gf2121 opened a new pull request, #14176:
URL: https://github.com/apache/lucene/pull/14176
**Background**
* https://github.com/apache/lucene/pull/541 tried to introduce the
vectorized decoding for BKD leaf blocks
* https://github.com/apache/lucene/pull/706 reverted the PR since we see
sgup432 commented on issue #14028:
URL: https://github.com/apache/lucene/issues/14028#issuecomment-2617663032
@kkewwei Great to see the improvement by changing the skip factor.
I see there have been many discussions around finding the right value for
`skip_factor` ([here](https://issu
gsmiller commented on code in PR #13974:
URL: https://github.com/apache/lucene/pull/13974#discussion_r1931291804
##
lucene/sandbox/src/java/org/apache/lucene/sandbox/search/SortedSetMultiRangeQuery.java:
##
@@ -0,0 +1,300 @@
+/*
+ * Licensed to the Apache Software Foundation (AS
benchaplin closed pull request #14085: Implement ACORN-1 search for HNSW
URL: https://github.com/apache/lucene/pull/14085
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To uns
ChrisHegarty opened a new issue, #14175:
URL: https://github.com/apache/lucene/issues/14175
Reproduces on _main_.
```
Reproduce with: gradlew :lucene:core:test --tests
"org.apache.lucene.search.TestKnnFloatVectorQuery.testFindFewer" -Ptests.jvms=6
"-Ptests.jvmargs=-XX:TieredStopAt
benwtrent commented on PR #14167:
URL: https://github.com/apache/lucene/pull/14167#issuecomment-2616760637
> This made me wonder if it would be a better trade-off to let just one
slice run on its own first, and then let all other N-1 slices run in parallel
with one another,
I really
navneet1v commented on PR #14163:
URL: https://github.com/apache/lucene/pull/14163#issuecomment-2616674699
@cheng66551 I have also seen similar behavior but not with ES, but with
Opensearch.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log
rmuir commented on PR #14132:
URL: https://github.com/apache/lucene/pull/14132#issuecomment-2616629072
I will look in on the ECJ side, last time I used it, their null analysis had
issues on lucene's enormous codebase. It was many years ago though, maybe it
has solidified.
error-prone
rmuir commented on PR #14132:
URL: https://github.com/apache/lucene/pull/14132#issuecomment-2616604765
One option we could do for correctness is to turn on ecj's null analysis.
ecj is pretty fast and runs as part of gradle checks already, and it is the
compiler often using this feature in t
benwtrent commented on issue #13640:
URL: https://github.com/apache/lucene/issues/13640#issuecomment-2616596850
I discovered two other weird behaviors digging into this test failure. But,
neither seemed to fix this inconsistency:
https://github.com/apache/lucene/pull/14174
--
Thi
benwtrent opened a new pull request, #14174:
URL: https://github.com/apache/lucene/pull/14174
This fixes two minor inconsistencies.
- Makes sure that connected components isn't called twice with concurrent
hnsw merger
- make sure that duplicate connections are also handled when a
vigyasharma opened a new pull request, #14173:
URL: https://github.com/apache/lucene/pull/14173
Another take at #12313
The following PR adds support for _independent_ multi-vectors, i.e.
scenarios where a single document is represented by multiple independent vector
values. The most
benwtrent opened a new pull request, #14172:
URL: https://github.com/apache/lucene/pull/14172
With the new connected components work, merging is unstable. It generally
seems that between indexing, there might be new connections as `vex` index size
increases.
I am not 100% sure the ca
jpountz commented on PR #14070:
URL: https://github.com/apache/lucene/pull/14070#issuecomment-2616495910
Updated benchmark results, there is still a speedup. I'll merge soon.
```
TaskQPS baseline StdDevQPS
my_modified_version StdDev
benwtrent opened a new issue, #13640:
URL: https://github.com/apache/lucene/issues/13640
### Description
All KNN formats are periodically failing `testMergeStability`.
I have verified its due to https://github.com/apache/lucene/pull/13566
The stability failure is due to
benwtrent commented on issue #13640:
URL: https://github.com/apache/lucene/issues/13640#issuecomment-2616417275
@msokolov its reared its head again.
```
./gradlew test --tests TestPerFieldKnnVectorsFormat.testMergeStability
-Dtests.seed=FF1182F3FC600FF -Dtests.locale=mni-Beng-IN
jpountz commented on PR #14167:
URL: https://github.com/apache/lucene/pull/14167#issuecomment-2616408185
Somewhat related, thinking out loud: I have been wondering about what is the
best way to parallelize top-k query processing. Lexical search has a similar
issue as knn search in that it i
jpountz commented on PR #14167:
URL: https://github.com/apache/lucene/pull/14167#issuecomment-2616390109
> I don't know of another query where multiple passes over a static dataset
can return different docs.
Currently, this does not happen because Lucene only enables so-called
"rank-
benwtrent commented on PR #14167:
URL: https://github.com/apache/lucene/pull/14167#issuecomment-2616250521
@mayya-sharipova maybe a search time flag is possible, but it would stink to
have a "inconsistent but fast" flag that users then have to worry about.
I don't know of another quer
msokolov commented on PR #14132:
URL: https://github.com/apache/lucene/pull/14132#issuecomment-2616198970
> Specifying that something is nullable doesn't provide any value: all types
are nullable by default already.
I guess these null-checking systems impose their own assumptions and
rmuir commented on PR #14132:
URL: https://github.com/apache/lucene/pull/14132#issuecomment-2616174314
For the java language:
Specifying that something is nullable doesn't provide any value: all types
are nullable by default already.
Specifying that something is NOT-nullable wo
benwtrent commented on code in PR #14094:
URL: https://github.com/apache/lucene/pull/14094#discussion_r1930637609
##
lucene/core/src/java/org/apache/lucene/search/HnswKnnCollector.java:
##
@@ -0,0 +1,24 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one or mor
chatman commented on PR #14131:
URL: https://github.com/apache/lucene/pull/14131#issuecomment-2615867677
> I want to start cleaning some of the outstanding items in this PR, but I
do not have push access to SearchScale:cuvs-integration-main. Can I get access,
or is there a better way to pro
chatman commented on PR #14131:
URL: https://github.com/apache/lucene/pull/14131#issuecomment-2615875009
>
restructure cuvs-java so that it compiles to a minimum JDK 21, with an mr
jar/version specific loading. Maybe it can even strip use class file version 65
and strip the preview
uschindler commented on PR #14131:
URL: https://github.com/apache/lucene/pull/14131#issuecomment-2615941784
One additional important thing: no public API in this new Lucene
module/codec must export any preview API, so it must all be private/pkg-private.
--
This is an automated message fro
ChrisHegarty commented on PR #14131:
URL: https://github.com/apache/lucene/pull/14131#issuecomment-2615930453
> P.P.S. Elasticsearch has a Gradle plugin to strip preview flags. Basically
it patches one byte in all class files that are created by Javac.
Yes, we can do this. Along with
ChrisHegarty commented on PR #14131:
URL: https://github.com/apache/lucene/pull/14131#issuecomment-2615927388
> Tha APIs are available in Java 21, too (with minimal changes regarding
some specific parts like string handling). If you omit those, you can compile
against java 21 and later stri
uschindler commented on PR #14131:
URL: https://github.com/apache/lucene/pull/14131#issuecomment-2615924479
P.P.S. Elasticsearch has a Gradle plugin to strip preview flags.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and u
benwtrent commented on PR #14131:
URL: https://github.com/apache/lucene/pull/14131#issuecomment-2615923918
@dweiss While I think that might work for unused/snapshot Lucene releases, I
think @chatman et. al. is aiming for usage in the current Lucene Main so that
Lucene focused search engines
uschindler commented on PR #14131:
URL: https://github.com/apache/lucene/pull/14131#issuecomment-2615909166
> > restructure cuvs-java so that it compiles to a minimum JDK 21, with an
mr jar/version specific loading. Maybe it can even strip use class file version
65 and strip the preview bit
dweiss commented on PR #14131:
URL: https://github.com/apache/lucene/pull/14131#issuecomment-2615902695
The third option is to bump the minimum Java requirement to Java 22 on main?
I know it's an interim release but maybe we should just do it, anticipating the
next major lts (due to be rele
benwtrent commented on code in PR #14094:
URL: https://github.com/apache/lucene/pull/14094#discussion_r1930581288
##
lucene/core/src/java/org/apache/lucene/search/HnswKnnCollector.java:
##
@@ -0,0 +1,24 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one or mor
jpountz commented on issue #14168:
URL: https://github.com/apache/lucene/issues/14168#issuecomment-2615747790
`PhraseWildcardSearch` is appealing, but its implementation makes trade-offs
to work around the fact that it doesn't work efficiently if any of the
wildcards expands to many terms.
msokolov commented on PR #14132:
URL: https://github.com/apache/lucene/pull/14132#issuecomment-2615754639
This seems reasonable to me, but it implies a future promise to maintain it,
and I don't know how we would ever know if we added some new usage that isn't
properly annotated
--
This
msokolov commented on issue #14127:
URL: https://github.com/apache/lucene/issues/14127#issuecomment-2615749814
the tests seem to have stopped failing, as expected, closing
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and us
msokolov commented on PR #14129:
URL: https://github.com/apache/lucene/pull/14129#issuecomment-2615746064
Thanks, @msfroh
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
T
msokolov closed issue #14127: TestBPReorderingMergePolicy fails
CheckIndex.testHnswGraph
URL: https://github.com/apache/lucene/issues/14127
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specif
msokolov merged PR #14129:
URL: https://github.com/apache/lucene/pull/14129
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: issues-unsubscr...@lucene.ap
mayya-sharipova merged PR #14169:
URL: https://github.com/apache/lucene/pull/14169
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: issues-unsubscr...@lu
ChrisHegarty commented on PR #14131:
URL: https://github.com/apache/lucene/pull/14131#issuecomment-2615430123
I want to start cleaning some of the outstanding items in this PR, but I do
not have push access to SearchScale:cuvs-integration-main. Can I get access, or
is there a better way to
ChrisHegarty commented on PR #14131:
URL: https://github.com/apache/lucene/pull/14131#issuecomment-2615395910
Hi, `cuvs-java-25.02` is currently compiled with JDK 22, so has a minimum
class file version of 66. Lucene compiles with a minimum of JDK 21, class file
version 65. The reason why
41 matches
Mail list logo
