_query_expr_compiler

Defines QueryExprCompiler for compiling query expressions.

Classes

QueryExprCompiler

Compiles a list of query expressions to a single measurement object.

class QueryExprCompiler(output_measure=PureDP())

Compiles a list of query expressions to a single measurement object.

Requires that each query is a groupby-aggregation on a sequence of transformations on a PrivateSource or PrivateView. If there is a PrivateView, the stability of the view is handled when the noise scale is calculated.

A QueryExprCompiler object compiles a list of QueryExpr objects into a single object (based on the privacy framework). The Measurement object can be run with a private data source to obtain DP answers to supplied queries.

Supported QueryExprs:

• PrivateSource

• Filter

• FlatMap

• Map

• Rename

• Select

• JoinPublic

• JoinPrivate

• GroupByCount

• GroupByCountDistinct

• GroupByBoundedSum

• GroupByBoundedAverage

• GroupByBoundedSTDEV

• GroupByBoundedVariance

• GroupByQuantile

Parameters

output_measure (Union[tmlt.core.measures.PureDP, tmlt.core.measures.ApproxDP, tmlt.core.measures.RhoZCDP]) –

__init__(output_measure=PureDP())

Constructor.

Parameters

output_measure (PureDP | ApproxDP | RhoZCDPUnion[PureDP, ApproxDP, RhoZCDP] (default: PureDP())) – Distance measure for measurement’s output.

property mechanism

Return the value of Core noise mechanism.

Return type

tmlt.core.measurements.aggregations.NoiseMechanism

property output_measure

Return the distance measure for the measurement’s output.

Return type

Union[tmlt.core.measures.PureDP, tmlt.core.measures.ApproxDP, tmlt.core.measures.RhoZCDP]

static query_schema(query, catalog)

Return the schema created by a given query.

Parameters

• query (tmlt.analytics.query_expr.QueryExpr) –

• catalog (tmlt.analytics._catalog.Catalog) –

Return type

tmlt.analytics._schema.Schema

__call__(queries, privacy_budget, stability, input_domain, input_metric, public_sources, catalog, table_constraints)

Returns a compiled DP measurement and its noise information.

Parameters

• queries (Sequence[tmlt.analytics.query_expr.QueryExpr]) – Queries representing measurements to compile.

• privacy_budget (tmlt.analytics.privacy_budget.PrivacyBudget) – The total privacy budget for answering the queries.

• stability (Any) – The stability of the input to compiled query.

• input_domain (tmlt.core.domains.collections.DictDomain) – The input domain of the compiled query.

• input_metric (tmlt.core.metrics.DictMetric) – The input metric of the compiled query.

• public_sources (Dict[str, pyspark.sql.DataFrame]) – Public data sources for the queries.

• catalog (tmlt.analytics._catalog.Catalog) – The catalog, used only for query validation.

• table_constraints (Dict[tmlt.analytics._table_identifier.Identifier, List[tmlt.analytics.constraints.Constraint]]) – A mapping of tables to the existing constraints on them.

Return type

Tuple[tmlt.core.measurements.base.Measurement, tmlt.analytics._noise_info.NoiseInfo]

build_transformation(query, input_domain, input_metric, public_sources, catalog, table_constraints)

Returns a transformation and reference for the query.

Supported QueryExprs:

• Filter

• FlatMap

• JoinPrivate

• JoinPublic

• Map

• PrivateSource

• Rename

• Select

Parameters

• query (tmlt.analytics.query_expr.QueryExpr) – A query representing a transformation to compile.

• input_domain (tmlt.core.domains.collections.DictDomain) – The input domain of the compiled query.

• input_metric (tmlt.core.metrics.DictMetric) – The input metric of the compiled query.

• public_sources (Dict[str, pyspark.sql.DataFrame]) – Public data sources for the queries.

• catalog (tmlt.analytics._catalog.Catalog) – The catalog, used only for query validation.

• table_constraints (Dict[tmlt.analytics._table_identifier.Identifier, List[tmlt.analytics.constraints.Constraint]]) – A mapping of tables to the existing constraints on them.

Return type

Tuple[tmlt.core.transformations.base.Transformation, tmlt.analytics._table_reference.TableReference, List[tmlt.analytics.constraints.Constraint]]