Field named "constructor"

[tc39]

Field named "constructor"

[littledan]

I don't see why a field named constructor (whether it's static or instance) would need any special treatment. We talked about special treatment for fields named __proto__ in the past, but really, there's no particular interaction in either case. If there's a public instance field named constructor, then a constructor property is simply added to the instance with [[DefineOwnProperty]] at the appropriate time in the [[Construct]] sequence. You can still get at the "real" constructor through instance.__proto__.constructor. Do you think we should ban this, though, to avoid confusion?

There's definitely a bug about PropName being misapplied, which I see in one place: when prohibiting static fields named prototype. I'll fix that up; thanks for the catch.

[littledan]

What do you think of this fix?

[waldemarhorwat]

It's not fixed. The PropName call and disallowance of "constructor" I was referring to are coming from NonConstructorMethodDefinitions, which is called by ClassDefinitionEvaluation.

[rwaldron]

• https://tc39.github.io/ecma262/#sec-static-semantics-nonconstructormethoddefinitions
• https://tc39.github.io/ecma262/#sec-runtime-semantics-classdefinitionevaluation

[waldemarhorwat]

Huh?

https://tc39.github.io/proposal-class-fields/#runtime-semantics-class-definition-evaluation
calls NonConstructorMethodDefinitions, which calls PropName on something on which it's not defined. And it will silently drop any fields named constructor.

[rwaldron]

I was providing links to the sections you mentioned in your comment, for others that might read this thread.

[littledan]

@waldemarhorwat Thanks, I wasn't thinking about that path (but I believe the other one was a bug as well); will fix.

[waldemarhorwat]

NonConstructorElementDefinitions currently starts with:
If ClassElement is any of ClassElement:; , FieldDefinition:; or , return a new empty List.
I don't know what that is referring to; that doesn't parse and there are no such grammar productions.

[littledan]

@waldemarhorwat Thanks for pointing this out; I should've checked the generated output more closely (the colon doesn't appear in the source but was inserted by grammardown). Uploading a fix in a minute.