Struct tikv::storage::txn::latch::Latch

[+] Show declaration

struct Latch {
    pub waiting: VecDeque<Option<(u64, u64)>>,
}

[−] Expand description

Latch which is used to serialize accesses to resources hashed to the same slot.

Latches are indexed by slot IDs. The keys of a command are hashed into unsigned numbers, then the command is added to the waiting queues of the latches.

If command A is ahead of command B in one latch, it must be ahead of command B in all the overlapping latches. This is an invariant ensured by the gen_lock, acquire and release.

Fields

waiting: VecDeque<Option<(u64, u64)>>

Implementations

impl Latch[−]

pub fn new() -> Latch[−]

Creates a latch with an empty waiting queue.

pub fn get_first_req_by_hash(&self, hash: u64) -> Option<u64>[−]

Find the first command ID in the queue whose hash value is equal to hash.

pub fn pop_front(&mut self, key_hash: u64) -> Option<(u64, u64)>[−]

Remove the first command ID in the queue whose hash value is equal to hash_key. If the element which would be removed does not appear at the front of the queue, it will leave a hole in the queue. So we must remove consecutive hole when remove the head of the queue to make the queue not too long.

pub fn wait_for_wake(&mut self, key_hash: u64, cid: u64)

fn maybe_shrink(&mut self)[−]

For some hot keys, the waiting list maybe very long, so we should shrink the waiting VecDeque after pop.

Trait Implementations

impl Clone for Latch[+]

[+] Show hidden undocumented items

fn clone(&self) -> Latch[−]

Returns a copy of the value.

pub fn clone_from(&mut self, source: &Self)[−]

Performs copy-assignment from source.