functional-streams-for-scala/fs2

Most discussion is on Typelevel Discord: https://discord.gg/bSQBZA3Ced

Fabio Labella

@SystemFw

You can just look at the code to see how they are implemented

that applies also to the exercises in the subsequent sections

they are all operations already in the library

Justin Heyes-Jones

@justinhj

Ah I see. Yes I can do that. As a new user to the library I found some of the implementations a bit intimidating because for example eval_ is implemented using eval which uses internal private types FromC and Algebra. I thought the intent would be to implement them more from the library users point of view.

Fabio Labella

@SystemFw

yeah, that's the point

you can treat eval as a primitive

and in fact, it is a primitive, that's why it goes to Algebra

so how do you implement eval_ using eval and the other things you have learned so far?

you are not supposed to implement eval_ from scratch :)

Justin Heyes-Jones

@justinhj

ah that makes sense. I did figure it out in the end I just wasn't sure whether I'd done it right, but I had.

Taylor Brown

@tbrown1979

Hi everyone, so I have a Stream[F, Byte] and I'd like to create a stream that emits 128kbps. I'd like to keep the type Stream[F, Byte]. To do this I figured I could make each Chunk 128kb and have each Chunk process every second.

Something like this:

      val perSecond = scheduler.awakeEvery(500 millis)

      val stream: Stream[F, Byte] =
        inputStream
          .chunks
          .zip(perSecond)
          .map(_._1)
          .flatMap(Stream.chunk(_).covary[F])
          .repeat

This works nicely (if there is a better way than this please let me know, I am an fs2 noob). The issue I'm having right now is making each Chunk the proper size. I want each chunk to contain 16000 Bytes, however when I read in my InputStream it comes in at ~8000 bytes per chunk. Is there a way I can condense these chunks so that they're all roughly 16k bytes? I've looked through the Stream code and I see a lot of things about limiting chunk size, but that's just not what I want. Is this a completely dumb way of getting the result I want? Any help is appreciated. Thanks

Christopher Davenport

@ChristopherDavenport

Take a look at rechunk or rechunkN, not near a computer but I believe that gets you what you are looking for.

Justin Heyes-Jones

@justinhj

Another newb question. I'm trying to write a Pull function that's similar to take(n) but that will take stream elements until has seen n elements that match a predicate. For example a stream of Char, I would like to take the stream elements until I have seen 5 'a's. I'm thinking it would be similar to the tk function in the samples using scanSegmentsOpt and take elements one at a time, incrementing the count when it sees an 'a'. The problem I'm having is how to look at the next element. It looks like peek1 may be what I want but I can’t seem to use it from a segment or find an example of its use.

Jose C

@jmcardon

@justinhj uncons1

Justin Heyes-Jones

@justinhj

haha short and sweet, thanks

Jose C

@jmcardon

@justinhj this is just for the sake of an exercise correct?

because the actual implementation of take is in terms of uncons

Justin Heyes-Jones

@justinhj

I'm just extending the exercise in a different way for the learning experience. I was taking the first implementation of tk (that uses scanSegmentsOpt) as a starting point rather than the second one that uses uncons

Jose C

@jmcardon

:+1: cool. yeah that's a good way to learn

Denis Mikhaylov

@notxcain

Hi! How do I change the effect type? I have a FunctionK[F, G] and a Stream[F, A], and I need Stream[G, A]

epifania villamor

@fgalaban_twitter

Denis Mikhaylov

@notxcain

Ah, nevermind, found a way around

Fabio Labella

@SystemFw

@notxcain generally you can do so with translate, or in the case of Pure ~> F, simply use covary

Denis Mikhaylov

@notxcain

@SystemFw nice, thank you!

Paulius Imbrasas

@CremboC

hey @SystemFw , we're running a service reading from Google PubSub, and we occassionaly this beautiful stack trace: https://gist.github.com/CremboC/6f3f158bd32720890e47519211a1a71d (sorry it's in one line, but you'll see a pattern quickly!). any ideas?

Fabio Labella

@SystemFw

@CremboC what version are you on?

this was fixed in 0.10.3

Paulius Imbrasas

@CremboC

ah, on 0.10.1

I assume they're backwards compatible?

Fabio Labella

@SystemFw

yep

Paulius Imbrasas

@CremboC

coolio, will update, danke!

Justin Heyes-Jones

@justinhj

I've been trying to write a function that streams until it has seen N objects equal to a provided instance of O. The last piece of the puzzle is how to emit the stream. I think I need to pull output but somehow need to join that to the tail of the stream in the recursive call? https://gist.github.com/justinhj/10fd8dc2a364a1fdbade36352b15abe0

Peter Aaser

@PeterAaser

I'm not sure what you are trying to do here

maintaining a list, but with Pipe[F,O,O] as type seems incompatible

Pull.output1 might be what you want to use

typically you compose Pulls with >>

how about

s.pull.uncons1.flatMap {

            case Some((o, tl)) =>

              if(o == thing) {
                output1(o) >> go(tl, n - 1)
              }
              else {
                output1(o) >> go(tl,n)
              }

            case None =>
              Pull.done

does that seem like what you are after @justinhj

Peter Aaser

@PeterAaser

Pull forms a monad in R btw, so >> is just myPull.flatMap{ _ => ... } if that helps any (but this is in no way necessary to understand for using pulls)

Jose C

@jmcardon

@PeterAaser sort of useful when you need to evaluate effects in a Pull though

since eval returns it in R not O

Fabio Labella

@SystemFw

@justinhj you need to Pull.output/output1 as @PeterAaser says, before recursing (using >> to tie things)

another problem is that you are accumulating a List[O] as your output

actually that's the confusion

you are used to the typical accumulator based thing, where you first accumulate the whole result and then you "return it"

but here you want to do this in a streaming fashion, so as soon as you find something that needs to be emitted, you just emit it

and Pull.output/output1 is the command that lets you do so

Fabio Labella

@SystemFw

I can explain the idea behind Pull a bit more in depth, if you want

Peter Aaser

@PeterAaser

To make it very clear, imagine if the stream never reaches enough elements to halt

Where communities thrive

People

Repo info

Activity