Re: Pattern for attribute matching

["Paul W. Abrahams" <abrahams@xxxxxxxxxxx>]

> The XSL-List Digest      Saturday, July 31 1999      Volume 02 : Number 234
>
>
> Date: Fri, 30 Jul 1999 16:16:11 -0400
> From: "Paul W. Abrahams" <abrahams@xxxxxxxxxxx>
> Subject: Pattern for attribute matching
>
> If I want to construct a template that matches a `foo' element with a
> `bar' attribute, the appropriate `match' attribute is:
>
>     match="foo[@bar]"
>
> Moreover, if I want to match any `bar' attribute at all, the appropriate
> `match' attribute is
>
>    match="@bar"
>
> But suppose I want to match a `bar' attribute that's attached to a `foo'
> element.  My first guess would be
>
>   match="foo/@bar"
>
> But that doesn't seem entirely right because attributes are not children
> of element nodes, but rather associated with element nodes.  So if that
> isn't correct, what would be correct?
>
> This question relates to a larger unresolved (apparently) issue in
> XPath: whether attribute nodes are really part of the document tree
> described by XPath, or related in some other way.  Sec. 5 (Data Model)
> says:
>
>    ``XPath operates on a XML document as a tree.  This section descripe
> how XPath models an XML document as a tree.''
>
> But in Sec. 5.3 (Attribute Nodes) it says:
>
>    ``Each element node has an associated set of attribute nodes.''
>
> So where does reality lie?
>
> Paul Abrahams
> abrahams@xxxxxxx
>
> ------------------------------
>
> Date: Fri, 30 Jul 1999 22:24:03 +0100 (BST)
> From: David Carlisle <davidc@xxxxxxxxx>
> Subject: Re: Pattern for attribute matching
>
>   match="foo/@bar"
>
> But that doesn't seem entirely right because attributes are not children
>
> it is right. / does not mean `children' it means `next step' and at each
> step you can change axis along which to travel, here you have @ which is
> short for attribute::, and no specified axis is short for child:: so the
> above is really
>
> child::foo/attribute::bar
>
> which means get bar attributes of foo children.
>
> David
>
> ------------------------
>
> ------------------------------
>
> Date: Fri, 30 Jul 1999 23:41:12 +0100 (BST)
> From: David Carlisle <davidc@xxxxxxxxx>
> Subject: Re: Pattern for attribute matching
>
> > The first match finds foo elements that have a bar attribute.
> > The third match finds children of foo elements that have a bar attribute.
>
> No.
>
> the first finds foo elements that have are children of the current
> node, and have a bar attribute
> the second finds bar attributes of foo children of the current node.
>
> if this is in an apply-templates, then the current node in the matching
> template will be the foo element in the first case but an _attribute
> node_ in the second.
>
> David
>

My intent in the third pattern is to match certain `bar' attributes,
namely, all those in the document that
are associated with `foo' elements.  I guess that the pattern

    foo/@bar

really does that (except for the possible case of the root node being a
`foo').

But the other question remains: is an attribute node a child of its
element node or merely associated with
it?  And if it's merely associated, how can it be part of the document
tree?

Paul Abrahams


 XSL-List info and archive:  http://www.mulberrytech.com/xsl/xsl-list