try a simple fix for wrongly ordered files in tar file

[bpinsard]

solves #519

[codecov]

Codecov Report

Merging #535 (fcee77a) into master (80a6538) will not change coverage.
The diff coverage is 100.00%.

@@           Coverage Diff           @@
##           master     #535   +/-   ##
=======================================
  Coverage   77.64%   77.64%           
=======================================
  Files          41       41           
  Lines        3167     3167           
=======================================
  Hits         2459     2459           
  Misses        708      708           

Impacted Files	Coverage Δ
heudiconv/parser.py	92.70% <100.00%> (ø)

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 80a6538...fcee77a. Read the comment docs.

[yarikoptic]

But isn't the underlying issue that we (or is it a heuristic?) have some order effect?

[yarikoptic]

File names could be 1, 2, ..., 10 and string sorting would screw it up

[bpinsard]

You're right, our files are alphanumerically sorted by acquisition/content-creation order because it uses Siemens UID, and that might not be the general case.
However, I think it is sorted as well when not pulled from a tar filed (eg obtained from a directory).

heudiconv/heudiconv/parser.py

Lines 163 to 164 in 80a6538

	files += sorted(find_files(
	'.*', topdir=f, exclude_vcs=True, exclude="/\.datalad/"))

In most cases the order of the files won't matter to the indexing nor to dcm2niix.

But for that particular case described in #519 : complex data (at least 2 dicoms with M and P ImageType) in a single series will give a varying ImageType to the seqinfo depending on the sorting, as the parse will choose the first one as the example dicom.

[yarikoptic]

your argument that we sort in non-tar handling convinces me. Let's proceed with this, and see if any fallout happens ;) Thank you @bpinsard !

[github-actions]

🚀 PR was released in v0.11.0 🚀