[
https://issues.apache.org/jira/browse/LUCENE-3661?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13195640#comment-13195640
]
Robert Muir commented on LUCENE-3661:
-------------------------------------
I'll open a followup issue (can be done in trunk after this is committed).
Just for reference, there are 3 bitvector versions that can be read:
unversioned/darkages: some 3.x versions
codec header version 0: some 3.x versions
codec header version 1: trunk
so trunk always writes a versioned codec header, even lucene 3.5 always writes
a versioned codec header, but i think the 3.x codec must at least support
unversioned files still for e.g. 3.0 indexes.
also, under the same issue we could maybe remove bitvector completely and the
codecs just use FastBitSet. there is no difference now that count() is unused.
This could be cleaner: just having different serialization forms for fastbitset.
> move deletes under codec
> ------------------------
>
> Key: LUCENE-3661
> URL: https://issues.apache.org/jira/browse/LUCENE-3661
> Project: Lucene - Java
> Issue Type: Task
> Affects Versions: 4.0
> Reporter: Robert Muir
> Attachments: LUCENE-3661.patch
>
>
> After LUCENE-3631, this should be easier I think.
> I haven't looked at it much myself but i'll play around a bit, but at a
> glance:
> * SegmentReader to have Bits liveDocs instead of BitVector
> * address the TODO in the IW-using ctors so that SegmentReader doesn't take a
> parent but just an existing core.
> * we need some type of minimal "MutableBits" or similar subinterface of bits.
> BitVector and maybe Fixed/OpenBitSet could implement it
> * BitVector becomes an impl detail and moves to codec (maybe we have a shared
> base class and split the 3.x/4.x up rather than the conditional backwards)
> * I think the invertAll should not be used by IndexWriter, instead we define
> the codec interface to say "give me a new MutableBits, by default all are
> set" ?
> * redundant internally-consistent checks in checkLiveCounts should be done in
> the codec impl instead of in SegmentReader.
> * plain text impl in SimpleText.
--
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]
