Module aggregates

Expand description

Aggregates functionalities

Modules§

group_values: GroupValues trait for storing and interning group keys
no_grouping 🔒: Aggregate without grouping columns
order
row_hash 🔒: Hash aggregation
topk 🔒: TopK functionality for aggregates
topk_stream 🔒: A memory-conscious aggregation implementation that limits group buckets to a fixed number

AggregateExec: Hash aggregate execution plan
PhysicalGroupBy: Represents GROUP BY clause in the plan (including the more general GROUPING SET) In the case of a simple GROUP BY a, b clause, this will contain the expression [a, b] and a single group [false, false]. In the case of GROUP BY GROUPING SETS/CUBE/ROLLUP the planner will expand the expression into multiple groups, using null expressions to align each group. For example, with a group by clause GROUP BY GROUPING SETS ((a,b),(a),(b)) the planner should create a PhysicalGroupBy like

AGGREGATION_HASH_SEED 🔒: Hard-coded seed for aggregations to ensure hash values differ from RepartitionExec, avoiding collisions.

aggregate_expressions: Returns physical expressions for arguments to evaluate against a batch.
concat_slices: Concatenates the given slices.
create_accumulators
create_schema 🔒
determine_finer 🔒
evaluate 🔒: Evaluates expressions against a record batch.
evaluate_group_by: Evaluate a group by expression against a RecordBatch
evaluate_many: Evaluates expressions against a record batch.
evaluate_optional 🔒
finalize_aggregation: returns a vector of ArrayRefs, where each entry corresponds to either the final value (mode = Final, FinalPartitioned and Single) or states (mode = Partial)
get_aggregate_expr_req 🔒: Determines the lexical ordering requirement for an aggregate expression.
get_finer_aggregate_exprs_requirement: Gets the common requirement that satisfies all the aggregate expressions. When possible, chooses the requirement that is already satisfied by the equivalence properties.
group_id_array 🔒
merge_expressions 🔒: uses state_fields to build a vec of physical column expressions required to merge the AggregateFunctionExpr’ accumulator’s state.