@SseProduces looks to fit nicely with the high-level declarative API I asked for recently. :-)
But for a low-level programmatic API as proposed by Oracle I think it does not fit to have a declaration that serves as a default for omitted programmatic declaration of the type. In this case I would think it is more streamlined to have some programmatic solution like sseContext.setDefaultMessageType(T) or something like that.
-Markus
From: Sergey Beryozkin [mailto:sberyozkin_at_talend.com]
Sent: Montag, 6. Februar 2017 13:14
To: jsr370-experts_at_jax-rs-spec.java.net
Subject: Re: Server-Sent Events API proposal
I'm not keen to spend much time on this possible improvement, but I'll summarize it here with the code example.
Typical JAX-RS (static):
// text/plain
@GET
@Produces("text/plain")
@GET
public Response getText() {
return Response.ok().entity("somedata").build();
}
Typical JAX-RS (overrides static):
// text/xml overrides text/plain
@GET
@Produces("text/plain")
public Response getText() {
return Response.ok().type("text/xml").entity("<somedata/>").build();
}
At this stage we know the users can override the static Produces if needed, but let the static Produces affect the media type
if the user does not want to repeat the obvious "text/plain" in the code
In SSE, one can do something similar:
@GET
// this is a media type for the HTTP client which invoked this method
@Produces("text/plain")
// this is a media type for the OutboundEvent.Builder
@SSEProduces ("someMediaType")
public Response getText() {
// sse event without having to set a media type on the builder
// for the HTTP client
return Response.ok().type("text/xml").entity("<somedata/>").build();
}
The examples show the basic string data, assuming in most cases people will want JSON, it would make sense to let the developes
just type something like @SSEProduces ("application/json") as opposed to having them to do that on the builder every time.
This is arguably a minor improvement, possibly part of the higher-level API idea...
Sergey
On 06/02/17 09:41, Sergey Beryozkin wrote:
I do not remember us agreeing at all.
You are referring to the media type required to initiate the flow, I'm referring to the one users have to set afterwards from the sse builders.
And why exactly you can not set that statically if it what users would prefer if they always use the same type ?
Sergey
On 05/02/17 19:10, Markus KARG wrote:
I can remember that thread and I thought we agreed that the current proposal from Oracle is correct, because to initiate the SSE cannel actually the @Produces defines the media type of SSE, not of the message payloads. As all messages can have different payloads, IMHO we cannot set the type statically.
-Markus
From: Sergey Beryozkin [mailto:sberyozkin_at_talend.com]
Sent: Sonntag, 5. Februar 2017 18:23
To: jsr370-experts_at_jax-rs-spec.java.net
Subject: Re: Server-Sent Events API proposal
Hi Pavel
There was an issue I was trying to raise when the SSE draft was originally proposed.
Specifically, it was about a default media type for SSE data.
Right now one needs to set a media type with the SSE builder. The question was if we could let people use @Produces,
so that they do not set it every time an SSE data is returned, similarly to the way it is done in the 'traditional' JAX-RS - one either
uses @Produces or uses Response to customize
What do you think ?
Thanks, Sergey
On 27/01/17 15:03, Sergey Beryozkin wrote:
I've only glanced so far over the proposed changes, it all looks very neat, but we will def provide more feedback later on.
As far as Flow is concerned, is the idea to use it only as part of SSE ? What will happen to it once the future JAX-RS API will be Java 9-only based ?
Cheers, Sergey
On 27/01/17 14:54, Sergey Beryozkin wrote:
Hi, I've been planning to move to m03 next week and see how the latest reactive API fits,
As far as SSE is concerned I've asked my colleague Andriy who did implement the initial SSE proposal to evaluate the latest one when he gets a chance and I will also have a look but unlikely next week,
I recall Andriy raised an issue on this list, something to do with initializing SSE Broadcasters, Marek said at a time he would look at it.
Thanks, Sergey
On 27/01/17 07:55, Pavel Bucek wrote:
No commends/feedback?
I understand that the change is not a small one - feel free to ask if you need to have something explained in more detail.
Please let me know if someone is planning to look into this topic. Ideally by the end of this week (start of the day Monday, CET). If there are no comments, we might need to move it forward even without any other input, but I don't really want to do that..
Thanks and regards,
Pavel
On 24/01/2017 20:54, Pavel Bucek wrote:
Dear experts,
please allow me to bring up another addition planned to be included in JAX-RS 2.1: Support for Server Sent Events.
The API was introduced a while back [1], but we did incorporate some changes, mostly alignments with Java SE 8 and another one, slightly more controversial - alignment with Java SE 9 Flow API.
Why we should use Flow API? We want to make the transition to Java 9 as smooth as possible. Please note that today, the JAX-RS sources do contain Flow class (1:1 copy from Java SE 9), but that is NOT a final state. We want to work with it as much as possible, but we will minimize and clean up the code before final release. How? Consider following example:
public interface SseBroadcaster extends AutoCloseable, Flow.Publisher<OutboundSseEvent> {
//...
}
could be modified to
public interface SseBroadcaster extends AutoCloseable {
//...
void subscribe(SseSubscriber<? super OutboundSseEvent> subscriber);
}
I hope you will like the attempt to be "forward-compatible", but I assume that this proposal alone will bring some opinions - please share them! Last note: currently, the Flow API is used only in SSE. We plan to use it heavily in Non-blocking I/O. If you think there are other areas where we can take advantage of Subscriber/Publisher methods, please let us know.
Let me get back to the SSE API.
Very brief description of the API is on the wiki [2], I will expand that once we'll have some feedback.
The proposal contains support for server and client side. All classes are in the package javax.ws.rs.sse [3]. There is no added class from the last proposal, but there are some changes:
- [server] SseBroadcaster extends Publisher<OutboundSseEvent>
- [server] replaced SseBroadcaster.Listener by SseBroadcaster#onException and SseBroadcaster#onClose
- [server] SseEventOutput extends Subscriber<OutboundSseEvent>
- [client] SseEventSource extends Publisher<InboundSseEvent>
Corresponding commit [4] looks like a bigger one, but most of it is renaming based on introduced interfaces, the core of the change is relatively small.
I look forward to your comments and suggestions.
I don't want to prolong this already too long email - please let me know if there are some areas which I should describe in more detail.
Thanks and regards,
Pavel & Santiago
[1]
https://java.net/projects/jax-rs-spec/lists/jsr370-experts/archive/2015-10/message/28
[2]
https://java.net/projects/jax-rs-spec/pages/ServerSentEvents
[3]
https://github.com/jax-rs/api/tree/master/jaxrs-api/src/main/java/javax/ws/rs/sse
[4]
https://github.com/jax-rs/api/commit/b2b8f3f4f20696558a3ff52b0de17fb04c343d02