You can query each node in the selectNodes() nodelist one at a time, but not 
the nodelist as a whole. 

To be honest, I think XSLT would be faster.  I've recently played around 
with a fairly large XML document (several MB), and tried several methods of 
processing nodes.  One selects a nodelist and passes it as a parameter to an 
XSLT, which then processes the nodelist.  This is quite a bit faster than 
iterating over the same nodelist using DOM methods.

A faster method passed the whole DOM to the XSLT, and I set several 
parameters, using xslProc.addParameter() with the filter criteria.  This 
worked fairly well even on large numbers of nodes.

The fastest method I found was to pass the whole DOM to the XSLT, and then 
pass a query string as a parameter to the XSLT.  The query string was parsed 
in a recursive template inside, and the results used to filter the original 
XML.  I was quite surprised how fast this was.

If the end result of this process is to generate an HTML page, then I'd 
*definitely* use XSLT for this.  Even if the goal is to process a set of 
nodes, and produce some other XML document, it's probably faster to use an 
identity transformation as your starting point, and do the filtering in the 
XSLT.

What platform are you doing this in?  Is this in Javascript in an InfoPath 
document?  

Regards,
Mike Sharp


"Greg Collins [InfoPath MVP]" wrote:

> I did a search on the newsgroup and couldn't find an answer to this...
> 
> I have a fairly large DOM that I'm am doing a series of 6 complex selectNodes on. It takes quites a number of seconds to complete.
> 
> Only 2 of the 6 calls need to be over the entire DOM. The next set only needs to query the original set of selected nodes, and the third set only needs to query the second query.
> 
> In other words, I need to do:
> 
> 1. Query entire DOM
> 2. Query results of #1.
> 3. Query results of #2.
> 
> 4. Query entire DOM.
> 5. Query results of #4.
> 6. Query results of #5.
> 
> I could cut out quite a bit of time if I were able to do a selectNodes on the previously selected Nodes. As far as I can tell this is not possible. It could also be faster if I could convert the results of a selectNodes into a document that I could then, once again, do a selectNodes on.
> 
> Are either of these possible without looping through some code?
> 
> -- 
> Greg Collins [InfoPath MVP]
> Please visit: http://www.InfoPathDev.com
> 
> 
> 
>