Support `@stream` directive extension

There is a proposal spec for @stream directive that supposed to address streaming list results:

• Similar to @defer, but for lists
• Will load each element of the list as it is ready
• Show users the first item in the list without waiting for the whole list

Request

query {
  talks @stream {
    title
    time
    speaker {
      name
    }
  }
}

Initial Response

{
  "data": {
    "talks": []
  }
}

Patch for first talk

{
  "path": ["talks", 0],
  "data":  {
    "title": "Opening Keynote",
    "time": "10:30-11:00",
    "speaker": {
      "name": "Sean Larkin"
    }
  }
}

Patch for second talk

{
  "path": ["talks", 1],
  "data":  {
    "title": "Knock knock, who’s there?",
    "time": "11:30-12:00",
    "speaker": {
      "name": "Sam Bellen"
    }
  }
}

Having native @stream directive support in graphql-java will be great for streaming large result sets without having to wait for the whole root list or any embedded list to resolve in memory for https://github.com/introproventures/graphql-jpa-query in order to minimize memory usage while fetching large results from the database.

I tried to use subscriptions for streaming large results via SSE or HTTP streaming, but it is very limited for queries type, i.e. https://github.com/introproventures/graphql-jpa-query/pull/229#issue-353230436

I can collaborate and contribute to graphql-java

[andimarek]

@igdianov please follow the development here: https://github.com/graphql/graphql-spec/blob/master/rfcs/DeferStream.md

GraphQL Java will not implement any stream/defer until the spec for has made some serious progress: we don’t have the resources to support an alpha version of it. We already implemented defer once (it is actually still available) but then the community never implemented it more widely.

[manuelfink]

May I ask, is there any progress on @stream directive and @defer?

For data heavy applications @stream and @defer are very interesting as they solve the issue when querying larger amounts of data, which are heavy and slow.

I guess many users are not using defer yet, as it is in experimental state. Side questions, but maybe not related to this issue: are there any plans to extend defer from experimental to stable? When do you normally declare a feature as stable?