feat(python) Series.index_of() and equivalent expression - Request for architectural review, take 2 #19408
Conversation
| input: vec![self, element], | ||
| function: FunctionExpr::IndexOf, | ||
| options: FunctionOptions { | ||
| // TODO which ApplyOptions, if any? |
There was a problem hiding this comment.
Is there a specific choice that makes sense here? I don't yet understand the enum options' meanings.
There was a problem hiding this comment.
I think wildcard expansion should be set to 2..
Codecov ReportAttention: Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## main #19408 +/- ##
==========================================
- Coverage 80.03% 80.00% -0.03%
==========================================
Files 1528 1534 +6
Lines 209657 210840 +1183
Branches 2416 2443 +27
==========================================
+ Hits 167791 168675 +884
- Misses 41317 41610 +293
- Partials 549 555 +6 ☔ View full report in Codecov by Sentry. |
That start's to feel like a join. |
| input: vec![self, element], | ||
| function: FunctionExpr::IndexOf, | ||
| options: FunctionOptions { | ||
| // TODO which ApplyOptions, if any? |
There was a problem hiding this comment.
I think wildcard expansion should be set to 2..
| T: PolarsNumericType, | ||
| { | ||
| fn index_of(&'a self, value: Option<T::Native>) -> Option<usize> { | ||
| // A NaN is never equal to anything, including itself. But we still want |
There was a problem hiding this comment.
This should be implemented over the chunks and not using an iterator over ChunkedArray's as they are really slow.
In the case we are looking for a None we can also fast path to first null.
Goal is to fix a decent chunk of #5503 when finished. At this point it is more in the "am I in the right direction?" stage, though it does work for the implemented cases!
search_sorted()has where it can search for multiple values in a single pass, so actually maybe better to leave it as is.Out of scope for this pass: