Here are some examples of how to use Emacs’s XML parser (xml.el). Other parsers are available: see XmlParsers.

Parse a simple XML string

How to parse a string containing the following fragment?

<post time="20050716234509" id="010101">
  <msg>Here is the message</msg>
  <info>My UA</info>

The parser will return a list of nodes. The post node is the only element in that list. The datastructure of the post node itself is very simple: symbol, alist of attributes, followed by child nodes. There are dedicated accessors for this information, however. We only need to remember when we’re getting at attribute values, for example.

Text nodes are simple strings. xml-get-children returns a list of children. Tag names are symbols. That should be enough to understand the following code:

(let* ((xml "<post time=\"20050716234509\" id=\"010101\">
               <msg>Here is the message</msg>
               <info>My UA</info>
       (root (with-temp-buffer
	       (insert xml)
	       (xml-parse-region (point-min) (point-max))))
       (post (car root))
       (attrs (xml-node-attributes post))
       (time (cdr (assq 'time attrs)))
       (msg (car (xml-get-children post 'msg)))
       (text (car (xml-node-children msg))))
  (message "time: %s, message '%s'" time text))

You can also walk the tree converting nodes to sexp’s. The following function transforms an xml document into a tree-widget using tree-widget.el. This could form the basis of a really cool xml editing interface for emacs.

(defun xml->tree-widget (root)
  (cond ((null root) nil)
	((listp root) (let ((elem (xml-node-name root))
			    (children (remove-if (function stringp) (xml-node-children root))))
			`(tree-widget :node (push-button
					     :tag ,(format "%s" elem)
					     :format "%[%t%]\n"
					     :xml-node ,root
					     :notify ,(lambda (widget &rest rest)
                                                        (message (format "%s" (widget-get widget :xml-node)))))
				      ,@(mapcar (lambda (x) (xml->tree-widget x)) children))))))

Parse an HTML file, walking over the nodes and filtering them

I wanted to prepare a list of national days and used the following web page as the source:

The data looked as follows:

        <table id="calendar" border="2" width="100%" cellspacing="0" cellpadding="0">
              <th align="left"/>
              <th align="left"/>
              <th align="left"/>
              <td valign="top">
                <table border="0">
                  <tr class="month">
                      <p class="monthName">January</p>
                      <p class="dayOfMonth">1</p>
                        <dt class="country">Cuba</dt>
                        <dd class="dayTitle">Liberation Day</dd>
                        <dt class="country">Haiti</dt>
                        <dd class="dayTitle">Independence Day</dd>
                        <dt class="country">Sudan</dt>
                        <dd class="dayTitle">National Day</dd>

And here is the code. When the node is a string, then it is not processed. When the node has the appropriate attribute, the first of its children (the string) is stored. When the node does not have the appropriate attribute, the function is called recursively for all children. The results are accumulated whenever the last piece of information is found.

    (defun walk (node)
      (when (listp node)
	(cond ((string= "monthName" (xml-get-attribute node 'class))
	       (setq m (first (xml-node-children node) )))
	      ((string= "dayOfMonth" (xml-get-attribute node 'class))
	       (setq d (first (xml-node-children node) )))
	      ((string= "country" (xml-get-attribute node 'class))
	       (setq c (first (xml-node-children node) )))
	      ((string= "dayTitle" (xml-get-attribute node 'class))
	       (setq f (first (xml-node-children node) ))
	       (setq result (cons (list m d c f) result)))
	       (mapc 'walk (xml-node-children node))))))
    (let (d m c f result
          (days (xml-parse-file "data.xml"))
      (walk (car days))
      (nreverse result))

And the result is something like this:

    (("January" "1" "Cuba" "Liberation Day")
     ("January" "1" "Haiti" "Independence Day")
     ("January" "1" "Sudan" "National Day")

Converting XML to XPath

This code helped me replace XHTML code with XPath statements in some unit tests.

The problem was that using a string match for HTML such as the following was brittle:

<a class="inter number" href="http://www.emacswiki.org/cgi-bin/oddmuse.pl?Link_Pattern"><span><span class="bracket">[</span>1<span class="bracket">]</span></span></a>

The attributes can come in any order. The code translates the above piece of XML into the following XPath statement:

//a[@class="inter number"][@href="http://www.emacswiki.org/cgi-bin/oddmuse.pl?Link_Pattern"]/span/span[@class="bracket"][text()="["]/following-sibling::text()[string()="1"]/following-sibling::span[@class="bracket"][text()="]"]

And here’s the code:

(require 'xml)

(defun wiki-attribute (beg end)
  (interactive "r")
  (let ((root (xml-parse-region beg end)))
    (delete-region beg end)
    (insert "//" (xpath-element (car root)))))

(defun xpath-element (elem)
  (if (stringp elem)
      (concat "text()[string()=\"" elem "\"]")
    (concat (xpath-element-name elem)
	    (xpath-element-attributes elem)
	    (xpath-element-children elem))))

(defun xpath-element-children (elem)
  (let ((children (xml-node-children elem)))
    (cond ((null children)
	  ((and (= (length children) 1)
		(stringp (car children)))
	   (concat "[text()=\"" (car children) "\"]"))
	   (concat "/" (mapconcat 'xpath-element

(defun xpath-element-name (elem)
  (symbol-name (xml-node-name elem)))

(defun xpath-element-attributes (elem)
  (mapconcat (lambda (attr)
	       (concat "[@" (symbol-name (car attr))
		       "=\"" (cdr attr) "\"]"))
	     (xml-node-attributes elem)

Format of lists

Note: The functions ‘xml-parse-file’, ‘xml-parse-region’ and ‘xml-parse-tag’ return a list with the following format:

    xml-list   ::= (node node ...)
    node       ::= (qname attribute-list . child_node_list)
    child_node_list ::= child_node child_node ...
    child_node ::= node | string
    qname      ::= (:namespace-uri . "name") | "name"
    attribute_list ::= ((qname . "value") (qname . "value") ...)
                       | nil
    string     ::= "..."

CategoryXML CategoryCode