All Downloads are FREE. Search and download functionalities are using the official Maven repository.

graphql.execution.instrumentation.threadpools.ExecutorInstrumentation Maven / Gradle / Ivy

There is a newer version: 230521-nf-execution
Show newest version
package graphql.execution.instrumentation.threadpools;

import com.google.common.annotations.Beta;
import graphql.Assert;
import graphql.Internal;
import graphql.TrivialDataFetcher;
import graphql.execution.Async;
import graphql.execution.instrumentation.InstrumentationState;
import graphql.execution.instrumentation.SimplePerformantInstrumentation;
import graphql.execution.instrumentation.parameters.InstrumentationFieldFetchParameters;
import graphql.schema.DataFetcher;
import graphql.schema.DataFetchingEnvironment;
import org.jetbrains.annotations.NotNull;

import java.util.concurrent.CompletableFuture;
import java.util.concurrent.CompletionStage;
import java.util.concurrent.Executor;
import java.util.function.Consumer;
import java.util.function.Function;
import java.util.function.Supplier;

import static graphql.execution.instrumentation.threadpools.ExecutorInstrumentation.Action.FETCHING;
import static graphql.execution.instrumentation.threadpools.ExecutorInstrumentation.Action.PROCESSING;

/**
 * This instrumentation can be used to control on what thread calls to {@link DataFetcher}s happen on.
 * 

* If your data fetching is inherently IO bound then you could use a IO oriented thread pool for your fetches and transfer control * back to a CPU oriented thread pool and allow graphql-java code to run the post-processing of results there. *

* An IO oriented thread pool is typically a multiple of {@link Runtime#availableProcessors()} while a CPU oriented thread pool * is typically no more than {@link Runtime#availableProcessors()}. *

* The instrumentation will use the {@link graphql.execution.instrumentation.Instrumentation#instrumentDataFetcher(DataFetcher, InstrumentationFieldFetchParameters, InstrumentationState)} * method to change your data fetchers, so they are executed on a thread pool dedicated to fetching (if you provide one). *

* Once the data fetcher value is returns it will transfer control back to a processing thread pool (if you provide one). *

* This code uses {@link CompletableFuture#supplyAsync(Supplier, Executor)} and {@link CompletableFuture#thenApplyAsync(Function, Executor)} to transfer * control between thread pools. */ @Internal @Beta public class ExecutorInstrumentation extends SimplePerformantInstrumentation { private static final Consumer NOOP = a -> { }; /** * This describes what action is currently being done. This is mostly intended for testing. */ enum Action {FETCHING, PROCESSING} private final Executor fetchExecutor; private final Executor processingExecutor; private final Consumer actionObserver; private ExecutorInstrumentation(Executor fetchExecutor, Executor processingExecutor, Consumer actionObserver) { this.fetchExecutor = fetchExecutor; this.processingExecutor = processingExecutor; this.actionObserver = actionObserver; } public Executor getFetchExecutor() { return fetchExecutor; } public Executor getProcessingExecutor() { return processingExecutor; } public static Builder newThreadPoolExecutionInstrumentation() { return new Builder(); } public static class Builder { Executor fetchExecutor; Executor processingExecutor; private Consumer actionObserver; public Builder fetchExecutor(Executor fetchExecutor) { this.fetchExecutor = fetchExecutor; return this; } public Builder processingExecutor(Executor processingExecutor) { this.processingExecutor = processingExecutor; return this; } /** * This is really intended for testing but this consumer will be called during * stages to indicate what is happening. * * @param actionObserver the observer code * * @return this builder */ public Builder actionObserver(Consumer actionObserver) { this.actionObserver = Assert.assertNotNull(actionObserver); return this; } public ExecutorInstrumentation build() { return new ExecutorInstrumentation(fetchExecutor, processingExecutor, actionObserver != null ? actionObserver : NOOP); } } @Override public @NotNull DataFetcher instrumentDataFetcher(DataFetcher originalDataFetcher, InstrumentationFieldFetchParameters parameters, InstrumentationState state) { if (originalDataFetcher instanceof TrivialDataFetcher) { return originalDataFetcher; } return environment -> { CompletableFuture> invokedCF; if (fetchExecutor != null) { // run the fetch asynchronously via the fetch executor // the CF will be left running on that fetch executors thread invokedCF = CompletableFuture.supplyAsync(invokedAsync(originalDataFetcher, environment), fetchExecutor); } else { invokedCF = invokedSync(originalDataFetcher, environment); } if (processingExecutor != null) { invokedCF = invokedCF.thenApplyAsync(processingControl(), processingExecutor); } else { invokedCF = invokedCF.thenApply(processingControl()); } return invokedCF.thenCompose(cs -> cs); }; } private Supplier> invokedAsync(DataFetcher originalDataFetcher, DataFetchingEnvironment environment) { return () -> { actionObserver.accept(FETCHING); return invokeOriginalDF(originalDataFetcher, environment); }; } private CompletableFuture> invokedSync(DataFetcher originalDataFetcher, DataFetchingEnvironment environment) { actionObserver.accept(FETCHING); return CompletableFuture.completedFuture(invokeOriginalDF(originalDataFetcher, environment)); } private Function, CompletionStage> processingControl() { return completionStage -> { actionObserver.accept(PROCESSING); return completionStage; }; } private CompletionStage invokeOriginalDF(DataFetcher originalDataFetcher, DataFetchingEnvironment environment) { Object value; try { value = originalDataFetcher.get(environment); } catch (Exception e) { return Async.exceptionallyCompletedFuture(e); } if (value instanceof CompletionStage) { return ((CompletionStage) value); } else { return CompletableFuture.completedFuture(value); } } }





© 2015 - 2024 Weber Informatics LLC | Privacy Policy