namespaces and xpath queries

[yawnmoth]

Say I have the following XML file:

<parent xmlns="uri://domain.tld/"><child /></parent>

In order to query the child node I need to register the uri://domain.tld/
namespace with a prefix, like 'parent'. I then need to do '//
parentarent//parent:child'. My question is this: if one node is in
a namespace are all children to that node assumed to be in the same
name space? In that case it seems like it's redundant to have to
specify the namespace for each and every successive child. I mean, is
there even a way to construct a child to parentarent such that '//
parentarent//child' would return a node?

Also, it doesn't seem to me like there's really a lot of point in
assigning a namespace to the root node? I mean, since there can only
be one root node it's not as if disambiguation is necessary. For
<root><parent xmlns="uri://a.a" /><parent /></root>, you'd need a
namespace to differentiate the two parent tags, but for my initial XML
it seems totally unnecessary? And yet I frequently see XML documents
that do just this - place the root node in it's own namespace.

 

[Mayeul]

Say I have the following XML file:

<parent xmlns="uri://domain.tld/"><child /></parent>

In order to query the child node I need to register the uri://domain.tld/
namespace with a prefix, like 'parent'.

Yes, but you should really keep the prefix short. Like p or par.

I then need to do '//
parentarent//parent:child'. My question is this: if one node is in
a namespace are all children to that node assumed to be in the same
name space?

More or less. The rules are more precise than that.
XML provides two ways to declare an element's namespace:
- default namespace
- prefixed namespace

Default namespace is declared as such:
xmlns="uri://domain.tld/"
Any element without prefix is in the default namespace.
Note that default namespaces cannot be declared for an XPath expression.

Prefixed namespaces are declared as such:
xmlns="uri://domain.tld/"
And an element is in this namespace if it is prefixed with p :
<parent xmlns="uri://domain.tld/">
<p:child />
</parent>

Both default namespace and prefixed namespaces are inherited by descendants.

Therefore, it is possible to declare this:
<parent xmlns="uri://domain.tld/" xmlns:c="http://anotheruri.tld">
<c:child>
<grandchild/>
</c:child>
</parent>

Here, <grandchild> is in the same namespace as <parent>, because they're
both in the declared default namespace.

<child> is in another namespace though, even though it is a descendant

In that case it seems like it's redundant to have to
specify the namespace for each and every successive child.

In practice it is, yes. Better keep your prefixes short.

I mean, is
there even a way to construct a child to parentarent such that '//
parentarent//child' would return a node?

Yes:
<parentarent xmlnsarent="uri://domain.tld/">
<child/>
</parentarent>

Here <child> is in the "no namespace" namespace.

Equivalent example:
<parent xmlns="uri://domain.tld/">

Also, it doesn't seem to me like there's really a lot of point in
assigning a namespace to the root node? I mean, since there can only
be one root node it's not as if disambiguation is necessary. For
<root><parent xmlns="uri://a.a" /><parent /></root>, you'd need a
namespace to differentiate the two parent tags, but for my initial XML
it seems totally unnecessary? And yet I frequently see XML documents
that do just this - place the root node in it's own namespace.

I guess I can't see a point to do that either.

However, I'd prefer the consistency: what sense would it make if every
elements is in a namespace but the root element is in no namespace?

Besides, the root element is a great place where to declare namespaces,
since they're inherited. So it would usually be in whatever default
namespace it declared, or on the contray, would need a prefix to *not*
be in *another* namespace than the default it declared.

 

[yawnmoth]

yawnmoth wrote:

Thanks - that was a very helpful response!

 

[Joe Kesselman]

Say I have the following XML file:

<parent xmlns="uri://domain.tld/"><child /></parent>

In order to query the child node I need to register the uri://domain.tld/
namespace with a prefix, like 'parent'. I then need to do '//
parentarent//parent:child'.

Yes, unfortunately XPath 1.0 does not have the concept of default
namespace, so your path must use prefixes to specify namespaced nodes.

> My question is this: if one node is in
a namespace are all children to that node assumed to be in the same
name space?

In general, namespace declarations are inherited to child elements until
another declaration overrides them. In specific, that includes the
default namespace declaration, which is what you've used here. So, yes,
your <child/> element is in the same namespace as <parent/>, because it
has the same prefix (none) and does not redeclare that prefix, and hence
is bound to the same namespace (the default)

> In that case it seems like it's redundant to have to
specify the namespace for each and every successive child.

See my first point. Sorry, but XPath 1.0 doesn't support that shortcut.

Also, it doesn't seem to me like there's really a lot of point in
assigning a namespace to the root node?

The namespace is part of the name, and the semantics, of the node. It
will affect how namespace-aware applications process the node, including
how it is validated against schemas, how it is processed by stylesheets,
and how other tools handle it. When there is *any* possiblity of
confusion about what kind of document you may be looking at, or simply
for code cleanliness reasons, putting the root node into a namespace
will be wise.

Generally, the namespace is related to a kind of document or data
structure -- a purchase order, for example, or a mailing address -- and
all elements (and occasionally some attributes) related to that kind of
data will use that namespace. That will be true whether the element
happens to be the root of this particular document or not. If you leave
off the namespace, you've changed the meaning of the element.

Also, it's often just as easy to do a few global namespace declarations
on the root element and have those prefixes (and/or that default)
available through the rest of the document via inheritance, rather than
having to repeatedly redeclare those bindings. As long as you're doing
them at the root anyway, putting the root element itself into the proper
namespace really doesn't add any significant amount of work.

--
Joe Kesselman,
http://www.love-song-productions.com/people/keshlam/index.html

{} ASCII Ribbon Campaign | "may'ron DaroQbe'chugh vaj bIrIQbej" --
/\ Stamp out HTML mail! | "Put down the squeezebox & nobody gets hurt."