Datashader support for points and shapes

[LucaMarconato]

Work from @Sonja-Stockhaus and @ckmah.

[LucaMarconato]

@Sonja-Stockhaus recently we introduced datashader support in spatialdata for points and shapes and you may find it useful to reuse some components.

In particular, you may find it useful to reuse the code in https://github.com/scverse/spatialdata/blob/b8fbc5c4ce7643455768a94665496ea4987294dd/src/spatialdata/_core/operations/rasterize.py#L604.

Technically, I would consider isolating the code used to compute cnv = ds.Canvas(...) into a private function of rasterize.py in spatialdata, and then I would use that private function from spatialdata-plot (using rasterize_shapes_points() directly may not be useful for spatialdata-plot since you probably want more control on the shading operation, so reusing some intermediate code by placing it on private functions may be the way to go).

Please let me know what you think about this. Also, if you prefer to merge this PR as it is and then eventually reuse the components from rasterize.py in a follow up refactoring, that would also be perfectly fine.

[Sonja-Stockhaus]

@Sonja-Stockhaus recently we introduced datashader support in spatialdata for points and shapes and you may find it useful to reuse some components.

In particular, you may find it useful to reuse the code in scverse/spatialdata@b8fbc5c/src/spatialdata/_core/operations/rasterize.py#L604.

Technically, I would consider isolating the code used to compute cnv = ds.Canvas(...) into a private function of rasterize.py in spatialdata, and then I would use that private function from spatialdata-plot (using rasterize_shapes_points() directly may not be useful for spatialdata-plot since you probably want more control on the shading operation, so reusing some intermediate code by placing it on private functions may be the way to go).

Please let me know what you think about this. Also, if you prefer to merge this PR as it is and then eventually reuse the components from rasterize.py in a follow up refactoring, that would also be perfectly fine.

Thanks for the heads up! I think fo now it would be best to finish and merge this PR and then do the refactoring later, since it's already so big.
Also regarding that @LucaMarconato can you tell me when the next pip release of spatialdata will be? Since this code is currently using some of the newer changes and that's why the tests keep failing right now

[LucaMarconato]

Your message comes at the right time 😊 I just made a new release 10 minutes ago!

I agree, let's rather merge this and refactor later.

[codecov-commenter]

Codecov Report

Attention: Patch coverage is 91.72932% with 11 lines in your changes missing coverage. Please review.

Please upload report for BASE (main@80dfe4b). Learn more about missing BASE report.
Report is 11 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #244   +/-   ##
=======================================
  Coverage        ?   80.19%           
=======================================
  Files           ?       11           
  Lines           ?     1737           
  Branches        ?        0           
=======================================
  Hits            ?     1393           
  Misses          ?      344           
  Partials        ?        0           

Files	Coverage Δ
src/spatialdata_plot/pl/render_params.py	100.00% <100.00%> (ø)
src/spatialdata_plot/pl/utils.py	69.76% <75.00%> (ø)
src/spatialdata_plot/pl/basic.py	90.90% <60.00%> (ø)
src/spatialdata_plot/pl/render.py	93.98% <93.22%> (ø)

[timtreis]

@Sonja-Stockhaus is this ready for review?

[timtreis]

Already looks fairly good! Can we also add examples to your speed improvement notebook to show these?

[timtreis]

Why a different one? Is 150 bad?

[Sonja-Stockhaus]

Not necessarily, but it's an arbitrary value that I chose to make everything better visible...