[analysis-icu] Allow setting unicodeSetFilter (#20814)

UnicodeSetFilter was only allowed in the icu_folding token filter.
It seems useful to expose this setting in icu_normalizer token filter
and char filter.

docs/plugins/analysis-icu.asciidoc

@@ -37,6 +37,10 @@ normalization can be specified with the `name` parameter, which accepts `nfc`,
 `nfkc`, and `nfkc_cf` (default).  Set the `mode` parameter to `decompose` to
 convert `nfc` to `nfd` or `nfkc` to `nfkd` respectively:
 
+Which letters are normalized can be controlled by specifying the
+`unicodeSetFilter` parameter, which accepts a
+http://icu-project.org/apiref/icu4j/com/ibm/icu/text/UnicodeSet.html[UnicodeSet].
+
 Here are two examples, the default usage and a customised character filter:
 
 
@@ -189,6 +193,10 @@ without any further configuration. The type of normalization can be specified
 with the `name` parameter, which accepts `nfc`, `nfkc`, and `nfkc_cf`
 (default).
 
+Which letters are normalized can be controlled by specifying the
+`unicodeSetFilter` parameter, which accepts a
+http://icu-project.org/apiref/icu4j/com/ibm/icu/text/UnicodeSet.html[UnicodeSet].
+
 You should probably prefer the <<analysis-icu-normalization-charfilter,Normalization character filter>>.
 
 Here are two examples, the default usage and a customised token filter:

plugins/analysis-icu/src/main/java/org/elasticsearch/index/analysis/IcuFoldingTokenFilterFactory.java

@@ -19,9 +19,8 @@
 
 package org.elasticsearch.index.analysis;
 
-import com.ibm.icu.text.FilteredNormalizer2;
 import com.ibm.icu.text.Normalizer2;
-import com.ibm.icu.text.UnicodeSet;
+
 import org.apache.lucene.analysis.TokenStream;
 import org.apache.lucene.analysis.icu.ICUFoldingFilter;
 import org.elasticsearch.common.settings.Settings;
@@ -41,31 +40,20 @@ import org.elasticsearch.index.IndexSettings;
  * @author kimchy (shay.banon)
  */
 public class IcuFoldingTokenFilterFactory extends AbstractTokenFilterFactory implements MultiTermAwareComponent {
-    private final String unicodeSetFilter;
+    /** Store here the same Normalizer used by the lucene ICUFoldingFilter */
+    private static final Normalizer2 ICU_FOLDING_NORMALIZER = Normalizer2.getInstance(
+            ICUFoldingFilter.class.getResourceAsStream("utr30.nrm"), "utr30", Normalizer2.Mode.COMPOSE);
+
+    private final Normalizer2 normalizer;
 
     public IcuFoldingTokenFilterFactory(IndexSettings indexSettings, Environment environment, String name, Settings settings) {
         super(indexSettings, name, settings);
-        this.unicodeSetFilter = settings.get("unicodeSetFilter");
+        this.normalizer = IcuNormalizerTokenFilterFactory.wrapWithUnicodeSetFilter(ICU_FOLDING_NORMALIZER, settings);
     }
 
     @Override
     public TokenStream create(TokenStream tokenStream) {
-
-        // The ICUFoldingFilter is in fact implemented as a ICUNormalizer2Filter.
-        // ICUFoldingFilter lacks a constructor for adding filtering so we implemement it here
-        if (unicodeSetFilter != null) {
-            Normalizer2 base = Normalizer2.getInstance(
-                    ICUFoldingFilter.class.getResourceAsStream("utr30.nrm"),
-                    "utr30", Normalizer2.Mode.COMPOSE);
-            UnicodeSet unicodeSet = new UnicodeSet(unicodeSetFilter);
-
-            unicodeSet.freeze();
-            Normalizer2 filtered = new FilteredNormalizer2(base, unicodeSet);
-            return new org.apache.lucene.analysis.icu.ICUNormalizer2Filter(tokenStream, filtered);
-        }
-        else {
-            return new ICUFoldingFilter(tokenStream);
-        }
+        return new org.apache.lucene.analysis.icu.ICUNormalizer2Filter(tokenStream, normalizer);
     }
 
     @Override

plugins/analysis-icu/src/main/java/org/elasticsearch/index/analysis/IcuNormalizerCharFilterFactory.java

@@ -21,6 +21,7 @@ package org.elasticsearch.index.analysis;
 
 
 import com.ibm.icu.text.Normalizer2;
+
 import org.apache.lucene.analysis.icu.ICUNormalizer2CharFilter;
 import org.elasticsearch.common.settings.Settings;
 import org.elasticsearch.env.Environment;
@@ -33,22 +34,22 @@ import java.io.Reader;
  * Uses the {@link org.apache.lucene.analysis.icu.ICUNormalizer2CharFilter} to normalize character.
  * <p>The <tt>name</tt> can be used to provide the type of normalization to perform.</p>
  * <p>The <tt>mode</tt> can be used to provide 'compose' or 'decompose'. Default is compose.</p>
+ * <p>The <tt>unicodeSetFilter</tt> attribute can be used to provide the UniCodeSet for filtering.</p>
  */
 public class IcuNormalizerCharFilterFactory extends AbstractCharFilterFactory implements MultiTermAwareComponent {
 
-    private final String name;
-
     private final Normalizer2 normalizer;
 
     public IcuNormalizerCharFilterFactory(IndexSettings indexSettings, Environment environment, String name, Settings settings) {
         super(indexSettings, name);
-        this.name = settings.get("name", "nfkc_cf");
+        String method = settings.get("name", "nfkc_cf");
         String mode = settings.get("mode");
         if (!"compose".equals(mode) && !"decompose".equals(mode)) {
             mode = "compose";
         }
-        this.normalizer = Normalizer2.getInstance(
-            null, this.name, "compose".equals(mode) ? Normalizer2.Mode.COMPOSE : Normalizer2.Mode.DECOMPOSE);
+        Normalizer2 normalizer = Normalizer2.getInstance(
+            null, method, "compose".equals(mode) ? Normalizer2.Mode.COMPOSE : Normalizer2.Mode.DECOMPOSE);
+        this.normalizer = IcuNormalizerTokenFilterFactory.wrapWithUnicodeSetFilter(normalizer, settings);
     }
 
     @Override

plugins/analysis-icu/src/main/java/org/elasticsearch/index/analysis/IcuNormalizerTokenFilterFactory.java

@@ -19,7 +19,10 @@
 
 package org.elasticsearch.index.analysis;
 
+import com.ibm.icu.text.FilteredNormalizer2;
 import com.ibm.icu.text.Normalizer2;
+import com.ibm.icu.text.UnicodeSet;
+
 import org.apache.lucene.analysis.TokenStream;
 import org.elasticsearch.common.settings.Settings;
 import org.elasticsearch.env.Environment;
@@ -28,26 +31,40 @@ import org.elasticsearch.index.IndexSettings;
 
 /**
  * Uses the {@link org.apache.lucene.analysis.icu.ICUNormalizer2Filter} to normalize tokens.
- * <p>The <tt>name</tt> can be used to provide the type of normalization to perform.
+ * <p>The <tt>name</tt> can be used to provide the type of normalization to perform.</p>
+ * <p>The <tt>unicodeSetFilter</tt> attribute can be used to provide the UniCodeSet for filtering.</p>
  *
  *
  */
 public class IcuNormalizerTokenFilterFactory extends AbstractTokenFilterFactory implements MultiTermAwareComponent {
 
-    private final String name;
+    private final Normalizer2 normalizer;
 
     public IcuNormalizerTokenFilterFactory(IndexSettings indexSettings, Environment environment, String name, Settings settings) {
         super(indexSettings, name, settings);
-        this.name = settings.get("name", "nfkc_cf");
+        String method = settings.get("name", "nfkc_cf");
+        Normalizer2 normalizer = Normalizer2.getInstance(null, method, Normalizer2.Mode.COMPOSE);
+        this.normalizer = wrapWithUnicodeSetFilter(normalizer, settings);
     }
 
     @Override
     public TokenStream create(TokenStream tokenStream) {
-        return new org.apache.lucene.analysis.icu.ICUNormalizer2Filter(tokenStream, Normalizer2.getInstance(null, name, Normalizer2.Mode.COMPOSE));
+        return new org.apache.lucene.analysis.icu.ICUNormalizer2Filter(tokenStream, normalizer);
     }
 
     @Override
     public Object getMultiTermComponent() {
         return this;
     }
+
+    static Normalizer2 wrapWithUnicodeSetFilter(final Normalizer2 normalizer, Settings settings) {
+        String unicodeSetFilter = settings.get("unicodeSetFilter");
+        if (unicodeSetFilter != null) {
+            UnicodeSet unicodeSet = new UnicodeSet(unicodeSetFilter);
+
+            unicodeSet.freeze();
+            return new FilteredNormalizer2(normalizer, unicodeSet);
+        }
+        return normalizer;
+    }
 }

plugins/analysis-icu/src/test/resources/rest-api-spec/test/analysis_icu/10_basic.yml

@@ -39,3 +39,50 @@
             tokenizer:    keyword
     - length: { tokens: 1 }
     - match:  { tokens.0.token: foo bar resume }
+---
+"Normalization with a UnicodeSet Filter":
+    - do:
+        indices.create:
+            index:  test
+            body:
+                settings:
+                    index:
+                        analysis:
+                            char_filter:
+                                charfilter_icu_normalizer:
+                                    type: icu_normalizer
+                                    unicodeSetFilter: "[^ß]"
+                            filter:
+                                tokenfilter_icu_normalizer:
+                                    type: icu_normalizer
+                                    unicodeSetFilter: "[^ßB]"
+                                tokenfilter_icu_folding:
+                                    type: icu_folding
+                                    unicodeSetFilter: "[^â]"
+    - do:
+        indices.analyze:
+          index:    test
+          body:
+            char_filter: ["charfilter_icu_normalizer"]
+            tokenizer:  keyword
+            text:     charfilter Föo Bâr Ruß
+    - length: { tokens: 1 }
+    - match:  { tokens.0.token: charfilter föo bâr ruß }
+    - do:
+        indices.analyze:
+          index:    test
+          body:
+            tokenizer:  keyword
+            filter: ["tokenfilter_icu_normalizer"]
+            text:     tokenfilter Föo Bâr Ruß
+    - length: { tokens: 1 }
+    - match:  { tokens.0.token: tokenfilter föo Bâr ruß }
+    - do:
+        indices.analyze:
+          index:    test
+          body:
+            tokenizer:  keyword
+            filter: ["tokenfilter_icu_folding"]
+            text:     icufolding Föo Bâr Ruß
+    - length: { tokens: 1 }
+    - match:  { tokens.0.token: icufolding foo bâr russ }