Dataloader return as async Fetcher about java-dataloader HOT 4 CLOSED

Have a look CompleteableFuture.thenCompose

https://docs.oracle.com/javase/8/docs/api/java/util/concurrent/CompletableFuture.html#thenCompose-java.util.function.Function-

Its effectively a flatmap on two CFs that means you get to chain together to steps that both produce CFs and present one

Also I think you need to change the data fetchers so that instead of wrapping them in AsyncDataFetcher.async() you code them to use the CompleteableFuture.supplyAsync

https://docs.oracle.com/javase/8/docs/api/java/util/concurrent/CompletableFuture.html#supplyAsync-java.util.function.Supplier-

        DataFetcher dataFetcher = new DataFetcher() {
            @Override
            public Object get(DataFetchingEnvironment environment) throws Exception {
                Supplier<String> fetchOfValue = new Supplier<String>() {
                    @Override
                    public String get() {
                        return "This is done async via a supplier and CompletableFuture.supplyAsync";
                    }
                };
                return CompletableFuture.supplyAsync(fetchOfValue);
            }
        };

HOWEVER....

if you are using DataLoader then typically you DONT need async data fetchers because the org.dataloader.BatchLoader is itself ready for an async value

Your BatchLoader of books should use CompletableFuture.supplyAsync and hence there is no need for AsyncDataFetcher.async to be used

from java-dataloader.

Thank you for your answer.
In fact, my project is a common project.
We abstract the MODEL and DATASOURCE .
MODEL like VOs, use by front-end such as web or app; and DATASOURCE like BOs, fetch data from database or other rpc.
Dataloader take charge of batch load and cache a DATASOURCE .

A fetcher will returns a MODEL, but MODEL may from two DATASOURCE or more, even use DAG to process DATASOURCE dependency.
So a fetcher may consume a long time. It's why I use AsyncDataFetcher.async to get a fetcher.

Back to this question, I think your first suggest apply to two CFs serializability , such as flapMap. but this fact is CF1<CF2>, the CF1 produce AsyncDataFetcher.async Fetchers , the CF2 product dataloader.load in the Fetchers.

And the 2nd suggest async in fetcher of Supplier, or use AsyncDataFetcher.async is no difference I think. because I want async fetcher and dataloader.load is async to wait graphql execution to dispatchAll() also.

Of course, my question is not only belong to dataloader. It's a graphql-java Invocation process question. But it's bring by dataloader, so we discuss it in here.

from java-dataloader.

Let me be more explicit.

When you use DataLoader you are trying to "batch" N calls by key to one batch load function.

Therefore your data fetchers should be just like

DataFetcher df  = env -> {
      return dataLoader.load(ev.getArg("myKey)
}

This will return a CF that is not resolved until dispatch is called and the batch load function is invoked

In your BatchLoad function you should do this

BatchLoader<K,V> batchLoader = keys -> {
   return CompleteableFuture.supplyAsync( () -> {
         return thisCallMakeTakeSomeTime(keys);
   });
}

This will be 100% async . There is no need for AsyncDataFectcher etc... at the field data fetcher site.

from java-dataloader.

Yeah, so I think the dataloader will above of DAG fetcher
Thx~

from java-dataloader.