lucene-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Doug Cutting <cutt...@apache.org>
Subject Re: svn commit: r332747 - in /lucene/java/trunk: ./ src/java/org/apache/lucene/search/regex/ src/test/org/apache/lucene/search/regex/
Date Thu, 17 Nov 2005 22:56:05 GMT
Yonik Seeley wrote:
> Hmmm, is .03->2000 really enough range?
> Seems like the choice is between that and .0005->2000000 will one less
> mantissa bit.

Consider the failure modes:

With the .0005->2000000 range we'll fail to distinguish close-scoring 
matches in more commmon score ranges, while more correctly 
distinguishing extreme scores.

With the .03->2000 range we'll fail to distinguish unusually high scores 
(overflow) and we'll over-distinguish unusually low scores (underflow), 
but we'll more accurately handle common score ranges.

I can't recall seeing many scores outside of the "normal" range, but 
then I have not used range or wildcard queries much.  It's easier to get 
huge scores with huge numbers of terms, although I still think it is 
unlikely.  You'd need several documents with lots of very rare terms.

Doug

---------------------------------------------------------------------
To unsubscribe, e-mail: java-dev-unsubscribe@lucene.apache.org
For additional commands, e-mail: java-dev-help@lucene.apache.org


Mime
View raw message