RE: [xsl] XPath for matching multiple child elements

Subject: RE: [xsl] XPath for matching multiple child elements
From: Wendell Piez <wapiez@xxxxxxxxxxxxxxxx>
Date: Fri, 29 Sep 2006 09:46:00 -0400
Hi Mike,

(Of course, anyone else with insight or perspective is welcome to contribute.)

At 05:48 PM 9/28/2006, you wrote:
> But it's free enough to support a bit of a workaround, even
> if you have no schema:
>
> <xsl:variable name="comment-elements"
> select="//comments//(h1|p|b|i)"/>
>
> <xsl:template match="*[exists(. intersect $comment-elements)]">
>     ...
> </xsl:template>
>
> I think this should work.

Yes, but it might have fairly horrible performance.

I don't know how other processors work, but with Saxon, patterns are a lot
more efficient if they identify the element names that match, because Saxon
basically does a hash lookup on the node kind and then on the node name. Any
match="*" pattern gets tested against every node; and in this case that will
involve a serial search through the list of h1|p|b|i elements, which is
distinctly O(n^2). Not worth the savings in keystrokes.

Ah, this is good to know. I suppose one could say "comments//*[etc]" to alleviate this a bit (though not changing the scaling issue).


I'm mostly running what are considered by many to be small documents, so I'm probably not sensitive to this sort of thing. (In my world, 0.5MB/UTF-8 is "quite large" and most documents are under 100KB.) I almost never see a performance bog-down in XSLT.

I suppose this would also be true of David's

match="comments//*[self::e1|self::e2|self::e3]"

... which I like even better than mine?

And while we're at it, could you comment on the performance implications of a near-identity first pass in a pipeline (in which, if you liked, you could rename the elements)? I guess this would generally tend simply to double the time of the transform? In sum, is your assessment that none of these tricks for maintainability (which all involve their tradeoffs) are worth it, and one is just better off biting the bullet?

Cheers,
Wendell

Current Thread