I'm having an issue with overload resolution for n...

# javascript

[Blake Anderson]

I'm having an issue with overload resolution for nested sealed classes with the same simpleName, in this case Expression.Access.Variable and Pattern.Variable. Here's some debug output from my testing:

visit Variable(name=x)
  > is Expression.Access.Variable: false
  > is Pattern.Variable: true
Matched is Pattern.Variable
  > func = function (p0) { return $boundThis.visit_ta6mxa_k$(p0); }
visit pattern variable: Variable(name=x)

visit Variable(qualifier=null, name=x)
  > is Expression.Access.Variable: true
  > is Pattern.Variable: false
Matched is Expression.Access.Variable
  > func = function (p0) { return $boundThis.visit_ta6mxa_k$(p0); }
visit pattern variable: Variable(qualifier=null, name=x)

So, the is check of my when statement seems to be working fine - it's the actual function visit(ast: Expression.Access.Variable) versus visit(ast: Pattern.Variable) that's off (via val func: Type -> T = ::visit). I'm guessing it's the function resolution, but could be an implementation layer. Continuing to dig around here to see what I can find out.

EDIT: Self-contained example in thread.

[Blake Anderson]

Here's an example, looks like it's function resolution that's off - both visit(obj = First.Common()) and visit(obj = Second.Common()) print visit(Second.Common), obj = .... I'm guessing there's probably not a workaround for this without renaming either the function names or the AST itself.

visit(Second.Common), obj = Common(name=First.Common)
visit(First.FirstUnique), obj = FirstUnique(name=First.FirstUnique)
visit(Second.Common), obj = Common(name=Second.Common)
visit(Second.SecondUnique), obj = SecondUnique(name=Second.SecondUnique)

https://gist.github.com/WillBAnders/73b128281de268aac726760d93338fbd

[Blake Anderson]

According to a friend this is only an issue for the IR compiler; they couldn't replicate it with the non-IR one.

[Blake Anderson]

Update on this, we think this is coming from NameTables.calculateJsFunctionSignature, specifically this line that adds argument types to the name. I'm not sure if it.type.asString() is supposed to include the entire package name - I'm expecting it should, but that doesn't appear to be the case.

I can't get indexing to work locally so I'm not able to track this down too well; can't tell what the type of anything is or where methods are defined. Would like to open a PR for this but not too sure where to start.

[Blake Anderson]

Here's the generated output for this; both functions end up with the same name so the second overrides the first.

[andylamax]

I have no idea what your code does, but to avoid overload resolution for JS, just mark one of the methods with @JsName("anotherName")

[Blake Anderson]

Genius, that's exactly the type of workaround I was hoping for - at the very least I don't have to restructure my code for it. Thanks again!