Exceptions when adding labels to backed SpatialData

[aeisenbarth]

When saving labels to a SpatialData that has already been written to disk (with or without overwrite=True), there are multiple exceptions.

Example code

import numpy as np
import spatialdata

sdata = spatialdata.SpatialData()
sdata.write("/tmp/spatialdata.zarr")
labels = spatialdata.models.Labels2DModel.parse(np.ones((2, 2)))
sdata.add_labels(name="my_label", labels=labels, overwrite=True)

1. UnboundLocalError: local variable 'elem_group' referenced before assignment

A variable is accessed (L415) which is only conditionally declared. It would be safer either to assign a default value, or to access it only in the same conditional branch.

I don't understand the reasoning of the comment, but to me it seems a fix would be to remove the condition above L411.

spatialdata/src/spatialdata/_core/spatialdata.py

Lines 409 to 416 in 4dd5250

	# not need to create the group for labels as it is already handled by ome-zarr-py
	if element_type != "labels":
	elem_group = root.create_group(name=element_type) if element_type not in root else root[element_type]
	if overwrite:
	if element_type == "labels" and element_type in root:
	elem_group = root[element_type]
	if name in elem_group:
	del elem_group[name]

Traceback (most recent call last):
 File "<ipython-input-6-6d9e217edb59>", line 1, in <module>
 sdata.add_labels(name="my_label", labels=labels, overwrite=True)
 File "…/spatialdata/_core/spatialdata.py", line 779, in add_labels
 elem_group = self._init_add_element(name=name, element_type="labels", overwrite=overwrite)
 File "…/spatialdata/_core/spatialdata.py", line 415, in _init_add_element
 if name in elem_group:
UnboundLocalError: local variable 'elem_group' referenced before assignment

2. AssertionError on type of dims

For backed labels, xdata is not a DataArray (where dims is ("y", "x")) but a DataTree where dims is Frozen({"y": 1633, "x": 1290}).

spatialdata/src/spatialdata/models/_utils.py

Lines 151 to 153 in 4dd5250

	xdata = e["scale0"].values().__iter__().__next__()
	dims_data = xdata.dims
	assert isinstance(dims_data, tuple)

This works with:

xdata = next(iter(e["scale0"].ds.values()))

or converting the dictionary to tuple:

 if isinstance(dims_data, Mapping):
 dims_data = tuple(dims_data.keys())

3. DataTree has no attribute shape

spatialdata/src/spatialdata/transformations/_utils.py

Lines 118 to 119 in 4dd5250

	xdata = list(node.values())[0]
	new_shape = np.array(xdata.shape)

Also here:

 xdata = next(iter(e["scale0"].ds.values()))

[LucaMarconato]

Thanks for reporting, I will look into this and continue from your PR.

[LucaMarconato]

Closing because the IO apis changed in main with #329 (please note that the PR is not in the current version v0.0.15).

I have opened a separate issue to track the other points #446.