A validated, compiled expression callable. Parses a Python expression string, validates it uses only safe constructs, and compiles it into a call

A single aggregation measure. Args: field: Input field name to aggregate. agg: The aggregation operator. alias: Output name for this me

Windowed grouped aggregation configuration. Args: window: Window configuration. group_by: Field names for grouping. Empty list means global

Complete metric computation specification. Defines how to transform raw row dicts into a single numeric metric value suitable for anomaly detecti

In-memory stand-in for a Beam ReadModifyWriteStateSpec, used to unit-test _RateLimitAlerts without a real pipeline. Matches the contract the DoFn

Per-key session-window debounce for alerts. ``window_end`` is "now". Fires iff first event for the key OR gap from the prior event ≥ cooldown. ``

Window configuration for metric aggregation. Args: type: FIXED or SLIDING window. size_seconds: Window size in seconds. period_seconds:

Renders an anomaly for Pub/Sub as UTF-8 bytes. If ``message_format`` is set, emits ``message_format.format(...)``. Otherwise emits a default JSON

Configuration for the RelativeChange detector. At least one of ``threshold_pct`` or ``absolute_threshold`` must be provided. If both are provided

Transforms raw row dicts into metric beam.Rows for anomaly detection. Takes a ``PCollection[dict]`` with event-time timestamps and produces eithe

Incremental sliding window mean tracker. Uses the same algorithm as ``apache_beam.ml.anomaly.univariate.mean.IncSlidingMeanTracker``: maintains

Converts all AnomalyResult elements to BQ row dicts.

Restriction carrying BQ Storage stream names for cross-worker safety. Unlike a plain OffsetRange(0, N), this restriction is self-contained: each

All mutable per-key state for the timestamp buffer. Stored as a single pickled object in ReadModifyWriteState. All timestamp fields use Beam ``Ti

A DoFn that copies input files to a GCS destination. It skips files that are already on GCS (starting with 'gs://'). It uses a bundle-level ran

Mapper-side pre-aggregation within each bundle. Mirrors PGBKCVOperation (operations.py) but as a user-level DoFn so Dataflow still sees the downs

POSTs each anomaly to ``webhook_spec.endpoint`` using ADC. String leaves in ``body`` and ``headers`` are format-substituted with ``anomaly fields

Emits AnomalyResult(label=1) when ``expression`` is truthy, else label=0. ``value`` is bound to the metric value. The output matches the shape of

Pre-aggregation column derivation via expression. Args: name: Name of the new field to create. expression: A compiled ``Expr`` callable, e.

Strategy for aggregation parallelism. Applies to both keyed and global. NONE: Plain Combine, no fanout. Relies on combiner lifting for mapper-side

MatchIsComputed is a Matcher for tfjson.SchemaAttribute Computed attribute.

Stateful DoFn that detects relative changes between windows. Subclasses ``TimestampBufferDoFnBag`` for the buffer/timer/trim machinery. Uses the

OffsetRestrictionTracker that allows checkpointing but prevents splitting. Checkpointing (fraction=0) is required for defer_remainder(). All other

Stage 1 wrapper: extract_output returns the raw accumulator.

Minimal stand-in for a requests Response used by DoFn tests.

Captures the most recent request() call for assertions.

Aggregation operator.

Buffer reader backed by a materialized BagState.

A PTransform that copies files to GCS. It takes a PCollection of objects with a `path` attribute and copies non-GCS files to the specified GCS de

Buffer reader backed by OrderedListState.

Key that pairs an optional grouping key with a window offset. Used to route each sliding-window offset to an independent detector state. Fixed wi

Streaming source for BigQuery change history. Continuously polls BigQuery APPENDS() or CHANGES() functions and emits changed rows as an unbounded

Window type for metric aggregation.

Parse each line of input text into words.

DoFn that evaluates a post-aggregation expression on combined results.

WatermarkEstimatorProvider that initializes the hold from _QueryResult. Uses range_start from the element to set the initial watermark hold. This

Stateful DoFn that deletes temp tables after all streams are read. Receives cleanup signals from the Read SDF as: (table_key, (streams_read_cou

Decompress and convert raw Arrow batches to timestamped row dicts. Receives individual (schema_bytes, batch_bytes) tuples after Reshuffle and con

Callable that evaluates derived field expressions on each row dict. Each derived field's ``expression`` is a compiled ``Expr`` callable. This cla

Executes a BQ CHANGES/APPENDS query from a _QueryRange instruction.

In-memory stand-in for a Beam BagStateSpec. Matches the contract AsyncWrapper expects (``add``, ``clear``, ``read``). Modeled directly on the hel

Logs each AnomalyResult at WARNING level for visibility in Dataflow.

SDF that periodically emits _QueryRange instructions. Uses defer_remainder() for poll timing and _PollWatermarkEstimator to control the watermark

Input element for the polling SDF. Only contains start_time (Beam Timestamp), which _PollWatermarkEstimatorProvider uses to initialize the waterm

Watermark estimator that tracks both a watermark hold and poll cursor. The watermark hold (reported via current_watermark) is set to start_ts: th

Provider for _PollWatermarkEstimator. Initializes with watermark hold at start_time and poll cursor at start_time (first poll will query from sta

Stage 2 wrapper: add_input merges an accumulator from Stage 1.

Lightweight instruction emitted by the polling SDF. Contains only the time range to query as Beam Timestamps (int microseconds internally). Stati

Bridges the Query step (query execution) to the Read SDF. After _ExecuteQueryFn runs a CHANGES/APPENDS query, it emits a _QueryResult pointing to

SDF that reads a temp table via BigQuery Storage Read API. Note on SDF lifecycle: the runner decomposes this SDF into three internal wrapper DoFn

Session stub that returns a scripted outcome per request() call. Each entry in ``outcomes`` is either an int HTTP status (returns a ``_StubRespon

Tracker for _StreamRestriction, delegating offset logic to OffsetRestrictionTracker.

Abstract class for JDBC source connectors. Handles sharded vs single instance migration.

Abstract class covering the common operations between Classic and Flex templates. <p>Generally, the methods here are the ones that focus more on the

Unit tests for {@link AbstractPipelineLauncher}.

{@code AccumulateListFn<InputT, OutputListT>} specifies how to combine a collection of input values of type {@code InputT} to a single list of values

Test class for {@link AccumulateListFn}.

Helper to {@link ReaderTransform} that expands the Pipeline Graph to flatten all the tableReaders and generate table completions. Please refer to {@li

Test class for {@link AccumulatingTableReader}.

Context for actions.

Factory providing indirection to action handler.

Helper class for parsing legacy json into {@link Action}. @deprecated use the current JSON format instead

Factory providing indirection to action handler.

Tests that ComputeMetric + ZScore pipeline produces correct aggregations. For each (agg_type, window_type, keyed) combination, we feed deterministi

And is a Matcher that applies when all of its containing Matcher instances Match.

Tests for _anomaly_id used by AsyncWrapper for per-element dedup.

Pipeline options for the anomaly monitor.

<h3>Configuration</h3> <p>The connector expects configuration to be passed via Properties files. By default, the connector will look for a properties

A Beam transform that applies the DDL statements passed in a Cloud Spanner database and outputs the original {@link Ddl}.

DoFn to apply schema overrides from a HOCON/JSON config file.

A {@link DoFn} that applies a JavaScript UDF to the document.

Generates a <a href=https://avro.apache.org/docs/1.8.2/spec.html#Decimal>Decimal</a> Avro Type.

Represents a single artifact. <p>An "artifact" is an entity in object storage, file storage, or block storage. Artifacts should be able to be stored

Interface for working with test artifacts. <p>It is the responsibility of implementations to make sure that artifacts are kept separate from each oth

Artifact registry image spec.

Type of job flow artifact object.

Utilities for working with test artifacts.

Artifacts for {@link ArtifactUtils}.

Subject that has assertion operations for artifact lists (GCS files), usually coming from the result of a template.

Combines all elements in the {@link PCollection} as list.

Tests for AsList class.