[jira] [Comment Edited] (OAK-2200) Tune cost calculation for lucene property index

["Chetan Mehrotra (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/OAK-2200?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14184991#comment-14184991 ] 

Chetan Mehrotra edited comment on OAK-2200 at 10/27/14 9:21 AM:
----------------------------------------------------------------

bq. Later on (let's see) we might also want to add more ways to configure the cost (costMin, costFactor, or similar), but I guess that's not urgent.

[~tmueller] Is the proposed approahc a short term solution or final one? As for ideal case index implementation should just report the right metric (avg cost, no of entries etc) and let query engine pickup the right index based on that. Tuning up the cost like above would be fragile. So a long term fix would be to let nodeType index report proper entryCount. 


was (Author: chetanm):
nq. Later on (let's see) we might also want to add more ways to configure the cost (costMin, costFactor, or similar), but I guess that's not urgent.

[~tmueller] Is the proposed approahc a short term solution or final one? As for ideal case index implementation should just report the right metric (avg cost, no of entries etc) and let query engine pickup the right index based on that. Tuning up the cost like above would be fragile. So a long term fix would be to let nodeType index report proper entryCount. 

> Tune cost calculation for lucene property index
> -----------------------------------------------
>
>                 Key: OAK-2200
>                 URL: https://issues.apache.org/jira/browse/OAK-2200
>             Project: Jackrabbit Oak
>          Issue Type: Sub-task
>          Components: oak-lucene
>            Reporter: Amit Jain
>            Assignee: Amit Jain
>             Fix For: 1.0.8, 1.2
>
>         Attachments: OAK-2200.patch
>
>
> Though lucene provides an accurate count with {{IndexReader#numDocs}}, we may need to apply heuristics to constrain the value so that it is properly selected for indexed properties Vs other inappropriate indexes e.g wrt nodeType index. 



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