Hi there,

I need to set up all the element range indexes to test the query.  Once I get
through I'll let you know .

Thanks, Helen

>>> Geert Josten <[email protected]> 04/07/10 11:09 AM >>>
Hi David,

> My *guess* is not that Helen is against using
> distinct-values() but that she's discovered, as all of us
> have, that using "pure xquery" in ML on large datasets often
> doesn't optimize as well as we like.

That wasn't my 'guess' either, but let us wait for Helen to comment herself..

> The solutions to this are various and will require more
> detail knowledge about the exact makeup of the dataset, such
> as how large, how many documents, if its fragmented or not, etc.

Yes.

> I've found generally that using cts:search() or related
> things is required in order to get decent performance out of
> ML, although sometimes (magically to me) I've stumbled on
> 'pure xquery' or 'pure xpath' expressions that happen to
> optimize well but I've never found a good way to know for
> sure without deep analysis of the specific query and lots of
> trial & error.

Yes, using cts or search library is a more certain way to rely on indexes, as
you address them explicitly.

> As for pre-processing.   Good question about performance against
> updates.
> This really comes down to how often are you updating vs how
> often do you query ?
> If your going to always need to create a tree version of the
> entire dataset its worth doing that upfront (IMHO), but if
> its occasional and you only need a tiny subset of the data
> turned into a tree at any time, especially if your updates
> are frequent compared to the queries then "on the fly" might be best.
> And there is always hybrid approaches like using a "cache" of
> the tree-formatted data and regenerating it only on first hit.

It could indeed be worth doing it upfront, but you still want to calculation to
be quick. Otherwise the task server might get too busy keeping this
precalculated data up to date (if there happen to be a lot of updates
simultaniously). From the looks of it, it should be possible to fully rely on
indexes, making it quick enough to perform 'on the fly'..

Kind regards,
Geert


drs. G.P.H. (Geert) Josten
Consultant


Daidalos BV
Hoekeindsehof 1-4
2665 JZ Bleiswijk

T +31 (0)10 850 1200
F +31 (0)10 850 1199

mailto:[email protected]
http://www.daidalos.nl/

KvK 27164984

P Please consider the environment before printing this mail.
De informatie - verzonden in of met dit e-mailbericht - is afkomstig van
Daidalos BV en is uitsluitend bestemd voor de geadresseerde. Indien u dit
bericht onbedoeld hebt ontvangen, verzoeken wij u het te verwijderen. Aan dit
bericht kunnen geen rechten worden ontleend.

_______________________________________________
General mailing list
[email protected]
http://xqzone.com/mailman/listinfo/general

_______________________________________________
General mailing list
[email protected]
http://xqzone.com/mailman/listinfo/general

Reply via email to