Describe the bug
pandas DataFrames not consistently typed with pandas 3.0.0. Even when a dataframe is a pandera-typed dataframe within a function, it reverts to a standard pandas dataframe in the calling context.

• I have checked that this issue has not already been reported.
• I have confirmed this bug exists on the latest version of pandera.
• (optional) I have confirmed this bug exists on the main branch of pandera.

Code Sample, a copy-pastable example

The following code no longer works with pandas 3.0. This will also affect tests using pandas.testing.assert_frame_equal(), which checks dataframe type by default.

import pandas as pd
import pandera.typing as pat
from pandera.pandas import DataFrameModel, check_types

DATA_DICT = {
    'a': [1, 2, 3],
    'b': [4, 5, 6]
}


class TestSchema(DataFrameModel):
    """A simple pandera schema for testing."""

    a: int
    b: int


@check_types
def generate_test_dataframe() -> pat.DataFrame[TestSchema]:
    """Generate a test DataFrame conforming to TestSchema."""
    df = pd.DataFrame(DATA_DICT)

    typed_df = df.pipe(pat.DataFrame[TestSchema])

    print(f'Return type: {type(typed_df)}')

    return typed_df


expected_df = pat.DataFrame[TestSchema](DATA_DICT)
return_df = generate_test_dataframe()

assert isinstance(return_df, type(expected_df)), f'Expected {type(expected_df)}, got {type(return_df)}'
print('Successful completion')

pandas 2.3.2 and pandera 0.28.1 ✅

• No mypy warnings/errors
• generate_test_dataframe() prints Return type: <class 'pandera.typing.pandas.DataFrame'>
• Successful completion is printed

pandas 2.3.3 and pandera 0.28.1 ✅

• No mypy warnings/errors
• generate_test_dataframe() prints Return type: <class 'pandera.typing.pandas.DataFrame'>
• Successful completion is printed

pandas 3.0.0 and pandera 0.28.1 ❌

• No mypy warnings/errors
• generate_test_dataframe() prints Return type: <class 'pandera.typing.pandas.DataFrame'>

File prints:

Traceback (most recent call last):
  File "c:\Users\m277249\Documents\56649_Speech_AI\CASLIM\pandera_typing_check.py", line 33, in <module>
    assert isinstance(return_df, type(expected_df)), f'Expected {type(expected_df)}, got {type(return_df)}'
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
AssertionError: Expected <class 'pandera.typing.pandas.DataFrame'>, got <class 'pandas.DataFrame'>

Expected behavior

I would expect the DataFrame type to be consistent within the function, when leaving the function (especially when checked via check_types), and in the calling context.

Desktop (please complete the following information):

• OS: Windows 11
• Version: pandera 0.28.1 and pandas 3.0.0