matplotlib-users Mailing List for matplotlib

Following the advice to ping you on the mailing list, I've submitted a 
suggestion for a patch to the EMF Backend. I've copied the contents 
here for convenience.
Patch ID is: 3232997
Content is:
With Matplotlib 1.0.1, When plotting a patch (e.g. text with a bbox with 
facecolor), the facecolor of the patch is passed to the backend as RGBa. 
EMFBrush can't handle RGBa and dies with the following:
File "/usr/local/lib/python2.7/dist-packages/matplotlib/patches.py", 
line 411, in draw
renderer.draw_path(gc, tpath, affine, rgbFace)
File 
"/usr/local/lib/python2.7/dist-packages/matplotlib/backends/backend_emf.py", 
line 260, in draw_path
self.select_brush(rgbFace)
File 
"/usr/local/lib/python2.7/dist-packages/matplotlib/backends/backend_emf.py", 
line 567, in select_brush
brush=EMFBrush(self.emf,rgb)
File 
"/usr/local/lib/python2.7/dist-packages/matplotlib/backends/backend_emf.py", 
line 107, in __init__
r,g,b=rgb
ValueError: too many values to unpack
Suggested patch is to change line 107 in backend_emf.py from r,g,b=rgb to:
r,g,b=rgb[:3]
which will drop the alpha channel. Given the EMF backend doesn't support 
alpha channels, it shouldn't cause any problems to do this silently.

Is there a way to set the style and scilimits to the colorbar axes? All my
attempts failed. For example, run the following script, I get the error
shown below:
### begin example script
import numpy
from matplotlib import pyplot
n = 5000000
x = numpy.random.standard_normal(n)
y = 2.0 + 3.0 * x + 4.0 * numpy.random.standard_normal(n)
xmin = x.min()
xmax = x.max()
ymin = y.min()
ymax = y.max()
hist, edges = numpy.histogramdd([y,x], bins=[25,25],
 range=[[ymin,ymax], [xmin,xmax]])
extent = [xmin, xmax, ymin, ymax]
fig = pyplot.figure()
ax = fig.add_subplot(1,1,1)
plt = ax.imshow(hist,
 extent = extent,
 origin = 'lower',
 interpolation = 'nearest',
 aspect = 'auto')
cb = fig.colorbar(plt, ax=ax)
# This causes an AttributeError exception
cb.ax.ticklabel_format(style='sci', scilimits=(0,4))
pyplot.show()
### end example script
> python cb_scilim_test.py
Traceback (most recent call last):
 File "cb_scilim_test.py", line 25, in <module>
 cb.ax.ticklabel_format(style='sci', scilimits=(0,4))
 File "/usr/lib64/python2.7/site-packages/matplotlib/axes.py", line 2117,
in ticklabel_format
 "This method only works with the ScalarFormatter.")
AttributeError: This method only works with the ScalarFormatter.
-- 
View this message in context: http://old.nabble.com/colorbar-and-scilimits-tp31201133p31201133.html
Sent from the matplotlib - users mailing list archive at Nabble.com.

--- On Sun, 3/20/11, Daniel Welling <dan...@gm...> wrote:
> The OSX backend used to have a bug where you cannot type a
> name in the file name text box.Since that has been fixed, ...
That was not a bug in the MacOSX backend (and therefore was not fixed), but is related to how Python is installed on your system: If your Python is not a framework installation, it will not interact correctly with Apple's windowing manager. This is due to OS X itself and is independent of the MacOSX backend.
I
 have found a new bug: every so often, when you go to save a file, the 
"save" and "cancel" button stop responding, trapping the user in file 
saving limbo. I'll have to play with it again to figure out what 
triggers this.
Can you check if your Python is built as a framework? If it is, MacOS.WMAvailable() should return True:
$ python
Python 2.7.1 (r271:86832, Mar 12 2011, 13:44:53) 
[GCC 4.0.1 (Apple Computer, Inc. build 5370)] on darwin
Type "help", "copyright", "credits" or "license" for more information.
>>> import MacOS
>>> MacOS.WMAvailable()
True
>>> 
Best,
--Michiel.
 

I've been visualising some data using the pyplot fill function, and whilst
overlapping polygons do mix their colours and allow for any overlaps to be
apparent, I thought that perhaps this would look much better in a heatmap
format to better show high levels of change in the same areas.
I've tested this using the histogram2d function in numpy using various
online examples, and whilst experiments with random data looked nice enough,
each set of results data is basically a series of 0's and 1's: In a heatmap
using the same code, this results in an incredibly long (since there's about
65000 results) heatmap that's only a pixel or two in height. Also, I can't
seem to use the tools to zoom in and navigate around this, which is pretty
much a requirement given the large number of data rows.
Does anyone know if there's a way to convert a series of polygon objects
into a decent looking heatmap? Any help is appreciated!
Cheers!
*Rohaq*

Hello, Michiel.
The OSX backend used to have a bug where you cannot type a name in the file
name text box. Since that has been fixed, I have found a new bug: every so
often, when you go to save a file, the "save" and "cancel" button stop
responding, trapping the user in file saving limbo. I'll have to play with
it again to figure out what triggers this. Other than that, the OSX backend
is light and quick, which is nice. When QT4 is working, however, I prefer
that. Sorry I can't give you more details; I'll see if I can get a clearer
picture.
-dw
On Sun, Mar 20, 2011 at 1:31 AM, Michiel de Hoon <mjl...@ya...>wrote:
> --- On *Wed, 3/16/11, Daniel Welling <dan...@gm...>* wrote:
>
> After playing with backends quite a bit, I have found that the best one in
> terms of speed, robustness, and features is Qt4Agg - especially on OSX,
> where the MacOSX backend is buggy and many others just don't plain work.
>
> Why do you think that the MacOSX backend is buggy?
>
> --Michiel.
>
>
>
>
>
>

On Sun, Mar 20, 2011 at 8:52 PM, Xavier Gnata <xav...@gm...> wrote:
> On 03/20/2011 07:19 PM, Darren Dale wrote:
>>
>> On Sun, Mar 20, 2011 at 2:08 PM, Xavier Gnata<xav...@gm...>
>> wrote:
>>>
>>> Hi,
>>>
>>> It looks like the
>>> http://matplotlib.svn.sourceforge.net/viewvc/matplotlib/branches/py3k/
>>> in dead.
>>> The last commit was 8months ago.
>>>
>>> Numpy is ok with python3.1, scipy is ok, nose is ok, ipython is usable.
>>> Ubuntu already provides python-tk for python3.x
>>> It would be nice to port matplotlib to python3.
>>> Is there a plan? another svn/git branch?
>>
>> https://github.com/matplotlib/matplotlib-py3
>
> Thanks!
> AFAICS it works already well with the tk backend :)
> Are you interested in more testing with another backend?
Mike D. put a wiki page together summarizing where things stand and
listing outstanding issues:
https://github.com/matplotlib/matplotlib-py3/wiki

On 03/20/2011 07:19 PM, Darren Dale wrote:
> On Sun, Mar 20, 2011 at 2:08 PM, Xavier Gnata<xav...@gm...> wrote:
>> Hi,
>>
>> It looks like the
>> http://matplotlib.svn.sourceforge.net/viewvc/matplotlib/branches/py3k/
>> in dead.
>> The last commit was 8months ago.
>>
>> Numpy is ok with python3.1, scipy is ok, nose is ok, ipython is usable.
>> Ubuntu already provides python-tk for python3.x
>> It would be nice to port matplotlib to python3.
>> Is there a plan? another svn/git branch?
> https://github.com/matplotlib/matplotlib-py3
Thanks!
AFAICS it works already well with the tk backend :)
Are you interested in more testing with another backend?
Xavier