[
https://issues.apache.org/jira/browse/LUCENE-3443?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13146677#comment-13146677
]
Michael McCandless commented on LUCENE-3443:
--------------------------------------------
bq. I'm arguing a specific point though: that we shouldn't double time time it
takes solr users to instantiate field caches (or other lucene users that
regularly use sortMissingLast). "we can do it later" is not a good argument to
remove existing functionality.
We're not doubling the time vs 3.x.
{quote}
I'd rather not do that; apps that don't use sort missing first/last shouldn't
be forced to spend the RAM (even if it's only a bit per doc).
bq. But it's trunk, right? There's no guarantee w/in trunk that performance
won't "change" ;-)
bq. Using 1/32 more memory (3%) is way less evil than doubling field cache
entry times.
{quote}
Again, compared to 3.x, that'd be worse. Ie, w/ the patch as it now
stands we could ship 4.0 and we'd match 3.x.
For a byte[] FC entry that's 12.5% extra RAM.
I think one-time FC init cost is the lesser evil vs "permanently"
tying up unused RAM.
{quote}
bq. Also, this (uniqueTermCount) is redundant with Terms.getUniqueTermCount()?
Not for numeric fields, where a precisionStep causes nTerms > nUniqueValues
{quote}
Ahh, true.
> Port 3.x FieldCache.getDocsWithField() to trunk
> -----------------------------------------------
>
> Key: LUCENE-3443
> URL: https://issues.apache.org/jira/browse/LUCENE-3443
> Project: Lucene - Java
> Issue Type: Improvement
> Components: core/search
> Reporter: Michael McCandless
> Assignee: Michael McCandless
> Fix For: 3.5, 4.0
>
> Attachments: LUCENE-3443.patch
>
>
> [Spinoff from LUCENE-3390]
> I think the approach in 3.x for handling un-valued docs, and making it
> possible to specify how such docs are sorted, is better than the
> solution we have in trunk.
> I like that FC has a dedicated method to get the Bits for docs with field
> -- easy for apps to directly use. And I like that the
> bits have their own entry in the FC.
> One downside is that it's 2 passes to get values and valid bits, but
> I think we can fix this by passing optional bool to FC.getXXX methods
> indicating you want the bits, and the populate the FC entry for the
> missing bits as well. (We can do that for 3.x and trunk). Then it's
> single pass.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators:
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
