Fix: DataFrame.replace incompatible with type "dict[str, dict[int, int]]"

[MarcoGorelli]

• Closes #xxxx (Replace xxxx with the Github issue number)
• Tests added: Please use assert_type() to assert the type of any return value

closes #1161

[Dr-Irv]

should be HashableT3 since it might not be the same type as HashableT2.

Same here and throughout

[MarcoGorelli]

I've made the change, but I think if someone writes df.replace({'a': 1}, {0: 9}) then that's likely a user error. Though there's likely some valid usage I can't think, so perhaps best to err on the side of not letting false positives through

[Dr-Irv]

I've made the change, but I think if someone writes df.replace({'a': 1}, {0: 9}) then that's likely a user error. Though there's likely some valid usage I can't think, so perhaps best to err on the side of not letting false positives through

Yes, I agree. While this is valid:

>>> import pandas as pd
>>> df = pd.DataFrame([["a", 1], ["b", 2]], columns=["ab", "num"])
>>> df
  ab  num
0  a    1
1  b    2
>>> df.replace({"a":"foo", 2: 20})
    ab  num
0  foo    1
1    b   20

it's not a great idea to be mixing types in one statement. With the stubs, I try to support "normal" usage of pandas. Otherwise, the stubs get too wide and don't catch common errors.

[Dr-Irv]

Looks good. When green, change from draft and ping me and I'll approve and merge

[Dr-Irv]

thanks @MarcoGorelli