Micha Reiser comments

Results 1092 comments of


                                            Micha Reiser

[ty] implement `typing.TypeGuard`

Thank you for your continuous work on this. Almost the entire team is out this week. It may take a few days before someone finds time to review your PR....

Respect `fmt: skip` for multiple statements on same line

Regarding the benchmark. I don't think using hyperfine will give you meaningful numbers. There's just too much overhead in ruff's cli (config discovery, directory walking, io) in addition to the...

[ty] infer function's return type

This PR seems to blow up the runtime for the large wall-time benchmark.

[ty] Handle covariant "deep" mutual typevar constraints

> When limited to TY_MAX_PARALLELISM=1, it doesn't time out at all. That's scary. It could mean that There's a fixpoint issue (where salsa fails to establish an owner thread). Or...

Go-to-definition jumps to typeshed stubs for datetime module import

Can you share an example? I'm not entirely sure if I understand what you mean by cache. The LSP also differentiates between go to definition and go to declaration: *Go...

Go-to-definition jumps to typeshed stubs for datetime module import

Thanks for the example. ty correctly jumps to `datetime.py` when clicking on the name after `from` but we always jump to the `pyi` file when clicking on `datetime` after the...

Go-to-definition jumps to typeshed stubs for datetime module import

Makes sense. I'm not sure then how Pylance resolves `datetime` to `_pydatetime.py`

Go-to-definition jumps to typeshed stubs for datetime module import

Oh, Pylance emulates: ```py try: from _datetime import * from _datetime import __doc__ except ImportError: from _pydatetime import * from _pydatetime import __doc__ ``` That's interesting

Go-to-definition jumps to typeshed stubs for datetime module import

This is related to [#2096](https://github.com/astral-sh/ty/issues/2096)

[ty] WIP: sort union types during cycle recovery to stabilize output

> Hmm, at least, this approach alone doesn't seem to eliminate the instability of the diagnostic itself, rather than the instability of the output type. I think you need to...