Is this correct or is the first error message incorrect? If this is right it’d be nice to have a short test.

The function signature is:

class FigureCanvasQT(QtWidgets.QWidget, FigureCanvasBase):

    # map Qt button codes to MouseEvent's ones:
    buttond = {QtCore.Qt.LeftButton: 1,
               QtCore.Qt.MidButton: 2,
               QtCore.Qt.RightButton: 3,
               # QtCore.Qt.XButton1: None,
               # QtCore.Qt.XButton2: None,
               }

    @_allow_super_init
    def __init__(self, figure):

which doesn't force kwargs

I copy pasted the error messages I got from the terminal. It definitely took me longer than I wanted to find what was wrong in my code.

Where do your tests go?

Tests go in lib/matplotlib/tests... But I'm not familiar enough with the low-level backend APIs to know exqctly where or how you should test this. However, if something this basic got through without tripping a test, then either its being called incorrectly and our documentation needs fixing, or it needs a test!

Yeah.... qt5agg uses a different code path that doesn't have the error. I added a test, but it doesn't cover the error since it is using the Agg backend.

I think the test should really go to test_backend_qt4agg (as there's no problem with qt5)?... which is not run on CI both because we can't install pyqt4 there and because we can't import pyqt4 and pyqt5 in the same process, but that's another can of worms.

Otherwise looks good. Doesn't seem to warrant backporting though.

Qt5 is not the same codepath as Qt5Agg.

The problem is with Qt5 (I'm using PySide, but the code change I made is pretty clear that it would happen with PyQt5 as well Edit: I guess not, it doesn't happen on PyQt5. I may have tried it in a different environment too, not too sure).

The bug was in lib/matplotlib/backends/backend_qt5.py. I'm not sure if it happens in Qt4. I'm pretty sure Qt4 is the same code path as Qt5 these days. I'm really not a matplotlib expert, this one was just easy to fix. I felt it would have taken longer to write the "Issue"

Now I'm confused by how exactly you got the error. "qt5" is not a backend by itself, qt5agg and qt5cairo are.

I don't see backends_qt5 importing backends_qt5Agg...

backends_qt5 has a different FigureCanvas object
https://github.com/matplotlib/matplotlib/blob/master/lib/matplotlib/backends/backend_qt5.py#L214
called FigureCanvasQT

The bug is in the decorator for FigureCanvasQT.

FigureCanvasQTAgg is distinct
https://github.com/matplotlib/matplotlib/blob/master/lib/matplotlib/backends/backend_qt5agg.py#L17

But I don't see any tests for the QT backend, only the QTAgg, so I don't really have anything to base myself on. And I'm just learning how to use Qt, so I'm no expert.

I'm still confused; on master, v2.2.2, and v2.2.3 , without this PR, I can run

from matplotlib.backends.backend_qt4agg import FigureCanvas
from matplotlib.figure import Figure

FigureCanvas(Figure())
FigureCanvas(figure=Figure())

without the errors in your original message for either qt4 or qt5. Can you specify what code triggers the error this PR is supposed to fix?

I don't know the history well enough, but I think backend_qt5 is a convenience abstraction of methods shared between the backend_qt5agg and backend_qt5cairo backends, and should not be used by itself. Why it was not made private is above my paygrade, but I don't think its meant to be used directly like this....

Perhaps a better explanation: https://stackoverflow.com/questions/47271291/matplotlib-backends-backend-qt5-vs-from-matplotlib-backends-backend-qt5agg

hmm ok, maybe that is why I had to add the navbar for things to draw.

that said, it is still a bug. the decorator should also pass *args.

I can't repro your bug...

In [1]: matplotlib.backends.backend_qt5.FigureCanvas(matplotlib.figure.Figure())
Out[1]: <matplotlib.backends.backend_qt5.FigureCanvasQT at 0x7f187ee094c8>

(regardless of the fact that this class shouldn't really be instantiated as noted by @jklymak above).

In fact the super-method is called in FigureCanvasQT.__init__ as

        super().__init__(figure=figure)

not posiitionally so it's unclear how you can get positional arguments to reach the patched super-init.

Maybe because I'm on PySide2....

PyQt5 just passes the original __init__

The bug happens because the decorator only passes the keyword args and drops the args.

Ah, I see.
The fix looks correct, but then the test is absolutely not testing the correct thing, and I would just remove it (not really worth it, the correct thing is testing instantiation of FigureCanvasQT which as stated above is not really useful anyways, etc.).

Ok i removed the tests.

Thanks for the persistence @hmaarrfk ! Sorry this was a confusing one...

Yeah it was. I'll try to use Qt5Agg, see how that works.

@meeseeksdev backport to v2.2.x

@meeseeksdev backport to v3.0.x

Owee, I'm MrMeeseeks, Look at me.

There seem to be a conflict, please backport manually. Here are approximate instructions:

1. Checkout backport branch and update it.

$ git checkout v2.2.x
$ git pull

2. Cherry pick the first parent branch of the this PR on top of the older branch:

$ git cherry-pick -m1 1ec1a9c4f9bd50bb5a078e7327024c718318bac6

3. You will likely have some merge/cherry-pick conflict here, fix them and commit:

$ git commit -am 'Backport PR #11900: Allow args to pass through _allow_super_init'

4. Push to a named branch :

git push YOURFORK v2.2.x:auto-backport-of-pr-11900-on-v2.2.x

5. Create a PR against branch v2.2.x, I would have named this PR:

"Backport PR #11900 on branch v2.2.x"

And apply the correct labels and milestones.

Congratulation you did some good work ! Hopefully your backport PR will be tested by the continuous integration and merged soon!

If these instruction are inaccurate, feel free to suggest an improvement.

Looks like this did not get backported to 2.2.x but did make it into 3.0.0, remilestoned to match that.