Summary

Fixes #3712 .

List of changes proposed in this PR (pull-request)

• try getting the len of idxs
• if it fails with TypeError (which happens when idxs is a 0D array)
• set nels to 1

@man-shu I'm not sure if this code actually gets tested by our tests. If you're testing on your end, please make sure that it works in cases where it was previously failing.

Codecov Report

Attention: coverage is 0% with 3 lines in your changes missing coverage. Please review.

Project coverage is 50.84%. Comparing base (fe20d45) to head (12f7023).

Files with missing lines	%	Lines
nipype/algorithms/misc.py	0.00%	3 Missing⚠️

❗ There is a different number of reports uploaded between BASE (fe20d45) and HEAD (12f7023). Click for more details.

HEAD has 21 uploads less than BASE

Flag	BASE (fe20d45)	HEAD (12f7023)
	22	1

Additional details and impacted files

@@             Coverage Diff             @@
##           master    #3713       +/-   ##
===========================================
- Coverage   73.07%   50.84%   -22.23%     
===========================================
  Files        1278     1278               
  Lines       59406    59367       -39     
===========================================
- Hits        43411    30187    -13224     
- Misses      15995    29180    +13185     

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

@man-shu I'm not sure if this code actually gets tested by our tests. If you're testing on your end, please make sure that it works in cases where it was previously failing.

Ok will update you on that.

@man-shu Have you had a chance to test this?

Sorry not yet! Our HPC has been down since the day I opened this PR. Our IT dept is telling us that it should be ok by next week, so I hope to get back to you then.

(I tried copying the original nipype cache locally and then run it, but then it just started a fresh run because the checksum didn't match.)

Hello @effigies ! I finally ran the full pipeline with and without this fix and both of them worked.
So I don't think either of them were hitting this case.

I do not have access to the cache for the failing run either.

Do you have any suggestions on how to proceed from here?

Just need to eyeball it, then.

It still looks right to me after a few weeks, so let's merge.