Compare two XML strings ignoring element order

My original answer is outdated. If I would have to build it again i would use xmlunit 2 and xmlunit-matchers. Please note that for xml unit a different order is always 'similar' not equals.

@Test
public void testXmlUnit() {
    String myControlXML = "<test><elem>a</elem><elem>b</elem></test>";
    String expected = "<test><elem>b</elem><elem>a</elem></test>";
    assertThat(myControlXML, isSimilarTo(expected)
            .withNodeMatcher(new DefaultNodeMatcher(ElementSelectors.byNameAndText)));
    //In case you wan't to ignore whitespaces add ignoreWhitespace().normalizeWhitespace()
    assertThat(myControlXML, isSimilarTo(expected)
        .ignoreWhitespace()
        .normalizeWhitespace()
        .withNodeMatcher(new DefaultNodeMatcher(ElementSelectors.byNameAndText)));
}  

If somebody still want't to use a pure java implementation here it is. This implementation extracts the content from xml and compares the list ignoring order.

public static Document loadXMLFromString(String xml) throws Exception {
    DocumentBuilderFactory factory = DocumentBuilderFactory.newInstance();
    DocumentBuilder builder = factory.newDocumentBuilder();
    InputSource is = new InputSource(new StringReader(xml));
    return builder.parse(is);
}

@Test
public void test() throws Exception {
    Document doc = loadXMLFromString("<test>\n" +
            "  <elem>b</elem>\n" +
            "  <elem>a</elem>\n" +
            "</test>");
    XPathFactory xPathfactory = XPathFactory.newInstance();
    XPath xpath = xPathfactory.newXPath();
    XPathExpression expr = xpath.compile("//test//elem");
    NodeList all = (NodeList) expr.evaluate(doc, XPathConstants.NODESET);
    List<String> values = new ArrayList<>();
    if (all != null && all.getLength() > 0) {
        for (int i = 0; i < all.getLength(); i++) {
            values.add(all.item(i).getTextContent());
        }
    }
    Set<String> expected = new HashSet<>(Arrays.asList("a", "b"));
    assertThat("List equality without order",
            values, containsInAnyOrder(expected.toArray()));
}

XMLUnit will do what you want, but you have to specify the elementQualifier. With no elementQualifier specified it will only compare the nodes in the same position.

For your example you want an ElementNameAndTextQualifer, this considers a node similar if one exists that matches the element name and it's text value, something like :

Diff diff = new Diff(control, toTest);
// we don't care about ordering
diff.overrideElementQualifier(new ElementNameAndTextQualifier());
XMLAssert.assertXMLEqual(diff, true);

You can read more about it here: http://xmlunit.sourceforge.net/userguide/html/ar01s03.html#ElementQualifier

For xmlunit 2.0 (I was looking for this) it is now done, by using DefaultNodeMatcher

Diff diff = Diffbuilder.compare(Input.fromFile(control))
   .withTest(Input.fromFile(test))
   .withNodeMatcher(new DefaultNodeMatcher(ElementSelectors.byNameAndText))
   .build()

Hope this helps this helps other people googling...