Struct PrefixProber 

pub struct PrefixProber<'a, D: DBAccess> { /* private fields */ }

A reusable prefix probe that avoids per-call iterator creation/destruction.

Use this when performing many prefix existence checks in a tight loop.

A prober reads the database as of the sequence number that was current when it was created, and it pins the memtables and SST files that were current then. Both are what refresh exists to move forward. A prober that is only used for a single burst of probes and then dropped needs neither.

Implementations

impl<D: DBAccess> PrefixProber<'_, D>

pub fn exists(&mut self, prefix: &[u8]) -> Result<bool, Error>

Returns true if any key exists with the given prefix. This performs a seek to the prefix and checks the current key.

Errors

Returns the RocksDB error if the seek failed. A seek that hit an error leaves the iterator invalid, so the key check below cannot observe one.

pub fn refresh(&mut self) -> Result<(), Error>

Moves the probe to the latest committed state of the database.

Call this before reusing a prober that has been sitting idle. Writes that land after the prober was created or last refreshed are invisible to it until then.

This is cheap when RocksDB’s superversion has not changed, because the existing merge tree over the memtables and SST files is kept and only the read sequence moves. A flush or a compaction bumps the superversion and forces that tree to be rebuilt, which costs roughly what building a new prober costs. Refreshing a prober that has not probed yet is a pessimisation, because it builds the tree that the first exists call would otherwise build lazily.

Refreshing also releases the memtables and SST files the prober was pinning, which is what lets a flush or a compaction reclaim them. A cached prober that is never refreshed holds them for as long as it lives, so pool them on a timer rather than on request arrival.

Errors

Returns the RocksDB error if the refresh failed.