constructors

[lucaswerkmeister]

We need to support constructors. See ceylon/ceylon-spec#796.

EDIT: Let’s make a task list.

• [x] ConstructorDefinition (I think we’ll inline the spec’s ConstructorHeader)
• [x] Update ClassBody (see ceylon/ceylon-spec#1194) – obsolete, since ConstructorDefinition extends Declaration
• [x] ConstructorExpression – nope, that’s a subset of MemberExpression.
• [x] ConstructorMeta – strictly speaking, not a subset of TypeMeta per-spec, but the RedHat compiler can’t parse memberQualifier.ConstructorName at all, so we’ll omit it too
• [x] ConstructorDec, and don’t forget to update the Dec.keyword documentation
• [x] Update ExtendedType (ceylon/ceylon-spec#1126) (I think we’ll inline the spec’s SuperclassOrConstructor) – edit: actually, update ClassInstantiation, which both ExtendedType and ClassSpecifier wrap
• [x] Update ClassDefinition (optional parameters)
• [x] ceylon.ast.create

[lucaswerkmeister]

Also includes ceylon/ceylon-spec#1126.

[lucaswerkmeister]

So what’s the supertype of ConstructorDefinition? Per spec, a ConstructorDefinition can only appear within a ClassBody, not within an InterfaceBody or Block. However, without complement types ({Statement|Declaration~ConstructorDefinition*} content), this would be a bit ugly to enforce within the type system… unless we say that ConstructorDefinition doesn’t extend Declaration.

But that’s probably too weird. Perhaps that’s a restriction we can either do via assertions, or just omit from ceylon.ast altogether.

[lucaswerkmeister]

Regarding ExtendedType: I think we’ll get rid of ClassInstantiation and inline it into ClassSpecifier (unchanged) and ExtendedType (with the added possibility of a TypeNameWithTypeArguments qualifier).

[lucaswerkmeister]

Actually… the compiler completely accepts:

class C {
    shared new C() {}
    shared new D() {}
}
class D() => C.D();

even though, per the spec, a ClassSpecifier can only be qualified by super:

ClassSpecifier: "=>" ("super" ".")? TypeNameWithArguments PositionalArguments

So perhaps that rule just wasn’t updated, and in fact ClassSpecifier and ExtendedType still have the same syntax, meaning we can keep ClassInstantiation?

@gavinking?

[gavinking]

@lucaswerkmeister fixed.

[lucaswerkmeister]

Thanks, then I’ll keep ClassInstantiation.

[lucaswerkmeister]

Alright, I think we’re done here! I hope I haven’t forgotten anything…