Re: [xslt] key() bug?

[Norman Walsh <ndw nwalsh com>]

/ Daniel Veillard <veillard@redhat.com> was heard to say:
| On Wed, Jan 07, 2004 at 10:12:50PM -0500, Norman Walsh wrote:
|> I believe the output of this stylesheet (with the same stylesheet as input)
|> should be:
|> 
|> Mode of FOO is input
|> Mode of FOO is fiddled
|> 
|> But xsltproc gives
|> 
|> Mode of FOO is input
|> Mode of FOO is input
|> 
|> Known bug?
|
|   keys on exslt:node-set() result, let's say that yes, it used to be
| an open issue, apparently this got confirmed by whoever maintains
| the EXSLT descriptions.
|   Bugzilla'ing it will avoid loosing track of it.

I tried to refresh my memory about the xsltproc bugzilla URI, but failed.
It's mentioned in passing on http://xmlsoft.org/XSLT/bugs.html, but no
URI is provided.

|   Implementing it will make running your stylesheets even slower,
| well I assume nobody cares anymore.

Maybe a lazy evaluation is possible: only build the keys if someone
attempts to use one?

The fact is that without keys, I have to resort to
//rng:define[@name=$name] in the real example I'm working on. Since I
do this many, many times and it is *a lot* slower than key(), my
stylesheet is currently taking tens of minutes if not longer to run
(with keys, it takes about 10 seconds). I'm willing to pay the
performance hit building a set keys at exslt:node-set() time.

                                        Be seeing you,
                                          norm

-- 
Norman Walsh <ndw@nwalsh.com> | An ill-humoured man is a prisoner at
http://nwalsh.com/            | the mercy of an enemy from whom he can
                              | never escape.-- Sa'di

PGP signature