Re: "dict"ionary typing?

On 5/12/2013, at 11:01 PM, Michael Hendricks wrote:
> I dislike declarations when they can be inferred from usage.

They can only be inferred from usage if you are willing either
(1) to follow an ML-style **strict** definition before use policy,
    which has never been the Prolog Way, or
(2) take two passes to process the source file,
    which is also contrary to Prolog practice.
>  I
> particularly dislike function declarations in Prolog because they imply
> that a predicate can only be used as a function in a single "mode".

There is no such implication.  Change it to ':- active', then.
They're the perfect place to put type information, even if it is
currently just treated as commentary.

>  One
> might reasonably use append/3 as a function in three different modes.  I
> haven't implemented it yet, but I'm considering a new feature for
> library(func) inspired by Ciao's notation.  Specifically, a single ~ marks
> the argument to be used as output for *this* call.
> 
>  ?- writeln(append([a,b],[c,d],~)).

It's just too bad that ~ is a perfectly good data term
as it stands and append([a,b],[c,d],~) is also a perfectly
good data term as it stands, so this query is perfectly
meaningful right now, meaning something else.  This is a
recipe for confusion.

I've seen something similar done in Eiffel "agent expressions",
and it's nearly as unattractive there.

You can get something of the same effect using definitions like

	ret1(G) = X :- arg(1, G, X), call(G).
	ret2(G) = X :- arg(2, G, X), call(G).
	ret3(G) = X :- arg(3, G, X), call(G).
...
	ret3(append([a,b], [c,d], _))


Oh yes, there is a >> logical << approach to functions in
Prolog called "narrowing".  There's been quite a lot written
about it.