Converting from pandas with non-string columns can drop columns

[FBruzzesi]

What happens?

When converting from a pandas DataFrame that has non-string columns (allowed in pandas), if the pair 1 and True and/ the pair 0 and False are present, then:

1. The first appearing column name of the pair is maintained
2. The latter appearing column values of the pair is maintained

To Reproduce

import duckdb
import pandas as pd

_df = pd.DataFrame({
    1: [0, 1],
    True: ["x", "y"],
    False: [4.2, 3.14],
    0: [True, False]
})
rel = duckdb.sql("select * from _df")

rel.columns
# ['1', 'False']

rel
┌─────────┬─────────┐
│    1    │  False  │
│ varchar │ boolean │
├─────────┼─────────┤
│ x       │ true    │
│ y       │ false   │
└─────────┴─────────┘

OS:

Compiler: Clang 22.1.1 ; OS: Darwin ; Release : 25.4.0 ; Machine : arm64 ; Processor : arm

DuckDB Package Version:

1.5.2

Python Version:

3.13.12

Full Name:

Francesco Bruzzesi

Affiliation:

narwhals-dev

What is the latest build you tested with? If possible, we recommend testing with the latest nightly build.

I have tested with a stable release

Did you include all relevant data sets for reproducing the issue?

Not applicable - the reproduction does not require a data set

Did you include all code required to reproduce the issue?

• Yes, I have

Did you include all relevant configuration to reproduce the issue?

• Yes, I have

[FBruzzesi]

Sorry my bad, I saw a ghost! It's from a python dict itself (as hash(0) == hash(False) and hash(1) == hash(True)) 🤦🏼‍♂️
If a pandas DataFrame has 0, 1, True and False as columns (which is still possible) these are properly maintained (and their names converted to their string representation)