use of com.torodb.concurrent.akka.BatchFlow in project torodb by torodb.
the class DefaultOplogApplier method createBatcherFlow.
/**
* Creates a flow that batches and analyze a input of {@link AnalyzedOplogBatch remote jobs}.
*
* This flow tries to accummulate several remote jobs into a bigger one and does not emit until:
* <ul>
* <li>A maximum number of operations are batched</li>
* <li>Or a maximum time has happen since the last emit</li>
* <li>Or the recived job is not {@link AnalyzedOplogBatch#isReadyForMore()}</li>
* </ul>
*
*/
private Flow<OplogBatch, AnalyzedStreamElement, NotUsed> createBatcherFlow(ApplierContext context) {
Predicate<OplogBatch> finishBatchPredicate = (OplogBatch rawBatch) -> !rawBatch.isReadyForMore();
ToIntFunction<OplogBatch> costFunction = (rawBatch) -> rawBatch.count();
Supplier<RawStreamElement> zeroFun = () -> RawStreamElement.INITIAL_ELEMENT;
BiFunction<RawStreamElement, OplogBatch, RawStreamElement> acumFun = (streamElem, newBatch) -> streamElem.concat(newBatch);
BatchAnalyzer batchAnalyzer = batchAnalyzerFactory.createBatchAnalyzer(context);
return Flow.of(OplogBatch.class).via(new BatchFlow<>(batchLimits.maxSize, batchLimits.maxPeriod, finishBatchPredicate, costFunction, zeroFun, acumFun)).filter(rawElem -> rawElem.rawBatch != null && !rawElem.rawBatch.isEmpty()).map(rawElem -> {
List<OplogOperation> rawOps = rawElem.rawBatch.getOps();
List<AnalyzedOplogBatch> analyzed = batchAnalyzer.apply(rawOps);
return new AnalyzedStreamElement(rawElem, analyzed);
});
}
Aggregations