Repository navigation
_robust_transform() after cropping gives problems with images and labels #167
Description
Probably this bug will disappear once scverse/spatialdata#318 is merged.
I have noticed that if I do a bounding box query on the blobs dataset, and I try to plot the shapes, I get an error related to the images and labels (btw strange, since I was not trying to plot raster data).
This can be reproduced as follows:
import spatialdata as sd
from spatialdata.models import TableModel
from spatialdata.datasets import blobs
import spatialdata_plot
from napari_spatialdata import Interactive
import numpy as np
sdata = blobs()
sdata_cropped = sdata.query.bounding_box(axes=('x', 'y'), min_coordinate=[200, 150], max_coordinate=[300, 250], target_coordinate_system='global')
sdata_cropped
# del sdata_cropped.images['blobs_image']
# del sdata_cropped.images['blobs_multiscale_image']
# del sdata_cropped.labels['blobs_labels']
# del sdata_cropped.labels['blobs_multiscale_labels']
sdata_cropped.pl.render_shapes('blobs_circles').pl.show()The last line will fail, but if we uncomment the 4 lines with del, which remove raster data from the sdata_cropped object, then plotting will work.
Activity
Sonja-Stockhaus commentedon Sep 29, 2023
@LucaMarconato I also noticed that there are bugs in this function, the problem is when it gets multi-scale images as input or labels with only [y, x] and not [c, y, x]. Leaving in
sdata_cropped.images['blobs_image']should be fine. I made some adjustments to the function so I hope the problem is gone when multiscale-image handling is finished (#164)timtreis commentedon Oct 2, 2023
Yeah, I'd prefer not to fix
robust_transformas it'll soon disappear. I'm just waiting for the fix of scverse/spatialdata#353 to finalise the scverse/spatialdata#318. Afterwards we plan to also do a release ofspatialdata-plot👍timtreis commentedon Nov 3, 2023
_robust_transformis no longer part of the codebase since #162