Improved overload resolution

Motivation

Current overloads have two main issues:

• They depend on code position of toplevel definitions (confusing for newbies, manageable for some, unacceptable for others, error-prone for all)
• They depend on module loading order (much more problematic)

Proposed implementation

Syntax:

various syntaxes are possible....

Overload blocks

"overload" [ "(" module_list ")" ] "{" [ function_definition ]* "}"

example:

// vectors.clay

// just make the shadow block toplevel, for more regular syntax ?
shadow {
    Vector(s:'S) | Sequence?('S) {
        ...
    }
    Vector(s:'S) | SizedSequence?('S) {
        ...
    }
    Vector(rvalue s:Vector['T]) {
        ...
    }
}

// randomaccesssequences.clay
overload {
Vector(s:'S) | RandomAccessSequence?('S) {
    ...
}

// forwardsequences.clay

Vector(s:'S) | ForwardSequence?('S) {
    ...
}

Overload chain statements

"overload" [ "(" module_list ")" ] function_definition [ "shadow" function_definition ]*

Semantics:

• overloads are a list of function definitions with a list the modules whose definition they shadow
• each module has only one overload list per callable
• definitions within one overload will overload each-other based on definition order (current semantics)
• function definitions outside overload blocks cannot overload other functions
• overload lists with no specified modules implicitly overload all modules

overloads thus form a DAG where every node points to the node it overloads if two definitions are valid, but are unordered according the the DAG, then that ambiguity is a static error

Discussion

Please add comments here.