Three of the hits look like the ternary join issue again:

Here's Spack:

def anchorify(data: dict | list) -> None:
    for key, value in data.items() if isinstance(data, dict) else enumerate(data):
        pass

Optuna:

def foo(values: list[str], names: list[str] | None):
    iterator = enumerate(values) if names is None else zip(names, values)
    for item in iterator:
        pass

PySpark:

from typing import Any

def drop(labels: Any):
    if labels is not None:
        return
    foo = []
    cols, labels = zip(*foo) if len(foo) > 0 else ([], [])
    [label for label in labels]

Ibis and psycopg show errors going away, and I think represent mypy inferring Any where it didn't before.

The error from Ibis going away looks like this:

from collections.abc import Iterable

class Value:
    def cases(self, *branches: tuple[Value, Value]) -> Value: ...

    def label(self, labels: Iterable[str]) -> Value:
        return self.cases(*enumerate(labels))

Take away the type: ignore from psycopg and the diff is:

< tests/types/test_datetime.py:785: error: Argument 1 to "time" has incompatible type "*map[int]"; expected "tzinfo | None"  [arg-type]
< tests/types/test_datetime.py:807: error: Argument 1 to "datetime" has incompatible type "*map[int]"; expected "tzinfo | None"  [arg-type]

Which comes from something like this:

import datetime

def as_time(s):
    datetime.time(*map(int, s.split(",")))

def as_naive_dt(s):
    datetime.datetime(*map(int, s.split(",")))

Following on from the revert and discussion of the related itertools MR (#12816 and #12853 ), I believe that all of these are blocking. I'll close out this MR in a day or two if nobody else has anything to say about it.

Moving this to open since the revert of #12816 wasn't necessary. I'm pretty sure the three new messages in mypy-primer will go away for mypy 1.12.

Let's update typeshed to latest mypy to confirm before merging this!

Also this class of change probably makes type checkers slower, since structural checks are more expensive than nominal checks

Diff from mypy_primer, showing the effect of this PR on open source code:

psycopg (https://github.com/psycopg/psycopg)
+ tests/types/test_datetime.py:785: error: Unused "type: ignore" comment  [unused-ignore]
+ tests/types/test_datetime.py:807: error: Unused "type: ignore" comment  [unused-ignore]

ibis (https://github.com/ibis-project/ibis)
- ibis/expr/types/numeric.py:1223: error: Argument 1 to "cases" of "Value" has incompatible type "*enumerate[str]"; expected "tuple[Value, Value]"  [arg-type]