[
https://issues.apache.org/jira/browse/LUCENE-5293?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13798541#comment-13798541
]
Adrien Grand commented on LUCENE-5293:
--------------------------------------
bq. DocIdSetIterator.cost has another purpose, so it's not worthwhile to use it
here I think.
I was mentioning it because it is the closest thing we have to a cardinality()
which is available for every DocIdSet.
bq. Does the faster build time of the EF DocIdSet (compared to WAH8 and FBS)
allow for an extra FBS to be built?
bq. That is not immediately clear from the benchmark results, but it could be
so.
My guess is that it is faster to build the in-memory doc id sets compared to
consuming the (eg. QueryWrapper) filter we need to cache, so indeed, it may
allow for building an additional FBS. Since WAH8DocIdSet computes it size
(available thorugh cardinality()), maybe we could build a WAH8DocIdSet in any
case and replace it with an EF doc id set when there are not many documents (?)
I can try to update the benchmarks to add the building of an additional FBS
before the EF doc id set. This also reminds me that I should look into the
building time of the WAH8 doc id set, there are probably things to improve...
I'm currently thinking it may be due to the fact that it keeps resizing buffers
but I may be completely wrong.
> Also use EliasFanoDocIdSet in CachingWrapperFilter
> --------------------------------------------------
>
> Key: LUCENE-5293
> URL: https://issues.apache.org/jira/browse/LUCENE-5293
> Project: Lucene - Core
> Issue Type: Improvement
> Components: core/search
> Reporter: Paul Elschot
> Priority: Minor
> Attachments: LUCENE-5293.patch
>
>
--
This message was sent by Atlassian JIRA
(v6.1#6144)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
