[jira] [Updated] (LUCENE-7019) explore two-phase iteration for GeoPoint query

["Robert Muir (JIRA)" <ji...@apache.org>]

     [ https://issues.apache.org/jira/browse/LUCENE-7019?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Robert Muir updated LUCENE-7019:
--------------------------------
    Attachment: LUCENE-7019.patch

Here's an initial patch, the tests seem happy.

In order to keep the optimization where we only verify edges, we have to use an additional SparseBitSet.

I didn't do any benchmarking yet: i expect that if you benchmark the query in isolation, it can only be very slightly slower (since it needs to set another bit for those edge terms). But I think it should be much better if e.g. AND'd with a termquery or similar.

> explore two-phase iteration for GeoPoint query
> ----------------------------------------------
>
>                 Key: LUCENE-7019
>                 URL: https://issues.apache.org/jira/browse/LUCENE-7019
>             Project: Lucene - Core
>          Issue Type: Bug
>            Reporter: Robert Muir
>         Attachments: LUCENE-7019.patch
>
>
> This query today uses an approximation+confirm approach, but it all happens when you call scorer(), in a termsEnum loop.
> This causes several problems (even after https://issues.apache.org/jira/browse/LUCENE-7018) because it can do too much work, if queries have multiple values since the doc can be "confirmed" more than once.
> I think it would be better to delay this confirmation as much as possible, so that other parts of the query (e.g. other filters, conjunctions, etc) can eliminate checks as well.



--
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