Update cmark-gfm + add support for front matter.

Author: samuel-williams-shopify

.gitmodules

@@ -1,4 +1,4 @@
 [submodule "cmark-gfm"]
 	path = cmark-gfm
-	url = https://github.com/github/cmark-gfm.git
+	url = https://github.com/socketry/cmark-gfm.git
 	ignore = dirty

bake.rb

@@ -46,7 +46,7 @@ def clean
 end
 
 def format
-	system("clang-format -style llvm -i ext/markly/*.c ext/markly/*.h")
+	system("clang-format -style llvm -i ext/markly/*.c ext/markly/*.h ext/markly/extensions/*.c ext/markly/extensions/*.h")
 end
 
 def benchmark
@@ -71,9 +71,9 @@ def synchronize_upstream
 	cmark_gfm = root/"cmark-gfm"
 
 	(cmark_gfm/"src").glob("**/*").copy(ext_markly)
-	(cmark_gfm/"extensions").glob("**/*").copy(ext_markly)
+	(cmark_gfm/"extensions").glob("**/*").copy(ext_markly/"extensions")
 	(cmark_gfm/"build/src").glob("config.h").copy(ext_markly)
 	(cmark_gfm/"build/src").glob("cmark-gfm_export.h").copy(ext_markly)
 	(cmark_gfm/"build/src").glob("cmark-gfm_version.h").copy(ext_markly)
-	(cmark_gfm/"build/extensions").glob("cmark-gfm-extensions_export.h").copy(ext_markly)
+	(cmark_gfm/"build/extensions").glob("cmark-gfm-extensions_export.h").copy(ext_markly/"extensions")
 end

cmark-gfm

@@ -23,6 +23,7 @@
 #include "houdini.h"
 #include "buffer.h"
 #include "footnotes.h"
+#include "front_matter.h"
 
 #define CODE_INDENT 4
 #define TAB_STOP 4
@@ -68,16 +69,12 @@ static CMARK_INLINE bool S_is_line_end_char(char c) {
   return (c == '\n' || c == '\r');
 }
 
-static CMARK_INLINE bool S_is_space_or_tab(char c) {
-  return (c == ' ' || c == '\t');
-}
-
 // Returns true if block is being finalized on the same line it ends.
 // This happens for:
 // - Document node (special case)
 // - Fenced code blocks (end on the closing fence line)
 // - Setext headings (end on the underline)
-// - HTML blocks types 1-5 per CommonMark spec §4.6 (end on the line
+// - HTML block types 1-5 per CommonMark spec §4.6 (end on the line
 //   containing the closing marker)
 // - Any block finalized on the same line it started (e.g., single-line blocks)
 static CMARK_INLINE bool S_ends_on_current_line(cmark_parser *parser, cmark_node *b) {
@@ -100,6 +97,10 @@ static CMARK_INLINE bool S_ends_on_current_line(cmark_parser *parser, cmark_node
          b->start_line == parser->line_number;
 }
 
+static CMARK_INLINE bool S_is_space_or_tab(char c) {
+  return (c == ' ' || c == '\t');
+}
+
 static void S_parser_feed(cmark_parser *parser, const unsigned char *buffer,
                           size_t len, bool eof);
 
@@ -159,6 +160,8 @@ static void cmark_parser_reset(cmark_parser *parser) {
 
   cmark_strbuf_init(parser->mem, &parser->curline, 256);
   cmark_strbuf_init(parser->mem, &parser->linebuf, 0);
+  cmark_strbuf_init(parser->mem, &parser->front_matter_buf, 0);
+  cmark_strbuf_init(parser->mem, &parser->front_matter_info, 0);
 
   cmark_node *document = make_document(parser->mem);
 
@@ -189,6 +192,8 @@ void cmark_parser_free(cmark_parser *parser) {
   cmark_parser_dispose(parser);
   cmark_strbuf_free(&parser->curline);
   cmark_strbuf_free(&parser->linebuf);
+  cmark_strbuf_free(&parser->front_matter_buf);
+  cmark_strbuf_free(&parser->front_matter_info);
   cmark_llist_free(parser->mem, parser->syntax_extensions);
   cmark_llist_free(parser->mem, parser->inline_syntax_extensions);
   mem->free(parser);
@@ -1244,7 +1249,8 @@ static void open_new_blocks(cmark_parser *parser, cmark_node **container,
                              parser->first_nonspace + 1);
       S_advance_offset(parser, input, input->len - 1 - parser->offset, false);
     } else if (!indented &&
-               parser->options & CMARK_OPT_FOOTNOTES &&
+               (parser->options & CMARK_OPT_FOOTNOTES) &&
+               depth < MAX_LIST_DEPTH &&
                (matched = scan_footnote_definition(input, parser->first_nonspace))) {
       cmark_chunk c = cmark_chunk_dup(input, parser->first_nonspace + 2, matched - 2);
 
@@ -1516,6 +1522,10 @@ static void S_process_line(cmark_parser *parser, const unsigned char *buffer,
 
   parser->line_number++;
 
+  if ((parser->options & CMARK_OPT_FRONT_MATTER) &&
+      cmark_front_matter_process_line(parser, &input))
+    goto finished;
+
   last_matched_container = check_open_blocks(parser, &input, &all_matched);
 
   if (!last_matched_container)
@@ -1556,12 +1566,20 @@ cmark_node *cmark_parser_finish(cmark_parser *parser) {
     cmark_strbuf_clear(&parser->linebuf);
   }
 
+  // If front matter scanning was still active when the document ended, no
+  // closing delimiter was found.  The entire document (after the opening ---)
+  // is treated as front matter.
+  if ((parser->options & CMARK_OPT_FRONT_MATTER) && parser->front_matter_scanning)
+    cmark_front_matter_process_line(parser, NULL);
+
   finalize_document(parser);
 
   cmark_consolidate_text_nodes(parser->root);
 
   cmark_strbuf_free(&parser->curline);
   cmark_strbuf_free(&parser->linebuf);
+  cmark_strbuf_free(&parser->front_matter_buf);
+  cmark_strbuf_free(&parser->front_matter_info);
 
 #if CMARK_DEBUG_NODES
   if (cmark_node_check(parser->root, stderr)) {

ext/markly/blocks.c

@@ -53,6 +53,7 @@ typedef enum {
   CMARK_NODE_HEADING        = CMARK_NODE_TYPE_BLOCK | 0x0009,
   CMARK_NODE_THEMATIC_BREAK = CMARK_NODE_TYPE_BLOCK | 0x000a,
   CMARK_NODE_FOOTNOTE_DEFINITION = CMARK_NODE_TYPE_BLOCK | 0x000b,
+  CMARK_NODE_FRONT_MATTER       = CMARK_NODE_TYPE_BLOCK | 0x000c,
 
   /* Inline */
   CMARK_NODE_TEXT          = CMARK_NODE_TYPE_INLINE | 0x0001,
@@ -768,6 +769,13 @@ char *cmark_render_latex_with_mem(cmark_node *root, int options, int width, cmar
  */
 #define CMARK_OPT_FULL_INFO_STRING (1 << 16)
 
+/** Parse front matter ("---" delimited block at the start of the document)
+ *  and expose it as a CMARK_NODE_FRONT_MATTER node.  The raw content between
+ *  the delimiters is available via cmark_node_get_literal(); how it is
+ *  interpreted (e.g. as YAML, TOML, JSON) is left to the caller.
+ */
+#define CMARK_OPT_FRONT_MATTER (1 << 18)
+
 /**
  * ## Version information
  */

ext/markly/cmark-gfm.h

@@ -7,7 +7,7 @@
 #include "cmark-gfm.h"
 #include "buffer.h"
 
-cmark_node_type CMARK_NODE_LAST_BLOCK = CMARK_NODE_FOOTNOTE_DEFINITION;
+cmark_node_type CMARK_NODE_LAST_BLOCK = CMARK_NODE_FRONT_MATTER;
 cmark_node_type CMARK_NODE_LAST_INLINE = CMARK_NODE_FOOTNOTE_REFERENCE;
 
 int cmark_version(void) { return CMARK_GFM_VERSION; }

ext/markly/cmark.c

@@ -492,6 +492,19 @@ static int S_render_node(cmark_renderer *renderer, cmark_node *node,
     }
     break;
 
+  case CMARK_NODE_FRONT_MATTER:
+    if (entering) {
+      const char *info = cmark_node_get_fence_info(node);
+      BLANKLINE();
+      LIT("---");
+      if (info && *info) { LIT(" "); OUT(info, false, LITERAL); }
+      LIT("\n");
+      OUT(cmark_node_get_literal(node), false, LITERAL);
+      LIT("---\n");
+      BLANKLINE();
+    }
+    break;
+
   default:
     assert(false);
     break;

ext/markly/commonmark.c

@@ -12,6 +12,11 @@
 
 append_cflags(["-O3", "-Wall", "-Wno-unknown-pragmas", "-std=c99"])
 
+# Include the extensions/ subdirectory, mirroring the cmark-gfm source layout.
+$INCFLAGS << " -I$(srcdir)/extensions"
+$VPATH    << "$(srcdir)/extensions"
+$srcs = (Dir["#{$srcdir}/*.c"] + Dir["#{$srcdir}/extensions/*.c"]).map { |f| File.basename(f) }
+
 gem_name = File.basename(__dir__)
 extension_name = "markly"