sszuev / concurrent-rdf-graph Goto Github PK

• java.util.concurrent.locks.Lock.tryLock with timeout
• to reduce memory consumption remove processed items from snapshot iterator immediately after Iterator.next
• memory limit (capacity for snapshots)
• optional callback to control cache overflow

As a library, it would be better it does not pull extra dependencies. + Right now there are no kotlin-specific features in use (coroutines are only in tests)

see #3

drawt: https://github.com/sszuev/concurrent-rdf-graph/blob/experimental/src/main/kotlin/experimental/KGraph.kt

• 1. change modify and releaseToSnapshot, snapshot collect operating could be divided on steps interspersed with each other, e.g. instead openIterators.keys.forEach { releaseToSnapshot(it) } something like while (openIterators.isNotEmpty) { if (toSnapshot(chunkSize, it)) openIterators.remove(...) }
• 2. intellectual processing, if there is open iterator for ANY X ANY and modify operation for Q W E, than don't need releaseToSnapshot for that iterator
• 3. check if iterator is started, if not the internal lock can be released after white operation
• 4. NoOp Lock while Iterator's processing. Modify operations should turn on ReentrantLock before modification and turn off after modification

Related to apache/jena#1961

GraphMem (single-thread), GraphTxn, SynchronizedGraph, ReadWriteLockingGraph