Repository navigation
TST: Daily cron with pandas fails with AttributeError: 'SoupStrainer' object has no attribute 'name' #16251
Description
pandas 2.2.1 was release back in Feb 23, so not sure why I didn't see this earlier.
pandas==2.2.1
...
_________________________ test_read_write_format[html] _________________________
...
def _parse_tables(self, document, match, attrs):
> element_name = self._strainer.name
E AttributeError: 'SoupStrainer' object has no attribute 'name'
pandas/io/html.py:601: AttributeError
Activity
neutrinoceros commentedon Mar 30, 2024
I took a look, and
pandasdoesn't seem to be the culprit.Rather, the breaking change appears to be in
beautifulsoup44.13.0b2, released on March the 20thFurther inspection is needed to assess what to do, but this is where I got so far.
neutrinoceros commentedon Mar 30, 2024
I couldn't find release notes for this beta release, or a changelog, or an existing report at pandas. I'll try to extract a minimal reproducer and report to pandas devs.
neutrinoceros commentedon Mar 30, 2024
reported upstream at pandas-dev/pandas#58086
pllim commentedon Mar 31, 2024
Thanks for investigating!
neutrinoceros commentedon Apr 1, 2024
The fix seemed simple enough so I also opened a PR upstream pandas-dev/pandas#58100
neutrinoceros commentedon Apr 1, 2024
My fix was just merged. However, we're not pulling in pandas nighties, so it might take a while before CI goes back to green by itself. I think we have three options:
What would you advise ?
pllim commentedon Apr 1, 2024
That job is for pre-release. Is pandas going to release that fix soon? We can wait a few weeks. If not, then maybe we can check for conflicting pandas + bs4 versions and then xfail it for that combo.
neutrinoceros commentedon Apr 1, 2024
Apparently the next release should be pandas 3.0, and was due today, so I guess it should come out sometimes soon, at least in a pre-release, which would work for this job.
pllim commentedon Apr 1, 2024
OK let's wait a bit then. Thanks!
neutrinoceros commentedon Apr 3, 2024
The fix is now backported to pandas 2.2.x, which means it should be available sooner !
neutrinoceros commentedon Apr 11, 2024
@pllim pandas 2.2.2 is out with the fix + numpy 2 compat, and daily cron is green again ! I think we can close this one.
pllim commentedon Apr 11, 2024
So it is. Thanks for the update!
https://github.com/astropy/astropy/actions/runs/8640882810