[IPython-dev] Qt api selection re. ipython and matplotlib

Evan Patterson epatters at enthought.com
Wed Jul 6 11:23:28 EDT 2011


[snip]

> Darren et al.,
> 
> Is all this good from your mpl/qt standpoint?  We could make mpl pay 
> attention to the ETS QT_API environment variable, if present, while 
> retaining the rcParam.  For example, QT_API could simply override the 
> rcParam, with the additional effect of forcing v2 in the PyQt4 case.  In 
> other words, if QT_API is defined, we would follow the original ETS 
> strategy; otherwise the rcParams strategy as in my last mpl pull 
> request.  I think that although this is a bit more complicated, and 
> would require one more iteration on the ipython side as well as on the 
> mpl side, it might last longer as a way to handle just about all 
> situations, including mpl mixed with ETS things.
> 
> Regarding PySide: I actually can't get it to work in interactive mode 
> (that is, display a plot) even from the straight python prompt; it is 
> working in interactive mode at present only in the ipython qtconsole. 
> As of April, PySide apparently did not use the input hook:
> 
> http://lists.pyside.org/pipermail/pyside/2011-April/002362.html
> 
> Google didn't turn up anything more recent.  So how is it that pylab 
> with PySide is working in qtconsole?
> 
> Eric

PySide works in the qtconsole because the ZMQ kernel actually starts the event loop, unlike the single process IPython shell. (It can do this, of course, because it does not need to prompt the user for any input.)

Evan


More information about the IPython-dev mailing list