lucene-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Michael McCandless (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (LUCENE-6630) Further improve memory of FieldInfos
Date Sun, 28 Jun 2015 18:01:05 GMT

    [ https://issues.apache.org/jira/browse/LUCENE-6630?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14604810#comment-14604810
] 

Michael McCandless commented on LUCENE-6630:
--------------------------------------------

+1, this looks nice!

Maybe add a comment above {{ return o1.number - o2.number;}} that this is safe (can't overflow
int) because {{number}} is non-negative?

We can increase the 1/16th heuristic now?  We picked that because that was the cutover to
minimize RAM, but with this change it suddenly becomes a time/space tradeoff because the sparse
representation is now always <= the non-sparse one.  Maybe we should always just be sparse,
unless it's complete dense?

> Further improve memory of FieldInfos
> ------------------------------------
>
>                 Key: LUCENE-6630
>                 URL: https://issues.apache.org/jira/browse/LUCENE-6630
>             Project: Lucene - Core
>          Issue Type: Bug
>            Reporter: Ryan Ernst
>         Attachments: LUCENE-6630.patch
>
>
> In LUCENE-6325, FieldInfos were improved to use an array except in very sparse cases.
In the sparse cases, a TreeMap is still used. However, even in the sparse case, we can still
use an array, and then use binary search. The perf will be equivalent, but memory use should
be much less.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

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


Mime
View raw message