Merge branch 'utfconversionfixsize' into 'master'

Use fixed-size type to hold a whole code point

See merge request OpenMW/openmw!4163

CHANGELOG.md

@@ -247,6 +247,7 @@
     Task #5896: Do not use deprecated MyGUI properties
     Task #6085: Replace boost::filesystem with std::filesystem
     Task #6149: Dehardcode Lua API_REVISION
+    Task #6505: UTF-8 support in Lua scripts
     Task #6624: Drop support for saves made prior to 0.45
     Task #7048: Get rid of std::bind
     Task #7113: Move from std::atoi to std::from_char

components/lua/utf8.cpp

@@ -6,8 +6,8 @@
 namespace
 {
     constexpr std::string_view UTF8PATT = "[%z\x01-\x7F\xC2-\xF4][\x80-\xBF]*"; // %z is deprecated in Lua5.2
-    constexpr uint32_t MAXUTF = 0x7FFFFFFFu;
-    // constexpr uint32_t MAXUNICODE = 0x10FFFFu;
+    // constexpr uint32_t MAXUTF = 0x7FFFFFFFu;
+    constexpr uint32_t MAXUNICODE = 0x10FFFFu;
 
     inline bool isNilOrNone(const sol::stack_proxy arg)
     {
@@ -96,16 +96,15 @@ namespace LuaUtf8
 
         utf8["char"] = [](const sol::variadic_args args) -> std::string {
             std::string result{};
-            std::wstring_convert<std::codecvt_utf8<wchar_t>> converter;
+            std::wstring_convert<std::codecvt_utf8<char32_t>, char32_t> converter;
             for (size_t i = 0; i < args.size(); ++i)
             {
                 int64_t codepoint = getInteger(args[i], (i + 1), "char");
-                if (codepoint < 0 || codepoint > MAXUTF)
+                if (codepoint < 0 || codepoint > MAXUNICODE)
                     throw std::runtime_error(
                         "bad argument #" + std::to_string(i + 1) + " to 'char' (value out of range)");
 
-                // this feels dodgy if wchar_t is 16-bit as MAXUTF won't fit in sixteen bits
-                result += converter.to_bytes(static_cast<wchar_t>(codepoint));
+                result += converter.to_bytes(static_cast<char32_t>(codepoint));
             }
             return result;
         };

scripts/data/integration_tests/test_lua_api/test.lua

@@ -1,4 +1,4 @@
-local testing = require('testing_util')
+local testing = require('testing_util')
 local core = require('openmw.core')
 local async = require('openmw.async')
 local util = require('openmw.util')
@@ -150,6 +150,42 @@ local function testRecordCreation()
         testing.expectEqual(record[key],value)
     end
 end
+local function testUTF8()
+    local utf8char = "😀"
+    local utf8str = "Hello, 你好, 🌎!"
+    local chars = {}
+
+    for codepoint = 0, 0x10FFFF do
+        local char = utf8.char(codepoint)
+        local charSize = string.len(char)
+
+        testing.expect(not chars[char], nil, "Duplicate UTF-8 character: " .. char)
+        chars[char] = true
+
+        if codepoint <= 0x7F then
+            testing.expectEqual(charSize, 1)
+        elseif codepoint <= 0x7FF then
+            testing.expectEqual(charSize, 2)
+        elseif codepoint <= 0xFFFF then
+            testing.expectEqual(charSize, 3)
+        elseif codepoint <= 0x10FFFF then
+            testing.expectEqual(charSize, 4)
+        end
+
+        testing.expectEqual(utf8.codepoint(char), codepoint)
+        testing.expectEqual(utf8.len(char), 1)
+    end
+
+    local str = ""
+    for utf_char in utf8str:gmatch(utf8.charpattern) do
+        str = str .. utf_char
+    end
+    testing.expectEqual(str, utf8str)
+
+    testing.expectEqual(utf8.codepoint(utf8char), 128512)
+    testing.expectEqual(utf8.len(utf8str), 13)
+    testing.expectEqual(utf8.offset(utf8str, 9), 11)
+end
 local function initPlayer()
     player:teleport('', util.vector3(4096, 4096, 867.237), util.transform.identity)
     coroutine.yield()
@@ -189,6 +225,7 @@ tests = {
     {'getGMST', testGetGMST},
     {'recordStores', testRecordStores},
     {'recordCreation', testRecordCreation},
+    {'utf8', testUTF8},
     {'mwscript', testMWScript},
 }