Re: [xslt] XSLT patterns without transformations?

[Magnus Lie Hetland <magnus hetland org>]

Daniel Veillard <veillard redhat com>:
>
[snip]
> 
>   Well if I wrote pattern.c for libxslt and later migrated this to
> libxml2 it's really because they work differently and their cost/benefit
> are really different. Except taking a precompiled form of xpath.c
> expressions and trying to interpret it in a different way. Currently
> there is no code which seems to satify your needs, statement which
> itself is hard to assert as you didn't specify them.

Yes, I know -- I'm not 100% certain about what I need myself, yet. I
do think I need basic location paths and some form of negation (e.g.,
"nodes that are not code nodes"), which I guess is most easily done
using e.g., 'node()[not(self::code)]'.

> And pattern.[ch] in libxml2 is a work in progress which may or may
> not suit your needs in a distant future, but again since you didn't
> specify what you want...

Right... :)

As I suspect my needs for efficiency might not be all too extreme, I
have -- for now -- simply used xmlXPathEvalExpression(...) along with
one loop traversing the ancestor chain and one looking through the
returned node-set (if any). Sort of a "brute force"-application of
XPath matching to get the XSLT pattern semantics (or something
similar), but it seems to work, so... (Of course, this doesn't enforce
the restricted subset of XPath, but that may not be too critical.)

-- 
Magnus Lie Hetland       Fallen flower I see / Returning to its branch
http://hetland.org       Ah! a butterfly.           [Arakida Moritake]