Re: [xsl] Some more stuff about selecting unique elements

Subject: Re: [xsl] Some more stuff about selecting unique elements
From: Jeni Tennison <mail@xxxxxxxxxxxxxxxx>
Date: Tue, 10 Apr 2001 14:41:54 +0100
David C. wrote:
> shouldn't you be using preceding-sibling rather than using preceding
> and then checking on the id of the parent?

I don't think Joerg can - he's going down two levels in the XPath -
through level2 to 'stuff'.  Unless you mean:

  level2/stuff[not(. = ../preceding-sibling::level2/stuff)]

or perhaps

  level2/stuff[not(. = (.. | ../preceding-sibling::level2)/stuff)]

> and .//stuff would apparently be better as stuff (unless  your documents
> are more complicated than shown.)

or rather as level2/stuff

>> Well, the problem i dislike the preceding axis is performance,
> well yes, but preceding-sibling would be more fun, and in either case
> you could probably optimise using xsl:key if needed.

Joerg - you should look at the Muenchian Method, I think -

>> I suppose in XSLT 1.1 (2.0) where RTFs are gone
> I'll let Jeni tell you about that....

<sigh /> For those of you who weren't at the conference, it's official
- XSLT 1.1 has been put 'on hold' (which seems to be a euphemism along
the same lines as 'retired' or 'downsized'). The next update to XSLT
that we can expect is XSLT 2.0.

This makes the case for EXSLT even stronger. Come and join us on
list@xxxxxxxxx and/or look at the website at
We'll definitely be adding xsl:document to EXSLT - Common, and
probably xsl:script to EXSLT - Functions, so most of the cool stuff
in XSLT 1.1 should be covered by community extensions.

> All (?) existing systems let you query into a RTF using an explicit
> coersion to a node set (xt saxon. msxsl, xalan etc all have a
> node-set() function in theor extension namespace)

And of course EXSLT has one in the EXSLT namespace, see



Jeni Tennison

 XSL-List info and archive:

Current Thread