dweiss commented on code in PR #15120:
URL: https://github.com/apache/lucene/pull/15120#discussion_r2299848291
##
lucene/core/src/java/org/apache/lucene/index/IndexWriter.java:
##
@@ -4777,6 +4782,21 @@ private void abortOneMerge(MergePolicy.OneMerge merge)
throws IOException {
dweiss commented on code in PR #15120:
URL: https://github.com/apache/lucene/pull/15120#discussion_r2299847152
##
lucene/core/src/test/org/apache/lucene/index/TestIndexWriterMerging.java:
##
@@ -461,4 +461,33 @@ public void run() {
directory.close();
}
+
+ public void
dweiss commented on PR #15110:
URL: https://github.com/apache/lucene/pull/15110#issuecomment-3222702445
Yeah... I'm no longer so convinced we should accept micro-benchmarking
results without looking at overall performance. When you run the code in a
different context it seems to compile dif
dnhatn opened a new pull request, #15125:
URL: https://github.com/apache/lucene/pull/15125
(no comment)
--
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-ma
github-actions[bot] commented on PR #15124:
URL: https://github.com/apache/lucene/pull/15124#issuecomment-3222332460
This PR does not have an entry in lucene/CHANGES.txt. Consider adding one.
If the PR doesn't need a changelog entry, then add the skip-changelog label to
it and you will stop
sgup432 opened a new pull request, #15124:
URL: https://github.com/apache/lucene/pull/15124
### Description
Related issue - https://github.com/apache/lucene/issues/15097
Instead of using the default 1024 bytes for query size, we try to use
RamUsageEstimator to calculate
easyice commented on PR #15045:
URL: https://github.com/apache/lucene/pull/15045#issuecomment-3222169221
No problem. I will update it.
--
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
ayinresh commented on PR #15112:
URL: https://github.com/apache/lucene/pull/15112#issuecomment-3222009047
Are there any plans to backport this PR considering the severity of the bug?
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to Gi
ayinresh commented on issue #15068:
URL: https://github.com/apache/lucene/issues/15068#issuecomment-3222001450
@uschindler
After investigating the production outage, you are absolutely correct that
we were illegally using a searcher that had already been released. I found an
error lo
jainankitk commented on PR #15123:
URL: https://github.com/apache/lucene/pull/15123#issuecomment-3221971679
That's a good point. I was initially thinking of doing that by adding
`relate` method similar to `PointRangeQuery`, but thought this could be
simpler. I overlooked this key detail reg
jpountz commented on PR #15110:
URL: https://github.com/apache/lucene/pull/15110#issuecomment-3221920602
I ran `PostingIndexInputBenchmark` with the vector module enabled to check
performance, but it seems to report a slowdown.
main:
```
Benchmark
jpountz commented on PR #15123:
URL: https://github.com/apache/lucene/pull/15123#issuecomment-3221874432
Javadocs of `LeafReader#numDocs` warn that it may run in O(maxDoc) time. I
wonder if we should instead extract a method that tells whether it matches all
or nothing (without calling numD
github-actions[bot] commented on PR #15123:
URL: https://github.com/apache/lucene/pull/15123#issuecomment-3221860977
This PR does not have an entry in lucene/CHANGES.txt. Consider adding one.
If the PR doesn't need a changelog entry, then add the skip-changelog label to
it and you will stop
jainankitk opened a new pull request, #15123:
URL: https://github.com/apache/lucene/pull/15123
### Description
Reusing count() for minor refactor in SortedNumericDocValuesRangeQuery
--
This is an automated message from the Apache Git Service.
To respond to the message, pleas
jainankitk commented on PR #14413:
URL: https://github.com/apache/lucene/pull/14413#issuecomment-3221761929
Thanks all for reviewing this PR. Planning to merge this PR by tomorrow, if
there is no new feedback. Again, thanks for helping improve this change with
your inputs!
--
This is an
uschindler merged PR #15116:
URL: https://github.com/apache/lucene/pull/15116
--
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.
jpountz commented on PR #15116:
URL: https://github.com/apache/lucene/pull/15116#issuecomment-3221698361
FWIW we break implementations quite regularly, which I think is fine. I
wouldn't worry about it. We should be careful with not breaking users who call
our (non-internal) APIs. In my opin
uschindler commented on PR #15116:
URL: https://github.com/apache/lucene/pull/15116#issuecomment-3221687618
I will add a change log when backporting this. I am not yet sure how to not
break implementations, so let's wait a bit and let it bake om main.
--
This is an automated message from
github-actions[bot] commented on PR #15116:
URL: https://github.com/apache/lucene/pull/15116#issuecomment-3221682280
This PR does not have an entry in lucene/CHANGES.txt. Consider adding one.
If the PR doesn't need a changelog entry, then add the skip-changelog label to
it and you will stop
jpountz commented on code in PR #15116:
URL: https://github.com/apache/lucene/pull/15116#discussion_r2299064975
##
lucene/core/src/java/org/apache/lucene/util/GroupVIntUtil.java:
##
@@ -56,7 +58,54 @@ public static void readGroupVInts(DataInput in, int[] dst,
int limit) throws
uschindler commented on PR #15116:
URL: https://github.com/apache/lucene/pull/15116#issuecomment-3221641553
Last version had more or less same performance on my laptop:
```
Benchmark(size)
Mode Cnt Score Error Unit
cwperks commented on code in PR #15120:
URL: https://github.com/apache/lucene/pull/15120#discussion_r2298967850
##
lucene/core/src/test/org/apache/lucene/index/TestIndexWriterMerging.java:
##
@@ -461,4 +461,33 @@ public void run() {
directory.close();
}
+
+ public voi
cwperks commented on code in PR #15120:
URL: https://github.com/apache/lucene/pull/15120#discussion_r2298966736
##
lucene/core/src/java/org/apache/lucene/index/IndexWriter.java:
##
@@ -3285,6 +3285,11 @@ public AddIndexesMergeSource(IndexWriter writer) {
}
public voi
cwperks commented on code in PR #15120:
URL: https://github.com/apache/lucene/pull/15120#discussion_r2298954182
##
lucene/core/src/java/org/apache/lucene/index/IndexWriter.java:
##
@@ -4777,6 +4782,21 @@ private void abortOneMerge(MergePolicy.OneMerge merge)
throws IOException
dweiss commented on code in PR #15120:
URL: https://github.com/apache/lucene/pull/15120#discussion_r2298924395
##
lucene/core/src/test/org/apache/lucene/index/TestIndexWriterMerging.java:
##
@@ -461,4 +461,33 @@ public void run() {
directory.close();
}
+
+ public void
benwtrent commented on PR #15003:
URL: https://github.com/apache/lucene/pull/15003#issuecomment-3221485260
> It's probably because of disconnectedness issue (Let me try to find
connectedness number of these graphs as well.)
I would think so. My gut is that we don't actually go through
RamakrishnaChilaka commented on PR #15110:
URL: https://github.com/apache/lucene/pull/15110#issuecomment-3221188494
@dweiss / @rmuir / @jpountz If no further comments, can we please merge the
PR. Thank you!
--
This is an automated message from the Apache Git Service.
To respond to the mes
mccullocht commented on PR #15021:
URL: https://github.com/apache/lucene/pull/15021#issuecomment-3221175027
Integrated this into the new off heap bulk scorers.
My intuition was that I'd want to use vector instructions for this but
layering it is kind of tricky since we don't want to u
cwperks commented on code in PR #15120:
URL: https://github.com/apache/lucene/pull/15120#discussion_r2298724035
##
lucene/core/src/java/org/apache/lucene/index/IndexWriter.java:
##
@@ -4777,6 +4782,21 @@ private void abortOneMerge(MergePolicy.OneMerge merge)
throws IOException
github-actions[bot] commented on PR #15021:
URL: https://github.com/apache/lucene/pull/15021#issuecomment-3221169788
This PR does not have an entry in lucene/CHANGES.txt. Consider adding one.
If the PR doesn't need a changelog entry, then add the skip-changelog label to
it and you will stop
dweiss commented on code in PR #15120:
URL: https://github.com/apache/lucene/pull/15120#discussion_r2298713344
##
lucene/core/src/java/org/apache/lucene/index/IndexWriter.java:
##
@@ -4777,6 +4782,21 @@ private void abortOneMerge(MergePolicy.OneMerge merge)
throws IOException {
dweiss commented on PR #15122:
URL: https://github.com/apache/lucene/pull/15122#issuecomment-3221143066
Thank you for the follow-up and checking, @RamakrishnaChilaka !
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use th
github-actions[bot] commented on PR #15021:
URL: https://github.com/apache/lucene/pull/15021#issuecomment-3221113522
This PR does not have an entry in lucene/CHANGES.txt. Consider adding one.
If the PR doesn't need a changelog entry, then add the skip-changelog label to
it and you will stop
benwtrent closed issue #15059: Add optimistic knn search to other vector query
types
URL: https://github.com/apache/lucene/issues/15059
--
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 c
benwtrent merged PR #15063:
URL: https://github.com/apache/lucene/pull/15063
--
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.a
uschindler commented on PR #15116:
URL: https://github.com/apache/lucene/pull/15116#issuecomment-3220931509
Of course in 10.x I would keep the methods now made private.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use t
uschindler commented on PR #15116:
URL: https://github.com/apache/lucene/pull/15116#issuecomment-3220927490
Should we maybe merge this branch to main and wait for Mike's benchmarks. If
we backport, we need to make sure people customizing IndexInputs don't fail too
worse or we need to add a
msokolov commented on PR #15120:
URL: https://github.com/apache/lucene/pull/15120#issuecomment-3220901546
Lucene doesn't really separate unit tests from integration tests, because it
doesn't ship a service. But you can write unit tests at different levels of
abstraction. I think what you've
jpountz commented on PR #15045:
URL: https://github.com/apache/lucene/pull/15045#issuecomment-3220744455
We don't actually need to allocate a FixedBitSet of size maxDoc, we could
copy slices of 1024 bits into a FixedBitSet(1024) to do the counting?
--
This is an automated message from the
cwperks commented on PR #15120:
URL: https://github.com/apache/lucene/pull/15120#issuecomment-3220723408
> so this moves the accounting from deep in `registerMerge` to the top? Did
the test fail before this change?
@msokolov merges registered in `public void
registerMerge(MergePolicy
msokolov commented on PR #15120:
URL: https://github.com/apache/lucene/pull/15120#issuecomment-3220636868
so this moves the accounting from deep in `registerMerge` to the top? Did
the test fail before this change?
--
This is an automated message from the Apache Git Service.
To respond to
RamakrishnaChilaka commented on PR #15122:
URL: https://github.com/apache/lucene/pull/15122#issuecomment-3220604700
```
TaskQPS baseline StdDevQPS
my_modified_version StdDevPct diff p-value
BrowseMonthSSDVFacets3.
RamakrishnaChilaka closed pull request #15122: DataInput: Unroll loop in
readVInt and readVLong
URL: https://github.com/apache/lucene/pull/15122
--
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 s
easyice commented on PR #15045:
URL: https://github.com/apache/lucene/pull/15045#issuecomment-3220493289
It s a nice idea! although it requires allocating an
`FixedBitSet(bits.length())`, it is still much faster than checking bits one by
one.
Here are some JMH numbers:
```
benwtrent commented on PR #14863:
URL: https://github.com/apache/lucene/pull/14863#issuecomment-3220475933
> The question I had was whether the benefits are compelling enough to
maintain these functions..
I think the goal for 4bit is that we just have the "compressed" version only.
benwtrent commented on issue #15060:
URL: https://github.com/apache/lucene/issues/15060#issuecomment-3220415362
Yeah @mikemccand ! I will close in deference to @dweiss 's issue and fix :).
--
This is an automated message from the Apache Git Service.
To respond to the message, please log o
benwtrent closed issue #15060: TestMultiIndexMergeScheduler tests run forever?
URL: https://github.com/apache/lucene/issues/15060
--
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.
benwtrent commented on PR #15021:
URL: https://github.com/apache/lucene/pull/15021#issuecomment-3220363568
Yeah! This is what I had in mind. I would expect the underlying bulk scorers
should be able to keep track of their max score pretty easily (without much
overhead), allowing for the cal
RamakrishnaChilaka commented on PR #15122:
URL: https://github.com/apache/lucene/pull/15122#issuecomment-3220343731
Got it, Thanks for checking the PR @dweiss.
I agree that keeping the code simple is important. I observed improvements
specifically in the microbenchmarks, but I haven’t
dweiss commented on PR #15122:
URL: https://github.com/apache/lucene/pull/15122#issuecomment-3220325237
Unless this shows improvement in macrobenchmarks, I would leave such
optimizations to the hotspot compiler. The code is simpler the way it's
currently written and I think hotspot will tak
github-actions[bot] commented on PR #15122:
URL: https://github.com/apache/lucene/pull/15122#issuecomment-3220311711
This PR does not have an entry in lucene/CHANGES.txt. Consider adding one.
If the PR doesn't need a changelog entry, then add the skip-changelog label to
it and you will stop
RamakrishnaChilaka opened a new pull request, #15122:
URL: https://github.com/apache/lucene/pull/15122
### Description
After reviewing the PR: https://github.com/apache/lucene/pull/15065, I
considered further improving it by unrolling the for loop.
This PR optimizes the perform
benwtrent commented on PR #15090:
URL: https://github.com/apache/lucene/pull/15090#issuecomment-3220284930
> Currently we do have outer HNSW format wrappers for each variation of flat
vectors format. But in order to do so we are creating multiple duplicates of
Lucene99HnswVectorsFormat with
uschindler merged PR #15112:
URL: https://github.com/apache/lucene/pull/15112
--
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.
uschindler closed issue #15106: RefCountedShatredArena is causing infinite
(spin-)loop on close of IndexInput/IndexReader/... due to exception handling
URL: https://github.com/apache/lucene/issues/15106
--
This is an automated message from the Apache Git Service.
To respond to the message, pl
dweiss commented on code in PR #15118:
URL: https://github.com/apache/lucene/pull/15118#discussion_r2297285454
##
lucene/core/src/test/org/apache/lucene/search/TestTermInSetQuery.java:
##
@@ -112,16 +112,23 @@ public void testAllDocsInFieldTerm() throws IOException {
public v
56 matches
Mail list logo
