posted by exoteric

C

posted by Charles

posted by AceHack

C

posted by Charles

posted by felix9

posted by AceHack

posted by Petr

val aSet = Set(1, 2, 3)
val result = aSet.filter(x => somePredicate(x))

In Scala the type of 'result' will be a Set[Int], in F# & C# it will be an IEnumerable. Scala is one of the few languages that has a type system that is powerfull enaugh to allow this trickery.

Of course Scala can have behave similarly as F# or C# if you want it to (aSet.iterator.filter... )

A more powerfull example:

val aString = "string"
val stringResult = aString.map(c => c.toUpper) // result is a String: "STRING"
val listResult = aString.map(c => c.toInt) // result is a Vector[Int]: Vector(115, 116, 114, 105, 110, 103)

 http://www.scala-lang.org/docu/files/collections-api/collections_2.html

posted by llemieng

posted by Petr

let filtered = set |> filter (fun elm -> elm = 2)?

(you can in Haskell, but I don't think F# has type classes or type constructors)

In scala filter is implemented once, for all collections (of course some collections override & specialize for performance), it still supports the uniform return type principle and you don't need to tell it specifically what function it has to call.

In F# you have to reïmplement filter for all collections and then you still can't handle my last example with strings.

posted by llemieng

posted by Petr

let aString = "string"

let stringResult = aString |> Seq.map (fun c -> Char.ToUpper(c)) // returns seq<char>, not String, not expected

let listResult = aString |> Seq.map(fun c -> int c) // returns seq<int> as expected

And in both cases you specifically had to point out the function that F# uses and you lost type information in the first case.

I don't agree with your comments on erasure & type classes. The reified types wouldn't help Scala one bit to achieve their collection library. (I'm not saying that reified types aren't usefull, just that they don't help here)

And I see now that by functional datastructures, they mean efficient persistent datastructures. So, does F# have datastructures (Lists, Sets, Maps) that are immutable and have a very efficient (close to O(1)) update and delete operations?

posted by llemieng

posted by Petr

The F# examples are not as idiomatic as they could be.

Example 1:

let aString = "string"
let stringResult = aString |> Seq.map Char.ToUpper
let listResult = aString |> Seq.map int

Example 2:

let set = Set([1; 2; 3])
let filtered = set |> Set.filter ((=) 2)
let mapped = set |> Set.map (fun elm -> elm.ToString())

A little bit of the "Intellisense effortlessness" is lost of course, because you're not dotting your way through functions but instead tell the compiler that it should use the filter function from the Set module, etc.

The |> pipeline operator does ensure that you can use the same fluent style though. 

posted by exoteric

posted by Richard.Hein

posted by JarleSt

Good points about the skill of the library writer vs the skill of the application writer vs the application writer as a library writer.

I do wonder what Anders meant with machine learning and its future in programming.

posted by exoteric

posted by Moondevil

22 hours ago, exoteric wrote

It was indeed very humorous.

Good points about the skill of the library writer vs the skill of the application writer vs the application writer as a library writer.

I do wonder what Anders meant with machine learning and its future in programming.

@exoteric:  Check out this session, where he mentions what Anders said there, how most people don't know what to do with machine learning in terms of language design, yet:

http://channel9.msdn.com/Events/Lang-NEXT/Lang-NEXT-2012/Reverend-Bayes-meet-Countess-Lovelace-Probabilistic-Programming-for-Machine-Learning

It's funny, he mentions Anders and esoteric in the same paragraph ... oh your exo ... still.

posted by Richard.Hein

2 days ago, Richard.Hein wrote

*snip*

@exoteric:  Check out this session, where he mentions what Anders said there, how most people don't know what to do with machine learning in terms of language design, yet:

http://channel9.msdn.com/Events/Lang-NEXT/Lang-NEXT-2012/Reverend-Bayes-meet-Countess-Lovelace-Probabilistic-Programming-for-Machine-Learning

It's funny, he mentions Anders and esoteric in the same paragraph ... oh your exo ... still.

(I am actually also esoteric , I just don't use that account anymore for technical reasons.)

Thanks for the reference. I'll check it out!

posted by exoteric

posted by Mike

posted by vesuvius

Apr 10, 2012 at 5:26 PM

It's very strange that this panel didn't mentioned F# at all.

I agree. In F# you can write same kind of code inside and outside a monad (="computational expression") where e.g. in C# you have to move from old imperative ("normal") code to LINQ-syntax.

posted by Thorium