Proposal: "function-default-params" scope

[mysticatea]

From #33, eslint/eslint#9335.

As the step 27.a of 9.2.12 FunctionDeclarationInstantiation( func, argumentsList ), eslint-scope should make new scope for the function body only if the parameters has any default value.

It's similar to function-expression-name scope which separates the variable of function expression names from function bodies. So I propose new function-default-params scope which separates the references in default parameters from function bodies.

1. If the current function has one or more default parameters in traversing, the referencer defines a function-default-params scope instead of function scope.
   • Note: the function-default-params scope has implicit arguments variable except if the function is arrow function expression.
   • Note: the function-default-params scope is the child scope of function-expression-name scope if the function is function expression.
2. The referencer evaluates the parameters.
3. The referencer defines function scope.
4. The referencer copies the all variables from the function-default-params scope to the function scope.
   • Note: those copies handle redeclaration such as function f(a = 1) { let a } correctly.
   • Note: at least, no-shadow rule has to handle those copies as special.
   • Note: I think we should use shallow copy here because Variable#defs and Variable#references should be shared with both scopes but Reference#resolved property cannot have multiple variable objects. For example, in function f(a, b = a) { g(a) } case, the two references of a should be aggregated into one Variable#references array.
5. The referencer evaluates the function body.
6. The referencer closes both function-default-params and function scopes at the end of the function.

[not-an-aardvark]

Is this still something we want to do?

[shannon]

I'm having a similar issue as estools/escope#130 without this.

[mysticatea]

Oh, I had forgotten this issue. Yes, I hope to do. I added the memorandum of 7.

@shannon eslint-scope package has fixed in #33.

[kaicataldo]

@mysticatea Can we close this issue if it's fixed in #33?

[not-an-aardvark]

I don't think this issue has been fixed. If I'm understanding correctly, #33 fixed estools/escope#130, which is a separate issue.

[mysticatea]

@kaicataldo No. The #33 was a superficial fix because I avoided breaking changes. And this is a follow‐up issue to make the correct scope structure.

But probably I should write an RFC.