[WASM] Add more complete handling toLowerCase/toUpperCase for non-str…

…ingref. This is closer to original Android based implementation but delegates to JS in difficult cases instead of ICU4C. PiperOrigin-RevId: 519809176
google · Mar 27, 2023 · 24a7b79 · 24a7b79
1 parent 5752855
commit 24a7b79
Show file tree

Hide file tree

Showing 3 changed files with 30 additions and 2 deletions.
diff --git a/jre/java/super-wasm-alt/java/lang/CaseMapper.java b/jre/java/super-wasm-alt/java/lang/CaseMapper.java
@@ -27,6 +27,9 @@ class CaseMapper {
       "\u000b\u0000\f\u0000\r"
           + "\u0000\u000e\u0000\u0000\u0000\u0000\u0000\u0000\u0000\u0000\u0000\u0000\u0000\u0000\u0000\u0000\u0000\u0000\u0000\u0000\u0000\u0000\u0000\u0000\u0000\u0000\u0000\u0000\u0000\u0000\u0000\u0000\u0000\u0000\u0000\u0000\u0000\u0000\u0000\u0000\u0000\u0000\u0000\u000f\u0010\u0011\u0012\u0013\u0014\u0015\u0016\u0017\u0018\u0019\u001a\u001b\u001c\u001d\u001e\u001f!\"#$%&'()*+,-./0123456789:;<=>\u0000\u0000?@A\u0000BC\u0000\u0000\u0000\u0000D\u0000\u0000\u0000\u0000\u0000EFG\u0000HI\u0000\u0000\u0000\u0000J\u0000\u0000\u0000\u0000\u0000KL\u0000\u0000MN\u0000\u0000\u0000\u0000\u0000\u0000\u0000\u0000\u0000\u0000OPQ\u0000RS\u0000\u0000\u0000\u0000\u0000\u0000\u0000\u0000\u0000\u0000TUV\u0000WX\u0000\u0000\u0000\u0000Y";
 
+  private static final char LATIN_CAPITAL_I_WITH_DOT = '\u0130';
+  private static final char GREEK_CAPITAL_SIGMA = '\u03a3';
+
   /*
    * Our current GC makes short-lived objects more expensive than we'd like. When that's fixed, this
    * class should be changed so that you instantiate it with the String and its value, offset, and
@@ -44,6 +47,12 @@ public static String toLowerCase(String s, char[] value, int offset, int count)
     int newCount = 0;
     for (int i = offset, end = offset + count; i < end; ++i) {
       char ch = value[i];
+      if (ch == LATIN_CAPITAL_I_WITH_DOT
+          || ch == GREEK_CAPITAL_SIGMA
+          || Character.isHighSurrogate(ch)) {
+        // Punt these hard cases.
+        return String.fromJsString(s.toJsString().toLowerCase());
+      }
       char newCh = charToLowerCase(ch);
       if (newValue == null && ch != newCh) {
         newValue = new char[count]; // The result can't be longer than the input.
@@ -107,6 +116,10 @@ public static String toUpperCase(String s, char[] value, int offset, int count)
     int i = 0;
     for (int o = offset, end = offset + count; o < end; o++) {
       char ch = value[o];
+      if (Character.isHighSurrogate(ch)) {
+        // Punt these hard cases.
+        return String.fromJsString(s.toJsString().toUpperCase());
+      }
       int index = upperIndex(ch);
       if (index == -1) {
         if (output != null && i >= output.length) {

diff --git a/jre/java/super-wasm-alt/java/lang/String.java b/jre/java/super-wasm-alt/java/lang/String.java
@@ -1009,6 +1009,10 @@ public static class NativeString {
 
     native NativeString replace(WasmExtern regex, NativeString replace);
 
+    native NativeString toLowerCase();
+
+    native NativeString toUpperCase();
+
     native NativeString toLocaleLowerCase();
 
     native NativeString toLocaleUpperCase();

diff --git a/jre/javatests/com/google/j2cl/jre/java/lang/StringTest.java b/jre/javatests/com/google/j2cl/jre/java/lang/StringTest.java
@@ -738,7 +738,12 @@ public void testLowerCase() {
   }
 
   public void testLowerCaseNonAscii() {
-    assertEquals("öçşğü", hideFromCompiler("ÖÇŞĞÜ").toLowerCase()); // a.k.a "Turkey Test"
+    assertEquals("i̇öçşğü", hideFromCompiler("İÖÇŞĞÜ").toLowerCase()); // a.k.a "Turkey Test"
+
+    // Greek sigma
+    assertEquals("\u03C3", hideFromCompiler("\u03A3").toLowerCase());
+    assertEquals("abc\u03C2", hideFromCompiler("ABC\u03A3").toLowerCase());
+    assertEquals("abc\u03C3abc", hideFromCompiler("ABC\u03A3ABC").toLowerCase());
   }
 
   public void testMatch() {
@@ -1064,7 +1069,13 @@ public void testUpperCase() {
   }
 
   public void testUpperCaseNonAscii() {
-    assertEquals("ÖÇŞĞÜ", hideFromCompiler("öçşğü").toUpperCase()); // a.k.a "Turkey Test"
+    assertEquals("İÖÇŞĞÜ", hideFromCompiler("i̇öçşğü").toUpperCase()); // a.k.a "Turkey Test"
+
+    assertEquals("SS", hideFromCompiler("ß").toUpperCase());
+    assertEquals("ʼN", hideFromCompiler("ŉ").toUpperCase());
+
+    // surrogate example
+    assertEquals("\uD801\uDC1c", hideFromCompiler("\uD801\uDC44").toUpperCase());
   }
 
   /*