RE: [xsl] Best performant way to selectively find and process one node

Subject: RE: [xsl] Best performant way to selectively find and process one node
From: "Michael Kay" <mike@xxxxxxxxxxxx>
Date: Mon, 21 May 2007 22:48:52 +0100
> One (IMO) non performant way of doing it is to have a 
> template that matches books, select book with the correct id 
> and pass it for processing to named template.

That sounds almost like a description of the following code:

<xsl:template match="books">
  <xsl:apply-templates select="book[id=$bookId]"/>
</xsl:template>

(except that I use a match template rather than a named template, because
it's more concise)

Why do you assume this is non-performant? Are you just guessing, or have you
made measurements? I would expect this to perform pretty well.

It depends of course on the processor. It's impossible to make reliable
statements about performance without knowing what processor you are talking
about.

> 
> Ideally, i would like to be able to do something like 
> <xslt:template match="bookstore/books/book[id=$book_id]">
>    <xslt:apply templates/>
> </xslt:template>

You can do that in XSLT 2.0, but I would expect that on most processors the
version given above will perform better. However, that's guesswork, and I've
just advised you against making guesses...

If you are performing repeated queries of the same kind on the same
document, then using keys will probably perform better, as most
implementations are likely to construct an index. But if you're only doing
one query, then building an index is an unnecessary cost.

Michael Kay
http://www.saxonica.com/

Current Thread