WIP/ENH: pyplot internals #906
Conversation
|
Invoking %pylab inline resolves our block issue. |
|
travis is not happy: File "/home/travis/virtualenv/python2.7_with_system_site_packages/lib/python2.7/site-packages/mne-0.7.git-py2.7.egg/mne/viz.py", line 1241, in plot_evoked |
|
This shows us that we don't control whether pyplot is really used. The pyplot figure hase a show method, the pylab hasn't ... |
|
Here's a helpful discussion of the problem. Will look into it tonight. |
|
@agramfort WDYT about my last commit. |
| if show: | ||
| plt.show() | ||
| if show and plt.get_backend() != 'agg': | ||
| fig.show() |
There was a problem hiding this comment.
how about if
hasattr(fig, 'show')
?
would that work?
There was a problem hiding this comment.
I think it's almost equivalent ...
The reasons leading to fig objects without show method are backend-related. E.g. 'Agg' as used in tests will produce such objects while plt.show calls would just pass and do nothing.
The variant I proposed would allow us to learn about additional backends / constellations where this fails ...
But I'm fine with going for attribute checking.
|
Point taken |
In part addresses #903
probably more examples need to be updated.
This also addresses the matplotlib issues with Anaconda (should now work without self compiled matplotlib) reported in #728 by @kambysese.
We're now able to understand the reasons and extract two insigths:
In addition many inconveniences can be avoided by explicitly invoking %pylab inline in a notebook.
This should be added to the docs.
You can checkout this notebook for exploring:
http://nbviewer.ipython.org/7533330
The multi-axis figures now work in the notebook.
The block thing still needs a fix.