Using Alt+Mousemove over an image used to update plot cross sections but is now broken

[RussBerg]

did a quick check and found the issue, XCrossSectionPlot and YCrossSectionPlot do not contain connect_plot functions,
I tested by adding the following to both classes and the behaviour was restored:

    def connect_plot(self, plot: BasePlot) -> None:
        """Connect plot to cross section plot"""
        XYCrossSectionMixin.connect_plot(self, plot)

PlotPy/plotpy/panels/csection/csplot.py

Line 439 in c5c1213

class XCrossSectionPlot(HorizontalCrossSectionPlot, XYCrossSectionMixin):

PlotPy/plotpy/panels/csection/csplot.py

Line 453 in c5c1213

class YCrossSectionPlot(VerticalCrossSectionPlot, XYCrossSectionMixin):

I can't create a branch on the repo otherwise I would have created a PR

[PierreRaybaut]

Thanks for the detailed diagnosis, it pointed straight at the root cause! You don't need write access to open a PR: just fork the repository to your own GitHub account, push your branch there, and open a pull request from your fork against PlotPyStack/PlotPy:master. That's the standard contribution flow and it's very welcome.

Regarding the fix itself: rather than duplicating connect_plot into both XCrossSectionPlot and YCrossSectionPlot, the underlying problem is an MRO ordering issue. The classes were declared as class XCrossSectionPlot(HorizontalCrossSectionPlot, XYCrossSectionMixin), so BaseCrossSectionPlot.connect_plot (inherited via HorizontalCrossSectionPlot) shadowed the mixin's connect_plot - the one that wires up SIG_MARKER_CHANGED and drives the Alt+Mousemove cross-section update. Putting the mixin first (class XCrossSectionPlot(XYCrossSectionMixin, HorizontalCrossSectionPlot)]) restores the correct precedence with no duplication. I've applied that fix.

This regression was introduced in a522dcb ("Cross section plots: fixed historic flaw"), first released in v2.1.0 (Feb 8, 2024).

[PierreRaybaut]

@RussBerg, can you please confirm that this fix works for you?

[RussBerg]

@RussBerg, can you please confirm that this fix works for you?

cool, yes I will test now and report back here

[RussBerg]

@RussBerg, can you please confirm that this fix works for you?

Yes this works now, thanks!

[PierreRaybaut]

Thanks again for the feedback!

[RussBerg]

My Pleasure, I have a project that dates back at least 15 years that uses plotpy (started with guiqwt), I had not updated the conda env I was using to the latest plotpy until last week (#shame) so that's when I noticed it.