Fs2: Add fs2-reactive to functional-streams-for-scala

[zainab-ali]

fs2-reactive is a reactive streams implementation for fs2.
It will be ready for a first release over the coming weeks.

Should this be published under the fs2 group id, and kept under the functional-streams-for-scala organization?

I am happy to be a core maintainer, but think this would help adoption and visibility.

[zainab-ali]

To clarify, fs2-reactive is not ready for release. There are plenty of outstanding issues and further tests, docs and code to be added. However, these are likely to be tackled over the coming weeks.

Once these are tackled, it would be worth thinking about what to publish it under.

[rossabaker]

This library is a good for the fs2 ecosystem. I don't enjoy Reactive Streams, but I enjoy the integrations it has enabled. I am glad to see this, and expect to integrate it as an optional module for http4s.

[adelbertc]

👍 for same reasons as Ross. I am by no means a proponent of reactive streams, my interest in this is primarily motivated for work reasons. I am looking into doing hot-reloading of configuration by providing a stream of confs and my tool of choice is FS2. A Java team here would like to re-use it instead of building their own, and abstracting over the streaming portion with Reactive Streams seems useful.

[mpilquist]

:+1: from me. Given the stability of rx api, I'm fine with creating the fs2-reactive project as a submodule in the main fs2 repository, with the following caveats:

• The only dependency is "org.reactivestreams" % "reactive-streams" % "1.0.0"
• The implementation is parametric in effect type

[zainab-ali]

At the moment there is an additional dependency on log4s, I'm somewhat reluctant to remove logging given the convoluted stateful nature of the reactive streams spec.

A parametric effect type would certainly be worth having - I'll test it out against the tck.

[pchiusano]

I would call it fs2-reactive-streams. It's a binding to reactive-streams.
Just "reactive" could mean anything...

I am 👎 on hosting this in main repo (even as submodule), but okay with it
being under fs2 org. The rx streams dependency may be very stable but there
are enough design choices with the binding that I think it's better to keep
separate and evolve independently. However I'll defer to @mpilquist on
decision.

IMO, the log4j dependency is not good. I'd find a better way of providing
monitoring of streams without baking in a dependency like that.

On Thu, Apr 13, 2017 at 9:33 AM zainab-ali [e-mail] wrote:

At the moment there is an additional dependency on log4s, I'm somewhat
reluctant to remove logging given the convoluted stateful nature of the
reactive streams spec.

A parametric effect type would certainly be worth having - I'll test it
out against the tck.

—
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHub
https://github.com/functional-streams-for-scala/fs2/issues/844#issuecomment-293896528,
or mute the thread
https://github.com/notifications/unsubscribe-auth/AAArQvbw8PtLbeOZuy91p8xzWtw41iIlks5rviQtgaJpZM4M4ES1
.

[pchlupacek]

I am sort of on same page like @pchiusano here. I think it will be excellent to list library in readme.md. My feeling is that (perhaps not correct) that anything that moved under fs2 repo is where all collaborators have to support such library / extension etc. Not sure if we shouldn't do this after a while library is there and we get first hands on.

[zainab-ali]

Thanks for the comments.

@mpilquist I have replaced Task with a parametric effect type.
@pchiusano I have removed the dependency on log4s - the core module only depends on reactivestreams. The docs module will have a dependency on other streaming libraries (e.g. akka streams) to keep it machine testable with tut, but that won't cause problems for end users.
I will also rename this to fs2-reactive-streams for clarity.

My own preference is still to have this within the fs2 org, purely because it's useful for early adopters. All other streaming libraries come with reactivestreams support out of the box, and this is something adopters tend to look for.

However, I do think it should be a separate project. The reactivestreams spec is not something encouraged in a functional architecture.

[zainab-ali]

Following the above comments, I've made a first release under com.github.zainab-ali and will reference this in the fs2 README

[mpilquist]

Reopening this as fs2-reactive-streams has proven to be an integral part of the wider ecosystem -- for example, it's used in http4s. Keeping it in a separate repository has been problematic, as breakages in the FS2 main repo have to be dealt with after the fact in the fs2-reactive-streams repository. Hence, I strongly advocate for bringing this in as a submodule to the core repository for 1.0.0 and adding @zainab-ali as a contributor if desired by Zainab. Adding the submodule will increase the maintenance burden but that's intentional in this case -- we need to keep this module working as FS2 evolves.

[rossabaker]

:+1: reactive-streams is unlikely to see a major version bump and does not require Scala bumps. It is a new dependency, but is about as safe as they come.

[SystemFw]

Adding the submodule will increase the maintenance burden but that's intentional in this case -- we need to keep this module working as FS2 evolves.

For all practical means, I'm already contributing heavily to reactive-streams to keep it in sync, so :+1: to this

[ngbinh]

This is great!

[zainab-ali]

:+1: I'm highly in favor of it! If there are no objections, I will submit a PR to fs2 with the new module added. It's worth noting that fs2-reactive-streams has at least one transient test failure which need to be addressed before this is merged.

[mpilquist]

@zainab-ali I checked with Pavel today and he's good with this too so let's go for it. I'll add you as a member now.