Add is_widget_backend() to the widget code path - #595

I tried this patch on NixOS with:

python -c 'import matplotlib.pyplot as plt; import scipp as sc; sc.arange("x", 5.0).plot(); plt.show()'

And it simply did nothing... I then tried:

python -c 'import matplotlib.pyplot as plt; import scipp as sc; print(sc.arange("x", 5.0).plot()); plt.show()'

Which printed:

<plopp.backends.matplotlib.figure.StaticFigure object at 0x78cce095af90>

Is this something I do wrong?

I think InteractiveFigure is a bit misleading? Shouldn't this be something like WidgetFigure, as InteractiveFigure subclasses ipywidgets.

I agree. By the same token, maybe StaticFigure is also a bad name? Because if you use the qt backend, it makes a StaticFigure but the figure isn't static at all. Should we call them WidgetFigure and ... something better? MplFigure?

Is this something I do wrong?

I usually did it via ipython, and using the fig.show() method instead of the plt.show(). Example in a ipython console:

import plopp as pp

f = pp.data.data1d().plot()
f.show()

works for me.

For the on-liner python -c, I think it makes the figure but closes it instantly so that I don't see it. Waiting for input at the end shows the figure:

python -c 'import scipp as sc; f = sc.arange("x", 5.0).plot(); f.show(); input()'

I am currently not sure why the matplotlib plt.show() does not work.

Example in a ipython console:

Doesn't work for me - I did created a Python file this time:

import scipp as sc
f = sc.arange("x", 5.0).plot()
f.show()

For the on-liner python -c, I think it makes the figure but closes it instantly so that I don't see it. Waiting for input at the end shows the figure:

Indeed! But this is very peculiar...

I am currently not sure why the matplotlib plt.show() does not work.

Yea for me too.

@MridulS do you want to proceed with the renaming or should I do it?
Would be nice to get this merged soon :-)

@MridulS do you want to proceed with the renaming or should I do it?

Didn't want to add the change to this pr, but feel free to push here and merge :)

I am currently not sure why the matplotlib plt.show() does not work.

Yea for me too.

Are you going to at least document this behavior somewhere? It could be hard to diagnose for future users like me...

For the on-liner python -c, I think it makes the figure but closes it instantly so that I don't see it. Waiting for input at > > the end shows the figure:

Indeed! But this is very peculiar...

You get the same behaviour with Matplotlib. Using plt.show()

python -c "import matplotlib.pyplot as plt; fig, ax = plt.subplots(); plt.show()"

will show the window and wait until you close it, while using fig.show()

python -c "import matplotlib.pyplot as plt; fig, ax = plt.subplots(); fig.show()"

will close the figure instantly so that you cannot see it.

There is not much we can do about this on our side. The only thing left is to figure out why plt.show() doesn't work in the scipp/plopp example.

Doesn't work for me - I did created a Python file this time:

Finally, for the record, putting that code in a file and running python test.py doesn't work for me either (the window is closed immediately).
It only works inside the ipython/python console.

The only thing left is to figure out why plt.show() doesn't work in the scipp/plopp example.

See #607