正则表达式/ lastIndex-意外行为

Question

I know there are a few regex/ lastIndex discrepancies but this one is new to me!

Expected behaviour : Creating a new regular expression (with literal/constructor) will, obviously, create a new RegExp object with a lastIndex property set to zero.

Actual behaviour : (in FF, Chrome): The lastIndex property seems to persist through multiple RegExp creations.

Eg

function foo(s) {

    // A *NEW* regular expression
    // is created on each call of foo():
    var regex = /ABC/g;

    document.write( regex.lastIndex + '<br/>' );

    // regex.test() updates lastIndex property
    regex.test(s);

    // This is where the regex's life should end...
    // (Why does it persist?)

}

foo('ABC');
foo('ABCABC');
foo('ABCABCABC');

See here: http://jsbin.com/otoze

---

A new RegExp object is being created on every function call (right?), so why is the following being written to the document?? -

0
3
6

???

Note, this weirdness appears to happen in FF(3) and Chrome(2), but, curiously not IE.

Is this expected behaviour, does IE get it wrong or right? Is this a well-known bug?

---

EDIT: this doesn't appear to happen when instantiating the regex with a constructor instead of a literal. Eg new RegExp('ABC','g'); ... Still, the literal should (theoretically) work, right?

Answer 1

var regex = new RegExp("ABC", "g"); doesn't have that problem, so I guess /ABC/g re-uses regexp objects.

EDIT: Apparently this is correct behavior according to the ECMAScript 3.0 specification, it's fixed in ECMAScript 3.1 - details

Answer 2

Try this:

function foo(s) {

    // A *NEW* regular expression
    // is created on each call of foo():
    var regex = new RegEx("ABC", "g");

    document.write( regex.lastIndex + '<br/>' );

    // regex.test() updates lastIndex property
    regex.test(s);

    // This is where the regex's life should end...
    // (Why does it persist?)

}

foo('ABC');
foo('ABCABC');
foo('ABCABCABC');