News for hackers. Stuff that matters.
Re:Syntax across languages by mdcox on Sunday February 01, 2015 @04:53PM (#8981813)
A cursory skim leaves the impression of just how alone Haskell is on many syntactic choices (though supposedly with alot in common with "Merd".) I never took the time to learn Haskell, but it always looks beautiful on an aesthetic level, which differentiates it from the other loners in the list.

Can anyone point me to any papers/documentation/discussion on WHY certain syntax decisions were made? Many could have been arbitrary choices to match the constructs, but the way the community presents itself makes me think there's more to it.

Re:Syntax across languages by Retra on Sunday February 01, 2015 @07:34PM (#8982253)
The best part of Haskell is that function application is just whitespace.

    f x = f(x)
I would attribute almost 90% of it's elegance to that idea alone. (This is not an exact number or measured estimate.)

I also have a major thing for pattern-matching, and it always seems nice when you know that's what you're looking at. For instance, it is far easier to understand what `int * * var` is in C if you see it as a kind of pattern de-structuring.

Re:Syntax across languages by tormeh on Sunday February 01, 2015 @08:07PM (#8982321)
>The best part of Haskell is that function application is just whitespace.

It's the worst part, in my opinion. Is A B C D equal to A(B(C(D))), A(B,C,D), A(B,C(D)), A(B(C),D) or A(B(C,D))? Often you need to disambiguate with parentheses, so I guess there is some default interpretation that you can assume when reading code, but in the end it just looks like an unnecessary mental translation step.

Lisp has weird syntax, but lisp has a reason to have that syntax. Haskell has a weird syntax because it wants to either be different for different's sake or to just be plain inaccessible.

Re:Syntax across languages by epidemian on Sunday February 01, 2015 @09:55PM (#8982538)
Consider that, in Haskell:

1. all functions take a single argument (i.e., are curried)

2. function application is left-associative

So "a b c d" is just applying the function "a b c" with the argument "d", the same as "(a b c) d". "a b c" is the application of "c" to the function "a b", which, in turn, is the application of "b" to "a". So "a b c d" is the same as "((a b) c) d".

I personally find using the space symbol for a core concept of the language (function application) very elegant, and not dissimilar to other languages that strive to have simple and consistent syntax, like Smalltalk, where "obj meth1 meth2 meth3" is the same as "((obj meth1) meth2) meth3", even though Haskell and Smalltalk might be as different as programming languages can get :)

Re:Syntax across languages by tormeh on Monday February 02, 2015 @04:40AM (#8983425)
How does currying help? You still have to figure out which argument belongs to which function. GHC demands parentheses if you try to write A(B(C(D))) as A B C D, so I'm pretty sure I'm right.
Re:Syntax across languages by tomp on Monday February 02, 2015 @04:54AM (#8983459)
No, you're wrong. `A(B(C(D)))` is different from `(((A B) C) D)` (which is the parenthesized version of `A B C D`).

Currying allows simple partial application; if A is a function that "takes 3 arguments" (i.e. it has type `w -> x -> y -> z`, which means that it takes an argument of type `w` and returns a function with type `x -> y -> z`, which takes an argument ...), you can say e.g.

  A_ = A B C
  E1 = A_ D1
  E2 = A_ D2
where `A_` is the partially applied function `A`.
Re:Syntax across languages by michaelcampbell on Monday February 02, 2015 @07:36AM (#8983857)
> No, you're wrong.

And you and he both gave evidence to the initial assertion that "it's confusing".

Re:Syntax across languages by gamegoblin on Monday February 02, 2015 @01:39PM (#8985852)
I'm not sure if "confusing" is fair -- just different than C-family languages.

The rule is simply "The first expression is a function, all following expressions are arguments to that function"

It's no different than "The first symbol is a function, all comma-separated, parentheses-bound expressions are arguments to that function"

Both are pretty easy syntactic rules, I think.

Re:Syntax across languages by JoeAltmaier on Monday February 02, 2015 @01:41PM (#8985865)
Its arguably confusing, due to the lack of parentheses factoring a list of functions into one or more invocations. How is the nesting done here? Something more subtle. A lifetime of reading parenthesized functions is going to be confounded by changing the convention - which equals 'confusing'
Re:Syntax across languages by gamegoblin on Monday February 02, 2015 @01:52PM (#8985949)
I think you're probably overestimating the difference. Here is an algorithm for mechanically transforming a C function call to a Haskell function call:

    for all arguments in the function
        if the argument is not a symbol surround it with parentheses
    delete all commas
    delete the two initial parentheses 
And an algorithm for converting Haskell calls to C calls:

    put an open parenthesis between the first expression and second expression
    put a close parenthesis at the end of the expression
    put commas between all expressions between these two parentheses
The fact that it's so easy to convert between them highlights to me that they are both pretty easy.
Re:Syntax across languages by tormeh on Monday February 02, 2015 @05:18AM (#8983518)
I understand currying and partial function application. That's not the problem. The problem is that function application whitespace demands an arbitrary application precedence and obscures function signature to the reader.
Re:Syntax across languages by tomp on Monday February 02, 2015 @07:52AM (#8983899)
ah, that. Well, you get used to it. Also, it helps that application has the highest precedence, so it's quite easy to read expressions like `f x y + g z` (i.e. `(f x y) + (g z)`).
Re:Syntax across languages by gamegoblin on Monday February 02, 2015 @01:34PM (#8985807)
You don't have to figure out which argument belongs to which function.

The syntax is "The first expression is a function, all following expressions are arguments to that function". Fullstop.

It would be unfair of me to say "C has confusing and ambiguous calling conventions, the compiler complains if I write A(B(C(D))) as A(B,C,D)"

Because you would respond "Well of course, that's just the calling conventions of the language!"

To which I say "Exactly"

Re:Syntax across languages by gamegoblin on Sunday February 01, 2015 @08:33PM (#8982379)
A B C D is necessarily equal to A(B,C,D) in C-family syntax due to the precedence rules, there is no ambiguity.

The reason whitespace is more elegant is because your functions can return other functions, which due to function currying and partial application is equivalent to your function simply taking more arguments.

I'll use Python as an example.

    def flip(f): return lambda a,b : f(b,a)
That is, it's a function that takes a function of two arguments, and returns a function that does the same thing, but with the argument order flipped.

To use these function, it would look like:

    flip(f)(a,b)
In Haskell, a direct translation with the lambda would be:

    flip f = \a b -> f b a
The type of which could be written as:

    flip :: (a -> b -> c) -> (b -> a -> c)
That is, a function which takes one argument, that argument is a function that takes two arguments of types a and b and yields a value of type c, and returns a function that takes two arguments of types b and a and yields a value of type c.

We could use this function like:

    (flip f) a b
But due to function precedence we could use

    flip f a b
This segues into an equivalent (and arguably better) way to write the function taking advantage of partial application and currying:

    flip :: (a -> b -> c) -> b -> a -> c
    flip f a b = f b a
Which would be used the exact same way:

    flip f a b
This illustrates the elegance of currying.

    a -> b -> c
Is equivalent to

    a -> (b -> c)
Re:Syntax across languages by TazeTSchnitzel on Sunday February 01, 2015 @08:42PM (#8982397)
All Haskell functions are curried, so there's no ambiguity.
Re:Syntax across languages by Retra on Monday February 02, 2015 @09:44AM (#8984326)
>Lisp has weird syntax, but lisp has a reason to have that syntax.

Lisp has that syntax because it provides a uniformity to parsing the language, and that's the same thing Haskell offers .

If I write

    a = b + c
You might as well parse that as `a(=(b(+(c))))` with the appropriate semantics.

In that sense, Haskell is just trying to do something like what Lisp does without all the extra parentheses.

Re:Syntax across languages by tormeh on Monday February 02, 2015 @05:03PM (#8987324)
That's... actually really neat. So there's no operator precedence?
Re:Syntax across languages by Retra on Tuesday February 03, 2015 @04:24PM (#8993489)
Yes, there is operator precedence, but it only matters for infix notation. Or rather, function application has highest precedence, so there is no ambiguity using it.

My broader point was that the space is the most common symbol in English because the function that it performs is the most fundamental action of the language. And since Haskell is a functional language, function application is fundamental, and thus you gain a lot of terseness by using spaces to denote it.

Lisp does the same thing: `(+ a b)`, it just puts parentheses around things to make the parsing rules simpler. If you parse it in a different way, you won't need that many parentheses.

Re:Syntax across languages by tome on Sunday February 01, 2015 @05:26PM (#8981933)
Haskell inherits a lot from Standard ML. This paper explains a lot of the design decisions of Haskell, syntax amongst them:

http://haskell.cs.yale.edu/wp-content/uploads/2011/02/histor...