Woot guys. That was a long reading :) Thanks for all the answers, I'll go back to prototyping. If i come up with something usefull, I would like to wrap up this discussion in a wiki page.
Thanks, Andrea On Wed, Mar 30, 2011 at 2:14 PM, Mattias Persson <matt...@neotechnology.com> wrote: > 2011/3/30 Craig Taverner <cr...@amanzi.com> > >> Agreed, Rick. My opinion is the main reason to role your own index is to >> make use of domain specific optimizations not available with generic >> indices. In my case the main win is the combination of statistics result >> and >> index that is possible. >> >> But I have to confess, the real reason I started using graphs as indexes >> was >> just that I thought the graph concept so cool, I did not want to pollute it >> with something non-graphy. Foolish ideology, I know, and I grew out of that >> more than a year ago, but it did influence many of my early neo4j decisions >> :-) >> > > I know that feeling :) > >> >> On Wed, Mar 30, 2011 at 1:49 PM, Rick Bullotta >> <rick.bullo...@thingworx.com>wrote: >> >> > My experience with using large graph trees for indexes has been mixed, >> > with performance issues under heavy read/write load, perhaps due to the >> many >> > potential locks required during insertions. We switched to the timeline >> > index, fwiw. >> > >> > >> > >> > ----- Reply message ----- >> > From: "Craig Taverner" <cr...@amanzi.com> >> > Date: Wed, Mar 30, 2011 7:43 am >> > Subject: [Neo4j] question >> > To: "Neo4j user discussions" <user@lists.neo4j.org> >> > >> > > >> > > > I think for that the TimelineIndex interface would have to be >> extended >> > to >> > > be able to hold additional data so that you can do compound >> > > queries< >> > > >> > >> http://docs.neo4j.org/chunked/milestone/indexing-lucene-extras.html#indexing-lucene-compound >> > > >to >> > > it and get exactly the functionality you're asking for with only one >> > > index. Another way is to just copy the LuceneTimeline code and roll >> this >> > > yourself, it's really small, mostly one-liners for each implemented >> > method. >> > > >> > >> > Alternatively just role your own graph-tree structure that provides the >> > same >> > capabilities. Then you can index any combination of properties together, >> to >> > suite your planned queries. This is obviously much more work than Mattias >> > suggestion, and does require that you know more about your domain (ie. >> less >> > general). But it does allow you to inspect the index itself with graph >> > traversals, gremlin or neoclipse, which is not possible with lucene. >> > _______________________________________________ >> > Neo4j mailing list >> > User@lists.neo4j.org >> > https://lists.neo4j.org/mailman/listinfo/user >> > >> _______________________________________________ >> Neo4j mailing list >> User@lists.neo4j.org >> https://lists.neo4j.org/mailman/listinfo/user >> > > > > -- > Mattias Persson, [matt...@neotechnology.com] > Hacker, Neo Technology > www.neotechnology.com > _______________________________________________ > Neo4j mailing list > User@lists.neo4j.org > https://lists.neo4j.org/mailman/listinfo/user > _______________________________________________ Neo4j mailing list User@lists.neo4j.org https://lists.neo4j.org/mailman/listinfo/user
