SLAXML is a pure-Lua SAX-like streaming XML parser. It is more robust than many (simpler) pattern-based parsers that exist (such as mine), properly supporting code like <expr test="5 > 7" />, CDATA nodes, comments, namespaces, and processing instructions.

It is currently not a truly valid XML parser, however, as it allows certain XML that is syntactically-invalid (not well-formed) to be parsed without reporting an error.

• Pure Lua in a single file (two files if you use the DOM parser).
• Streaming parser does a single pass through the input and reports what it sees along the way.
• Supports processing instructions (<?foo bar?>).
• Supports comments (<!-- hello world -->).
• Supports CDATA sections (<![CDATA[ whoa <xml> & other content as text ]]>).
• Supports namespaces, resolving prefixes to the proper namespace URI (<foo xmlns="bar"> and <wrap xmlns:bar="bar"><bar:kittens/></wrap>).
• Supports unescaped greater-than symbols in attribute content (a common failing for simpler pattern-based parsers).
• Unescapes named XML entities (< > & " ') and numeric entities (e.g. 
) in attributes and text nodes (but—properly—not in comments or CDATA). Properly handles edge cases like &.
• Optionally ignore whitespace-only text nodes (as appear when indenting XML markup).
• Includes an optional DOM parser that is both a convenient way to pull in XML to use as well as a nice example of using the streaming parser.
  • DOM module also provides DOM-to-XML serialization, including options for pretty-printing and sorting (making plain-text diffs sane). Parse XML, modify Lua tables, and then round-trip the results back to XML.
• Does not add any keys to the global namespace.

local SLAXML = require 'slaxml'

local myxml = io.open('my.xml'):read('*all')

-- Specify as many/few of these as you like
parser = SLAXML:parser{
  startElement = function(name,nsURI,nsPrefix)       end, -- When "<foo" or <x:foo is seen
  attribute    = function(name,value,nsURI,nsPrefix) end, -- attribute found on current element
  closeElement = function(name,nsURI)                end, -- When "</foo>" or </x:foo> or "/>" is seen
  text         = function(text,cdata)                end, -- text and CDATA nodes (cdata is true for cdata nodes)
  comment      = function(content)                   end, -- comments
  pi           = function(target,content)            end, -- processing instructions e.g. "<?yes mon?>"
}

-- Ignore whitespace-only text nodes and strip leading/trailing whitespace from text
-- (does not strip leading/trailing whitespace from CDATA)
parser:parse(myxml,{stripWhitespace=true})

If you just want to see if it will parse your document correctly, you can simply do:

local SLAXML = require 'slaxml'
SLAXML:parse(myxml)

…which will cause SLAXML to use its built-in callbacks that print the results as they are seen.

If you simply want to build tables from your XML, you can alternatively:

local SLAXML = require 'slaxdom' -- also requires slaxml.lua; be sure to copy both files
local doc = SLAXML:dom(myxml)

The returned table is a 'document' composed of tables for elements, attributes, text nodes, comments, and processing instructions. See the following documentation for what each supports.

• Document - the root table returned from the SLAXML:dom() method.
  • doc.type : the string "document"
  • doc.name : the string "#doc"
  • doc.kids : an array table of child processing instructions, the root element, and comment nodes.
  • doc.root : the root element for the document
• Element
  • someEl.type : the string "element"
  • someEl.name : the string name of the element (without any namespace prefix)
  • someEl.nsURI : the namespace URI for this element; nil if no namespace is applied
  • someAttr.nsPrefix : the namespace prefix string; nil if no prefix is applied
  • someEl.attr : a table of attributes, indexed by name and index
    • local value = someEl.attr['attribute-name'] : any namespace prefix of the attribute is not part of the name
    • local someAttr = someEl.attr[1] : a single attribute table (see below); useful for iterating all attributes of an element, or for disambiguating attributes with the same name in different namespaces
  • someEl.kids : an array table of child elements, text nodes, comment nodes, and processing instructions
  • someEl.el : an array table of child elements only
  • someEl.parent : reference to the parent element or document table
• Attribute
  • someAttr.type : the string "attribute"
  • someAttr.name : the name of the attribute (without any namespace prefix)
  • someAttr.value : the string value of the attribute (with XML and numeric entities unescaped)
  • someAttr.nsURI : the namespace URI for the attribute; nil if no namespace is applied
  • someAttr.nsPrefix : the namespace prefix string; nil if no prefix is applied
  • someAttr.parent : reference to the owning element table
• Text - for both CDATA and normal text nodes
  • someText.type : the string "text"
  • someText.name : the string "#text"
  • someText.cdata : true if the text was from a CDATA block
  • someText.value : the string content of the text node (with XML and numeric entities unescaped for non-CDATA elements)
  • someText.parent : reference to the parent element table
• Comment
  • someComment.type : the string "comment"
  • someComment.name : the string "#comment"
  • someComment.value : the string content of the attribute
  • someComment.parent : reference to the parent element or document table
• Processing Instruction
  • somePI.type : the string "pi"
  • somePI.name : the string name of the PI, e.g. <?foo …?> has a name of "foo"
  • somePI.value : the string content of the PI, i.e. everything but the name
  • somePI.parent : reference to the parent element or document table

The following function can be used to calculate the "inner text" for an element:

function elementText(el)
  local pieces = {}
  for _,n in ipairs(el.kids) do
    if n.type=='element' then pieces[#pieces+1] = elementText(n)
    elseif n.type=='text' then pieces[#pieces+1] = n.value
    end
  end
  return table.concat(pieces)
end

local xml  = [[<p>Hello <em>you crazy <b>World</b></em>!</p>]]
local para = SLAXML:dom(xml).root
print(elementText(para)) --> "Hello you crazy World!"

If you want the DOM tables to be easier to inspect you can supply the simple option via:

local dom = SLAXML:dom(myXML,{ simple=true })

In this case the document will have no root property, no table will have a parent property, elements will not have the el collection, and the attr collection will be a simple array (without values accessible directly via attribute name). In short, the output will be a strict hierarchy with no internal references to other tables, and all data represented in exactly one spot.

You can serialize any DOM table to an XML string by passing it to the SLAXML:xml() method:

local SLAXML = require 'slaxdom'
local doc = SLAXML:dom(myxml)
-- ...modify the document...
local xml = SLAXML:xml(doc)

The xml() method takes an optional table of options as its second argument:

local xml = SLAXML:xml(doc,{
  indent = 2,    -- each pi/comment/element/text node on its own line, indented by this many spaces
  indent = '\t', -- ...or, supply a custom string to use for indentation
  sort   = true, -- sort attributes by name, with no-namespace attributes coming first
  omit   = {...} -- an array of namespace URIs; removes elements and attributes in these namespaces
})

When using the indent option, you likely want to ensure that you parsed your DOM using the stripWhitespace option. This will prevent you from having whitespace text nodes between elements that are then placed on their own indented line.

Some examples showing the serialization options:

local xml = [[
<!-- a simple document showing sorting and namespace culling -->
<r c="1" z="3" b="2" xmlns="uri1" xmlns:x="uri2" xmlns:a="uri3">
  <e a:foo="f" x:alpha="a" a:bar="b" alpha="y" beta="beta" />
  <a:wrap><f/></a:wrap>
</r>
]]

local dom = SLAXML:dom(xml, {stripWhitespace=true})

print(SLAXML:xml(dom))
--> <!-- a simple document showing sorting and namespace culling --><r c="1" z="3" b="2" xmlns="uri1" xmlns:x="uri2" xmlns:a="uri3"><e a:foo="f" x:alpha="a" a:bar="b" alpha="y" beta="beta"/><a:wrap><f/></a:wrap></r>

print(SLAXML:xml(dom, {indent=2}))
--> <!-- a simple document showing sorting and namespace culling -->
--> <r c="1" z="3" b="2" xmlns="uri1" xmlns:x="uri2" xmlns:a="uri3">
-->   <e a:foo="f" x:alpha="a" a:bar="b" alpha="y" beta="beta"/>
-->   <a:wrap>
-->     <f/>
-->   </a:wrap>
--> </r>

print(SLAXML:xml(dom.root.kids[2]))
--> <a:wrap><f/></a:wrap>
-- NOTE: you can serialize any DOM table node, not just documents

print(SLAXML:xml(dom.root.kids[1], {indent=2, sort=true}))
--> <e alpha="y" beta="beta" a:bar="b" a:foo="f" x:alpha="a"/>
-- NOTE: attributes with no namespace come first

print(SLAXML:xml(dom, {indent=2, omit={'uri3'}}))
--> <!-- a simple document showing sorting and namespace culling -->
--> <r c="1" z="3" b="2" xmlns="uri1" xmlns:x="uri2">
-->   <e x:alpha="a" alpha="y" beta="beta"/>
--> </r>
-- NOTE: Omitting a namespace omits:
--       * namespace declaration(s) for that space
--       * attributes prefixed for that namespace
--       * elements in that namespace, INCLUDING DESCENDANTS

print(SLAXML:xml(dom, {indent=2, omit={'uri3', 'uri2'}}))
--> <!-- a simple document showing sorting and namespace culling -->
--> <r c="1" z="3" b="2" xmlns="uri1">
-->   <e alpha="y" beta="beta"/>
--> </r>

print(SLAXML:xml(dom, {indent=2, omit={'uri1'}}))
--> <!-- a simple document showing sorting and namespace culling -->
-- NOTE: Omitting namespace for the root element removes everything

Serialization of elements and attributes ignores the nsURI property in favor of the nsPrefix attribute. As such, you can construct DOM's that serialize to invalid XML:

local el = {
  type="element",
  nsPrefix="oops", name="root",
  attr={
    {type="attribute", name="xmlns:nope", value="myuri"},
    {type="attribute", nsPrefix="x", name="wow", value="myuri"}
  }
}
print( SLAXML:xml(el) )
--> <oops:root xmlns:nope="myuri" x:wow="myuri"/>

So, if you want to use a foo prefix on an element or attribute, be sure to add an appropriate xmlns:foo attribute defining that namespace on an ancestor element.

• Does not require or enforce well-formed XML. Certain syntax errors are silently ignored and consumed. For example:
  • foo="yes & no" is seen as a valid attribute
  • <foo></bar> invokes startElement("foo") followed by closeElement("bar")
  • <foo> 5 < 6 </foo> is seen as valid text contents
• No support for custom entity expansion other than the standard XML entities (< > " ' &) and numeric entities (e.g. 
 or <)
• XML Declarations (<?xml version="1.x"?>) are incorrectly reported as Processing Instructions
• No support for DTDs
• No support for extended (Unicode) characters in element/attribute names
• No support for charset
• No support for XInclude
• Does not ensure that the reserved xml prefix is never redefined to an illegal namespace
• Does not ensure that the reserved xmlns prefix is never used as an element prefix

• Updating rockspec to prepare for LuaRocks

• Adds SLAXML:xml() to serialize the DOM back to XML.
• Adds nsPrefix properties to the DOM tables for elements and attributes (needed for round-trip serialization)
• Fixes test suite to work on Lua 5.2, 5.3.
• Fixes Issue #10, allowing DOM parser to handle comments/PIs after the root element.
• Fixes Issue #11, causing DOM parser to preserve whitespace text nodes on the document.
• Backwards-incompatible change: Removes doc.root key from DOM when simple=true is specified.

• Decodes entities above 127 as UTF8 (decimal and hexadecimal).
  • The encoding specified by the document is (still) ignored. If you parse an XML file encoded in some other format, that intermixes 'raw' high-byte characters with high-byte entities, the result will be a broken encoding.

• Fixes Issue #6, adding support for ASCII hexadecimal entities (e.g. <). (Thanks Leorex/Ben Bishop)

• Fixes Issue #5 (and more): Namespace prefixes defined on element are now properly applied to the element itself and any attributes using them when the definitions appear later in source than the prefix usage. (Thanks Oliver Kroth.)
• The streaming parser now supplies the namespace prefix for elements and attributes.

• Fixes Issue #3: The reserved xml prefix may be used without pre-declaring it. (Thanks David Durkee.)

• Lua 5.2 compatible
• Parser now errors if it finishes without finding a root element, or if there are unclosed elements at the end. (Proper element pairing is not enforced by the parser, but is—as in previous releases—enforced by the DOM builder.)

• <foo xmlns="bar"> now directly generates startElement("foo","bar") with no post callback for namespace required.

• Use the local SLAXML=require 'slaxml' pattern to prevent any pollution of the global namespace.

• Bugfix to allow empty attributes, i.e. foo=""
• closeElement no longer includes namespace prefix in the name, includes the nsURI

• DOM adds .parent references
• SLAXML.ignoreWhitespace is now :parse(xml,{stripWhitespace=true})
• "simple" mode for DOM parsing

• Support namespaces for elements and attributes
  • <foo xmlns="barURI"> will call startElement("foo",nil) followed by namespace("barURI") (and then attribute("xmlns","barURI",nil)); you must apply the namespace to your element after creation.
  • Child elements without a namespace prefix that inherit a namespace will receive startElement("child","barURI")
  • <xy:foo> will call startElement("foo","uri-for-xy")
  • <foo xy:bar="yay"> will call attribute("bar","yay","uri-for-xy")
  • Runtime errors are generated for any namespace prefix that cannot be resolved
• Add (optional) DOM parser that validates hierarchy and supports namespaces

• Supports expanding numeric entities e.g. " -> "
• Utility functions are local to parsing (not spamming the global namespace)

• Option to ignore whitespace-only text nodes
• Supports unescaped > in attributes
• Supports CDATA
• Supports Comments
• Supports Processing Instructions

Copyright © 2013 Gavin Kistner

Licensed under the MIT License. See LICENSE.txt for more details.