Allow identifier code points from supplementary multilingual planes.

src/char-predicates.h

Index: src/char-predicates.h
diff --git a/src/char-predicates.h b/src/char-predicates.h
index bfe7fe18bf6f048ee85d57d1fb38134ea2b0035c..5ecb07de992a6f62b1ad605ee97ce7f81c04a624 100644
--- a/src/char-predicates.h
+++ b/src/char-predicates.h
@@ -22,13 +22,24 @@ inline bool IsBinaryDigit(uc32 c);
 inline bool IsRegExpWord(uc32 c);
 inline bool IsRegExpNewline(uc32 c);
 
+
+struct SupplementaryPlanes {
+  static bool IsIDStart(uc32 c);
+  static bool IsIDPart(uc32 c);
+};
+
+
 // ES6 draft section 11.6
 // This includes '_', '$' and '\', and ID_Start according to
 // http://www.unicode.org/reports/tr31/, which consists of categories
 // 'Lu', 'Ll', 'Lt', 'Lm', 'Lo', 'Nl', but excluding properties
 // 'Pattern_Syntax' or 'Pattern_White_Space'.
+// For code points in the SMPs, we can resort to ICU (if available).
 struct IdentifierStart {
-  static inline bool Is(uc32 c) { return unibrow::ID_Start::Is(c); }
+  static inline bool Is(uc32 c) {
+    if (c > 0xFFFF) return SupplementaryPlanes::IsIDStart(c);
+    return unibrow::ID_Start::Is(c);
+  }
 };
 
 
@@ -37,8 +48,10 @@ struct IdentifierStart {
 // http://www.unicode.org/reports/tr31/, which consists of ID_Start,
 // the categories 'Mn', 'Mc', 'Nd', 'Pc', but excluding properties
 // 'Pattern_Syntax' or 'Pattern_White_Space'.
+// For code points in the SMPs, we can resort to ICU (if available).
 struct IdentifierPart {
   static inline bool Is(uc32 c) {
+    if (c > 0xFFFF) return SupplementaryPlanes::IsIDPart(c);
     return unibrow::ID_Start::Is(c) || unibrow::ID_Continue::Is(c);
   }
 };
@@ -49,6 +62,7 @@ struct IdentifierPart {
 // \u180e stops being one as of Unicode 6.3.0, but ES6 adheres to Unicode 5.1,
 // so it is also included.
 // Further included are \u0009, \u000b, \u0020, \u00a0, \u000c, and \ufeff.
+// There are no category 'Zs' code points in the SMPs.
 struct WhiteSpace {
   static inline bool Is(uc32 c) { return unibrow::WhiteSpace::Is(c); }
 };