Refactor Dict function type arguments to Mapping

[rexledesma]

When writing sqlglot code against pyright, I encounter errors in my local IDE. The argument types for lineage (and probably others) are too narrow w.r.t. Dict/Mapping types.

From https://docs.python.org/3/library/typing.html#typing.Dict:

Note that to annotate arguments, it is preferred to use an abstract collection type such as Mapping rather than to use dict or typing.Dict.

Fully reproducible code snippet

from typing import Dict

from sqlglot import exp, parse_one
from sqlglot.lineage import lineage


sqlglot_sources: Dict[str, exp.Query] = {}

lineage(column="y", sql=parse_one("SELECT y from x"), sources=sqlglot_sources)

Results in the following error:

Argument of type "dict[str, Query]" cannot be assigned to parameter "sources" of type "Dict[str, str | Query] | None" in function "lineage"
  Type "dict[str, Query]" cannot be assigned to type "Dict[str, str | Query] | None"
    "dict[str, Query]" is incompatible with "Dict[str, str | Query]"
      Type parameter "_VT@dict" is invariant, but "Query" is not the same as "str | Query"
      Consider switching from "dict" to "Mapping" which is covariant in the value type
    "dict[str, Query]" is incompatible with "None"Pylance[reportArgumentType](https://github.com/microsoft/pyright/blob/main/docs/configuration.md#reportArgumentType)

[georgesittas]

If I understand this correctly, this change in the lineage prototype should suffice, right?

-    sources: t.Optional[t.Dict[str, str | exp.Query]] = None,
+    sources: t.Optional[t.Mapping[str, str | exp.Query]] = None,

[rexledesma]

I believe so. But I'm surprised there isn't an warning/error thrown by the mypy CI check. I would want the fix to incorporate that. The type signature is an easy change, but is prone to change in the future if not enforced.

FWIW, I'm using typeCheckingMode: "basic" in pyright, so I don't have any fancy configuration: https://microsoft.github.io/pyright/#/configuration?id=main-configuration-options

[georgesittas]

I believe so. But I'm surprised there isn't an warning/error thrown by the mypy CI check. I would want the fix to incorporate that. The type signature is an easy change, but is prone to change in the future if not enforced.

FWIW, I'm using typeCheckingMode: "basic" in pyright, so I don't have any fancy configuration: https://microsoft.github.io/pyright/#/configuration?id=main-configuration-options

mypy does complain if I run it against your example, we just probably didn't have a use case where the dictionary was actually restricted to only Query values so the issue wasn't noticed.