[MAINT] transfer tests from pybids

[Remi-Gau]

• transfer tests cases that were run in pybids
• split between valid and invalid tests

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 93.19%. Comparing base (9acd81f) to head (bb5f82d).
Report is 7 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main       #8      +/-   ##
==========================================
+ Coverage   85.90%   93.19%   +7.28%     
==========================================
  Files           3        3              
  Lines         149      191      +42     
  Branches       43       55      +12     
==========================================
+ Hits          128      178      +50     
+ Misses          9        5       -4     
+ Partials       12        8       -4     

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

[effigies]

I'm surprised this is passing.

[effigies]

In retrospect, I don't like parametrizing on matches. I would rather just split it into two functions, e.g.,

@pytest.mark.parametrize('fname', valid_list)
def test_is_top_level(validator, fname):
    assert validator.is_top_level(fname)

@pytest.mark.parametrize('fname', invalid_list)
def test_is_not_top_level(validator, fname):
    assert not validator.is_top_level(fname)

If you've got the energy, would you do this? If not, I'll just create an issue, and might sooner or later get around to it.

[Remi-Gau]

do you want to apply this to all tests?

[effigies]

Yes.

[effigies]

@Remi-Gau I assume you're looking through the failing entries, but I can if you'd rather.

[Remi-Gau]

yeah I will have a look

[effigies]

I suspect that what's happening is that more targeted regexes were expected, and valid BIDS files were being passed to see whether they triggered only the right ones. Checking is_bids() would be expected to pass.

[Remi-Gau]

OK from my understanding it seems that all the failures are due to the fact that the validaton in pybids used to include methods methods such as is_anat which were stricter than the simple is_bids here.

I would suggest just removing the test case that fail unless we have any intention of adding those methods back at some point (and even then adding those tests would be easy, so I am very much in favor of removing those test cases).

[Remi-Gau]

I suspect that what's happening is that more targeted regexes were expected, and valid BIDS files were being passed to see whether they triggered only the right ones. Checking is_bids() would be expected to pass.

hahaha look at us independently coming to the same conclusion #reproducibility

[effigies]

Yeah, no we definitely aren't going to make more and more precise groupings of regexes, so let's get rid of these.

I see the regexes as a stopgap until we implement parsing that can tell people what they got wrong, not just "is it BIDS?". Possibly we'll keep them around for fast filtering of non-BIDS files from a layout, as in pybids.

[Remi-Gau]

why the failure on 3.8 with minimum dependencies ?

https://github.com/bids-standard/python-validator/actions/runs/10403948653/job/28811412557?pr=8#step:12:439

[effigies]

Bad regex generation in earlier bidsschematools. Bump to 0.11 in pyproject.toml and tox.ini.

[effigies]

It's pleasing that the min check exposed exactly the overlookable edge case it was meant to. I forgot about it, myself.

[Remi-Gau]

nothing else to do on my side for this one

[effigies]

Thanks!