Re: Functional composition operator?

From: Pierre Weis (
Date: Tue Dec 08 1998 - 18:02:21 MET

From: Pierre Weis <>
Message-Id: <>
Subject: Re: Functional composition operator?
To: (John Whitley)
Date: Tue, 8 Dec 1998 18:02:21 +0100 (MET)
In-Reply-To: <13932.45585.522844.651609@hadar.cs.Buffalo.EDU> from "John Whitley" at Dec 8, 98 00:23:59 am

> Andrew Kay <> wrote, in the caml-list archives:
> > We are in the process of converting our Caml code into OCaml, and
> > have a problem choosing an infix syntax for function composition
> > [...] What do other OCaml people use for function composition? Is
> > there standard emerging?
> I found no answer in the archives, so I'd like to raise the same
> question again: is there a consensus for choice of infix composition
> operator? Failing that, is there some design principle that warranted
> its omission?
> Thanks,
> John Whitley

The normal infix operator should be a o, or more precisely a $\circ$
symbol. Unfortunately if we add o in the syntax of Caml, this will be
a bit strange to have this identifier as an infix operation (moreover
this implies difficult to explain syntax errors in programs).

In fact we discourage the usage of functional composition as a general
programming tool, since:

 -- it only save a few characters in programs
      let h = f o g
      let h x = f (g x);;)
 -- it breaks the polymorphism
    (if defined as
     let h = f o g
     h is not generalized, since its definition is a function
     application, whereas inline expansion of functional composition
     let h x = f (g x)
     being the definition of a function is properly generalized.)
 -- it is not so clear, especially in case of composition of curried
     let f x y z = x + y + z
     then the compositions
     f o (f 2 3)
     or (f 1) o (f 2 3))

It is still possible to define a composition operator to use in
trivial cases. So you may choose any multi-character infix operator
such as ++, if you really need functional composition.

Best regards,

Pierre Weis

INRIA, Projet Cristal,,

This archive was generated by hypermail 2b29 : Sun Jan 02 2000 - 11:58:17 MET