[epidemian]

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 :)

[tormeh]

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.

[tomp]

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`.

[michaelcampbell]

> No, you're wrong.

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

[gamegoblin]

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.

[JoeAltmaier]

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'

[gamegoblin]

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.

[tormeh]

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.

[tomp]

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)`).

[gamegoblin]

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"