functional-streams-for-scala/fs2

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

Fabio Labella

@SystemFw

my mental model for Timer is still "there's a Time[F] typeclass, the runtime gives you an implementation", e.g. IOApp

but I'm open to chaining my mind and more importantly I'm exhausted by having to think about it :joy: :joy:

Michael Pilquist

@mpilquist

oh ContextShift is okay to pass implicitly

trait ContextShift[F[_]] {
  def shift: F[Unit]
  def evalOn[A](ec: ExecutionContext)(fa: F[A]): F[A]
}

Fabio Labella

@SystemFw

ah, cool

it's defo a typeclass then in my mind

Christopher Davenport

@ChristopherDavenport

So happy to get ContextShift

Fabio Labella

@SystemFw

oh yeah, it's awesome

Thanks Pavel ;)

Calvin Lee Fernandes

@calvinlfer

hey guys, I have a question about the Pull datatype, what exactly does the R type represent? Can you think of it as some kind of Resource that you can use to produce Os?

Fabio Labella

@SystemFw

it's the return type

you can imagine it as a computation the emits some values and ~~emits~~ returns a new thing

it's actually similar to pagination ;)

which is indeed pull-based

Calvin Lee Fernandes

@calvinlfer

so there's two output types? O and R?

Fabio Labella

@SystemFw

R is not an output type, it's a return type

note that to convert a Pull to a Stream

R needs to Unit

Calvin Lee Fernandes

@calvinlfer

oh i see

Fabio Labella

@SystemFw

i.e. a Stream doesn't have the ability to terminate with a value

Calvin Lee Fernandes

@calvinlfer

so R is used to transfer some state between Pulls

but its not used in Stream

Fabio Labella

@SystemFw

Pull and Stream are two "moral" newtype over Algebra

the Stream monad instance (which is on O)

is better for control flow

the Pull monad instance (which is on R) is better for stateful combinators (so for the actual "streaming" aspect)

you can obviously write a combinator in Pull

by recursing and "transferring some state", as you say

and once you're done you convert to Stream

that's how Stream's stateful combinators are done

Calvin Lee Fernandes

@calvinlfer

ahhh

Fabio Labella

@SystemFw

look at uncons on Pull

it returns Pull[F, Nothing, Option(Chunk[O], Stream[F, O])]

Calvin Lee Fernandes

@calvinlfer

thanks! i was using uncons1 and unconsChunk so i was confused what that R type was about

so that emits nothing?

Fabio Labella

@SystemFw

so uncons doesn't emit anything, but returns the first chunk of elements, and the rest of the Stream

similar to list uncons

then you flatMap to get that

Calvin Lee Fernandes

@calvinlfer

ahhh

Fabio Labella

@SystemFw

you can Pull.output on the chunk to emit it

and you have the stream to decide what to do next (recursively)

Calvin Lee Fernandes

@calvinlfer

so is it right to think of R to transfer state between Pulls?

Fabio Labella

@SystemFw

pull-based Streams are iterators on steroid

yeah, in most cases that's how it's used

Calvin Lee Fernandes

@calvinlfer

Streams are crazy!

Fabio Labella

@SystemFw

really focus on the types on Pull, and try writing some stuff with them

note that in Haskell most streaming libraries have 3 type params

one for output, one for result

Where communities thrive

People

Repo info

Activity