Allow separate MRIs per session (for longitudinal studies)

[drammock]

WIP. Could use some advice about the best approach here; so far it should work (but yet untested!) for my use case but definitely won't work for the "normal" case that is currently supported on main. See the (small) diff for a prominent comment explaining the issue.

Before merging …

• [ ] Changelog has been updated (docs/source/changes.md)

[drammock]

eh, this is not going to work as-is. There is an assumption baked-in in too many places that there's only a single folder level between subjects_dir and the various freesurfer files. I'll explore making the folders something like sub-116_ses-a/ rather than sub-116/ses-a/

[larsoner]

Makes sense

[drammock]

I think this is ready for review. Probably needs some addition to the docs so that folks other than me know how to actually make this work... I can add that along with whatever other changes y'all want.

[larsoner]

I think next you'd want to add docs/source/v1.9.md.inc to say that per-session MRIs are now supported.

Any idea if there is a dataset on openneuro that we could add to the test suite that would hit the new code path?

[drammock]

I think next you'd want to add docs/source/v1.9.md.inc to say that per-session MRIs are now supported.

Done. As for adding more thorough documentation elsewhere, I can't actually see where I would add it (since it doesn't involve a config value, and there's no dataset to add that could serve as an example).

Any idea if there is a dataset on openneuro that we could add to the test suite that would hit the new code path?

I doubt it; OpenNeuro has zero datasets with modality=MEG and study_type=Longitudinal :(

[drammock]

Just checked the EEG Longitudinal studies too; none of them involve source reconstruction, only 1 even includes MRI at all, and it's only at 1 timepoint (https://openneuro.org/datasets/ds004024/versions/1.0.1)

[larsoner]

@hoechenberger any idea about the error:

    raise RuntimeError(
RuntimeError: Error 404 when trying to download /home/circleci/mne_data/ds000248/derivatives/freesurfer/subjects/fsaverage/label/Yeo_Brainmap_fsaverage_README from https://s3.amazonaws.com/openneuro.org/ds000248/derivatives/freesurfer/subjects/fsaverage/label/Yeo_Brainmap_fsaverage_README?versionId=WuMze1X4S_cLoRCbEs5IpCIOJMe2VT41

Do we need to raise a bug for the openneuro folks again?

[hoechenberger]

yes seems like it! do you want to take a look? just follow what i did in my previous bug reports :)

[drammock]

FWIW the CIs were all green on the commit prior to the changelog entry, so it's probably safe to merge this (though def. should raise the download issue with the openneuro folks)

[larsoner]

https://github.com/OpenNeuroOrg/openneuro/issues/3145 and https://github.com/hoechenberger/openneuro-py/pull/187, openneuro folks usually resolve quickly so I say let's wait a day or two then restart. If they don't respond by then let's just merge.

[hoechenberger]

Issue has already been addressed upstream by OpenNeuro, restarting CI!

[hoechenberger]

@larsoner This is green. i won't have time to review, please move along if you're happy

[larsoner]

Thanks @drammock !