Struct AggGroup

pub struct AggGroup<S: StateStore, Strtg: Strategy> {
    group_key: Option<GroupKey>,
    states: Vec<AggState>,
    prev_outputs: Option<OwnedRow>,
    row_count_index: usize,
    _phantom: PhantomData<(S, Strtg)>,
}

AggGroup manages agg states of all agg calls for one group_key.

Fields

group_key: Option<GroupKey>

Group key.

states: Vec<AggState>

Current managed states for all AggCalls.

prev_outputs: Option<OwnedRow>

Previous outputs of aggregate functions. Initializing with None.

row_count_index: usize

Index of row count agg call (count(*)) in the call list.

_phantom: PhantomData<(S, Strtg)>

Implementations

impl<S: StateStore, Strtg: Strategy> AggGroup<S, Strtg>

pub async fn create( version: PbAggNodeVersion, group_key: Option<GroupKey>, agg_calls: &[AggCall], agg_funcs: &[BoxedAggregateFunction], storages: &[AggStateStorage<S>], intermediate_state_table: &StateTable<S>, pk_indices: &PkIndices, row_count_index: usize, extreme_cache_size: usize, input_schema: &Schema, ) -> StreamExecutorResult<Self>

Create AggGroup for the given AggCalls and group_key. For crate::executor::SimpleAggExecutor, the group_key should be None.

pub fn create_eowc( version: PbAggNodeVersion, group_key: Option<GroupKey>, agg_calls: &[AggCall], agg_funcs: &[BoxedAggregateFunction], storages: &[AggStateStorage<S>], encoded_states: &OwnedRow, pk_indices: &PkIndices, row_count_index: usize, extreme_cache_size: usize, input_schema: &Schema, ) -> StreamExecutorResult<Self>

Create a group from encoded states for EOWC. The previous output is set to None.

pub fn group_key(&self) -> Option<&GroupKey>

pub fn group_key_row(&self) -> OwnedRow

fn prev_row_count(&self) -> usize

fn curr_row_count(&self) -> usize

Get current row count of this group.

pub(crate) fn is_uninitialized(&self) -> bool

pub async fn apply_chunk( &mut self, chunk: &StreamChunk, calls: &[AggCall], funcs: &[BoxedAggregateFunction], visibilities: Vec<Bitmap>, ) -> StreamExecutorResult<()>

Apply input chunk to all managed agg states.

mappings contains the column mappings from input chunk to each agg call. visibilities contains the row visibility of the input chunk for each agg call.

fn reset( &mut self, funcs: &[BoxedAggregateFunction], ) -> StreamExecutorResult<()>

Reset all in-memory states to their initial state, i.e. to reset all agg state structs to the status as if they are just created, no input applied and no row in state table.

pub fn encode_states( &self, funcs: &[BoxedAggregateFunction], ) -> StreamExecutorResult<OwnedRow>

Encode intermediate states.

async fn get_outputs( &mut self, storages: &[AggStateStorage<S>], funcs: &[BoxedAggregateFunction], ) -> StreamExecutorResult<(usize, OwnedRow, AggStateCacheStats)>

Get the outputs of all managed agg states, without group key prefix. Possibly need to read/sync from state table if the state not cached in memory. This method is idempotent, i.e. it can be called multiple times and the outputs are guaranteed to be the same.

pub async fn build_change( &mut self, storages: &[AggStateStorage<S>], funcs: &[BoxedAggregateFunction], ) -> StreamExecutorResult<(Option<Record<OwnedRow>>, AggStateCacheStats)>

Build aggregation result change, according to previous and current agg outputs. The saved previous outputs will be updated to the latest outputs after this method.

Trait Implementations

impl<S: StateStore, Strtg: Strategy> Debug for AggGroup<S, Strtg>

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl<S: StateStore, Strtg: Strategy> EstimateSize for AggGroup<S, Strtg>

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.

impl<S: StateStore> EstimateSize for Box<AggGroup<S, OnlyOutputIfHasInput>>

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.