[KinoSearch] Wildcards

Marvin Humphrey marvin at rectangular.com
Sun Jan 27 20:15:54 PST 2008




On Jan 25, 2008, at 12:32 PM, Nathan Kurz wrote:

> But my instinct would to figure out a way simply make the search work,
> rather than throwing it out as an exception.  Supporting a search for
> "a* lovelace" seems reasonable, and shouldn't actually be that
> expensive if implemented lazily.

If you want correct results, you have to cruise through all the docs  
that match "a*" no matter what, because you won't know what the top  
scorers are until you've seen everything.

> If one was to have a limit, it should probably be on the total length
> of the records that need to be searched, not on the number of terms
> involved.

Or perhaps by introducing search timeouts.

   https://issues.apache.org/jira/browse/LUCENE-997

Unfortunately, it's not easy to integrate a bulletproof timeout  
mechanism into KS.  I think the most efficient approach would be to  
use threads: have a timer thread that checks back every once in a  
while to see if the query finishes and throws an exception if time  
runs out.  However, KS doesn't support threads.

I don't think we should get hung up on this detail, though.  For  
small collections, the cost won't be high enough to matter.

Marvin Humphrey
Rectangular Research
http://www.rectangular.com/

_______________________________________________
KinoSearch mailing list
KinoSearch at rectangular.com
http://www.rectangular.com/mailman/listinfo/kinosearch




More information about the kinosearch mailing list