(this is the discussion page for typeof)

Previous proposal and discussion

We are adding some new fundamental types in edition 4. What should the result of typeof be when a value of one of these new types is its operand?

Type Result Comment
Undefined “undefined”
Null “null” es4 bug fix
Boolean “boolean”
Number “number”
String “string”
XML “xml” new to e4x, but with no special case for Namespace
XMLList “xml” ditto
Object (native and doesn’t implement [Call]) “object”
Object (native and implements [Call]) “function”
Class “type” or “class”, typeof String returns “function” in es3
Interface “type” or “interface”
Namespace “namespace” e4x chose “object”, does it deserve special status?
int “number”
uint “number”


From bug fixes where the typeof fixing started, more about the typeof null === “null” fix:

  • Spidering the web shows no dependencies on typeof null === “object”.
  • Spidering discloses a number of pages that will suffer a ReferenceError if null is passed.
    • I.e., these pages contain scripts that use typeof foo == “object” to guard foo dereferences.

Note that if we make RegExp have a [Call] internal method, typeof /hi/ === “function”. In IE, typeof alert === “object” even though alert is callable (but of course not a function object, so not applyable, etc.). Do we really want all callable objects to be of type “function”?

In general, typeof seems like a mess that will be hard to reform sensibly. If we provide a builtin::type() function, or perhaps reflect::type(), we could do something much better, and stop investing in typeof. Comments?

Brendan Eich 2006/03/31 15:13

I tend to agree that we should deprecate typeof. But can we ignore the fact that we are introducing new builtin types that are peers of function, number, etc? Maybe that argues for associating Class and Interface with a more generic name like “type” (i think “object” is too generic) and leaving Namespace with “object”. Also, I’d rather break compatibility with E3 typeof String and other constructors giving “function” and let them instead give “type” (or “class”) in E4. Is that going too far? Maybe it is.

Jeff Dyer 2006/03/31 16:08

We have reason to believe typeof null === “object” is a bug that could bite real content, from our spidering of the web. We have no such evidence in favor of breaking typeof String === “function”, and plenty of reason to fear incompatibility. It might be best to leave typeof utterly alone and deprecate it, but I’m still in favor of the null bug-fix.

In any event, reflect::type() or something like it wins. (BTW, I don’t mean to suggest that a type built-in function or unary operator should be part of the optional reflection package – I do think it will be hard to reserve type, so we might rather put it in a namespace that is open by default but prioritized somehow such that user-defined function type... or var type... hide it.)

Brendan Eich 2006/03/31 16:34

 
discussion/typeof.txt · Last modified: 2006/05/31 06:55 by graydon
 
Recent changes RSS feed Creative Commons License Donate Powered by PHP Valid XHTML 1.0 Valid CSS Driven by DokuWiki