Skip to content

tc39/proposal-class-brand-check

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

72 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

class.hasInstance()

This proposal adds a class.hasInstance() brand checking syntactic form for ECMAScript code classes.

class C {
  static isC(o) {
    return class.hasInstance(o)
  }
}
let c = new C()
let o = Object.create(c)
let p = new Proxy(c, {})
C.isC(c) // true
C.isC(o) // false
C.isC(p) // false

Stage: 1

Authors: HE Shi-Jun (@hax), XU Tian-Yang (@XGHeaven), Tu Qiang(@YuriTu)

Champions: HE Shi-Jun, XU Tian-Yang, Tu Qiang

Experimental Implementations

Motivation

There is long discussion history of brand check, and various efforts on this area. The recent related proposal is "brand check for private fields". In the discussion of that proposal, the author of tc39/proposal-private-fields-in-in#13 propose an alternative design direction. Instead of duck-type style checking (#field in obj), we could provide a much general class brand check (class.hasInstance(obj)), a "real" instanceof check. In many cases it is much closer to programmer's intension and give a much similar concept model to other class-based OOP programming languages.

class C {
  static isC(o) {
    return o instanceof C
  }
}

instanceof is prototype-based by default, so C.isC(Object.create(C.prototype)) would pass the test. Note, there are many programmers still use instanceof, especiallly those from other OOP languages. Even they know the limitation of instanceof and there is Symbol.hasInstance which could be used to "customize" instanceof, people just ignore them because there is no easy way to "fix" it.

class C {
  #x
  static isC(o) {
    return #x in o
  }
}

As private-in proposal, people can utilize private field and private in syntax to do brand check, but it's just abusing of private fields. People need to add a private field even they don't really want it. And unfortunately, private field have a significant semantic differences to normal data properaty, so introducing of private field would introduce a breaking change, and potentially decrease the interoperability of the code with many libraries/frameworks.

Even you already use private fields for good reasons, this also cause the problem:

class C {
  #x
  #y // add a new field
  static isC(o) {
    return #x in o // do we need to update it to
    // return #x in o && #y in o
  }
}

Should the programmer update the implementation of isC? If you consider #x in o as a similar thing as duck type check, you should add && #y in o, but in almost all cases, this code is reduantant. But, in some edge cases (the initialzier of #y throw, as current private fields draft, u get an object with #x but no #y) u need to consider again. This cause confusion and unessarry burden to coders and code reviewers.

The essential problem is, in most cases, what people want is a general "instanceof" test, testing fields existence one by one can't really convery the intention of the programmers.

A better solution:

class C {
  static isC(o) {
    return class.hasInstance(o) // class.hasInstance is a meta method to check whether o have the class brand of C
  }
}

Actually it also allow programmers use Symbol.hasInstance to "fix" instanceof if they will:

class C {
  static [Symbol.hasInstance](o) {
    return class.hasInstance(o) // class.hasInstance is a meta method to check whether o have the class brand of C
  }
}

Presentations

Analysis of edge cases of interactions of using class brand check with private fields

TBD

Syntax and naming issue

class.hasInstance meta syntax conflict with class property access expression proposal, there is already an issue for it. If we decide to leave class.foo for class properties, we need to find some other syntax, for example class::hasInstance.

About naming, I choose .hasInstance which follow Symbol.hasInstance, though there could always be other choices: .hasBrand or .checkBrand.

About

JavaScript proposal to provide class brand checks

Resources

License

Code of conduct

Security policy

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages