[mercury-users] inst of list in non-empty branch

[Ian MacLarty]

On Fri, Mar 30, 2007 at 12:02:08PM +1000, Julien Fischer wrote:
> 
> On Thu, 29 Mar 2007 doug.auclair at logicaltypes.com wrote:
> 
> >This leads me to suspect that the mode for Kids is set to only 'in', even
> >in this non-empty-list clause.  ... I have seen the mode system set the
> >inst properly for other user-defined discriminated union types, 
> >particularly
> >when implementing different modes for different clauses (as illustrated in
> >the refman, sect 4.4) so it's puzzling why this particular example is being
> >uncooperative.
> >
> >Is there a pragma or assertion where I can assign the inst to Kids? Is
> >there some call-out-to-C hackery that would force the inst assignment?
> 
> We usually resort to something like:
> 
>     :- pred unsafe_cast_to_foo(type::in, type::out(foo)) is det.
> 
>     :- pragma foreign_proc("C",
>         unsafe_cast_to_foo(A::in, B::out(foo)),
>         [will_not_call_mercury, promise_pure, thread_safe],
>     "
>         B = A;
>     ").
> 

The following would be safer:

:- pred det_cast_to_foo(t::in, t::out(foo)) is det.

det_cast_to_foo(T, F) :-
    ( T = foo ->
        F = foo
    ;
        error("T not a foo")
    ).

or 

:- pred det_cast_to_foo(t).
:- mode det_cast_to_foo(ground >> foo).

det_cast_to_foo(T) :-
    ( T = foo
    ; T = bar,
        error("bar")
    ).

although in your case the solution I posted earlier should work I think.

Ian.
--------------------------------------------------------------------------
mercury-users mailing list
Post messages to:       mercury-users at csse.unimelb.edu.au
Administrative Queries: owner-mercury-users at csse.unimelb.edu.au
Subscriptions:          mercury-users-request at csse.unimelb.edu.au
--------------------------------------------------------------------------