Metadata auto-detection should ensure primary keys are unique (special sdtypes are not exempt from this rule!)

[npatki]

Environment Details

• SDV version: 1.11.0

Error Description

In metadata auto-detection script: If a column is a real-world sdtype (such as latitude), then the auto-detection seems willing to make it a primary key even if its values are not unique. This is a problem because primary keys are all expected to be unique. Metadata auto-detection should never make something a primary key if it contains repeating values.

Steps to reproduce

In the example below, column latitude has repeating values. There is no primary key:

import pandas as pd
import numpy as np

from sdv.metadata import SingleTableMetadata

data = pd.DataFrame(data={
    'Age': [int(i) for i in np.random.uniform(low=0, high=100, size=100)],
    'Sex': np.random.choice(['Male', 'Female'], size=100),
    'latitude': [round(i, 2) for i in np.random.uniform(low=-90, high=+90, size=50)] * 2
})

metadata = SingleTableMetadata()
metadata.detect_from_dataframe(data)

The metadata detection marks latitude as a primary key column even though it contains repeating values. This should not be possible. As a result, the validation fails.

metadata.validate_data(data)

InvalidDataError: The provided data does not match the metadata:
Key column 'latitude' contains repeating values: [-1.43, -12.66, -18.24, '+ 47 more']

Other Context

1. Another rule for primary keys is that they must be non-null. (Seee Metadata auto-detection should not assign a primary key if there are NaN values in it #1740). This rule is always working for all sdtypes, as expected

2. This issue only happens with real-world sdtypes. If I change the name of latitude column to something else, then the detected sdtype is no longer latitude but numerical. In this case, the script correctly exempts this column from being a primary key.