In Python 3.8, use the stdlib ast instead of typed_ast #6539
Conversation
…ctual test output (Python 3.8's ast changed a whole bunch of error messages from *can't* to *cannot*.)
…conditional on version
… in Python 3.8 We shouldn't need to test parser errors anyways
I changed visit_NameConstant() to set the line number, and this caused the HTML report to flag the line 'assert False' as precise (correct) rather than as empty (wrong).
(My tox-fu is too weak to make this work -- it still tries to run Python 3.7.) This reverts commit 8c24cd8.
gvanrossum
changed the title
|
At some point tomorrow I hope the Python 3.8-dev build will include python/cpython#12295 and then the tests should all pass. |
|
Oh, my review crossed with some updates and this is no more WIP. |
|
After a job restart, the tests passed (meaning python/cpython#12295 made it into Travis's "3.8-dev"). Maybe @jukka can greenlight this? |
|
@ilevkivskyi Can you review this one more time? @JukkaL told me offline he's fine with it. |
| @@ -1094,7 +1177,7 @@ def invalid_type(self, node: AST, note: Optional[str] = None) -> RawExpressionTy | |||
| def visit(self, node: ast3.expr) -> Type: ... | |||
|
|
|||
| @overload # noqa | |||
| def visit(self, node: Optional[AST]) -> Optional[Type]: ... | |||
| def visit(self, node: Optional[AST]) -> Optional[Type]: ... # noqa | |||
There was a problem hiding this comment.
It's sad this #noqa is still needed. The issue PyCQA/pyflakes#434 was fixed two weeks ago, so we will be able to update flake8 and remove these soon.
|
Does it makes sense to add a |
Remove the conditional method definition added in #6539, which mypyc doesn't support.
|
No, it’s still needed for 2.7. |
|
2.7 is included in The idea is to avoid needlessly installing a package when it's not needed, similar to how the |
|
@bluetech, I believe Guido means "It's needed to parse 2.7 on Python 3". Mypy only runs under Python 3, which means to type check Python 2 code we still need typed_ast. |
|
Ha, I misunderstood. That makes sense. |
Fixes #3871. This change fixes a few things concerning class/function definition line numbers. For users running mypy with Python 3.8+: - Function definitions are now reported on the correct line, rather than the line of the first decorator plus the number of decorators. - Class definitions are now reported on the correct line, rather than the line of the first decorator. - **These changes are fully backward compatible with respect to `# type: ignore` lines**. In other words, existing comments will _not_ need to be moved. This is accomplished by defining an ignore "scope" just like we added to expressions with #6648. See the recent discussion at #3871 for reasoning. For other users (running mypy with Python 3.7 or earlier): - Class definition lines are reported by using the same decorator-offset estimate as functions. This is both more accurate, and necessary to get 15 or so tests to pass for both pre- and post-3.8 versions. This seems fine, since there [doesn't appear](#6539 (comment)) to be a compelling reason why it was different in the first place. - **This change is also backward-compatible with existing ignores, as above.** About 15 tests had to have their error messages / nodes moved down one line, and 4 new regression tests have been added to `check-38.test`.
This required mostly changes to fastparse.py, plus a smattering of other things. The tests still don't all pass, but I believe all remaining errors are due to differences in error messages for
SyntaxErrorbetween typed_ast and Python 3.8's ast (where a bunch of improvements and spelling changes were made, including "can't" -> "cannot").To test, you need python/cpython#12295 (else it crashes due to
ast.Constantnot having akindfield). UPDATE: That landed today, March 13.UPDATE: This also requires a typeshed sync to incorporate python/typeshed#2859; I've send a PR to typeshed as #6540. UPDATE^2: Those have landed.