LAYOUT?-CF not accepted in Sdf3 Statix about spoofax-pie

[Gohla]

If this production is in a kernel syntax section (syntax), that error will not occur. However, then you will get an error that the sort of the production is not of kind kernel. Sorts of kind kernel would be definable in a sorts section, but it seems that the sorts section is desugared into context-free sorts in SDF3, making it impossible to define kernel sorts. Maybe a quick fix would be to have -CF and -LEX symbols be of kind kernel.

[Gohla]

Solved by metaborg/sdf@3579ddd. It is possible to define the production without errors as follows:

sorts

  Section SectionName Definition

syntax

  Section-CF.Section = "section" SectionName-CF LAYOUT?-CF Definition*-CF

There is no checking whether the -CF sorts are context-free sorts though.

[Virtlink]

Okay, previously doing this would give an error because Section-CF would be a context-free kernel sort declared in a (non-context-free) kernel syntax section, requiring the section to be changed to context-free syntax.

[Gohla]

There seems to be a bug where the SDF3->Stratego signature generator fails on LAYOUT?-CF in those kernel syntax productions, I will have a look at that tomorrow.

[Virtlink]

In Spoofax 3 it generates the following Stratego for it:

OpDecl("Section", FunType(
    [ ConstType(SortNoArgs("SectionName"))
    , OPT(LAYOUT(){OfSort(SORT("TYPE"))}){OfSort(SORT("TYPE"))}
    , ConstType(Sort("List", [ConstType(SortNoArgs("Definition"))]))
    ], ConstType(SortNoArgs("Section"))
))

In Spoofax 2 it seems to be:

OpDecl("Section", FunType(
    [ ConstType(SortNoArgs("SectionName"))
    , ConstType(Sort("List", [SortNoArgs("Definition")]))
    ], ConstType(SortNoArgs("Section"))
))

Spoofax 2 is correct. Apparently they are not using the same version of the signature generator?

[Gohla]

Should be fixed metaborg/sdf@d88c75a (untested)