Showing legend by defaults and allowing richer customizability

[LucaMarconato]

From @MeyerBender:

• legends: in my opinion, it would be great to show legends for the channels by default. Alternatively, one could consider implementing something like pl.render_legend() to give full customizability to the user.

This is an example of plot from spatialproteomics. Notebook here.

and the same plot from spatialdata. Notebook here

[MeyerBender]

Example code snippet.

import spatialdata as sd
import spatialdata_plot
import matplotlib.pyplot as plt
import numpy as np

# example data
image = sd.models.Image2DModel.parse(
    np.random.rand(4, 100, 100), transformations=None, dims=("c", "x", "y"), c_coords=["DAPI", "CD4", "CD8", "FOXP3"]
)
sdata = sd.SpatialData(images={"image": image})

fig, ax = plt.subplots(1, 2)

# plotting a single channel
sdata.pl.render_images(channel='DAPI').pl.show(ax=ax[0])

# plotting multiple channels
sdata.pl.render_images(channel=['CD4', 'CD8', 'FOXP3']).pl.show(ax=ax[1])

Currently returns this:

[timtreis]

Hey @MeyerBender, does #576 address your feature request?

[MeyerBender]

Hi @timtreis, yes, this is looking really good! One minor thing, I think if you call it channels_as_categories, it needs to be really clear what it does in the docs. Wondering if calling it something like legend or channel_legend would make more sense, but that might be confusing with the colorbar argument for single-channel images...

[timtreis]

How about channels_as_legend?

[MeyerBender]

Yeah that works :)