rmuir commented on issue #11676:
URL: https://github.com/apache/lucene/issues/11676#issuecomment-1296934205
Why so quick to jump to wall time? Please, no wall time, for any reason
whatsoever.
Surely, nanoTime can be used.
--
This is an automated message from the Apache Git Service.
xingdong015 commented on issue #11887:
URL: https://github.com/apache/lucene/issues/11887#issuecomment-1296990074
It looks like gradle initialization takes a lot of time
donnerpeter commented on code in PR #11893:
URL: https://github.com/apache/lucene/pull/11893#discussion_r1009428049
##
lucene/analysis/common/src/java/org/apache/lucene/analysis/hunspell/GeneratingSuggester.java:
##
@@ -60,7 +64,11 @@ private List>>
findSimilarDictionaryEntries
donnerpeter commented on code in PR #11893:
URL: https://github.com/apache/lucene/pull/11893#discussion_r1009429083
##
lucene/analysis/common/src/java/org/apache/lucene/analysis/hunspell/GeneratingSuggester.java:
##
@@ -70,10 +78,10 @@ char transformChar(char c) {
}
donnerpeter commented on code in PR #11893:
URL: https://github.com/apache/lucene/pull/11893#discussion_r1009429597
##
lucene/analysis/common/src/java/org/apache/lucene/analysis/hunspell/GeneratingSuggester.java:
##
@@ -87,45 +95,31 @@ char transformChar(char c) {
s
donnerpeter commented on code in PR #11893:
URL: https://github.com/apache/lucene/pull/11893#discussion_r1009431207
##
lucene/analysis/common/src/java/org/apache/lucene/analysis/hunspell/WordStorage.java:
##
@@ -179,11 +186,7 @@ void processSuggestibleWords(
}
donnerpeter commented on code in PR #11893:
URL: https://github.com/apache/lucene/pull/11893#discussion_r1009432252
##
lucene/analysis/common/src/java/org/apache/lucene/analysis/hunspell/WordStorage.java:
##
@@ -54,7 +55,8 @@ class WordStorage {
private static final int COLLI
donnerpeter commented on code in PR #11893:
URL: https://github.com/apache/lucene/pull/11893#discussion_r1009433258
##
lucene/analysis/common/src/java/org/apache/lucene/analysis/hunspell/WordStorage.java:
##
@@ -54,7 +55,8 @@ class WordStorage {
private static final int COLLI
donnerpeter commented on code in PR #11893:
URL: https://github.com/apache/lucene/pull/11893#discussion_r1009433258
##
lucene/analysis/common/src/java/org/apache/lucene/analysis/hunspell/WordStorage.java:
##
@@ -54,7 +55,8 @@ class WordStorage {
private static final int COLLI
donnerpeter commented on PR #11893:
URL: https://github.com/apache/lucene/pull/11893#issuecomment-1297113922
With the cache, about 2x memory is used (~850MB for ~190 dictionaries). The
caching gives me about 1.5x speedup for en/ru/de.
--
This is an automated message from the Apache Git Se
donnerpeter commented on code in PR #11893:
URL: https://github.com/apache/lucene/pull/11893#discussion_r1009435738
##
lucene/analysis/common/src/test/org/apache/lucene/analysis/hunspell/TestPerformance.java:
##
@@ -86,7 +86,7 @@ public void de() throws Exception {
@Test
benwtrent commented on issue #10665:
URL: https://github.com/apache/lucene/issues/10665#issuecomment-1297146958
I opened a PR for ann-benchmarks:
https://github.com/erikbern/ann-benchmarks/pull/315
I tested PyLucene locally, comparing it to @msokolov's "batch" methodology
(writing to
jtibshirani commented on issue #10665:
URL: https://github.com/apache/lucene/issues/10665#issuecomment-1297299587
Thanks @benwtrent, it's great to see that PyLucene works well and has low
overhead! It feels more solid than what we were doing before.
+1 to preparing a new version. As I
vigyasharma commented on issue #11676:
URL: https://github.com/apache/lucene/issues/11676#issuecomment-1297303476
Sorry, I meant it'll add dependence on `nanoTime()`. I thought we use
wallTime to refer to both `currentTimeInMillis` and `nanoTime`.
If nanotime is acceptable, I can use
reta commented on PR #11875:
URL: https://github.com/apache/lucene/pull/11875#issuecomment-1297316017
> Looks good to me. I'll wait for a few days before merging, in case people
have comments/concerns with the public visibility for `TimeLimitingBulkScorer`
Thanks a lot @vigyasharma !
rmuir commented on issue #11676:
URL: https://github.com/apache/lucene/issues/11676#issuecomment-1297328617
> Sorry, I meant it'll add dependence on `nanoTime()`. I thought we use
wallTime to refer to both `currentTimeInMillis` and `nanoTime`.
nanoTime (at least on linux) uses the mon
rmuir commented on issue #11887:
URL: https://github.com/apache/lucene/issues/11887#issuecomment-1297341214
I don't think profiler is helpful because test is not doing anything, except
sleeping on `Object.wait`. I used `jstack` while the test was hung:
```
"TEST-TestDocumentsWriter
rmuir commented on issue #11887:
URL: https://github.com/apache/lucene/issues/11887#issuecomment-1297345527
It runs much faster with this patch:
```
---
a/lucene/core/src/test/org/apache/lucene/index/TestDocumentsWriterStallControl.java
+++
b/lucene/core/src/test/org/apache/lucene/
dweiss commented on code in PR #11893:
URL: https://github.com/apache/lucene/pull/11893#discussion_r1009622594
##
lucene/analysis/common/src/java/org/apache/lucene/analysis/hunspell/Hunspell.java:
##
@@ -647,8 +671,23 @@ Root findStem(
if (!hasGoodSuggestions && dictionar
ovalhub commented on issue #10665:
URL: https://github.com/apache/lucene/issues/10665#issuecomment-1297370722
On Mon, 31 Oct 2022, Benjamin Trent wrote:
> I tested PyLucene locally, comparing it to @msokolov's "batch" methodology
> (writing to disk and spinning up a Java proces
rmuir commented on issue #11887:
URL: https://github.com/apache/lucene/issues/11887#issuecomment-1297392725
The condition where this test takes minutes isn't that rare, I ran the test
10 times and hit the slow condition 3 out of 10 executions:
* 151s
* 158s
* 32s
With the pat
rmuir opened a new pull request, #11894:
URL: https://github.com/apache/lucene/pull/11894
The current test has ~ minute runtimes approximately 30% of the time.
Closes #11887
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitH
benwtrent commented on issue #10665:
URL: https://github.com/apache/lucene/issues/10665#issuecomment-1297430223
@ovalhub `numpy` collections are already native. To use them, I have to pull
them into python collections and then cast them to be native again.
Example:
```
X = X.to
benwtrent commented on issue #10665:
URL: https://github.com/apache/lucene/issues/10665#issuecomment-1297440415
> It'd also be great to compare the results against hnswlib as part of the
submission. We can double-check that recall is the same for a given set of
parameters. This would give c
ovalhub commented on issue #10665:
URL: https://github.com/apache/lucene/issues/10665#issuecomment-1297463120
On Mon, 31 Oct 2022, Benjamin Trent wrote:
> @ovalhub `numpy` collections are already native. To use them, I have to
> pull them into python collections and then cast t
donnerpeter commented on code in PR #11893:
URL: https://github.com/apache/lucene/pull/11893#discussion_r1009734992
##
lucene/analysis/common/src/java/org/apache/lucene/analysis/hunspell/Hunspell.java:
##
@@ -72,10 +77,29 @@ public Hunspell(Dictionary dictionary) {
* or
benwtrent commented on code in PR #11860:
URL: https://github.com/apache/lucene/pull/11860#discussion_r1009762545
##
lucene/core/src/java/org/apache/lucene/codecs/lucene95/Lucene95HnswVectorsReader.java:
##
@@ -0,0 +1,505 @@
+/*
+ * Licensed to the Apache Software Foundation (AS
jtibshirani commented on code in PR #11860:
URL: https://github.com/apache/lucene/pull/11860#discussion_r1009765612
##
lucene/backward-codecs/src/java/org/apache/lucene/backward_codecs/lucene94/package-info.java:
##
@@ -0,0 +1,422 @@
+/*
+ * Licensed to the Apache Software Found
jtibshirani commented on PR #239:
URL: https://github.com/apache/lucene/pull/239#issuecomment-1297711772
@harishankar-gopalan sorry for the slow response! Your overall understanding
is right. In Lucene, deletions are handled by marking a document as deleted
using a 'tombstone'. The index st
jpountz commented on PR #11875:
URL: https://github.com/apache/lucene/pull/11875#issuecomment-1298076313
Sorry for the lag I'm on vacation.
The problem with "this class may be useful outside of Lucene" to me is that
it could apply to any class in Lucene. We did indeed make some classe
30 matches
Mail list logo
