Ported support for raw string literals from c++ parser.

It is a features of GNU C and supported by gcc and clang.

Support is copied from tree-sitter/tree-sitter-cpp, scanner.c was copied
directly and I applied minimal changes on top of it to work with C
parser. And I ported pieces related to raw string literals from c++
grammar.js to C grammar.js with minimal changes to file.

grammar.js

@@ -35,6 +35,11 @@ const PREC = {
 module.exports = grammar({
   name: 'c',
 
+  externals: $ => [
+    $.raw_string_delimiter,
+    $.raw_string_content,
+  ],
+ 
   conflicts: $ => [
     [$.type_specifier, $._declarator],
     [$.type_specifier, $._declarator, $.macro_type_specifier],
@@ -977,9 +982,25 @@ module.exports = grammar({
 
     _string: $ => prec.left(choice(
       $.string_literal,
+      $.raw_string_literal,
       $.concatenated_string,
     )),
 
+    raw_string_literal: $ => seq(
+      choice('R"', 'LR"', 'uR"', 'UR"', 'u8R"'),
+      choice(
+        seq(
+          field('delimiter', $.raw_string_delimiter),
+          '(',
+          $.raw_string_content,
+          ')',
+          $.raw_string_delimiter,
+        ),
+        seq('(', $.raw_string_content, ')'),
+      ),
+      '"',
+    ),
+
     comma_expression: $ => seq(
       field('left', $.expression),
       ',',
@@ -1298,11 +1319,11 @@ module.exports = grammar({
     // Identifier is added to parse macros that are strings, like PRIu64.
     concatenated_string: $ => prec.right(seq(
       choice(
-        seq($.identifier, $.string_literal),
-        seq($.string_literal, $.string_literal),
-        seq($.string_literal, $.identifier),
+        seq($.identifier, choice($.string_literal, $.raw_string_literal)),
+        seq(choice($.string_literal, $.raw_string_literal), choice($.string_literal, $.raw_string_literal)),
+        seq(choice($.string_literal, $.raw_string_literal), $.identifier),
       ),
-      repeat(choice($.string_literal, $.identifier)),
+      repeat(choice($.string_literal, $.raw_string_literal, $.identifier)),
     )),
 
     string_literal: $ => seq(

src/scanner.c

@@ -0,0 +1,148 @@
+#include "tree_sitter/alloc.h"
+#include "tree_sitter/parser.h"
+
+#include <assert.h>
+#include <string.h>
+#include <wctype.h>
+
+enum TokenType { RAW_STRING_DELIMITER, RAW_STRING_CONTENT };
+
+/// The spec limits delimiters to 16 chars
+#define MAX_DELIMITER_LENGTH 16
+
+typedef struct {
+    uint8_t delimiter_length;
+    wchar_t delimiter[MAX_DELIMITER_LENGTH];
+} Scanner;
+
+static inline void advance(TSLexer *lexer) { lexer->advance(lexer, false); }
+
+static inline void reset(Scanner *scanner) {
+    scanner->delimiter_length = 0;
+    memset(scanner->delimiter, 0, sizeof scanner->delimiter);
+}
+
+/// Scan the raw string delimiter in R"delimiter(content)delimiter"
+static bool scan_raw_string_delimiter(Scanner *scanner, TSLexer *lexer) {
+    if (scanner->delimiter_length > 0) {
+        // Closing delimiter: must exactly match the opening delimiter.
+        // We already checked this when scanning content, but this is how we
+        // know when to stop. We can't stop at ", because R"""hello""" is valid.
+        for (int i = 0; i < scanner->delimiter_length; ++i) {
+            if (lexer->lookahead != scanner->delimiter[i]) {
+                return false;
+            }
+            advance(lexer);
+        }
+        reset(scanner);
+        return true;
+    }
+
+    // Opening delimiter: record the d-char-sequence up to (.
+    // d-char is any basic character except parens, backslashes, and spaces.
+    for (;;) {
+        if (scanner->delimiter_length >= MAX_DELIMITER_LENGTH || lexer->eof(lexer) || lexer->lookahead == '\\' ||
+            iswspace(lexer->lookahead)) {
+            return false;
+        }
+        if (lexer->lookahead == '(') {
+            // Rather than create a token for an empty delimiter, we fail and
+            // let the grammar fall back to a delimiter-less rule.
+            return scanner->delimiter_length > 0;
+        }
+        scanner->delimiter[scanner->delimiter_length++] = lexer->lookahead;
+        advance(lexer);
+    }
+}
+
+/// Scan the raw string content in R"delimiter(content)delimiter"
+static bool scan_raw_string_content(Scanner *scanner, TSLexer *lexer) {
+    // The progress made through the delimiter since the last ')'.
+    // The delimiter may not contain ')' so a single counter suffices.
+    for (int delimiter_index = -1;;) {
+        // If we hit EOF, consider the content to terminate there.
+        // This forms an incomplete raw_string_literal, and models the code
+        // well.
+        if (lexer->eof(lexer)) {
+            lexer->mark_end(lexer);
+            return true;
+        }
+
+        if (delimiter_index >= 0) {
+            if (delimiter_index == scanner->delimiter_length) {
+                if (lexer->lookahead == '"') {
+                    return true;
+                }
+                delimiter_index = -1;
+            } else {
+                if (lexer->lookahead == scanner->delimiter[delimiter_index]) {
+                    delimiter_index += 1;
+                } else {
+                    delimiter_index = -1;
+                }
+            }
+        }
+
+        if (delimiter_index == -1 && lexer->lookahead == ')') {
+            // The content doesn't include the )delimiter" part.
+            // We must still scan through it, but exclude it from the token.
+            lexer->mark_end(lexer);
+            delimiter_index = 0;
+        }
+
+        advance(lexer);
+    }
+}
+
+void *tree_sitter_c_external_scanner_create() {
+    Scanner *scanner = (Scanner *)ts_calloc(1, sizeof(Scanner));
+    memset(scanner, 0, sizeof(Scanner));
+    return scanner;
+}
+
+bool tree_sitter_c_external_scanner_scan(void *payload, TSLexer *lexer, const bool *valid_symbols) {
+    Scanner *scanner = (Scanner *)payload;
+
+    if (valid_symbols[RAW_STRING_DELIMITER] && valid_symbols[RAW_STRING_CONTENT]) {
+        // we're in error recovery
+        return false;
+    }
+
+    // No skipping leading whitespace: raw-string grammar is space-sensitive.
+    if (valid_symbols[RAW_STRING_DELIMITER]) {
+        lexer->result_symbol = RAW_STRING_DELIMITER;
+        return scan_raw_string_delimiter(scanner, lexer);
+    }
+
+    if (valid_symbols[RAW_STRING_CONTENT]) {
+        lexer->result_symbol = RAW_STRING_CONTENT;
+        return scan_raw_string_content(scanner, lexer);
+    }
+
+    return false;
+}
+
+unsigned tree_sitter_c_external_scanner_serialize(void *payload, char *buffer) {
+    static_assert(MAX_DELIMITER_LENGTH * sizeof(wchar_t) < TREE_SITTER_SERIALIZATION_BUFFER_SIZE,
+                  "Serialized delimiter is too long!");
+
+    Scanner *scanner = (Scanner *)payload;
+    size_t size = scanner->delimiter_length * sizeof(wchar_t);
+    memcpy(buffer, scanner->delimiter, size);
+    return (unsigned)size;
+}
+
+void tree_sitter_c_external_scanner_deserialize(void *payload, const char *buffer, unsigned length) {
+    assert(length % sizeof(wchar_t) == 0 && "Can't decode serialized delimiter!");
+
+    Scanner *scanner = (Scanner *)payload;
+    scanner->delimiter_length = length / sizeof(wchar_t);
+    if (length > 0) {
+        memcpy(&scanner->delimiter[0], buffer, length);
+    }
+}
+
+void tree_sitter_c_external_scanner_destroy(void *payload) {
+    Scanner *scanner = (Scanner *)payload;
+    ts_free(scanner);
+}