As I said before in the js-bra list, this method is everything but optimized. In fact, it performs poorly, mainly because of the regular expressions.

One possible optimization would be caching the RegExp object used on the second replace. But that probably wouldn’t do much, since you would still have two replaces in there.

Besides, your method is broken for arrays which have semicolons on its values, i.e.:

>>> removeItem(['foo;bar', 'baz'], 'foo')
["bar", "baz"]


Can’t beat the simplicity of a `for’ loop.

function arrRemove(arr, v)
{
for(var l = arr.length; --l >= 0; )
{
if(arr[l] === v)
{
arr.splice(l, 1);
return arr;
}
}
return arr;
}

This method performs only one function call, and returns as soon as it finds the value to be removed. So, in its best scenario, performs as O(1), and in its worse, O(n). Besides from being type-safe, while yours relies on the implicit `toString’ call.

Cheers.

Thanks!