The `Cardinality` property is not returning a DataFrame

Environment Details

SDMetrics version: Dev release in master branch (for upcoming SDMetrics 0.11.0)

Error Description

In the new version of SDmetrics, we are renaming the 'Table Relationships' property to 'Cardinality' but it should otherwise behave the same. However, I am seeing that get_details for this property is returning a dictionary instead of a DataFrame, as expected.

Steps to reproduce

Reproduce using the demo data.

from sdmetrics import load_demo
from sdmetrics.reports.multi_table import QualityReport

real_data, synthetic_data, metadata = load_demo(modality='multi_table')

report = QualityReport()
report.generate(real_data, synthetic_data, metadata)
report.get_details(
    property_name='Cardinality',
    table_name='sessions'
)

Output:

{('sessions', 'transactions'): {'score': 0.7}}

(Note that one of the relationships is also missing from this output. But this is being covered in #404)

Expected Behavior

Nothing should change from the older version except for the property name ('Table Relationships' --> 'Cardinality'). In the older version, the same method call returns this:

@npatki What is the expected behavior for Column Pair Trends and Column Shapes when no table_name is passed? Because right now those also return a dictionary of dataframes, so it's weird that Cardinality returns a dataframe and the others don't.

Below are examples of what's returned, just to clarify things.

    # Assert Column Shapes details
    pd.testing.assert_frame_equal(details[0], pd.DataFrame({
        'Column': ['col2', 'col3'],
        'Metric': ['TVComplement', 'TVComplement'],
        'Score': [.75, .75]
    }))

    # Assert Column Pair Trends details
    pd.testing.assert_frame_equal(details[1], pd.DataFrame({
        'Column 1': ['col2'],
        'Column 2': ['col3'],
        'Metric': ['ContingencySimilarity'],
        'Score': [.25],
        'Real Correlation': [np.nan],
        'Synthetic Correlation': [np.nan],
    }))

    # Assert Cardinality details, with and without table name
    expected = pd.DataFrame({
        'Child Table': ['table1'],
        'Parent Table': ['table2'],
        'Metric': ['CardinalityShapeSimilariy'],
        'Quality Score': [.75]
    })
    pd.testing.assert_frame_equal(details[2], expected)
    pd.testing.assert_frame_equal(details[5], expected)

    # Assert Column Shapes details without table_name. NOTE: this is a dict
    pd.testing.assert_frame_equal(details[3]['table1'], pd.DataFrame({
        'Column': ['col2', 'col3'],
        'Metric': ['TVComplement', 'TVComplement'],
        'Score': [.75, .75]
    }))
    pd.testing.assert_frame_equal(details[3]['table2'], pd.DataFrame({
        'Column': ['col4', 'col5', 'col7'],
        'Metric': ['KSComplement', 'KSComplement', 'KSComplement'],
        'Score': [.75, .75, 1]
    }))

    # Assert Column Pair Trends details without table_name. NOTE: this is a dict
    pd.testing.assert_frame_equal(details[4]['table1'], pd.DataFrame({
        'Column 1': ['col2'],
        'Column 2': ['col3'],
        'Metric': ['ContingencySimilarity'],
        'Score': [.25],
        'Real Correlation': [np.nan],
        'Synthetic Correlation': [np.nan],
    }))
    pd.testing.assert_frame_equal(details[4]['table2'], pd.DataFrame({
        'Column 1': ['col4', 'col4', 'col5'],
        'Column 2': ['col5', 'col7', 'col7'],
        'Metric': ['CorrelationSimilarity', 'CorrelationSimilarity', 'CorrelationSimilarity'],
        'Score': [0.9901306731066666, 0.9853027960145061, 0.9678805694257717],
        'Real Correlation': [0.946664, 0.966247, 0.862622],
        'Synthetic Correlation': [0.926925, 0.936853, 0.798384],
    }))

@fealho, according to our docs, we say:

If [table_name is] provided, you'll receive filtered results for the table.

So the expected behavior is to return unfiltered results if you don't provide a table_name. That is, return 1 single dataframe with all the results from all tables. I think the discrepancy is that we're missing a Table column in all cases to identify the table name.

If you provide a table_name, the Table column will contain a constant value
If you do not provide a table_name, you will get all results so the Table column will vary

Note that for single table, we should not add the Table column (leave as is)

sdv-dev / SDMetrics