On Nov 7, 2007 1:49 PM, mortee <[EMAIL PROTECTED]> wrote:
> I just set out to do some simple measurements to see how fast libxml may
> be compared to hpricot.
>
> I made a little script with a ~4 megs XML document appended after __END__.
>
> $ uname -s
> CYGWIN_NT-5.1
> $ gem list libxml
>
> *** LOCAL GEMS ***
>
> libxml-ruby (0.5.2.0)
> LibXML2 bindings for Ruby
> $ head -18 ./xml-bm2.rb
> #!/usr/bin/env ruby
> require 'benchmark'
> require 'hpricot'
> require 'xml/libxml'
>
> xml = DATA.read
>
> Benchmark.bmbm { |b|
> b.report('hpricot') do
> Hpricot::XML(xml).search('data').each{}
> end
> b.report('libxml') do
> XML::Parser.string(xml).parse.find('//data').each{}
> end
> }
>
> __END__
> <?xml version="1.0" encoding="UTF-8" standalone="no"?>
> $ ./xml-bm2.rb
> Rehearsal -------------------------------------------
> hpricot 14.407000 0.157000 14.564000 ( 15.686000)
> libxml 0.796000 0.093000 0.889000 ( 42.462000)
> --------------------------------- total: 15.453000sec
>
> user system total real
> hpricot 13.797000 0.000000 13.797000 ( 15.578000)
> libxml 0.859000 0.016000 0.875000 ( 41.091000)
>
> As you can see, hpricot has finished with parsing the XML previously
> loaded into memory about three times faster in real time than libxml.
> Also the other figures for libxml are pretty interesting. To this comes
> the fact that while hpricot processes the document, my CPU maxes out all
> the way through - however during the libxml phase, it's virtually idle.
>
> Does anyone have any clue as to why this may happen, and how to have
> libxml live up to its potential?...
I've seen this kind of things before too. I was getting faster speeds
out of Rexml! libxml is fast, but something in the Ruby binding is
putting the kabosh on it. We really need a performance analysis run.
How hard is it to profile your code? Have you tried profile.rb?
I would like to get a release of Dan's great work out soon. But maybe
we should address this issue first.
T.
_______________________________________________
libxml-devel mailing list
libxml-devel@rubyforge.org
http://rubyforge.org/mailman/listinfo/libxml-devel
