Support for Server-Side Events / Streaming Endpoint

[x80486]

I'm not sure if this is possible, but these days it's very common for an endpoint to provide a stream of data. In order to simulate this, it would require that API Simulator should be able to keep the connection open while real time data is streamed through an HTTP response/endpoint.

For the most part, I read that there is support for HTTP/2, so I guess this could be doable.

[apisim]

Hi @x80486

As of now, API Simulator supports Request/Response/Callback (aka Request/Acknowledge/Callback) message exchange pattern. Notice that the Callback is over a new connection that API Simulator opens to the Callback endpoint's server; the protocol could be HTTP/1.1 or HTTP/2.

As far as streaming, API Simulator supports streaming of large files in response to request to simulate, for example, (potentially large) file download.

Not sure if any of the above is what you a looking for or something more in line with server-sent events or HTTP/2 Server Push.

Are there some concrete, practical Use Cases you are looking to implement or just exploring new features we can eventually add to API Simulator, if not supported already?

Cheers!

[x80486]

It's more on the realm of the SSE or HTTP/2 server push; if you try to mock/simulate most of the APIs from Twitter or Facebook are available as a stream of events they sent back to the caller after some data is requested.

That was what I was thinking initially when I submitted this one, basically to support those scenarios where they do such things, and also these days more people are jumping into the reactive bandwagon, so this would be sort of the standard at some point.

[apisim]

Off the top of my head, I'm thinking that streaming back distinct events over the same HTTP/1.1 connection may be doable today with a bit of scripting and/or use of scriptlets. It won't be a scalable solution, though, as a thread will be dedicated to the long-running response and with using Thread.sleep to simulate delay between the events. The existing latency simulation is for a whole response or for a callback and they get scheduled for future execution using a separate thread pool to assure API Simulator can scale under load (API Simulator is often used in performance tests to create a controlled environment).

The foundation for HTTP/2 Server Push exists today but will have to be exposed "natively" in API Simulator as it is at a very low protocol level.

In either case, some thinking will have to go into how to configure this in the DSL and where the events will come from if they are not from randomly generated data.