parquet support for visium HD

[ktpolanski]

• [x] Closes #2973
• [ ] Tests included or not required because:

• [ ] Release notes not necessary because:

Visium HD stores its coordinates in a .parquet file. This loads said file.

[codecov[bot]]

Codecov Report

Attention: Patch coverage is 40.00000% with 3 lines in your changes are missing coverage. Please review.

Project coverage is 75.49%. Comparing base (9c8c095) to head (8e23732).

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #2992      +/-   ##
==========================================
- Coverage   75.51%   75.49%   -0.02%     
==========================================
  Files         117      117              
  Lines       12955    12959       +4     
==========================================
+ Hits         9783     9784       +1     
- Misses       3172     3175       +3     

Files	Coverage Δ
scanpy/readwrite.py	67.72% <40.00%> (-0.43%)	:arrow_down:

[Rafael-Silva-Oliveira]

Also refer to https://github.com/scverse/scanpy/issues/2982

[ktpolanski]

We ran some data through spaceranger 3.0.1 locally, and in doing so found that 10X have reduced redundancy in the spatial folder of the binned outputs by moving the tissue images to a new, central location. This understandably breaks the existing loader.

A hotfix is to copy the images back into the appropriate subdirectory, but that's not a feasible expectation on users. I added an optional argument spaceranger_image_path to point to the new folder if need be, which should hopefully be robust with regard to any sort of further restructuring 10X may choose to do in the future.

The code is currently included in bin2cell in case anybody needs it or just wants to take it out for a spin, but I think it belongs in a more central location.

[giovp]

hi @ktpolanski ,thank you for the heads up

A hotfix is to copy the images back into the appropriate subdirectory, but that's not a feasible expectation on users. I added an optional argument spaceranger_image_path to point to the new folder if need be, which should hopefully be robust with regard to any sort of further restructuring 10X may choose to do in the future.

The code is currently included in bin2cell in case anybody needs it or just wants to take it out for a spin, but I think it belongs in a more central location.

@LucaMarconato do we have any datasets that test for spaceranger 3.0.1 ?

[LucaMarconato]

Last week together with Wouter we identified some good candidate datasets (some of them recommended by 10x). I believe it's 3.0.0 not 3.0.1 though. The list is in the top part of my comment here https://github.com/scverse/spatialdata/issues/700#issuecomment-2326790050