修复上游 issues 第一批：TypesettingTools#491 TypesettingTools#446 TypesettingTools#443 TypesettingTools#538

- 修复TypesettingTools#491: Lua删除崩溃，ObjectIndexWrite delete分支修正Lua栈操作
- 修复TypesettingTools#446: 样式重命名损坏，StyleRenamer添加modified_current标志
- 修复TypesettingTools#443: margin解析崩溃，std::from_chars替代boost::lexical_cast
- 修复TypesettingTools#538: CPS字符计数错误，IGNORE_BLOCKS模式使用TokenizeDialogueBody
- 新增6个回归测试用例验证#538修复

Author: mojie126

libaegisub/common/character_count.cpp

@@ -84,24 +84,26 @@ size_t CharacterCount(std::string_view str, int ignore) {
 	if ((ignore & agi::IGNORE_BLOCKS) == 0)
 		return count_in_range(str, mask);
 
+	// 使用分词器正确处理 \N、\n、\h 等 ASS 转义序列
+	auto tokens = agi::ass::TokenizeDialogueBody(str);
+	agi::ass::MarkDrawings(str, tokens);
+
 	size_t characters = 0;
-	while (!str.empty()) {
-		auto pos = str.find('{');
-		if (pos == str.npos) break;
-
-		// if there's no trailing }, the rest of the string counts as characters,
-		// including the leading {
-		auto end = str.find('}');
-		if (end == str.npos) break;
-
-		if (pos > 0)
-			characters += count_in_range(str.substr(0, pos), mask);
-		str.remove_prefix(end + 1);
+	size_t pos = 0;
+	for (auto token : tokens) {
+		if (token.type == agi::ass::DialogueTokenType::TEXT)
+			characters += count_in_range(str.substr(pos, token.length), mask);
+		else if (token.type == agi::ass::DialogueTokenType::LINE_BREAK) {
+			// \h 是硬空格，非忽略空白时计为 1 个字符
+			if (str[pos + 1] == 'h') {
+				if (!(mask & U_GC_Z_MASK))
+					characters += 1;
+			}
+			// \N 和 \n 是换行符，不计入字符数
+		}
+		pos += token.length;
 	}
 
-	if (!str.empty())
-		characters += count_in_range(str, mask);
-
 	return characters;
 }
 

src/ass_dialogue.cpp

@@ -36,6 +36,7 @@
 #include <libaegisub/string.h>
 
 #include <algorithm>
+#include <charconv>
 
 #include <boost/algorithm/string/predicate.hpp>
 #include <boost/lexical_cast.hpp>
@@ -111,8 +112,12 @@ void AssDialogue::Parse(std::string const& raw) {
 	End = agi::Trim(tkn.next_tok());
 	Style = tkn.next_str_trim();
 	Actor = tkn.next_str_trim();
-	for (int& margin : Margin)
-		margin = std::clamp(boost::lexical_cast<int>(tkn.next_tok()), -9999, 99999);
+	for (int& margin : Margin) {
+		auto tok = agi::Trim(tkn.next_tok());
+		int val = 0;
+		std::from_chars(tok.data(), tok.data() + tok.size(), val);
+		margin = std::clamp(val, -9999, 99999);
+	}
 	Effect = tkn.next_str_trim();
 
 	std::string text{tkn.next_tok().begin(), str.end()};

src/auto4_lua_assfile.cpp

@@ -478,9 +478,9 @@ namespace Automation4 {
 				AssignLine(n - 1, std::move(e));
 			}
 			else {
-				// delete
-				lua_remove(L, 1);
-				lua_remove(L, 1);
+				// delete: 清空栈后压入索引，供 ObjectDelete 使用
+				lua_settop(L, 0);
+				lua_pushinteger(L, n);
 				ObjectDelete(L);
 			}
 		}

src/dialog_style_editor.cpp

@@ -114,15 +114,18 @@ class StyleRenamer {
 	agi::Context *c;
 	bool found_any = false;
 	bool do_replace = false;
+	bool modified_current = false; ///< 当前行的 \r 标签是否被修改
 	std::string source_name;
 	std::string new_name;
 
-	/// Process a single override parameter to check if it's \r with this style name
+	/// @brief 处理单个覆写参数，检查是否为引用目标样式的 \r 标签
 	static void ProcessTag(std::string const& tag, AssOverrideParameter* param, void *userData) {
 		StyleRenamer *self = static_cast<StyleRenamer*>(userData);
 		if (tag == "\\r" && param->GetType() == VariableDataType::TEXT && param->Get<std::string>() == self->source_name) {
-			if (self->do_replace)
+			if (self->do_replace) {
 				param->Set(self->new_name);
+				self->modified_current = true;
+			}
 			else
 				self->found_any = true;
 		}
@@ -140,10 +143,11 @@ class StyleRenamer {
 					found_any = true;
 			}
 
+			modified_current = false;
 			auto blocks = diag.ParseTags();
 			for (auto block : blocks | agi::of_type<AssDialogueBlockOverride>())
 				block->ProcessParameters(&StyleRenamer::ProcessTag, this);
-			if (replace)
+			if (replace && modified_current)
 				diag.UpdateText(blocks);
 
 			if (found_any) return;

tests/tests/character_count.cpp

@@ -67,6 +67,21 @@ TEST(lagi_character_count, ignore_blocks_unclosed) {
 	EXPECT_EQ(6, agi::CharacterCount("{hello", agi::IGNORE_BLOCKS));
 }
 
+/// \brief 验证 IGNORE_BLOCKS 模式下 ASS 转义序列的正确处理 (#538)
+TEST(lagi_character_count, ignore_blocks_line_breaks) {
+	// \N 和 \n 是换行符，不计入字符数
+	EXPECT_EQ(10, agi::CharacterCount("{\\b1}hello\\Nworld", agi::IGNORE_BLOCKS));
+	EXPECT_EQ(10, agi::CharacterCount("{\\b1}hello\\nworld", agi::IGNORE_BLOCKS));
+	// \h 是硬空格，计为 1 个字符
+	EXPECT_EQ(11, agi::CharacterCount("{\\b1}hello\\hworld", agi::IGNORE_BLOCKS));
+	// 忽略空白时 \h 不计入
+	EXPECT_EQ(10, agi::CharacterCount("{\\b1}hello\\hworld", agi::IGNORE_BLOCKS | agi::IGNORE_WHITESPACE));
+	// 多个连续换行
+	EXPECT_EQ(10, agi::CharacterCount("{\\b1}hello\\N\\Nworld", agi::IGNORE_BLOCKS));
+	// 文本中混合块和换行
+	EXPECT_EQ(10, agi::CharacterCount("{\\b1}hello{\\i1}\\Nworld", agi::IGNORE_BLOCKS));
+}
+
 TEST(lagi_character_count, line_length) {
 	EXPECT_EQ(5, agi::MaxLineLength("hello", agi::IGNORE_NONE));
 	EXPECT_EQ(5, agi::MaxLineLength("hello\\Nasdf", agi::IGNORE_NONE));