Skip to content

Latest commit

 

History

History
93 lines (75 loc) · 5.84 KB

CHANGELOG.rst

File metadata and controls

93 lines (75 loc) · 5.84 KB
Raw

ifpress-solr-plugin CHANGELOG

  • 1.7.0 - Upgrade to Solr v8.11.3 - METACON-2476:

    • TermContext has been renamed to TermStates.
    • StandardFilter and StandardFilterFactory have been removed.
    • Deprecated method IndexSearcher#createNormalizedWeight() has been removed and replaced with createWeight() adding a booster of 1.

  • 1.6.0

    • Upgrade to Solr v7.6.0
    • SOLR-10700: PostingsHighlighter and its related classes (deprecated) replaced with the suggested UnifiedHighlighter.
      • UnifiedHighlighter has options to work almost exactly like PostingsHighlighter did, as Unified was a more performant extension of Postings.
      • UnifiedHighlighterTest: UnifiedHighlighter object creation requires an IndexAnalyzer, using SafariAnalyzer.
      • SolrUnifiedHighlighterTest: Unlike Postings, UnifiedHighlighter unifies matching sentences near each other, despite the SENTENCE break type. It relies on 'fragsize' to determine to return multiple sentences of just the highest matching sentence.
    • LUCENE-7867: analysis.Token (deprecated in Solr 2) moved to solr.spelling as that should be the only place it is used.
    • UpdateDocValuesProcessor: LUCENE-7407, NumericDocValues uses an iterator, must call iterator method advanceExact to iterate to desired doc value.
    • SafariQueryParserTest: - Single clause BooleanQueries no longer wrapped with an additional Occur.MUST - LUCENE-7369: BooleanQuery's disableCoords field was removed
    • SOLR-10585: defaultSearchField (deprecated in Solr 3) removed from schema use, using 'df' instead.
    • SOLR-10584: Setting <solrQueryParser defaultOperator="..."/> in schema now causes an exception, must use suggested "q.op" parameter on the request instead

  • 1.5.0

    • Upgrade to Solr v6.6.5

    • Removed SafariBlockJoinQuery and ScoringParentQParser. They were replaced with field collapsing for SPIDR-962

    • Update to UpdateDocValuesProcessor: - SolrIndexSearcher's getLeafReader() replaced with getSlowAtomicReader(), which is not recommend to use. Instead, leafContexts are used instead to get the leaf reader, only calling getSlowAtomicReader() if necessary.

    • Change to SafariQueryParserTest to handle an update to SafariQueryParser's super class ExtendedDismaxQParser: - SafariQueryParser now wraps BooleanQueries with an additional Occur.MUST (+) clause. Lucene 6.5.0 brought in this change: https://issues.apache.org/jira/browse/SOLR-9185 The whitespace change that this ticket focuses on does not directly change anything in SafariQueryParser.

      The condition statements in QueryParser.java lines 232-238 were changed, which causes the extra Boolean clause. This change should not have an impact on SafariQueryParser (wrapping original result with an extra MUST boolean clause does not change functionality) and it is impossible to override because QueryParser.java is such a deep, internal class.

      Call stack for this change:

      ExtendedDismaxQParser.parse() > ExtendedDismaxQParser.parseOriginalQuery() > SolrQueryParserBaseClass.parse() > QueryParser.TopLevelQuery() > QueryParser.Query()

  • 1.4.0

    • Upgrade to Solr v5.5.5. All changes in this version were necessary for Solr v5.5.5 compatibility.

    • solrconfig.xml: - Disabled updateLog because of issue introduced with TransactionLog. https://issues.apache.org/jira/browse/SOLR-8866 TransactionLog's resolver now throws an exception (previous versions it would return the object) for Lucene's TextField: "org.apache.solr.common.SolrException: TransactionLog doesn't know how to serialize class org.apache.lucene.document.TextField; try implementing ObjectResolver?" The ONLY way around this is to make TextField implement ObjectResolver, but TextField is a Lucene internal class. AND we cannot create a custom class that Extends TextField (and Implements ObjectResolver) because TextField is marked final (cannot create subclasses of final classes). - AdminHandlers no longer needed to be defined (redundant).

    • schema.xml: EdgeNGram's 'side=front' functionality was removed

    • HitCount.java: parse method handles Query.extractTerms being replaced with Weight.extractTerms

    • MultiSuggester.java: commit method now uses BytesRefBuilder instead of BytesRef

    • SafariInfixSuggester.java: No longer need to specify the Lucene version unless want to use a different version than current.

    • ScoringParentQParser.java (changes were similar to the parser used for Lucene's ToParentBlockJoin): - Can no longer use FixedBitSetCachingWrapperFilter, QueryBitSetProducer recommended to use instead. - FixedBitSet, BitDocIdSet, and Filter classes went through major class inheritance change. - Created a BitDocIdSetFilterWrapper to be able to use QueryBitSetProducer with filtering.

    • SafariBlockJoinQuery.java: - Changes to Explanations. - Iterator methods needed wrapped in an Iterator class. - A custom TwoPhaseIterator needed for filtering - Uses getLiveDocs() to check for deleted docs since acceptDocs can no longer be used.

    • SafariBlockJoinTest.java: Small issue with filter query on block join, not sure if even possible to resolve. SBJQ used to check 'acceptDocs' to make sure parent doc was not deleted or filtered out.

      Lucene v5.3.0 includes a huge rework on 'acceptDocs' for LUCENE-6553, removing 'acceptDocs' from Weight class.

      SBJQ no longer has access to 'acceptDocs'. It was forced to use reader.getLiveDocs() instead to make sure the parent docs were not deleted.

      Because of this, SBJQ can filter out children of deleted parent docs (see testOrphanedDocs) but does not filter out children of parents who do not meet filter query criteria.

  • 1.3.10

    • Provides more meaningful query explanations to SafariBlockJoinQuery

  • 1.3.9

    • Was the latest version used 2015 - June 2018
    • Adds checks for ArrayIndexOutOfBoundsException in SBJQ