lucene-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Martijn van Groningen (JIRA)" <>
Subject [jira] [Commented] (SOLR-2524) Adding grouping to Solr 3x
Date Wed, 18 May 2011 18:56:49 GMT


Martijn van Groningen commented on SOLR-2524:

bq. Maybe rename group.cache.maxSize -> .maxSizeMB? (So it's clear what the units are).
Yes that is a more descriptive name.

bq. Should we default group.cache to true? (It's false now?).
That makes sense. 

I think that if the cachedCollector.isCached() returns false we should put something in the
response indication that the cache wasn't used because it hit the cache.maxSizeMB limit. Otherwise
the nobody will no whether the cache was utilized.

When I was playing around with the cache options I noticed that searching without cache (~350
ms) was faster then with cache (~500 ms) on a 10M index with 1711 distinct group values. This
is not what I'd expect.

bq. When you get the top groups from collector2, should you pass in offset instead of 0? (Hmm
– maybe groupOffset? It seems like you're using offset for both the first & second phase
collectors? Maybe I'm confused...).
I know that is confusing, but the DocSlice expects offset + len documents. So that was a quick
of doing that. I will clean that up.

bq. This matches how Solr does grouping on trunk right?
Yes it does. I'm already thinking about a new collector that collects all most relevant documents
of all groups. This collector should produce something like an OpenBitSet. We can use the
OpenBitSet to create a DocSet. I think this should be implemented in a different issue.

> Adding grouping to Solr 3x
> --------------------------
>                 Key: SOLR-2524
>                 URL:
>             Project: Solr
>          Issue Type: New Feature
>    Affects Versions: 3.2
>            Reporter: Martijn van Groningen
>            Assignee: Michael McCandless
>         Attachments: SOLR-2524.patch
> Grouping was recently added to Lucene 3x. See LUCENE-1421 for more information.
> I think it would be nice if we expose this functionality also to the Solr users that
are bound to a 3.x version.
> The grouping feature added to Lucene is currently a subset of the functionality that
Solr 4.0-trunk offers. Mainly it doesn't support grouping by function / query.
> The work involved getting the grouping contrib to work on Solr 3x is acceptable. I have
it more or less running here. It supports the response format and request parameters (expect:
group.query and group.func) described in the FieldCollapse page on the Solr wiki.
> I think it would be great if this is included in the Solr 3.2 release. Many people are
using grouping as patch now and this would help them a lot. Any thoughts?

This message is automatically generated by JIRA.
For more information on JIRA, see:

To unsubscribe, e-mail:
For additional commands, e-mail:

View raw message