augur/parse.py: fix test failure with pandas 2.2.

[emollier]

Description of proposed changes

This change fixes another case of change in pandas internals causing a breakage in the date parser; initial report is Debian bug #1071122.

Here is a copy of the test failure message for your convenience:

=================================== FAILURES ===================================
___________________________ TestParse.test_fix_dates ___________________________

self = <test_parse.TestParse object at 0x7fe8586a3590>
capsys = <_pytest.capture.CaptureFixture object at 0x7fe8582e7ce0>

    def test_fix_dates(self, capsys):
        full_date = "4-5-2020"
>       assert parse.fix_dates(full_date) == "2020-05-04"

tests/test_parse.py:14:
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _

d = '4-5-2020', dayfirst = True

    def fix_dates(d, dayfirst=True):
        '''
        attempt to parse a date string using pandas date parser. If ambiguous,
        the argument 'dayfirst' determines whether month or day is assumed to be
        the first field. Incomplete dates will be padded with XX.
        On failure to parse the date, the function will return the input.
        '''
        try:
>           from pandas.core.tools.datetimes import parsing
E           ImportError: cannot import name 'parsing' from 'pandas.core.tools.datetimes' (/usr/lib/python3/dist-packages/pandas/core/tools/datetimes.py)

augur/parse.py:34: ImportError

I agree adjusting the date parser to stick to pandas public api should be preferred; this change is available if need be.

Related issue(s)

• test failure beyond pandas 2.0 #1303
• parse: Support pandas 2.x #1436

Checklist

• Checks pass: note I verified tests kept passing in Debian sid context with pandas 2.1, and were now passing with pandas 2.2 from Debian experimental.
• If making user-facing changes, add a message in CHANGES.md summarizing the changes in this PR I did not make user facing changes I believe.

[victorlin]

@emollier thanks for surfacing these issues to us. I'll review this later, but for now I've opened an issue for us to support pandas v2 officially to prevent stuff like this from coming up on your side.

[corneliusroemer]

Thanks, this is the same fix I arrived at. We should really stop using that internal pandas functionality though, I've made an issue for it #1670