> anyways... if you want performance avoid extra function calls when
> they aren't needed. all of the above could be done much simpler - the
> following example uses only 1 hasClass call:
>
> Element.implement({
> removeAllClassExcept: function(klass) {
> this.className = this.hasClass(klass) ? klass : '';
> return this;
> }
> });
>
>
I already wondered why always using set/get for class. It's good if you want
to keep the moo style, but bad for performance in this case. The
"Precompiled RexExp"-version is damn fast with className:
(1000 items, almost the same for 100&10)
time1::: 0.000040 <= RegExp+set/get, my method
time1a: 0.000026 <=precompiled RegExp+set/get
*time1b: 0.000007* <=precompiled RegExp+className (yeah, thats one zero
more ;) )
time2::: 0.000040 <= get/set+Array.filter, Roman
time3::: 0.000031 <= hasClass?:+get/set/erease, hazlema
time4::: 0.000020 <= hasClass?:+className, rasmusfl0e
http://mootools.net/shell/bmfGu/
I didn't expected RegExp to be faster, but actually it is in this case ;-)
Just wonder why Sanford got other results...perhaps because className was
empty most time in his test...
Thanks for all which did the benchmark and contributed, such
tests/benchmarks are really interesting.
Yann
>
> On Feb 23, 4:08 pm, Yann Leretaille <[email protected]>
> wrote:
> > could somebody test this one with 100,100 & 10 Elements?
> >
> > My results:
> > n
> > FF 3.6, Linux, 64Bit FF 2.5.8,WinXP, 32Bit
> > 1000
> > time1::: 0.000033
> > *time1a: 0.000020*
> > time2::: 0.000030
> > time3::: 0.000024time1::: 0.000055
> > *time1a: 0.000036*
> > time2::: 0.000048
> > time3::: 0.000038
> > 100
> > time1::: 0.000045
> > *time1a: 0.000027*
> > time2::: 0.000040
> > time3::: 0.000032time1::: 0.000058
> > *time1a: 0.000037*
> > time2::: 0.000051
> > time3::: 0.000040
> > 10
> > time1::: 0.000052
> > *time1a: 0.000026*
> > time2::: 0.000045
> > time3::: 0.000042**time1::: 0.000077
> > *time1a: 0.000046*
> > time2::: 0.000072
> > time3::: 0.000059
> >
> > http://mootools.net/shell/J7ySB/
> >
> > On Mon, Feb 22, 2010 at 17:45, Yann Leretaille
> > <[email protected]>wrote:
> >
> > > I understand your point. Actually, benchmarking seems particualry
> difficult
> > > for this one.
> > > I created a mooshell wich creates a list with 1000 Elements, wich have
> 0-3
> > > classes, with or without base_class.
> > > The list ist cloned for each benchmark and i ran each methos is runned
> 10
> > > times.
> > > Here are the results:
> >
> > > - Method1 (my method):
> > > - return this.set('class',new
> RegExp(klass).exec(this.get('class')))
> > > ;
> >
> > > - Method1a (my method, *precompiled RegExp* )
> > > - regexp=new RegExp(klass)
> > > - this.each(function(e) {
> e.set('class',regexp.exec(e.get('class')))
> > > });
> >
> > > - Method2 (Roman):
> > > - return this.set('class', this.get('class').split(' ').filter(
> > > function(c){return c == this },klass)[0]);
> > > - Method3 (hazlema):
> > > - return this.hasClass(klass) ? this.set("class", klass) :
> this.erase(
> > > "class");
> >
> > > Results (1000 items):
> >
> > > - time1: 0.0000340
> > > - *time1a: 0.0000216*
> > > - time2: 0.0000306
> > > - time3: 0.0000234
> >
> > > It seems to depend on how much items you have what the fastet method
> is.
> > > The strange thing is that i tried it with 1,50,10,100,500,1000 items
> and now
> > > Method1a is always faster! Perhaps Firefox is caching the RegExp...
> >
> > >http://mootools.net/shell/J7ySB/
> >
> > > Yann
> >
> > > On Mon, Feb 22, 2010 at 06:47, Sanford Whiteman <
> > > [email protected]> wrote:
> >
> > >> > - Method1: 0.000265s
> > >> > - Method2: 0.000297s
> > >> > - Method3: 0.000279s
> >
> > >> > Looks like if my method is slightly faster...
> >
> > >> Mmm, something is wrong with your benchmark.
> >
> > >> Comparing your regex
> >
> > >> [f1] t.set('class',new RegExp('base_class').exec(t.get('class')));
> >
> > >> with the straightforward
> >
> > >> [f2] t.hasClass('base_class') ? t.set('class','base_class') :
> > >> t.erase('class');
> >
> > >> averaging over 10 runs of 1000 function calls each, the regex approach
> > >> is markedly slower.
> >
> > >> [f1]:
> >
> > >> .016 when base_class is present on the element
> > >> .012 when base_class is not present
> >
> > >> [f2]:
> >
> > >> .007 when base_class is present
> > >> .004 when base_class is not present
> >
> > >> You can eke out more performance by compiling the RegEx, but the basic
> > >> approach still beats it out significantly, 50-100% faster depending on
> > >> whether the target class is present.
> >
> > >> [f1-compiled]:
> >
> > >> .011 when base_class is present on the element
> > >> .008 when base_class is not present
> >
> > >> Seehttp://mootools.net/shell/B8pPc/7/.
> >
> > >> - S.
>
