silx resetting matplotlib Qt binding with Qt6

[malte-storm]

In silx, the matplotlib backend is set to Qt5Agg also for Qt6 bindings:

https://github.com/silx-kit/silx/blob/main/src/silx/gui/utils/matplotlib.py:

# This must be performed before any import from matplotlib
if qt.BINDING in ("PySide6", "PyQt6", "PyQt5"):
    matplotlib.use("Qt5Agg", force=False)
    from matplotlib.backends.backend_qt5agg import FigureCanvasQTAgg  # noqa

with similar code in fabio https://github.com/silx-kit/fabio/blob/main/src/fabio/qt/matplotlib.py

whereas pyFAI (in https://github.com/silx-kit/pyFAI/blob/main/src/pyFAI/gui/matplotlib.py) explicitly uses the QtAgg binding:

elif qt.BINDING in ('PyQt6', 'PySide6', 'PyQt5', 'PySide2'):
    _configure('QtAgg', check=_check_matplotlib)
    from matplotlib.backends.backend_qtagg import FigureCanvasQTAgg  # noqa

When executing the simple code with for example QT_API=pyside6:

import silx.gui.utils.matplotlib
import pyFAI.gui.matplotlib

This raises a single warning on the command line

WARNING:pyFAI.gui.matplotlib:Matplotlib already loaded with backend `Qt5Agg`, setting its backend to `QtAgg` may not work!

or two in ipython:

Cannot switch Qt versions for this session; will use qt6.
WARNING:pyFAI.gui.matplotlib:Matplotlib already loaded with backend `Qt5Agg`, setting its backend to `QtAgg` may not work!

I assume the use of the Qt5Agg backend is not deliberate qith Qt6 bindings? If there are no further dependencies, I would be happy to create a pull request for a fix, but I wanted to check whether there is a specific reason for pinning the backend first.

[t20100]

I assume the use of the Qt5Agg backend is not deliberate qith Qt6 bindings?

No this looks like legacy. It works because force=False I guess, but it is not great.

Pull request welcomed!

[t20100]

closed by #4481