grafeo_core/execution/
mod.rs

1//! Vectorized query execution engine.
2//!
3//! Grafeo uses vectorized processing - instead of one row at a time, we process
4//! batches of ~1024 rows. This unlocks SIMD and keeps the CPU busy.
5//!
6//! | Module | Purpose |
7//! | ------ | ------- |
8//! | [`chunk`] | Batched rows (DataChunk = multiple columns) |
9//! | [`vector`] | Single column of values |
10//! | [`factorized_vector`] | Factorized vectors for avoiding Cartesian products |
11//! | [`factorized_chunk`] | Multi-level factorized chunks |
12//! | [`selection`] | Bitmap for filtering without copying |
13//! | [`operators`] | Physical operators (scan, filter, join, etc.) |
14//! | [`pipeline`] | Push-based execution (data flows through operators) |
15//! | [`parallel`] | Morsel-driven parallelism |
16//! | [`spill`] | Disk spilling when memory is tight |
17//! | [`adaptive`] | Adaptive execution with runtime cardinality feedback |
18//! | [`collector`] | Generic collector pattern for parallel aggregation |
19//!
20//! The execution model is push-based: sources push data through a pipeline of
21//! operators until it reaches a sink.
22
23pub mod adaptive;
24pub mod chunk;
25pub mod chunk_state;
26pub mod collector;
27pub mod factorized_chunk;
28pub mod factorized_iter;
29pub mod factorized_vector;
30pub mod memory;
31pub mod operators;
32pub mod parallel;
33pub mod pipeline;
34pub mod selection;
35pub mod sink;
36pub mod source;
37pub mod spill;
38pub mod vector;
39
40pub use adaptive::{
41    AdaptiveCheckpoint, AdaptiveContext, AdaptiveEvent, AdaptiveExecutionConfig,
42    AdaptiveExecutionResult, AdaptivePipelineBuilder, AdaptivePipelineConfig,
43    AdaptivePipelineExecutor, AdaptiveSummary, CardinalityCheckpoint, CardinalityFeedback,
44    CardinalityTrackingOperator, CardinalityTrackingSink, CardinalityTrackingWrapper,
45    ReoptimizationDecision, SharedAdaptiveContext, evaluate_reoptimization, execute_adaptive,
46};
47pub use chunk::{ChunkZoneHints, DataChunk};
48pub use collector::{
49    Collector, CollectorStats, CountCollector, LimitCollector, MaterializeCollector,
50    PartitionCollector, StatsCollector,
51};
52pub use memory::{ExecutionMemoryContext, ExecutionMemoryContextBuilder};
53pub use parallel::{
54    CloneableOperatorFactory, MorselScheduler, ParallelPipeline, ParallelPipelineConfig,
55    ParallelSource, RangeSource,
56};
57pub use pipeline::{ChunkCollector, ChunkSizeHint, Pipeline, PushOperator, Sink, Source};
58pub use selection::SelectionVector;
59pub use sink::{CollectorSink, CountingSink, LimitingSink, MaterializingSink, NullSink};
60pub use source::{ChunkSource, EmptySource, GeneratorSource, OperatorSource, VectorSource};
61pub use spill::{SpillFile, SpillFileReader, SpillManager};
62pub use vector::ValueVector;
63
64// Factorized execution types
65pub use chunk_state::{ChunkState, FactorizationState, FactorizedSelection, LevelSelection};
66pub use factorized_chunk::{ChunkVariant, FactorizationLevel, FactorizedChunk};
67pub use factorized_iter::{PrecomputedIter, RowIndices, RowView, StreamingIter};
68pub use factorized_vector::{FactorizedState, FactorizedVector, UnflatMetadata};
69pub use operators::{FactorizedData, FlatDataWrapper};
grafeo_core/execution/mod.rs

grafeo_core/execution/
mod.rs
