matplotlib-users Mailing List for matplotlib

I am writing a GUI application in wxPython and using MPL that plots figures,
saves them to image files, and then closes them all from within a thread to
keep the GUI responsive. Using the Enthought 2.3 install I would
sporatically have crashes that were hard to reproduce, but now that I've
upgrade to Enthought 2.5 (Python 2.5 and MPL 0.91.2.0001) it started
throwing exceptions when attempting to close a figure that I don't
understand. I've created an example script that excites the exception. I'm
not sure if the syptoms are a combination of problems caused by my use of
threads but this problem does not occur if threading is removed from the
script. I've read of others experiencing memory leaks and such when closing
figures, but no situations quite like this.

I am working from a desktop in Windows XP and I use the WXAgg backend.

Does anyone have any ideas, or any feedback on my use of threads? Thanks,
Josh

-------------------------------------------------------------------------------------------
#!/usr/bin/python

import wx
import time
from threading import Thread

import matplotlib
matplotlib.interactive(False)
matplotlib.use('WXAgg')
import matplotlib.pylab

class TestThread(Thread):
    def __init__(self, notify_window):
        Thread.__init__(self)

    def run(self):
        while True:  # Continuous loop that plots and destroys a figure
            matplotlib.pylab.figure(1)
            matplotlib.pylab.plot(range(0,5), range(0,5))
            matplotlib.pylab.show()
            time.sleep(1)
            matplotlib.pylab.close(1)  ### ERROR OCCURS HERE

class MainPanel(wx.Panel):
    def __init__(self, *args, **kwargs):
        wx.Panel.__init__(self, *args, **kwargs)
        self.test_thread = None

        # GUI objects and bindings
        self.run_test_button = wx.Button(self, -1, "Run")
        self.top_sizer = wx.BoxSizer(wx.VERTICAL)
        self.top_sizer.Add(self.run_test_button,      0, wx.EXPAND|wx.ALL,
border=2)
        self.SetSizer(self.top_sizer)
        self.Bind(wx.EVT_BUTTON,   self.run_test_button_bind,   id =
self.run_test_button.GetId())

    ### Binding Methods
    def run_test_button_bind(self, event):
        self.test_thread = TestThread(self)
        self.test_thread.start()
        #self.test_thread.run()
        
class MainFrame(wx.Frame):
    def __init__(self, *args, **kwargs):
        mFrame = wx.Frame.__init__(self, *args, **kwargs)
        self.mPanel = MainPanel(self, -1)
    
class mainApp(wx.App):
    def OnInit(self):
        self.frame = MainFrame(None, title="Test", size=wx.Size(100, 100))
        self.frame.Show()
        self.SetTopWindow(self.frame)
        return True

if __name__ == "__main__":
    app = mainApp()
    app.MainLoop()

--------------------------------------------------------------------------------------------------

Exception in thread Thread-1:
Traceback (most recent call last):
  File "C:\Python25\lib\threading.py", line 486, in __bootstrap_inner
    self.run()
  File ".\test.py", line 22, in run
    matplotlib.pylab.close(1)  ### ERROR OCCURS HERE
  File
"C:\python25\lib\site-packages\matplotlib-0.91.2.0001-py2.5-win32.egg\matplotlib\pyplot.py",
line 245, in close
    _pylab_helpers.Gcf.destroy(arg)
  File
"C:\python25\lib\site-packages\matplotlib-0.91.2.0001-py2.5-win32.egg\matplotlib\_pylab_helpers.py",
line 28, in destroy
    figManager.destroy()
  File
"C:\python25\lib\site-packages\matplotlib-0.91.2.0001-py2.5-win32.egg\matplotlib\backends\backend_wx.py",
line 1406, in destroy
    self.frame.Destroy()
  File
"C:\python25\lib\site-packages\matplotlib-0.91.2.0001-py2.5-win32.egg\matplotlib\backends\backend_wx.py",
line 1365, in Destroy
    wxapp.Yield()
  File
"C:\python25\lib\site-packages\wxpython-2.8.4.0.0003_s-py2.5-win32.egg\wx\_core.py",
line 7119, in Yield
    return _core_.PyApp_Yield(*args, **kwargs)
PyAssertionError: C++ assertion "wxThread::IsMain()" failed at
..\..\src\msw\evtloop.cpp(244) in wxEventLoop::Dispatch(): only the main
thread can process Windows messages
-- 
View this message in context: http://www.nabble.com/Problems-plotting-closing-figures-within-a-thread-tp19789227p19789227.html
Sent from the matplotlib - users mailing list archive at Nabble.com.

I have a Python module that I've written with some pylab plotting  
routines in it.  One of them loops over a list of objects, generating  
some plots based on the data associated with each list member in  
turn.  It's meant to be an interactive way to step through the list  
and see what the things look like.

I have the loop set up to wait for user input before going on to the  
next object, using

x = raw_input("press return for next fit: ")

but for some reason, the plot window never updates.  I've tried  
putting both draw() and show() immediately
before the raw_input() line, and neither works.  The plot only seems  
to update when I ctrl-C out of the loop, and otherwise, I just get the  
spinning beachball when I mouseover the plot.

I'm using Matplotlib 0.98.3 on OS X (10.5.5) with the built-in python  
from Apple I think (2.5.1).  I'm calling the plotting routine from  
ipython 0.8.3 using the -pylab option.

The same script previously worked fine under Matplotlib 0.98.1

Other plotting routines from the same module seem to work fine - it's  
only this plot within a loop that's giving me trouble.

Any ideas?

--
Zane Selvans
Amateur Earthling
za...@id...
303/815-6866
http://zaneselvans.org
PGP Key: 55E0815F

Hello,
I am making some plots using the upperlimit keyword in the errorbar 
command as in the examples.
The arrows are drawn correctly but I notice that the exact limit point 
is in the middle of the arrow and not where the arrow starts.
Is it any way to improve it and have the beginning of the arrow 
represent the actual limit measure?
Also is there any way to increase the thickness of the arrow?

Thanks,
Nino

-- 

Antonino Cucchiara
PhD candidate
Department of Astronomy&Astrophysics
Penn State University
website: www.astro.psu.edu/~cucchiara/

Florian Koelling wrote:
> Hello again!
> 
> I think it's not a namespace problem.
> I tried:
> 
> from pylab import plot as pplot
> import pylab as p
> 
> As result I received busted SD Files. I ' ve this problem nearly for

I am not familiar with pybel and I don't know what an SD file is, so I 
have no idea what you mean above.  It would be nice to know where the 
interference is coming from, even if you can work around it by using the 
  OO api or some other suggestion.

First, for plotting capability, don't use pylab, use pyplot:

import matplotlib.pyplot as plt
plt.plot(...)
or
fig = plt.figure()
ax = fig.add_subplot(111)
ax.plot(...)
plt.show()

Many of the examples have been converted to use this recommended idiom.

Pyplot has the plotting part of pylab without dumping numpy into the 
namespace.  I suspect, however, that in your case the result of 
importing matplotlib.pyplot will be the same

Second, do you need plots on the screen, or is it enough to generate 
plot files?  If the latter, see the agg_oo.py example of how to 
eliminate pylab/pyplot completely (watch out for spurious line breaks in 
the following paste-in):

from matplotlib.backends.backend_agg import FigureCanvasAgg as FigureCanvas
from matplotlib.figure import Figure

fig = Figure()
canvas = FigureCanvas(fig)
ax = fig.add_subplot(111)
ax.plot([1,2,3])
ax.set_title('hi mom')
ax.grid(True)
ax.set_xlabel('time')
ax.set_ylabel('volts')
canvas.print_figure('test')

This makes png files. If you want ps or pdf, the method would be the 
same but with the appropriate backend.

Eric

> three months - so I switched to Gnuplot - now recognizing that it is
> doesn't  have
> enough performance to make a large amount of plots.
> 
> (I tried local imports in a function as well -- same result)
> 
> Actually I'm not using any matplotlib code -- just the import statement
> -- and it's enough that all goes to hell.
> 
> With "import matplotlib" I do not have  those  problems - most plotting
> examples are dealing with pylab -- what can I use instead?
> 
> Maybe one of you Harry Plotters  has  an idea!
> 
> Thanks,
> 
> Florian
> 
> 
> Gary Pajer wrote:
>> On Wed, Oct 1, 2008 at 12:34 PM, Florian Koelling
>> <flo...@tu... <mailto:flo...@tu...>> wrote:
>>
>>     Hi Folks!
>>
>>     Is there a way to plot simple x/y Data in matplotlib WITHOUT using the
>>     "import pylab" command? (in all examples I only saw pylab).
>>
>>     I' working with openbabel(pybel) as well and I've got serious problems
>>     whenever I use "from pylab import plot" or similar commands -
>>
>>     (if I do, SDF files aren't read out anymore properly...)
>>
>>
>>     Thanks alot
>>
>>
>> If the problem is "namespace pollution"  (one package overriding names
>> defined earlier) then the following should work:
>>
>> from pylab import plot as pplot  (where "pplot" is an unused name)
>>
>> or
>>
>> import pylab as p
>>
>> then use, p.plot() instead of plot()
>> (this latter method is what I use.)
>>
>> hth,
>> gary

I see that my tcl/tk email made it through this time, but that this 
list has failed to achieve at the modest task given to them.  All of 
you are unwilling, or unable, to rise to the challenge of finding 
tcl/tk.  You leave me no option but to denounce and deprecate you.  I 
demand a full refund!

All tongue in cheek :)

But I would like some guidance.

Dave

Florian Koelling wrote:
> Actually I'm not using any matplotlib code -- just the import statement
> -- and it's enough that all goes to hell.

> With "import matplotlib" I do not have  those  problems

It sounds like you have a problem with whatever back-end is the default. 
Try changing that in your .matplotlibrc file, or try:

import matplotlib
matplotlib.use('Agg') # or 'wxagg', or 'whatever'

import pylab

that should do it.

What back end - are you/do you want to - use?

despite all the pylab examples, the OO inteface has a lot of advantages...

-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...

Florian Koelling wrote:
> Hello again!
>
> I think it's not a namespace problem.
> I tried:
>
> from pylab import plot as pplot
> import pylab as p
>
> As result I received busted SD Files. I ' ve this problem nearly for
> three months - so I switched to Gnuplot - now recognizing that it is
> doesn't  have
> enough performance to make a large amount of plots.
>
> (I tried local imports in a function as well -- same result)
>
> Actually I'm not using any matplotlib code -- just the import statement
> -- and it's enough that all goes to hell.
>
> With "import matplotlib" I do not have  those  problems - most plotting
> examples are dealing with pylab -- what can I use instead?
>
> Maybe one of you Harry Plotters  has  an idea!
>   
As I said before, look in the examples/api directory (available here 
http://matplotlib.sourceforge.net/matplotlib_examples_0.98.3.zip) for 
examples of using the non-pyplot API.

Cheers,
Mike

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

Andrew Charles wrote:
> The latest beta of CDAT (Climate Data Analysis Tools, widely used in
> the atmospheric/oceanographic sciences) is distributed with the
> matplotlib source included.
>
> On OSX, while it seems to compile without a hitch, plotting
> interactively (or calling show() ) with the TkAgg backend results in a
> seg fault. The non-interactive backends have so far worked just fine.
>
> The output from a simple interactive hello world line plot with
> --verbose-helpful set is:
>
> =================
> Executing /opt/cdat/bin/python
> $HOME=/Users/acharles
> CONFIGDIR=/Users/acharles/.matplotlib
> matplotlib data path /opt/cdat/lib/python2.5/site-packages/matplotlib/mpl-data
> loaded rc file /opt/cdat/lib/python2.5/site-packages/matplotlib/mpl-data/matplotlibrc
> matplotlib version 0.98.3
> verbose.level helpful
> interactive is False
> units is False
> platform is darwin
> Using fontManager instance from /Users/acharles/.matplotlib/fontManager.cache
> backend TkAgg version 8.5
> Testing matplotlib
> Segmentation fault
> =================
>
> CDAT builds almost everything from source. I've always found building
> matplotlib for OSX to be tricky, so tend to just install binary eggs
> these days. Nevertheless I have tried a couple of solutions:
>
> 1. setting export PKG_CONFIG_PATH=/usr/X11/lib/pkgconfig
> 2. editing setupext.py to put the system X11 libraries ahead of any
> fink/macports versions
> 3. setting CFLAGS="-Os -arch i386 -arch ppc" LDFLAGS="-Os -arch i386 -arch ppc"
>
> but to no avail. Is there a simple solution that can be applied to the
> script that builds matplotlib?
>
>   

Andrew:   I think I ran into this before.  Correct me if I'm wrong ..

1) CDAT builds a unix version (not a framework build) of python for you, 
and installs itself in the unix python.
2) CDAT links against X11 versions of all the GUI toolkits (including Tk).

I think the segfault happens because you're picking up either the 
headers or libs for the framework version of Tk included with MacOS X.  
You need to use an X11 build.  To fix this, you need to edit the 
matplotlib setupext.py file to make sure that it's not finding the 
framework version of Tk (i.e, make sure the variable tk_framework ends 
up being zero or False).

HTH,

-Jeff

-- 
Jeffrey S. Whitaker         Phone : (303)497-6313
NOAA/OAR/CDC  R/PSD1        FAX   : (303)497-6449
325 Broadway                Boulder, CO, USA 80305-3328

Dean N. Williams wrote:
> Hi Jeff,
>
>     You are correct on both assumptions below. On the Mac, we know how 
> to build CDAT using framework and have done so for various projects. 
> Is there a huge advantage for building CDAT on the Mac using 
> framework?  Perhaps in the future we will do this as the default.
>
> Thanks for the help on this.
>
> Best regards,
>     Dean

Dean:  If you use a framework build, you can use the builtin mac-native 
Tk and Wx.  If you use unix python, you must use X11 versions of the 
toolkits.  Personally, I don't mind that, but a lot of people don't want 
to mess with X11.  I like it because you can run stuff on your mac 
remotely and display windows over an ssh tunnel.

-Jeff
>
> On Oct 2, 2008, at 4:39 AM, Jeff Whitaker wrote:
>
>> Andrew Charles wrote:
>>> The latest beta of CDAT (Climate Data Analysis Tools, widely used in
>>> the atmospheric/oceanographic sciences) is distributed with the
>>> matplotlib source included.
>>>
>>> On OSX, while it seems to compile without a hitch, plotting
>>> interactively (or calling show() ) with the TkAgg backend results in a
>>> seg fault. The non-interactive backends have so far worked just fine.
>>>
>>> The output from a simple interactive hello world line plot with
>>> --verbose-helpful set is:
>>>
>>> =================
>>> Executing /opt/cdat/bin/python
>>> $HOME=/Users/acharles
>>> CONFIGDIR=/Users/acharles/.matplotlib
>>> matplotlib data path 
>>> /opt/cdat/lib/python2.5/site-packages/matplotlib/mpl-data
>>> loaded rc file 
>>> /opt/cdat/lib/python2.5/site-packages/matplotlib/mpl-data/matplotlibrc
>>> matplotlib version 0.98.3
>>> verbose.level helpful
>>> interactive is False
>>> units is False
>>> platform is darwin
>>> Using fontManager instance from 
>>> /Users/acharles/.matplotlib/fontManager.cache
>>> backend TkAgg version 8.5
>>> Testing matplotlib
>>> Segmentation fault
>>> =================
>>>
>>> CDAT builds almost everything from source. I've always found building
>>> matplotlib for OSX to be tricky, so tend to just install binary eggs
>>> these days. Nevertheless I have tried a couple of solutions:
>>>
>>> 1. setting export PKG_CONFIG_PATH=/usr/X11/lib/pkgconfig
>>> 2. editing setupext.py to put the system X11 libraries ahead of any
>>> fink/macports versions
>>> 3. setting CFLAGS="-Os -arch i386 -arch ppc" LDFLAGS="-Os -arch i386 
>>> -arch ppc"
>>>
>>> but to no avail. Is there a simple solution that can be applied to the
>>> script that builds matplotlib?
>>>
>>>
>>
>> Andrew:   I think I ran into this before.  Correct me if I'm wrong ..
>>
>> 1) CDAT builds a unix version (not a framework build) of python for 
>> you, and installs itself in the unix python.
>> 2) CDAT links against X11 versions of all the GUI toolkits (including 
>> Tk).
>>
>> I think the segfault happens because you're picking up either the 
>> headers or libs for the framework version of Tk included with MacOS 
>> X.  You need to use an X11 build.  To fix this, you need to edit the 
>> matplotlib setupext.py file to make sure that it's not finding the 
>> framework version of Tk (i.e, make sure the variable tk_framework 
>> ends up being zero or False).
>>
>> HTH,
>>
>> -Jeff
>>
>> -- 
>> Jeffrey S. Whitaker         Phone : (303)497-6313
>> NOAA/OAR/CDC  R/PSD1        FAX   : (303)497-6449
>> 325 Broadway                Boulder, CO, USA 80305-3328
>>
>


-- 
Jeffrey S. Whitaker         Phone : (303)497-6313
NOAA/OAR/CDC  R/PSD1        FAX   : (303)497-6449
325 Broadway                Boulder, CO, USA 80305-3328

Hello again!

I think it's not a namespace problem.
I tried:

from pylab import plot as pplot
import pylab as p

As result I received busted SD Files. I ' ve this problem nearly for
three months - so I switched to Gnuplot - now recognizing that it is
doesn't  have
enough performance to make a large amount of plots.

(I tried local imports in a function as well -- same result)

Actually I'm not using any matplotlib code -- just the import statement
-- and it's enough that all goes to hell.

With "import matplotlib" I do not have  those  problems - most plotting
examples are dealing with pylab -- what can I use instead?

Maybe one of you Harry Plotters  has  an idea!

Thanks,

Florian


Gary Pajer wrote:
> On Wed, Oct 1, 2008 at 12:34 PM, Florian Koelling
> <flo...@tu... <mailto:flo...@tu...>> wrote:
>
>     Hi Folks!
>
>     Is there a way to plot simple x/y Data in matplotlib WITHOUT using the
>     "import pylab" command? (in all examples I only saw pylab).
>
>     I' working with openbabel(pybel) as well and I've got serious problems
>     whenever I use "from pylab import plot" or similar commands -
>
>     (if I do, SDF files aren't read out anymore properly...)
>
>
>     Thanks alot
>
>
> If the problem is "namespace pollution"  (one package overriding names
> defined earlier) then the following should work:
>
> from pylab import plot as pplot  (where "pplot" is an unused name)
>
> or
>
> import pylab as p
>
> then use, p.plot() instead of plot()
> (this latter method is what I use.)
>
> hth,
> gary
>  
>
>
>
>     -------------------------------------------------------------------------
>     This SF.Net email is sponsored by the Moblin Your Move Developer's
>     challenge
>     Build the coolest Linux based applications with Moblin SDK & win
>     great prizes
>     Grand prize is a trip for two to an Open Source event anywhere in
>     the world
>     http://moblin-contest.org/redirect.php?banner_id=100&url=/
>     <http://moblin-contest.org/redirect.php?banner_id=100&url=/>
>     _______________________________________________
>     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 Moblin Your Move Developer's challenge
> Build the coolest Linux based applications with Moblin SDK & win great prizes
> Grand prize is a trip for two to an Open Source event anywhere in the world
> http://moblin-contest.org/redirect.php?banner_id=100&url=/
> ------------------------------------------------------------------------
>
> _______________________________________________
> Matplotlib-users mailing list
> Mat...@li...
> https://lists.sourceforge.net/lists/listinfo/matplotlib-users
>   

The latest beta of CDAT (Climate Data Analysis Tools, widely used in
the atmospheric/oceanographic sciences) is distributed with the
matplotlib source included.

On OSX, while it seems to compile without a hitch, plotting
interactively (or calling show() ) with the TkAgg backend results in a
seg fault. The non-interactive backends have so far worked just fine.

The output from a simple interactive hello world line plot with
--verbose-helpful set is:

=================
Executing /opt/cdat/bin/python
$HOME=/Users/acharles
CONFIGDIR=/Users/acharles/.matplotlib
matplotlib data path /opt/cdat/lib/python2.5/site-packages/matplotlib/mpl-data
loaded rc file /opt/cdat/lib/python2.5/site-packages/matplotlib/mpl-data/matplotlibrc
matplotlib version 0.98.3
verbose.level helpful
interactive is False
units is False
platform is darwin
Using fontManager instance from /Users/acharles/.matplotlib/fontManager.cache
backend TkAgg version 8.5
Testing matplotlib
Segmentation fault
=================

CDAT builds almost everything from source. I've always found building
matplotlib for OSX to be tricky, so tend to just install binary eggs
these days. Nevertheless I have tried a couple of solutions:

1. setting export PKG_CONFIG_PATH=/usr/X11/lib/pkgconfig
2. editing setupext.py to put the system X11 libraries ahead of any
fink/macports versions
3. setting CFLAGS="-Os -arch i386 -arch ppc" LDFLAGS="-Os -arch i386 -arch ppc"

but to no avail. Is there a simple solution that can be applied to the
script that builds matplotlib?

-------------------------
Andrew Charles