Root namespace without prefix breaks xpath

[lukaseder]

XML documents that have a root xmlns with no prefix cannot be parsed with xpath.

Here is a unit test that produces this problem:

    @Test
    public void testNamespace() throws Exception {
        String xml = "<root xmlns:foo=\"http://foo.com\"><bar>baz</bar></root>";
        // This passes
        assertEquals("baz", $(xml).namespace("foo", "http://foo.com").xpath("//root/bar").content());

        // Now remove the prefix
        xml = "<root xmlns=\"http://foo.com\"><bar>baz</bar></root>";
        // But this fails
        assertEquals("baz", $(xml).xpath("//root/bar").content());
        // This fails too
        assertEquals("baz", $(xml).namespace("", "http://foo.com").xpath("//root/bar").content());
    }

It should be possible to use xpath on documents with a root xmlns declaration with no prefix, but it just doesn't work.

Joox 1.2.0 on OSX.

---

Originally reported here:
https://code.google.com/p/joox/issues/detail?id=121

[lukaseder]

Unfortunately, this is not how namespaces work. Just because your <root/> element uses a default namespace, this doesn't mean that there is automatically a binding for a default namespace in your XPath expression. For example:

        // But this fails
        assertEquals("baz", $(xml).xpath("//root/bar").content());

This can't work because you're querying "unnamespaced" root and bar elements. Imagine the default namespace was set on the bar element, rather than the root element:

<root><bar xmlns="http://foo.com">baz</bar></root>

This:

        // This fails too
        assertEquals("baz", $(xml).namespace("", "http://foo.com").xpath("//root/bar").content());

The correct solution here would be to express the following XPath:

//:root/:bar

The best idea would be to bind explicit namespaces for your XPath:

assertEquals("baz", $(xml).namespace("x", "http://foo.com").xpath("//x:root/x:bar").content());

[lukaseder]

Note, we also have #102 in our roadmap to simply remove namespaces from documents. Often, this will make queries much more easy without necessarily changing semantics

[breedx2]

Oooh, thanks for the clarification. This was totally my fault in not understanding. Thanks for taking the time to explain and point at #102 .

[lukaseder]

@breedx2 : You're welcome. No worries, I constantly have to review namespaces myself. The rule of thumb is that once you add namespaces to your document, everything changes drastically