XPath subtree pattern matching

[ahogue at theory dot lcs dot mit dot edu]

Hello -

Is there any way to match complex subtree patterns with XPath? The functions I see all seem to match along a single path from root to leaf. I would like to match full subtrees.

For example, given the XHTML:

<html>
<body>
<p>
<a>#text</a>
<br/>
#text
<b>#text</b>
#text
<br/>
<font>
<a>#text</a>
</font>
</p>
<p>
<a>#text</a>
<br/>
#text
<br/>
<font>
<a>#text</a>
</font>
</p>
</body>
</html>

I would like to construct a "pattern" using XPath to match all subtrees like:

<p>
<a>*</a>
<br/>
*
(<b>*</b>)?
(*)?
<br/>
<font>
<a>*</a>
</font>
</p>

where the "*" means that any text can be matched, and the "?" means that 0 or 1 instances of the item may be matched, similar to a regular expression.

Is there an easy way to do this kind of "subtree pattern matching" in XPath? Would I be better off writing a wrapper over XPath and using several XPath queries to represent and retreive my pattern?

Thanks in advance,

Andrew Hogue

 

[Sebastian Schaffert]

"ahogue at theory dot lcs dot mit dot edu" <"ahogue at theory dot lcs dot

Hello -

Is there any way to match complex subtree patterns with XPath? The functions I see all seem to match along a single path from root to leaf. I would like to match full subtrees.

XPath is basically a tree language, not a path language, so you *can* specify tree patterns. This is usually done by using qualifiers. To match e.g.

<f>
<a/>
<b>Text</b>
<c>Other Text</c>
</f>

and select "Text", an XPath expression could be used as follows:
f/b[preceding-sibling::a][following-sibling::c]

However, Tree matching in XPath has two restrictions:
1. It is not "nice", since you basically encode the tree in a linear representation which is not straightforward, as it does not resemble the XML document
2. It is not possible to select content at several positions (e.g. "Text" and "Other Text" together)

I don't want to make too much advertisement again, but you might want to have a look at http://www.xcerpt.org if you want to have a look at a language with "real" tree patterns.

 

[Dimitre Novatchev]

As easy as:

node[count(ancestor-or-self::someNode | theRoot-someNode) = count(ancestor-or-self::someNode )
]

This matches all nodes of the tree with root theRoot-someNode, which is a specific "someNode" element.

In case we want simply to select all nodes of a given tree, we can use the following simpler XPath expression, which is not a match pattern, because the location steps (not the predicates) of a match pattern may only contain the child and attribute axis:

theRoot-someNode//descendant-or-self::node

This selects all nodes of the tree with root a "theRoot-someNode" element.

=====
Cheers,

Dimitre Novatchev.
http://fxsl.sourceforge.net/ -- the home of FXSL

"ahogue at theory dot lcs dot mit dot edu" <"ahogue at theory dot lcs dot mit dot edu"> wrote: