> Hi, > > On Jan 21, 2008 12:51 PM, Ard Schrijvers > <[EMAIL PROTECTED]> wrote: > > > Jukka Zitting wrote: > > > It should be fairly easy to optimize this by just updating the > > > changed path references in the search index for any nodes that > > > haven't been changed during the move. > > > > I suppose updating the nodes that *have* changed.
> Jukka Zitting wrote: > Yes. :-) What I was trying to express were nodes that are > moved but not otherwise changed. I.e. the only modification > is the path change. Clear :-) > > Good point. Though it shouldn't be too difficult to enhance > Lucene to reuse a previously analyzed and indexed document as > the basis of a slightly modified version. There's also the > new document payload feature that might be useful in this regard. I am not aware of the 'new document payload' feature, but I think reusing an already indexed and analyzed document might be a handy lucene feature anyway. OTOH I thought it was really an important Jackrabbit part to not store any hierarchical data whatsoever because inefficient moves (though indexing it might not be seen as storing it and if we enhance lucene somewhat...). What I do like about your idea is that we can implement some queries much more efficient when we store path info: the current havy DescendantSelfAxisWeight/ChildAxisQuery can be implemented easier because we might be able to find all nodes that are valid according its path with a single lucene query (see JCR-1196 'Optimize queries for DescendantSelfAxisWeight/ChildAxisQuery' ) -Ard > > BR, > > Jukka Zitting >
