Factorize copy BIDS files (PR 1)

[MaximeBICMTL]

Extracted from #1335

Description

Factorize the "get BIDS file path in LORIS" and "copy BIDS file to LORIS" operations in the lib.import_bids_dataset.copy_files, and adopt these factorizations in the Eeg class. The goal of this PR is to be able to share the same logic/function across all modalities (MRI / EEG / MEG) once the other PRs land.

Notes

• There was previously an inheritance parameter to the Eeg.copy_file_to_loris_bids_dir but it was actually never used.
• I am not sure our rename strategy works in all cases: we rename BIDS files manually, but their might be cross-references to these files across the dataset. In any case, this logic is not worse than the previous one (it's the same logic). Let's factorize first and fix later IMO.

[jeffersoncasimir]

Not sure if it stems from this PR, but a session directory with bids_default_vl was created, but ses-V1 was not injected as an entity in the downstream file names (nor scans.tsv content).

Note: Ran with #1394 patch

[MaximeBICMTL]

Thanks for the catch @jeffersoncasimir !

The first bug with the missing session label for sessionless datasets was introduced in this PR, my bad!
The second bug with the missing session label in scans.tsv in particular seems to already be present on main/27.0.

I added a commit that fixes both of these bugs, and makes the renaming logic more robust in general.
I think there might still be another bug where the name of file in the scans.tsv file is not renamed, but that is out of the scope of this PR IMO. I think the whole scans.tsv logic needs to be refactored (and it will be in a future PR!).

[jeffersoncasimir]

Entities are injected in the filenames but the scans.tsv is copied at the wrong level. This file should be copied to the injected session.

Resulting dataset:

Example dataset with 2 sessions:

[MaximeBICMTL]

Another good catch of yours @jeffersoncasimir, I looked at the file name but forgot to look at the whole path...

I pushed another commit that fixes this bug. I can confirm that this bug is also present in main/27.0 (basically the copy_scans_tsv_file_to_loris_bids_dir function that I copied from ScansTSV).

Anyway, I'll add it to my list of follow-ups:

• Add a scans.tsv file in the integration tests.
• Add a root events.json file in the integration tests (and fix the path concatenation bug).

[jeffersoncasimir]

Probably the final issue.

The chunks folder created does not contain the injected default session: sub-ABC0035_task-resting_eeg.chunks, and the chunks are created, but this line triggers the following error:

ERROR: Chunk creation failed, directory '/data/Loris-MRI/data/bids_imports/EEGManyLabs_Resting_State_Dataset_Eimer1996_BIDSVersion_1.9.0_chunks/sub-ABC0035_ses-V1_task-resting_eeg.chunks' does not exist.

[MaximeBICMTL]

Okay @jeffersoncasimir, I managed to reproduce the bug. I really should have tested more with the default session label, sorry for that...

Anyway, I can confirm you that the bug is solved by merging #1395. The reason is simple: before #1395, the chunks path was determined from the input file name (which was an oopsie form my part), whereas after it is determined with the LORIS file name.

EDIT: Actually I went ahead and merged #1395 and rebased this branch. I tested a sessionless dataset and it now works as expected.

[jeffersoncasimir]

Good to merge!