Fix unicode issue which caused wrong emit start/end

https://github.com/robert-bor/aho-corasick/issues/39

- Some unicode characters when lowercased in a string became more than a character
- This leads to problems in the proper size (start, end) in Emit
- Fix builds upon fix of https://github.com/robert-bor/aho-corasick/issues/8

src/main/java/org/ahocorasick/trie/Trie.java

@@ -32,13 +32,10 @@ public class Trie {
             return;
         }
         State currentState = this.rootState;
-        for (Character character : keyword.toCharArray()) {
-            if (trieConfig.isCaseInsensitive()) {
-                character = Character.toLowerCase(character);
-            }
+        String caseAdjustedKeyword = adjustCase(keyword).toString();
+        for (Character character : caseAdjustedKeyword.toCharArray())
             currentState = currentState.addState(character);
-        }
-        currentState.addEmit(trieConfig.isCaseInsensitive() ? keyword.toLowerCase() : keyword);
+        currentState.addEmit(caseAdjustedKeyword);
     }
 
     public Collection<Token> tokenize(String text) {
@@ -99,11 +96,9 @@ public class Trie {
 
     public void parseText(CharSequence text, EmitHandler emitHandler) {
         State currentState = this.rootState;
-        for (int position = 0; position < text.length(); position++) {
-            Character character = text.charAt(position);
-            if (trieConfig.isCaseInsensitive()) {
-                character = Character.toLowerCase(character);
-            }
+        CharSequence caseAdjustedText = adjustCase(text);
+        for (int position = 0; position < caseAdjustedText.length(); position++) {
+            Character character = caseAdjustedText.charAt(position);
             currentState = getState(currentState, character);
             if (storeEmits(position, currentState, emitHandler) && trieConfig.isStopOnHit()) {
                 return;
@@ -122,11 +117,9 @@ public class Trie {
         } else {
             // Fast path. Returns first match found.
             State currentState = this.rootState;
-            for (int position = 0; position < text.length(); position++) {
-                Character character = text.charAt(position);
-				if (trieConfig.isCaseInsensitive()) {
-					character = Character.toLowerCase(character);
-				}
+            CharSequence caseAdjustedText = adjustCase(text);
+            for (int position = 0; position < caseAdjustedText.length(); position++) {
+                Character character = caseAdjustedText.charAt(position);
                 currentState = getState(currentState, character);
                 Collection<String> emitStrs = currentState.emit();
                 if (emitStrs != null && !emitStrs.isEmpty()) {
@@ -229,6 +222,17 @@ public class Trie {
         return emitted;
     }
 
+    private CharSequence adjustCase(CharSequence text) {
+        if(trieConfig.isCaseInsensitive()){
+            char[] textChars = text.toString().toCharArray();
+            char[] adjustedTextChars = new char[textChars.length];
+            for(int i = 0; i < textChars.length; i++)
+                adjustedTextChars[i] = Character.toLowerCase(textChars[i]);
+            return new String(adjustedTextChars);
+        }
+        return text;
+    }
+
     public static TrieBuilder builder() {
         return new TrieBuilder();
     }

src/test/java/org/ahocorasick/trie/TrieTest.java

@@ -415,6 +415,23 @@ public class TrieTest {
         checkEmit(firstMatch, 5, 8, "this");
     }
 
+    @Test
+    public void unicodeIssueBug39ReportedByHumanzz(){
+        // Problem: "İ".length => 1, "İ".toLowerCase().length => 2. This causes all sorts of unexpected behaviors
+        // and bugs where the Emit will have a size different from the original string.
+        // Soln: As in issue #8, convert at character level Character.toLowerCase('İ') => 'i'  + make sure
+        // that emit gets the properly cased keyword.
+        String upperLengthOne = "İnt";
+        Trie trie = Trie.builder()
+                .caseInsensitive()
+                .onlyWholeWords()
+                .addKeyword(upperLengthOne)
+                .build();
+        Collection<Emit> emits = trie.parseText("İnt is good");
+        assertEquals(1, emits.size());
+        checkEmit(emits.iterator().next(), 0, 2, "int");
+    }
+
     @Test
     public void partialMatchWhiteSpaces() {
         Trie trie = Trie.builder()