ENH: Split assert_frame_equal's check_exact flag for floats and integers

[kcwerther]

Feature Type

• Adding new functionality to pandas

• Changing existing functionality in pandas

• Removing existing functionality in pandas

Problem Description

Consider the following 3 dataframes:
reference_df = pd.DataFrame([{"id": 8000000, "value": 0.000000123}])
correct_df = pd.DataFrame([{"id": 8000000, "value": 0.000000124}])
incorrect_df = pd.DataFrame([{"id": 8000001, "value": 0.000000123}])

I would like to be able to use assert_frame_equal with the following results:
assert_frame_equal(reference_df, correct_df) passes
assert_frame_equal(reference_df, incorrect_df) fails

Feature Description

I would like to split out the check_exact flag so that check_exact can be independently set for floats and integers, with separate tolerance values for each. The new flags would be:
check_exact_float=False
rtol_float=1e-5
atol_float=1e-5
check_exact_int=False
rtol_int=1e-5
atol_int=1

With these new flags, the above problem would be solved with the following:
assert_frame_equal(reference_df, correct_df, check_exact_int=True) passes
assert_frame_equal(reference_df, incorrect_df, check_exact_int=True) fails

Alternative Solutions

I'm not aware of any existing code that would do this, but one can convert the integers into strings to allow exact integers but not exact floats, as shown below.

def compare_exact_ints_non_exact_floats(df1, df2):
    df1_workaround = df1.astype({col: "string" for col in df1.select_dtypes("int").columns.to_list()})
    if str(df1_workaround.index.dtype).startswith("int"):
        df1_workaround.index = df1_workaround.index.astype("string")

    df2_workaround = df2.astype({col: "string" for col in df2.select_dtypes("int").columns.to_list()})
    if str(df2_workaround.index.dtype).startswith("int"):
        df2_workaround.index = df2_workaround.index.astype("string")

    pd.testing.assert_frame_equal(df1_workaround, df2_workaround)

Additional Context

No response