Performance inquiry

[Berislav Lopac]

Imagine three options:

a) There is an XML file. A PHP5 script loads it and puts it into a DOM
document object. Then, given an input string, a function looks via XPath
for a node with that name (or possibly value of an attribute) and returns
the node's content.

b) There is a text file containing a serialized array. PHP script opens it
and deserializes the array into a variable; then, given an input string,
looks for an array element with the same key and returns the element's
value.

c) There is a PHP file containing just a definition for an array. PHP
script includes it and, given an input string, looks for an array element
with the same key and returns the element's value.

Does anyone have any idea which option would be the fastest, and which the
slowest of the three?

Thanks,

Berislav

[Tony Marston]

Re: Performance inquiry

As you are clever enough to come up with that question why can't you write
the code to test each option yourself?

--
Tony Marston

http://www.tonymarston.net

"Berislav Lopac" <berislav.lopac @dimedia.hr> wrote in message
news:k9hxyjv8s1 pu.1xt3kn262k71 c.dlg@40tude.ne t...[color=blue]
> Imagine three options:
>
> a) There is an XML file. A PHP5 script loads it and puts it into a DOM
> document object. Then, given an input string, a function looks via XPath
> for a node with that name (or possibly value of an attribute) and returns
> the node's content.
>
> b) There is a text file containing a serialized array. PHP script opens it
> and deserializes the array into a variable; then, given an input string,
> looks for an array element with the same key and returns the element's
> value.
>
> c) There is a PHP file containing just a definition for an array. PHP
> script includes it and, given an input string, looks for an array element
> with the same key and returns the element's value.
>
> Does anyone have any idea which option would be the fastest, and which the
> slowest of the three?
>
> Thanks,
>
> Berislav[/color]

[Ward]

Re: Performance inquiry

"Tony Marston" <tony@NOSPAM.de mon.co.uk> wrote in
news:cj8ogn$fh2 $1$8302bc10@new s.demon.co.uk:
[color=blue]
> Path:
> phobos.telenet-ops.be!ossa.tel enet-ops.be!feed1.ne ws.be.easynet.n et!new
> s2.telebyte.nl! news.glorb.com! border1.nntp.dc a.giganews.com! nntp.gigane
> ws.com!peer01.c ox.net!cox.net! peer-uk.news.demon.n et!kibo.news.de mon.ne
> t!news.demon.co .uk!demon!not-for-mail From: "Tony Marston"
> <tony@NOSPAM.de mon.co.uk> Newsgroups: comp.lang.php
> Subject: Re: Performance inquiry
> Date: Mon, 27 Sep 2004 11:03:03 +0100
> Lines: 36
> Message-ID: <cj8ogn$fh2$1$8 302bc10@news.de mon.co.uk>
> References: <k9hxyjv8s1pu.1 xt3kn262k71c.dl g@40tude.net>
> Reply-To: "Tony Marston" <tony@marston-home.demon.co.u k>
> NNTP-Posting-Host: marston-home.demon.co.u k
> X-Trace: news.demon.co.u k 1096279383 15906 80.177.120.119 (27 Sep 2004
> 10:03:03 GMT) X-Complaints-To: abuse@demon.net
> NNTP-Posting-Date: Mon, 27 Sep 2004 10:03:03 +0000 (UTC)
> X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2900.2180
> X-Priority: 3
> X-RFC2646: Format=Flowed; Original
> X-MSMail-Priority: Normal
> X-Newsreader: Microsoft Outlook Express 6.00.2900.2180
> Xref: ossa.telenet-ops.be comp.lang.php:6 9100
>
> As you are clever enough to come up with that question why can't you
> write the code to test each option yourself?
>[/color]

and post the results here ;-)

[Thomas Stian Bergheim]

Re: Performance inquiry

On Mon, 27 Sep 2004 10:57:16 +0200, Berislav Lopac
<berislav.lopac @dimedia.hr> wrote:
[color=blue]
> Imagine three options:
>
> a) There is an XML file. A PHP5 script loads it and puts it into a DOM
> document object. Then, given an input string, a function looks via XPath
> for a node with that name (or possibly value of an attribute) and returns
> the node's content.
>
> b) There is a text file containing a serialized array. PHP script opens
> it
> and deserializes the array into a variable; then, given an input string,
> looks for an array element with the same key and returns the element's
> value.
>
> c) There is a PHP file containing just a definition for an array. PHP
> script includes it and, given an input string, looks for an array element
> with the same key and returns the element's value.
>
> Does anyone have any idea which option would be the fastest, and which
> the
> slowest of the three?
>
> Thanks,
>
> Berislav[/color]

c > b > a


--
Using Opera's revolutionary e-mail client: http://www.opera.com/m2/

[Berislav Lopac]

Re: Performance inquiry

On Mon, 27 Sep 2004 13:47:15 +0200, Thomas Stian Bergheim wrote:
[color=blue]
> c > b > a[/color]

Sorry for splitting hairs here, but is there any empirical data behind it,
or is it just gut feeling (mine says so too)?

Berislav

[rush]

Re: Performance inquiry

"Berislav Lopac" <berislav.lopac @dimedia.hr> wrote in message
news:1hsnmld88p qoa$.1cjdqlsuih vrk.dlg@40tude. net...[color=blue]
> On Mon, 27 Sep 2004 13:47:15 +0200, Thomas Stian Bergheim wrote:
>[color=green]
> > c > b > a[/color]
>
> Sorry for splitting hairs here, but is there any empirical data behind it,
> or is it just gut feeling (mine says so too)?[/color]

Unless you count " .. or I will eat my hat if it ain't so.." as empirical
expirience ;) .

I would expect the xml parsing to be the slowest, then array
deserialization , and hardcoded constant to be the fastest. Benchmarking xml
and deserialization should give some meaningfull results, hardcoded constant
should be more trickey, since zend or turk cache could be playing major
role.

rush
--

http://www.templatetamer.com/

[Gordan]

Re: Performance inquiry

Berislav Lopac wrote:[color=blue]
> Imagine three options:[/color]
....cut...

Aren't we missing something here :-)
Are we talking about 10 nods/elements/whatever things or are we talking
about 10^10 elements?
I believe it does make a diference

gordan

[Berislav Lopac]

Re: Performance inquiry

On Mon, 27 Sep 2004 15:47:07 +0200, Gordan wrote:
[color=blue]
> Berislav Lopac wrote:[color=green]
>> Imagine three options:[/color]
> ...cut...
>
> Aren't we missing something here :-)
> Are we talking about 10 nods/elements/whatever things or are we talking
> about 10^10 elements?
> I believe it does make a diference[/color]

For multi-level hierarchies, absolutely. I am right now interested only in
one-level ones, but I'd like to hear if anyone has some hard data for
either.

Berislav

[Chung Leong]

Re: Performance inquiry

"rush" <pipa@rush.aval on.hr> wrote in message
news:cj93ok$6n1 $1@ls219.htnet. hr...[color=blue]
> "Berislav Lopac" <berislav.lopac @dimedia.hr> wrote in message
> news:1hsnmld88p qoa$.1cjdqlsuih vrk.dlg@40tude. net...[color=green]
> > On Mon, 27 Sep 2004 13:47:15 +0200, Thomas Stian Bergheim wrote:
> >[color=darkred]
> > > c > b > a[/color]
> >
> > Sorry for splitting hairs here, but is there any empirical data behind[/color][/color]
it,[color=blue][color=green]
> > or is it just gut feeling (mine says so too)?[/color]
>
> Unless you count " .. or I will eat my hat if it ain't so.." as empirical
> expirience ;) .
>
> I would expect the xml parsing to be the slowest, then array
> deserialization , and hardcoded constant to be the fastest. Benchmarking[/color]
xml[color=blue]
> and deserialization should give some meaningfull results, hardcoded[/color]
constant[color=blue]
> should be more trickey, since zend or turk cache could be playing major
> role.
>
> rush
> --
> http://www.templatetamer.com/
>[/color]

Deserialization should be faster than an array in PHP code actually, since
there's no tokenizatin/compilation involved. The data is most densely packed
as well.

[rush]

Re: Performance inquiry

"Chung Leong" <chernyshevsky@ hotmail.com> wrote in message
news:Po-dnT2s4eJIssfcRV n-rg@comcast.com. ..[color=blue]
> Deserialization should be faster than an array in PHP code actually, since
> there's no tokenizatin/compilation involved. The data is most densely[/color]
packed[color=blue]
> as well.[/color]

Well, turk cache or something similar would change that other way around,
that's why I have mentioned that benchmarking constants would be tricky.

rush
--

http://www.templatetamer.com/

[Chung Leong]

Re: Performance inquiry

"rush" <pipa@rush.aval on.hr> wrote in message news:<cjdqff$d9 a$1@ls219.htnet .hr>...[color=blue]
> "Chung Leong" <chernyshevsky@ hotmail.com> wrote in message
> news:Po-dnT2s4eJIssfcRV n-rg@comcast.com. ..[color=green]
> > Deserialization should be faster than an array in PHP code actually, since
> > there's no tokenizatin/compilation involved. The data is most densely[/color]
> packed[color=green]
> > as well.[/color]
>
> Well, turk cache or something similar would change that other way around,
> that's why I have mentioned that benchmarking constants would be tricky.
>
> rush[/color]

I doubt it. Even if caching is present, you wouldn't have the array
just sitting in memory. PHP is not C after all. A array declaration is
a chunk of code, where each element is an expression. To instancize
the array, the PHP interpreter has to evaluate the code, which
obviously take time.

The following is a little benchmark comparing unserialize() with
array():

$x = serialize(array ( 'Hello', 'Hello', 'Hello', 'Hello', 'Hello',
'Hello', 'Hello', 'Hello', 'Hello', 'Hello', 'Hello', 'Hello',
'Hello', 'Hello', 'Hello', 'Hello', 'Hello', 'Hello', 'Hello',
'Hello', 'Hello', 'Hello', 'Hello', 'Hello', 'Hello', 'Hello',
'Hello', 'Hello', 'Hello', 'Hello', 'Hello', 'Hello', 'Hello'
));

$s = time();

for($i = 0; $i < 2000000; $i++) {
$a = unserialize($x) ;
}

$e = time();
$d = $e - $s;

echo "$d<br>";

$s = time();

for($i = 0; $i < 2000000; $i++) {
$a = array( 'Hello', 'Hello', 'Hello', 'Hello', 'Hello',
'Hello', 'Hello', 'Hello', 'Hello', 'Hello', 'Hello', 'Hello',
'Hello', 'Hello', 'Hello', 'Hello', 'Hello', 'Hello', 'Hello',
'Hello', 'Hello', 'Hello', 'Hello', 'Hello', 'Hello', 'Hello',
'Hello', 'Hello', 'Hello', 'Hello', 'Hello', 'Hello', 'Hello' );
}

$e = time();
$d = $e - $s;

echo "$d<br>";

The first loop required 30 sec while the second took 26 sec. Even
under this ideal situation the array declaration isn't that much
faster. Add in any cache overhead (mtime check, for instance) and
it'll end up being slower.

[rush]

Re: Performance inquiry

"Chung Leong" <chernyshevsky@ hotmail.com> wrote in message
news:a164f4b5.0 409291026.66a55 0e7@posting.goo gle.com...[color=blue]
> "rush" <pipa@rush.aval on.hr> wrote in message[/color]
news:<cjdqff$d9 a$1@ls219.htnet .hr>...[color=blue][color=green]
> > "Chung Leong" <chernyshevsky@ hotmail.com> wrote in message
> > news:Po-dnT2s4eJIssfcRV n-rg@comcast.com. ..[/color]
>
> I doubt it. Even if caching is present, you wouldn't have the array
> just sitting in memory. PHP is not C after all. A array declaration is
> a chunk of code, where each element is an expression. To instancize
> the array, the PHP interpreter has to evaluate the code, which
> obviously take time.[/color]

Well I can agree that issue is at least very tricky. But, I was doing quite
a lot of testing for TT, to find out if it is more effcient to load template
string from file, or as constant in php code. The results were not a clear
cut, but when php accelerator was present, code variant seemed to be faster
on average.

rush
--

http://www.templatetamer.com/