Better narrowing #20564

Activity

1. hauntsaninja

   added

   feature

   metaIssues tracking a broad area of workIssues tracking a broad area of work

   on Jan 11, 2026
2. hauntsaninja commented on Jan 11, 2026

   

   hauntsaninja

   on Jan 11, 2026

   Last edited by hauntsaninja

   CollaboratorAuthor

   More actions

   The first PR is:

   • Rework narrowing logic for equality and identity #20492

   This is the foundation for everything else and fixes several issues. It's been open for two weeks, so I will probably merge it in another week or so if there aren't comments
3. hauntsaninja commented on Jan 14, 2026

   

   hauntsaninja

   on Jan 14, 2026

   CollaboratorAuthor

   More actions

   In case useful, for now I have several of my queued changes visible together on this branch: hauntsaninja#5
4. hauntsaninja commented on Jan 16, 2026

   

   hauntsaninja

   on Jan 16, 2026

   Last edited by hauntsaninja

   CollaboratorAuthor

   More actions

   The second PR is pure refactoring (as a follow up to #20492 ):

   • Refactor equality and identity narrowing for clarity #20595

   I also opened a third PR with some more test cases:

   • Add some additional narrowing test cases #20598
5. hauntsaninja commented on Jan 16, 2026

   

   hauntsaninja

   on Jan 16, 2026

   Last edited by hauntsaninja

   CollaboratorAuthor

   More actions

   Note the first PR does introduce three regressions that I haven't yet solved:

   • Regression with narrowing with multiple hasattr #20596
   • Regression with narrowing with negative intersection #20597
   • Regression with narrowing with TypeVar intersection #20628
6. hauntsaninja commented on Jan 16, 2026

   

   hauntsaninja

   on Jan 16, 2026

   Last edited by hauntsaninja

   CollaboratorAuthor

   More actions

   A fourth PR is better narrowing for NotImplemented:

   • Treat NotImplemented as a singleton type #20601
7. hauntsaninja commented on Jan 16, 2026

   

   hauntsaninja

   on Jan 16, 2026

   Last edited by hauntsaninja

   CollaboratorAuthor

   More actions

   This next one is a useful one! Fifth PR is narrowing for collection containment:

   • Narrow types based on collection containment #20602
8. hauntsaninja commented on Jan 16, 2026

   

   hauntsaninja

   on Jan 16, 2026

   Last edited by hauntsaninja

   CollaboratorAuthor

   More actions

   Contributor PRs to follow up on as stack is merged: #20603 (done) #20277 (open) #20538 (done)
9. hauntsaninja commented on Jan 17, 2026

   

   hauntsaninja

   on Jan 17, 2026

   CollaboratorAuthor

   More actions

   Sixth PR improves logic with enum subclasses:

   • Improve narrowing logic for Enum int and str subclasses #20609
10. hauntsaninja commented on Jan 22, 2026

    

    hauntsaninja

    on Jan 22, 2026

    Last edited by hauntsaninja

    CollaboratorAuthor

    More actions

    Add this test: #20602 (comment) (done in stack)
11. hauntsaninja commented on Jan 22, 2026

    

    hauntsaninja

    on Jan 22, 2026

    CollaboratorAuthor

    More actions

    Seventh PR:

    • Refactor and improve narrowing for type(x) == t checks #20634
12. hauntsaninja commented on Jan 22, 2026

    

    hauntsaninja

    on Jan 22, 2026

    Last edited by hauntsaninja

    CollaboratorAuthor

    More actions

    Eighth PR (exploratory for now, do not review):

    • [do not review] Avoid narrowing TypeType #20636
13. hauntsaninja commented on Jan 22, 2026

    

    hauntsaninja

    on Jan 22, 2026

    CollaboratorAuthor

    More actions

    Ninth PR is trivial testing related changes:

    • Move tests to check-narrowing , improve them slightly #20637
14. hauntsaninja commented on Jan 23, 2026

    

    hauntsaninja

    on Jan 23, 2026

    Last edited by hauntsaninja

    CollaboratorAuthor

    More actions

    Tenth PR adds support for narrowing in type(x) is type(y), which seems pretty common in primer:

    • Narrow for type expr comparisons to type exprs #20639
15. 22 remaining items

    Load more

16. hauntsaninja commented on Feb 3, 2026

    

    hauntsaninja

    on Feb 3, 2026

    CollaboratorAuthor

    More actions

    Thirty second PR improves logic within conditional_types for unions:

    • Fix narrowing for unions #20728
17. hauntsaninja commented on Feb 5, 2026

    

    hauntsaninja

    on Feb 5, 2026

    CollaboratorAuthor

    More actions

    Thirty third PR improves narrowing with final type objects:

    • Fix narrowing with final type objects #20743
18. hauntsaninja commented on Feb 5, 2026

    

    hauntsaninja

    on Feb 5, 2026

    CollaboratorAuthor

    More actions

    Thirty fourth improves match statement narrowing against unions:

    • Further improve match statement narrowing against unions #20744
19. hauntsaninja commented on Feb 6, 2026

    

    hauntsaninja

    on Feb 6, 2026

    CollaboratorAuthor

    More actions

    Thirtyfifth fixes an issue with custom equality:

    • Fix bug when narrowing union containing custom eq against custom eq #20754
20. hauntsaninja commented on Feb 6, 2026

    

    hauntsaninja

    on Feb 6, 2026

    CollaboratorAuthor

    More actions

    Thirtysixth is unsound but we see there are some primer hits now:

    • Unsoundly narrow away from None with custom eq #20756
21. hauntsaninja commented on Feb 8, 2026

    

    hauntsaninja

    on Feb 8, 2026

    CollaboratorAuthor

    More actions

    Thirtyseventh avoids narrowing NewType:

    • Avoid narrowing to NewType #20766
22. hauntsaninja commented on Feb 11, 2026

    

    hauntsaninja

    on Feb 11, 2026

    CollaboratorAuthor

    More actions

    Thirtyeighth for irrefutable sequence patterns:

    • Better match narrowing for irrefutable sequence patterns #20782
23. hauntsaninja commented on Feb 23, 2026

    

    hauntsaninja

    on Feb 23, 2026

    CollaboratorAuthor

    More actions

    Fortieth PR is better narrowing for callables with generic return types:

    • Narrow Callable generic return types #20868
24. hauntsaninja commented on Feb 23, 2026

    

    hauntsaninja

    on Feb 23, 2026

    CollaboratorAuthor

    More actions

    Fortyfirst PR for better match narrowing for type objects:

    • Better match narrowing for type objects #20872
25. hauntsaninja commented on Feb 25, 2026

    

    hauntsaninja

    on Feb 25, 2026

    Last edited by hauntsaninja

    CollaboratorAuthor

    More actions

    Fortysecond PR improves narrowing reachability for tuple as patterns:

    • Fix match statement narrowing reachability for tuples #20896

    Fixes more issues than I expected, guess people really like matching tuples
26. hauntsaninja commented on Feb 26, 2026

    

    hauntsaninja

    on Feb 26, 2026

    CollaboratorAuthor

    More actions

    Fortythird PR for match narrowing of unions of type objects:

    • Better match narrowing for unions of type objects #20905
27. hauntsaninja commented on Feb 26, 2026

    

    hauntsaninja

    on Feb 26, 2026

    CollaboratorAuthor

    More actions

    Fortyfourth PR for match narrowing of irrefutable map patterns:

    • Better match narrowing for irrefutable mapping patterns #20906
28. hauntsaninja commented on Mar 11, 2026

    

    hauntsaninja

    on Mar 11, 2026

    CollaboratorAuthor

    More actions

    Time for changelog post: #21006
29. hauntsaninja

    closed this as completedin 537740bon Mar 11, 2026
30. hyperkai

    added a commit that references this issue on Mar 11, 2026

    Update changelog for narrowing changes (python#21006)

    ...

    537740b