[mercury-users] functions vs predicates (was Re: State variable problem

[Peter Moulder]

On Wed, May 31, 2006 at 11:29:44AM +1000, Ralph Becket wrote:

> I used to think the same way [that function calls are just an
> alternative syntax for predicate calls], but I've been convinced
> otherwise.  The main thing is that with functions it is natural to
> assume referential transparency.

That may well be a natural assumption, but it's unfortunately a false
one in Mercury: that is, the code

   f \= f

can succeed (where f is a function application in the usual forward
direction) even though

   Z = f,
   Z \= Z

always fails.  (Even if the NaN problem were fixed.)

This problem arises with function applications that fail: f \= f is
converted to not(f = f), which becomes not(f = Tmp1, f = Tmp2, Tmp1 =
Tmp2), where the `f = Tmp' calls are treated like predicate calls:
failure of the call means success of the `not' goal, i.e. success of
f \= f.


This is one reason that it's discouraged for the forward mode of a
function (i.e. f(in, in, ..., in) = out) to have determinism semidet (or
failure), and why the standard library would usually use a predicate
instead for such cases.

pjrm.
--------------------------------------------------------------------------
mercury-users mailing list
post:  mercury-users at csse.unimelb.edu.au
administrative address: owner-mercury-users at csse.unimelb.edu.au
unsubscribe: Address: mercury-users-request at csse.unimelb.edu.au Message: unsubscribe
subscribe:   Address: mercury-users-request at csse.unimelb.edu.au Message: subscribe
--------------------------------------------------------------------------