Drawing shapes in napari_spatialdata results in Z-Polygons that cannot be plotted with pl.render_shapes

[pakiessling]

sdata.shapes["Test_shapes"].iloc[0]
> geometry    POLYGON Z ((477.545 825.718 19.000, 433.553 82....)
sdata.pl.render_shapes("Test_shapes").pl.show()
>ValueError: 'vertices' must be 2D with shape (N, 2), but your input has shape (19, 3)
-----
anndata             0.10.6
geopandas           0.14.3
matplotlib          3.8.3
napari_spatialdata  0.3.1
pandas              2.2.1
session_info        1.0.0
shapely             2.0.3
spatialdata         0.0.16
spatialdata_plot    0.1.0

Heya,
Is this expected behavior?

Ty!

[LucaMarconato]

Hi @pakiessling, yes our data model currently assumes shapes (circles, polygons/multiploygons) to be 2D. I will add a check on the parser and validator in ShapesModel and explain this in a new tutorial notebook that I am currently preparing. I suggest to remove the z component. Please try and let me know if it works for you.

If you created the polygon as a napari-spatialdata annotation and you found that it included the z component by default please let as know. We had this problem in the past but it seemed to be fixed; I could look at it again.

[pakiessling]

I created the problematic polygon in napari-spatialdata.
It came out like this after shift+e.
You mention "napari annotation". I created a new shape layer in napari but there seems to also be the option to create an annotation layer.
Is that different?

[LucaMarconato]

Thanks, I'll look into this. I suggest to remove the z component, replace the shapes object in-memory and proceed with the analysis.

I will fix the napari behaviour; tracker here: scverse/napari-spatialdata#218.

With "napari annotation" I was referring to the shape layer.

[pakiessling]

Perfect, thank you!

[LucaMarconato]

I fixed it with scverse/napari-spatialdata#228, now napari will only save 2D points and 2D shapes 😊