perf: Reorder conditions in is_leap_year

[janpipek]

This is a completely random shot as I was just reading polars source code for fun and have no insights into it. I can imagine the rust compiler optimises this anyway but I'd rule out 75 % of cases in the first check rather than 0.25 %.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 79.84%. Comparing base (1eb2fcc) to head (2fc93a8).
Report is 14 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main   #19602   +/-   ##
=======================================
  Coverage   79.84%   79.84%           
=======================================
  Files        1536     1536           
  Lines      211405   211405           
  Branches     2445     2445           
=======================================
+ Hits       168790   168797    +7     
+ Misses      42060    42053    -7     
  Partials      555      555           

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

[ritchie46]

@MarcoGorelli can you take a look?

[alexander-beedie]

Makes sense to me that this will be faster; to codify the degree of speedup (and to confirm that the compiler isn't psychic) I checked it with a dist-release build - testing df.select(pl.col("dtm").dt.is_leap_year()) on ~9 million datetimes gets about 8% faster with the proposed change, so gets my vote 👌

[MarcoGorelli]

sure, thanks @janpipek , and @alexander-beedie for timing it!

[janpipek]

🙏 Thanks for approval and testing. I am glad that it really helped (8% is very nice!). As said, it was more of a random observation.

[alexander-beedie]

🙏 Thanks for approval and testing. I am glad that it really helped (8% is very nice!). As said, it was more of a random observation.

Thanks - feel free to make more ;)