Trim Mac Roundup (updated)

At the blog Flagrant Badassery I found some very thorough javascript trim research. I was really amazed by his findings. Did you know there’s a (performance) difference between /\s+/ and /\s\s*/? I thought the regular expression engines would optimize these differences away.

Please do read the article, for it’s pretty interesting. After reading the piece, I had two questions. What was the situation on Mac OS, and of course I had to test my own implementation. After coding, I realized the implementation was similar to the trim10 implementation. The code:

var trim12=function(s) {
  var whitespace = ' \n\r\t\v\f\u00a0\u2000\u2001\u2002\u2003\u2004\u2005\u2006\u2007\u2008\u2009\u200a\u200b\u2028\u2029\u3000';
  var i=0, j=s.length-1;
  while(i<s.length && whitespace.indexOf(s.charAt(i))!=-1) i++;
  while(j>i && whitespace.indexOf(s.charAt(j))!=-1) j--;
  return s.substring(i,j+1);
}

It turns out, this code is pretty fast. The difference with trim10 is it only makes one copy of the string.

But now to more interesting territory: How is the Mac doing? These figures are from my old iMac G4 @ 700Mhz, which is serving me well. I couldn’t test for Internet Explorer, since I installed OS X 10.4 from scratch. But it’s dead anyways. So is Netscape 4.

The results:

Layout engine	WebCore		Gecko		Presto
method\browser	Safari 2.0.4	Webkit 24064	Firefox 2.0.0.4	Camino 1.5	Opera 9.21
trim1	11.9ms	12.3ms	6.5ms	6.8ms	28.7ms
trim2	12.8ms	13.8ms	10.6ms	10.3ms	30.6ms
trim3	17.1ms	17.7ms	15.5ms	16.6ms	44.3ms
trim4	17.9ms	18.4ms	16.0ms	14.8ms	43.7ms
trim5	29.4ms	43.3ms	29.4ms	27.8ms	39.3ms
trim6	29.6ms	44.8ms	38.8ms	31.9ms	84.2ms
trim7	29.3ms	43.4ms	25.9ms	26.6ms	41.1ms
trim8	12.2ms	12.6ms	29.8ms	27.7ms	33.2ms
trim9	17.5ms	11.3ms	32.1ms	32.5ms	1038.5ms
trim10	0.4ms	0.4ms	0.3ms	0.1ms	0.3ms
trim11	0.7ms	0.6ms	0.9ms	0.8ms	1.5ms
trim12	0.3ms	0.3ms	0.3ms	0.2ms	0.3ms

I ran every test 20 times. For every test, I did a “cold start” (loaded the page fresh into memory). After pressing start, I didn’t touch my computer until it was finished. One of the very odd things is, trim9 really takes forever on Opera. Further, there are no surprises. Except maybe the observation, all Mac browsers (well, all three layout engines) perform pretty much the same.

Conclusions:

one doesn’t need a javascript library, to find something time consuming
trim1 is still the best all-rounder
avoid trim9 for our Opera friends
trim12 is well suited for trimming big strings in all browsers

You can do the tests yourself on the testpage.

updated: The browser doesn’t hang now when you feed it an empty or all whitespace string

javascript