Handling empty prefix

[hiraksarkar]

https://github.com/scverse/spatialdata-io/blame/bc17abc065c39e07d6b17affaf2c4b95adbaafea/src/spatialdata_io/readers/visium_hd.py#L357

Currently visium_hd function can't handle empty file prefixes. In other words,

binned_outputs       feature_slice.h5     molecule_info.h5              probe_set.csv  web_summary.html
cloupe_008um.cloupe  metrics_summary.csv  possorted_genome_bam.bam.bai  spatial

This would be unreadable by the function. Can this be fixed. Alternatively I raise a pull request.

[ddiez]

Same problem with me. The problem is here where the assignment of the dataset_id is done. Perhaps if dataset_id is None and the guessing returns empty then do not add "_". Something like this?

if dataset_id is None:
        dataset_id = _infer_dataset_id(path)

if dataset_id is "":
  filename_prefix = ""
else
  filename_prefix = f"{dataset_id}_"

@hiraksarkar A temporary solution for me to avoid renaming the files was to create a symlink to feature_slice.h5. Something like ln -s feature_slice.h5 id_ feature_slice.h5 where id is whatever label works for your project.

[ddiez]

@hiraksarkar just realized you offered a PR so probably didn't need my "temporary solution". Sorry for the noise although perhaps it is useful for someone else until this is fixed!

[LucaMarconato]

Thanks for reporting the bug and for the discussion. @hiraksarkar, if you could make a PR it would be wonderful, thank you! Happy to review it 😊

[ddiez]

Thanks @LucaMarconato sounds great. @hiraksarkar will you do the PR? If not, I can do it myself although I have not tested yet my proposed solution.

[ddiez]

Was the PR for this finally submitted? @LucaMarconato @hiraksarkar

[LucaMarconato]

No, I think there is still no PR for this. @hiraksarkar if you already have been working on a branch (even if not complete), please open a draft PR. If no code is available, @ddiez you are very welcome to try making a PR.

In any case, please open a draft PR when you are working on this so that we avoid duplicating the efforts. I am currently working on some other issues, but I want to get back to this after I finish with them.

[ddiez]

Opened draft #264. Didn't know how to open draft without having any commits (possible at all?) so added my proposal above, which fails for the reasons indicated. There are some options to consider. Happy to hear comments.

[LucaMarconato]

The issue should now be addressed thanks to #264. Please if you test the new fixed code, it would be great to learn your feedback.

[ddiez]

@LucaMarconato thanks. I tested quickly this. What I can see is:

1. If dataset_id is not provided, it will be inferred from the prefix in {prefix}_feature_slice.h5. If no prefix exists, then an error is thrown. If the prefix exists, this is used as dataset_id.

2. If dataset_id is passed feature_slice.h5 will be used.

3. If dataset_id is passed but feature_slice.h5 does not exist, dataset_id will be used as prefix in {prefix}_feature_slice.h5. In this situation if the prefix in {prefix}_feature_slice.h5 and dataset_id are different it will throw an error.

I tested these 3 situations successfully. Is my understanding of the expected behavior above correct?

[LucaMarconato]

Hi @ddiez, point 1 is correct, point 2 and 3 are slightly different. This is the behavior implemented (I also rewrote 1 for clarity):

1. If dataset_id is not provided, it will be inferred from the prefix in {dataset_id}_feature_slice.h5. If no dataset_id can be inferred, then an error is thrown and the user is asked to specify it.

2. If dataset_id is passed and{dataset_id}_feature_slice.h5 exists, then this file will be used. and filename_prefix is set to {dataset_id}_.

3. If dataset_id is passed and{dataset_id}_feature_slice.h5 does not exists, then feature_slice.h5 will be used. and filename_prefix is set to ``

4. filename_prefix dataset_id is used to name the elements of the constructed SpatialData object.