[m-users.] Inst problem

[Volker Wysk]

Am Sonntag, dem 02.11.2025 um 12:19 +1100 schrieb Mark Brown:
> On Sat, Nov 1, 2025 at 9:45 PM Volker Wysk <post at volker-wysk.de> wrote:
> > 
> > Hi!
> > 
> > I have this:
> > 
> > ------------
> > :- type iosupp.res(T)
> >     ---> ok(T)
> >     ;    error(iosupp.error).
> > 
> > :- inst iosupp.res(I)
> >     ---> ok(I)
> >     ;    error(ground).
> > 
> > :- type path ---> path(
> >     abs :: bool,
> >     comps :: list(string)
> > ).
> > 
> > :- inst abs_path == bound(path(bound(yes), ground)).
> > 
> > :- pred absolute_path_p(path, iosupp.res(path),   io, io).
> > :- mode absolute_path_p(in,   out(res(abs_path)), di, uo) is det.
> > 
> > :- pred current_directory_p(
> >     iosupp.res(path)::out(iosupp.res(abs_path)),
> >     io::di, io::uo
> > ) is det.
> > 
> > absolute_path_p(Path, Res, !IO) :-
> >     (
> >         if   Path ^ abs = yes
> >         then Res = iosupp.ok(Path)
> >         else dirsupp.current_directory_p(Res1, !IO),
> >              (
> >                  Res1 = ok(Cwd),
> >                  Res  = ok(path(yes, Cwd ^ comps ++ Path ^ comps))
> >              ;
> >                  Res1 = iosupp.error(Err),
> >                  Res  = iosupp.error(Err)
> >              )
> >     ).
> > ------------
> > 
> > And I get this error message from the compiler:
> > 
> > iosupp.m:1065: In clause for `absolute_path_p(in,
> > out(iosupp.res(iosupp.abs_path)), di, uo)':
> > iosupp.m:1065:   mode error: argument 2 had the wrong instantiatedness.
> > iosupp.m:1065:   Final instantiatedness of `Res' was
> > iosupp.m:1065:     unique(
> > iosupp.m:1065:       error(ground)
> > iosupp.m:1065:     ;
> > iosupp.m:1065:       ok(
> > iosupp.m:1065:         bound(
> > iosupp.m:1065:           path(ground, ground)
> > iosupp.m:1065:         )
> > iosupp.m:1065:       )
> > iosupp.m:1065:     ),
> > iosupp.m:1065:   expected final instantiatedness was
> > iosupp.m:1065:     named inst res(abs_path)
> > iosupp.m:1065:     which expands to
> > iosupp.m:1065:       bound(
> > iosupp.m:1065:         error(ground)
> > iosupp.m:1065:       ;
> > iosupp.m:1065:         ok(
> > iosupp.m:1065:           named inst abs_path
> > iosupp.m:1065:           which expands to
> > iosupp.m:1065:             bound(
> > iosupp.m:1065:               path(
> > iosupp.m:1065:                 bound(
> > iosupp.m:1065:                   yes
> > iosupp.m:1065:                 ),
> > iosupp.m:1065:                 ground
> > iosupp.m:1065:               )
> > iosupp.m:1065:             )
> > iosupp.m:1065:         )
> > iosupp.m:1065:       ).
> > 
> > 
> > I think I understand the final instantiatedness of `Res', and that it
> > doesn't match the mode line for absolute_path_p. But how *should* it be
> > done?
> 
> I would do this:
> 
> :- type path
>     --->    absolute_path(list(string))
>     ;       relative_path(list(string)).
> 
> Or something else that better reflects what the type actually is :-)

This looks like a good idea. No more fightig with insts. But I already have
much code which uses my presend form of type "path"...

Thanks, Volker