Module aggregate 

Aggregation functions for query results.

Implements core aggregation operations on Arrow arrays: count, sum, min, max, avg, first, last.

Supports group-by aggregation partitioned by tag columns.

Memory model

StreamingAggregator is the form to reach for over a lazy scan: batches are pushed in and dropped, and the state held is one accumulator set per group. Peak memory is therefore O(groups × fields) and independent of how many rows were aggregated — a SELECT avg(x) over a week of 1 s data produces its single row without ever holding the week.

The whole-slice entry points (aggregate_batch, aggregate_grouped, aggregate_grouped_streaming) are conveniences for callers that already hold their input.

The output is still a single RecordBatch, so a group-by over a very high-cardinality column is bounded by the group count rather than the row count. Pass a MemoryTracker to fail early rather than OOM in that case.

Structs

DecimalTrack

The exact half of an accumulator, allocated only for decimal columns.

StreamingAggregator

Constant-memory aggregation state that batches are pushed into.

Enums

AggFn

Aggregation function type.

AggResult

One aggregate’s result, in the type that keeps it exact.

AggregationStrategy

Aggregation strategy chosen based on estimated group-key cardinality.

Num

A value pulled out of a column, in the type that keeps it exact.

Constants

AVG_EXTRA_SCALE

Extra fractional digits an average of a decimal column is computed to.

Functions

agg_fn_name

Human-readable function name for column naming.

agg_result_column

Build one output column from a column of results.

aggregate_batch

Apply aggregation to a RecordBatch without grouping.

aggregate_grouped

Apply aggregation with group-by support.

aggregate_grouped_streaming

Streaming group-by aggregation over multiple RecordBatches.

aggregate_sorted

Sort-based group-by aggregation.

choose_strategy

Choose an aggregation strategy based on estimated cardinality.