lucene-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Michael McCandless" <m...@mikemccandless.com>
Subject Re: [jira] Created: (LUCENE-854) Create merge policy that doesn't periodically inadvertently optimize
Date Fri, 11 May 2007 14:00:40 GMT

"Yonik Seeley" <yonik@apache.org> wrote:
> On 5/2/07, Michael McCandless <mail@mikemccandless.com> wrote:
> > It would merge based on size (not # docs), would be free to merge
> > adjacent segments (not just rightmost segments), and would merge N
> > (configurable) at a time.
> 
> Hopefully it will always be easy to switch the meaning of "size" so it
> could be in bytes or number of documents.  Just keeping in mind the
> ParallelReader...

Good point.  I had been thinking we would keep the current merge
policy as a choice that users of ParallelReader could use.

But we could in addition give this new merge policy a "switch" as
whether size is measured by #docs vs by #bytes.  The benefit for users
of ParallelReader would then be that the policy takes deletes into
account and has freedom to merge adjacent (not just rightmost)
segments vs the current merge policy.  This seems like the way to go.

Mike
-- 
  Michael McCandless
  mail@mikemccandless.com


---------------------------------------------------------------------
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