[
https://issues.apache.org/jira/browse/SOLR-764?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12630136#action_12630136
]
Lars Kotthoff commented on SOLR-764:
------------------------------------
As far as I could see, "listFacets" wasn't used anywhere in the code --
everything was put into "topFacets" (which behaved exactly the same as
"listFacets") and sorted in the final stage. I guess the idea was to have a
data structure which automatically sorts entries as new ones are inserted.
Optimising the combining of sorted data from the shards can't be done in the
general case, as e.g. for sort by count the data returned affects the order --
merging the results from 2 shards can change the order completely compared
because you're adding up the counts.
If you specify facet.sort=false in a non-distributed setup, the facet values
are ordered by term number as well -- which happens to be the lexicographic
order. Also, sorting by term name only really makes sense for string or text
fields.
> Support facet.sort=false (index order) with distributed search
> --------------------------------------------------------------
>
> Key: SOLR-764
> URL: https://issues.apache.org/jira/browse/SOLR-764
> Project: Solr
> Issue Type: New Feature
> Affects Versions: 1.3
> Reporter: Wojtek Piaseczny
> Fix For: 1.4
>
> Attachments: SOLR-764.patch
>
>
> Distributed search does not currently support sorting facets by index order
> (facet.sort=false).
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
