Oldest version tests failing on SIMBAD utils mismatch

[keflavich]

This error affects #3459. It's a bit weird, as it looks like a change in a value that I would not expect under any conditions, no matter what underlying packages change. 28.93617776179148 is being read in as ...177761791473, apparently. Maybe I'm missing the underlying issue, but that doesn't seem possible.

https://github.com/astropy/astroquery/actions/runs/19459612807/job/55680512989?pr=3459

=================================== FAILURES ===================================
_ test_transpiler[region(GAL,180 0,2d) & otype = 'G' & (nbref >= 10|bibyear >= 2000)-CONTAINS(POINT('ICRS', ra, dec), CIRCLE('ICRS', 86.40498828654475, 28.93617776179148, 2.0)) = 1 AND otypes.otype = 'G' AND (nbref >= 10 OR bibyear >= 2000)] _

test = "region(GAL,180 0,2d) & otype = 'G' & (nbref >= 10|bibyear >= 2000)"
result = "CONTAINS(POINT('ICRS', ra, dec), CIRCLE('ICRS', 86.40498828654475, 28.93617776179148, 2.0)) = 1 AND otypes.otype = 'G' AND (nbref >= 10 OR bibyear >= 2000)"

    @pytest.mark.parametrize("test, result", [
        ("region(GAL,180 0,2d) & otype = 'G' & (nbref >= 10|bibyear >= 2000)",
         ("CONTAINS(POINT('ICRS', ra, dec), CIRCLE('ICRS', 86.40498828654475, 28.93617776179148, 2.0)) = 1"
          " AND otypes.otype = 'G' AND (nbref >= 10 OR bibyear >= 2000)")),
        ("author ∼ 'egret*'", "regexp(author, '^egret.*$') = 1"),
        ("cat in ('hd','hip','ppm')", "cat IN ('hd','hip','ppm')"),
        ("author !~ 'test'", "regexp(author, '^test$') = 0"),
        ("sptype < F4", "sp_type < 'F4'"),
        ("umag < 1", "allfluxes.u_ < 1"),
        ("Vmag = 10", "allfluxes.V = 10"),
        ("otypes != 'Galaxy'", "otypes.otype != 'Galaxy..'"),
        ("maintype=SNR", "basic.otype = 'SNR'"),
        ("maintypes=SNR", "basic.otype = 'SNR..'")
    ])  # these are the examples from http://simbad.cds.unistra.fr/guide/sim-fsam.htx
    # plus added examples
    def test_transpiler(test, result):
        # to regenerate transpiler after a change in utils.py, delete `criteria_parsetab.py`
        # and run this test file again.
        translated = CriteriaTranslator.parse(test)
>       assert translated == result
E       AssertionError: assert 'CONTAINS(POI...year >= 2000)' == 'CONTAINS(POI...year >= 2000)'
E         Skipping 73 identical leading characters in diff, use -v to show
E         - 17776179148, 2.0)) = 1 AND otypes.otype = 'G' AND (nbref >= 10 OR bibyear >= 2000)
E         ?           ^
E         + 177761791473, 2.0)) = 1 AND otypes.otype = 'G' AND (nbref >= 10 OR bibyear >= 2000)
E         ?           ^^

../../.tox/py39-test-oldestdeps-alldeps/lib/python3.9/site-packages/astroquery/simbad/tests/test_utils.py:103: AssertionError
=========================== short test summary info ============================
FAILED ../../.tox/py39-test-oldestdeps-alldeps/lib/python3.9/site-packages/astroquery/simbad/tests/test_utils.py::test_transpiler[region(GAL,180 0,2d) & otype = 'G' & (nbref >= 10|bibyear >= 2000)-CONTAINS(POINT('ICRS', ra, dec), CIRCLE('ICRS', 86.40498828654475, 28.93617776179148, 2.0)) = 1 AND otypes.otype = 'G' AND (nbref >= 10 OR bibyear >= 2000)] - AssertionError: assert 'CONTAINS(POI...year >= 2000)' == 'CONTAINS(POI...year >= 2000)'
  Skipping 73 identical leading characters in diff, use -v to show
  - 17776179148, 2.0)) = 1 AND otypes.otype = 'G' AND (nbref >= 10 OR bibyear >= 2000)
  ?           ^
  + 177761791473, 2.0)) = 1 AND otypes.otype = 'G' AND (nbref >= 10 OR bibyear >= 2000)
  ?           ^^
=========== 1 failed, 1868 passed, 2095 skipped, 2 xfailed in 45.06s ===========
py39-test-oldestdeps-alldeps: exit 1 (47.09 seconds) /home/runner/work/astroquery/astroquery/.tmp/py39-test-oldestdeps-alldeps> pytest --pyargs astroquery /home/runner/work/astroquery/astroquery/docs '' pid=2537
  py39-test-oldestdeps-alldeps: FAIL code 1 (84.37=setup[37.03]+cmd[0.26,47.09] seconds)
  evaluation failed :( (84.48 seconds)

[keflavich]

Ah, no, I'm mistaken, it's clear how it's happening:

from astropy.coordinates import SkyCoord
print(SkyCoord(180, 0, unit='deg', frame='galactic').icrs.dec.deg)"
# 28.93617776179148

Apparently there was a slightly different conversion on earlier versions of astropy?

Then the confusing thing is why this failure did not appear earlier.

[keflavich]

I can't run these tests locally to verify... so I'm not sure if it's astropy or numpy. The tox environment won't build because of weird mac / numpy wheel things that I can't circumvent.

[ManonMarchand]

Hum yes, very strange.
However we don't need that level of precision for the test, we just want to check that astropy's conversion was called. I'd say we can safely round to a few digits

[ManonMarchand]

I'm stuck in meetings today but I can find some time tomorrow

[keflavich]

I don't think this is really a SIMBAD issue, but maybe we can change the test to truncate some more digits. I'd like to understand why it started recently before we make any changes - is it because we changed the minimum supported version? Or did something else under the hood change?

[ManonMarchand]

I tried to bisect it but I cannot install most of astropy commits because of:

ModuleNotFoundError: No module named 'setuptools.dep_util'

Which is supposed to be only broken with setuptools 69 (see pypa/setuptools#4126), but changing versions up or down did not solve my issue at all :/

[keflavich]

I'm OK with a fix that just silences the error, then. It doesn't seem super important as it's only affecting older versions, which we're going to stop supporting sometime soon anyway.

[bsipocz]

I think this something off with 1) test plugin 2) version combination; and but really a Sinbad issue (though I don't understand why it always feels like a transient, or why it even runs without remote data)