Datafetchers as completable futures don't run in parallel

[prograhammer]

I created two simple datafetchers as completable futures but they don't run in parallel. What am I missing here?

Simple schema:

Users DataFetcher:

Products DataFetcher:

They don't run in parallel:

I could create the futures and join them within a single datafetcher but that defeats the nice annotation ability of DGS right? It is because one is Query and the other is an Object Type (on different levels, so-to-speak)?

[paulbakker]

You’ll need to provide an Executor as the second argument when creating the CompletableFuture

…

On Mon, Aug 2, 2021 at 5:36 PM David Graham ***@***.***> wrote: I created two simple datafetchers as completable futures but they don't run in parallel. What am I missing here? Simple schema: [image: image] <https://user-images.githubusercontent.com/4474660/127939638-1896f998-eaa0-4cc6-a4de-51e583594931.png> Users DataFetcher: [image: image] <https://user-images.githubusercontent.com/4474660/127939698-3c53cdf8-ead1-41a9-ba99-308c9887f93e.png> Products DataFetcher: [image: image] <https://user-images.githubusercontent.com/4474660/127939731-8c509bbc-86f0-42ef-a214-1c1862bc4fa6.png> They don't run in parallel: [image: image] <https://user-images.githubusercontent.com/4474660/127939755-dcd7e14f-3d76-486e-ad54-f77eaf4c51f9.png> I could create the futures and join them within a single datafetcher but that defeats the nice annotation ability of DGS right? It is because one is Query and the other is an Object Type (on different levels, so-to-speak)? — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#509>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAA2XLAHI57VFYMXDTGT3W3T242XVANCNFSM5BNYDSSA> .

[prograhammer]

Hi @paulbakker

I updated to explicitly pass in an executor (to both datafetchers) but still they don't run in parallel:

    Executor executor = Executors.newFixedThreadPool(10);

    return CompletableFuture.supplyAsync(() -> {
      try {
          System.out.println("User task execution started.");
          Thread.sleep(5000);
          System.out.println("User task execution stopped.");
      } catch (Exception e) {
          e.printStackTrace();
      }
      return users;
    }, executor);

Am I doing this wrong?

[prograhammer]

Oh I probably need to do this at startup and then use the same executor across datafetchers.....let me try that...

[prograhammer]

No this didn't help:

Executor bean:

@Service
public class ExecutorService {

    private final Executor executor;

    @Autowired
    public ExecutorService() {
        this.executor = Executors.newFixedThreadPool(10);
    }

    public Executor getExecutor() {
        return this.executor;
    }
    
}

Updated to User (and I also updated Product) datafetcher:

  @Autowired
  private ExecutorService executorService;

  @DgsData(parentType = "Query", field = "users")
  public CompletableFuture<List<User>> getUsers(DataFetchingEnvironment dfe) {

    String userId = dfe.getArgument("userId");

    return CompletableFuture.supplyAsync(() -> {
      try {
          System.out.println("User task execution started.");
          Thread.sleep(5000);
          System.out.println("User task execution stopped.");
      } catch (Exception e) {
          e.printStackTrace();
      }
      return users;
    }, executorService.getExecutor());

Still does not run in parallel.

[prograhammer]

Any example of how to use an executor @paulbakker to get datafetchers to actually work in parallel?

[paulbakker]

This was answered in #512:

It took me a minute to figure out what's going on. The problem comes from the relation between User and Product. Because products is a field of User, it first needs to resolve the parent (User in this case) before it can continue to resolve subfields.
Imagine for example that your products datafetcher would use the getSource method to get its parent (the User). That wouldn't work if User wasn't resolved yet.

If products and user were two unrelated fields on Query, your code would work. I also found that my suggestion about using an Executor isn't actually required.

We have a simple test in the example embedded in the framework's repository that demonstrates this: https://github.com/Netflix/dgs-framework/blob/master/graphql-dgs-example-shared/src/test/java/com/netflix/graphql/dgs/example/shared/ConcurrentDataFetcherTest.java

Of course, that doesn't answer the question how you could make such a scenario parallel. I think the best option is a "look ahead" data fetcher as described here: https://netflix.github.io/dgs/advanced/context-passing/#pre-loading.

I'm closing this issue because I don't think it's further actionable, but feel free to ask more questions.