Breakage when changing categorical data underpinning the color attribute

[hamelin]

On version 0.22.0 freshly installed off of PyPI, and using Python 3.13, I have found that, when coloring by categorical data and using my own color map, changing the data frame so that the categorical values are updated raises an exception. This may be related to issues #191 and #189 involving the mapping of categorical data to colormap entries. Please find attached an example notebook. I had to change its extension to .json to have Github accept linking it here.

bug_colormaps_jupyter_scatter.json

[hamelin]

I have found a workaround to this issue:

1. Use plot.color to change the plot coloring to another column.
2. Change the data frame using plot.data.
3. Use plot.color once again to set coloring to the categorical column, yielding an opportunity to set a new color map (which works around Unused categories leads to wrong color and legend #189).

This is annoying, however, as in the best case, the user sees a flicker in the plot coloring. In the worst case, the messaging system is too fast, and the second color change does not trigger a repainting of the plot, so the transient other-column color remains. Re-showing the plot paints it correctly, but this should not be necessary.

[flekschas]

Sorry about that! I got a little side tracked since the last update. I'll look into the issue and try to get it fixed quickly. Thanks for sharing the notebook!