Confusing error when passing in an empty dataframe (with constraints)

[npatki]

Problem Description

If I accidentally pass an empty DataFrame into a fit function, I receive an error that doesn't really describe what is happening.

data = pd.DataFrame({'a': [], 'b': [], 'c': []})
metadata = SingleTableMetadata.load_from_dict({
    'columns': {
        'a': { 'sdtype': 'numerical' },
        'b': { 'sdtype': 'numerical' },
        'c': { 'sdtype': 'numerical' }
    }
})
constraint = {
    'constraint_class': 'Inequality',
    'constraint_parameters': {
        'low_column_name': 'a',
        'high_column_name': 'b'
    }
}

synthesizer = GaussianCopulaSynthesizer(metadata)
synthesizer.add_constraints(constraints=[constraint])
synthesizer.fit(data)

Output:

AggregateConstraintsError: 
index 0 is out of bounds for axis 0 with size 0

Expected behavior

The error message should be more explicit.

ValueError: Your dataset is empty.

Context

The error seems to be correct if I do not add any constraints.

data = pd.DataFrame({'a': [], 'b': [], 'c': []})
metadata = SingleTableMetadata.load_from_dict({
    'columns': {
        'a': { 'sdtype': 'numerical' },
        'b': { 'sdtype': 'numerical' },
        'c': { 'sdtype': 'numerical' }
    }
})

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

Output:

ValueError: Your dataset is empty.