Gammo - A pure-Ruby HTML5 parser

Build Status GitHub issues GitHub forks GitHub stars GitHub license Documentation

Gammo is an implementation of the HTML5 parsing algorithm which conforms the WHATWG specification, without any dependencies. Given an HTML string, Gammo parses it and builds DOM tree based on the tokenization and tree-construction algorithm defined in WHATWG parsing algorithm.

Gammo, its naming is inspired by Gumbo. But Gammo is a fried tofu fritter made with vegetables.

require 'gammo'
require 'open-uri'

parser = open('') { |f| }
parser.parse #=> #<Gammo::Node::Document>




Gammo::Tokenizer implements the tokenization algorithm in WHATWG. You can get tokens in order by calling Gammo::Tokenizer#next_token.

Here is a simple example for performing only the tokenizer.

def dump_for(token)
  puts "data: #{}, class: #{token.class}"

tokenizer ='<!doctype html><input type="button"><frameset>')
dump_for tokenizer.next_token #=> data: html, class: Gammo::Tokenizer::DoctypeToken
dump_for tokenizer.next_token #=> data: input, class: Gammo::Tokenizer::StartTagToken
dump_for tokenizer.next_token #=> data: frameset, class: Gammo::Tokenizer::StartTagToken
dump_for tokenizer.next_token #=> data: end of string, class: Gammo::Tokenizer::ErrorToken

The parser described below depends on this tokenizer, it applies the WHATWG parsing algorithm to the tokens extracted by this tokenization in order.

Token types

The tokens generated by the tokenizer will be categorized into one of the following types:

Token type Description
Gammo::Tokenizer::ErrorToken Represents an error token, it usually means end-of-string.
Gammo::Tokenizer::TextToken Represents a text token like "foo" which is inner text of elements.
Gammo::Tokenizer::StartTagToken Represents a start tag token like <a>.
Gammo::Tokenizer::EndTagToken Represents an end tag token like </a>.
Gammo::Tokenizer::SelfClosingTagToken Represents a self closing tag token like <img />
Gammo::Tokenizer::CommentToken Represents a comment token like <!-- comment -->.
Gammo::Tokenizer::DoctypeToken Represents a doctype token like <!doctype html>.


Gammo::Parser implements processing in the tree-construction stage based on the tokenization described above.

A successfully parsed parser has the document accessor as the root document (this is the same as the return value of the Gammo::Parser#parse). From the document accessor, you can traverse the DOM tree constructed by the parser.

require 'gammo'
require 'pp'

document ='<!doctype html><input type="button">').parse

def dump_for(node, strm)
  strm << node.to_h
  return unless node && (child = node.first_child)
  while child
    dump_for(child, (strm.last[:children] ||= []))
    child = child.next_sibling

pp dump_for(document, [])


Currently, it's not possible to traverse the DOM tree with css selector or xpath like Nokogiri. However, Gammo plans to implement these features in the future.


The nodes generated by the parser will be categorized into one of the following types:

Node type Description
Gammo::Node::Error Represents error node, it usually means end-of-string.
Gammo::Node::Text Represents the text node like "foo" which is inner text of elements.
Gammo::Node::Document Represents the root document type. It's always returned by Gammo::Parser#document.
Gammo::Node::Element Represents any elements of HTML like <p>.
Gammo::Node::Comment Represents comments like <!-- foo -->
Gammo::Node::Doctype Represents doctype like <!doctype html>

For some nodes such as Gammo::Node::Element and Gammo::Node::Document, they contain pointers to nodes that can be referenced by itself, such as Gammo::Node#next_sibling or Gammo::Node#first_child. In addition, APIs such as Gammo::Node#append_child and Gammo::Node#remove_child that perform operations defined in DOM living standard are also provided.

DOM Tree Traversal

Currently, XPath 1.0 is the only way for traversing DOM tree built by Gammo. CSS selector support is also planned but not having any ETA.

XPath 1.0 (experimental)

Gammo has an original lexer/parser for XPath 1.0, it's provided as a helper in the DOM tree built by Gammo. Here is a simple example:

document ='<!doctype html><input type="button">').parse
node_set = document.xpath('//input[@type="button"]') #=> "<Gammo::XPath::NodeSet>"

node_set.length #=> 1
node_set.first #=> "<Gammo::Node::Element>"

Since this is implemented by full scratch, Gammo is providing this support as a very experimental feature. Please file an issue if you find bugs.


Before proceeding at the details of XPath support, let's have a look at a few simple examples. Given a sample HTML text and its DOM tree:

document =<<-EOS).parse
<!DOCTYPE html>
  <p class="description">Here is a sample web site.</p>
  <ul id="links">
    <li>Google <a href=""></a></li>
    <li>GitHub <a href=""></a></li>

The following XPath expression gets all li elements and prints those text contents:

document.xpath('//li').each do |elm|
  puts elm.inner_text

The following XPath expression gets all li elements under the ul element having the id=links attribute:

document.xpath('//ul[@id="links"]/li').each do |elm|
  puts elm.inner_text

The following XPath expression gets each text node for each li element under the ul element having the id=links attribute:

document.xpath('//ul[@id="links"]/li/text()').each do |elm|

Axis Specifiers

In the combination with Gammo, the axis specifier indicates navigation direction within the DOM tree built by Gammo. Here is list of axes. As you can see, Gammo fully supports the all of axes.

Full Syntax Abbreviated Syntax Supported Notes
ancestor yes
ancestor-or-self yes
attribute @ yes @abc is the alias for attribute::abc
child yes abc is the short for child::abc
descendant yes
descendant-or-self // yes // is the alias for /descendant-or-self::node()/
following yes
following-sibling yes
namespace yes
parent .. yes .. is the alias for parent::node()
preceding yes
preceding-sibling yes
self . yes . is the alias for self::node()

Node Test

Node tests consist of specific node names or more general expressions. Although particular syntax like : should work for specifying namespace prefix in XPath, Gammo does not support it yet as it's not a core feature in HTML5.

Full Syntax Supported Notes
text() yes Finds a node of type text, e.g. hello in

hello world

comment() yes Finds a node of type comment, e.g.
node() yes Finds any node at all.

Also note that the processing-instruction is not supported. There is no plan to support it.


  • The /, // and [] are used in the path expression.
  • The union operator | forms the union of two node sets.
  • The boolean operators: and, or
  • The arithmetic operators: +, -, *, div and mod
  • Comparison operators: =, !=, <, >, <=, >=


XPath 1.0 defines four data types (nodeset, string, number, boolean) and there are various functions based on the types. Gammo supports those functions partially, please check it to be supported before using functions.

Node set functions
Function Name Supported Specification
last() yes
position() yes
count(node-set) yes
String Functions
Function Name Supported Specification
string(object?) yes
concat(string, string, string*) yes
starts-with(string, string) yes
contains(string, string) yes
substring-before(string, string) yes
substring-after(string, string) yes
substring(string, number, number?) no
string-length(string?) no
normalize-space(string?) no
translate(string, string, string) no
Boolean Functions
Function Name Supported Specification
boolean(object) yes
not(object) yes
true() yes
false() yes
lang() no
Number Functions
Function Name Supported Specification
number(object?) no
sum(node-set) no
floor(number) no
ceiling(number) yes
round(number) no

CSS Selector



As mentioned in the features at the beginning, Gammo doesn't prioritize its performance. Thus, for example, Gammo is not suitable for very performance-sensitive applications (e.g. performing Gammo parsing synchronously from an incoming request from an end user). Instead, the goal is to work well with batch processing such as crawlers. Gammo places the highest priority on making it easy to parse HTML by peforming it without depending on native-extensions and external gems.


This was developed with reference to the following softwares.

  • x/net/html: I've been working on this package, it gave me strong reason to make this happen.
  • Blink: Blink gave me great impression about tree construction.
  • html5lib-tests: Gammo relies on this test.


The gem is available as open source under the terms of the MIT License.

Release History

  • v0.2.0
    • XPath 1.0 support #4
  • v0.1.0
    • Initial Release