Changes to apply/call this coercion (was: ES3.1 Draft: 27 Oct 2008 version available)

Mark S. Miller erights at google.com
Mon Oct 27 17:33:38 PDT 2008


On Mon, Oct 27, 2008 at 12:04 PM, Robert Sayre <sayrer at gmail.com> wrote:
> Pratap Lakshman (VJ#SDK) wrote:
>>
>> I have uploaded to the wiki (link <http://wiki.ecmascript.org/doku.php?id=es3.1:es3.1_proposal_working_draft>) the 27 Oct 2008 draft of the specification for ES3.1. This is in the form of in-place edits and markups to the ES3 specification. Revision history is at the end of the document.
>
>
> (I see multiple problems with these edits--I will address them separately)
>
> These edits to call and apply look like unacceptable, incompatible
> changes to ECMAScript-262 3rd edition.
>
> The following test passes in JS shells from Mozilla, WebKit, and Chrome:


You are completely correct, and all the folks editing the ES3.1 spec
are in complete agreement. The bug you point out in the spec is my
fault. It is an artifact of a partially completed refactoring of the
spec, exactly along the lines we've all already discussed on the
es*-discuss lists and verbally at the various meetings. The *intent*
is the following:


* All the coercions of the this-value that were scattered all over the
spec, including in the apply and call language, be removed.
This-arguments become this-values with no coercion.

* When a function is called as a function, the implicit this-value
provided is |undefined|.

* A non-strict function's [[Call]] method (rather than it's |call| or
|apply| methods) coerce the this-value on entry to preserve ES3
semantics. Since the time of coercion is unobservable (the coercion
has no observable side effects), the intended result is no observable
difference for non-strict functions. The non-strict this-coercions
are:

** null -> global object
** undefined -> global object
** primitives -> wrappers (via ToObject)

* Strict functions don't coerce their this-value at all. If your |f|
and |g| functions were strict, then your test case must indeed fail.
But if they're non-strict, your test case must pass


And again, sorry for the confusion!


-- 
    Cheers,
    --MarkM



More information about the es5-discuss mailing list