Final v0.9.0 Tasks (and Namespacing)

[jdantonio]

TL;DR: I believe we are ready to lock in the public API, release v0.9.0, and begin work on v1.0.

With PRs #322 and #323 (and the BiasedConditionVariable requested by the Rails team) I believe we are ready to release v0.9.0. This issue is to discuss final thoughts before that release.

When we first began discussing v0.9.0, in Issue #257, our goal was to stabilize the API so that we could focus on a stable, reliable, performant 1.0 release. Although many of the changes were of huge value to our users (creating an Edge gem, the common synchronization layer, bug fixes, etc.) the scope of the release began to grow. Many of the changes we undertook were internal changes only that had no functional benefit (newer pattern for runtime-specific subclasses, submodules for private abstractions). As a result we pushed the release date numerous times. I believe this was a mistake. We lost sight of the fact that "shipped" software is always better than "perfect" software.

One of the major discussions concerned the "namespace" modules for internal classes (see #261). At issue was whether or not we should make internal changes for greater consistency and to reflect common Ruby conventions. Although these suggestions had merit, they also represented a significant effort and a departure from what was already working an in production for many if our users. I began to reorganize some of what was discussed in PR #304 but had to abandon that effort due to merge conflicts. Some of that work was revisited in PRs #316, #317, #319, #320, and #321.

At this time I believe very strongly that we should release v0.9.0, move forward with the original plan for v1.0, and revisit namespace reorganization after v1.0.

According to the rules of Semantic Versioning, only changes to the public API must result in changes to the major and minor version numbers. Changes that are private are exempt. The namespace changes we discussed in #261 are private changes. We can make those changes, slowly and deliberately, once v1.0 is released an not violate the rules of Semantic Versioning. To facilitate this I have tagged all internal, private abstractions with the appropriate Yardoc tag (see PR #322). This PR (and any subsequent updates) should satisfy the requirements of Semantic Versioning.

Additional thoughts? Agree/disagree? @pitr-ch @mighe @chrisseaton @rkday @obrok @lucasallan @headius

[pitr-ch]

I think we should have another pre-release with the complete change-log (so users like @SebastianEdwards have time to try it before we release). Since we are releasing edge for the first time I would also like #314 to be merged (still few thinks to do there) before pre2.

We should also talk about how do we edge releases in future. Since it shares README file and some other pieces I think we should always bump patch (or minor?) version of core and do appropriate semantic bump on edge.

When would you like to add the BiasedCondition 0.9 or 1.0? It will require more time to understand it and to decide if and how do we add it to concurrent-ruby.

Nice work on the #316, #317, #319, #320, and #321! Before we release I would like to do a pass on whole public API if there is something we should consider changing, I'll comment here again after it's done. I am fine with doing rest of the changes on private parts later (preferably between 0.9-1.0 though).

My todos:

• Changelog updates 0.9.pre2
• Finish and merge Futures: passing args on construction, interoperability with channel, documentation #314 0.9.pre2
• go through whole public API 0.9.pre2
• resolve BiasedCondition 1.0

[jdantonio]

In no particular order:

• I'm comfortable doing another pre-release of 0.9.
  • We should discuss Prevent Ruby thread pools from creating too many threads. #326 and discuss a final solution.
  • We can certainly merge Futures: passing args on construction, interoperability with channel, documentation #314, too, but I'd like to get that done as soon as reasonably possible. (Any idea how long that may take?)
• We can wait until 1.0 or later for BiasedCondition. I didn't look at the feature request too closely, I just assumed (perhaps incorrectly) that it wouldn't be difficult. We should definitely wait if necessary.
• I'd like to freeze the current feature set, unless there is something absolutely critical still outstanding. (I've probably been the worst about adding new features into this release.)
• I like the idea of releasing Edge in conjunction with Core, the same way we plan to release Ext and Core simultaneously. We can document something in the README later.
• We definitely need more people to review the private/public API. As we discussed in Marked private API abstractions. #322, I want to revisit the Yardoc tags with an eye to generating the best documentation.
• I'd like some help updating the CHANGELOG in PR Began updating README and CHANGELOG #323.

[sschepens]

Are you planning on pushing pre-releases to rubygems? I wish to test them

[jdantonio]

@sschepens Yes, I will be publishing v0.9.0.pre2 this weekend. Are you OK downloading the gem files from GitHub, or do you need them pushed to Rubygems?

[sschepens]

@jdantonio it would be great if you could push them to Rubygems, i'm developing a gem that would be using concurrent-ruby as a dependecy so i need it to be in Rubygems :)

[jdantonio]

@sschepens 🆒 I'll push to Rubygems and make a note on this thread when I build the next pre-release. Do you mind if I ask what you are building? I'm very curious to see how others use this gem.

[sschepens]

I'm building an Amazon SQS message processor, it's based on shoryuken which uses Celluloid. I forked it because i was having memory leaks and wanted to change some things, I tried to use Celluloid but was having leak issues too and some other problems.
So I decided to give concurrent-ruby a try!

[jdantonio]

Awesome! Let us know how it goes and if we can do anything to help.

[pitr-ch]

@sschepens: exactly what @jdantonio said. Any feedback is very valuable and appreciated.

[sschepens]

@jdantonio @pitr-ch thanks!
Actor reset loop issue snipped and moved to #328

[jdantonio]

@pitr-ch I would like to release v0.9.0 this weekend. It seems like all the known bug have been addressed. Thoughts?