Repository navigation
Add more informative error message when pl.render_shapes color table columnhas a missmatch of length #316
Description
I ran into the following:
sc.pp.filter_cells(sdata.tables["table"],min_counts=10)
sdata.pl.render_shapes("cellpose_boundaries",color="total_counts").pl.show()
File /work/rwth1209/enviroments/spatial_data/lib/python3.9/site-packages/spatialdata_plot/pl/utils.py:264, in _get_collection_shape.<locals>.assign_fill_and_outline_to_row(shapes, fill_c, outline_c, row, idx)
262 row["outline_c"] = outline_c
263 else:
--> 264 row["fill_c"] = fill_c[idx]
265 row["outline_c"] = outline_c[idx]
IndexError: index 3424 is out of bounds for axis 0 with size 3424
sdata.shapes["cellpose_boundaries"] = sdata.shapes["cellpose_boundaries"].loc[sdata.tables["table"].obs.index]
sdata.pl.render_shapes("cellpose_boundaries",color="total_counts").pl.show()
fixed!
I removed some cells from the table but their polygons remained. When plotting this led to a generic error.
Could be a "me" problem but I think it might be good idea to add a more explicit error message that tells the user what is going on. An alternative might be to plot the missmatched polygons with a NaN color
Activity
[-]Add more informative error message when pl.render_shapes color table category has a missmatch of length[/-][+]Add more informative error message when pl.render_shapes color table columnhas a missmatch of length[/+]timtreis commentedon Aug 15, 2024
Hey @pakiessling, I'd actually say that in this case it's a "you" problem 😅 You're directly passing the
AnnDataobject, soscanpycannot really know it's dealing with aspatialdataobject andspatialdatawould have to do regular consistency checks or sth to notice that you modified the containedAnnData. I added a more explicit error message though.timtreis commentedon Aug 15, 2024
@flying-sheep are there any plans for
scanpyto natively work withSpatialDataobjects? Then this could happen automagicallytimtreis commentedon Aug 15, 2024
@LucaMarconato do we have a similar way of filtering in
spatialdata?LucaMarconato commentedon Aug 15, 2024
Yes, we have the user-friendly match_element_to_table() and match_table_to_element() functions, which are subcases of the more powerful API join_spatialelement_table().
LucaMarconato commentedon Aug 15, 2024
Anyway, I think that the bug reported may be related to this other bug: #169 and not fixable by the user itself. I'll write a MRE in the linked PR.
LucaMarconato commentedon Aug 15, 2024
Same bug, continuing the discussion there. Workaround: manually filtering the shapes so that there are no shapes not in the table.