You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@lucene.apache.org by "Shai Erera (JIRA)" <ji...@apache.org> on 2013/11/04 10:59:24 UTC
[jira] [Commented] (LUCENE-5326) Add enum facet method to Lucene
facet module
[ https://issues.apache.org/jira/browse/LUCENE-5326?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13812721#comment-13812721 ]
Shai Erera commented on LUCENE-5326:
------------------------------------
Why does the app need to explicitly add the facets as "enum"? I mean, if you use FacetFields, it already adds the drilldown terms (effectively as StringField), and I thought the 'enum' method (like EnumFacetsAccumulator) will use the drilldown terms instead of the category list? At least, I don't see FF.addEnumFields doing something special when it adds the fields...
I also think that it's wrong that the facets aren't added to the taxonomy. E.g. if you start w/ a dimension that's small and you choose to optimize faceted search with 'enum', but then the dimension grows, so much that enum stops performing or costs too much RAM, what will you do - reindex the content?
If it's possible to index facets as usual, only optimize their accumulation, I think that's best. It will be like aggregating from Lucene45DVF, DirectDVF or CachedOrds (RAM grows with each, but performance improves) - it's a search-time decision, not indexing.
I haven't done a deep review of the patch yet, but in EnumFacetsAccumulator's jdocs you write "you should index normally e.g. StringField..." -- that's wrong? Currently you have to index using FF.addEnumFields?
> Add enum facet method to Lucene facet module
> --------------------------------------------
>
> Key: LUCENE-5326
> URL: https://issues.apache.org/jira/browse/LUCENE-5326
> Project: Lucene - Core
> Issue Type: Improvement
> Components: modules/facet
> Reporter: Michael McCandless
> Assignee: Michael McCandless
> Fix For: 4.6, 5.0
>
> Attachments: LUCENE-5326.patch
>
>
> I've been testing Solr facet performance, and the enum method works
> very well for low cardinality (not many unique values) fields. So I
> think we should fold a similar option into Lucene's facet module.
--
This message was sent by Atlassian JIRA
(v6.1#6144)
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@lucene.apache.org
For additional commands, e-mail: dev-help@lucene.apache.org