[m-rev.] for review: separate type_details_sub from type_details_du

[Peter Wang]

On Sat, 02 Oct 2021 05:46:22 +1000 "Zoltan Somogyi" <zoltan.somogyi at runbox.com> wrote:
> Separate subtypes from du types in parse trees.

> diff --git a/compiler/check_type_inst_mode_defns.m b/compiler/check_type_inst_mode_defns.m
> index 6c5751eef..8185be599 100644
> --- a/compiler/check_type_inst_mode_defns.m
> +++ b/compiler/check_type_inst_mode_defns.m
> @@ -319,7 +327,55 @@ check_type_ctor_defns(InsistOnDefn,
>          map.det_insert(TypeCtor, CheckedDefn, !TypeCtorCheckedMap)
>      else if
>          % TypeCtor is NOT defined as a solver or equivalence type.
> -        % Is there a du type definition ...
> +        % Is there a subtype definition ...
> +        ( if
> +            % ... in the interface?
> +            IntSubMaybeDefn = yes(IntSubDefn)
> +        then
> +            SubDefn = IntSubDefn,
> +            SubSection = ms_interface,
> +            SubWhere = "interface"
> +        else if
> +            % ... in the implementation?
> +            ImpSubMaybeDefn = yes(ImpSubDefn)
> +        then
> +            SubDefn = ImpSubDefn,
> +            SubSection = ms_implementation,
> +            SubWhere = "implementation"
> +        else
> +            fail
> +        )
> +    then
> +        report_any_redundant_abstract_type_in_imp(TypeCtor, SubWhere,
> +            ImpAbstractStdMaybeDefn, !Specs),
> +        list.foldl(
> +            report_any_incompatible_type_definition(TypeCtor,
> +                SubDefn ^ td_context, "subtype", SubWhere),
> +            [IntContextDu, IntContextAbstractSolver,
> +            ImpContextDu, ImpContextAbstractSolver],
> +            !Specs),
> +
> +        % A subtype's representation is controlled entirely by the
> +        % representation of its base type. It cannot have its own
> +        % representation specifications for its constructors.
> +        ForeignDefns = get_maybe_type_defns([IntMaybeDefnC, ImpMaybeDefnC,
> +            IntMaybeDefnJava, ImpMaybeDefnJava,
> +            IntMaybeDefnCsharp, ImpMaybeDefnCsharp]),
> +        list.foldl(subtype_report_any_foreign_type(TypeCtor, SubDefn),
> +            ForeignDefns, !Specs),
> +        list.foldl(subtype_report_any_foreign_enum(TypeCtor, SubDefn),
> +            ImpEnums, !Specs),
> +
> +        decide_subtype_status(TypeCtor, SubDefn, SubSection,
> +            IntAbstractStdMaybeDefn, Status, SrcDefnsInt, SrcDefnsImp),
> +        CheckedStdDefn = std_mer_type_subtype(Status, SubDefn),
> +        SrcDefns = src_defns_std(SrcDefnsInt, SrcDefnsImp, []),
> +
> +        CheckedDefn = checked_defn_std(CheckedStdDefn, SrcDefns),
> +        map.det_insert(TypeCtor, CheckedDefn, !TypeCtorCheckedMap)
> +    else if
> +        % TypeCtor is NOT defined as a solver type, equivalence type,
> +        % or subtype Is there a du type definition ...

Add ".".

> diff --git a/compiler/parse_type_defn.m b/compiler/parse_type_defn.m
> index 0f3fa91d2..d28ac438d 100644
> --- a/compiler/parse_type_defn.m
> +++ b/compiler/parse_type_defn.m
> @@ -211,33 +224,72 @@ parse_du_type_defn(ModuleName, VarSet, HeadTerm, BodyTerm, Context, SeqNum,
...
> +            (
> +                MaybeDirectArgIs = no
> +            ;
> +                MaybeDirectArgIs = yes(_),
> +                DirectArgTypeCtor =
> +                    type_ctor(TypeSymName, list.length(Params)),
> +                DirectArgPieces = [words("Error: the subtype"),
> +                    unqual_type_ctor(DirectArgTypeCtor),
> +                    words("is not allowed to have its own"),
> +                    quote("where direct_arg is"), words("annotation"),

Add ";".

> +                    words("it must inherit its representation"),
> +                    words("from its supertype."), nl],
> +                DirectArgSpec = simplest_spec($pred, severity_error,
> +                    phase_parse_tree_to_hlds, Context, DirectArgPieces),
> +                !:Specs = [DirectArgSpec | !.Specs]
> +            )
> +        ;

The rest looks fine.

Peter