Flexible plotting class for sc.DataArray that maps dimensions to color, marker & linestyle, with a concise legend

[doronbehar]

Very often I find myself dealing with DataArrays with 3 or more dimensions, with sizes such as:

{
    "x": LARGE_INTEGER,
    "alpha": INTEGER_LESS_THEN_10,
    "beta": INTEGER_LESS_THEN_5,
}

And maybe with another dimension of 2-3 values. Plotting a combination of y v.s x for all combinations of alphas and betas, and adding a legend for it, will explode in this case to $\sim 50$ legend entries. The correct thing to do is to put only INTEGER_LESS_THEN_10+INTEGER_LESS_THEN_5 legend entries such as:

seaborn sort of provides this feature: https://seaborn.pydata.org/generated/seaborn.lineplot.html , but it lacks customization of the legend and the error bar computation is too a bit too locked down IMO (Not to mention upstream's ignorance there!). Since scipp has excellent error propagation capabilities, it is easier to wire these to plots with errorbars, while not locking the legend too much.

plopp has a single example here that demonstrates something simple like this, but it seems no plopp function truly implements this mapping like the class I wrote below.

ggplot has this feature, but it locks down the legend too so it seems, not to mention other matplotlib elements it simply sidesteps (example)

Here are a few plot examples:

The code is available here:

https://gitlab.com/doronbehar/lab-ideal-scientific-workflows/-/blob/b71eddc9a59d46445d768134c4cdd1bbabd2d894/SciPlot.py

Disclaimer: I used Gemini to generate that script, but I read and edited thoroughly the class, and added many features completely myself after the initial AI attempt.

Comments and ideas are more then welcome!