Implementation:Sdv dev SDV FixedCombinations Init

Knowledge Sources	SDV SDV Documentation
Domains	Data_Quality, Constraint_Satisfaction
Last Updated	2026-02-14 00:00 GMT

Overview

Concrete tool for enforcing fixed value combinations across columns in synthetic data, provided by the SDV library.

Description

The FixedCombinations constraint ensures only value tuples seen during training appear in synthetic output. It requires at least two columns and works with categorical or boolean columns.

Usage

Create a FixedCombinations instance with the list of column names and add it to a synthesizer.

Code Reference

Source Location

• Repository: SDV
• File: sdv/cag/fixed_combinations.py
• Lines: L21-56

Signature

class FixedCombinations(BaseConstraint):
    def __init__(self, column_names, table_name=None):
        """
        Args:
            column_names (list[str]): Columns that must maintain fixed combinations.
                Must contain at least two columns.
            table_name (str or None): Target table (multi-table). Defaults to None.
        """

Import

from sdv.cag import FixedCombinations

I/O Contract

Inputs

Name	Type	Required	Description
column_names	list[str]	Yes	List of columns (minimum 2) with fixed valid combinations
table_name	str or None	No	Target table for multi-table data

Outputs

Name	Type	Description
instance	FixedCombinations	Constraint object ready to add to a synthesizer

Usage Examples

from sdv.cag import FixedCombinations

constraint = FixedCombinations(column_names=['country', 'city'])

synthesizer.add_constraints([constraint])
synthesizer.fit(data)
synthetic_data = synthesizer.sample(num_rows=1000)

Related Pages

Implements Principle

• Principle:Sdv_dev_SDV_Fixed_Combinations_Constraint

Requires Environment

• Environment:Sdv_dev_SDV_Python_Runtime