causality

Documentation

Index

• type BlockStrategy
• type ConflictDetector
  • func NewConflictDetector[Txn txnEvent](numSlots uint64, opt TxnCacheOption) *ConflictDetector[Txn]
  • func (d *ConflictDetector[Txn]) Add(txn Txn)
  • func (d *ConflictDetector[Txn]) Close()
  • func (d *ConflictDetector[Txn]) GetOutChByCacheID(id int64) <-chan TxnWithNotifier[Txn]
• type TxnCacheOption
• type TxnWithNotifier

Types

type BlockStrategy

type BlockStrategy string

BlockStrategy is the strategy to handle the situation when the cache is full.

const (
	// BlockStrategyWaitAvailable means the cache will block until there is an available slot.
	BlockStrategyWaitAvailable BlockStrategy = "waitAvailable"
	// BlockStrategyWaitEmpty means the cache will block until all cached txns are consumed.
	BlockStrategyWaitEmpty = "waitEmpty"
)

type ConflictDetector

type ConflictDetector[Txn txnEvent] struct {
	// contains filtered or unexported fields
}

ConflictDetector implements a logic that dispatches transaction to different worker cache channels in a way that transactions modifying the same keys are never executed concurrently and have their original orders preserved. Transactions in different channels can be executed concurrently.

func NewConflictDetector

func NewConflictDetector[Txn txnEvent](
	numSlots uint64, opt TxnCacheOption,
) *ConflictDetector[Txn]

NewConflictDetector creates a new ConflictDetector.

func (*ConflictDetector[Txn]) Add

func (d *ConflictDetector[Txn]) Add(txn Txn)

Add pushes a transaction to the ConflictDetector.

NOTE: if multiple threads access this concurrently, Txn.ConflictKeys must be sorted by the slot index.

func (*ConflictDetector[Txn]) Close

func (d *ConflictDetector[Txn]) Close()

Close closes the ConflictDetector.

func (*ConflictDetector[Txn]) GetOutChByCacheID

func (d *ConflictDetector[Txn]) GetOutChByCacheID(id int64) <-chan TxnWithNotifier[Txn]

GetOutChByCacheID returns the output channel by cacheID. Note txns in single cache should be executed sequentially.

type TxnCacheOption

type TxnCacheOption struct {
	// Count controls the number of caches, txns in different caches could be executed concurrently.
	Count int
	// Size controls the max number of txns a cache can hold.
	Size int
	// BlockStrategy controls the strategy when the cache is full.
	BlockStrategy BlockStrategy
}

TxnCacheOption is the option for creating a cache for resolved txns.

type TxnWithNotifier

type TxnWithNotifier[Txn txnEvent] struct {
	TxnEvent Txn
	// The PostTxnExecuted will remove the txn related Node in the conflict detector's
	// dependency graph and resolve related dependencies for these transacitons
	// which depend on this executed txn.
	//
	// NOTE: the PostTxnExecuted() must be called after the txn executed.
	PostTxnExecuted func()
}

TxnWithNotifier is a wrapper of txnEvent with a PostTxnExecuted.