Metadata auto-detection should not assign a primary key if there are NaN values in it

[R-Palazzo]

Environment Details

• SDV version: 1.9.0

Error Description

This snippet code is currently crashing because the email column is assigned as primary key while it has missing values

import pandas as pd
from sdv.metadata import SingleTableMetadata
from sdv.single_table import GaussianCopulaSynthesizer

data = pd.DataFrame({
    'email': ['[email protected]', '[email protected]', '[email protected]', None],
    'numerical': [0, 1, 2, 1],
})

metadata = SingleTableMetadata()
metadata.detect_from_dataframe(data)
synthesizer = GaussianCopulaSynthesizer(metadata)
synthesizer.fit(data)

Expected behavior

The metadata auto-detection should not set the primary key to a column that has missing values.