Yes please, but it will need a few more modifications.
if we are going to us qtpy then we should delete our entire compatibility shim and add it as a hard dependency
I am not a fan of the if statement in the init, either we should fully merge the qt4/qt5 code bases or run the qt5 + pyside2 through the qt4 code paths (which already account for the non-cooperative super classes
Is there a reasonable path where we have a shim backend called 'qtagg' which imports from qt_compat and then decides if it needs to use the cooperative or non-cooperative versions of the classes?
It is really disappointing that pyside2 is not cooperative 😞 .
Activity
tacaswell commented on May 15, 2017
Is pyside2 not cooperative? If not we need to run the it through the qt4Agg code path instead of directly through the qt5 code path.
t-bltg commented on May 20, 2017
I have made progress and was able to build matplotlib with qt4/pyside and qt5/pyside2
Here is the patch: t-bltg@13a5d5b
I used qtpy from spyder as proposed in #6118
I ran the test suite python -c 'import matplotlib; matplotlib.test()', and it passed without issues on this part.
Should I make a pull request ?
tacaswell commented on May 22, 2017
Yes please, but it will need a few more modifications.
This probably crosses paths with #8613
Is there a reasonable path where we have a shim backend called 'qtagg' which imports from qt_compat and then decides if it needs to use the cooperative or non-cooperative versions of the classes?
It is really disappointing that pyside2 is not cooperative 😞 .
anntzer commented on Aug 26, 2017
Closed by #9052.