On Fri, 11 May 2012 06:30:37 -0700, Mike Sokolov <[email protected]> wrote:
> Thanks for that, Mary - It's good to know that position indexes could > help here - I had always assumed positions were only relevant for > element-query, and phrases, but I guess I shouldn't be surprised you all > have found other uses for them! > > I also didn't know you were doing lexicon expansion internally - do you > know when that came into the product? I seem to recall having to do > that myself just a couple of years ago. > > -Mike It has been contiuously evolving for some time; much of this was put in in 2009. There is certainly some flavour of wildcard optimization using the lexicons in all current versions of 4.0, 4.1, and 5.0. Cheers //Mary > > On 05/10/2012 09:51 PM, Mary Holstege wrote: >> >> The e* -> e, e?, e??, e??* rewrite is now done automatically. >> >> If you turn on 3 character wildcards and positions and put >> in place a word lexicon, a lot of wildcard queries can >> be resolved accurately by using lexicon expansion, possibly >> in combination with reducing the query to the set of unique >> 3-character prefixes and suffixes to make it more efficient >> (positions needed here). >> >> There are still a few heuristic gaps: since lexicon expansion >> can be expensive, we try to avoid it if it looks like there >> will be too many matches. >> >> //Mary >> _______________________________________________ >> General mailing list >> [email protected] >> http://developer.marklogic.com/mailman/listinfo/general _______________________________________________ General mailing list [email protected] http://developer.marklogic.com/mailman/listinfo/general
