stream

stream copied to clipboard

Elm package: http://package.elm-lang.org/packages/naddeoa/stream/latest

A Stream is kind of like a stream in Java 8 and kind of like a lazy list. It is a potentially infinite stream of items that can be transformed and collected back into other structures.

This library was made because Elm currently doesn't have great options for lazyiness. There are a few implementations that exist but they are vulnerable to stack overflows. This library uses tail call recursion for collecting streams into lists. Also, instead of modeling the streams recursively it uses an approach closer to the Elm architecture. The Lazy core library is not used. The deepest stack you will have is equal to the amount of transformations you apply on a stream. For example, in the following snippet:

let
    finalStream =
        Stream.fibonocci
            |> Stream.limit 100
            |> Stream.filter isEven
            |> Stream.map toString

Every time next is called on finalStream a stack of size three is generated before a value can be returned. It scales with the number of transforms, not the size of the list. You would have to apply a high enough number of transforms to a stream in order to get a stack overflow. That number is high enough that you won't have to worry about it. That said, in other libraries when you attempt to turn an infinite stream into a list you get a stack overflow. In this library, it will just run forever, so, keep that in mind.

One useful difference from Java 8 streams that we get just for being functional is stream reuse. Things like the following are possible now.

let
    baseStream =
        Stream.naturalNumbers
            |> Stream.limit 10

    firstList =
        Stream.toList baseStream

    list =
        baseStream
            |> Stream.map toString
            |> Stream.toList

This is obvious to anyone who feels comfortable with functional programming, but if you're coming from Java 8 then you're aware that you can't reuse streams that have been collected.

Changelog

2.3.0

• flatten - Convert a Stream (Stream a) -> Stream a

2.2.0

Added a bunch of async stuff, inspired from my use of rxjs. See the Demo.elm file for some examples. You'll have to run elm install first to get elm-lang/html added as a dependency so see it in elm reactor. I don't want to actually depend on it and elm doesn't currently allow for dev/test dependencies.

• StreamResult - container type for async operations
• deferNext - like the next function, but happens after some time
• deferNextN - like the nextN function, but happens after some time
• every - Subscribe to stream values until the stream is empty