Does ReactiveSwift have any plan to support defer? about reactiveswift

[samidalouche]

I don't feel comfortable enough with ReactiveSwift's codebase to put together a concrete proposal, but I can gather a summary of the different back pressure mechanisms used across the various reactive platforms so that it can serve as a a start for the discussions.

[andersio]

SignalProducer?

[samidalouche]

using stream1.concat(...) with a SignalProducer will start the SignalProducer before the client actually gets to consume the stream1 entirely, which is useless in the context of backpressure (that can be used to implement paging of infinite resources in the case of a slow consumer)

[andersio]

If you don't mind, would you please provide a more concrete context or examples of the obstacles you are facing? Otherwise, it would be hard to understand what you are looking for.

Your original question on RxSwift.deferred does already have an answer: it works in a similar semantic as SignalProducer as a type. It defers the creation of the signal until it is started, and creates a signal for each start attempts, thus being called "producer".

http://reactivex.io/documentation/operators/defer.html

[samidalouche]

Let's say that I have an HTTP API that provides a virtually unlimited amount of data through paging.

The basic use case is that I would like to provide a SignalProducer that will provide all the elements from the HTTP API. However, let's say that the consumer of the SignalProducer is much slower than the fetching of the HTTP API (and that it runs in a different scheduler). What will happen with ReactiveSwift is that an unbounded buffer of elements will accumulate until it exhausts all the memory.

What I would love to see is any way in ReactiveSwift to circumvent this issue and make the producer as slow as the consumer. Does that make sense?

[mdiep]

We don't have a plan to support defer, but I'm definitely interested in providing some back pressure mechanisms. Would you be interested in putting together a proposal or PR? That'd be the best way to move something like this forward. Or even just putting together a survey of different back pressure mechanisms used across different reactive platforms.

[RuiAAPeres]

Closing due to inactivity. If you are still interested in discussing this, please open a new Issue with your purpose (or a set of links with different approaches for the mechanism).