Everybody is more than welcome to contribute to Oga, no matter how small the change. To keep everything running smoothly there are a bunch of guidelines that one should follow.

Code of Conduct

The code of conduct (“CoC”) can be found in the file “”. Everybody participating in this project must adhere to the rules and guidelines stated in this CoC.


  • When changing code make sure to write RSpec tests for the changes.
  • Document code using YARD. At the very least the method arguments and return value(s) should be documented.
  • Use raise for raising errors instead of fail. You’re raising errors after all, not failing them.

Submitting Changes

Before making any big changes it’s best to open a Github issue to discuss the matter, this saves you from potentially spending hours on something that might ultimately be rejected.

When making changes please stick to the existing style and patterns as this keeps the codebase consistent. If a certain pattern or style is getting in your way please open a separate issue about this so it can be discussed.

Every commit and every pull request made is carefully reviewed. Chances are I’ll spend more time reviewing it than the time an author spent on their changes. This should ensure that Oga’s codebase is stable, of high quality and easy to maintain. As such please take my feedback into consideration (or discuss it in a civilized manner) instead of just dismissing it with comments such as “But I fixed the problem so your feedback is irrelevant” or “This is my way of doing things”.

Finally, and this will sound harsh: I will not merge pull requests if the author(s) simply disregard the feedback I’ve given them or if there are other problems with the pull request. Do not expect me to just blindly accept whatever changes are submitted.

Some examples of good pull request:



Git commits should have a <= 50 character summary, optionally followed by a blank line and a more in depth description of 72 characters per line. For example:

Use blacklists/whitelists for HTML closing rules

This allows for more fine grained control over when to close certain
elements. For example, an unclosed <tr> element should be closed first
when bumping into any element other than <td> or <th>. Using the old
NodeNameSet this would mean having to list every possible HTML element
out there. Using this new setup one can just create a whitelist of the
<td> and <th> elements.

Please, please write meaningful commit messages. Writing a good commit messages is just as important as writing good code. If you’re having trouble writing a commit message you should try to break the commits up into smaller chunks. You can do so using a git rebase.

Editor Setup

Use spaces for indentation, tabs are not accepted. The usage of spaces ensures the indentation is identical no matter what program or system is used to view the source code.

Hard wrap lines at 80 characters per line. Most modern editors can easily handle this, if not you should get a better editor. For example, in Vim you can select text in visual mode (using v) and press gq to automatically re-wrap the selected text.

It’s OK if a line is a few characters longer than 80 but please keep it as close to 80 characters as possible. Typically I do this when wrapping the line results in several extra lines without it being much more readable.

I often have multiple windows vertically next to each other and 80 characters per line is the only setup that lets me do so, even on smaller screen resolutions. For example, my typical setup is 1 file browser and two vertical windows. Using 80 characters per line ensures all code fits in that space along with some slight padding to make reading more pleasant.

To make this process easier Oga comes with an EditorConfig configuration file. If your editor supports this it will automatically apply the required settings for you.

Hacking on Oga

Before you start hacking on Oga make sure the following libraries/tools are installed:

  • Ragel 6.x (6.9 recommended)
  • gunzip (to unpack the fixtures)
  • javac (only when using JRuby)

Assuming you have the above tools installed and a local Git clone of Oga, lets install the required Gems:

bundle install

Next up, compile the required files and run the tests:


You can compile the various parsers/extensions by running:

rake generate

For more information about the available tasks, run rake -T.

Running Benchmarks

Benchmarks are located in the benchmark directory. Some of these require fixture files which can be generated by running rake fixtures. Running a benchmark is just a matter of running a Ruby script, for example:

ruby benchmark/xml/parser/parser_bench.rb


Tests are written using RSpec and use the “should” syntax instead of the “expect” syntax (for as long as RSpec keeps supporting this). This means that assertions are written as following:

some_object.should == some_value

instead of this:

expect(some_object).to eq(some_value)

Specification blocks should be written using it, grouping should be done using describe. Specification descriptions should be meaningful and human friendly English. For example:

describe Oga::XML::Entities do
  describe 'decode' do
    it 'decodes &lt; into <' do
      # ...

Typically the top-level describe block is used to describe a method name. In such a case use describe 'foo' for class methods and describe '#foo' for instance methods.

Do not use let for creating data re-used between specifications, instead use a before block that sets an instance variable. In other words, use this:

before do
  @number = 10

instead of this:

let(:number) { 10 }

Instance variables stand out much more and they don’t require one to also understand what exactly let does which in turn simplifies the process of reading and writing specifications.

Whenever adding new specifications please keep them in the existing style unless I indicate otherwise. There’s nothing more annoying than inconsistent specifications.

If you insist on changing the structure/style of specifications please open an issue and ask about it before making any changes. I am very picky about how I want things and it would be a shame for somebody to spend hours working on something that isn’t going to be merged in any way.

Continuous Integration

Two continuous integration services are used to ensure the tests of Oga pass at all times:

Please note that I will not accept patches that break any tests unless stated otherwise.

Extension Setup

Oga uses native extensions for the XML lexer. This is due to Ruby sadly not being fast enough to chew through large amounts of XML (at least when using Ragel). For example, the benchmark benchmark/lexer/big_xml_time.rb would take around 6 seconds to complete on MRI 2.1.1. The native extensions on the other hand can complete this benchmark in roughly 600 milliseconds.

Oga has two native extensions: one for MRI/Rubinius (written in C) and one for JRuby (written in Java). Both extensions share the same Ragel grammar, found in ext/ragel/base_lexer.rl. This grammar is set up in such a way that the syntax is compatible with both C and Java. Specific details on how the grammar is used can be found in the documentation of said grammar file.

The native extensions call back in to Ruby to actually perform the task of creating tokens, validating input and so forth. As a result of this you’ll most likely never have to touch the C and/or Java code when changing the behaviour of the lexer.

To compile the extensions run rake generate using your Ruby implementation of choice. Note that extensions compiled for MRI can not be used on Rubinius and vice-versa. To compile the JRuby extension you’ll have to switch your active Ruby version to JRuby first.

Thread Safety

To ensure Oga remains thread-safe for as much as possible the usage of global objects and/or state is forbidden. This means that you should only use constants/class methods for static/read-only data (e.g. an Array of static Strings). In other words, this is fine:

NUMBERS = [10, 20, 30]

NUMBERS.each do |number|


But this is not:


output = TOOL.replace(input, 'foo', 'bar')

The exception here are libraries that are designed to be thread-safe, clearly state this and can prove it (e.g. by simply using a mutex). Even then global state is highly frowned upon.

Loading Libraries

All require calls should be placed in lib/oga.rb. Any require calls specific to a Ruby implementation (e.g. JRuby) should be wrapped in a conditional. For example:

if RUBY_PLATFORM == 'java'

For loading files in Oga itself require should be used. Don’t modify $LOAD_PATH, instead run any scripts using ruby -I lib.


In case you have any further questions or would like to receive feedback before submitting a change, feel free to contact me. You can either open an issue, send a tweet to @yorickpeterse or send an Email to