LibCpp: Lex before processing the source in the Preprocessor

Previously, the preprocessor first split the source into lines, and then
processed and lexed each line separately.

This patch makes the preprocessor first lex the source, and then do the
processing on the tokenized representation.

This generally simplifies the code, and also fixes an issue we
previously had with multiline comments (we did not recognize them
correctly when processing each line separately).

Userland/Libraries/LibCpp/Preprocessor.cpp

@@ -16,47 +16,48 @@ Preprocessor::Preprocessor(const String& filename, const StringView& program)
     : m_filename(filename)
     , m_program(program)
 {
-    GenericLexer program_lexer { m_program };
-    for (;;) {
-        if (program_lexer.is_eof())
-            break;
-        auto line = program_lexer.consume_until('\n');
-        bool has_multiline = false;
-        while (line.ends_with('\\') && !program_lexer.is_eof()) {
-            auto continuation = program_lexer.consume_until('\n');
-            line = StringView { line.characters_without_null_termination(), line.length() + continuation.length() + 1 };
-            // Append an empty line to keep the line count correct.
-            m_lines.append({});
-            has_multiline = true;
-        }
-
-        if (has_multiline)
-            m_lines.last() = line;
-        else
-            m_lines.append(line);
-    }
 }
 
 Vector<Token> Preprocessor::process_and_lex()
 {
-    for (; m_line_index < m_lines.size(); ++m_line_index) {
+    Lexer lexer { m_program };
-        auto& line = m_lines[m_line_index];
+    lexer.set_ignore_whitespace(true);
+    auto tokens = lexer.lex();
 
-        bool include_in_processed_text = false;
+    for (size_t token_index = 0; token_index < tokens.size(); ++token_index) {
-        if (line.starts_with("#")) {
+        auto& token = tokens[token_index];
-            auto keyword = handle_preprocessor_line(line);
+        m_current_line = token.start().line;
-            if (m_options.keep_include_statements && keyword == "include")
+        if (token.type() == Token::Type::PreprocessorStatement) {
-                include_in_processed_text = true;
+            handle_preprocessor_statement(token.text());
-        } else if (m_state == State::Normal) {
+            continue;
-            include_in_processed_text = true;
         }
 
-        if (include_in_processed_text) {
+        if (m_state != State::Normal)
-            process_line(line);
+            continue;
+
+        if (token.type() == Token::Type::IncludeStatement) {
+            if (token_index >= tokens.size() - 1 || tokens[token_index + 1].type() != Token::Type::IncludePath)
+                continue;
+            handle_include_statement(tokens[token_index + 1].text());
+            if (m_options.keep_include_statements) {
+                m_processed_tokens.append(tokens[token_index]);
+                m_processed_tokens.append(tokens[token_index + 1]);
+            }
+            continue;
         }
+
+        if (token.type() == Token::Type::Identifier) {
+            if (auto defined_value = m_definitions.find(token.text()); defined_value != m_definitions.end()) {
+                auto last_substituted_token_index = do_substitution(tokens, token_index, defined_value->value);
+                token_index = last_substituted_token_index;
+                continue;
+            }
+        }
+
+        m_processed_tokens.append(token);
     }
 
-    return m_tokens;
+    return m_processed_tokens;
 }
 
 static void consume_whitespace(GenericLexer& lexer)
@@ -85,7 +86,7 @@ static void consume_whitespace(GenericLexer& lexer)
     }
 }
 
-Preprocessor::PreprocessorKeyword Preprocessor::handle_preprocessor_line(const StringView& line)
+void Preprocessor::handle_preprocessor_statement(StringView const& line)
 {
     GenericLexer lexer(line);
 
@@ -94,23 +95,25 @@ Preprocessor::PreprocessorKeyword Preprocessor::handle_preprocessor_line(const S
     consume_whitespace(lexer);
     auto keyword = lexer.consume_until(' ');
     if (keyword.is_empty() || keyword.is_null() || keyword.is_whitespace())
-        return {};
+        return;
 
     handle_preprocessor_keyword(keyword, lexer);
-    return keyword;
+}
+
+void Preprocessor::handle_include_statement(StringView const& include_path)
+{
+    m_included_paths.append(include_path);
+    if (definitions_in_header_callback) {
+        for (auto& def : definitions_in_header_callback(include_path))
+            m_definitions.set(def.key, def.value);
+    }
 }
 
 void Preprocessor::handle_preprocessor_keyword(const StringView& keyword, GenericLexer& line_lexer)
 {
     if (keyword == "include") {
-        consume_whitespace(line_lexer);
+        // Should have called 'handle_include_statement'.
-        auto include_path = line_lexer.consume_all();
+        VERIFY_NOT_REACHED();
-        m_included_paths.append(include_path);
-        if (definitions_in_header_callback) {
-            for (auto& def : definitions_in_header_callback(include_path))
-                m_definitions.set(def.key, def.value);
-        }
-        return;
     }
 
     if (keyword == "else") {
@@ -218,25 +221,6 @@ void Preprocessor::handle_preprocessor_keyword(const StringView& keyword, Generi
     }
 }
 
-void Preprocessor::process_line(StringView const& line)
-{
-    Lexer line_lexer { line, m_line_index };
-    line_lexer.set_ignore_whitespace(true);
-    auto tokens = line_lexer.lex();
-
-    for (size_t i = 0; i < tokens.size(); ++i) {
-        auto& token = tokens[i];
-        if (token.type() == Token::Type::Identifier) {
-            if (auto defined_value = m_definitions.find(token.text()); defined_value != m_definitions.end()) {
-                auto last_substituted_token_index = do_substitution(tokens, i, defined_value->value);
-                i = last_substituted_token_index;
-                continue;
-            }
-        }
-        m_tokens.append(token);
-    }
-}
-
 size_t Preprocessor::do_substitution(Vector<Token> const& tokens, size_t token_index, Definition const& defined_value)
 {
     if (defined_value.value.is_null())
@@ -265,7 +249,7 @@ size_t Preprocessor::do_substitution(Vector<Token> const& tokens, size_t token_i
             continue;
         token.set_start(original_tokens.first().start());
         token.set_end(original_tokens.first().end());
-        m_tokens.append(token);
+        m_processed_tokens.append(token);
     }
     return macro_call->end_token_index;
 }
@@ -322,7 +306,7 @@ Optional<Preprocessor::Definition> Preprocessor::create_definition(StringView li
 
     Definition definition;
     definition.filename = m_filename;
-    definition.line = m_line_index;
+    definition.line = m_current_line;
 
     definition.key = tokens.first().text();
 
@@ -356,11 +340,21 @@ Optional<Preprocessor::Definition> Preprocessor::create_definition(StringView li
     }
 
     if (token_index < tokens.size())
-        definition.value = line.substring_view(tokens[token_index].start().column);
+        definition.value = remove_escaped_newlines(line.substring_view(tokens[token_index].start().column));
 
     return definition;
 }
 
+String Preprocessor::remove_escaped_newlines(StringView const& value)
+{
+    AK::StringBuilder processed_value;
+    GenericLexer lexer { value };
+    while (!lexer.is_eof()) {
+        processed_value.append(lexer.consume_until("\\\n"));
+    }
+    return processed_value.to_string();
+}
+
 String Preprocessor::evaluate_macro_call(MacroCall const& macro_call, Definition const& definition)
 {
     if (macro_call.arguments.size() != definition.parameters.size()) {

Userland/Libraries/LibCpp/Preprocessor.h

@@ -49,10 +49,11 @@ public:
     Function<Definitions(StringView)> definitions_in_header_callback { nullptr };
 
 private:
-    using PreprocessorKeyword = StringView;
+    void handle_preprocessor_statement(StringView const&);
-    PreprocessorKeyword handle_preprocessor_line(StringView const&);
+    void handle_include_statement(StringView const&);
     void handle_preprocessor_keyword(StringView const& keyword, GenericLexer& line_lexer);
-    void process_line(StringView const& line);
+    String remove_escaped_newlines(StringView const& value);
+
     size_t do_substitution(Vector<Token> const& tokens, size_t token_index, Definition const&);
     Optional<Definition> create_definition(StringView line);
 
@@ -69,13 +70,12 @@ private:
 
     String m_filename;
     String m_program;
-    Vector<StringView> m_lines;
 
-    Vector<Token> m_tokens;
+    Vector<Token> m_processed_tokens;
     Definitions m_definitions;
     Vector<Substitution> m_substitutions;
 
-    size_t m_line_index { 0 };
+    size_t m_current_line { 0 };
     size_t m_current_depth { 0 };
     Vector<size_t> m_depths_of_taken_branches;
     Vector<size_t> m_depths_of_not_taken_branches;