ChrisHegarty commented on PR #12311:
URL: https://github.com/apache/lucene/pull/12311#issuecomment-1555874325
Just dumping an initial round of benchmark results, etc, based on what is
currently in this PR.
Benchmark source (derived from Robert's)
```
davekim$ cat sr
uschindler commented on PR #12311:
URL: https://github.com/apache/lucene/pull/12311#issuecomment-1555877992
Would also no interstin to run Mike's benchmark (the vector part).
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and
uschindler commented on PR #12311:
URL: https://github.com/apache/lucene/pull/12311#issuecomment-1555892033
Hi,
I rewrote the API stub extractor to use 2 passes. I also made the modules to
be exported configurable. It now only extracts vectors for Java 20.
The resulting Java 20 apijar
uschindler commented on code in PR #12311:
URL: https://github.com/apache/lucene/pull/12311#discussion_r1199602161
##
gradle/testing/defaults-tests.gradle:
##
@@ -119,10 +119,10 @@ allprojects {
if (rootProject.runtimeJavaVersion < JavaVersion.VERSION_16) {
jvmAr
uschindler commented on code in PR #12311:
URL: https://github.com/apache/lucene/pull/12311#discussion_r1199602226
##
gradle/testing/defaults-tests.gradle:
##
@@ -119,10 +119,10 @@ allprojects {
if (rootProject.runtimeJavaVersion < JavaVersion.VERSION_16) {
jvmAr
uschindler commented on PR #12311:
URL: https://github.com/apache/lucene/pull/12311#issuecomment-1555895724
There's something strange on my computer when running core tests with Java
20:
:lucene:core:test (SUCCESS): 5730 test(s), 193 skipped
The slowest tests (exceeding 500 ms) dur
uschindler commented on code in PR #12311:
URL: https://github.com/apache/lucene/pull/12311#discussion_r1199608565
##
gradle/testing/defaults-tests.gradle:
##
@@ -119,10 +119,10 @@ allprojects {
if (rootProject.runtimeJavaVersion < JavaVersion.VERSION_16) {
jvmAr
uschindler commented on PR #12311:
URL: https://github.com/apache/lucene/pull/12311#issuecomment-1555906079
> There's something strange on my computer when running core tests with Java
20:
>
> ```
> :lucene:core:test (SUCCESS): 5730 test(s), 193 skipped
> The slowest tests (exce
almogtavor opened a new issue, #12318:
URL: https://github.com/apache/lucene/issues/12318
### Description
I'd like to use Lucene Monitor in a non-blocking IO application, and I'd
like to know what is the recommended way of doing that. Currently, I match
queries with the `ParallelMatc
rmuir commented on PR #12311:
URL: https://github.com/apache/lucene/pull/12311#issuecomment-1555915665
Thanks for re-benchmarking @ChrisHegarty ! It has been a few years and an
older JDK version since this stuff was developed. I had in mind to do a couple
more runs eventually:
* try it o
uschindler commented on PR #12311:
URL: https://github.com/apache/lucene/pull/12311#issuecomment-1555926627
I imporved the extractor even more (it had a small bug). It now only uses
one pass, it just delays writing out the class file to apijar until all visible
classes were collected into a
gsmiller commented on PR #12310:
URL: https://github.com/apache/lucene/pull/12310#issuecomment-1555927824
@mikemccand where's the build method you're referencing? I took a pass at
creating a "direct to binary" version of the Daciuk-Mihov algorithm in #12312.
We could fold that into this wor
gsmiller opened a new issue, #12319:
URL: https://github.com/apache/lucene/issues/12319
### Description
`DaciukMihovAutomatonBuilder#build` requires sorted input but accepts a
`Collection` argument. Since `Collection` generally doesn't guarantee any
ordering as part of its contract,
uschindler commented on code in PR #12311:
URL: https://github.com/apache/lucene/pull/12311#discussion_r1199625873
##
lucene/core/src/java/org/apache/lucene/util/VectorUtilProvider.java:
##
@@ -0,0 +1,85 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one or mo
uschindler commented on code in PR #12311:
URL: https://github.com/apache/lucene/pull/12311#discussion_r1199625964
##
lucene/core/src/java/org/apache/lucene/util/VectorUtilProvider.java:
##
@@ -0,0 +1,85 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one or mo
uschindler commented on code in PR #12311:
URL: https://github.com/apache/lucene/pull/12311#discussion_r1199626630
##
lucene/core/src/java/org/apache/lucene/util/VectorUtilProvider.java:
##
@@ -0,0 +1,85 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one or mo
uschindler commented on code in PR #12311:
URL: https://github.com/apache/lucene/pull/12311#discussion_r1199630748
##
lucene/core/src/java/org/apache/lucene/util/VectorUtilProvider.java:
##
@@ -0,0 +1,85 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one or mo
uschindler commented on code in PR #12311:
URL: https://github.com/apache/lucene/pull/12311#discussion_r1199630782
##
lucene/core/src/java/org/apache/lucene/util/VectorUtilProvider.java:
##
@@ -0,0 +1,85 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one or mo
uschindler commented on PR #12311:
URL: https://github.com/apache/lucene/pull/12311#issuecomment-1555945898
I fixed the warnings to give correct instructions if incubator module is
missing or the default locale f*cks up.
--
This is an automated message from the Apache Git Service.
To resp
ChrisHegarty commented on PR #12311:
URL: https://github.com/apache/lucene/pull/12311#issuecomment-1555947416
@uschindler - Thanks for all the work, cleanup, log messages, etc. Looks
great.
Here's what I'm hoping to get to, might be tomorrow at this stage.
1. I'm in the process
uschindler commented on PR #12311:
URL: https://github.com/apache/lucene/pull/12311#issuecomment-1555962105
Hi,
> 1. I'm in the process of preparing a luceneutil run - still downloading on
my Linux box. Then I'll try to get some comparative numbers from the benchmark
- I'm not quite
kwatters commented on issue #12302:
URL: https://github.com/apache/lucene/issues/12302#issuecomment-1556050926
For what it's worth, I did an alternative impl of vector utils using nd4j to
compute cosine similarity... And shockingly, it was not any faster. I am very
much in favor of having
gsmiller opened a new pull request, #12320:
URL: https://github.com/apache/lucene/pull/12320
### Description
Adds the ability to directly build a binary automaton for a string union
using the Daciuk-Mihov algorithm, and uses it to make the
`TermInSetQuery#visit` implementation a litt
gsmiller commented on PR #12310:
URL: https://github.com/apache/lucene/pull/12310#issuecomment-1556069606
@mikemccand the `build` method you reference above in
`DaciukMihovAutomatonBuilder` build an automaton with code points as transition
labels, but I think we need a "compiled" binary aut
gsmiller opened a new issue, #12321:
URL: https://github.com/apache/lucene/issues/12321
### Description
There's some good suggestions/discussion around renaming this class in
#12310, but I wonder if we should consider making it pkg-private and exposing
the `build` functionality throu
gsmiller commented on PR #12310:
URL: https://github.com/apache/lucene/pull/12310#issuecomment-1556069972
Separately, in terms of renaming this class, I'm still in favor of it but I
also wonder if we should just consider making it pkg-private and proxying the
build functionality through `Au
26 matches
Mail list logo
