Lucene/Solr 9; Java 17; Tomcat 10 #526
Closed
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This contains the more "interesting" migrations (more than just a removed function or import path) - poms updated to use latest version of jackson/jersey/xml-bind (might miss a runtime dependency still). Servlet-api still on v4 due to packaged jetty server in SOLR being stuck on v4. - implement most simple (oneliner) migrations, should be doublechecked - one remaining compilation error in BLSpanOrQuery that should be solved.
This gives it access to the package-private class used by SpanOrQuery. Might seem like a hack, but these classes are so intertwined with how Lucene works internally that they might as well live in the same package. Arguably most subclasses of BLSpanQuery and BLSpans should be moved to the same package eventually. We might need to still double-check that BLSpanTermQuery and BLSpanMultiTermQueryWrapper are up to date with their Lucene 9 counterparts.
update
1 solr test failing
exclude solr module untill jakarta compatible version is out
# Conflicts: # engine/pom.xml # engine/src/main/java/nl/inl/blacklab/indexers/config/saxon/XPathFinder.java # engine/src/main/java/nl/inl/blacklab/search/lucene/SpansCaptureRelationsWithinSpan.java
no try catch needed LoggingWatcher
…riment/tomcat-10 # Conflicts: # engine/pom.xml # engine/src/main/java/nl/inl/blacklab/codec/BlackLab40StoredFieldsReader.java # engine/src/main/java/nl/inl/blacklab/search/SingleDocIdFilter.java # engine/src/main/java/nl/inl/blacklab/search/lucene/BLConjunctionSpans.java # engine/src/main/java/nl/inl/blacklab/search/lucene/BLConjunctionSpansInBuckets.java # engine/src/main/java/nl/inl/blacklab/search/lucene/BLFilterDocsSpans.java # engine/src/main/java/nl/inl/blacklab/search/lucene/BLSpanMultiTermQueryWrapper.java # engine/src/main/java/nl/inl/blacklab/search/lucene/BLSpanQuery.java # engine/src/main/java/nl/inl/blacklab/search/lucene/SpanQueryAnyToken.java # engine/src/main/java/nl/inl/blacklab/search/lucene/SpanQueryNoHits.java # engine/src/main/java/nl/inl/blacklab/search/lucene/SpanQuerySequence.java # engine/src/main/java/nl/inl/blacklab/search/lucene/SpansAnd.java # engine/src/main/java/nl/inl/blacklab/search/lucene/SpansFiltered.java # engine/src/main/java/nl/inl/blacklab/search/results/HitsInternal.java # engine/src/main/java/nl/inl/util/DocValuesUtil.java # engine/src/main/java/org/apache/lucene/queries/spans/BLSpanOrQuery.java # pom.xml # proxy/jaxb/src/main/java/org/ivdnt/blacklab/proxy/representation/ParsePatternResponse.java # proxy/jaxb/src/main/java/org/ivdnt/blacklab/proxy/representation/SummaryTextPattern.java # solr/pom.xml # solr/src/main/java/org/ivdnt/blacklab/solr/DocSetFilter.java # wslib/src/main/java/nl/inl/blacklab/server/exceptions/BadRequest.java # wslib/src/main/java/nl/inl/blacklab/server/lib/Response.java
correct analyzers version solr.XSLTResponseWriter not found
correct analyzers version solr.XSLTResponseWriter not found
full classname for xsltresponsewriter new lucene version => rebuild index in test resources jakarta needed in solr tests
|
After making sure BlackLab40Codec uses Lucene87 as the delegate codec (previously it requested the default codec from Lucene, which is obvisouly different in Lucene 9), we still have a problem reading back our custom terms file. The cause seems to be that DataInput/DataOutput have been switched to little-endian. Presumably we need to update our custom codec to explicitly read big-endian, which is how these indexes were written. We should also add a new codec version, e.g. BlackLab41, that will use the new 9.x Lucene codec as delegate (or even adapts, using the default when creating, and recording what delegate codec was used so it can be reinstantiated when reading later). That version of the codec can use little-endian for the custom terms files as well. |
When reading an index written using Lucene 8, we need to make sure we use EndiannessReverserUtil when opening the file, because Lucene 9 switched to little-endian for DataInput/Output.
API documentation was grouped by subject. BLS configuration docs were updated and expanded. Squashed commit of the following: commit 0a0cf33 Author: Jan Niestadt <[email protected]> Date: Wed Jul 9 09:03:52 2025 +0200 Redirect. commit 269459a Author: Jan Niestadt <[email protected]> Date: Tue Jun 24 16:02:39 2025 +0200 Configuration, more. commit 876bb1b Author: Jan Niestadt <[email protected]> Date: Mon Jun 23 14:40:20 2025 +0200 Typo commit 639cdd0 Author: Jan Niestadt <[email protected]> Date: Thu Jun 19 15:34:00 2025 +0200 Work on API docs. commit 8e8125a Author: Jan Niestadt <[email protected]> Date: Wed Jun 18 15:25:34 2025 +0200 Use Docker tag dev instead of latest. commit d95bffe Author: Jan Niestadt <[email protected]> Date: Tue Jun 17 16:27:50 2025 +0200 Restructure API reference. commit c7b1baa Author: Jan Niestadt <[email protected]> Date: Tue Jun 17 13:26:52 2025 +0200 API v4/5.
Values larger than 32K cannot be indexed in Lucene, so we truncate them even if no maxValueLength was set. Created WarnOnce to be able to issue warnings only once during indexing (where a problem may often occur many times, flooding the logs).
EphemeralHit's instance variables had the same name as the getter methods, which could be confusing. Reduced access to package private as well, and only used direct access to variables in HitsInternal classes. In other classes, we use the getter methods. The JVM should normally inline these anyway for hot code.
When you catch this exception and don't re-throw it, you should re-set the thread's interrupted flag so the status is not lost.
|
|
I've created a new branch |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Thanks to @eduarddrenth for this branch, which updates our previous experiment and solves more issues.
CURRENT STATUS: working, experimental. Will probably be merged in after releasing v4 soon
Old comments: