From cd4fad54c0be9fb7fca57e8e03228b8b649b5b51 Mon Sep 17 00:00:00 2001
From: Roland Reichwein <mail@reichwein.it>
Date: Wed, 27 Jan 2021 22:21:04 +0100
Subject: Bugfixes, tests

---
 include/unicode.h | 40 +++++++++++++++++++++-------------------
 1 file changed, 21 insertions(+), 19 deletions(-)

(limited to 'include')

diff --git a/include/unicode.h b/include/unicode.h
index a55eac3..f539e6b 100644
--- a/include/unicode.h
+++ b/include/unicode.h
@@ -1,5 +1,4 @@
 // libunicode
-// Copyright (C) 2021 Roland Reichwein
 
 #pragma once
 
@@ -7,11 +6,20 @@
 #include <stdexcept>
 #include <string>
 
-#ifdef __has_cpp_attribute
-#if __has_cpp_attribute(__cpp_char8_t)
+#ifdef __cpp_char8_t
 // char8_t available
 #endif
-#endif
+
+namespace unicode {
+
+ // usually, char32_t, uint32_t etc.
+ template<typename T>
+ static inline bool is_valid_unicode(const T& value)
+ {
+   return value <= 0x10FFFF && (value <= 0xD7FF || value >= 0xE000);
+ }
+
+}
 
 namespace {
 
@@ -50,6 +58,8 @@ namespace {
   template<typename T1>
   void calculate_value()
   {
+   static_assert(sizeof(T1) == 4);
+
    size_t remaining{remaining_code_units()};
 
    if (!remaining)
@@ -57,7 +67,7 @@ namespace {
 
    value = get_code_unit<0>();
    
-   if (value > 0x10FFFF || (value > 0xD7FF && value < 0xE000))
+   if (!unicode::is_valid_unicode(value))
     throw std::invalid_argument("Invalid Unicode character: "s + std::to_string(static_cast<uint32_t>(value)));
 
    sequence_length = 1;
@@ -88,7 +98,7 @@ namespace {
   template<typename... Targs>
   inline static char32_t continuation_value(T b, Targs... Fargs)
   {
-   return continuation_value(b) << 6 | continuation_value(Fargs...);
+   return continuation_value(b) << (6 * sizeof...(Targs)) | continuation_value(Fargs...);
   }
 
   template<size_t n>
@@ -159,7 +169,7 @@ namespace {
     if ((unit0 & 0xFC00) != 0xD800 || (unit1 & 0xFC00) != 0xDC00)
      throw std::invalid_argument("Bad input: 2 malformed UTF-16 surrogates");
 
-    value = static_cast<char32_t>(unit0 & 0x03FF) << 10 | (unit1 & 0x03FF);
+    value = (static_cast<char32_t>(unit0 & 0x03FF) << 10 | (unit1 & 0x03FF)) + 0x10000;
     sequence_length = 2;
    }
   }
@@ -185,7 +195,7 @@ namespace {
   typename string_type::const_iterator iterator;
   typename string_type::const_iterator end_iterator;
 
-  value_type value{};
+  char32_t value{}; // always save complete unicode code point at this point
   size_t sequence_length{};
  };
 
@@ -276,8 +286,9 @@ namespace {
    if (value <= 0xFFFF) { // expect value to be already valid Unicode values
     s.push_back(value);
    } else {
-    s.push_back((value >> 10) + 0xD800);
-    s.push_back((value & 0x3FF) + 0xDC00);
+    char32_t value_reduced{value - 0x10000};
+    s.push_back((value_reduced >> 10) + 0xD800);
+    s.push_back((value_reduced & 0x3FF) + 0xDC00);
    }
    return *this;
   }
@@ -317,14 +328,5 @@ std::basic_string<To> utf_to_utf(const std::basic_string<From>& s)
  return result;
 }
 
-//std::u8string utf16_to_utf8(const std::u16string& s)
-//{
-// std::u8string result;
-//
-// std::transform(utf16_begin(s), utf16_end(s), std::back_inserter(result));
-//
-// return result;
-//}
-
 } // namespace unicode
 
-- 
cgit v1.2.3