[ENH] Recompress files where dcm2niix failed

[octomike]

As threatened in #802 an alternative approach.

• try to open and read the nifit to verify it wasn't truncated
• recompress with gzip
• fail if we couldn't - as opposed to just emit invalid file names (at least fmriprep fails to handle nii.gz that are not gz at all)

I decided to leverage the fact that heudiconv blindly moves all converted output files to nii.gz, even if they are just nii by renaming the successfully re-compressed image to .nii. This feels ugly, but results in the shortest/least-invasive PR.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 83.64%. Comparing base (9fc2318) to head (22db16e).
⚠️ Report is 7 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #803      +/-   ##
==========================================
+ Coverage   83.24%   83.64%   +0.40%     
==========================================
  Files          42       42              
  Lines        4404     4488      +84     
==========================================
+ Hits         3666     3754      +88     
+ Misses        738      734       -4     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[yarikoptic]

why to delay this crash here? this way you are hiding from user information on individual file and corresponding error -- why not just to just call recompress here and let error (if any happens) to bubble up?

I would also just loop here and have recompress to operate on a single file.

FWIW, I think it would be worth adding a test scenario to test this code path. I would have just mock patched the nipype_convert so that I would decompress some output file(s) and thus triggering this code path.

[octomike]

Thanks, I'll look into it!

[yarikoptic]

Hi @octomike , have time to finish this one up?

[octomike]

Wow, completely forgot. Will definitely look again

[octomike]

Took me just 6 months..

I addressed your suggestions and also added two test cases as you described (disclaimer: those were generated in a claude session - I don't know the policy here, but I'm open to redo those manually). The test cases check that recompress_failed actually recompresses the non-gz nifti returned from a mocked nipype_convert and also assert an error when loading a truncated/corrupt file because in the implementation I validate the uncompressed files are actually okay with nibabel.load().get_fdata().