[m-users.] Cannot execute a predicate from a Discriminated Union

[Charles Shuller]

On Sat, Jun 27, 2015 at 11:27 PM, Mark Brown <mark at mercurylang.org> wrote:

> Hi Charles,
>
> On Sun, Jun 28, 2015 at 7:47 AM, Charles Shuller
> <charles.shuller at gmail.com> wrote:
> > Hello All,
> >
> > I can't get a predicate to execute from a discriminated union after
> passing
> > the type as a parameter into a predicate.  Though it works just fine from
> > main.
> >
> > I've tried everything I can think of with modes and insts on the DU, but
> > nothing has worked, any insight would be much appreciated.
>
> To make any call Mercury needs to know the type, mode and determinism
> of what is being called. For higher order terms, this information
> comes from the inst of the term (see reference manual, 8.3). Functions
> have an obvious default they can use, but for predicates this
> information must be given somewhere.
>
> In the signature of execute_test/4, the Test argument is declared
> 'in', meaning that its inst is ground. Mercury can't determine what
> mode and determinism you intend the call to Test^test_pred to be made
> in (and neither can I, for that matter). The problem is not the
> discriminated union, but the fact that it is declared to be ground.
>
> >
> > :- type test
> >     ---> test(test_name::string,
> >           test_pred::pred(disposition)).
>
> The inst you need to declare would look something like:
>
> :- inst test == bound(ground, pred(m) is d).
>
> where m and d represent the mode and determinism.
>
>
No joy.  bound/2 is not defined.


:- inst test
    ---> test(ground, pred(out) is semidet).

Is as close as I'd managed to get on my own, but even with that declaration
it still doesn't seem able to figure out inst of test_pred and I get the
same error message.


Also, I got the idea that the :-inst name == bound(...;...;...)  form of
inst was creating a name for an inst-tree which could later be applied as a
mode, was that completely wrong??





> >
> > :- pred execute_test(test::in, disposition::out, io::di, io::uo) is det.
>
> Then you can say:
>
> :- pred execute_test(test::in(test), disposition::out, io::di, io::uo) is
> det.
>
> > execute_test(Test, Disposition, !IO) :-
> >     call(Test^test_pred, Disposition), %% Compilation fails here
> >     Disposition = fail(Test^test_name ++ " Failed to execute").
> >
> > main(!IO) :-
> >     Test = test("PassTest", pass_test),
> >     call(Test^test_pred, Disposition). %% This works just fine though
>
> This would work because Mercury infers the correct inst of Test from
> the mode and determinism of pass_test/1.
>
> Hope this helps,
>


A bit ;)

I now know for sure it's a problem getting the inst declared on test_pred,
and I just need to find the correct syntax.

I also know it's not a compiler bug, and that there was just more context
for inference in the call in main(...).



Thanks!


Charles



>
> Mark.
>
> >
> >
> >
> > Compiler Output (14.01.1)
> >
> ----------------------------------------------------------------------------
> >
> > manual_test.m:044: In clause for `execute_test(in, out, di, uo)':
> > manual_test.m:044:   in argument 1 (i.e. the predicate term) of
> higher-order
> > manual_test.m:044:   predicate call:
> > manual_test.m:044:   mode error: variable `V_10' has instantiatedness
> > `ground',
> > manual_test.m:044:   expecting higher-order pred inst (of arity 1).
> >
> >
> > Thanks!
> >
> > Charles
> >
> > _______________________________________________
> > users mailing list
> > users at lists.mercurylang.org
> > https://www.mercurylang.org/lists/listinfo/users
> >
> _______________________________________________
> users mailing list
> users at lists.mercurylang.org
> https://www.mercurylang.org/lists/listinfo/users
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mercurylang.org/archives/users/attachments/20150628/2143e710/attachment.html>