A SAX-based XML parser for parsing large files into manageable chunks
Latest commit 43dbe93 Feb 25, 2017 @soulcutter Version bump


Saxerator Build Status Code Climate

Saxerator is a streaming xml-to-hash parser designed for working with very large xml files by giving you Enumerable access to manageable chunks of the document.

Each xml chunk is parsed into a JSON-like Ruby Hash structure for consumption.

You can parse any valid xml in 3 simple steps.

  1. Initialize the parser
  2. Specify which tag you care about using a simple DSL
  3. Perform your work in an each block, or using any Enumerable method


  1. gem install saxerator
  2. Choose an xml parser
    • (default) Use ruby's built-in REXML parser - no other dependencies necessary
    • gem install nokogiri
    • gem install ox
  3. If not using the default, specify your adapter in the Saxerator configuration


The DSL consists of predicates that may be combined to describe which elements the parser should enumerate over. Saxerator will only enumerate over chunks of xml that match all of the combined predicates (see Examples section for added clarity).

Predicate Explanation
all Returns the entire document parsed into a hash. Cannot combine with other predicates
for_tag(name) Elements whose name matches the given name
for_tags(names) Elements whose name is in the names Array
at_depth(n) Elements n levels deep inside the root of an xml document. The root element itself is n = 0
within(name) Elements nested anywhere within an element with the given name
child_of(name) Elements that are direct children of an element with the given name
with_attribute(name, value) Elements that have an attribute with a given name and value. If no value is given, matches any element with the specified attribute name present
with_attributes(attrs) Similar to with_attribute except takes an Array or Hash indicating the attributes to match


parser = Saxerator.parser(File.new("rss.xml"))

parser.for_tag(:item).each do |item|
  # where the xml contains <item><title>...</title><author>...</author></item>
  # item will look like {'title' => '...', 'author' => '...'}
  puts "#{item['title']}: #{item['author']}"

# a String is returned here since the given element contains only character data
puts "First title: #{parser.for_tag(:title).first}"

Attributes are stored as a part of the Hash or String object they relate to

# author is a String here, but also responds to .attributes
primary_authors = parser.for_tag(:author).select { |author| author.attributes['type'] == 'primary' }

You can combine predicates to isolate just the tags you want.

require 'saxerator'

parser = Saxerator.parser(bookshelf_xml)

# You can chain predicates
parser.for_tag(:name).within(:book).each { |book_name| puts book_name }

# You can re-use intermediary predicates
bookshelf_contents = parser.within(:bookshelf)

books = bookshelf_contents.for_tag(:book)
magazines = bookshelf_contents.for_tag(:magazine)

books.each do |book|
  # ...

magazines.each do |magazine|
  # ...


Certain options are available via a configuration block at parser initialization.

Saxerator.parser(xml) do |config|
  config.output_type = :xml
Setting Default Values Description
adapter :nokogiri :nokogiri, :ox, :rexml The XML parser used by Saxerator
output_type :hash :hash, :xml The type of object generated by Saxerator's parsing. :hash generates a Ruby Hash, :xml generates a REXML::Document
symbolize_keys! n/a n/a Call this method if you want the hash keys to be symbols rather than strings
ignore_namespaces! n/a n/a Call this method if you want to treat the XML document as if it has no namespace information. It differs slightly from strip_namespaces! since it deals with how the XML is processed rather than how it is output
strip_namespaces! n/a user-specified Called with no arguments this strips all namespaces, or you may specify an arbitrary number of namespaces to strip, i.e. config.strip_namespaces! :rss, :soapenv
put_attributes_in_hash! n/a n/a Call this method if you want xml attributes included as elements of the output hash - only valid with output_type = :hash

Known Issues

  • JRuby closes the file stream at the end of parsing, therefor to perform multiple operations which parse a file you will need to instantiate a new parser with a new File object.

Other Documentation


Why the name 'Saxerator'?

It's a combination of SAX + Enumerator.

Why use Saxerator over regular SAX parsing?

Much of the SAX parsing code I've written over the years has fallen into a pattern that Saxerator encapsulates: marshall a chunk of an XML document into an object, operate on that object, then move on to the next chunk. Saxerator alleviates the pain of marshalling and allows you to focus solely on operating on the document chunk.

Why not DOM parsing?

DOM parsers load the entire document into memory. Saxerator only holds a single chunk in memory at a time. If your document is very large, this can be an important consideration.

When I fetch a tag that has one or more elements, sometimes I get an Array, and other times I get a Hash or String. Is there a way I can treat these consistently?

You can treat objects consistently as arrays using Ruby's built-in array conversion method in the form Array(element_or_array)

Generally you should not need to convert a parsed element to a String or Hash. One case it occasionally comes up is for elements that are sometimes-empty. Empty elements behave mostly like an empty Hash, however you may convert it to a more String-like object via #to_s


For running tests for all parsers run rake spec:adapters


Saxerator was inspired by - but not affiliated with - nori and Gregory Brown's Practicing Ruby

Legal Stuff

Copyright © 2012-2017 Bradley Schaefer. MIT License (see LICENSE file).