ICU-2906 make unescapeAt() handle escaped surrogate pairs

X-SVN-Rev: 13342

icu4c/source/common/ustring.c

@@ -1481,6 +1481,20 @@ u_unescapeAt(UNESCAPE_CHAR_AT charAt,
         if (result < 0 || result >= 0x110000) {
             goto err;
         }
+        // If an escape sequence specifies a lead surrogate, see if
+        // there is a trail surrogate after it, either as an escape or
+        // as a literal.  If so, join them up into a supplementary.
+        if (*offset < length && U16_IS_LEAD(result)) {
+            int32_t ahead = *offset + 1;
+            c = charAt(*offset, context);
+            if (c == 0x5C /*'\\'*/ && ahead < length) {
+                c = (UChar) u_unescapeAt(charAt, &ahead, length, context);
+            }
+            if (U16_IS_TRAIL(c)) {
+                *offset = ahead;
+                result = U16_GET_SUPPLEMENTARY(result, c);
+            }
+        }
         return result;
     }
 

icu4c/source/test/intltest/usettest.cpp

@@ -58,6 +58,7 @@ UnicodeSetTest::runIndexedTest(int32_t index, UBool exec,
         CASE(16,TestEscapePattern);
         CASE(17,TestInvalidCodePoint);
         CASE(18,TestSymbolTable);
+        CASE(19,TestSurrogate);
         default: name = ""; break;
     }
 }
@@ -1259,6 +1260,32 @@ void UnicodeSetTest::TestSymbolTable() {
     }
 }
 
+void UnicodeSetTest::TestSurrogate() {
+    const char* DATA[] = {
+        // These should all behave identically
+        "[abc\\uD800\\uDC00]",
+        // "[abc\uD800\uDC00]", // Can't do this on C -- only Java
+        "[abc\\U00010000]",
+        0
+    };
+    for (int i=0; DATA[i] != 0; ++i) {
+        UErrorCode ec = U_ZERO_ERROR;
+        logln((UnicodeString)"Test pattern " + i + " :" + DATA[i]);
+        UnicodeSet set(DATA[i], ec);
+        if (U_FAILURE(ec)) {
+            errln("FAIL: UnicodeSet constructor");
+            continue;
+        }
+        expectContainment(set,
+                          CharsToUnicodeString("abc\\U00010000"),
+                          CharsToUnicodeString("\\uD800;\\uDC00")); // split apart surrogate-pair
+        if (set.size() != 4) {
+            errln((UnicodeString)"FAIL: " + DATA[i] + ".size() == " + 
+                  set.size() + ", expected 4");
+        }
+    }
+}
+
 void UnicodeSetTest::TestExhaustive() {
     // exhaustive tests. Simulate UnicodeSets with integers.
     // That gives us very solid tests (except for large memory tests).

icu4c/source/test/intltest/usettest.h

@@ -72,6 +72,8 @@ private:
 
     void TestSymbolTable(void);
 
+    void TestSurrogate();
+
 private:
 
     UBool toPatternAux(UChar32 start, UChar32 end);