[CPS-devel] Re: Lucene Product Suite

[Christian Klinger]

> 
> yes, this Zope code is terrible for everybody... I know. I'd advice to
> just add the methods you need on the CPSBrain but don't inherit from
> this code...
> 
>> I don´t have a svn repository so far. But if you like i can email the
>> package.
> 
> no was just asking by curiosity.
> 
>> Does CPS have any perfomace tests for a "Lucene based Catalog"?
> 
> We don't have strict benchmarks but our first customer with NXLucene has
> for the moment 750k documents with an index store  of almost 2Gb. We
> should reach the million in the following weeks.
> 
> Reindexing *from CPS* to NXLucene (reindexObject()), gives in between
> 15-20 documents / secs with a store of 100k documents.
> 
> This is really depending of what you are indexing. We have around 50
> fields including fulltext.
> 
> It's been enough for us so far.
> 
> 	J.

Hi agin,

The Problem with the CPSBrain was relative easy. The Class has to 
inherit from Acquisation.Implicit instead of Acquisation.Explicit.

class CPSBrain(Item,Acquisition.Implicit):
     """Simple light brain.      ********

Unfortunatly Plone use the Product ExtendedPathIndex for the path index 
in PortalCatalog. There are many little differences between the lucene 
path index and the ExtendedPathIndex. But i try to get things done.

I think the real Question for Performance is what is the 
break-even-point for using a lucene-based-catalog instead of 
zodb-based-catalog???



Christian