Document.evaluate()

var xpathResult = document.evaluate(
 xpathExpression,
 contextNode,
 namespaceResolver,
 resultType,
 result
);

• xpathExpression - строка, описывающая XPath, который должен быть исполнен.
• contextNode указываетконтекстный узел для запроса (см. [http://www.w3.org/TR/xpath спецификация XPath). В качестве данного аргумента может быть задан объект document.
• namespaceResolver - функция, которой будут переданы все префиксы пространств имён. Она должна возвращать строку, описывающую URI, ассоциированный с данным префиксом. It will be used to resolve prefixes within the XPath itself, so that they can be matched with the document. null is common for HTML documents or when no namespace prefixes are used.
• resultType - число, описывающее тип возвращаемого XPathResult (см. ниже). Используйте именные свойства-константы конструктора класса XPathResult (эквивалентно численным значениям от 0 до 9), как например XPathResult.ANY_TYPE.
• result - экземпляр объекта типа XPathResult, используемого для хранения результатов поиска по данному xpathExpression. Может принимать значение null

var headings = document.evaluate(
  "/html/body//h2",
  document,
  null,
  XPathResult.ANY_TYPE,
  null,
);
/* Найти в документе все элементы h2
 * В качестве результата будет получен узловой итератор. */
var thisHeading = headings.iterateNext();
var alertText = "В данном документе заголовками 2-го уровня являются:\n";
while (thisHeading) {
  alertText += thisHeading.textContent + "\n";
  thisHeading = headings.iterateNext();
}
alert(alertText); // Показывает alert со всеми найденными элементами h2

Note, in the above example, a more verbose XPath is preferred over common shortcuts such as //h2. Generally, more specific XPath selectors as in the above example usually gives a significant performance improvement, especially on very large documents. This is because the evaluation of the query spends does not waste time visiting unnecessary nodes. Using // is generally slow as it visits every node from the root and all subnodes looking for possible matches.

Further optimization can be achieved by careful use of the context parameter. For example, if you know the content you are looking for is somewhere inside the body tag, you can use this:

document.evaluate(".//h2", document.body, null, XPathResult.ANY_TYPE, null);

Notice in the above document.body has been used as the context instead of document so the XPath starts from the body element. (In this example, the "." is important to indicate that the querying should start from the context node, document.body. If the "." was left out (leaving //h2) the query would start from the root node (html) which would be more wasteful.)

Более детально данный материал описан в статье Introduction to using XPath in JavaScript (en-US).

• Выражения XPath могут быть интерпретированы в HTML- и XML-документах.
• While using document.evaluate() works in FF2, in FF3 one must use someXMLDoc.evaluate() if evaluating against something other than the current document.

These are supported values for the resultType parameter of the evaluate method:

Results of NODE_ITERATOR types contain references to nodes in the document. Modifying a node will invalidate the iterator. After modifying a node, attempting to iterate through the results will result in an error.

Results of NODE_SNAPSHOT types are snapshots, which are essentially lists of matched nodes. You can make changes to the document by altering snapshot nodes. Modifying the document doesn't invalidate the snapshot; however, if the document is changed, the snapshot may not correspond to the current state of the document, since nodes may have moved, been changed, added, or removed.

• DOM:document.createExpression (en-US)
• XPath Code Snippets
• Check for browser support