Add ReferentialIntegrity metric

[frances-h]

Problem Description

As a user, I would like a metric that checks the integrity of my inter-table relationships.

Expected behavior

• Add a new column_pairs metric that calculates the percent of foreign keys values that reference a real parent value.
• This metric takes in primary key and foreign key column pairs.

Attributes

The metric should have the following attributes:

• name: 'ReferentialIntegrity'
• goal: Goal.MAXIMIZE
• min_value: 0.0
• max_value: 1.0

Methods

The metric should also define the following methods

• compute(real_data, synthetic_data): Compute the score for the metric. The returned score should be the percent of foreign key values that reference a value in the primary key column.
  • Parameters:
    • (required) real_data: a tuple of 2 pandas.Series objects. The first is the primary key column and the second is the foreign key column from the real data. (Note that this is different than other column_pair metrics)
    • (required) synthetic_data: a tuple of 2 pandas.Series objects. The first is the primary key column and the second is the foreign key column from the synthetic data. (Note that this is different than other column_pair metrics)
  • Returns: The score for the metric

>>> from sdmetrics.column_pairs import ReferentialIntegrity
>>> ReferentialIntegrity.compute(
	real_data=(real['users']['id'], real['sessions']['user_id']),
	synthetic_data=(synth['users']['id'], synth['sessions']['user_id'])
1.0
>>> ReferentialIntegrity.compute_breakdown(
	real_data=(real['users']['id'], real['sessions']['user_id']),
	synthetic_data=(synth['users']['id'], synth['sessions']['user_id'])
{ 'score': 1.0 }