Multiple 1.4.1 type stub issues

[dangotbanned]

Downstream in narwhals, I've got a PR open which is trying to adjust to the new stubs:

• fix(DONT MERGE): duckdb>=1.4.1 typing & warnings narwhals-dev/narwhals#3189

From #96 the issues we have seem to stem from (236713a) and (f5d4e42) after which we get a lot of false negatives on:

• ConstantExpression
• LambdaExpression
• DuckDBPyType

Question

Does anyone have a preference for splitting these into individual issues/PRs?

I think the fixes I have for ConstantExpression + LambdaExpression would be pretty minor.

There's 1 or 2 examples for ConstantExpression in the test suite already 😉

Show ConstantExpression

DuckDBPyType might be a little more involved, since it's an issue with children.
Also the way I've worked around it in narwhals probably wouldn't be the same thing I'd write in the stubs

Anyways, thought I'd ask first before opening lots of issues 😄

[evertlammerts]

Hey @dangotbanned, thanks for the extensive discovery! I think a single issue / PR is completely fine for this.

Expressions are kind of a minefield I guess, typewise.

[evertlammerts]

Thanks again for this report. It became #131, and #323 fixed it in February. I rechecked on 2.0.0: ConstantExpression now accepts any value, a duckdb.Value included, and LambdaExpression takes a tuple of names or expressions, both clean under mypy in strict mode. The one loose spot left is DuckDBPyType.children, which is still a flat union per entry; #349 tracks making it precise. I am closing this discussion in favor of that one.