[mercury-users] Pred defns

[Ralph Becket]

Peter Phelan wrote on 25 Mar:
> Hello again
> 
> I'm confused by pred definitions in Mercury.
> 
> The following is snipped from calculator.m in samples.
> 
> :- pred fullexpr(expr::out, list(char)::in, list(char)::out)
> is semidet.
> fullexpr(X) -->
>  expr(X),
>  ['\n'].
> 
> How come the declarative definition takes just one argument,
> whilst there are 3 in the type declaration?
> 
> It's something to do with input/output, but this is also
> difficult to understand.

With the best will in the world, RTFM.  Have a look at

http://www.cs.mu.oz.au/research/mercury/doc/reference_manual_2.html#SEC11

Often one writes predicates that take some `state' as input and
produce a transformed `state' as output.  When you're doing a lot of
IO, for example, this can be a real pain:

say_hi(Mum, IOStateIn, IOStateOut) :-
	io__write_string("Hello ", IOStateIn,   TmpIOState0),
	io__write_string(Mum,      TmpIOState0, TmpIOState1),
	io__write_string("!",      TmpIOState1, TmpIOState2),
	io__nl(                    TmpIOState2, IOStateOut)

(of course, you wouldn't write it like this, but you get the idea).
DCGs are a formalism for adding the pair of `state' thread arguments
to the end of the head and the goals within a clause.  That's what the
--> is for, rather than :-

say_hi(Mum, IOStateIn, IOStateOut) -->
	io__write_string("Hello "),
	io__write_string(Mum),
	io__write_string("!"),
	io__nl.

However, you still have to declare the implicit extra pair of
arguments in pred and mode declarations so the compiler knows the
right type and mode for them.

Ralph

-- 
Ralph Becket  |  rwab1 at cam.sri.com  |  http://www.cam.sri.com/people/becket.html