Re: Where Is Method Call Precedence?

"Marcin 'Qrczak' Kowalczyk" <[e-mail]> wrote in message

> f     Ada, Eiffel, Haskell, Mercury, Perl, Pliant, Ruby, Tcl, Pascal
> f()   Awk, C, C#, C++, Java, Lua, Python
> f()   merd, OCaml, SML (there really is a parameter which is the empty
tuple)

You can also add OCaml to the first line but in this case the function would
only ever be evaluated once - henceforth the return value would be used
(strict evaluation).

Be careful with OCaml and parentheses:

For the sake of example lets define the min and max functions first:
let min a b = a < b then a else b;; let max a b = a > b then a else b;;

OCaml does not have parentheses around function arguments. arguments follow
the function separated by space: "min 2 3" this yields "2" as expected.

C-style "min(x, max(y, z))" becomes "min x (max y z)" in OCaml. The
parentheses are used to correctly group the arguments. Really "(max y z)"
means the single value tuple made up of the return value from max - but a
single value is the same as the single value tuple so the parentheses works
like normal parentheses grouping.

"min(2,3)" yields something strange - a partially evaluated function taking
a two integer tuple as argument as if it was defined like
"let min_strange (x, y) = if (2, 3) < (x, y) then (2, 3) else (x, y);;"
(you can compare tuples, so the function makes sense).

You could also define min to work on a two tuple:
"let min' (a, b) = a < b then a else b" and you would get the expected
result when writing "min(2,3)". This would be the same as writing a Ruby
function expecting an array as argument to a min function.

It's correct that the empty tuple () can be applied to a function. But a
function can also be called without arguments at al (depending on its type).
A function without arguments evaluate to a constant and is only executed
once. There you sometimes write functions that take dummy arguments such as
the empty tuple to allow multiple evalutions (which is useful when printing,
say, linebreaks to an out stream).

The arguments are applied to OCaml differs from standard ML (SML) and it is
primary source of trouble to newcomers (i.e. me). (Someone said OCaml is big
on currying or something).

There is a reason for OCamls strange syntax: partially evaluated functions
and higher order functions (functions taking functions as arguments) becomes
very natural to work with.

If a function isn't provided with all arguments, the return value is a new
function that accepts the remaining arguments: "(min 2)" yields a new
function taking one integer and returns the smaller value of 2 and the
argument: "(min 2) 4" is the same as "min 2 4" but it happened was evaluated
in two steps (in bytecode this may matter, in assembly it's optimized away).

In conclusion: In OCaml you generally put the left parentheses before the
function name, rather than after it. This may look at bit like Lisp, but
isn't quite.

Tuples are very powerful datastructures in OCaml. They are not lists and
they are not arrays. They do, for example, make it much easier to handle
values in parser reductions. In Ruby you would use arrays or hashes, but
they have a much higher overhead. It would be nice with tuples in Ruby.

Mikkel