in vain to get a better font on a plot than the fixed 
width 
> > serif one that always appears. I've got lib-freetype6 installed (on Ubuntu 
> > server), and ran fc-cache after. The .matplotlib/ttfont.cache contains 
entries 
> > for the Free* fonts. The fonts themselves are 
> > in /usr/share/fonts/truetype/freefont. The sample code below would seem 
(to 
> > me) to have used FreeSans, but the png has the same bad looking fixed-
serif 
> > font.
> > When running the matplotlib examples/fonts_demo.py the result is the same 
font 
> > all over that .png figure too (difference only in sizes).
> >
> > I usually work on XP, and it all works fine there. Does using only Agg as 
the 
> > backend make any difference whatsoever?
> >
> > Paul
> >
> > ubuntu 7.10
> > python 2.5
> > matplotlib 0.90.1
> >
> > ----
> > Some sample code;
> >
> > import matplotlib
> > from matplotlib import rc
> > rc('font',**{'family':'sans-serif','sans-serif':['Arial','FreeSans']})
> > matplotlib.use('Agg')
> > from pylab import *
> >
> > plot(arange(100))
> > ax=gca()
> > ax.set(xlabel='Useless',ylabel='Pointless')
> > draw()
> > show()
> > savefig('test')
> > l=ax.xaxis.get_label()
> > print 'font prop: ',l.get_font_properties()
> > print 'font name: ',l.get_fontname()
> >
> > ----
> > Produces this output;
> >
> > font prop: 
(['Arial', 'FreeSans'], 'normal', 'normal', 'normal', 'normal', 12)
> > font name: FreeSans
Hi Michael,
I put in the rc line you suggested below into fonts_demo.py but didn't see it 
print any extra info (but did confirm in ipython that rcParams showed 
verbose.level had changed to "annoying"). It just quietly finished otherwise. 
Did I miss something here? 
I've linked the output of fonts_demo.py to;
https://www.box.net/shared/static/o693hq3soo.png
If I need to upgrade matplotlib above 0.90.1 I guess I'll have to build it 
separately unfortunately. We're trying to stick to the standard ubuntu 
packages that get pulled in for that release.
 
Thanks for your help,
Paul

On Wed, Apr 23, 2008 at 1:49 PM, Glen W. Mabey <Gle...@sw...> wrote:
> Hello,
>
> I'm using today's svn source and I'm surprised that the following loop
> does not get redrawn 10 times.
>
> for it in range( 10 ):
> plot( arange( it ) )
> draw()
> raw_input();
>
> That is, within a 'ipython -pylab' session. Is this really a question
> for the ipython folks? Or am I missing something here?
Can you explain what is happening in your code? First off, in pylab
mode this should be drawn 20 times, because in interactive mode plot
forces a draw and draw does too, so you should get 20 draws.
Secondly, ipython in pylab mode for some backends (Qt, GTK and WX) run
the GUI mainloop in a thread and I am not sure what will happen with
the "raw_input" which you expect to be blocking (it is not supported).
Folks have had various successes with getting blocking input for
matplotlib in interactive sessions, but Gael has made the most
progress. See the ginput function in svn
 -- "ginput'' in pylab or matplotlib.pyplot
 -- matplotlib.figure.Figure.ginput in the API
from pylab import arange, plot, sin, ginput, show
t = arange(10)
plot(t, sin(t))
print "Please click"
x = ginput(3, verbose=True)
show()
JDH

On Thursday 24 April 2008 08:53:47 am you wrote:
> Although the 'align' environment inside the figure environment does not
> cause any error,
> tt seems like that the 'put' command from the 'picture' environment in
> LaTeX does not
> accept \begin{something} ~ \end{something} or \[~\].
>
> As you pointed out, my problem is coming from the limitation in LaTeX.
>
> However, there are two things I do not understand still.
>
> First, saving a plot with the align environment in 'eps' format had worked
> in the previous
> version of matplotlib (0.90.1 from ubuntu gutsy 7.10). What has been
> changed?
Nothing with the latex support, as far as I can tell from the changelogs. 
> Secondly, saving in a 'png' or 'pdf' format works still. What's the
> difference?
Saving a ps or eps file goes through the procedure of embedding the figure in 
a latex document in order for PSFrags to do its job and convert some tags 
into actual text. That way you get scalable output. png just renders the text 
as images, not as text. pdf output uses dviread.py to parse the dvi files 
created by latex, get the font layout information, and place the glyphs. 
dviread was not available when we added support to the eps/*agg backends, it 
was contributed later by Jouni Seppanen. dviread is a simpler and more 
elegant approach than the way it is done in eps/*agg, but there are some 
subtle and difficult to resolve limitations of dviread (like rendering greek 
letters in math mode) that have prevented us from using it everywhere. Jouni, 
would you care to comment?

Thanks. Your suggested changes (slightly modified) are in SVN r5070. 
Thanks for the warning output -- I'll try to tackle some of those as well.
Cheers,
Mike
Martin Spacek wrote:
> It worked! I had to make a few changes, and there's lots of warnings, 
> but it's now compiling. I've attached a patch. In backend_agg.cpp, I 
> had to replace an 'and' with '&&', and I had to replace a few 
> 'round()' calls with your 'my_round()'. I don't think my_round is 
> directly being included in backend_agg.cpp, it's still hanging around 
> from its definition in agg_py_path_iterator.h. Don't know if something 
> should be done about that.
>
> I've also attached the full output with all the warnings.
>
> Thanks Michael!
>
> Martin
>
> Michael Droettboom wrote:
>> It doesn't appear to exist in VS2008 -- at least I can't find it in 
>> the online docs. Maybe someone else can enlighten me.
>>
>> In the meantime, I just wrote my own round function and switched to 
>> use that. Please update from SVN and let me know how it goes. And 
>> again, thank you for your patience.
>>
>> Cheers,
>> Mike
>>
-- 
Michael Droettboom
Science Software Branch
Operations and Engineering Division
Space Telescope Science Institute
Operated by AURA for NASA

I don't suspect another distro will make a difference. The bug is in 
matplotlib's Cairo backend (it's interface to Cairo), not in Cairo 
itself. Change your backend to PDF and you can avoid this bug. In the 
meantime, we'll need to fix the Cairo backend in matplotlib. Sorry I 
wasn't clearer the first time.
Cheers,
Mike
Yong-Duk Jin wrote:
> Well, I may have to try it from another distro.
> Thanks ^^
>
> On Thu, Apr 24, 2008 at 3:12 AM, Michael Droettboom <md...@st... 
> <mailto:md...@st...>> wrote:
>
> This bug seems to be Cairo-backend specific. The Python PDF
> backend seems to work fine. Most likely a problem in the Cairo
> backend and not Cairo itself, but I haven't dug any further.
>
> Cheers,
> Mike
>
> Joshua Lippai wrote:
>
> I just tested it on OS X using the wxAgg backend, and it
> appears to be
> fine. I've atached the pdf image. What GUI are you having
> matplotlib
> use?
>
> Josh
>
> On Wed, Apr 23, 2008 at 8:50 AM, Yong-Duk Jin
> <ne...@gm... <mailto:ne...@gm...>> wrote:
> 
>
> I'm using matplotlib 0.91.2 from ubuntu hardy 8.04
>
> When I saved a plot in a pdf format the legend is not
> properly drawn.
> I attached the figure to demonstrate the problem. The
> legend box(?)
> is smaller than the legend label.
>
> I used the following code to draw this test plot.
>
> from pylab import *
> x = []; y = []
> figure(figsize=(2,2))
> Angle = '0.0'
> plot(x,y,label='0.0deg'); axis([0,1,0,1]); legend()
> savefig('test.pdf')
>
> Any comment?
>
> --
> Yong-Duk Jin
> -------------------------------------------------------------------------
> This SF.net email is sponsored by the 2008 JavaOne(SM)
> Conference
> Don't miss this year's exciting event. There's still time
> to save 100ドル.
> Use priority code J8TL2D2.
>
> http://ad.doubleclick.net/clk;198757673;13503038;p?http://java.sun.com/javaone
> _______________________________________________
> Matplotlib-users mailing list
> Mat...@li...
> <mailto:Mat...@li...>
> https://lists.sourceforge.net/lists/listinfo/matplotlib-users
>
>
> 
> ------------------------------------------------------------------------
>
>
>
> -------------------------------------------------------------------------
> This SF.net email is sponsored by the 2008 JavaOne(SM)
> Conference Don't miss this year's exciting event. There's
> still time to save 100ドル. Use priority code J8TL2D2.
> http://ad.doubleclick.net/clk;198757673;13503038;p?http://java.sun.com/javaone
> ------------------------------------------------------------------------
>
> _______________________________________________
> Matplotlib-users mailing list
> Mat...@li...
> <mailto: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
>
>
>
>
> -- 
> Yong-Duk Jin 
-- 
Michael Droettboom
Science Software Branch
Operations and Engineering Division
Space Telescope Science Institute
Operated by AURA for NASA

The font lookup mechanism has been much improved in 0.91.2 -- you may 
want to try using that. In 0.90.x, often if you don't get a perfectly 
exact match for a font, it reverts back to the default "Vera Sans". 
Vera Sans, however, is not a fixed-width font. Can you provide the png 
file of fonts_demo.py so I can be sure of what is happening? One way to 
diagnose this is to do
 rc("verbose", level="debug-annoying")
which will print out a bunch of stuff related to font lookup. Attach 
the output here and I'll have a look at it to try to figure out what may 
be going wrong.
Cheers,
Mike
Paul Smith wrote:
> Hi all,
>
> I've been trying in vain to get a better font on a plot than the fixed width 
> serif one that always appears. I've got lib-freetype6 installed (on Ubuntu 
> server), and ran fc-cache after. The .matplotlib/ttfont.cache contains entries 
> for the Free* fonts. The fonts themselves are 
> in /usr/share/fonts/truetype/freefont. The sample code below would seem (to 
> me) to have used FreeSans, but the png has the same bad looking fixed-serif 
> font.
> When running the matplotlib examples/fonts_demo.py the result is the same font 
> all over that .png figure too (difference only in sizes).
>
> I usually work on XP, and it all works fine there. Does using only Agg as the 
> backend make any difference whatsoever?
>
> Paul
>
> ubuntu 7.10
> python 2.5
> matplotlib 0.90.1
>
> ----
> Some sample code;
>
> import matplotlib
> from matplotlib import rc
> rc('font',**{'family':'sans-serif','sans-serif':['Arial','FreeSans']})
> matplotlib.use('Agg')
> from pylab import *
>
> plot(arange(100))
> ax=gca()
> ax.set(xlabel='Useless',ylabel='Pointless')
> draw()
> show()
> savefig('test')
> l=ax.xaxis.get_label()
> print 'font prop: ',l.get_font_properties()
> print 'font name: ',l.get_fontname()
>
> ----
> Produces this output;
>
> font prop: (['Arial', 'FreeSans'], 'normal', 'normal', 'normal', 'normal', 12)
> font name: FreeSans
>
>
>
> -------------------------------------------------------------------------
> This SF.net email is sponsored by the 2008 JavaOne(SM) Conference 
> Don't miss this year's exciting event. There's still time to save 100ドル. 
> Use priority code J8TL2D2. 
> http://ad.doubleclick.net/clk;198757673;13503038;p?http://java.sun.com/javaone
> _______________________________________________
> 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

On Wednesday 23 April 2008 02:49:46 pm Glen W. Mabey wrote:
> Hello,
>
> I'm using today's svn source and I'm surprised that the following loop
> does not get redrawn 10 times.
>
> for it in range( 10 ):
> plot( arange( it ) )
> draw()
> raw_input();
>
> That is, within a 'ipython -pylab' session. Is this really a question
> for the ipython folks? Or am I missing something here?
That also surprises me, but it is not specific to the qt4 backend, nor to 
recent changes in svn. I see similar behavior with the gtkcairo backend in 
0.91.2. Unfortunately, I dont have time to look into it further.
Darren

The clipping rectangle was using inverted y-coordinates (origin at 
bottom), rather than origin at top. This has been fixed in SVN trunk r5067.
FWIW, this seems to be specific to the Wx rendering backend, and doesn't 
happen with Agg, Gtk, Cairo etc.
Cheers,
Mike
Brian Blais wrote:
> On Apr 23, 2008, at Apr 23:9:17 PM, Joshua Lippai wrote:
>
>> I should note that I'm doing this on OS X 10.5.2, but I don't imagine
>> Tiger would have a problem Leopard doesn't have.
>
> you do indeed have the problem. If you change the subplot(211) to 
> subplot(111) you'll see the proper plot (a sine wave). For some 
> reason, when you try to subplot with 2 rows and 2 columns (and 
> probably more) the line doesn't appear on the plot. what it should 
> look like, and does with .91.2, is attached.
>
>
> bb
>
> -- 
> Brian Blais
> bb...@br... <mailto:bb...@br...>
> http://web.bryant.edu/~bblais <http://web.bryant.edu/%7Ebblais>
>
>
> ------------------------------------------------------------------------
>
> -------------------------------------------------------------------------
> This SF.net email is sponsored by the 2008 JavaOne(SM) Conference 
> Don't miss this year's exciting event. There's still time to save 100ドル. 
> Use priority code J8TL2D2. 
> http://ad.doubleclick.net/clk;198757673;13503038;p?http://java.sun.com/javaone
> ------------------------------------------------------------------------
>
> _______________________________________________
> 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

It worked! I had to make a few changes, and there's lots of warnings, but it's 
now compiling. I've attached a patch. In backend_agg.cpp, I had to replace an 
'and' with '&&', and I had to replace a few 'round()' calls with your 
'my_round()'. I don't think my_round is directly being included in 
backend_agg.cpp, it's still hanging around from its definition in 
agg_py_path_iterator.h. Don't know if something should be done about that.
I've also attached the full output with all the warnings.
Thanks Michael!
Martin
Michael Droettboom wrote:
> It doesn't appear to exist in VS2008 -- at least I can't find it in the 
> online docs. Maybe someone else can enlighten me.
> 
> In the meantime, I just wrote my own round function and switched to use 
> that. Please update from SVN and let me know how it goes. And again, 
> thank you for your patience.
> 
> Cheers,
> Mike
> 

I'm confused, because I don't see any place where self.canvas.draw is
called in the code1 version. Also, when I resize the figure, the
background region changes, so the plot gets messed up as I have
noticed before with this method. Does anyone know a good way to
recapture a clean background, in particular when the image is resized?
Thanks,
Glenn
On Mon, Apr 21, 2008 at 6:21 AM, John Hunter <jd...@gm...> wrote:
> On Sun, Apr 20, 2008 at 9:13 PM, hjc520070 <jia...@16...> wrote:
> >
> > Thanks for your help. I get it work. However, an interesting thing appears.
> > The following two codes(code 1 and code 2) makes different result??? Only
> > set background in different place. Anyone can tell me why? I am eager to
> > know it.
>
>
> In code2, you copy the background before the canvas is drawn, which is
> wrong. code 1 is the correct approach. Calling ax.plot is not enough
> to cause the background to be drawn. You must call fig.canvas.draw
> first.
>
> JDH
>
>
>
> -------------------------------------------------------------------------
> This SF.net email is sponsored by the 2008 JavaOne(SM) Conference
> Don't miss this year's exciting event. There's still time to save 100ドル.
> Use priority code J8TL2D2.
> http://ad.doubleclick.net/clk;198757673;13503038;p?http://java.sun.com/javaone
> _______________________________________________
> Matplotlib-users mailing list
> Mat...@li...
> https://lists.sourceforge.net/lists/listinfo/matplotlib-users
>

Hello Glen,
I'm not sure, but maybe it would help to use
 ax = subplot(111)
and
 ax.plot(arange(it))
instead of plot(arange(it))
or to call a second draw() after plotting.
regards Matthias
On Wednesday 23 April 2008 20:49:46 Glen W. Mabey wrote:
> Hello,
>
> I'm using today's svn source and I'm surprised that the following loop
> does not get redrawn 10 times.
>
> for it in range( 10 ):
> plot( arange( it ) )
> draw()
> raw_input();
>
> That is, within a 'ipython -pylab' session. Is this really a question
> for the ipython folks? Or am I missing something here?
>
> Thank you,
> Glen Mabey
>
> -------------------------------------------------------------------------
> This SF.net email is sponsored by the 2008 JavaOne(SM) Conference
> Don't miss this year's exciting event. There's still time to save 100ドル.
> Use priority code J8TL2D2.
> http://ad.doubleclick.net/clk;198757673;13503038;p?http://java.sun.com/java
>one _______________________________________________
> Matplotlib-users mailing list
> Mat...@li...
> https://lists.sourceforge.net/lists/listinfo/matplotlib-users

Hi all,
I've been trying in vain to get a better font on a plot than the fixed width 
serif one that always appears. I've got lib-freetype6 installed (on Ubuntu 
server), and ran fc-cache after. The .matplotlib/ttfont.cache contains entries 
for the Free* fonts. The fonts themselves are 
in /usr/share/fonts/truetype/freefont. The sample code below would seem (to 
me) to have used FreeSans, but the png has the same bad looking fixed-serif 
font.
When running the matplotlib examples/fonts_demo.py the result is the same font 
all over that .png figure too (difference only in sizes).
I usually work on XP, and it all works fine there. Does using only Agg as the 
backend make any difference whatsoever?
Paul
ubuntu 7.10
python 2.5
matplotlib 0.90.1
----
Some sample code;
import matplotlib
from matplotlib import rc
rc('font',**{'family':'sans-serif','sans-serif':['Arial','FreeSans']})
matplotlib.use('Agg')
from pylab import *
plot(arange(100))
ax=gca()
ax.set(xlabel='Useless',ylabel='Pointless')
draw()
show()
savefig('test')
l=ax.xaxis.get_label()
print 'font prop: ',l.get_font_properties()
print 'font name: ',l.get_fontname()
----
Produces this output;
font prop: (['Arial', 'FreeSans'], 'normal', 'normal', 'normal', 'normal', 12)
font name: FreeSans

On Apr 23, 2008, at Apr 23:9:17 PM, Joshua Lippai wrote:
> I should note that I'm doing this on OS X 10.5.2, but I don't imagine
> Tiger would have a problem Leopard doesn't have.
you do indeed have the problem. If you change the subplot(211) to 
subplot(111) you'll see the proper plot (a sine wave). For some 
reason, when you try to subplot with 2 rows and 2 columns (and 
probably more) the line doesn't appear on the plot. what it should 
look like, and does with .91.2, is attached.
		bb
-- 
Brian Blais
bb...@br...
http://web.bryant.edu/~bblais


I should note that I'm doing this on OS X 10.5.2, but I don't imagine
Tiger would have a problem Leopard doesn't have.
Josh
On Wed, Apr 23, 2008 at 6:16 PM, Joshua Lippai <dis...@gm...> wrote:
> I'm using SVN matplotlib and wxPython 2.8.7.1, and I'm not running
> into the problem. Though I may be interpreting your problem
> incorrectly. I've attached a screen capture of the window that opens
> up when I run your program.
>
> Josh
>
>
>
> On Wed, Apr 23, 2008 at 5:33 PM, Brian Blais <bb...@br...> wrote:
> >
> >
> > On Apr 23, 2008, at Apr 23:6:56 PM, Brian Blais wrote:
> >
> > On Apr 23, 2008, at Apr 23:4:33 PM, Brian Blais wrote:
> >
> > I just upgraded a number of things on my Mac OS X (Tiger) machine, including
> > to the latest version of wx and matplotlib. I found that there is a bug in
> > the display of dynamic plots with subplots. If I change the subplot line in
> > the examples/dynamic_demo_wx.py to:
> >
> >
> > a = self.fig.add_subplot(221)
> >
> > instead of
> >
> >
> > a = self.fig.add_subplot(111)
> >
> > the plot never shows.
> >
> >
> > In the quest to discover the problem, I found that the problem exists in svn
> > (version .98pre) but not in .91.2. Both with version wx-2.8.7.1.
> >
> >
> >
> >
> > bb
> >
> >
> >
> > --
> > Brian Blais
> > bb...@br...
> > http://web.bryant.edu/~bblais
> >
> >
> >
> >
> > -------------------------------------------------------------------------
> > This SF.net email is sponsored by the 2008 JavaOne(SM) Conference
> > Don't miss this year's exciting event. There's still time to save 100ドル.
> > Use priority code J8TL2D2.
> >
> > http://ad.doubleclick.net/clk;198757673;13503038;p?http://java.sun.com/javaone
> > _______________________________________________
> > Matplotlib-users mailing list
> > Mat...@li...
> > https://lists.sourceforge.net/lists/listinfo/matplotlib-users
> >
> >
>