Erlang improvement?

Robert Virding <>
Wed Sep 19 23:36:43 CEST 2001


Helmut Enck-Radana <> writes:
>At 19:25 19-09-01, Kostis wrote:
>>  > There is no ambiguity; to perhaps obfuscate things even more, the
>>  > above notation is just "syntactic sugar" for
>>[...]
>>
>>t2(M,F) ->
>>   apply(M,F,[bar,42]).
>
>I thought the question was whether f(X):g(Y) should mean f(X):(g(Y)) or 
>(f(X):g)(Y).

This is indeed what was probably meant.  However if you try to write
f(X):g(Y) you will find that is not allowed.  I did this on purpose.
The main reason being that while is is easy for the parser it can
often be difficult for a human to work out what is meant.  Otherwise
you could get beauties like:

	f(M):g(N):h(O):i(P)

This means that you need parenthesis around the module and function
parts of a remote call if they themselves are calls.  So you would
write:

	(f(X)):g(Y)

Also you can't really create a functional object X:Y as the ':' is
only valid in a call.  It is a bug that the parser allows while
erl_lint in the compiler disallows it.  It is my laziness which has
not fixed the parser to disallow it.

	Robert



More information about the erlang-questions mailing list