Metadata detection crashes when the column names are integers (AttributeError: 'int' object has no attribute 'lower')

[npatki]

Environment Details

• SDV version: 1.12.0 (latest)

Error Description

If my datafame's column names are represented with integers (instead of with strings), then the metadata detection crashes with a TypeError.

Steps to reproduce

import pandas as pd
import numpy as np

from sdv.metadata import SingleTableMetadata

# generate a dataframe with integer column names
num_rows = 100
num_cols = 20
values = {i+1: np.random.randint(0, 100, size=num_rows) for i in range(num_cols)}
data = pd.DataFrame(values)

metadata = SingleTableMetadata()
metadata.detect_from_dataframe(data=df)

AttributeError: 'int' object has no attribute 'lower'

stack_trace.txt

Additional Context

When would you have a DataFrame with integer column names? One use case is if I have a CSV that doesn't have any column names listed. If I tell pandas that there is no header in my data, then it will create column names with index values 0, 1, ...

df = pd.read_csv('my_data.csv', header=None)

Workaround

A simple workaround would be to rename the columns using strings instead of integers.