Test spatialelement table join

[LucaMarconato]

Closes #181

Adds a test for checking that the plot is correct even when some rows of the table are not annotating instances of the spatial element or when the spatial element contains instances not in the table.

Also, checks that this works both for the case when the instance_key column has dtype int and when it's a str.

Eventually, the instance_key column will only have dtype int, as discussed here: scverse/spatialdata#158, but for the moment it would be easier if both were supported, as this PR will check.

[LucaMarconato]

I would remove this and adjust the asserts in the test.

[LucaMarconato]

Some additional comments, I will put them here while I test.

• single_molecule from the merfish dataset can't be visualized.
  The error is

--> 431 subsample_points = points.iloc[subsample] if subsample else points
        points =                    x            y     cell_type
0        1278.683956  6020.642260  outside_VISp
1        1326.970330  6023.884788  outside_VISp
2        1292.026844  6059.337093  outside_VISp
3        1300.886241  6097.786264  outside_VISp
4        1232.410068  6102.884182  outside_VISp
...              ...          ...           ...
3714637  3161.427603  5192.594981  outside_VISp
3714638  3099.698528  5221.596008  outside_VISp
3714639  3084.582240  5297.234605  outside_VISp
3714640  3054.192051  5342.142346   VISp_II/III
3714641  3058.963217  5348.150185   VISp_II/III

[3714642 rows x 3 columns]
        subsample = <class 'numpy.ndarray'> (100000,) int64
    432 if subsample:
    433     adata = _left_join_spatialelement_table(
    434         {"points": {original_name: subsample_points}}, table_name, match_rows="left"
    435     )

ValueError: The truth value of an array with more than one element is ambiguous. Use a.any() or a.all()

[LucaMarconato]

• Unfortunately also genes from the mibitof dataset can't be visualized (I checked them because they contain labels/table mismatches).

File ~/embl/projects/basel/napari-spatialdata/src/napari_spatialdata/_widgets.py:217, in AListWidget._(self=<napari_spatialdata._widgets.AListWidget object>, vec=cell_id
2      -0.042887
4      -0.004494
5     ...-0.039209
Name: vec, Length: 1023, dtype: float32, **kwargs={'key': 'CD31'})
    215 vec = pd.Series(vec, name="vec", index=instance_key_col)
    216 layer_meta = self.model.layer.metadata if self.model.layer is not None else None
--> 217 element_indices = pd.Series(layer_meta["indices"], name="element_indices")
        layer_meta = {'sdata': SpatialData object with:
├── Images
│     ├── 'point8_image': SpatialImage[cyx] (3, 1024, 1024)
│     ├── 'point16_image': SpatialImage[cyx] (3, 1024, 1024)
│     └── 'point23_image': SpatialImage[cyx] (3, 1024, 1024)
├── Labels
│     ├── 'point8_labels': SpatialImage[yx] (1024, 1024)
│     ├── 'point16_labels': SpatialImage[yx] (1024, 1024)
│     └── 'point23_labels': SpatialImage[yx] (1024, 1024)
└── Tables
      └── 'table': AnnData (3309, 36)
with coordinate systems:
▸ 'point8', with elements:
        point8_image (Images), point8_labels (Labels)
▸ 'point16', with elements:
        point16_image (Images), point16_labels (Labels)
▸ 'point23', with elements:
        point23_image (Images), point23_labels (Labels), 'adata': AnnData object with n_obs × n_vars = 1023 × 36
    obs: 'row_num', 'point', 'cell_id', 'X1', 'center_rowcoord', 'center_colcoord', 'cell_size', 'category', 'donor', 'Cluster', 'batch', 'library_id'
    uns: 'spatialdata_attrs'
    obsm: 'X_scanorama', 'X_umap', 'spatial', 'region_key': 'library_id', 'instance_key': 'cell_id', 'table_names': ['table'], 'name': 'point16_labels', '_active_in_cs': {'point16'}, '_current_cs': 'point16'}
        pd = <module 'pandas' from '/Users/macbook/miniconda3/envs/ome/lib/python3.10/site-packages/pandas/__init__.py'>
    218 diff_element_table = set(instance_key_col).symmetric_difference(element_indices)
    219 merge_vec = pd.merge(element_indices, vec, left_on="element_indices", right_index=True, how="left")[
    220     "vec"
    221 ].fillna(0, axis=0)
    ```

[LucaMarconato]

I could reproduce the bug appearing with the mibitof dataset in test_partial_table_matching_with_arbitrary_ordering() (it appears both for labels and multiscale_labels`. I could not reproduce the points bug with the test (because no subsampling is happening).

I verified the robustness against shuffling and it works. I will now check the get_extent() code for the automatic determination of the default points size.