From: Xavier Leroy <xleroy@pomerol>
Message-Id: <9210281018.AA08184@pomerol.inria.fr>
Subject: Re: What about infix operator?
To: jcmoreno@dia.ucm.es (Juan C. Gonzalez Moreno)
Date: Wed, 28 Oct 92 11:18:50 MET
In-Reply-To: <36*jcmoreno@dia.ucm.es>; from "Juan C. Gonzalez Moreno" at Oct 27, 92 11:51 am
Caml Light allows you to define infix operators (that is,
functions), but not infix constructors. Constructors are not
functions. Concerning the :: infix constructor is built-in at the
level of the parser. What you propose is the best way to circumvent
this limitation:
type rational = Frac of int * int;;
let frac x y = Frac(x,y);;
#infix "frac";;
1 frac 2;;
However, you won't be able to use "frac" in left-hand sides of
pattern-matching.
As you have noticed, you are allowed to do:
let Frac x y = Frac(x,y);;
though this changes the status of "Frac" from constructor to function.
Actually, this is a bug in the current implementation. The phrase
above should be rejected, just as the semantically equivalent fomr
below is rejected:
let Frac = fun x y -> Frac(x,y);;
This might be fixed some day. So, don't rely on this behavior and name
differently the constructor "Frac" and the infix operator "frac".
> There is any possibility to define as
> prefix operators any combination of symbols, not only identifiers?.
Presently, no. Operators are restricted to be identifiers. Allowing more
operators (e.g. ++ or */ ) would be nice, but requires some changes in
the lexical conventions of Caml Light. For instance: 1+-2 without
blanks is currently parsed as 1 + (-2), but would probably have to be
parsed as prefix +- 1 2 with this extension.
Regards,
- Xavier Leroy