Faster implementation of utf8::advance not copying iterators at each step

src/utf8.hh

@@ -17,6 +17,14 @@ template<typename Iterator>
 [[gnu::always_inline]]
 inline char read(Iterator& it) { char c = *it; ++it; return c; }
 
+// return true if it points to the first byte of a (either single or
+// multibyte) character
+[[gnu::always_inline]]
+inline bool is_character_start(char c)
+{
+    return (c & 0xC0) != 0x80;
+}
+
 // returns an iterator to next character first byte
 template<typename Iterator>
 Iterator next(Iterator it, const Iterator& end)
@@ -52,27 +60,28 @@ Iterator previous(Iterator it, const Iterator& begin)
 template<typename Iterator>
 Iterator advance(Iterator it, const Iterator& end, CharCount d)
 {
+    if (it == end)
+        return it;
+
     if (d < 0)
     {
-       while (it != end and d++)
+        while (it != end and d != 0)
-           it = utf8::previous(it, end);
-    }
-    else
         {
-        while (it != end and d--)
+            if (is_character_start(*--it))
-           it = utf8::next(it, end);
+                ++d;
+        }
+    }
+    else if (d > 0)
+    {
+        while (it != end and d != 0)
+        {
+            if (is_character_start(*++it))
+                --d;
+        }
     }
     return it;
 }
 
-// return true if it points to the first byte of a (either single or
-// multibyte) character
-[[gnu::always_inline]]
-inline bool is_character_start(char c)
-{
-    return (c & 0xC0) != 0x80;
-}
-
 // returns the character count between begin and end
 template<typename Iterator>
 CharCount distance(Iterator begin, const Iterator& end)