constraints

Defines Constraint types.

Constraints are necessary for most aggregations on tables using the AddRowsWithID ProtectedChange.

Illustrated examples using constraints can be found in the Working with privacy IDs tutorial.

Functions

simplify_constraints(constraints)

Remove redundant constraints from a list of constraints.

Given a list of the constraints on a table, produce a copy which simplifies it as much as possible by removing or combining constraints which provide overlapping information. The original list is not modified.

Parameters:

constraints (List[tmlt.analytics.constraints._base.Constraint])

Return type:

List[tmlt.analytics.constraints._base.Constraint]

Classes

Constraint	A known, enforceable fact about a table.
MaxRowsPerID	A constraint limiting the number of rows associated with each ID in a table.
MaxGroupsPerID	A constraint limiting the number of distinct groups per ID.
MaxRowsPerGroupPerID	A constraint limiting rows per unique (ID, grouping column) pair in a table.

class Constraint

Bases: abc.ABC

A known, enforceable fact about a table.

Constraints provide information about the contents of a table to help produce differentially-private results. For example, a constraint might say that each ID in a table corresponds to no more than two rows in that table (the MaxRowsPerID constraint). Constraints are applied via the QueryBuilder.enforce() method.

This class is a base class for all constraints, and cannot be used directly.

class MaxRowsPerID

Bases: tmlt.analytics.constraints._base.Constraint

A constraint limiting the number of rows associated with each ID in a table.

This constraint limits how many times each distinct value may appear in the ID column of a table with the AddRowsWithID protected change. For example, MaxRowsPerID(5) guarantees that each ID appears in at most five rows. It cannot be applied to tables with other protected changes.

max: int

The maximum number of times each distinct value may appear in the column.

class MaxGroupsPerID

Bases: tmlt.analytics.constraints._base.Constraint

A constraint limiting the number of distinct groups per ID.

This constraint limits how many times a distinct value may appear in the grouping column for each distinct value in the table’s ID column. For example, MaxGroupsPerID("grouping_column", 4) guarantees that there are at most four distinct values of grouping_column for each distinct value of ID_column.

grouping_column: str

The name of the grouping column.

max: int

The maximum number of distinct values in the grouping column for each distinct value in the ID column.

class MaxRowsPerGroupPerID

Bases: tmlt.analytics.constraints._base.Constraint

A constraint limiting rows per unique (ID, grouping column) pair in a table.

For example, MaxRowsPerGroupPerID("group_col", 5) guarantees that each ID appears in at most five rows for each distinct value in group_col.

grouping_column: str

Name of column defining the groups to truncate.

max: int

The maximum number of times each distinct value may appear in the column.