feat(python,rust): Add many more auto-inferable datetime formats for str.to_datetime

[Julian-J-S]

fix #16115

also as @alexander-beedie mentioned (#16115 (comment)) I added support for many missing patterns so that all separators (-, ., /) have the same patterns.

❗ Needs Decision

There are currently 3 failing tests because they have datetime texts with fractional seconds that do not have 3/6/9 digits but something else like

• 2018-09-05T04:24:01.9 -> 1 digit
• 2018-09-05T04:24:02.11 -> 2 digits

usually fractional seconds should only be in milli, micro or nano seconds. Not sure if there are actually any formats out there apart from the tests?!?
Still, it would be possible to support that using the %.f specifier (little slower probably) instead of %.3f, %.6f, %.9f

Need Feedback if we want to support those (strange?) seconds fractions (-> will adjust patterns accordingly) or not (-> will adjust tests)

[MarcoGorelli]

thanks @JulianCologne

for the failing tests, is that because you've removed "%FT%H:%M:%S%.f"? tbh I'd be ok with just keeping that one in at the end of the list, no big deal

[Julian-J-S]

thanks @JulianCologne

for the failing tests, is that because you've removed "%FT%H:%M:%S%.f"? tbh I'd be ok with just keeping that one in at the end of the list, no big deal

yes, I removed that on purpose because it makes the patterns unbalanced.
With all the different separators and formats we have like ~30 patterns but then only 1(!) should support different number of fractional digits? Feels very wrong and arbitrary to me

Format %FT%H:%M:%S%.f is just an abbreviation where %F=%Y-%m-%d

My suggestion would be either to not support this (have never seen this irl) or support any fractional digits for all patterns wich can be achieved by either

• adding %.f (any number of digits) after all the %.3f, %.6f, %.9f patterns, or
• use %.f (any number of digits) instead of all the %.3f, %.6f, %.9f patterns

[MarcoGorelli]

Feels very wrong

It's just an extra pattern :) For context, it comes from #3195 . %Y-%m-%dT%H:%M:%S%.f is the common ISO8601 format, IMO it's ok for that one to be a bit laxxer as a last-resort, so long as it's in the "other" section at the end of the formats (as it currently is)

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 81.53%. Comparing base (d7b4f72) to head (20fa8c8).
Report is 20 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main   #16634      +/-   ##
==========================================
+ Coverage   81.51%   81.53%   +0.01%     
==========================================
  Files        1414     1414              
  Lines      185906   186398     +492     
  Branches     3027     3014      -13     
==========================================
+ Hits       151535   151972     +437     
- Misses      33840    33896      +56     
+ Partials      531      530       -1     

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

[Julian-J-S]

@MarcoGorelli
my last commit added the missing pattern again and all checks are passing.
Ready from my side 😃

[MarcoGorelli]

cool, open to expanding these, thanks @JulianCologne !

[MarcoGorelli]

this one would previously get parsed as the year 0021, which is almost surely not what anybody wanted...good call here