combine(R) vs concatenate(Py) for SpatialData objects

[Artur-man]

CC @LucaMarconato

The merging utility in scverse/spatialdata (i.e. concatenate) forces users to define SpatialData objects in a dictionary to avoid duplicated elements (by manually giving names for deduplication). This is currently not necessary in HelenaLC/spatialdataR (i.e. combine) and a default set of keys are applied to element names.

The R approach I favor since the method works more smoothly, but we can also follow a similar approach giving names to sd's, thus both calls below would work but in second the user defines these keys.

combine(sdata)
combine(list(s1 = sdata, s2 = sdata)

Here are behaviours of two implementations:

In scverse/spatialdata:

>>> from spatialdata import concatenate
>>> concatenate([sdata, sdata])
KeyError: 'Images must have unique names across the SpatialData objects to concatenate. Please pass a `dict[str, SpatialData]` to `concatenate()` to address this (see docstring).'
>>> concatenate({"s1": sdata, "s2": sdata})
SpatialData object
├── Images
│     ├── 'blobs_image-s1': DataArray[cyx] (3, 64, 64)
│     ├── 'blobs_image-s2': DataArray[cyx] (3, 64, 64)
│     ├── 'blobs_multiscale_image-s1': DataTree[cyx] (3, 64, 64), (3, 32, 32), (3, 16, 16)
│     └── 'blobs_multiscale_image-s2': DataTree[cyx] (3, 64, 64), (3, 32, 32), (3, 16, 16)
├── Labels
│     ├── 'blobs_labels-s1': DataArray[yx] (64, 64)
│     ├── 'blobs_labels-s2': DataArray[yx] (64, 64)
│     ├── 'blobs_multiscale_labels-s1': DataTree[yx] (64, 64), (32, 32), (16, 16)
│     └── 'blobs_multiscale_labels-s2': DataTree[yx] (64, 64), (32, 32), (16, 16)
...

In HelenaLC/spatialdataR:

> combine(sdata, sdata)
class: SpatialData   
- images(4):
  - blobs_image (3,64,64)
  - blobs_multiscale_image (3,64,64)
  - blobs_image.1 (3,64,64)
  - blobs_multiscale_image.1 (3,64,64)
...

[HelenaLC]

I agree!! I like the default, but exactly what u say was on my todo list :))

[HelenaLC]

Revised to accept both, combine(x, y) and combine(list(a, b, c, ...)). Example output below. Close if this seems alright (for now).

• If list is partially named, only named elements will be prefixed.
• A final make.unique() call assures all remains unique in case of prepending giving duplicates (unlikely).

> y <- combine(list(A=x, B=x))
> y
class: SpatialData
- images(4):
  - A.blobs_image (3,64,64)
  - A.blobs_multiscale_image (3,64,64)
  - B.blobs_image (3,64,64)
  - B.blobs_multiscale_image (3,64,64)
- labels(4):
  - A.blobs_labels (64,64)
  - A.blobs_multiscale_labels (64,64)
  - B.blobs_labels (64,64)
  - B.blobs_multiscale_labels (64,64)
- points(2):         
  - A.blobs_points (200)
  - B.blobs_points (200)
- shapes(6):
  - A.blobs_circles (5,circle)
  - A.blobs_multipolygons (2,polygon)
  - A.blobs_polygons (5,polygon)
  - B.blobs_circles (5,circle)
  - B.blobs_multipolygons (2,polygon)
  - B.blobs_polygons (5,polygon)
- tables(0):
coordinate systems(5):
- global(16): A.blobs_image A.blobs_multiscale_image ...
  A.blobs_points B.blobs_points
- scale(2): A.blobs_labels B.blobs_labels
- translation(2): A.blobs_labels B.blobs_labels
- affine(2): A.blobs_labels B.blobs_labels
- sequence(2): A.blobs_labels B.blobs_labels