Support multiple-entry flowables

[badgerwithagun]

Been playing around with some RX patterns and wondered if this was possible in a generic way. Turns out it is, sort of.

With these changes, the following spits out each item as a string on a new line as the items arrive, all using async:

        @GET
        @Path("stringStream")
        @Streamable // Activates the new code
        public Flowable<String> stringStream() {
            return Flowable.range(1, 49)
                .zipWith(Flowable.interval(1, SECOND), (it, interval) -> it)
                .map(i -> "Item " + i + "\n");
        }

What I've put together is also compatible with Jackson's JsonCreator, so you can just use @Streamable(writer = SomeCustomWriter.class and call JsonCreator from there:

        @GET
        @Path("objectStream")
        @Streamable(writer = SomeCustomWriter.class)
        public Flowable<Tuple> objectStream() {
            return Flowable.range(1, 49)
                .zipWith(Flowable.interval(1, SECOND), (it, interval) -> it)
                .map(Tuple::new);
        }

Not sure it's perfect, and I need to test it a while lot more, but I thought I'd pop it here here for thoughts.

[badgerwithagun]

Demo app: https://github.com/badgerwithagun/async-all-the-way-down

[alex-shpak]

Hi, thank you for your work!
I will need to check more deeply. In general when I think about emitting multiple values from flowable I see two alternatives:

1. Flowable could be streamed as list automatically;
2. Or Flowable could emit server-sent events: https://eclipse-ee4j.github.io/jersey.github.io/documentation/latest/sse.html

Perhaps if time between emits is large it is not expected by client to keep listen on connection.

[badgerwithagun]

In my use-case, I am using JsonFactory in Jackson to stream a JSON array back to the client. This allows a fully-reactive pipeline:

• Asynchronous JAX-RS entrypoint
• Non-blocking DB query loading millions of records (e.g. using R2DBC)
• Each record is piped to the client as it arrives
• Client can pipe those to its own processing code as they arrive
• Server and client are now processing the same data stream simultaneously
• No-one is blocking any threads :)