utf8.h: Remove a branch in macro for Unicode surrogates

By masking, this macro can be written so it only has one branch.
Presumably an optimizing compiler would do the same, but not necessarily
so.

diff --git a/utf8.h b/utf8.h
index cc22942..cd7e5bc 100644 (file)
--- a/utf8.h
+++ b/utf8.h
@@ -662,8 +662,11 @@ case any call to string overloading updates the internal UTF-8 encoding flag.
 #define UNICODE_ALLOW_SUPER    0
 #define UNICODE_ALLOW_ANY      0
 
-#define UNICODE_IS_SURROGATE(uv)        ((UV) (uv) >= UNICODE_SURROGATE_FIRST && \
-                                         (UV) (uv) <= UNICODE_SURROGATE_LAST)
+/* This matches the 2048 code points between UNICODE_SURROGATE_FIRST (0xD800) and
+ * UNICODE_SURROGATE_LAST (0xDFFF) */
+#define UNICODE_IS_SURROGATE(uv)        (((UV) (uv) & (~0xFFFF | 0xF800))       \
+                                                                    == 0xD800)
+
 #define UNICODE_IS_REPLACEMENT(uv)     ((UV) (uv) == UNICODE_REPLACEMENT)
 #define UNICODE_IS_BYTE_ORDER_MARK(uv) ((UV) (uv) == UNICODE_BYTE_ORDER_MARK)
 #define UNICODE_IS_NONCHAR(uv)         (((UV) (uv) >= 0xFDD0 && (UV) (uv) <= 0xFDEF) \