Napari visualization of AnnData Visium containing only a single library is broken

[ilia-kats]

Following the documentation and tutorials, code like

image = sq.im.ImageContainer(adata.uns["spatial"]["spaceranger100_count_30458_ST8059048_mm10-3_0_0_premrna"]["images"]["hires"], scale=adata.uns["spatial"]["spaceranger100_count_30458_ST8059048_mm10-3_0_0_premrna"]["scalefactors"]["tissue_hires_scalef"])
viewer = image.interactive(adata)

should work and has worked in the past. However, with Squidpy 1.2.0, it aborts with

KeyError: "Unable to get the spot diameter from `adata.uns['spatial']['0']['scalefactors']['spot_diameter_fullres']]`"

Explicitly passing the library id, i.e.

viewer = image.interactive(adata, library_id="spaceranger100_count_30458_ST8059048_mm10-3_0_0_premrna")

results in the same error. This is due to the passed library ID being overwritten by self.container.library_ids if library_key is None, which would be fine if self.container.library_ids had a sensible value instead of [0]. According to the documentation, library_key is optional and corresponds to the column in .obs containing Z coordinates. A single visium slice does not have such a column.

[michalk8]

Hi @ilia-kats , you can also pass library_id in the __init__, i.e. this should work:

lid = "spaceranger100_count_30458_ST8059048_mm10-3_0_0_premrna"
image = sq.im.ImageContainer(
    adata.uns["spatial"][lid]["images"]["hires"],
    scale=adata.uns["spatial"][lid]["scalefactors"]["tissue_hires_scalef"]
)

viewer = image.interactive(adata, library_id=lid)

Alternatively, there's a function called .from_adata that's slightly more convenient:

img = sq.im.ImageContainer.from_adata(adata,
    img_key="hires",
    library_id="spaceranger100_count_30458_ST8059048_mm10-3_0_0_premrna"
)

when passing the library_id.

[michalk8]

This PR might be relevant #531 (adata subsetting bug when interactively visualizing it)

[ilia-kats]

Ah, thanks, I didn't know about from_adata, this works. The docs could definitely use some work though, neither library_id nor **kwargs are documented in the ImageContainer constructor, and the docs for ImageContainer.interactive say

library_id (Union[str, Sequence[str], None]) – Subset of library ids to visualize. If None, visualize all library ids.

which is obviously not what's happening.

[michalk8]

**kwargs are documented in the ImageContainer constructor, and the docs for ImageContainer.interactive

There was a typo (an extra whitespace caused the add_img not to be included) that, fixed in #531 .

[giovp]

should be working at scverse/napari-spatialdata