These benchmarks were written for the article "XML Parsing in Ruby" on the Engine Yard blog.

http://www.engineyard.com/blog/2009/xml-parsing-in-ruby/

Usage

To use them, first ensure you have the Hpricot and Nokogiri gems installed.

gem install nokogiri hpricot

Benchmarks are run with Rake. To run a benchmark, run one of the following:

• rake bench:all Run all available benchmarks
• rake bench: Run one package's benchmarks (where is the name of one of the subdirectories in lib/xmlbench)
• rake bench BENCH= Run one or more individual benchmarks.

Pass N= on the command-line to change the number of iterations.

To add more data files, edit the Rakefile and change the URLs hash at the top. Be aware that some of the benchmarks make assumptions about the documents taking the form of an Atom document, so those benchmarks may break or be invalid.

Creating a new benchmark

A benchmark is an object responds to a method named #perform(xml_input). The input is an IO-like stream containing the document to be parsed. A benchmark can also respond to a method named #prepare(xml_stream). The prepare method is called once before the benchmark measurement and can be used to setup any state whose creation shouldn't be included in the measurements.

Each benchmark should be in its own file and have an overload for the Harness.parser factory method as follows:

class Harness
  def self.parser
    MyParser.new
  end
end

Notes

If you're running benchmarks with JRuby, be aware of the following issues:

• To run the Nokogiri benchmarks on JRuby, you need to enable objectspace (jruby -X+O; as of Nokogiri release 1.4)
• However, turning on objectspace slows down overall JRuby performance, so it's recommended to only enable it for Nokogiri.
• To run the jaxp/stream benchmark, you need Java 6 or higher or you need to manually download and install the StAX parser and APIs on your classpath.