Prepare fixup for lazy deserialization#21170
Merged
ilevkivskyi merged 5 commits intopython:masterfrom Apr 9, 2026
Merged
Conversation
This comment has been minimized.
This comment has been minimized.
JukkaL
approved these changes
Apr 9, 2026
| self.kind, self._node, self.module_public, self.implicit, self.module_hidden | ||
| ) | ||
| new.cross_ref = self.cross_ref | ||
| new.unfixed = self.unfixed |
Collaborator
There was a problem hiding this comment.
Do we need to copy stored_info here?
Member
Author
There was a problem hiding this comment.
Good catch! I think that although it is a "short lived" attribute, for consistency we should.
Contributor
|
According to mypy_primer, this change doesn't affect type check results on a corpus of open source code. ✅ |
Member
Author
|
I am going to merge this now and will enable proper lazy deserialization after #21158 is merged as well. |
ilevkivskyi
added a commit
that referenced
this pull request
Apr 13, 2026
This is a follow-up for #21170 Now that we have `modules` available as part of the global state (for the purposes of lazy deserialization), I propose to gradually move away from (IMO awkward) architecture where we pass around a lot of `named_type` callbacks, and instead use: * A library of lookup functions (we already started migrating lookup functions to `lookup.py`) * And the newly added global modules state I am not sure yet what is the best way to use instance cache in a most robust way. For now I am starting with something a bit ad-hoc. We can experiment with this, and see if this works. Note I only migrate few most problematic cases of the old callback-based pattern (one of those actually used an invalid callback). I also rename the global state file added in #21170 to have more generic name. This also removes one function-level import and few nested functions, which may make code faster with mypyc (I measure 0.5% improvement, but this is at the noise level).
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.
Together with #21158, this gets us 95% there for lazy desrialization. Implementation is generally straightforward. At some point I wanted to unify the new global state with (parts of) the checker state, and potentially get rid of a bunch of
named_typecallbacks we pass around, but decided to do this later in a separate PR, since this is not strictly necessary for parallel checking.Although this PR itself only makes fixup phase (which is the smaller part of deserialization) lazy, it already shows small performance improvements:
torchwith 4 workerstorchcc @JukkaL