Make the EEG chunker into a standalone Python package

[MaximeBICMTL]

Description

Make the LORIS EEG chunker into its own Python package. This allows to better isolate this pipeline from the rest of LORIS Python by having it declare its own dependencies and make sure it does not depend on the rest of LORIS Python. It also makes the LORIS EEG chunker installable without the rest of LORIS Python using pip install.

Details

• Added the loris-eeg-chunker package metadata.
• Structure the loris-eeg-chunker package using the conventional src-layout.
• Make sure the main LORIS Python package depends on loris-eeg-chunker.

Testing instructions

If you want to test this PR manually, either use:

• pip install . in the LORIS-MRI root directory, which will install both LORIS Python along with the LORIS EEG chunker as a dependency.
• pip install python/loris_eeg_chunker which will only install the LORIS EEG chunker (the main codebase can still rely on environment variables if it is not installed as a package).

Migration

Note that the LORIS EEG chunker scripts have been renamed using the [project.scripts] Python feature (instead of the PATH environment variable):

• edf_to_chunks.py to edf-to-chunks
• eeglab_to_chunks.py to eeglab-to-chunks

The LORIS Python scripts that call these subscripts have been updated, so this is only notable if a project calls these scripts directly.

[jeffersoncasimir]

Tested via bids_import.py. Works seamlessly.

The only issue is that the EEGChunksPath config is not in core, but in an unmerged draft PR.

[MaximeBICMTL]

Thanks for the review Jefferson!

I did not know about EEGChunksPath but upon inspection I don't think it is relevant to this PR 🤔. There is a fallback option and that seemed to work well in my tests and #1343:
https://github.com/aces/Loris-MRI/blob/main/python/lib/physiological.py#L1223-L1229

[jeffersoncasimir]

My mistake, the EEGChunksPath problem was not introduced by this PR and I may have confused it with electrophysiology_chunked_dataset_path.

Here is what happened:

• The chunks were created (in {datasetsDir}/bids_imports/{dataset}_chunks/)
• The electrophysiology_chunked_dataset_path entries did not get added to physiological_parameter_file
• I added electrophysiology_chunked_dataset_path entries manually and visualized the recordings

Maybe important detail: I also had #1342 pulled for this test run

[MaximeBICMTL]

Hhhhm, I rebased this PR to benefit from integration tests and I also ran it manually (same thing for #1342), everything is working correctly for me, all the files are created and I can visualize the result in the EEG browser. Let's merge #1342 first and see for this one later.

[MaximeBICMTL]

@jeffersoncasimir Can you re-test this PR ? I rebased it on the other merged PRs (integration tests, os.path in EEG) and everything works well in my manual tests, the files are correctly created and I can visualize the recording in the EEG browser.

[jeffersoncasimir]

#1342 (which I approved) changed the expected chunk path. It seems that the os.path replacement must not have been equivalent.

My EEGChunksPath: /data/Loris-MRI/data/bids_imports_chunks/. Chunks ended up in {EEGChunksPath}/sub-PIDCC0821_ses-V03_task-FACE_acq-eeg_eeg.chunks but should have ended up in {EEGChunksPath}/{DatasetName}_chunks/sub-PIDCC0821_ses-V03_task-FACE_acq-eeg_eeg.chunks

My physiological_parameter_file.Value for electrophysiology_chunked_dataset_path ended up as sub-PIDCC0821_ses-V03_task-FACE_acq-eeg_eeg.chunks, but should have been bids_imports/{DatasetName}_chunks/sub-PIDCC0821_ses-V03_task-FACE_acq-eeg_eeg.chunks to match RB examples

So the front-end is looking at the wrong place for the index.json and the chunks are also not saved in the right place

Let me know how your setup looks and ends up, if you think the problem is on my end. The PR is otherwise ready, but it seems like the right time to get the above straightened out.

[MaximeBICMTL]

Oooohokay I think I understand better thanks @jeffersoncasimir. In my tests I did not have EEGChunksPath configured and it worked well. But in your case you seem to have it configured and it causes some issues with the code. I will look into this second setup thanks.

[jeffersoncasimir]

Tested with #1345 and works great!

Ready to merge