exec

package
v1.1.0-beta.0...-e953a09 Latest Latest
Warning

This package is not in the latest version of its module.

 Go to latest Published: Sep 30, 2026 License: Apache-2.0 Imports: 26 Imported by: 0

Documentation

Index

Constants

This section is empty.

Variables

This section is empty.

Functions

func Close 

func Close(e Executor) (err error)

Close is a wrapper function on e.Close(), it handles some common codes.

func NewFirstChunk 

func NewFirstChunk(e Executor) *chunk.Chunk

NewFirstChunk creates a new chunk to buffer current executor's result.

func Next 

func Next(ctx context.Context, e Executor, req *chunk.Chunk) (err error)

Next is a wrapper function on e.Next(), it handles some common codes.

func Open 

func Open(ctx context.Context, e Executor) (err error)

Open is a wrapper function on e.Open(), it handles some common codes.

func RetTypes 

func RetTypes(e Executor) []*types.FieldType

RetTypes returns all output column types.

func TryNewCacheChunk 

func TryNewCacheChunk(e Executor) *chunk.Chunk

TryNewCacheChunk tries to get a cached chunk

Types

type AdaptiveLimitConfig 

type AdaptiveLimitConfig struct {
	// DemandRows is the final-row demand derived from LIMIT offset plus count.
	DemandRows uint64
	// InitialOuterWindow is the starting outer-row admission window.
	InitialOuterWindow uint64
	// MaxOuterWindow is the upper bound of the outer-row admission window.
	MaxOuterWindow uint64
	// InitialLookupWindow is the starting lookup-handle admission window.
	InitialLookupWindow uint64
	// MaxLookupWindow is the upper bound of the lookup-handle admission window.
	MaxLookupWindow uint64
	// InitialLookupBatchSize is the starting number of handles assigned to one
	// lookup task.
	InitialLookupBatchSize uint64
	// MaxLookupBatchSize is the upper bound on handles assigned to one lookup task.
	MaxLookupBatchSize uint64
}

AdaptiveLimitConfig defines the immutable bounds of one statement-local adaptive LIMIT controller. Outer windows are measured in outer rows; lookup windows and batches are measured in index handles.

type AdaptiveLimitController 

type AdaptiveLimitController struct {
	// contains filtered or unexported fields
}

AdaptiveLimitController bounds speculative work for an early-stop LIMIT. It is owned by one executor tree and learns only from the current execution.

The controller maintains two independent admission lifecycles: 

outer rows:     ReserveOuter -> CommitOuter -> ObserveJoinProgress
lookup handles: ReserveLookup -> lookup task -> CompleteLookup / AbortLookup

Both lifecycles are active for an IndexJoin: lookup admission bounds the index handles used to produce outer rows, while outer admission bounds the rows consumed by the join to produce final rows. A direct IndexLookUp uses only the lookup lifecycle.

Outer admission is measured in IndexJoin outer rows. Lookup admission is measured in index handles. The accounting must remain separate because a table-side filter can make many handles produce few or no rows. Every reservation must eventually be committed, completed, aborted, or cleared by Stop. All mutable state below is protected by mu.

func NewAdaptiveLimitController 

func NewAdaptiveLimitController(config AdaptiveLimitConfig) *AdaptiveLimitController

NewAdaptiveLimitController creates a statement-local admission controller.

func NewAdaptiveLimitLookupController 

func NewAdaptiveLimitLookupController(config AdaptiveLimitConfig) *AdaptiveLimitController

NewAdaptiveLimitLookupController creates a controller for a direct ordered IndexLookUp under LIMIT. Direct lookup has no Join outer stage, so only its lookup handle budget is initialized.

func (*AdaptiveLimitController) AbortLookup 

func (c *AdaptiveLimitController) AbortLookup(handles int)

AbortLookup releases lookup admission without using the task as a yield sample. It is used for extraction errors, cancellation, and undispatched work.

func (*AdaptiveLimitController) CommitOuter 

func (c *AdaptiveLimitController) CommitOuter(reserved, fetched int)

CommitOuter converts an outer reservation into rows actually fetched.

func (*AdaptiveLimitController) CompleteLookup 

func (c *AdaptiveLimitController) CompleteLookup(reserved, handles, rows int)

CompleteLookup releases a fully consumed lookup task and learns its handle-to-row yield. reserved is admission accounting; handles is the task's actual input. Callers must report tasks in result-consumption order.

func (*AdaptiveLimitController) ObserveJoinProgress 

func (c *AdaptiveLimitController) ObserveJoinProgress(consumedRows, outputRows int)

ObserveJoinProgress updates input consumption and output production under one lock. Output produced before an outer row is fully consumed is retained until it can be paired with completed outer-row input.

func (*AdaptiveLimitController) ReserveLookup 

func (c *AdaptiveLimitController) ReserveLookup(ctx context.Context, maxHandles int) (int, bool, error)

ReserveLookup bounds handles admitted to the double-read table lookup stage. The bool is false after LIMIT completion; context cancellation is returned as an error.

func (*AdaptiveLimitController) ReserveOuter 

func (c *AdaptiveLimitController) ReserveOuter(ctx context.Context, maxRows int) (int, bool, error)

ReserveOuter waits until up to maxRows can be admitted to the join outer pipeline. The bool is false after LIMIT completion; context cancellation is returned as an error.

func (*AdaptiveLimitController) Reset 

func (c *AdaptiveLimitController) Reset()

Reset prepares the controller for another Open/Next/Close lifecycle. Callers must ensure all producers from the previous lifecycle have exited.

func (*AdaptiveLimitController) Snapshot 

func (c *AdaptiveLimitController) Snapshot() AdaptiveLimitSnapshot

Snapshot returns the current controller counters for tests and diagnostics.

func (*AdaptiveLimitController) Stop 

func (c *AdaptiveLimitController) Stop()

Stop prevents future admission and wakes all blocked producers.

func (*AdaptiveLimitController) SuggestedBatchSize 

func (c *AdaptiveLimitController) SuggestedBatchSize(ceiling int) int

SuggestedBatchSize returns the current execution batch bounded by the caller's configured batch ceiling.

type AdaptiveLimitSnapshot 

type AdaptiveLimitSnapshot struct {
	// DemandRows is the final-row demand derived from LIMIT offset plus count.
	DemandRows uint64
	// OutputRows is the cumulative number of final rows observed by the controller.
	OutputRows uint64
	// OuterFetched is the cumulative number of committed IndexJoin outer rows.
	OuterFetched uint64
	// OuterConsumed is the cumulative number of outer rows fully consumed by the join.
	OuterConsumed uint64
	// OuterReserved is the number of admitted outer rows not yet committed or aborted.
	OuterReserved uint64
	// OuterWindow is the current logical outer-row admission limit.
	OuterWindow uint64
	// OuterOutstandingAtStop captures fetched-but-unconsumed and reserved outer
	// rows when the controller stops.
	OuterOutstandingAtStop uint64
	// LookupReserved is the number of admitted handles not yet completed or aborted.
	LookupReserved uint64
	// LookupHandles is the cumulative number of handles in fully consumed lookup tasks.
	LookupHandles uint64
	// LookupRows is the cumulative number of table rows returned by those lookup tasks.
	LookupRows uint64
	// LookupWindow is the current logical lookup-handle admission limit.
	LookupWindow uint64
	// LookupBatchSize is the current execution-task size in index handles.
	LookupBatchSize uint64
	// LookupPhysicalWindow is LookupWindow rounded to executable whole batches
	// without exceeding the configured maximum window.
	LookupPhysicalWindow uint64
	// LookupOutstandingAtStop captures reserved lookup handles when the controller stops.
	LookupOutstandingAtStop uint64
	// OuterAdmissionBlocked is the wall-clock time with at least one blocked
	// outer reserver.
	OuterAdmissionBlocked time.Duration
	// LookupAdmissionBlocked is the wall-clock time with at least one blocked
	// lookup reserver.
	LookupAdmissionBlocked time.Duration
	// Stopped reports whether the controller rejects future admission.
	Stopped bool
}

AdaptiveLimitSnapshot is a point-in-time view of an AdaptiveLimitController. Outer fields are measured in outer rows. Lookup reservation, handle, window, and batch fields are measured in handles; LookupRows is measured in table rows. DemandRows and OutputRows are measured in final output rows. For an IndexJoin, output rows are observed by the join; for a direct IndexLookUp, they are observed when a lookup task is fully consumed. AdmissionBlocked fields are wall-clock durations with overlapping waits in the same stage counted once.

type BaseExecutor 

type BaseExecutor struct {
	BaseExecutorV2
	// contains filtered or unexported fields
}

BaseExecutor holds common information for executors.

func NewBaseExecutor 

func NewBaseExecutor(ctx sessionctx.Context, schema *expression.Schema, id int, children ...Executor) BaseExecutor

NewBaseExecutor creates a new BaseExecutor instance.

func (*BaseExecutor) AllChildren 

func (e *BaseExecutor) AllChildren() []Executor

AllChildren returns all children.

func (*BaseExecutor) Children 

func (e *BaseExecutor) Children(idx int) Executor

Children returns the children for an executor.

func (*BaseExecutor) ChildrenLen 

func (e *BaseExecutor) ChildrenLen() int

ChildrenLen returns the length of children.

func (*BaseExecutor) Ctx 

func (e *BaseExecutor) Ctx() sessionctx.Context

Ctx return ```sessionctx.Context``` of Executor

func (*BaseExecutor) EmptyChildren 

func (e *BaseExecutor) EmptyChildren() bool

EmptyChildren judges whether the children is empty.

func (*BaseExecutor) GetSchema 

func (e *BaseExecutor) GetSchema() *expression.Schema

GetSchema gets the schema.

func (*BaseExecutor) GetSysSession 

func (e *BaseExecutor) GetSysSession() (sessionctx.Context, error)

GetSysSession gets a system session context from executor.

func (*BaseExecutor) HandleSQLKillerSignal 

func (e *BaseExecutor) HandleSQLKillerSignal() error

func (*BaseExecutor) ID 

func (e *BaseExecutor) ID() int

ID returns the id of an executor.

func (*BaseExecutor) InitCap 

func (e *BaseExecutor) InitCap() int

InitCap returns the initial capacity for chunk

func (*BaseExecutor) MaxChunkSize 

func (e *BaseExecutor) MaxChunkSize() int

MaxChunkSize returns the max chunk size.

func (*BaseExecutor) NewChunk 

func (e *BaseExecutor) NewChunk() *chunk.Chunk

NewChunk creates a new chunk according to the executor configuration

func (*BaseExecutor) NewChunkWithCapacity 

func (e *BaseExecutor) NewChunkWithCapacity(fields []*types.FieldType, capacity int, maxCachesize int) *chunk.Chunk

NewChunkWithCapacity allows the caller to allocate the chunk with any types, capacity and max size in the pool

func (*BaseExecutor) RegisterSQLAndPlanInExecForTopProfiling 

func (e *BaseExecutor) RegisterSQLAndPlanInExecForTopProfiling()

RegisterSQLAndPlanInExecForTopProfiling registers the current SQL and Plan on top profiling.

func (*BaseExecutor) ReleaseSysSession 

func (e *BaseExecutor) ReleaseSysSession(ctx context.Context, sctx sessionctx.Context)

ReleaseSysSession releases a system session context to executor.

func (*BaseExecutor) RetFieldTypes 

func (e *BaseExecutor) RetFieldTypes() []*types.FieldType

NewChunkWithCapacity allows the caller to allocate the chunk with any types, capacity and max size in the pool

func (*BaseExecutor) RuntimeStats 

func (e *BaseExecutor) RuntimeStats() *execdetails.BasicRuntimeStats

RuntimeStats returns the runtime stats of an executor.

func (*BaseExecutor) Schema 

func (e *BaseExecutor) Schema() *expression.Schema

Schema returns the current BaseExecutor's schema. If it is nil, then create and return a new one.

func (*BaseExecutor) SetAllChildren 

func (e *BaseExecutor) SetAllChildren(children []Executor)

SetAllChildren sets the children for an executor.

func (*BaseExecutor) SetChildren 

func (e *BaseExecutor) SetChildren(idx int, ex Executor)

SetChildren sets a child for an executor.

func (*BaseExecutor) SetInitCap 

func (e *BaseExecutor) SetInitCap(c int)

SetInitCap sets the initial capacity for chunk

func (*BaseExecutor) SetMaxChunkSize 

func (e *BaseExecutor) SetMaxChunkSize(size int)

SetMaxChunkSize sets the max chunk size.

func (*BaseExecutor) UpdateDeltaForTableID 

func (e *BaseExecutor) UpdateDeltaForTableID(id int64)

UpdateDeltaForTableID updates the delta info for the table with tableID.

type BaseExecutorV2 

type BaseExecutorV2 struct {
	// contains filtered or unexported fields
}

BaseExecutorV2 is a simplified version of `BaseExecutor`, which doesn't contain a full session context

func NewBaseExecutorV2 

func NewBaseExecutorV2(vars *variable.SessionVars, schema *expression.Schema, id int, children ...Executor) BaseExecutorV2

NewBaseExecutorV2 creates a new BaseExecutorV2 instance.

func (*BaseExecutorV2) AllChildren 

func (e *BaseExecutorV2) AllChildren() []Executor

AllChildren returns all children.

func (*BaseExecutorV2) BuildNewBaseExecutorV2 

func (e *BaseExecutorV2) BuildNewBaseExecutorV2(stmtRuntimeStatsColl *execdetails.RuntimeStatsColl, schema *expression.Schema, id int, children ...Executor) BaseExecutorV2

BuildNewBaseExecutorV2 builds a new `BaseExecutorV2` based on the configuration of the current base executor. It's used to build a new sub-executor from an existing executor. For example, the `IndexLookUpExecutor` will use this function to build `TableReaderExecutor`

func (*BaseExecutorV2) Children 

func (e *BaseExecutorV2) Children(idx int) Executor

Children returns the children for an executor.

func (*BaseExecutorV2) ChildrenLen 

func (e *BaseExecutorV2) ChildrenLen() int

ChildrenLen returns the length of children.

func (*BaseExecutorV2) Close 

func (e *BaseExecutorV2) Close() error

Close closes all executors and release all resources.

func (*BaseExecutorV2) Detach 

func (*BaseExecutorV2) Detach() (Executor, bool)

Detach detaches the current executor from the session context.

func (*BaseExecutorV2) EmptyChildren 

func (e *BaseExecutorV2) EmptyChildren() bool

EmptyChildren judges whether the children is empty.

func (*BaseExecutorV2) GetSchema 

func (e *BaseExecutorV2) GetSchema() *expression.Schema

GetSchema gets the schema.

func (*BaseExecutorV2) HandleSQLKillerSignal 

func (e *BaseExecutorV2) HandleSQLKillerSignal() error

func (*BaseExecutorV2) ID 

func (e *BaseExecutorV2) ID() int

ID returns the id of an executor.

func (*BaseExecutorV2) InitCap 

func (e *BaseExecutorV2) InitCap() int

InitCap returns the initial capacity for chunk

func (*BaseExecutorV2) MaxChunkSize 

func (e *BaseExecutorV2) MaxChunkSize() int

MaxChunkSize returns the max chunk size.

func (*BaseExecutorV2) NewChunk 

func (e *BaseExecutorV2) NewChunk() *chunk.Chunk

NewChunk creates a new chunk according to the executor configuration

func (*BaseExecutorV2) NewChunkWithCapacity 

func (e *BaseExecutorV2) NewChunkWithCapacity(fields []*types.FieldType, capacity int, maxCachesize int) *chunk.Chunk

NewChunkWithCapacity allows the caller to allocate the chunk with any types, capacity and max size in the pool

func (*BaseExecutorV2) Next 

func (*BaseExecutorV2) Next(_ context.Context, _ *chunk.Chunk) error

Next fills multiple rows into a chunk.

func (*BaseExecutorV2) Open 

func (e *BaseExecutorV2) Open(ctx context.Context) error

Open initializes children recursively and "childrenResults" according to children's schemas.

func (*BaseExecutorV2) RegisterSQLAndPlanInExecForTopProfiling 

func (e *BaseExecutorV2) RegisterSQLAndPlanInExecForTopProfiling()

RegisterSQLAndPlanInExecForTopProfiling registers the current SQL and Plan on top profiling.

func (*BaseExecutorV2) RetFieldTypes 

func (e *BaseExecutorV2) RetFieldTypes() []*types.FieldType

NewChunkWithCapacity allows the caller to allocate the chunk with any types, capacity and max size in the pool

func (*BaseExecutorV2) RuntimeStats 

func (e *BaseExecutorV2) RuntimeStats() *execdetails.BasicRuntimeStats

RuntimeStats returns the runtime stats of an executor.

func (*BaseExecutorV2) Schema 

func (e *BaseExecutorV2) Schema() *expression.Schema

Schema returns the current BaseExecutor's schema. If it is nil, then create and return a new one.

func (*BaseExecutorV2) SetAllChildren 

func (e *BaseExecutorV2) SetAllChildren(children []Executor)

SetAllChildren sets the children for an executor.

func (*BaseExecutorV2) SetChildren 

func (e *BaseExecutorV2) SetChildren(idx int, ex Executor)

SetChildren sets a child for an executor.

func (*BaseExecutorV2) SetInitCap 

func (e *BaseExecutorV2) SetInitCap(c int)

SetInitCap sets the initial capacity for chunk

func (*BaseExecutorV2) SetMaxChunkSize 

func (e *BaseExecutorV2) SetMaxChunkSize(size int)

SetMaxChunkSize sets the max chunk size.

type Executor 

type Executor interface {
	NewChunk() *chunk.Chunk
	NewChunkWithCapacity(fields []*types.FieldType, capacity int, maxCachesize int) *chunk.Chunk

	RuntimeStats() *execdetails.BasicRuntimeStats

	HandleSQLKillerSignal() error
	RegisterSQLAndPlanInExecForTopProfiling()

	AllChildren() []Executor
	SetAllChildren([]Executor)
	Open(context.Context) error
	Next(ctx context.Context, req *chunk.Chunk) error

	// `Close()` may be called at any time after `Open()` and it may be called with `Next()` at the same time
	Close() error
	Schema() *expression.Schema
	RetFieldTypes() []*types.FieldType
	InitCap() int
	MaxChunkSize() int

	// Detach detaches the current executor from the session context without considering its children.
	//
	// It has to make sure, no matter whether it returns true or false, both the original executor and the returning executor
	// should be able to be used correctly.
	Detach() (Executor, bool)
}

Executor is the physical implementation of an algebra operator.

In TiDB, all algebra operators are implemented as iterators, i.e., they support a simple Open-Next-Close protocol. See this paper for more details:

"Volcano-An Extensible and Parallel Query Evaluation System"

Different from Volcano's execution model, a "Next" function call in TiDB will return a batch of rows, other than a single row in Volcano. NOTE: Executors must call "chk.Reset()" before appending their results to it.

type IndexUsageReporter 

type IndexUsageReporter struct {
	// contains filtered or unexported fields
}

IndexUsageReporter is a toolkit to report index usage

func NewIndexUsageReporter 

func NewIndexUsageReporter(reporter *indexusage.StmtIndexUsageCollector,
	runtimeStatsColl *execdetails.RuntimeStatsColl,
	statsMap *stmtctx.UsedStatsInfo) *IndexUsageReporter

NewIndexUsageReporter creates an index usage reporter util

func (*IndexUsageReporter) ReportCopIndexUsage 

func (e *IndexUsageReporter) ReportCopIndexUsage(tableID int64, physicalTableID int64, indexID int64, planID int)

ReportCopIndexUsage reports the index usage to the inside collector. The index usage will be recorded in the `tableID+indexID`, but the percentage is calculated using the size of the table specified by `physicalTableID`.

func (*IndexUsageReporter) ReportCopIndexUsageForHandle 

func (e *IndexUsageReporter) ReportCopIndexUsageForHandle(tbl table.Table, planID int)

ReportCopIndexUsageForHandle wraps around `ReportCopIndexUsageForTable` to get the `indexID` automatically from the `table.Table` if the table has a clustered index or clustered integer primary key. If the table has a non-clustered PK or no PK, this function simply returns without reporting index usage.

func (*IndexUsageReporter) ReportCopIndexUsageForTable 

func (e *IndexUsageReporter) ReportCopIndexUsageForTable(tbl table.Table, indexID int64, planID int)

ReportCopIndexUsageForTable wraps around `ReportCopIndexUsage` to get `tableID` and `physicalTableID` from the `table.Table`. If it's expected to calculate the percentage according to the size of partition, the `tbl` argument should be a `table.PhysicalTable`, or the percentage will be calculated using the size of whole table.

func (*IndexUsageReporter) ReportPointGetIndexUsage 

func (e *IndexUsageReporter) ReportPointGetIndexUsage(tableID int64, physicalTableID int64, indexID int64, kvRequestTotal, rows int64)

ReportPointGetIndexUsage reports the index usage of a point get or batch point get

func (*IndexUsageReporter) ReportPointGetIndexUsageForHandle 

func (e *IndexUsageReporter) ReportPointGetIndexUsageForHandle(tblInfo *model.TableInfo, physicalTableID int64, kvRequestTotal, rows int64)

ReportPointGetIndexUsageForHandle wraps around `ReportPointGetIndexUsage` to get the `indexID` automatically from the `table.Table` if the table has a clustered index or integer primary key.

Source Files

View all Source files

Jump to

Keyboard shortcuts

? : This menu
/ : Search site
f or F : Jump to
y or Y : Canonical URL
go.dev uses cookies from Google to deliver and enhance the quality of its services and to analyze traffic. Learn more.