Re: [xml] Minor documentation path for libxml/xpath.h

["James Dennett" <jdennett technocom-wireless com>]

> -----Original Message-----
> From: Daniel Veillard [mailto:veillard redhat com]
> Sent: Wednesday, March 21, 2007 9:51 AM
> To: James Dennett
> Cc: xml gnome org
> Subject: Re: [xml] Minor documentation path for libxml/xpath.h
> 
> On Wed, Mar 21, 2007 at 09:16:40AM -0700, James Dennett wrote:
> > ... the output is attached to this message as xpathpatch.txt 
> > ...  If this is still the wrong format, let me know.
> 
>   Nah, that worked just fine, applied and commited, thanks a lot !

Taking a little time to prepare patches is a small price
for the value I/we get from libxml2!

[snip]

> > It's not that I *want* to modify the context.  It's that certain
libxml2
> > functions do so in ways that may not be documented -- in particular,
I
> > noted that evaluating an expression using a context appears to
change
> > (at least) the current node within that context.  I might change my
> > wrapper to restore the node after the expression has been evaluated,
but
> > without documenting the contract of the libxml2 functions I can't
know
> > that such behavior is sufficient or appropriate.
> >
> > (Our current workaround, and probably a good idea in any case, is
for
> > our XPath expressions to be absolute ones.)
> 
>   Right evaluation may change the node (and possibly the doc but only
in
> an XSLT context so it should not affect you). In any case if you have
> relative expressions, then you definitely need to set up the node
before
> calling the evaluation. Other items which influence evaluation are
> namespaces/nsNr the context namespaces, here/origin used in XPointer,
> registered functions and the dictionnary if any . But those should not
> be modified by evaluation.

OK, perfect; I'll make our code set the node, but also caution in its
documentation against using relative expressions in any case -- there
should be no good reason for using relative expressions when there's no
clearly-defined node for them to be relative *to*.

-- James