matplotlib-devel Mailing List for matplotlib

Thanks again for pointing out the issue (to my embarrassment).
The rev5 patch addresses the issue. I had missed an essential parameter in
pyplot.autogen_docstring. I've tested this new patch, and I'm able to call
help(pyplot.plot) and execute pyplot.plot([1,2]) without any errors. I
believe this corrects the issue.
Please let me know if you encounter any additional issues.
Regards,
Jason
-----Original Message-----
From: Eric Firing [mailto:ef...@ha...] 
Sent: Thursday, 13 August, 2009 21:33
To: Jason R. Coombs
Cc: matplotlib development list
Subject: Re: [matplotlib-devel] kwdoc processing with decorators
Jason,
There is a problem with rev4, running "ipython -pylab":

Hi,
There are a few new features I have committed into the svn. Nothing
significant, but hopefully useful (and fun to play with). Here is a
quick summary with screenshots ( I made a separate page because it
involves number of images).
http://abitofpythonabitofastronomy.blogspot.com/2009/08/few-new-things-in-mpl-svn.html
Give them a try when you have a chance. The api is still experimental
and any suggestion will be welcomed. On the other hand, I needed to
make some minor changes to the existing code. While I hope it does not
break any, let me know if anything odd happens.
Regards,
-JJ

Michael Droettboom wrote:
> I think this is just a vanilla bug that set_clip_on is being ignored for 
> collections. That patch is rather straightforward.
> 
> Other developers: do you agree this should be fixed, or is there a good 
> reason for current behavior that I'm missing?
It certainly looks to me like an annoying bug, so if your patch fixes 
it, great!
Eric
> 
> Cheers,
> Mike
> 
> Index: lib/matplotlib/collections.py
> ===================================================================
> --- lib/matplotlib/collections.py (revision 7486)
> +++ lib/matplotlib/collections.py (working copy)
> @@ -207,8 +207,7 @@
> transform, transOffset, offsets, paths = self._prepare_points()
> 
> gc = renderer.new_gc()
> - gc.set_clip_rectangle(self.get_clip_box())
> - gc.set_clip_path(self.get_clip_path())
> + self._set_gc_clip(gc)
> 
> renderer.draw_path_collection(
> gc, transform.frozen(), paths, self.get_transforms(),
> @@ -1211,8 +1210,7 @@
> transOffset = transOffset.get_affine()
> 
> gc = renderer.new_gc()
> - gc.set_clip_rectangle(self.get_clip_box())
> - gc.set_clip_path(self.get_clip_path())
> + self._set_gc_clip(gc)
> 
> if self._shading == 'gouraud':
> triangles, colors = self.convert_mesh_to_triangles(
> 
> 
> jas...@cr... wrote:
>> On this thread:
>>
>> http://www.mail-archive.com/mat...@li.../msg05383.html
>>
>> clip_on was a suggested way of getting around the clipping that happens 
>> at the edge of a frame. In the Sage project, we are always setting the 
>> limits on the axes via set_xlim and set_ylim. Is there any way to get 
>> the lines and circles that pass across the edge of the usual clip 
>> boundary to still be drawn, even if we have set the xlim and ylim of the 
>> axis?
>>
>> Basically (using an example from the gallery), is there a way to get the 
>> scatter plot circles below not clipped, while still having the y-axis 
>> only go from -2 to 2? If not, is there a way to easily calculate the 
>> protrusion of the various objects, like the circles below, so we know 
>> how much to adjust the axes to just include the circles?
>>
>> import matplotlib.pyplot as plt
>> import numpy as np
>> fig = plt.figure()
>> x = np.linspace(0r,2*np.pi,100r)
>> y = 2*np.sin(x)
>> ax = fig.add_subplot(1,1,1)
>> q=ax.scatter(x,y)
>> ax.set_ylim([-2,2])
>> q.set_clip_on(False)
>> ax.set_clip_on(False)
>> ax.spines['left'].set_position(('outward',10))
>> ax.spines['bottom'].set_position(('outward',10))
>> ax.spines['top'].set_color('none')
>> ax.spines['right'].set_color('none')
>> ax.xaxis.set_ticks_position('bottom')
>> ax.yaxis.set_ticks_position('left')
>> fig.savefig('test.png')
>>
>> Thanks,
>>
>> Jason
>>
>>
>>
>> ------------------------------------------------------------------------------
>> Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day 
>> trial. Simplify your report design, integration and deployment - and focus on 
>> what you do best, core application coding. Discover what's new with 
>> Crystal Reports now. http://p.sf.net/sfu/bobj-july
>> _______________________________________________
>> Matplotlib-devel mailing list
>> Mat...@li...
>> https://lists.sourceforge.net/lists/listinfo/matplotlib-devel
>> 
> 

Oh, so you want _everything_ to work after the patch? GG
Thanks for reporting this. I'll track it down and get it fixed.
-----Original Message-----
From: Eric Firing [mailto:ef...@ha...] 
Sent: Thursday, 13 August, 2009 21:33
To: Jason R. Coombs
Cc: matplotlib development list
Subject: Re: [matplotlib-devel] kwdoc processing with decorators
Jason R. Coombs wrote:
> I'm about to upload a new patch that implements some of the ideas John and
> Darren have sent. Would you mind running the performance tests against
that
> one also? This new change has the potential to increase performance drag.
Jason,
There is a problem with rev4, running "ipython -pylab":
In [1]:plot([1,2])
---------------------------------------------------------------------------
AttributeError Traceback (most recent call last)
/home/efiring/<ipython console> in <module>()
/usr/local/lib/python2.6/dist-packages/matplotlib/docstring.pyc in 
<lambda>(target)
 334 # language" - GVR. I think functools might help when 
Python 2.5
 335 # is required.
--> 336 return lambda target: dedent(copy(source)(target))
 337 from matplotlib import cbook
 338
/usr/local/lib/python2.6/dist-packages/matplotlib/docstring.pyc in 
do_copy(target)
 422 def do_copy(target):
 423 if source.__doc__:
--> 424 target.__doc__ = source.__doc__
 425 return target
 426 return do_copy
AttributeError: 'list' object attribute '__doc__' is read-only
Eric

Eric Firing <ef...@ha...> writes:
> 3) Question: how does this affect startup time? Jouni tested two 
> approaches to his work, and the decorator approach was significantly 
> slower. 
No, actually the difference was barely measurable. The reason I chose
the boilerplate approach of generating source files instead of using
magic wrappers to generate functions at runtime was that I thought the
latter were too magical - difficult to get right, and probably more
difficult to debug when something breaks.
-- 
Jouni K. Seppänen
http://www.iki.fi/jks

I just happened to type getp(gca()) on matplotlib 0.99.0, and the output
looks all garbled:
>>> getp(gca())
/Library/Frameworks/Python.framework/Versions/2.5/lib/python2.5/site-packages/matplotlib/axes.py:1269: DeprecationWarning: use ax.patch instead
 warnings.warn('use ax.patch instead', DeprecationWarning)
 stable = box
 a = 1.0
 or = C
 ated = False
 ct = auto
 scale_on = True
 scalex_on = True
 scaley_on = True
 = Axes(0.125,0.1;0.775x0.8)
 _locator = None
 _bgcolor = w
[...]
It's been a long time since I last tried this, but does anyone have an
idea what changes could have caused this? Could it be related to the
ReST formatting in the docstrings?
Querying single attributes as in getp(gca(), 'yscale') seems to work
fine, it's just this listing of all attributes that seems to be broken.
-- 
Jouni K. Seppänen
http://www.iki.fi/jks

I think this is just a vanilla bug that set_clip_on is being ignored for 
collections. That patch is rather straightforward.
Other developers: do you agree this should be fixed, or is there a good 
reason for current behavior that I'm missing?
Cheers,
Mike
Index: lib/matplotlib/collections.py
===================================================================
--- lib/matplotlib/collections.py (revision 7486)
+++ lib/matplotlib/collections.py (working copy)
@@ -207,8 +207,7 @@
 transform, transOffset, offsets, paths = self._prepare_points()
 gc = renderer.new_gc()
- gc.set_clip_rectangle(self.get_clip_box())
- gc.set_clip_path(self.get_clip_path())
+ self._set_gc_clip(gc)
 renderer.draw_path_collection(
 gc, transform.frozen(), paths, self.get_transforms(),
@@ -1211,8 +1210,7 @@
 transOffset = transOffset.get_affine()
 gc = renderer.new_gc()
- gc.set_clip_rectangle(self.get_clip_box())
- gc.set_clip_path(self.get_clip_path())
+ self._set_gc_clip(gc)
 if self._shading == 'gouraud':
 triangles, colors = self.convert_mesh_to_triangles(
jas...@cr... wrote:
> On this thread:
>
> http://www.mail-archive.com/mat...@li.../msg05383.html
>
> clip_on was a suggested way of getting around the clipping that happens 
> at the edge of a frame. In the Sage project, we are always setting the 
> limits on the axes via set_xlim and set_ylim. Is there any way to get 
> the lines and circles that pass across the edge of the usual clip 
> boundary to still be drawn, even if we have set the xlim and ylim of the 
> axis?
>
> Basically (using an example from the gallery), is there a way to get the 
> scatter plot circles below not clipped, while still having the y-axis 
> only go from -2 to 2? If not, is there a way to easily calculate the 
> protrusion of the various objects, like the circles below, so we know 
> how much to adjust the axes to just include the circles?
>
> import matplotlib.pyplot as plt
> import numpy as np
> fig = plt.figure()
> x = np.linspace(0r,2*np.pi,100r)
> y = 2*np.sin(x)
> ax = fig.add_subplot(1,1,1)
> q=ax.scatter(x,y)
> ax.set_ylim([-2,2])
> q.set_clip_on(False)
> ax.set_clip_on(False)
> ax.spines['left'].set_position(('outward',10))
> ax.spines['bottom'].set_position(('outward',10))
> ax.spines['top'].set_color('none')
> ax.spines['right'].set_color('none')
> ax.xaxis.set_ticks_position('bottom')
> ax.yaxis.set_ticks_position('left')
> fig.savefig('test.png')
>
> Thanks,
>
> Jason
>
>
>
> ------------------------------------------------------------------------------
> Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day 
> trial. Simplify your report design, integration and deployment - and focus on 
> what you do best, core application coding. Discover what's new with 
> Crystal Reports now. http://p.sf.net/sfu/bobj-july
> _______________________________________________
> Matplotlib-devel mailing list
> Mat...@li...
> https://lists.sourceforge.net/lists/listinfo/matplotlib-devel
> 
-- 
Michael Droettboom
Science Software Branch
Operations and Engineering Division
Space Telescope Science Institute
Operated by AURA for NASA

jas...@cr... wrote:
> On this thread:
>
> http://www.mail-archive.com/mat...@li.../msg05383.html
>
> clip_on was a suggested way of getting around the clipping that happens 
> at the edge of a frame. In the Sage project, we are always setting the 
> limits on the axes via set_xlim and set_ylim. Is there any way to get 
> the lines and circles that pass across the edge of the usual clip 
> boundary to still be drawn, even if we have set the xlim and ylim of the 
> axis?
>
> Basically (using an example from the gallery), is there a way to get the 
> scatter plot circles below not clipped, while still having the y-axis 
> only go from -2 to 2? If not, is there a way to easily calculate the 
> protrusion of the various objects, like the circles below, so we know 
> how much to adjust the axes to just include the circles?
> 
Another idea would be to disable clipping specifically for any markers
at xlim[0] <= markerx <= xlim[1].
I'll talk about this with John and other interested parties at the SciPy
conference (will you be there?) and hopefully address this issue during
the sprint, where I'm planning to work on MPL.
-Andrew

It is now stored in the parent Axes object.
 axis.axes.transAxes
Same is true of transData.
Thanks for pointing this out. I will update the docs.
Cheers,
Mike
jas...@cr... wrote:
> In the documentation for the Axis object, I see that there is supposed 
> to be a transAxis attribute. However, when grepping for it in the 
> source, the only place it appears is in the documentation:
>
> grout@tiny:~/sage/local/lib/python2.6/site-packages/matplotlib$ grep -r 
> transAxis *
> axis.py: * :attr:`transAxis` - transform axis coords to display coords
> Binary file axis.pyc matches
>
>
> Has this attribute been removed? This is with version 0.99.0.
>
> Thanks,
>
> Jason
>
>
>
> ------------------------------------------------------------------------------
> Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day 
> trial. Simplify your report design, integration and deployment - and focus on 
> what you do best, core application coding. Discover what's new with 
> Crystal Reports now. http://p.sf.net/sfu/bobj-july
> _______________________________________________
> Matplotlib-devel mailing list
> Mat...@li...
> https://lists.sourceforge.net/lists/listinfo/matplotlib-devel
> 
-- 
Michael Droettboom
Science Software Branch
Operations and Engineering Division
Space Telescope Science Institute
Operated by AURA for NASA

This is a great idea, Eric. I don't think this is a SF bug, as I recall 
having seen this many since I "got here". I've usually just stopped 
looking before the 0.91 release or so. But you're right -- let's clean 
this stuff out.
I've cleaned out a few already, but are there any takers for me to 
assign Mac-specific and Windows-specific bugs? I'm happy to take any 
Linux-specific ones.
Cheers,
Mike
Eric Firing wrote:
> After ignoring it for a long time, I took a look at the Sourceforge bug 
> tracker, and closed a couple bugs. Our bug list is a bit embarrassing; 
> there are 52 open ones going back to 2005. Is any of this an artifact 
> of all the reworking going on at SF? Did some closed bugs get reopened 
> by accident? In any case, it would be nice to get the list of open bugs 
> down to a much smaller number, and ensure that the oldest is not very 
> old. I suspect many of the old ones have long since either been fixed 
> or rendered irrelevant by other changes.
>
> For all reading this list: even if you don't have mpl svn commit access, 
> if you can review a bug and propose a fix, or conclude it is obsolete, 
> or provide some other way of moving us towards getting it closed, your 
> effort would be much appreciated.
>
> Thank you.
>
> Eric
>
> ------------------------------------------------------------------------------
> Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day 
> trial. Simplify your report design, integration and deployment - and focus on 
> what you do best, core application coding. Discover what's new with 
> Crystal Reports now. http://p.sf.net/sfu/bobj-july
> _______________________________________________
> Matplotlib-devel mailing list
> Mat...@li...
> https://lists.sourceforge.net/lists/listinfo/matplotlib-devel
> 
-- 
Michael Droettboom
Science Software Branch
Operations and Engineering Division
Space Telescope Science Institute
Operated by AURA for NASA

In the documentation for the Axis object, I see that there is supposed 
to be a transAxis attribute. However, when grepping for it in the 
source, the only place it appears is in the documentation:
grout@tiny:~/sage/local/lib/python2.6/site-packages/matplotlib$ grep -r 
transAxis *
axis.py: * :attr:`transAxis` - transform axis coords to display coords
Binary file axis.pyc matches
Has this attribute been removed? This is with version 0.99.0.
Thanks,
Jason

I just noticed that the following two functions seem inconsistent:
axes.get_xticklabels has a "minor" argument, which defaults to False
However, axes.get_xticklines has no such argument, so in order to get 
the minor ticklines, one has to go down to the axes.xaxis object. This 
isn't a huge problem, but the inconsistency seems a little odd. Can we 
make a minor argument for axes.get_xticklines?
Thanks,
Jason

On this thread:
http://www.mail-archive.com/mat...@li.../msg05383.html
clip_on was a suggested way of getting around the clipping that happens 
at the edge of a frame. In the Sage project, we are always setting the 
limits on the axes via set_xlim and set_ylim. Is there any way to get 
the lines and circles that pass across the edge of the usual clip 
boundary to still be drawn, even if we have set the xlim and ylim of the 
axis?
Basically (using an example from the gallery), is there a way to get the 
scatter plot circles below not clipped, while still having the y-axis 
only go from -2 to 2? If not, is there a way to easily calculate the 
protrusion of the various objects, like the circles below, so we know 
how much to adjust the axes to just include the circles?
import matplotlib.pyplot as plt
import numpy as np
fig = plt.figure()
x = np.linspace(0r,2*np.pi,100r)
y = 2*np.sin(x)
ax = fig.add_subplot(1,1,1)
q=ax.scatter(x,y)
ax.set_ylim([-2,2])
q.set_clip_on(False)
ax.set_clip_on(False)
ax.spines['left'].set_position(('outward',10))
ax.spines['bottom'].set_position(('outward',10))
ax.spines['top'].set_color('none')
ax.spines['right'].set_color('none')
ax.xaxis.set_ticks_position('bottom')
ax.yaxis.set_ticks_position('left')
fig.savefig('test.png')
Thanks,
Jason

After ignoring it for a long time, I took a look at the Sourceforge bug 
tracker, and closed a couple bugs. Our bug list is a bit embarrassing; 
there are 52 open ones going back to 2005. Is any of this an artifact 
of all the reworking going on at SF? Did some closed bugs get reopened 
by accident? In any case, it would be nice to get the list of open bugs 
down to a much smaller number, and ensure that the oldest is not very 
old. I suspect many of the old ones have long since either been fixed 
or rendered irrelevant by other changes.
For all reading this list: even if you don't have mpl svn commit access, 
if you can review a bug and propose a fix, or conclude it is obsolete, 
or provide some other way of moving us towards getting it closed, your 
effort would be much appreciated.
Thank you.
Eric

Jason R. Coombs wrote:
> I'm about to upload a new patch that implements some of the ideas John and
> Darren have sent. Would you mind running the performance tests against that
> one also? This new change has the potential to increase performance drag.
Jason,
There is a problem with rev4, running "ipython -pylab":
In [1]:plot([1,2])
---------------------------------------------------------------------------
AttributeError Traceback (most recent call last)
/home/efiring/<ipython console> in <module>()
/usr/local/lib/python2.6/dist-packages/matplotlib/docstring.pyc in 
<lambda>(target)
 334 # language" - GVR. I think functools might help when 
Python 2.5
 335 # is required.
--> 336 return lambda target: dedent(copy(source)(target))
 337 from matplotlib import cbook
 338
/usr/local/lib/python2.6/dist-packages/matplotlib/docstring.pyc in 
do_copy(target)
 422 def do_copy(target):
 423 if source.__doc__:
--> 424 target.__doc__ = source.__doc__
 425 return target
 426 return do_copy
AttributeError: 'list' object attribute '__doc__' is read-only
Eric