Replies: 5 comments 23 replies
-
|
I think Akka will drop support to 2.12.x once build tools moved to Scala 3. |
Beta Was this translation helpful? Give feedback.
-
|
My original plan was for Cats Effect to do the same. This would be Sbt 2, basically. But @SethTisue is effectively arguing that we don't need to wait that long. |
Beta Was this translation helpful? Give feedback.
-
|
I think once Mill and Sbt move to Scala 3, Mission critical projects like CE can move drop 2.12.x Because I think that's a sign of Scala 3 is mutual enough for more adoption. |
Beta Was this translation helpful? Give feedback.
-
|
I don't know about Pekko, but the Akka libraries dropped 2.12 quite some time ago. |
Beta Was this translation helpful? Give feedback.
-
|
I'm using Pekko from Java with Scala 3 now. But as Akka is a commercial project now, We can see Akka's customers have move to 2.13.x or 3.3+. I think 2.12.x should support Java 25. The next LTS, and then , a good time to EOL. |
Beta Was this translation helpful? Give feedback.
-
|
I believe @mpilquist noted some of the ongoing pain felt by Fs2 due to the poor GADT handling in 2.12. That would be a pretty pragmatic motivation for dropping support in that project, and if Fs2 drops support, there's very little reason for Cats Effect to cling to it. For our part, 2.12 cross-building is very rarely a source of any meaningful heartache in Cats Effect. It does contort one or two things (most notably So I think what I'd like to see as far as homework is first, on a 1-10 scale where 10 is "Scala 2.10" and 1 is "meh", how much does @mpilquist and co want to nuke 2.12 from the matrix? Second, I'd love to see a bit of a semi-formal survey, probably based on the community build, of sbt transitive build dependencies. Is Laika the only major source of Cats Effect usage within the Sbt plugin ecosystem, or is there something else lurking out there we don't know about? It'd be good to get at least a semi-vague feel for this. Would also love to hear from @eed3si9n on any warm fuzzy around Sbt 2's progress. If Sbt 1 is going to move into "LTS but not current" state in the near-ish future (for some definition of "near"), that is an important variable. |
Beta Was this translation helpful? Give feedback.
-
|
Just to answer the question, i maintain a sbt plugin Stryker4s which has around 40k monthly downloads according to sonatype. However accurate that may be. It's heavily dependent on cats-effect. I'd love to drop 2.12 from the build matrix but until sbt 2 is used more than 1 I don't see it happening. What are the downloads for 2.12 like compared to other scala versions? That should be the first question I think. If keeping 2.12 is a relatively small maintenance burden my vote (not that I have one) would be to keep it in until some big blocker does arise. 2.12 still gets new releases and I think cats-effect is such a core ecosystem library that long support is something a lot of people look for. As long as it is reasonably maintainable |
Beta Was this translation helpful? Give feedback.
-
There exists sbt 2.0.0-M3 on Maven Central, which is not nothing. But even if we released the sbt 2.0 sometime in 2025, there's a long road ahead for the plugins and the builds to convert fully into sbt 2.x. Generally this conversation does feel like a déjà vu of people asking about sbt 0.13 to sbt 1.x transition to sunset Scala 2.10, so maybe this is a good sign. |
Beta Was this translation helpful? Give feedback.
-
|
Yeah the 2.10 situation was very similar. |
Beta Was this translation helpful? Give feedback.
-
|
@hugo-vrijswijk is there any reason your plugin couldn't just stay on the last Cats Effect minor version series with 2.12 support? To my knowledge, nobody is proposing dropping 2.12 entirely at present, only dropping it on the main branch. The previous minor version could still get critical updates. |
Beta Was this translation helpful? Give feedback.
-
Not particularly. Different versions for projects that cross-build is a lot isn't too uncommon, I suppose. From my pov cats-effect does seem to not have any crazy breaking changes in the near future while sbt 1 would still be a target |
Beta Was this translation helpful? Give feedback.
-
|
Pulling this up to the top level… The download statistics for Cats Effect (Core) fluctuate a bit, but the present ratios are roughly the following:
Scala 3 is trending upwards, and this seems to be coming mostly at the expense of Scala 2.12 (trending down) but also very slightly at the expense of Scala 2.13. Scala 2.12 downloads are pretty evenly distributed across a large number of recent-ish releases, while Scala 3 is heavily biased towards the absolute latest stable release (almost 50% of the Scala 3 downloads are 3.5.7. As an aside, this is the very first time I've looked at these numbers where Scala 3 usage was higher than Scala 2.12. I'm just eyeballing the last few months though; this isn't a statistically rigorous analysis. Also I'm fairly certain there are some odd error bars and artifacts in Sonatype's numbers, since I pulled the same numbers for Cats Effect Kernel and got meaningfully different results. I'm going to take Core as authoritative, but clearly this means there are some issues with transitive dependency download counting. |
Beta Was this translation helpful? Give feedback.
-
|
We all know why 2.12 support is important. Comments in this issue already state that it is not a burden to continue to support it. The answer is clear but I'll request it more formally: Please continue to support 2.12 until Spark is available in 2.13, as well as the cloud offerings for Spark (e.g. Databricks Runtimes, AWS EMR, etc). I personally maintain libraries that require 2.12 support due to spark. This choice would leave me unable to meaningfully update dependencies for little benefit. |
Beta Was this translation helpful? Give feedback.
-
|
Yes, runtimes do not support Finally, similar to the sbt discussion, nobody is saying we should delete |
Beta Was this translation helpful? Give feedback.
-
|
I think you meant that Runtimes don't support 2.13. They've supported 2.12 for a few years, I was personally pushing for these. And yes, Spark 4 should help make 2.12 obsolete with these vendors. Though I think 2.13 support in Databricks is still planned prior to the Spark 4 release. We may not have to wait that long. Unsure about the AWS EMR support. On your point with 3.7 not being released for a while, i understand it. But that also means there's no rush to do it. If runtimes support 2.13 before 3.7 is released, then we can introduce work to strip 2.12 support from the 3.7 branch. No need to get ahead of ourselves there. |
Beta Was this translation helpful? Give feedback.
-
Right, sorry, typo.
Probably, but again, IMHO, users would be fine with CE
Except it would add maintenance burden during the development of |
Beta Was this translation helpful? Give feedback.
-
|
There is not much of a maintenance burden as already noted in this thread. Not sure what the rush is with that in mind. We can just remove it when the vendors support 2.13. |
Beta Was this translation helpful? Give feedback.
-
|
Not much , but the burden must be pay |
Beta Was this translation helpful? Give feedback.
-
|
I just want to drop a link here to the Reddit discussion on the same topic: https://www.reddit.com/r/scala/comments/1i2zieq/should_cats_effect_drop_support_for_scala_212/ I don't think anything really unique has been said there (so far?), but it's additional data on what the community sentiment is. |
Beta Was this translation helpful? Give feedback.
-
Perhaps the next minor release of Cats Effect could drop Scala 2.12?
If any critical bugs or security issues are found in the previous minor release, more patch releases in that series could still be published.
previous similar discussions:
We had a long discussion about this on Discord the other day, beginning here: https://discord.com/channels/632277896739946517/632277897448652844/1328175562442801283
@djspiewak's main concern was that sbt 1.x is still on Scala 2.12 and as a result people might be using Cats Effect on 2.12 in their builds, even if their project code is on 2.13 and/or 3.
sbt 2.x will be on Scala 3, but it's still in milestones and we don't know how long the final release will take.
My own take was:
but I don't know the Typelevel ecosystem as well as some others.
Beta Was this translation helpful? Give feedback.
All reactions