Google sync #1040
Merged
Google sync #1040
Conversation
This file contains 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
Allows setting up a dependency tree before checking a file. Python files in the dependency tree are analysed in order, and the corresponding pyi files generated. The pythonpath for all files is set to the root of the dependency tree. Also converts one test file as a more extended test of the new functionality. PiperOrigin-RevId: 406256121
Error.message contains Error._message, _details, and _traceback combined. So copy_from() passing in one error's message attribute as another's _message attribute caused a lot of information to be duplicated. PiperOrigin-RevId: 406272568
The skip in overlays/typing_overlay.py was only needed until the typed_ast-based pyi parser went live, which happened months ago. Once a type annotation is added to the long list in pytd/slots.py, pytype analysis completes in a negligible amount of time. PiperOrigin-RevId: 406371373
Such aliases aren't currently common enough for this to be a problem, but an upcoming bug fix that adds a bunch of aliases exposes these issues: * In LateType resolution, we need to first check if the entire late type is a module name before assuming it's a module attribute. * convert._create_module needs to strip only the module prefix; taking off everything before the last dotted part is wrong for nested classes. * The pyi parser shouldn't allow imports like 'from x import typing' to change the meaning of 'from typing import ...'. * The pyi parser was rewriting __builtin__ to builtins too aggressively; we only want to get rid of imports like "import __builtin__"; lines like "import builtins as __builtin__" should not be changed. * The pytd printer needs to handle the case when it is passed a partially resolved ast in which modules are still represented as NamedType, not Module. * Also in the pytd printer, we were accidentally outputting "from x import y as y" when we meant to just output "from x import y". * serialize_ast needs to undo the pytd printer's application of module aliases. (It would be better if serialize_ast could just use the existing ast rather than printing and reparsing it, but fixing that is a much bigger endeavor.) PiperOrigin-RevId: 406413484
…g check. We have some logic that checks whether a TypeVar rename is ok, but it would emit [not-supported-yet] when trying to compare two bounded/constrained TypeVars against each other. The matcher is able to handle this case now, so do the comparison instead of complaining. PiperOrigin-RevId: 406424691
pytype/pyi:parser_test is still timing out. It seems to be a single test in particular that's very slow, so pulling that test out into its own file to ensure that nothing else gets put into the same shard as it seems to work. PiperOrigin-RevId: 406433991
The vm subclass now lives in tracer_vm.py, not analyze.py. PiperOrigin-RevId: 406867564
This kicks in whenever an enum class has its own `__init__` or `__new__` that adds attributes to members, and that enum is used in an annotation or other context that causes a canonical member to be created. PiperOrigin-RevId: 406868258
* Recognize LateTypes as modules in the pytd printer. Otherwise, it's inconsistent whether an aliased import gets printed as an import or an alias. * Also in the pytd printer, add a heuristic for figuring out which prefix of a dotted NamedType is the module name. If we don't do this, we generate illegal imports containing class names for pb2 nested classes. * Fix a bug where pytype crashed on an empty Awaitable type. Improving pyi output exposes this bug by precisely typing code in which types were previously all Any. PiperOrigin-RevId: 406869267
Currently, pytype adds "x: module" constants to pyi files, and all attributes on builtins.module are typed as Any, so accessing whatever_module.x.attribute always produces Any. This is fixed by instead adding "import x" statements and changing the pyi parser to always generate aliases for import statements. This is a very large change because of all the changes to test output and the places we need to plumb through --gen-stub-imports so that the pyi parser sees it. The actually interesting changes are in: * pytype/output.py * pytype/pyi/modules.py * pytype/pytd/printer.py * pytype/pytd/pytd_visitors.py PiperOrigin-RevId: 406940492
PiperOrigin-RevId: 407145181
This is to unblock python/typeshed#6221. Also see #786. PiperOrigin-RevId: 407168815
PiperOrigin-RevId: 407173206
PiperOrigin-RevId: 407176581
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
No description provided.