Confusing error when multiple dataframe columns have the same name

[Yura52]

Description

The following snippet is a minimal example demonstrating the issue:

import numpy as np
import pandas as pd
from xgboost import DMatrix

df_numeric = pd.DataFrame(np.random.randn(10, 2))
df_categorical = pd.DataFrame(np.random.randint(0, 2, (10, 2))).astype('category')

df = pd.concat(
    [df_numeric, df_categorical],
    axis=1,
    # ignore_index=True  # <-- Uncomment to fix the issue
)
DMatrix(df, enable_categorical=True)

The above code triggers the following exception:

...
AttributeError: 'DataFrame' object has no attribute 'dtype'

The above error message does not clearly point to the root cause of the exception. Note that passing ignore_index=True to pd.concat fixes the issue, so it seems that columns with the same names (which happens without passing ignore_index) is the problem. My suggestion is to raise a more user-friendly exception in this scenario.

Software

python==3.12.9
xgboost==3.1.2

[trivialfis]

Thank you for sharing, looks like a convoluted error. I'm not sure if XGBoost internal data handling is the best place to validate duplicated column names. From my perspective, an error should be raised during the concatenation by pandas.