BUG: millis res on parsed DatetimeIndex slice

[tapiralec]

Added support for millis resolution on a parsed DatetimeIndex slice.

There was no elif statement for when _parsed_string_to_bounds(self,
reso, parsed) was given reso == 'millisecond'

• passes git diff upstream/master -u -- "*.py" | flake8 --diff

[pep8speaks]

Hello @tapiralec! Thanks for submitting the PR.

• There are no PEP8 issues in the file pandas/core/indexes/datetimes.py !

[codecov]

Codecov Report

Merging #23010 into master will decrease coverage by <.01%.
The diff coverage is 33.33%.

@@ Coverage Diff @@ ## master #23010 +/- ## ========================================== - Coverage 92.19% 92.19% -0.01%  ========================================== Files 169 169 Lines 50835 50838 +3 ========================================== + Hits 46868 46869 +1  - Misses 3967 3969 +2

Flag	Coverage Δ
#multiple	90.61% <33.33%> (-0.01%)	⬇️
#single	42.35% <0%> (-0.01%)	⬇️

Impacted Files	Coverage Δ
pandas/core/indexes/datetimes.py	95.53% <33.33%> (-0.25%)	⬇️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update d523d9f...8109682. Read the comment docs.

[jreback]

is this fixing an issue? this would need tests, I am not sure what doesn't work now.

[tapiralec]

Sorry for the lack of good explanation earlier. I couldn't find any existing bug in the issues that listed this problem. Basically there exists support for slicing data with a DatetimeIndex using strings for the range in core/indexes/datetimes.py (line 1263):

_, parsed, reso = parsing.parse_time_string(label, freq) lower, upper = self._parsed_string_to_bounds(reso, parsed) 

_libs/tslibs/parsing.pyx returns a resolution of milliseconds when microsecond % 1000 == 0 (line 334)
_parsed_string_to_bounds() has if/elif's on the resolution argument for year, month, ... second, microsecond, but doesn't have one for millisecond.

I found this when I was working with data that had a precision of tenths of milliseconds (hundreds of microseconds). If you pass in a string bound with, e.g. .9871 for the fraction of the second, it slices as expected since 987100 % 1000 = 100 != 0, so resolution is "microseconds". But if you pass in a string bound with .9870, the parsed resolution is determined to be milliseconds and _parsed_string_to_bounds() will raise a KeyError.

With regards to the tests: I'm not familiar yet with pytest / numpy.testing, but I can take a crack at writing the tests to cover this case.

[jreback]

needs tests and a whatsnew

[jreback]

closing as stale. if you'd like to continue, pls ping.