From: Jean-Hugues ROBERT <jean_hugues_robert@...>
Date: 2002-04-16T07:19:29+09:00
Subject: RCR: xx: yy as shorter notation for :xx => yy



Hello,

I am considering making a RCR but first I would like to make sure it has 
some value. Here it is:

I would like
   xxx: yyy
to be equivalent to
   :xxx => yyy.

Example A: Call with named parameters. Instead of
   obj.do_something :some_arg => value1, :some_other_arg => value 2
you would write:
   obj.do_something some_arg: value1, some_other_arg: value2

Example B: Hash literal. Instead of
   a_hash = {
     :if => "page 4",
     :return => "page 5"
   }
you would write
   a_hash = {
     if: "page 4"
     return: "page 5"
   }

This syntax form is similar (with a whole different meaning) to the C 
construct where, instead of
   (*xxx).field          (in Ruby: :xxx => yyy)
you can write
   xxx->field            (in Ruby?: xxx: yyy)

I don't know if it should be restricted to symbol (i.e. must xxx be a 
symbol in xxx: yyy) or can it be extended to any expressions (i.e. 
(an_expression): yyy). Of course, this later one makes ":" become an 
operator, instead of a syntax sugar. And this has more impacts I guess 
(i.e. more work) on the Ruby parser.

Rationnal:
   1) xxx: yyy is shorter than :xxx => yyy
   2) Promote the usage of symbols (efficient)
   3) Compares well to similar constructs in some other languages that have 
either "named arguments" or "named values" (or both)
   4) Natural, xxx: yyy is the syntax used in many internet protocols where 
the name of an attribute/field/element is followed by ":" and then the 
value of that attribute/field/element.

Difficulties:
   I don't know for sure but there might be some conflicts with the "?:" 
ternary operator. Example: (a ? b:c) and (a ? b:c : d:e). However in this 
case I don't think that => would be considered valid either: (a ? :b => c : 
:d => e) but I may be wrong.
   I don't know for sure what can follow ":" in :xxx today. Ideal any xxx 
that is valid in :xxx should be ok in the xxx: form too.

Implementation:
   The implementation requires some look ahead (at lexical level), not much 
probably. i.e. What may first appear has a reference to a local variable 
xxx or a function xxx turns out to be the xxx symbol (when ":" is 
encountered). When that happens, "=>" should be pushed back in the input 
stream (that later one is easy I guess).
   The implementation could focus on the simple cases first, when xxx is an 
identifier, and address more complex cases (when xxx is something more 
complex then an identifier, an arbitrary expression) later.

Jean-Hugues