DataFrame.apply / GroupBy.apply: first parameter named `f` in stubs but `func` at runtime

[sshishov]

Describe the bug

The first parameter of DataFrame.apply (and DataFrameGroupBy.apply) is named f in pandas-stubs, but the runtime pandas signature names it func. This makes the keyword-argument form inconsistent between the stubs and runtime:

• df.apply(func=...) — valid at runtime, but rejected by the stubs (no overload variant matches).
• df.apply(f=...) — accepted by the stubs, but raises TypeError at runtime.

Only the positional form (df.apply(callable, ...)) works with both. So code that correctly calls apply(func=...) fails type checking, and code written to satisfy the stubs (apply(f=...)) crashes at runtime.

The stubs declare f (pandas-stubs/core/frame.pyi):

def apply(self, f: Callable[..., ListLikeExceptSeriesAndStr | Series], axis: AxisIndex = ...): ...

while runtime pandas declares func (pandas/core/frame.py):

def apply(self, func: AggFuncType, axis: Axis = 0, ...): ...

To Reproduce

1. Minimal example — correct at runtime, but rejected by the stubs:

import pandas as pd

df = pd.DataFrame({"a": [1, 2, 3]})
result = df.apply(func=lambda row: row["a"] * 2, axis="columns")  # runs fine at runtime

2. Type checker: mypy (2.1.0)

3. Error message from mypy:

error: No overload variant of "apply" of "DataFrame" matches argument types "Callable[[Any], Any]", "str"  [call-overload]
note:     def apply(self, f: Callable[..., ...], axis: Literal['index', 0] = ..., ...) -> ...

Conversely, the form the stubs accept crashes at runtime:

df.apply(f=lambda row: row["a"] * 2, axis="columns")   # mypy: Success: no issues found
# runtime: TypeError: DataFrame.apply() missing 1 required positional argument: 'func'

System specifications

• OS and its version: macOS 26.5.2 (Darwin 25.5.0)
• python version: 3.14.6
• type checker and its version: mypy 2.1.0
• version of installed pandas-stubs: 3.0.5.260730 (with pandas 3.0.5)

Additional context

Renaming the stub parameter from f to func (matching the runtime signature) would make the keyword form consistent across all DataFrame.apply overloads. The same mismatch affects DataFrameGroupBy.apply, where the parameter is additionally declared positional-only (/), so no keyword form works there at all.

[cmp0xff]

Thanks for the report. I can confirm in

https://github.com/pandas-dev/pandas/blob/7b0bc4da28cfd654be81e964654cf64de5f8a411/pandas/core/frame.py#L14569

and

pandas-stubs/pandas-stubs/core/frame.pyi

Line 1452 in 3570645

f: Callable[..., ListLikeExceptSeriesAndStr | Series],

PR with tests welcome.