[ty] Inform literal promotion with variance of inferred argument types#22322
Closed
ibraheemdev wants to merge 5 commits intomainfrom
Closed
[ty] Inform literal promotion with variance of inferred argument types#22322ibraheemdev wants to merge 5 commits intomainfrom
ibraheemdev wants to merge 5 commits intomainfrom
Conversation
Diagnostic diff on typing conformance testsNo changes detected when running ty on typing conformance tests ✅ |
|
CodSpeed Performance ReportMerging this PR will degrade performance by 52.98%Comparing Summary
Performance Changes
Footnotes
|
ibraheemdev
force-pushed
the
ibraheem/variance-in-argument
branch
2 times, most recently
from
f5cfc73 to
e169795
Compare
|
| Lint rule | Added | Removed | Changed |
|---|---|---|---|
invalid-argument-type |
16 | 9 | 63 |
invalid-return-type |
3 | 12 | 7 |
invalid-assignment |
0 | 6 | 7 |
possibly-missing-attribute |
11 | 1 | 1 |
unresolved-attribute |
0 | 11 | 0 |
unused-ignore-comment |
0 | 2 | 0 |
no-matching-overload |
0 | 1 | 0 |
redundant-cast |
1 | 0 | 0 |
unsupported-operator |
0 | 0 | 1 |
| Total | 31 | 42 | 79 |
ibraheemdev
force-pushed
the
ibraheem/variance-in-argument
branch
5 times, most recently
from
8268bc6 to
06f17fd
Compare
ibraheemdev
force-pushed
the
ibraheem/variance-in-argument
branch
from
e046945 to
fba4ac1
Compare
Member
Author
|
There seems to be some subtle changes in the ecosystem that I have yet to look into, but I think this is ready for an initial review. The overall problem that this is trying to solve is: if we infer a type variable The approach taken in this PR transforms the formal and actual types to contain synthetic type variables and recursing one generic layer at a time, keeping track of the variance of the inferred type along the way. This is very similar to how we perform reverse inference. Alternatively, we might be able to transform the actual type such that each type is uniquely identified, e.g., wrapping each type in a synthetic type alias?. This seems to have a quite significant performance and memory usage impact (though there are some optimizations that could be made here), but I'm wondering if there's a simpler approach that I am missing. |
ibraheemdev
requested review from
AlexWaygood,
carljm,
dcreager and
sharkdp
as code owners
ibraheemdev
commented
Jan 9, 2026
|
|
||
| # TODO: we should avoid literal promotion here. | ||
| # error: [invalid-argument-type] "Argument to function `f12` is incorrect: Expected `(int, /) -> None`, found `def takes_literal[T](x: Literal[1]) -> Unknown`" | ||
| reveal_type(f12(takes_literal)) # revealed: Invariant[int] | None |
Member
Author
There was a problem hiding this comment.
This is a regression from main because we don't currently support the synthetic type mapping on Callable types, but this shouldn't be too hard to fix.
Contributor
|
@dcreager I wonder if you can take a look at this? |
This was referenced Jan 30, 2026
ibraheemdev
added a commit
that referenced
this pull request
Feb 19, 2026
…23107) Updates our literal promotion heuristics to only promote implicitly inferred literal values, i.e., literal values inferred without an explicit literal annotation. This requires us to track promotability on the literal value type itself, and so a lot of the changes here are churn from having to split the literal value variants of `Type` into a separate `LiteralValue` type. Another consequence of this change is that the same literal value does not compare equal in its promotable and unpromotable form, so there are a few cases where we relied on type equality (or hashing) that need to be changed. Resolves astral-sh/ty#2664, astral-sh/ty#1815, and supersedes #22322.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
This also indirectly implements #21905. Similar how to we perform reverse inference, we can perform regular inference by recursing one generic layer at a time, and keep track of the variance manually.
Resolves astral-sh/ty#1815.