[
https://issues.apache.org/jira/browse/LUCENE-9855?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17315813#comment-17315813
]
Robert Muir commented on LUCENE-9855:
-------------------------------------
{quote}
For postings and doc values, we make it relatively easy via
PerFieldPostingsFormat and PerFieldDocValuesFormat, as well as
Lucene90Codec#getPostingsFormatForField and
Lucene90Codec#getDocValuesFormatForField. It still requires touching the codec,
but in a less daunting way than writing a codec from scratch and then
registering it.
{quote}
+1: I see the "strategy" stuff as just an inconsistent way of doing this,
versus postings and docvalues, but with a lot of downsides. It would be cleaner
to provide different algos via PerField???Format. Instead of having one
"mega-format" that is really trying to support 3 different algorithms, create 3
subclasses of ???Format. We should be able to experiment/randomize these
algorithms in unit tests (just like docvalues/postings). Having a simpler api
makes it easier to add exciting new stuff to codecs/ package and start baking
it with tests without having long terms backwards compatibility committment.
It makes it easier to optimize the memory usage and data structures to support
a single algorithm rather than trying to create a wonder-do-it-all rube
goldberge machine. It also makes it easier to add not-fully-optimized
implementations or ones with caveats. The implementation in the default index
format should be biased heavily towards production use (cpu efficiency, low
memory usage, etc).
> Reconsider codec name VectorFormat
> ----------------------------------
>
> Key: LUCENE-9855
> URL: https://issues.apache.org/jira/browse/LUCENE-9855
> Project: Lucene - Core
> Issue Type: Improvement
> Components: core/codecs
> Affects Versions: main (9.0)
> Reporter: Tomoko Uchida
> Assignee: Tomoko Uchida
> Priority: Blocker
> Fix For: main (9.0)
>
> Time Spent: 20m
> Remaining Estimate: 0h
>
> There is some discussion about the codec name for ann search.
> https://lists.apache.org/thread.html/r3a6fa29810a1e85779de72562169e72d927d5a5dd2f9ea97705b8b2e%40%3Cdev.lucene.apache.org%3E
> Main points here are 1) use plural form for consistency, and 2) use more
> specific name for ann search (second point could be optional).
> A few alternatives were proposed:
> - VectorsFormat
> - VectorValuesFormat
> - NeighborsFormat
> - DenseVectorsFormat
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
