Struct MaterializedInputState

pub struct MaterializedInputState {
    arg_col_indices: Vec<usize>,
    state_table_arg_col_indices: Vec<usize>,
    order_col_indices: Vec<usize>,
    state_table_order_col_indices: Vec<usize>,
    cache: Box<dyn AggStateCache + Send + Sync>,
    output_first_value: bool,
    cache_key_serializer: OrderedRowSerde,
}

Aggregation state as a materialization of input chunks.

For example, in string_agg, several useful columns are picked from input chunks and stored in the state table when applying chunks, and the aggregation result is calculated when need to get output.

Fields

arg_col_indices: Vec<usize>

Argument column indices in input chunks.

state_table_arg_col_indices: Vec<usize>

Argument column indices in state table, group key skipped.

order_col_indices: Vec<usize>

The columns to order by in input chunks.

state_table_order_col_indices: Vec<usize>

The columns to order by in state table, group key skipped.

cache: Box<dyn AggStateCache + Send + Sync>

Cache of state table.

output_first_value: bool

Whether to output the first value from cache.

cache_key_serializer: OrderedRowSerde

Serializer for cache key.

Implementations

impl MaterializedInputState

pub fn new( version: PbAggNodeVersion, agg_call: &AggCall, pk_indices: &PkIndices, order_columns: &[ColumnOrder], col_mapping: &StateTableColumnMapping, extreme_cache_size: usize, input_schema: &Schema, ) -> StreamExecutorResult<Self>

Create an instance from AggCall.

pub fn apply_chunk(&mut self, chunk: &StreamChunk) -> StreamExecutorResult<()>

Apply a chunk of data to the state cache. This method should never involve any state table operations.

pub async fn get_output( &mut self, state_table: &StateTable<impl StateStore>, group_key: Option<&GroupKey>, func: &BoxedAggregateFunction, ) -> StreamExecutorResult<(Datum, AggStateCacheStats)>

Get the output of the state. We may need to read from the state table into the cache to get the output.

Trait Implementations

impl EstimateSize for MaterializedInputState

fn estimated_heap_size(&self) -> usize

The estimated heap size of the current struct in bytes.

fn estimated_size(&self) -> usize

where Self: Sized,

The estimated total size of the current struct in bytes, including the estimated_heap_size and the size of Self.