Hibernate search sorting with collation

I upgraded Hibernate search from version - 4.3.0.Final to the latest stable version - 5.4.12.Final. All is good except sorting norwegian words. In the old version of hibernate there was SortField with locale in the constructor:

/** Creates a sort, possibly in reverse, by terms in the given field sorted
   * according to the given locale.
   * @param field  Name of field to sort by, cannot be <code>null</code>.
   * @param locale Locale of values in the field.
   */
  public SortField (String field, Locale locale, boolean reverse) {
    initFieldType(field, STRING);
    this.locale = locale;
    this.reverse = reverse;
  }

But in the new hibernate search SortField does not have locale. According to hibernate reference documentation (https://docs.jboss.org/hibernate/stable/search/reference/en-US/html_single/#_analysis) for sort words words in foreign languages we should use CollationKeyFilterFactory with normalizer. But there is no such class in this version of hibernate search. Maven pom:

<dependency>
   <groupId>org.hibernate</groupId>
   <artifactId>hibernate-search-orm</artifactId>
   <version>5.11.5.Final</version>
</dependency>

The question: What should I use/create in the hibernate search for sort norwegian words?

Now I have such sort order:

> atest, btest, ctest, ztest, åtest, ætest, øtest

The correct order:

> atest, btest, ctest, ztest, ætest, øtest, åtest

There is CollationKeyAnalyzer class, but I do not know how to use this for sorting:

  public final class CollationKeyAnalyzer extends Analyzer {
  private final CollationAttributeFactory factory;
  
  /**
   * Create a new CollationKeyAnalyzer, using the specified collator.
   *
   * @param collator CollationKey generator
   */
  public CollationKeyAnalyzer(Collator collator) {
    this.factory = new CollationAttributeFactory(collator);
  }

  @Override
  protected TokenStreamComponents createComponents(String fieldName) {
    KeywordTokenizer tokenizer = new KeywordTokenizer(factory, KeywordTokenizer.DEFAULT_BUFFER_SIZE);
    return new TokenStreamComponents(tokenizer, tokenizer);
  }
}

Very similar question without answer: https://stackoverflow.com/questions/39264308/how-to-do-case-insensitive-sorting-of-norwegian-characters-%c3%86-%c3%98-and-%c3%85-using-h

I'm not sure how much it helps you but the CollationKeyFilterFactory was deprecated and indeed removed.

In the class' Javadoc it says:

>Deprecated.
>use CollationKeyAnalyzer instead.

You can find the Javadoc here.

> But there is no such class in this version of hibernate search.

This part of the documentation looks obsolete, I'll look into updating it.

I found CollationKeyAnalyzer, but the javadoc states that it's obsolete and that ICUCollationKeyAnalyzer should be used instead.

Try adding this dependency to your POM:

<dependency>
   <groupId>org.apache.lucene</groupId>
   <artifactId>lucene-analyzers-icu</artifactId>
   <version>5.5.5</version>
</dependency>

Then create your own analyzer class that re-implements ICUCollationKeyAnalyzer with a hard-coded locale:

public class MyCollationKeyAnalyzer extends Analyzer {
    private final ICUCollationAttributeFactory factory;

    public MyCollationKeyAnalyzer(Version luceneVersion) {
        this.factory = new ICUCollationAttributeFactory( Collactor.getInstance( Locale.getInstance( "nb_NO" ) ) );
    }

    @Override
    protected TokenStreamComponents createComponents(String fieldName) {
        KeywordTokenizer tokenizer = new KeywordTokenizer(factory, KeywordTokenizer.DEFAULT_BUFFER_SIZE);
        return new TokenStreamComponents(tokenizer, tokenizer);
    }
}

Then create your field:

@Entity
@Indexed
public class MyEntity {

    // ...

    @Field(name = "title_sort", index = Index.NO, normalizer = @Normalizer(impl = MyCollationKeyAnalyzer.class))
    @SortableField(forField = "title_sort")
    private String title;

   // ...
}

Then sort on that field like this:

FullTextEntityManager ftEm = Search.getFullTextEntityManager( entityManager );
QueryBuilder qb = ...; // The usual
Query luceneQuery = ...; // The usual
FullTextQuery ftQuery = ftEm.createFullTextQuery( luceneQuery, MyEntity.class );
ftQuery.setSort( qb.sort().byField( "title_sort" ).createSort() );
ftQuery.setMaxResults( 20 );
List<MyEntity> hits = ftQuery.getResultList();

I didn't try this though, so let us know if it worked for you.

In order to fix sorting I created my own NorwegianCollationFactory. It is not perfect solution as I copied code from old version of Hibernate Search (IndexableBinaryStringTools.class), but it is working fine.<br>
NorwegianCollationFactory class:

import org.apache.lucene.analysis.TokenStream;
import org.apache.lucene.analysis.util.TokenFilterFactory;

import java.text.Collator;
import java.util.Locale;
import java.util.Map;

public final class NorwegianCollationFactory extends TokenFilterFactory {

    public NorwegianCollationFactory(Map<String, String> args) {
        super(args);
    }

    @Override
    public TokenStream create(TokenStream input) {
        Collator norwegianCollator = Collator.getInstance(new Locale("no", "NO"));
        return new CollationKeyFilter(input, norwegianCollator);
    }

}

CollationKeyFilter class:

import org.apache.lucene.analysis.TokenFilter;
import org.apache.lucene.analysis.TokenStream;
import org.apache.lucene.analysis.tokenattributes.CharTermAttribute;

import java.io.IOException;
import java.text.Collator;
import java.util.Objects;

public final class CollationKeyFilter extends TokenFilter {

    // This code is copied from IndexableBinaryStringTools.class from the old version of hibernate search  4.3.0.Final
    private static final CollationKeyFilter.CodingCase[] CODING_CASES = {
            new CollationKeyFilter.CodingCase(7, 1),
            new CollationKeyFilter.CodingCase(14, 6, 2),
            new CollationKeyFilter.CodingCase(13, 5, 3),
            new CollationKeyFilter.CodingCase(12, 4, 4),
            new CollationKeyFilter.CodingCase(11, 3, 5),
            new CollationKeyFilter.CodingCase(10, 2, 6),
            new CollationKeyFilter.CodingCase(9, 1, 7),
            new CollationKeyFilter.CodingCase(8, 0)
    };

    private final Collator collator;
    private final CharTermAttribute termAtt = addAttribute(CharTermAttribute.class);

    public CollationKeyFilter(TokenStream input, Collator collator) {
        super(input);
        this.collator = (Collator) collator.clone();
    }

    @Override
    public boolean incrementToken() throws IOException {
        if (input.incrementToken()) {
            byte[] collationKey = collator.getCollationKey(termAtt.toString()).toByteArray();
            int encodedLength = getBinaryStringEncodedLength(collationKey.length);
            termAtt.resizeBuffer(encodedLength);
            termAtt.setLength(encodedLength);
            encodeToBinaryString(collationKey, collationKey.length, termAtt.buffer());
            return true;
        } else {
            return false;
        }
    }

    // This code is copied from IndexableBinaryStringTools class from the old version of hibernate search  4.3.0.Final
    private void encodeToBinaryString(byte[] inputArray, int inputLength, char[] outputArray) {
        if (inputLength > 0) {
            int inputByteNum = 0;
            int caseNum = 0;
            int outputCharNum = 0;
            CollationKeyFilter.CodingCase codingCase;
            for (; inputByteNum + CODING_CASES[caseNum].numBytes <= inputLength; ++outputCharNum) {
                codingCase = CODING_CASES[caseNum];
                if (codingCase.numBytes == 2) {
                    outputArray[outputCharNum] = (char) (((inputArray[inputByteNum] & 0xFF) << codingCase.initialShift)
                            + (((inputArray[inputByteNum + 1] & 0xFF) >>> codingCase.finalShift) & codingCase.finalMask) & (short) 0x7FFF);
                } else {
                    outputArray[outputCharNum] = (char) (((inputArray[inputByteNum] & 0xFF) << codingCase.initialShift)
                            + ((inputArray[inputByteNum + 1] & 0xFF) << codingCase.middleShift)
                            + (((inputArray[inputByteNum + 2] & 0xFF) >>> codingCase.finalShift) & codingCase.finalMask) & (short) 0x7FFF);
                }
                inputByteNum += codingCase.advanceBytes;
                if (++caseNum == CODING_CASES.length) {
                    caseNum = 0;
                }
            }
            codingCase = CODING_CASES[caseNum];
            if (inputByteNum + 1 < inputLength) {
                outputArray[outputCharNum++] = (char) ((((inputArray[inputByteNum] & 0xFF) << codingCase.initialShift)
                        + ((inputArray[inputByteNum + 1] & 0xFF) << codingCase.middleShift)) & (short) 0x7FFF);
                outputArray[outputCharNum] = (char) 1;
            } else if (inputByteNum < inputLength) {
                outputArray[outputCharNum++] = (char) (((inputArray[inputByteNum] & 0xFF) << codingCase.initialShift) & (short) 0x7FFF);
                outputArray[outputCharNum] = caseNum == 0 ? (char) 1 : (char) 0;
            } else {
                outputArray[outputCharNum] = (char) 1;
            }
        }
    }

    // This code is copied from IndexableBinaryStringTools class from the old version of hibernate search 4.3.0.Final
    private int getBinaryStringEncodedLength(int inputLength) {
        return (int) ((8L * inputLength + 14L) / 15L) + 1;
    }

    // This code is copied from IndexableBinaryStringTools class from the old version of hibernate search 4.3.0.Final
    private static class CodingCase {
        int numBytes;
        int initialShift;
        int middleShift;
        int finalShift;
        int advanceBytes = 2;
        short middleMask;
        short finalMask;

        CodingCase(int initialShift, int middleShift, int finalShift) {
            this.numBytes = 3;
            this.initialShift = initialShift;
            this.middleShift = middleShift;
            this.finalShift = finalShift;
            this.finalMask = (short) ((short) 0xFF >>> finalShift);
            this.middleMask = (short) ((short) 0xFF << middleShift);
        }

        CodingCase(int initialShift, int finalShift) {
            this.numBytes = 2;
            this.initialShift = initialShift;
            this.finalShift = finalShift;
            this.finalMask = (short) ((short) 0xFF >>> finalShift);
            if (finalShift != 0) {
                advanceBytes = 1;
            }
        }
    }

    @Override
    public boolean equals(Object o) {
        if (this == o) {
            return true;
        }
        if (o == null || getClass() != o.getClass()) {
            return false;
        }
        if (!super.equals(o)) {
            return false;
        }
        CollationKeyFilter that = (CollationKeyFilter) o;
        return Objects.equals(collator, that.collator) &&
                Objects.equals(termAtt, that.termAtt);
    }

    @Override
    public int hashCode() {
        return Objects.hash(super.hashCode(), collator, termAtt);
    }

}

Entity mapping example:

@Entity
@NormalizerDef(name = "textSortNormalizer",
        filters = {
                @TokenFilterDef(factory = LowerCaseFilterFactory.class),
                @TokenFilterDef(factory = PatternReplaceFilterFactory.class, params = {
                        @Parameter(name = "pattern", value = "('-&\\.,\\(\\))"),
                        @Parameter(name = "replacement", value = " "),
                        @Parameter(name = "replace", value = "all")
                }),
                @TokenFilterDef(factory = PatternReplaceFilterFactory.class, params = {
                        @Parameter(name = "pattern", value = "([^0-9\\p{L} ])"),
                        @Parameter(name = "replacement", value = ""),
                        @Parameter(name = "replace", value = "all")
                }),
                @TokenFilterDef(factory = NorwegianCollationFactory.class)
        }
)
public class Entity {

    @Field(name = "name_for_sort", normalizer = @Normalizer(definition = "textSortNormalizer"))
    @SortableField(forField = "name_for_sort")
    private String name;

}