Array.prototype.remove(item)
Isiah Meadows
isiahmeadows at gmail.com
Fri Nov 10 13:13:16 UTC 2017
My proposed semantics does similar, except I just read until it
matches, then I skip it and incrementally copy the rest in a separate
loop (avoiding an extra branch). I `delete` the remaining parts in the
polyfill, but you wouldn't need to do that with normal arrays (just
non-array array-likes).
As for performance, branch prediction matters a lot more than local
variable assignment, and sequential array iteration usually hits the
CPU cache line in any JIT.
-----
Isiah Meadows
me at isiahmeadows.com
Looking for web consulting? Or a new website?
Send me an email and we can get started.
www.isiahmeadows.com
On Fri, Nov 10, 2017 at 7:24 AM, Bob Myers <rtm at gol.com> wrote:
> Thanks for your optimization. In one of my library routines I further
> optimize this with
>
> ```js
> if (elt === item) {
> changed = true;
> } else {
> if (changed) { array[j] = elt; }
> j++;
> }
> ```
>
> To avoid unnecessary assignments (which might be expensive--I don't know,
> are they?) while you're still in the portion of the array before the first
> element to be removed.
>
> On Fri, Nov 10, 2017 at 5:39 PM, T.J. Crowder
> <tj.crowder at farsightsoftware.com> wrote:
>>
>> On Fri, Nov 10, 2017 at 11:41 AM, Bob Myers <rtm at gol.com> wrote:
>> >
>> > What's wrong with this?
>>
>> I had the impression he was trying to avoid callbacks, just using `===`.
>> But other than a missing `const` on the `for-of`, it looks nice and
>> efficient -- except that [it doesn't seem like `for-of` on arrays with the
>> default iterator is much optimized yet][1]. FWIW:
>>
>> ```js
>> function removeFromArray(array, item) {
>> let changed = false;
>> let j, i, len, elt;
>>
>> for (j = i = 0, len = array.length; i < len; ++i) {
>> elt = array[i];
>> if (elt === item) {
>> changed = true;
>> } else {
>> array[j++] = elt;
>> }
>> }
>>
>> array.length = j;
>> return changed;
>> }
>> ```
>>
>> Clunkier but apparently we're optimizing for speed...
>>
>> -- T.J. Crowder
>>
>> [1]: https://jsperf.com/for-of-vs-for-with-const-binding/1
>
>
More information about the es-discuss
mailing list