Historically we had no way to separate object value properties expression and entity attribute expressions.
I decided to separate them using the following notion:
object value property expression:
- computed: identifier[property]
- not computed: identifier.property
entity attribute expression:
- computed: identifier[.property]
- not computed: identifier..property
I'm going to attach a patch against silme AST and parser that shows the proposed changes.
Created attachment 535755 [details] [diff] [review]
requesting feedback from Axel.
I'm changing the way Idref works (removing multipart Idrefs in favor of nested Indexes)
I'm reusing the AST model from JS with their MemberExpressions with computed property.
Despite allowing for nested combinations (obj..attr.prop or obj[.obj2..attr2]) it also gives serializer ability to recognize the way ObjectIndex is being used depending on it's computed value.
fixed long ago in the l20n python tree