This guide walks you through the steps to create asynchronous queries to GitHub. The focus is on the asynchronous part, a feature often used when scaling services.
You’ll build a lookup service that queries GitHub user information and retrieves data through GitHub’s API. One approach to scaling services is to run expensive jobs in the background and wait for the results using Java’s CompletableFuture interface. Java’s CompletableFuture is an evolution from the regular Future. It makes it easy to pipeline multiple asynchronous operations merging them into a single asynchronous computation.
Before you can create a GitHub lookup service, you need to define a representation for the data you’ll retrieve through GitHub’s API.
To model the user representation, you create a resource representation class. Provide a plain old Java object with fields, constructors, and accessors:
src/main/java/hello/User.java
link:complete/src/main/java/hello/User.java[role=include]Spring uses the Jackson JSON library to convert GitHub’s JSON response into a User object. The @JsonIgnoreProperties annotation signals Spring to ignore any attributes not listed in the class. This makes it easy to make REST calls and produce domain objects.
In this guide, we are only grabbing the name and the blog URL for demonstration purposes.
Next you need to create a service that queries GitHub to find user information.
src/main/java/hello/GitHubLookupService.java
link:complete/src/main/java/hello/GitHubLookupService.java[role=include]The GitHubLookupService class uses Spring’s RestTemplate to invoke a remote REST point
(api.github.com/users/), and then convert the answer into a User object. Spring Boot automatically provides a RestTemplateBuilder that customizes the defaults with any auto-configuration bits (i.e. MessageConverter).
The class is marked with the @Service annotation, making it a candidate for Spring’s component scanning to detect it and add it to the application context.
The findUser method is flagged with Spring’s @Async annotation, indicating it will run on a separate thread. The method’s return type is CompletableFuture<User> instead of User, a requirement for any asynchronous service. This code uses the completedFuture method to return a CompletableFuture instance which is already completed with result of the GitHub query.
|
Note
|
Creating a local instance of the GitHubLookupService class does NOT allow the findUser method to run asynchronously. It must be created inside a @Configuration class or picked up by @ComponentScan.
|
The timing for GitHub’s API can vary. To demonstrate the benefits later in this guide, an extra delay of one second has been added to this service.
To run a sample, you can create an executable jar. Spring’s @Async annotation works with web apps, but you don’t need all the extra steps of setting up a web container to see its benefits.
src/main/java/hello/Application.java
link:complete/src/main/java/hello/Application.java[role=include]The @EnableAsync annotation switches on Spring’s ability to run @Async methods in a background thread pool. This class also customizes the used Executor by defining a new bean. Here the method is named taskExecutor since this is the specific method name Spring will search for. In our case, we want to limit the number of concurrent threads to 2 and limit the size of the queue to 500. There are many more things you can tune. If the user does not define an Executor bean Spring will create a SimpleAsyncTaskExecutor and use that.
There is also a CommandLineRunner that injects the GitHubLookupService and calls that service 3 times to demonstrate the method is executed asynchronously.
src/main/java/hello/AppRunner.java
link:complete/src/main/java/hello/AppRunner.java[role=include]https://raw.githubusercontent.com/spring-guides/getting-started-macros/master/build_an_executable_jar_subhead.adoc https://raw.githubusercontent.com/spring-guides/getting-started-macros/master/build_an_executable_jar_with_both.adoc
Logging output is displayed, showing each query to GitHub. With the help of allOf factory method we create an array of CompletableFutures on which by calling the join
method makes it possible to wait for the completion of all of the CompletableFutures.
2016-09-01 10:25:21.295 INFO 17893 --- [ GithubLookup-2] hello.GitHubLookupService : Looking up CloudFoundry 2016-09-01 10:25:21.295 INFO 17893 --- [ GithubLookup-1] hello.GitHubLookupService : Looking up PivotalSoftware 2016-09-01 10:25:23.142 INFO 17893 --- [ GithubLookup-1] hello.GitHubLookupService : Looking up Spring-Projects 2016-09-01 10:25:24.281 INFO 17893 --- [ main] hello.AppRunner : Elapsed time: 2994 2016-09-01 10:25:24.282 INFO 17893 --- [ main] hello.AppRunner : --> User [name=Pivotal Software, Inc., blog=http://pivotal.io] 2016-09-01 10:25:24.282 INFO 17893 --- [ main] hello.AppRunner : --> User [name=Cloud Foundry, blog=https://www.cloudfoundry.org/] 2016-09-01 10:25:24.282 INFO 17893 --- [ main] hello.AppRunner : --> User [name=Spring, blog=http://spring.io/projects]
Note that the first two calls happen in separate threads (GithubLookup-2, GithubLookup-1) and the third one is parked until one of the two threads became available. To compare how long this takes without the asynchronous feature, try commenting out the @Async annotation and run the service again. The total elapsed time should increase noticeably because each query takes at least a second. You can also tune the Executor to increase the corePoolSize attribute for instance.
Essentially, the longer the task takes and the more tasks are invoked simultaneously, the more benefit you will see with making things asynchronous. The trade off is handling the CompletableFuture interface. It adds a layer of indirection because you are no longer dealing directly with the results.
Congratulations! You’ve just developed an asynchronous service that lets you scale multiple calls at once.
The following guides may also be helpful: