How to resolve Union types in pd.concat?

[tvdboom]

Hi,

When running this code with mypy (note I am using the code in main, which accepts None as type for pd.concat):

data: pd.DataFrame | None = pd.DataFrame()

pd.concat([pd.DataFrame(), data])

it returns this error:

error: List item 1 has incompatible type "DataFrame | None"; expected "DataFrame" [list-item]

Now, I understand where the error comes from, the overloaded method has type Iterable[DataFrame | None] and data has type Union[DataFrame, None] instead of only DataFrame or None, hence it's incompatible. But common sense says this should work, and indeed, at runtime the code works just fine. Is there a clean/standard way of solving this error on mypy?

I understand I could just write an if-else statement to check if data is None, but I am looking for a more structural solution that could solve other similar issues, for example when data can also be a pd.Series, e.g., data: pd.Series | pd.DataFrame | None.

Thank you.

[Dr-Irv]

I just tried your example on what is currently in main and it works fine with mypy and pyright.

[tvdboom]

You are right, the example doesn't work, my bad. This example does work though, and the same question still applies.

data: pd.DataFrame | pd.Series = pd.Series()

pd.concat([pd.DataFrame(), data])

returns

error: List item 0 has incompatible type "DataFrame"; expected "None"  [list-item]
error: List item 1 has incompatible type "DataFrame | Series[Any]"; expected "None"  [list-item]

[Dr-Irv]

I created an issue #879