use Unicode normalization in String.prototype.localeCompare

2 files changed, 100 insertions, 33 deletions

diff --git a/quickjs.c b/quickjs.c
index c80a300..dfcacda 100644
--- a/quickjs.c
+++ b/quickjs.c
@@ -41109,26 +41109,6 @@ static BOOL test_final_sigma(JSString *p, int sigma_pos)
     return !lre_is_cased(c1);
 }
 
-static JSValue js_string_localeCompare(JSContext *ctx, JSValueConst this_val,
-                                       int argc, JSValueConst *argv)
-{
-    JSValue a, b;
-    int cmp;
-
-    a = JS_ToStringCheckObject(ctx, this_val);
-    if (JS_IsException(a))
-        return JS_EXCEPTION;
-    b = JS_ToString(ctx, argv[0]);
-    if (JS_IsException(b)) {
-        JS_FreeValue(ctx, a);
-        return JS_EXCEPTION;
-    }
-    cmp = js_string_compare(ctx, JS_VALUE_GET_STRING(a), JS_VALUE_GET_STRING(b));
-    JS_FreeValue(ctx, a);
-    JS_FreeValue(ctx, b);
-    return JS_NewInt32(ctx, cmp);
-}
-
 static JSValue js_string_toLowerCase(JSContext *ctx, JSValueConst this_val,
                                      int argc, JSValueConst *argv, int to_lower)
 {
@@ -41214,23 +41194,38 @@ static JSValue JS_NewUTF32String(JSContext *ctx, const uint32_t *buf, int len)
     return JS_EXCEPTION;
 }
 
+static int js_string_normalize1(JSContext *ctx, uint32_t **pout_buf, 
+                                JSValueConst val,
+                                UnicodeNormalizationEnum n_type)
+{
+    int buf_len, out_len;
+    uint32_t *buf, *out_buf;
+    
+    buf_len = JS_ToUTF32String(ctx, &buf, val);
+    if (buf_len < 0)
+        return -1;
+    out_len = unicode_normalize(&out_buf, buf, buf_len, n_type,
+                                ctx->rt, (DynBufReallocFunc *)js_realloc_rt);
+    js_free(ctx, buf);
+    if (out_len < 0)
+        return -1;
+    *pout_buf = out_buf;
+    return out_len;
+}
+
 static JSValue js_string_normalize(JSContext *ctx, JSValueConst this_val,
                                    int argc, JSValueConst *argv)
 {
     const char *form, *p;
     size_t form_len;
-    int is_compat, buf_len, out_len;
+    int is_compat, out_len;
     UnicodeNormalizationEnum n_type;
     JSValue val;
-    uint32_t *buf, *out_buf;
+    uint32_t *out_buf;
 
     val = JS_ToStringCheckObject(ctx, this_val);
     if (JS_IsException(val))
         return val;
-    buf_len = JS_ToUTF32String(ctx, &buf, val);
-    JS_FreeValue(ctx, val);
-    if (buf_len < 0)
-        return JS_EXCEPTION;
 
     if (argc == 0 || JS_IsUndefined(argv[0])) {
         n_type = UNICODE_NFC;
@@ -41256,22 +41251,96 @@ static JSValue js_string_normalize(JSContext *ctx, JSValueConst this_val,
             JS_FreeCString(ctx, form);
             JS_ThrowRangeError(ctx, "bad normalization form");
         fail1:
-            js_free(ctx, buf);
+            JS_FreeValue(ctx, val);
             return JS_EXCEPTION;
         }
         JS_FreeCString(ctx, form);
     }
 
-    out_len = unicode_normalize(&out_buf, buf, buf_len, n_type,
-                                ctx->rt, (DynBufReallocFunc *)js_realloc_rt);
-    js_free(ctx, buf);
+    out_len = js_string_normalize1(ctx, &out_buf, val, n_type);
+    JS_FreeValue(ctx, val);
     if (out_len < 0)
         return JS_EXCEPTION;
     val = JS_NewUTF32String(ctx, out_buf, out_len);
     js_free(ctx, out_buf);
     return val;
 }
-#endif /* CONFIG_ALL_UNICODE */
+
+/* return < 0, 0 or > 0 */
+static int js_UTF32_compare(const uint32_t *buf1, int buf1_len,
+                            const uint32_t *buf2, int buf2_len)
+{
+    int i, len, c, res;
+    len = min_int(buf1_len, buf2_len);
+    for(i = 0; i < len; i++) {
+        /* Note: range is limited so a subtraction is valid */
+        c = buf1[i] - buf2[i];
+        if (c != 0)
+            return c;
+    }
+    if (buf1_len == buf2_len)
+        res = 0;
+    else if (buf1_len < buf2_len)
+        res = -1;
+    else
+        res = 1;
+    return res;
+}
+
+static JSValue js_string_localeCompare(JSContext *ctx, JSValueConst this_val,
+                                       int argc, JSValueConst *argv)
+{
+    JSValue a, b;
+    int cmp, a_len, b_len;
+    uint32_t *a_buf, *b_buf;
+    
+    a = JS_ToStringCheckObject(ctx, this_val);
+    if (JS_IsException(a))
+        return JS_EXCEPTION;
+    b = JS_ToString(ctx, argv[0]);
+    if (JS_IsException(b)) {
+        JS_FreeValue(ctx, a);
+        return JS_EXCEPTION;
+    }
+    a_len = js_string_normalize1(ctx, &a_buf, a, UNICODE_NFC);
+    JS_FreeValue(ctx, a);
+    if (a_len < 0) {
+        JS_FreeValue(ctx, b);
+        return JS_EXCEPTION;
+    }
+
+    b_len = js_string_normalize1(ctx, &b_buf, b, UNICODE_NFC);
+    JS_FreeValue(ctx, b);
+    if (b_len < 0) {
+        js_free(ctx, a_buf);
+        return JS_EXCEPTION;
+    }
+    cmp = js_UTF32_compare(a_buf, a_len, b_buf, b_len);
+    js_free(ctx, a_buf);
+    js_free(ctx, b_buf);
+    return JS_NewInt32(ctx, cmp);
+}
+#else /* CONFIG_ALL_UNICODE */
+static JSValue js_string_localeCompare(JSContext *ctx, JSValueConst this_val,
+                                       int argc, JSValueConst *argv)
+{
+    JSValue a, b;
+    int cmp;
+
+    a = JS_ToStringCheckObject(ctx, this_val);
+    if (JS_IsException(a))
+        return JS_EXCEPTION;
+    b = JS_ToString(ctx, argv[0]);
+    if (JS_IsException(b)) {
+        JS_FreeValue(ctx, a);
+        return JS_EXCEPTION;
+    }
+    cmp = js_string_compare(ctx, JS_VALUE_GET_STRING(a), JS_VALUE_GET_STRING(b));
+    JS_FreeValue(ctx, a);
+    JS_FreeValue(ctx, b);
+    return JS_NewInt32(ctx, cmp);
+}
+#endif /* !CONFIG_ALL_UNICODE */
 
 /* also used for String.prototype.valueOf */
 static JSValue js_string_toString(JSContext *ctx, JSValueConst this_val,
diff --git a/test262_errors.txt b/test262_errors.txt
index 77ee6cc..31eeb6e 100644
--- a/test262_errors.txt
+++ b/test262_errors.txt
@@ -5,8 +5,6 @@ test262/test/built-ins/RegExp/lookahead-quantifier-match-groups.js:27: Test262Er
 test262/test/built-ins/RegExp/lookahead-quantifier-match-groups.js:27: strict mode: Test262Error: Expected [a, abc] and [a, undefined] to have the same contents. ? quantifier
 test262/test/built-ins/RegExp/unicode_full_case_folding.js:20: Test262Error: \u0390 does not match \u1fd3
 test262/test/built-ins/RegExp/unicode_full_case_folding.js:20: strict mode: Test262Error: \u0390 does not match \u1fd3
-test262/test/built-ins/String/prototype/localeCompare/15.5.4.9_CE.js:62: Test262Error: String.prototype.localeCompare considers ö (\u006f\u0308) ≠ ö (\u00f6).
-test262/test/built-ins/String/prototype/localeCompare/15.5.4.9_CE.js:62: strict mode: Test262Error: String.prototype.localeCompare considers ö (\u006f\u0308) ≠ ö (\u00f6).
 test262/test/built-ins/TypedArray/prototype/sort/sort-tonumber.js:30: TypeError: ArrayBuffer is detached (Testing with Float64Array.)
 test262/test/built-ins/TypedArray/prototype/sort/sort-tonumber.js:30: strict mode: TypeError: ArrayBuffer is detached (Testing with Float64Array.)
 test262/test/language/expressions/assignment/target-member-computed-reference-null.js:32: Test262Error: Expected a DummyError but got a TypeError