Python Buzz Forum - Lies, Damned Lies and [Benchmarks]

Articles |
News |
Weblogs |
Books |
Forums

Artima Forums | Articles | Weblogs | Java Answers | News

Sponsored Link •

Python Buzz Forum
Lies, Damned Lies and [Benchmarks]

2 replies on 1 page. Most recent reply: Jan 4, 2005 7:22 AM by Aaron Brady

Welcome Guest
Sign In

Back to Topic List

Reply to this Topic

Search Forum

Threaded View


Previous Topic		Next Topic

Flat View: This topic has 2 replies on 1 page

Aaron Brady

Posts: 576
Nickname: insommeuk
Registered: Aug, 2003

Aaron Brady is lead developer for Crestsource

Lies, Damned Lies and [Benchmarks]

Posted: Jan 1, 2005 4:14 PM

This post originated from an RSS feed registered with Python Buzz by Aaron Brady.
Original Post: Lies, Damned Lies and [Benchmarks] Feed Title: insom.me.uk Feed URL: http://feeds2.feedburner.com/insommeuk Feed Description: Posts related to using Python. Some tricks and tips, observations, hacks, and the Brand New Things.	Latest Python Buzz Posts Latest Python Buzz Posts by Aaron Brady Latest Posts From insom.me.uk

My experience screen-scraping with HTML Tidy and 4DOM. And why pyRXP is a tiny-god.

Read: Lies, Damned Lies and [Benchmarks]

Marius Gedminas

Posts: 1
Nickname: mgedmin
Registered: Jan, 2005

Re: Lies, Damned Lies and [Benchmarks]

Posted: Jan 1, 2005 6:38 PM

I wanted to know how libxml2 (www.xmlsoft.org) compares in this benchmark, so I adapted your benchmark script to use libxml2: http://mg.pov.lt/test-libxml.py

Benchmarks on my 666 MHz Thinkpad (measured by running time python test-foo.py > /dev/null three times and taking the lowest value):


   4dom:     0m6.091s (variance: 0.004s)
   minidom:  0m0.616s (variance: 0.011s)
   libxml2:  0m0.232s (variance: 0.006s)
   pyRXP:    0m0.107s (variance: 0.001s)

libxml2 is faster than minidom, but slower than pyRXP. libxml2 supports Unicode and XPath, however it is not very Pythonic and forces you to do a certain amount of manual memory management (wrappers can help here, e.g. [1] or [2]). Comparing test-r.py and test-libxml.py, I'd choose libxml2 -- readability of the source code (xpath!) beats the slight speed disadvantage.

[1] http://source.schooltool.org/viewcvs/trunk/schooltool/src/schooltool/rest/xmlparsing.py?rev=1730&view=markup

[2] http://codespeak.net/svn/lxml/trunk/

Aaron Brady

Posts: 576
Nickname: insommeuk
Registered: Aug, 2003

Re: Lies, Damned Lies and [Benchmarks]

Posted: Jan 4, 2005 7:22 AM

Thanks for the post! I might consider libxml2 in the future, I've heard good things, but the manual memory management scared me off :)

I've done plenty of programming in languages without a GC, but when I'm in Python I don't think about free()ing objects. Will check out the wrappers, though.

Flat View: This topic has 2 replies on 1 page

Previous Topic

Next Topic


	Web Artima.com