Fixes and inference improvements to ZSink combinators #1686
Conversation
1fd0468 to
0218383
Compare
0218383 to
4f3df9d
Compare
|
|
||
| s1.mergeWith(s2)(_.toString, _.toString) | ||
| .run(Sink.succeed("done")) | ||
| .run(Sink.succeed[String, String]("done")) |
There was a problem hiding this comment.
Arg. This is one of the places where it's a loss.
There was a problem hiding this comment.
@jdegoes Oh it's good to have the grandmaster of variance and inference here.
There's a really annoying issue that the Stream's element type parameter isn't used to infer the Sink's input type parameter - in run, transduce and others. Any thoughts on that?
There was a problem hiding this comment.
Can you paste signatures and I'll make it work (if it's possible 😆 )? Thank you! 🙏
There was a problem hiding this comment.
trait ZSink[-R, +E, A, +B]
object ZSink {
final def succeed[A, B](b: B): ZSink[Any, Nothing, A, B] = new ZSink[Any, Nothing, A, B] {}
}
trait ZStream[-R, +E, +A] {
def run[R1 <: R, E1 >: E, A1 >: A, B](sink: ZSink[R1, E1, A1, B]): Unit = ()
}
val stream = new ZStream[Any, Nothing, String] {}
stream.run(ZSink.succeed("OK"))
// cmd0.sc:11: type mismatch;
// found : ammonite.$sess.cmd0.ZSink[Any,Nothing,Nothing,String]
// required: ammonite.$sess.cmd0.ZSink[Any,Nothing,String,String]There was a problem hiding this comment.
I think this case is beyond help. The covariance on A in ZStream is interacting with the invariance on ZSink to cause the problem, and we certainly don't want to eliminate covariance on ZStream
| * Creates a sink halting with the specified `Throwable`. | ||
| */ | ||
| final def die(e: Throwable): ZSink[Any, Nothing, Nothing, Any, Nothing] = | ||
| final def die(e: Throwable): ZSink[Any, Nothing, Any, Nothing] = |
There was a problem hiding this comment.
Just to mark them for having Any, die, dieMessage and drain.
| * the predicate `p`. | ||
| */ | ||
| final def ignoreWhile[A](p: A => Boolean): ZSink[Any, Nothing, A, A, Unit] = | ||
| final def ignoreWhile[A](p: A => Boolean): ZSink[Any, Nothing, A, Unit] = |
There was a problem hiding this comment.
I'd like this to be a SinkPure and not delegate through ignoreWhileM. Ignore if you think it's not worth it.
There was a problem hiding this comment.
Yeah we'll do that in a follow-up. There are several combinators that can be overriden in SinkPure.
|
Another day another Initial review done. I will be around for follow up changes. Looking great. |
|
Hmm. So, I went over the changes again and I admit I am not sold on whether the inference changes are a win or not. So I'm going to backpedal on this and keep |
|
I stand by your decision. Would you be willing to fork out the |
|
@vasilmkd I moved the definitions of combinators that use |
|
@iravid I saw that and like it a lot. Big 👍. It will take some more consideration when implementing future combinators but the removal of implicit compiler evidence is worth it in my opinion. Can I bother you to go through the tests again and try to remove as much of the type parameters as possible? Just as a testament to the improved inference. E.g. |
| * see [[ZSink.succeed]] | ||
| */ | ||
| final def succeed[B](b: B): Sink[Nothing, Nothing, Any, B] = | ||
| final def succeed[A, B](b: B): Sink[Nothing, A, A, B] = |
There was a problem hiding this comment.
A should be reverted now. 😄
There was a problem hiding this comment.
It's actually required because of the leftovers.
There was a problem hiding this comment.
Sorry yes, I skimmed over the implementation.
| object ZSink extends ZSinkPlatformSpecific { | ||
|
|
||
| implicit class InputRemainderOps[R, E, A, B](val sink: ZSink[R, E, A, A, B]) extends AnyVal { | ||
| implicit class InputRemainderOps[R, E, A, B](private val sink: ZSink[R, E, A, A, B]) { |
There was a problem hiding this comment.
Can't work because of the inner classes that are defined. Not a big loss.
There was a problem hiding this comment.
Sorry, can you point me to the defined classes? I can't find anything.
There was a problem hiding this comment.
case class State in collectAllN
| * Filters the inputs fed to this sink. | ||
| */ | ||
| def filter(f: A => Boolean): ZSink[R, E, A0, A, B] = | ||
| self match { |
There was a problem hiding this comment.
Nitpicking. Not sure how I feel about having SinkPure implementations in this file from a discoverability and maintainability standpoint.
There was a problem hiding this comment.
Well, since filter was moved out to an implicit class to improve inference, it can no longer be overriden in `SinkPure.
There was a problem hiding this comment.
I'm well aware... we'll live with it. :-)
| } | ||
| } | ||
|
|
||
| private[ZSink] object internal { |
There was a problem hiding this comment.
Can the Optional ADT be migrated to ? as well? OTOH Not sure if Side is used in multiple combinators.
| def stepPure(state: State, a: Any) = () | ||
| def extractPure(state: State) = Right((b, Chunk.empty)) | ||
| def cont(state: State) = false | ||
| final def succeed[A, B](b: B): ZSink[Any, Nothing, A, A, B] = |
There was a problem hiding this comment.
I'm guessing this cannot be inferred from the sink before this one? :/
There was a problem hiding this comment.
No, unfortunately not. Same problem as with ZStream#run
- Rewrite ZSink#collectAllWhileWith,collectAllN,untilOutput - Move some combinators to implicit classes for improved inference - Fix ZSink#takeWhile's continuation predicate
a438f12 to
7841938
Compare
- Rewrite ZSink#collectAllWhileWith,collectAllN,untilOutput - Move some combinators to implicit classes for improved inference - Fix ZSink#takeWhile's continuation predicate
Resolves #1682
Resolves #1676
@vasilmkd - could you have a look? Thanks!