On Jan 7, 2010, at 12:21 PM, Marvin Humphrey wrote:
On Thu, Jan 07, 2010 at 10:13:00AM -0500, Ryan McKinley wrote:
With the new flexible indexing stuff, would it be possible to
natively
write an rtree to disk in the index process?
The question I'd have is, how would you handle interleaving of hits
from
different segments? Meaning, if you have a priority queue with 10
slots, 10
hits from one segment, and 10 hits from another segment, how do you
determine
which his win and which fall out the bottom of the queue?
(I am not familiar enough with the segment internals, so I may be way
off base...)
Perhaps each segment has its own r-tree. For each query, the r-tree
would only be used to say if something could be in the results, it
does not contribute to the score, so I don't think interleaving would
be any different then how it currently works (without knowing how it
currently works) Potentially, the node depth could influence the
score, but lets ignore that for now :)
ryan
---------------------------------------------------------------------
To unsubscribe, e-mail: java-user-unsubscr...@lucene.apache.org
For additional commands, e-mail: java-user-h...@lucene.apache.org
