GQLGW-5297-optimise-incremental-part-execution-for-defer-requests - #4174
Conversation
Sorry, something went wrong.
| initialResult.incrementalItemPublisher.subscribe(sub) | ||
|
|
||
| then: | ||
| Awaitility.await().untilTrue(sub.isDone()) |
There was a problem hiding this comment.
Needs an assert on the contents of the Publisher
Sorry, something went wrong.
| && deferredExecutionSupport.deferredFieldsCount() > 0 | ||
| && executionContext.getGraphQLContext().getBoolean(IncrementalExecutionContextKeys.ENABLE_EAGER_DEFER_START, false)) { | ||
|
|
||
| executionContext.getIncrementalCallState().startDeferredCalls(); |
There was a problem hiding this comment.
This call here is only need as a side effect to create the memoizing supplier (of the Publisher)
I think we should invent a new method that does that and hides the side effect
executionContext.getIncrementalCallState().startEarlyDraining();
Sorry, something went wrong.
There was a problem hiding this comment.
Hang on you have that - startDrainingNow() and it calls publisher.get()
I dont think executionContext.getIncrementalCallState().startDeferredCalls(); is needed
Sorry, something went wrong.
|
|
||
| } | ||
|
|
||
|
|
There was a problem hiding this comment.
We need to create a new entry in the unusual config - lets put it in
graphql.GraphQLUnusualConfiguration.IncrementalSupportConfig
/**
* This controls whether @defer field execution starts as early as possible.
*/
@ExperimentalApi
public IncrementalSupportConfig enableEarlyFieldExecution(boolean enable) {
contextConfig.put(YOUR_KEY_HERE, enable);
return this;
}
Sorry, something went wrong.
| } | ||
|
|
||
| public void startDrainingNow() { | ||
| startDeferredCalls(); |
There was a problem hiding this comment.
no need to call start Deferred calls as it just creates the publisher
Sorry, something went wrong.
This is to optimise defer execution for incremental results to begin processing as soon as the first incremental call is detected rather than on completion of the initial result.