Struct ShardedLockTable 

pub struct ShardedLockTable { /* private fields */ }

Sharded lock table for row-level locks 256 shards reduce contention by ~256x for uniform key distribution

Implementations

impl ShardedLockTable

pub fn new() -> Self

Create a new sharded lock table

pub fn try_lock( &self, row_id: RowId, mode: LockMode, txn_id: TxnId, ) -> LockResult

Try to acquire a lock on a row

pub fn unlock(&self, row_id: RowId, txn_id: TxnId) -> bool

Release a lock on a row

pub fn unlock_all(&self, txn_id: TxnId) -> usize

Release all locks held by a transaction

WARNING: This method scans all 256 shards, acquiring each mutex. For a transaction holding k locks across s shards, it still visits all 256 shards at ~15ns per uncontested mutex = ~7.7µs overhead.

For better performance, use try_lock_tracked + unlock_all_tracked which only visits the shards that actually hold locks for this txn.

pub fn try_lock_tracked( &self, row_id: RowId, mode: LockMode, txn_id: TxnId, lock_set: &mut TransactionLockSet, ) -> LockResult

Acquire a lock and record it in the transaction’s lock set

This is the preferred API over try_lock because it enables O(k) unlock via unlock_all_tracked instead of O(256) unlock via unlock_all.

pub fn unlock_all_tracked( &self, txn_id: TxnId, lock_set: &TransactionLockSet, ) -> usize

Release all locks tracked in a TransactionLockSet

O(k) where k is the number of locks held, instead of O(256) for unlock_all. For a typical 3-lock transaction: ~90ns instead of ~7.7µs (85× improvement).

pub fn stats(&self) -> &LockTableStats

Get statistics

Trait Implementations

impl Default for ShardedLockTable

fn default() -> Self

Returns the “default value” for a type.