From b8af8792d1aba480b68baa36a2d687355f6f6bd1 Mon Sep 17 00:00:00 2001
From: Mathieu Duponchelle <mathieu.duponchelle@opencreed.com>
Date: Sun, 28 Feb 2016 20:47:26 +0100
Subject: [PATCH 01/20] Implement and expose block parsing API.

+ Document the parser structure for internal usage.
---
 src/CMakeLists.txt        |   2 +
 src/blocks.c              |  36 ++++++++++
 src/cmark_extension_api.h | 145 ++++++++++++++++++++++++++++++++++++++
 src/parser.h              |  15 ++++
 4 files changed, 198 insertions(+)
 create mode 100644 src/cmark_extension_api.h

diff --git a/src/CMakeLists.txt b/src/CMakeLists.txt
index 2de501aa1..f4fda4386 100644
--- a/src/CMakeLists.txt
+++ b/src/CMakeLists.txt
@@ -3,6 +3,7 @@ set(LIBRARY "libcmark")
 set(STATICLIBRARY "libcmark_static")
 set(HEADERS
   cmark.h
+  cmark_extension_api.h
   parser.h
   buffer.h
   node.h
@@ -133,6 +134,7 @@ install(TARGETS ${PROGRAM} ${LIBRARY} ${STATICLIBRARY}
 
 install(FILES
   cmark.h
+  cmark_extension_api.h
   ${CMAKE_CURRENT_BINARY_DIR}/cmark_export.h
   ${CMAKE_CURRENT_BINARY_DIR}/cmark_version.h
   DESTINATION include
diff --git a/src/blocks.c b/src/blocks.c
index 2386bb184..904a6ffc6 100644
--- a/src/blocks.c
+++ b/src/blocks.c
@@ -1185,3 +1185,39 @@ cmark_node *cmark_parser_finish(cmark_parser *parser) {
 #endif
   return parser->root;
 }
+
+int cmark_parser_get_line_number(cmark_parser *parser) {
+  return parser->line_number;
+}
+
+bufsize_t cmark_parser_get_offset(cmark_parser *parser) {
+  return parser->offset;
+}
+
+bufsize_t cmark_parser_get_column(cmark_parser *parser) {
+  return parser->column;
+}
+
+int cmark_parser_get_first_nonspace(cmark_parser *parser) {
+  return parser->first_nonspace;
+}
+
+int cmark_parser_get_first_nonspace_column(cmark_parser *parser) {
+  return parser->first_nonspace_column;
+}
+
+int cmark_parser_get_indent(cmark_parser *parser) {
+  return parser->indent;
+}
+
+int cmark_parser_is_blank(cmark_parser *parser) {
+  return parser->blank;
+}
+
+int cmark_parser_has_partially_consumed_tab(cmark_parser *parser) {
+  return parser->partially_consumed_tab;
+}
+
+int cmark_parser_get_last_line_length(cmark_parser *parser) {
+  return parser->last_line_length;
+}
diff --git a/src/cmark_extension_api.h b/src/cmark_extension_api.h
new file mode 100644
index 000000000..ca45adffa
--- /dev/null
+++ b/src/cmark_extension_api.h
@@ -0,0 +1,145 @@
+#ifndef CMARK_EXTENSION_API_H
+#define CMARK_EXTENSION_API_H
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#include <cmark.h>
+
+/** Return the index of the line currently being parsed, starting with 1.
+ */
+CMARK_EXPORT
+int cmark_parser_get_line_number(cmark_parser *parser);
+
+/** Return the offset in bytes in the line being processed.
+ *
+ * Example:
+ *
+ * ### foo
+ *
+ * Here, offset will first be 0, then 5 (the index of the 'f' character).
+ */
+CMARK_EXPORT
+int cmark_parser_get_offset(cmark_parser *parser);
+
+/**
+ * Return the offset in 'columns' in the line being processed.
+ *
+ * This value may differ from the value returned by
+ * cmark_parser_get_offset() in that it accounts for tabs,
+ * and as such should not be used as an index in the current line's
+ * buffer.
+ *
+ * Example:
+ *
+ * cmark_parser_advance_offset() can be called to advance the
+ * offset by a number of columns, instead of a number of bytes.
+ *
+ * In that case, if offset falls "in the middle" of a tab
+ * character, 'column' and offset will differ.
+ *
+ * ```
+ * foo                 \t bar
+ * ^                   ^^
+ * offset (0)          20
+ * ```
+ *
+ * If cmark_parser_advance_offset is called here with 'columns'
+ * set to 'true' and 'offset' set to 22, cmark_parser_get_offset()
+ * will return 20, whereas cmark_parser_get_column() will return
+ * 22.
+ *
+ * Additionally, as tabs expand to the next multiple of 4 column,
+ * cmark_parser_has_partially_consumed_tab() will now return
+ * 'true'.
+ */
+CMARK_EXPORT
+int cmark_parser_get_column(cmark_parser *parser);
+
+/** Return the absolute index in bytes of the first nonspace
+ * character coming after the offset as returned by
+ * cmark_parser_get_offset() in the line currently being processed.
+ *
+ * Example:
+ *
+ * ```
+ *   foo        bar            baz  \n
+ * ^               ^           ^
+ * 0            offset (16) first_nonspace (28)
+ * ```
+ */
+CMARK_EXPORT
+int cmark_parser_get_first_nonspace(cmark_parser *parser);
+
+/** Return the absolute index of the first nonspace column coming after 'offset'
+ * in the line currently being processed, counting tabs as multiple
+ * columns as appropriate.
+ *
+ * See the documentation for cmark_parser_get_first_nonspace() and
+ * cmark_parser_get_column() for more information.
+ */
+CMARK_EXPORT
+int cmark_parser_get_first_nonspace_column(cmark_parser *parser);
+
+/** Return the difference between the values returned by
+ * cmark_parser_get_first_nonspace_column() and
+ * cmark_parser_get_column().
+ *
+ * This is not a byte offset, as it can count one tab as multiple
+ * characters.
+ */
+CMARK_EXPORT
+int cmark_parser_get_indent(cmark_parser *parser);
+
+/** Return 'true' if the line currently being processed has been entirely
+ * consumed, 'false' otherwise.
+ *
+ * Example:
+ *
+ * ```
+ *   foo        bar            baz  \n
+ * ^
+ * offset
+ * ```
+ *
+ * This function will return 'false' here.
+ *
+ * ```
+ *   foo        bar            baz  \n
+ *                 ^
+ *              offset
+ * ```
+ * This function will still return 'false'.
+ *
+ * ```
+ *   foo        bar            baz  \n
+ *                                ^
+ *                             offset
+ * ```
+ *
+ * At this point, this function will now return 'true'.
+ */
+CMARK_EXPORT
+int cmark_parser_is_blank(cmark_parser *parser);
+
+/** Return 'true' if the value returned by cmark_parser_get_offset()
+ * is 'inside' an expanded tab.
+ *
+ * See the documentation for cmark_parser_get_column() for more
+ * information.
+ */
+CMARK_EXPORT
+int cmark_parser_has_partially_consumed_tab(cmark_parser *parser);
+
+/** Return the length in bytes of the previously processed line, excluding potential
+ * newline (\n) and carriage return (\r) trailing characters.
+ */
+CMARK_EXPORT
+int cmark_parser_get_last_line_length(cmark_parser *parser);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif
diff --git a/src/parser.h b/src/parser.h
index 0c5033bd2..ef2e747ec 100644
--- a/src/parser.h
+++ b/src/parser.h
@@ -14,20 +14,35 @@ extern "C" {
 
 struct cmark_parser {
   struct cmark_mem *mem;
+  /* A hashtable of urls in the current document for cross-references */
   struct cmark_reference_map *refmap;
+  /* The root node of the parser, always a CMARK_NODE_DOCUMENT */
   struct cmark_node *root;
+  /* The last open block after a line is fully processed */
   struct cmark_node *current;
+  /* See the documentation for cmark_parser_get_line_number() in cmark.h */
   int line_number;
+  /* See the documentation for cmark_parser_get_offset() in cmark.h */
   bufsize_t offset;
+  /* See the documentation for cmark_parser_get_column() in cmark.h */
   bufsize_t column;
+  /* See the documentation for cmark_parser_get_first_nonspace() in cmark.h */
   bufsize_t first_nonspace;
+  /* See the documentation for cmark_parser_get_first_nonspace_column() in cmark.h */
   bufsize_t first_nonspace_column;
+  /* See the documentation for cmark_parser_get_indent() in cmark.h */
   int indent;
+  /* See the documentation for cmark_parser_is_blank() in cmark.h */
   bool blank;
+  /* See the documentation for cmark_parser_has_partially_consumed_tab() in cmark.h */
   bool partially_consumed_tab;
+  /* Contains the currently processed line */
   cmark_strbuf curline;
+  /* See the documentation for cmark_parser_get_last_line_length() in cmark.h */
   bufsize_t last_line_length;
+  /* FIXME: not sure about the difference with curline */
   cmark_strbuf linebuf;
+  /* Options set by the user, see the Options section in cmark.h */
   int options;
   bool last_buffer_ended_with_cr;
 };

From f2e9eff2eb9aafc31b472029694f87e94f5bb0fc Mon Sep 17 00:00:00 2001
From: Mathieu Duponchelle <mathieu.duponchelle@opencreed.com>
Date: Sun, 28 Feb 2016 05:49:46 +0100
Subject: [PATCH 02/20] [API]: parser: Expose 'setter' methods in cmark_parser.

As opposed to the previous commit, where I exposed getters for
the private parser structure, this exposes two methods that
will influence parsing, cmark_parser_add_child and
cmark_parser_advance_offset.
---
 src/blocks.c              | 16 ++++++++++++++++
 src/cmark_extension_api.h | 23 +++++++++++++++++++++++
 2 files changed, 39 insertions(+)

diff --git a/src/blocks.c b/src/blocks.c
index 904a6ffc6..94434752f 100644
--- a/src/blocks.c
+++ b/src/blocks.c
@@ -1221,3 +1221,19 @@ int cmark_parser_has_partially_consumed_tab(cmark_parser *parser) {
 int cmark_parser_get_last_line_length(cmark_parser *parser) {
   return parser->last_line_length;
 }
+
+cmark_node *cmark_parser_add_child(cmark_parser *parser,
+                                   cmark_node   *parent,
+                                   cmark_node_type block_type,
+                                   int start_column) {
+  return add_child(parser, parent, block_type, start_column);
+}
+
+void cmark_parser_advance_offset(cmark_parser *parser,
+                                 const char *input,
+                                 int count,
+                                 int columns) {
+  cmark_chunk input_chunk = cmark_chunk_literal(input);
+
+  S_advance_offset(parser, &input_chunk, count, columns);
+}
diff --git a/src/cmark_extension_api.h b/src/cmark_extension_api.h
index ca45adffa..bf0d97f0d 100644
--- a/src/cmark_extension_api.h
+++ b/src/cmark_extension_api.h
@@ -138,6 +138,29 @@ int cmark_parser_has_partially_consumed_tab(cmark_parser *parser);
 CMARK_EXPORT
 int cmark_parser_get_last_line_length(cmark_parser *parser);
 
+/** Add a child to 'parent' during the parsing process.
+ *
+ * If 'parent' isn't the kind of node that can accept this child,
+ * this function will back up till it hits a node that can, closing
+ * blocks as appropriate.
+ */
+CMARK_EXPORT
+cmark_node*cmark_parser_add_child(cmark_parser *parser,
+                                  cmark_node *parent,
+                                  cmark_node_type block_type,
+                                  int start_column);
+
+/** Advance the 'offset' of the parser in the current line.
+ *
+ * See the documentation of cmark_parser_get_offset() and
+ * cmark_parser_get_column() for more information.
+ */
+CMARK_EXPORT
+void cmark_parser_advance_offset(cmark_parser *parser,
+                                 const char *input,
+                                 int count,
+                                 int columns);
+
 #ifdef __cplusplus
 }
 #endif

From 96f966e1d730a85b38465a0843bf0c04edc20af1 Mon Sep 17 00:00:00 2001
From: Mathieu Duponchelle <mathieu.duponchelle@opencreed.com>
Date: Sun, 28 Feb 2016 06:53:23 +0100
Subject: [PATCH 03/20] [API]: node: implement and expose cmark_node_set_type.

This is necessary for extensions implementing vertical rules.

An example is setext headings:

A heading
---------

When cmark parses 'A heading', it first creates a paragraph
block to contain it, it's only when cmark parses the second line
that the type of the block is changed to the
CMARK_NODE_TYPE_HEADING type.
---
 src/cmark_extension_api.h |  6 +++
 src/node.c                | 77 +++++++++++++++++++++++++++------------
 2 files changed, 59 insertions(+), 24 deletions(-)

diff --git a/src/cmark_extension_api.h b/src/cmark_extension_api.h
index bf0d97f0d..40101a4c8 100644
--- a/src/cmark_extension_api.h
+++ b/src/cmark_extension_api.h
@@ -161,6 +161,12 @@ void cmark_parser_advance_offset(cmark_parser *parser,
                                  int count,
                                  int columns);
 
+/** Change the type of 'node'.
+ *
+ * Return 0 if the type could be changed, 1 otherwise.
+ */
+CMARK_EXPORT int cmark_node_set_type(cmark_node *node, cmark_node_type type);
+
 #ifdef __cplusplus
 }
 #endif
diff --git a/src/node.c b/src/node.c
index 62acf14b9..d5e09d239 100644
--- a/src/node.c
+++ b/src/node.c
@@ -105,35 +105,41 @@ cmark_node *cmark_node_new(cmark_node_type type) {
   return cmark_node_new_with_mem(type, &DEFAULT_MEM_ALLOCATOR);
 }
 
+static void free_node_as(cmark_node *node) {
+  switch (node->type) {
+  case CMARK_NODE_CODE_BLOCK:
+    cmark_chunk_free(NODE_MEM(node), &node->as.code.info);
+    cmark_chunk_free(NODE_MEM(node), &node->as.code.literal);
+    break;
+  case CMARK_NODE_TEXT:
+  case CMARK_NODE_HTML_INLINE:
+  case CMARK_NODE_CODE:
+  case CMARK_NODE_HTML_BLOCK:
+    cmark_chunk_free(NODE_MEM(node), &node->as.literal);
+    break;
+  case CMARK_NODE_LINK:
+  case CMARK_NODE_IMAGE:
+    cmark_chunk_free(NODE_MEM(node), &node->as.link.url);
+    cmark_chunk_free(NODE_MEM(node), &node->as.link.title);
+    break;
+  case CMARK_NODE_CUSTOM_BLOCK:
+  case CMARK_NODE_CUSTOM_INLINE:
+    cmark_chunk_free(NODE_MEM(node), &node->as.custom.on_enter);
+    cmark_chunk_free(NODE_MEM(node), &node->as.custom.on_exit);
+    break;
+  default:
+    break;
+  }
+}
+
 // Free a cmark_node list and any children.
 static void S_free_nodes(cmark_node *e) {
   cmark_node *next;
   while (e != NULL) {
     cmark_strbuf_free(&e->content);
-    switch (e->type) {
-    case CMARK_NODE_CODE_BLOCK:
-      cmark_chunk_free(NODE_MEM(e), &e->as.code.info);
-      cmark_chunk_free(NODE_MEM(e), &e->as.code.literal);
-      break;
-    case CMARK_NODE_TEXT:
-    case CMARK_NODE_HTML_INLINE:
-    case CMARK_NODE_CODE:
-    case CMARK_NODE_HTML_BLOCK:
-      cmark_chunk_free(NODE_MEM(e), &e->as.literal);
-      break;
-    case CMARK_NODE_LINK:
-    case CMARK_NODE_IMAGE:
-      cmark_chunk_free(NODE_MEM(e), &e->as.link.url);
-      cmark_chunk_free(NODE_MEM(e), &e->as.link.title);
-      break;
-    case CMARK_NODE_CUSTOM_BLOCK:
-    case CMARK_NODE_CUSTOM_INLINE:
-      cmark_chunk_free(NODE_MEM(e), &e->as.custom.on_enter);
-      cmark_chunk_free(NODE_MEM(e), &e->as.custom.on_exit);
-      break;
-    default:
-      break;
-    }
+
+    free_node_as(e);
+
     if (e->last_child) {
       // Splice children into list
       e->last_child->next = e->next;
@@ -159,6 +165,29 @@ cmark_node_type cmark_node_get_type(cmark_node *node) {
   }
 }
 
+int cmark_node_set_type(cmark_node * node, cmark_node_type type) {
+  cmark_node_type initial_type;
+
+  if (type == node->type)
+    return 1;
+
+  initial_type = node->type;
+  node->type = type;
+
+  if (!S_can_contain(node->parent, node)) {
+    node->type = initial_type;
+    return 0;
+  }
+
+  /* We rollback the type to free the union members appropriately */
+  node->type = initial_type;
+  free_node_as(node);
+
+  node->type = type;
+
+  return 1;
+}
+
 const char *cmark_node_get_type_string(cmark_node *node) {
   if (node == NULL) {
     return "NONE";

From 8075d6411a8baeb5d11a2dcfdfa460a2c7c74d8a Mon Sep 17 00:00:00 2001
From: Mathieu Duponchelle <mathieu.duponchelle@opencreed.com>
Date: Sun, 28 Feb 2016 08:05:54 +0100
Subject: [PATCH 04/20] [API]: node: implement and expose
 cmark_node_set_user_data_free_func.

Ideally, this would be passed in set_user_data, but this would
break API.
---
 src/cmark.h |  7 +++++++
 src/node.c  | 12 ++++++++++++
 src/node.h  |  1 +
 3 files changed, 20 insertions(+)

diff --git a/src/cmark.h b/src/cmark.h
index 45d7bc324..5c03b9e87 100644
--- a/src/cmark.h
+++ b/src/cmark.h
@@ -87,6 +87,8 @@ typedef struct cmark_node cmark_node;
 typedef struct cmark_parser cmark_parser;
 typedef struct cmark_iter cmark_iter;
 
+typedef void (*cmark_free_func) (void *user_data);
+
 /**
  * ## Custom memory allocator support
  */
@@ -253,6 +255,11 @@ CMARK_EXPORT void *cmark_node_get_user_data(cmark_node *node);
  */
 CMARK_EXPORT int cmark_node_set_user_data(cmark_node *node, void *user_data);
 
+/** Set free function for user data */
+CMARK_EXPORT
+int cmark_node_set_user_data_free_func(cmark_node *node,
+                                        cmark_free_func free_func);
+
 /** Returns the type of 'node', or `CMARK_NODE_NONE` on error.
  */
 CMARK_EXPORT cmark_node_type cmark_node_get_type(cmark_node *node);
diff --git a/src/node.c b/src/node.c
index d5e09d239..081e533a8 100644
--- a/src/node.c
+++ b/src/node.c
@@ -138,6 +138,9 @@ static void S_free_nodes(cmark_node *e) {
   while (e != NULL) {
     cmark_strbuf_free(&e->content);
 
+    if (e->user_data && e->user_data_free_func)
+      e->user_data_free_func(e->user_data);
+
     free_node_as(e);
 
     if (e->last_child) {
@@ -297,6 +300,15 @@ int cmark_node_set_user_data(cmark_node *node, void *user_data) {
   return 1;
 }
 
+int cmark_node_set_user_data_free_func(cmark_node *node,
+                                        cmark_free_func free_func) {
+  if (node == NULL) {
+    return 0;
+  }
+  node->user_data_free_func = free_func;
+  return 1;
+}
+
 const char *cmark_node_get_literal(cmark_node *node) {
   if (node == NULL) {
     return NULL;
diff --git a/src/node.h b/src/node.h
index 65d857f0b..bafa730fd 100644
--- a/src/node.h
+++ b/src/node.h
@@ -61,6 +61,7 @@ struct cmark_node {
   struct cmark_node *last_child;
 
   void *user_data;
+  cmark_free_func user_data_free_func;
 
   int start_line;
   int start_column;

From 0839b4bc274a5337d71bbcd8be4273aeac794857 Mon Sep 17 00:00:00 2001
From: Mathieu Duponchelle <mathieu.duponchelle@opencreed.com>
Date: Sun, 28 Feb 2016 08:21:02 +0100
Subject: [PATCH 05/20] [API]: node: Expose string_content.

By implementing cmark_node_get_string_content and
cmark_node_set_string_content.

This is useful for vertical rules in extensions, as they may need
to access it in order to decide whether to update the block.

Unfortunately, this overlaps with get_literal and set_literal.

As far as I can tell we should deprecate these functions, and
have them follow the get_string_content code path and
set_string_content for the while.
---
 src/cmark_extension_api.h | 10 ++++++++++
 src/node.c                |  9 +++++++++
 2 files changed, 19 insertions(+)

diff --git a/src/cmark_extension_api.h b/src/cmark_extension_api.h
index 40101a4c8..c097bb185 100644
--- a/src/cmark_extension_api.h
+++ b/src/cmark_extension_api.h
@@ -167,6 +167,16 @@ void cmark_parser_advance_offset(cmark_parser *parser,
  */
 CMARK_EXPORT int cmark_node_set_type(cmark_node *node, cmark_node_type type);
 
+/** Return the string content for all types of 'node'.
+ *  The pointer stays valid as long as 'node' isn't freed.
+ */
+CMARK_EXPORT const char *cmark_node_get_string_content(cmark_node *node);
+
+/** Set the string 'content' for all types of 'node'.
+ *  Copies 'content'.
+ */
+CMARK_EXPORT int cmark_node_set_string_content(cmark_node *node, const char *content);
+
 #ifdef __cplusplus
 }
 #endif
diff --git a/src/node.c b/src/node.c
index 081e533a8..8e7c3dec2 100644
--- a/src/node.c
+++ b/src/node.c
@@ -355,6 +355,15 @@ int cmark_node_set_literal(cmark_node *node, const char *content) {
   return 0;
 }
 
+const char *cmark_node_get_string_content(cmark_node *node) {
+  return (char *) node->content.ptr;
+}
+
+int cmark_node_set_string_content(cmark_node *node, const char *content) {
+  cmark_strbuf_sets(&node->content, content);
+  return true;
+}
+
 int cmark_node_get_heading_level(cmark_node *node) {
   if (node == NULL) {
     return 0;

From 19ffd43cae38663ebe255545f019672d694271ed Mon Sep 17 00:00:00 2001
From: Mathieu Duponchelle <mathieu.duponchelle@opencreed.com>
Date: Sun, 6 Mar 2016 01:30:12 +0100
Subject: [PATCH 06/20] [API]: node: set and get code blocks fenced state.

---
 src/cmark.h |  9 +++++++++
 src/node.c  | 32 ++++++++++++++++++++++++++++++++
 2 files changed, 41 insertions(+)

diff --git a/src/cmark.h b/src/cmark.h
index 5c03b9e87..318db498f 100644
--- a/src/cmark.h
+++ b/src/cmark.h
@@ -339,6 +339,15 @@ CMARK_EXPORT const char *cmark_node_get_fence_info(cmark_node *node);
  */
 CMARK_EXPORT int cmark_node_set_fence_info(cmark_node *node, const char *info);
 
+/** Sets code blocks fencing details
+ */
+CMARK_EXPORT int cmark_node_set_fenced(cmark_node * node, int fenced,
+    int length, int offset, char character);
+
+/** Returns code blocks fencing details
+ */
+CMARK_EXPORT int cmark_node_get_fenced(cmark_node *node, int *length, int *offset, char *character);
+
 /** Returns the URL of a link or image 'node', or an empty string
     if no URL is set.
  */
diff --git a/src/node.c b/src/node.c
index 8e7c3dec2..00e599125 100644
--- a/src/node.c
+++ b/src/node.c
@@ -530,6 +530,38 @@ int cmark_node_set_fence_info(cmark_node *node, const char *info) {
   }
 }
 
+int cmark_node_get_fenced(cmark_node *node, int *length, int *offset, char *character) {
+  if (node == NULL) {
+    return 0;
+  }
+
+  if (node->type == CMARK_NODE_CODE_BLOCK) {
+    *length = node->as.code.fence_length;
+    *offset = node->as.code.fence_offset;
+    *character = node->as.code.fence_char;
+    return node->as.code.fenced;
+  } else {
+    return 0;
+  }
+}
+
+int cmark_node_set_fenced(cmark_node * node, int fenced,
+    int length, int offset, char character) {
+  if (node == NULL) {
+    return 0;
+  }
+
+  if (node->type == CMARK_NODE_CODE_BLOCK) {
+    node->as.code.fenced = fenced;
+    node->as.code.fence_length = length;
+    node->as.code.fence_offset = offset;
+    node->as.code.fence_char = character;
+    return 1;
+  } else {
+    return 0;
+  }
+}
+
 const char *cmark_node_get_url(cmark_node *node) {
   if (node == NULL) {
     return NULL;

From 9d8a9dc9a677755e10633120ca7d0d7d3ef7d1f6 Mon Sep 17 00:00:00 2001
From: Mathieu Duponchelle <mathieu.duponchelle@opencreed.com>
Date: Sat, 27 Feb 2016 02:22:24 +0100
Subject: [PATCH 07/20] Check in and expose a linked list.

---
 src/CMakeLists.txt |  1 +
 src/cmark.h        | 37 +++++++++++++++++++++++++++++++++++++
 src/linked_list.c  | 37 +++++++++++++++++++++++++++++++++++++
 3 files changed, 75 insertions(+)
 create mode 100644 src/linked_list.c

diff --git a/src/CMakeLists.txt b/src/CMakeLists.txt
index f4fda4386..8baf7ed23 100644
--- a/src/CMakeLists.txt
+++ b/src/CMakeLists.txt
@@ -38,6 +38,7 @@ set(LIBRARY_SOURCES
   houdini_html_e.c
   houdini_html_u.c
   cmark_ctype.c
+  linked_list.c
   ${HEADERS}
   )
 
diff --git a/src/cmark.h b/src/cmark.h
index 318db498f..5e04f0c81 100644
--- a/src/cmark.h
+++ b/src/cmark.h
@@ -102,6 +102,43 @@ typedef struct cmark_mem {
 	void (*free)(void *);
 } cmark_mem;
 
+/*
+ * ## Basic data structures
+ *
+ * To keep dependencies to the strict minimum, libcmark implements
+ * its own versions of "classic" data structures.
+ */
+
+/**
+ * ### Linked list
+ */
+
+/** A generic singly linked list.
+ */
+typedef struct _cmark_llist
+{
+  struct _cmark_llist *next;
+  void         *data;
+} cmark_llist;
+
+/** Append an element to the linked list, return the possibly modified
+ * head of the list.
+ */
+CMARK_EXPORT
+cmark_llist * cmark_llist_append    (cmark_llist       * head,
+                                     void              * data);
+
+/** Free the list starting with 'head', calling 'free_func' with the
+ *  data pointer of each of its elements
+ */
+CMARK_EXPORT
+void          cmark_llist_free_full (cmark_llist       * head,
+                                     cmark_free_func     free_func);
+
+/** Free the list starting with 'head'
+ */
+CMARK_EXPORT
+void          cmark_llist_free      (cmark_llist       * head);
 
 /**
  * ## Creating and Destroying Nodes
diff --git a/src/linked_list.c b/src/linked_list.c
new file mode 100644
index 000000000..7d6690dae
--- /dev/null
+++ b/src/linked_list.c
@@ -0,0 +1,37 @@
+#include <stdlib.h>
+
+#include "cmark.h"
+
+cmark_llist *cmark_llist_append(cmark_llist *head, void *data) {
+  cmark_llist *tmp;
+  cmark_llist *new_node = (cmark_llist *) malloc(sizeof(cmark_llist));
+
+  new_node->data = data;
+  new_node->next = NULL;
+
+  if (!head)
+    return new_node;
+
+  for (tmp = head; tmp->next; tmp=tmp->next);
+
+  tmp->next = new_node;
+
+  return head;
+}
+
+void cmark_llist_free_full(cmark_llist *head, cmark_free_func free_func) {
+  cmark_llist *tmp, *prev;
+
+  for (tmp = head; tmp;) {
+    if (free_func)
+      free_func(tmp->data);
+
+    prev = tmp;
+    tmp = tmp->next;
+    free(prev);
+  }
+}
+
+void cmark_llist_free(cmark_llist *head) {
+  cmark_llist_free_full(head, NULL);
+}

From d2f829ff2b44e40b843a107db56da87833cbdd87 Mon Sep 17 00:00:00 2001
From: Mathieu Duponchelle <mathieu.duponchelle@opencreed.com>
Date: Thu, 28 Apr 2016 00:46:22 +0200
Subject: [PATCH 08/20] Define syntax extensions

And expose and implement block parsing hooks
---
 src/CMakeLists.txt        |   2 +
 src/blocks.c              |  52 ++++++++++++++++-
 src/cmark_extension_api.h | 120 ++++++++++++++++++++++++++++++++++++++
 src/node.c                |  17 ++++++
 src/node.h                |   3 +
 src/parser.h              |   1 +
 src/syntax_extension.c    |  27 +++++++++
 src/syntax_extension.h    |  13 +++++
 8 files changed, 233 insertions(+), 2 deletions(-)
 create mode 100644 src/syntax_extension.c
 create mode 100644 src/syntax_extension.h

diff --git a/src/CMakeLists.txt b/src/CMakeLists.txt
index 8baf7ed23..9f01ef4f6 100644
--- a/src/CMakeLists.txt
+++ b/src/CMakeLists.txt
@@ -16,6 +16,7 @@ set(HEADERS
   houdini.h
   cmark_ctype.h
   render.h
+  syntax_extension.h
   )
 set(LIBRARY_SOURCES
   cmark.c
@@ -39,6 +40,7 @@ set(LIBRARY_SOURCES
   houdini_html_u.c
   cmark_ctype.c
   linked_list.c
+  syntax_extension.c
   ${HEADERS}
   )
 
diff --git a/src/blocks.c b/src/blocks.c
index 94434752f..7d057efed 100644
--- a/src/blocks.c
+++ b/src/blocks.c
@@ -10,6 +10,7 @@
 #include <stdio.h>
 
 #include "cmark_ctype.h"
+#include "syntax_extension.h"
 #include "config.h"
 #include "parser.h"
 #include "cmark.h"
@@ -80,6 +81,12 @@ static cmark_node *make_document(cmark_mem *mem) {
   return e;
 }
 
+int cmark_parser_attach_syntax_extension(cmark_parser *parser,
+                                      cmark_syntax_extension *extension) {
+  parser->syntax_extensions = cmark_llist_append(parser->syntax_extensions, extension);
+  return 1;
+}
+
 cmark_parser *cmark_parser_new_with_mem(int options, cmark_mem *mem) {
   cmark_parser *parser = (cmark_parser *)mem->calloc(1, sizeof(cmark_parser));
   parser->mem = mem;
@@ -103,6 +110,7 @@ cmark_parser *cmark_parser_new_with_mem(int options, cmark_mem *mem) {
   parser->last_line_length = 0;
   parser->options = options;
   parser->last_buffer_ended_with_cr = false;
+  parser->syntax_extensions = NULL;
 
   return parser;
 }
@@ -117,6 +125,7 @@ void cmark_parser_free(cmark_parser *parser) {
   cmark_strbuf_free(&parser->curline);
   cmark_strbuf_free(&parser->linebuf);
   cmark_reference_map_free(parser->refmap);
+  cmark_llist_free(parser->syntax_extensions);
   mem->free(parser);
 }
 
@@ -765,6 +774,21 @@ static bool parse_html_block_prefix(cmark_parser *parser,
   return res;
 }
 
+static bool parse_extension_block(cmark_parser *parser,
+                                  cmark_node *container,
+                                  cmark_chunk *input)
+{
+  bool res = false;
+
+  if (container->extension->last_block_matches) {
+    if (container->extension->last_block_matches(
+        container->extension, parser, input->data, input->len, container))
+      res = true;
+  }
+
+  return res;
+}
+
 /**
  * For each containing node, try to parse the associated line start.
  *
@@ -786,6 +810,12 @@ static cmark_node *check_open_blocks(cmark_parser *parser, cmark_chunk *input,
 
     S_find_first_nonspace(parser, input);
 
+    if (container->extension) {
+      if (!parse_extension_block(parser, container, input))
+        goto done;
+      continue;
+    }
+
     switch (cont_type) {
     case CMARK_NODE_BLOCK_QUOTE:
       if (!parse_block_quote_prefix(parser, input))
@@ -981,9 +1011,27 @@ static void open_new_blocks(cmark_parser *parser, cmark_node **container,
       (*container)->as.code.fence_length = 0;
       (*container)->as.code.fence_offset = 0;
       (*container)->as.code.info = cmark_chunk_literal("");
-
     } else {
-      break;
+      cmark_llist *tmp;
+      cmark_node *new_container = NULL;
+
+      for (tmp = parser->syntax_extensions; tmp; tmp=tmp->next) {
+        cmark_syntax_extension *ext = (cmark_syntax_extension *) tmp->data;
+
+        if (ext->try_opening_block) {
+          new_container = ext->try_opening_block(
+              ext, indented, parser, *container, input->data, input->len);
+
+          if (new_container) {
+            *container = new_container;
+            break;
+          }
+        }
+      }
+
+      if (!new_container) {
+        break;
+      }
     }
 
     if (accepts_lines(S_type(*container))) {
diff --git a/src/cmark_extension_api.h b/src/cmark_extension_api.h
index c097bb185..8a2164a99 100644
--- a/src/cmark_extension_api.h
+++ b/src/cmark_extension_api.h
@@ -7,6 +7,106 @@ extern "C" {
 
 #include <cmark.h>
 
+/**
+ * ## Extension Support
+ *
+ * While the "core" of libcmark is strictly compliant with the
+ * specification, an API is provided for extension writers to
+ * hook into the parsing process.
+ *
+ * It should be noted that the cmark_node API already offers
+ * room for customization, with methods offered to traverse and
+ * modify the AST, and even define custom blocks.
+ * When the desired customization is achievable in an error-proof
+ * way using that API, it should be the preferred method.
+ *
+ * The following API requires a more in-depth understanding
+ * of libcmark's parsing strategy, which is exposed
+ * [here](http://spec.commonmark.org/0.24/#appendix-a-parsing-strategy).
+ *
+ * It should be used when "a posteriori" modification of the AST
+ * proves to be too difficult / impossible to implement correctly.
+ *
+ * It can also serve as an intermediary step before extending
+ * the specification, as an extension implemented using this API
+ * will be trivially integrated in the core if it proves to be
+ * desirable.
+ */
+
+
+/** A syntax extension that can be attached to a cmark_parser
+ * with cmark_parser_attach_syntax_extension().
+ *
+ * Extension writers should assign functions matching
+ * the signature of the following 'virtual methods' to
+ * implement new functionality.
+ *
+ * Their calling order and expected behaviour match the procedure outlined
+ * at <http://spec.commonmark.org/0.24/#phase-1-block-structure>:
+ *
+ * During step 1, cmark will call the function provided through
+ * 'cmark_syntax_extension_set_match_block_func' when it
+ * iterates over an open block created by this extension,
+ * to determine  whether it could contain the new line.
+ * If no function was provided, cmark will close the block.
+ *
+ * During step 2, if and only if the new line doesn't match any
+ * of the standard syntax rules, cmark will call the function
+ * provided through 'cmark_syntax_extension_set_open_block_func'
+ * to let the extension determine whether that new line matches
+ * one of its syntax rules.
+ * It is the responsibility of the parser to create and add the
+ * new block with cmark_parser_make_block and cmark_parser_add_child.
+ * If no function was provided is NULL, the extension will have
+ * no effect at all on the final block structure of the AST.
+ */
+typedef struct cmark_syntax_extension cmark_syntax_extension;
+
+/** Should create and add a new open block to 'parent_container' if
+ * 'input' matches a syntax rule for that block type. It is allowed
+ * to modify the type of 'parent_container'.
+ *
+ * Should return the newly created block if there is one, or
+ * 'parent_container' if its type was modified, or NULL.
+ */
+typedef cmark_node * (*cmark_open_block_func) (cmark_syntax_extension *extension,
+                                       int indented,
+                                       cmark_parser *parser,
+                                       cmark_node *parent_container,
+                                       unsigned char *input,
+                                       int len);
+
+/** Should return 'true' if 'input' can be contained in 'container',
+ *  'false' otherwise.
+ */
+typedef int (*cmark_match_block_func)        (cmark_syntax_extension *extension,
+                                       cmark_parser *parser,
+                                       unsigned char *input,
+                                       int len,
+                                       cmark_node *container);
+
+/** Free a cmark_syntax_extension.
+ */
+CMARK_EXPORT
+void cmark_syntax_extension_free               (cmark_syntax_extension *extension);
+
+/** Return a newly-constructed cmark_syntax_extension, named 'name'.
+ */
+CMARK_EXPORT
+cmark_syntax_extension *cmark_syntax_extension_new (const char *name);
+
+/** See the documentation for 'cmark_syntax_extension'
+ */
+CMARK_EXPORT
+void cmark_syntax_extension_set_open_block_func(cmark_syntax_extension *extension,
+                                                cmark_open_block_func func);
+
+/** See the documentation for 'cmark_syntax_extension'
+ */
+CMARK_EXPORT
+void cmark_syntax_extension_set_match_block_func(cmark_syntax_extension *extension,
+                                                 cmark_match_block_func func);
+
 /** Return the index of the line currently being parsed, starting with 1.
  */
 CMARK_EXPORT
@@ -161,6 +261,16 @@ void cmark_parser_advance_offset(cmark_parser *parser,
                                  int count,
                                  int columns);
 
+/** Attach the syntax 'extension' to the 'parser', to provide extra syntax
+ *  rules.
+ *  See the documentation for cmark_syntax_extension for more information.
+ *
+ *  Returns 'true' if the 'extension' was successfully attached,
+ *  'false' otherwise.
+ */
+CMARK_EXPORT
+int cmark_parser_attach_syntax_extension(cmark_parser *parser, cmark_syntax_extension *extension);
+
 /** Change the type of 'node'.
  *
  * Return 0 if the type could be changed, 1 otherwise.
@@ -177,6 +287,16 @@ CMARK_EXPORT const char *cmark_node_get_string_content(cmark_node *node);
  */
 CMARK_EXPORT int cmark_node_set_string_content(cmark_node *node, const char *content);
 
+/** Get the syntax extension responsible for the creation of 'node'.
+ *  Return NULL if 'node' was created because it matched standard syntax rules.
+ */
+CMARK_EXPORT cmark_syntax_extension *cmark_node_get_syntax_extension(cmark_node *node);
+
+/** Set the syntax extension responsible for creating 'node'.
+ */
+CMARK_EXPORT int cmark_node_set_syntax_extension(cmark_node *node,
+                                                  cmark_syntax_extension *extension);
+
 #ifdef __cplusplus
 }
 #endif
diff --git a/src/node.c b/src/node.c
index 00e599125..7cfa7147e 100644
--- a/src/node.c
+++ b/src/node.c
@@ -694,6 +694,23 @@ int cmark_node_set_on_exit(cmark_node *node, const char *on_exit) {
   return 0;
 }
 
+cmark_syntax_extension *cmark_node_get_syntax_extension(cmark_node *node) {
+  if (node == NULL) {
+    return NULL;
+  }
+
+  return node->extension;
+}
+
+int cmark_node_set_syntax_extension(cmark_node *node, cmark_syntax_extension *extension) {
+  if (node == NULL) {
+    return 0;
+  }
+
+  node->extension = extension;
+  return 1;
+}
+
 int cmark_node_get_start_line(cmark_node *node) {
   if (node == NULL) {
     return 0;
diff --git a/src/node.h b/src/node.h
index bafa730fd..93932aae0 100644
--- a/src/node.h
+++ b/src/node.h
@@ -9,6 +9,7 @@ extern "C" {
 #include <stdint.h>
 
 #include "cmark.h"
+#include "cmark_extension_api.h"
 #include "buffer.h"
 #include "chunk.h"
 
@@ -70,6 +71,8 @@ struct cmark_node {
   uint16_t type;
   uint16_t flags;
 
+  cmark_syntax_extension *extension;
+
   union {
     cmark_chunk literal;
     cmark_list list;
diff --git a/src/parser.h b/src/parser.h
index ef2e747ec..1129f3bae 100644
--- a/src/parser.h
+++ b/src/parser.h
@@ -45,6 +45,7 @@ struct cmark_parser {
   /* Options set by the user, see the Options section in cmark.h */
   int options;
   bool last_buffer_ended_with_cr;
+  cmark_llist *syntax_extensions;
 };
 
 #ifdef __cplusplus
diff --git a/src/syntax_extension.c b/src/syntax_extension.c
new file mode 100644
index 000000000..11a92c020
--- /dev/null
+++ b/src/syntax_extension.c
@@ -0,0 +1,27 @@
+#include <stdlib.h>
+
+#include "cmark.h"
+#include "syntax_extension.h"
+#include "buffer.h"
+
+void cmark_syntax_extension_free(cmark_syntax_extension *extension) {
+  free(extension->name);
+  free(extension);
+}
+
+cmark_syntax_extension *cmark_syntax_extension_new(const char *name) {
+  cmark_syntax_extension *res = (cmark_syntax_extension *) calloc(1, sizeof(cmark_syntax_extension));
+  res->name = (char *) malloc(sizeof(char) * (strlen(name)) + 1);
+  strcpy(res->name, name);
+  return res;
+}
+
+void cmark_syntax_extension_set_open_block_func(cmark_syntax_extension *extension,
+                                                cmark_open_block_func func) {
+  extension->try_opening_block = func;
+}
+
+void cmark_syntax_extension_set_match_block_func(cmark_syntax_extension *extension,
+                                                 cmark_match_block_func func) {
+  extension->last_block_matches = func;
+}
diff --git a/src/syntax_extension.h b/src/syntax_extension.h
new file mode 100644
index 000000000..5565b4523
--- /dev/null
+++ b/src/syntax_extension.h
@@ -0,0 +1,13 @@
+#ifndef SYNTAX_EXTENSION_H
+#define SYNTAX_EXTENSION_H
+
+#include "cmark.h"
+#include "cmark_extension_api.h"
+
+struct cmark_syntax_extension {
+  cmark_match_block_func   last_block_matches;
+  cmark_open_block_func    try_opening_block;
+  char                   * name;
+};
+
+#endif

From 1bc58d0938b5f4b4f206e997d2c6710028a90237 Mon Sep 17 00:00:00 2001
From: Mathieu Duponchelle <mathieu.duponchelle@opencreed.com>
Date: Thu, 28 Apr 2016 00:52:02 +0200
Subject: [PATCH 09/20] Implement plugin loading and discovery

Linux-only support
---
 api_test/CMakeLists.txt   |   2 +-
 src/CMakeLists.txt        |   8 +++
 src/cmark.c               |  11 +++
 src/cmark.h               |  16 +++++
 src/cmark_extension_api.h |  62 +++++++++++++++++
 src/config.h.in           |   2 +
 src/libcmark.pc.in        |   2 +-
 src/plugin.c              |  33 +++++++++
 src/plugin.h              |  34 +++++++++
 src/registry.c            | 141 ++++++++++++++++++++++++++++++++++++++
 src/registry.h            |  18 +++++
 11 files changed, 327 insertions(+), 2 deletions(-)
 create mode 100644 src/plugin.c
 create mode 100644 src/plugin.h
 create mode 100644 src/registry.c
 create mode 100644 src/registry.h

diff --git a/api_test/CMakeLists.txt b/api_test/CMakeLists.txt
index 3151ccccb..5c247aba0 100644
--- a/api_test/CMakeLists.txt
+++ b/api_test/CMakeLists.txt
@@ -8,7 +8,7 @@ include_directories(
   ${PROJECT_SOURCE_DIR}/src
   ${PROJECT_BINARY_DIR}/src
 )
-target_link_libraries(api_test libcmark)
+target_link_libraries(api_test libcmark ${CMAKE_DL_LIBS})
 
 # Compiler flags
 if(MSVC)
diff --git a/src/CMakeLists.txt b/src/CMakeLists.txt
index 9f01ef4f6..a4f673bd0 100644
--- a/src/CMakeLists.txt
+++ b/src/CMakeLists.txt
@@ -16,7 +16,9 @@ set(HEADERS
   houdini.h
   cmark_ctype.h
   render.h
+  registry.h
   syntax_extension.h
+  plugin.h
   )
 set(LIBRARY_SOURCES
   cmark.c
@@ -41,6 +43,8 @@ set(LIBRARY_SOURCES
   cmark_ctype.c
   linked_list.c
   syntax_extension.c
+  registry.c
+  plugin.c
   ${HEADERS}
   )
 
@@ -85,6 +89,8 @@ set(CMAKE_LINKER_DEBUG "${CMAKE_LINKER_FLAGS_DEBUG}")
 set(CMAKE_C_FLAGS_PROFILE "${CMAKE_C_FLAGS_RELEASE} -pg")
 set(CMAKE_LINKER_PROFILE "${CMAKE_LINKER_FLAGS_RELEASE} -pg")
 
+add_definitions(-DLIBDIR=\"${CMAKE_BINARY_DIR}\")
+
 if (${CMAKE_MAJOR_VERSION} GREATER 1 AND ${CMAKE_MINOR_VERSION} GREATER 8)
   set(CMAKE_C_VISIBILITY_PRESET hidden)
   set(CMAKE_VISIBILITY_INLINES_HIDDEN 1)
@@ -113,6 +119,8 @@ else()
     VERSION ${PROJECT_VERSION})
 endif(MSVC)
 
+target_link_libraries(cmark ${CMAKE_DL_LIBS})
+
 set_property(TARGET ${LIBRARY}
   APPEND PROPERTY MACOSX_RPATH true)
 
diff --git a/src/cmark.c b/src/cmark.c
index d5fef6382..ad2554cf8 100644
--- a/src/cmark.c
+++ b/src/cmark.c
@@ -1,6 +1,7 @@
 #include <stdlib.h>
 #include <assert.h>
 #include <stdio.h>
+#include "registry.h"
 #include "node.h"
 #include "houdini.h"
 #include "cmark.h"
@@ -35,3 +36,13 @@ char *cmark_markdown_to_html(const char *text, size_t len, int options) {
 
   return result;
 }
+
+int cmark_init(void) {
+  cmark_discover_plugins();
+  return 1;
+}
+
+int cmark_deinit(void) {
+  cmark_release_plugins();
+  return 1;
+}
diff --git a/src/cmark.h b/src/cmark.h
index 5e04f0c81..6322b5094 100644
--- a/src/cmark.h
+++ b/src/cmark.h
@@ -140,6 +140,22 @@ void          cmark_llist_free_full (cmark_llist       * head,
 CMARK_EXPORT
 void          cmark_llist_free      (cmark_llist       * head);
 
+/**
+ * ## Initialization
+ */
+
+/** Initialize the cmark library. This will discover available plugins.
+ *  Returns 'true' if initialization was successful, 'false' otherwise.
+ */
+CMARK_EXPORT
+int cmark_init(void);
+
+/** Deinitialize the cmark library. This will release all plugins.
+ *  Returns true if deinitialization was successful, 'false' otherwise.
+ */
+CMARK_EXPORT
+int cmark_deinit(void);
+
 /**
  * ## Creating and Destroying Nodes
  */
diff --git a/src/cmark_extension_api.h b/src/cmark_extension_api.h
index 8a2164a99..e74f5676b 100644
--- a/src/cmark_extension_api.h
+++ b/src/cmark_extension_api.h
@@ -33,6 +33,7 @@ extern "C" {
  * desirable.
  */
 
+typedef struct cmark_plugin cmark_plugin;
 
 /** A syntax extension that can be attached to a cmark_parser
  * with cmark_parser_attach_syntax_extension().
@@ -62,6 +63,67 @@ extern "C" {
  */
 typedef struct cmark_syntax_extension cmark_syntax_extension;
 
+/**
+ * ### Plugin API.
+ *
+ * Extensions should be distributed as dynamic libraries,
+ * with a single exported function named after the distributed
+ * filename.
+ *
+ * When discovering extensions (see cmark_init), cmark will
+ * try to load a symbol named "init_{{filename}}" in all the
+ * dynamic libraries it encounters.
+ *
+ * For example, given a dynamic library named myextension.so
+ * (or myextension.dll), cmark will try to load the symbol
+ * named "init_myextension". This means that the filename
+ * must lend itself to forming a valid C identifier, with
+ * the notable exception of dashes, which will be translated
+ * to underscores, which means cmark will look for a function
+ * named "init_my_extension" if it encounters a dynamic library
+ * named "my-extension.so".
+ *
+ * See the 'cmark_plugin_init_func' typedef for the exact prototype
+ * this function should follow.
+ *
+ * For now the extensibility of cmark is not complete, as
+ * it only offers API to hook into the block parsing phase
+ * (<http://spec.commonmark.org/0.24/#phase-1-block-structure>).
+ *
+ * See 'cmark_plugin_register_syntax_extension' for more information.
+ */
+
+/** The prototype plugins' init function should follow.
+ */
+typedef int (*cmark_plugin_init_func)(cmark_plugin *plugin);
+
+/** Register a syntax 'extension' with the 'plugin', it will be made
+ * available as an extension and, if attached to a cmark_parser
+ * with 'cmark_parser_attach_syntax_extension', it will contribute
+ * to the block parsing process.
+ *
+ * See the documentation for 'cmark_syntax_extension' for information
+ * on how to implement one.
+ *
+ * This function will typically be called from the init function
+ * of external modules.
+ *
+ * This takes ownership of 'extension', one should not call
+ * 'cmark_syntax_extension_free' on a registered extension.
+ */
+CMARK_EXPORT
+int cmark_plugin_register_syntax_extension(cmark_plugin *plugin,
+                                            cmark_syntax_extension *extension);
+
+/** This will search for the syntax extension named 'name' among the
+ *  registered syntax extensions.
+ *
+ *  It can then be attached to a cmark_parser
+ *  with the cmark_parser_attach_syntax_extension method.
+ */
+CMARK_EXPORT
+cmark_syntax_extension *cmark_find_syntax_extension(const char *name);
+
 /** Should create and add a new open block to 'parent_container' if
  * 'input' matches a syntax rule for that block type. It is allowed
  * to modify the type of 'parent_container'.
diff --git a/src/config.h.in b/src/config.h.in
index 0f72d6a1f..06c74c305 100644
--- a/src/config.h.in
+++ b/src/config.h.in
@@ -77,6 +77,8 @@ CMARK_INLINE int c99_snprintf(char *outBuf, size_t size, const char *format, ...
 # include <sys/types.h>
 #endif
 
+#define EXTENSION_DIR LIBDIR "/extensions"
+
 #ifdef __cplusplus
 }
 #endif
diff --git a/src/libcmark.pc.in b/src/libcmark.pc.in
index 9c3a9a9e6..8419df378 100644
--- a/src/libcmark.pc.in
+++ b/src/libcmark.pc.in
@@ -6,5 +6,5 @@ includedir=@CMAKE_INSTALL_PREFIX@/include
 Name: libcmark
 Description: CommonMark parsing, rendering, and manipulation
 Version: @PROJECT_VERSION@
-Libs: -L${libdir} -lcmark
+Libs: -L${libdir} -lcmark -ldl
 Cflags: -I${includedir}
diff --git a/src/plugin.c b/src/plugin.c
new file mode 100644
index 000000000..39c361ac7
--- /dev/null
+++ b/src/plugin.c
@@ -0,0 +1,33 @@
+#include <stdlib.h>
+
+#include "plugin.h"
+
+int cmark_plugin_register_syntax_extension(cmark_plugin    * plugin,
+                                        cmark_syntax_extension * extension) {
+  plugin->syntax_extensions = cmark_llist_append(plugin->syntax_extensions, extension);
+  return 1;
+}
+
+cmark_plugin *
+cmark_plugin_new(void) {
+  cmark_plugin *res = malloc(sizeof(cmark_plugin));
+
+  res->syntax_extensions = NULL;
+
+  return res;
+}
+
+void
+cmark_plugin_free(cmark_plugin *plugin) {
+  cmark_llist_free_full(plugin->syntax_extensions,
+                        (cmark_free_func) cmark_syntax_extension_free);
+  free(plugin);
+}
+
+cmark_llist *
+cmark_plugin_steal_syntax_extensions(cmark_plugin *plugin) {
+  cmark_llist *res = plugin->syntax_extensions;
+
+  plugin->syntax_extensions = NULL;
+  return res;
+}
diff --git a/src/plugin.h b/src/plugin.h
new file mode 100644
index 000000000..b9e9d2994
--- /dev/null
+++ b/src/plugin.h
@@ -0,0 +1,34 @@
+#ifndef CMARK_PLUGIN_H
+#define CMARK_PLUGIN_H
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#include "cmark.h"
+#include "cmark_extension_api.h"
+
+/**
+ * cmark_plugin:
+ *
+ * A plugin structure, which should be filled by plugin's
+ * init functions.
+ */
+struct cmark_plugin {
+  cmark_llist *syntax_extensions;
+};
+
+cmark_llist *
+cmark_plugin_steal_syntax_extensions(cmark_plugin *plugin);
+
+cmark_plugin *
+cmark_plugin_new(void);
+
+void
+cmark_plugin_free(cmark_plugin *plugin);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif
diff --git a/src/registry.c b/src/registry.c
new file mode 100644
index 000000000..8f7b9c4e1
--- /dev/null
+++ b/src/registry.c
@@ -0,0 +1,141 @@
+#include <dirent.h>
+#include <dlfcn.h>
+#include <stdint.h>
+#include <stdlib.h>
+#include <string.h>
+
+#include "config.h"
+#include "cmark.h"
+#include "syntax_extension.h"
+#include "registry.h"
+#include "plugin.h"
+
+
+static cmark_llist *syntax_extensions = NULL;
+static cmark_llist *plugin_handles = NULL;
+
+static cmark_plugin *scan_file(char* filename) {
+  char* last_slash = strrchr(filename, '/');
+  char* name_start = last_slash ? last_slash + 1 : filename;
+  char* last_dot = strrchr(filename, '.');
+  cmark_plugin *plugin = NULL;
+  char *init_func_name = NULL;
+  int i;
+  void *libhandle;
+  char *libname = NULL;
+
+  if (!last_dot || strcmp(last_dot, ".so"))
+      goto done;
+
+  libname = malloc(sizeof(char) * (strlen(EXTENSION_DIR) + strlen(filename) + 2));
+  snprintf(libname, strlen(EXTENSION_DIR) + strlen(filename) + 2, "%s/%s",
+      EXTENSION_DIR, filename);
+  libhandle = dlopen(libname, RTLD_NOW);
+  free(libname);
+
+  if (!libhandle) {
+      printf("Error loading DSO: %s\n", dlerror());
+      goto done;
+  }
+
+  name_start[last_dot - name_start] = '\0';
+
+  for (i = 0; name_start[i]; i++) {
+    if (name_start[i] == '-')
+      name_start[i] = '_';
+  }
+
+  init_func_name = malloc(sizeof(char) * (strlen(name_start) + 6));
+
+  snprintf(init_func_name, strlen(name_start) + 6, "init_%s", name_start);
+
+  cmark_plugin_init_func initfunc = (cmark_plugin_init_func)
+      (intptr_t) dlsym(libhandle, init_func_name);
+  free(init_func_name);
+
+  plugin = cmark_plugin_new();
+
+  if (initfunc) {
+    if (initfunc(plugin)) {
+      plugin_handles = cmark_llist_append(plugin_handles, libhandle);
+    } else {
+      cmark_plugin_free(plugin);
+      printf("Error Initializing plugin %s\n", name_start);
+      plugin = NULL;
+      dlclose(libhandle);
+    }
+  } else {
+    printf("Error loading init function: %s\n", dlerror());
+    dlclose(libhandle);
+  }
+
+done:
+  return plugin;
+}
+
+static void scan_path(char *path) {
+  DIR *dir = opendir(path);
+  struct dirent* direntry;
+
+  if (!dir)
+    return;
+
+  while ((direntry = readdir(dir))) {
+    cmark_plugin *plugin = scan_file(direntry->d_name);
+    if (plugin) {
+      cmark_llist *syntax_extensions_list = cmark_plugin_steal_syntax_extensions(plugin);
+      cmark_llist *tmp;
+
+      for (tmp = syntax_extensions_list; tmp; tmp=tmp->next) {
+        syntax_extensions = cmark_llist_append(syntax_extensions, tmp->data);
+      }
+
+      cmark_llist_free(syntax_extensions_list);
+      cmark_plugin_free(plugin);
+    }
+  }
+
+ closedir(dir);
+}
+
+void cmark_discover_plugins(void) {
+  cmark_release_plugins();
+  scan_path(EXTENSION_DIR);
+}
+
+static void
+release_plugin_handle(void *libhandle) {
+  dlclose(libhandle);
+}
+
+void cmark_release_plugins(void) {
+  if (syntax_extensions) {
+    cmark_llist_free_full(syntax_extensions,
+        (cmark_free_func) cmark_syntax_extension_free);
+    syntax_extensions = NULL;
+  }
+
+  cmark_llist_free_full(plugin_handles, release_plugin_handle);
+  plugin_handles = NULL;
+}
+
+cmark_llist *cmark_list_syntax_extensions(void) {
+  cmark_llist *tmp;
+  cmark_llist *res = NULL;
+
+  for (tmp = syntax_extensions; tmp; tmp = tmp->next) {
+    res = cmark_llist_append(res, tmp->data);
+  }
+  return res;
+}
+
+cmark_syntax_extension *cmark_find_syntax_extension(const char *name) {
+  cmark_llist *tmp;
+
+  for (tmp = syntax_extensions; tmp; tmp = tmp->next) {
+    cmark_syntax_extension *ext = (cmark_syntax_extension *) tmp->data;
+    if (!strcmp(ext->name, name))
+      return ext;
+  }
+  return NULL;
+}
diff --git a/src/registry.h b/src/registry.h
new file mode 100644
index 000000000..bc566e010
--- /dev/null
+++ b/src/registry.h
@@ -0,0 +1,18 @@
+#ifndef CMARK_REGISTRY_H
+#define CMARK_REGISTRY_H
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#include "cmark.h"
+
+void cmark_discover_plugins(void);
+void cmark_release_plugins(void);
+cmark_llist *cmark_list_syntax_extensions(void);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif

From e7075293f660148811a4e6c15f2273b70653d148 Mon Sep 17 00:00:00 2001
From: Mathieu Duponchelle <mathieu.duponchelle@opencreed.com>
Date: Sat, 27 Feb 2016 02:31:43 +0100
Subject: [PATCH 10/20] cmark executable: add extension switches

Allow listing and attaching extensions.

Also cleanup valgrind a little by removing exits and using
cleanup gotos
---
 src/main.c | 110 +++++++++++++++++++++++++++++++++++++++++------------
 1 file changed, 85 insertions(+), 25 deletions(-)

diff --git a/src/main.c b/src/main.c
index 88a4f322f..c6e2508c0 100644
--- a/src/main.c
+++ b/src/main.c
@@ -5,6 +5,9 @@
 #include "config.h"
 #include "memory.h"
 #include "cmark.h"
+#include "cmark_extension_api.h"
+#include "syntax_extension.h"
+#include "registry.h"
 
 #if defined(_WIN32) && !defined(__CYGWIN__)
 #include <io.h>
@@ -23,20 +26,22 @@ typedef enum {
 void print_usage() {
   printf("Usage:   cmark [FILE*]\n");
   printf("Options:\n");
-  printf("  --to, -t FORMAT  Specify output format (html, xml, man, "
+  printf("  --to, -t FORMAT                Specify output format (html, xml, man, "
          "commonmark, latex)\n");
-  printf("  --width WIDTH    Specify wrap width (default 0 = nowrap)\n");
-  printf("  --sourcepos      Include source position attribute\n");
-  printf("  --hardbreaks     Treat newlines as hard line breaks\n");
+  printf("  --width WIDTH                  Specify wrap width (default 0 = nowrap)\n");
+  printf("  --sourcepos                    Include source position attribute\n");
+  printf("  --hardbreaks                   Treat newlines as hard line breaks\n");
   printf("  --nobreaks       Render soft line breaks as spaces\n");
-  printf("  --safe           Suppress raw HTML and dangerous URLs\n");
-  printf("  --smart          Use smart punctuation\n");
-  printf("  --normalize      Consolidate adjacent text nodes\n");
-  printf("  --help, -h       Print usage information\n");
-  printf("  --version        Print version\n");
+  printf("  --safe                         Suppress raw HTML and dangerous URLs\n");
+  printf("  --smart                        Use smart punctuation\n");
+  printf("  --normalize                    Consolidate adjacent text nodes\n");
+  printf("  -e, --extension EXTENSION_NAME Specify an extension name to use\n");
+  printf("  --list-extensions              List available extensions and quit\n");
+  printf("  --help, -h                     Print usage information\n");
+  printf("  --version                      Print version\n");
 }
 
-static void print_document(cmark_node *document, writer_format writer,
+static bool print_document(cmark_node *document, writer_format writer,
                            int options, int width) {
   char *result;
 
@@ -58,23 +63,44 @@ static void print_document(cmark_node *document, writer_format writer,
     break;
   default:
     fprintf(stderr, "Unknown format %d\n", writer);
-    exit(1);
+    return false;
   }
+
   printf("%s", result);
   free(result);
+
+  return true;
+}
+
+static void print_extensions(void) {
+  cmark_llist *syntax_extensions;
+  cmark_llist *tmp;
+
+  printf ("Available extensions:\n");
+
+  syntax_extensions = cmark_list_syntax_extensions();
+  for (tmp = syntax_extensions; tmp; tmp=tmp->next) {
+    cmark_syntax_extension *ext = (cmark_syntax_extension *) tmp->data;
+    printf("%s\n", ext->name);
+  }
+
+  cmark_llist_free(syntax_extensions);
 }
 
 int main(int argc, char *argv[]) {
   int i, numfps = 0;
   int *files;
   char buffer[4096];
-  cmark_parser *parser;
+  cmark_parser *parser = NULL;
   size_t bytes;
-  cmark_node *document;
+  cmark_node *document = NULL;
   int width = 0;
   char *unparsed;
   writer_format writer = FORMAT_HTML;
   int options = CMARK_OPT_DEFAULT;
+  int res = 1;
+
+  cmark_init();
 
 #if defined(_WIN32) && !defined(__CYGWIN__)
   _setmode(_fileno(stdin), _O_BINARY);
@@ -87,7 +113,10 @@ int main(int argc, char *argv[]) {
     if (strcmp(argv[i], "--version") == 0) {
       printf("cmark %s", CMARK_VERSION_STRING);
       printf(" - CommonMark converter\n(C) 2014-2016 John MacFarlane\n");
-      exit(0);
+      goto success;
+    } else if (strcmp(argv[i], "--list-extensions") == 0) {
+      print_extensions();
+      goto success;
     } else if (strcmp(argv[i], "--sourcepos") == 0) {
       options |= CMARK_OPT_SOURCEPOS;
     } else if (strcmp(argv[i], "--hardbreaks") == 0) {
@@ -105,7 +134,7 @@ int main(int argc, char *argv[]) {
     } else if ((strcmp(argv[i], "--help") == 0) ||
                (strcmp(argv[i], "-h") == 0)) {
       print_usage();
-      exit(0);
+      goto success;
     } else if (strcmp(argv[i], "--width") == 0) {
       i += 1;
       if (i < argc) {
@@ -113,11 +142,11 @@ int main(int argc, char *argv[]) {
         if (unparsed && strlen(unparsed) > 0) {
           fprintf(stderr, "failed parsing width '%s' at '%s'\n", argv[i],
                   unparsed);
-          exit(1);
+          goto failure;
         }
       } else {
         fprintf(stderr, "--width requires an argument\n");
-        exit(1);
+        goto failure;
       }
     } else if ((strcmp(argv[i], "-t") == 0) || (strcmp(argv[i], "--to") == 0)) {
       i += 1;
@@ -134,27 +163,48 @@ int main(int argc, char *argv[]) {
           writer = FORMAT_LATEX;
         } else {
           fprintf(stderr, "Unknown format %s\n", argv[i]);
-          exit(1);
+          goto failure;
         }
       } else {
         fprintf(stderr, "No argument provided for %s\n", argv[i - 1]);
-        exit(1);
+        goto failure;
       }
+    } else if ((strcmp(argv[i], "-e") == 0) || (strcmp(argv[i], "--extension") == 0)) {
+      i += 1; // Simpler to handle extensions in a second pass, as we can directly register
+              // them with the parser.
     } else if (*argv[i] == '-') {
       print_usage();
-      exit(1);
+      goto failure;
     } else { // treat as file argument
       files[numfps++] = i;
     }
   }
 
   parser = cmark_parser_new(options);
+
+  for (i = 1; i < argc; i++) {
+    if ((strcmp(argv[i], "-e") == 0) || (strcmp(argv[i], "--extension") == 0)) {
+      i += 1;
+      if (i < argc) {
+        cmark_syntax_extension *syntax_extension = cmark_find_syntax_extension(argv[i]);
+        if (!syntax_extension) {
+          fprintf(stderr, "Unknown extension %s\n", argv[i]);
+          goto failure;
+        }
+        cmark_parser_attach_syntax_extension(parser, syntax_extension);
+      } else {
+        fprintf(stderr, "No argument provided for %s\n", argv[i - 1]);
+        goto failure;
+      }
+    }
+  }
+
   for (i = 0; i < numfps; i++) {
     FILE *fp = fopen(argv[files[i]], "rb");
     if (fp == NULL) {
       fprintf(stderr, "Error opening file %s: %s\n", argv[files[i]],
               strerror(errno));
-      exit(1);
+      goto failure;
     }
 
     while ((bytes = fread(buffer, 1, sizeof(buffer), fp)) > 0) {
@@ -178,13 +228,23 @@ int main(int argc, char *argv[]) {
   }
 
   document = cmark_parser_finish(parser);
-  cmark_parser_free(parser);
 
-  print_document(document, writer, options, width);
+  if (!print_document(document, writer, options, width))
+    goto failure;
+
+
+success:
+  res = 0;
+
+failure:
+  if (parser)
+    cmark_parser_free(parser);
 
-  cmark_node_free(document);
+  if (document)
+    cmark_node_free(document);
 
   free(files);
+  cmark_deinit();
 
-  return 0;
+  return res;
 }

From fe08f83202c61abcd09041280f1874cc74babe25 Mon Sep 17 00:00:00 2001
From: Mathieu Duponchelle <mathieu.duponchelle@opencreed.com>
Date: Sun, 28 Feb 2016 18:07:34 +0100
Subject: [PATCH 11/20] Define blocks constituting a table.

We have no syntax rules yet for creating them natively, but
future extensions may provide some.
---
 src/blocks.c     | 12 +++++--
 src/cmark.h      | 12 ++++++-
 src/commonmark.c | 27 ++++++++++++++++
 src/html.c       | 63 ++++++++++++++++++++++++++++++++++++-
 src/latex.c      | 38 ++++++++++++++++++++++
 src/man.c        | 34 ++++++++++++++++++++
 src/node.c       | 82 ++++++++++++++++++++++++++++++++++++++++++++++++
 src/node.h       | 10 ++++++
 8 files changed, 274 insertions(+), 4 deletions(-)

diff --git a/src/blocks.c b/src/blocks.c
index 7d057efed..d0e958575 100644
--- a/src/blocks.c
+++ b/src/blocks.c
@@ -153,7 +153,14 @@ static bool is_blank(cmark_strbuf *s, bufsize_t offset) {
 }
 
 static CMARK_INLINE bool can_contain(cmark_node_type parent_type,
-                                     cmark_node_type child_type) {
+                               cmark_node_type child_type) {
+  if (parent_type == CMARK_NODE_TABLE) {
+    return child_type == CMARK_NODE_TABLE_ROW;
+  }
+
+  if (parent_type == CMARK_NODE_TABLE_ROW)
+    return child_type == CMARK_NODE_TABLE_CELL;
+
   return (parent_type == CMARK_NODE_DOCUMENT ||
           parent_type == CMARK_NODE_BLOCK_QUOTE ||
           parent_type == CMARK_NODE_ITEM ||
@@ -168,7 +175,8 @@ static CMARK_INLINE bool accepts_lines(cmark_node_type block_type) {
 
 static CMARK_INLINE bool contains_inlines(cmark_node_type block_type) {
   return (block_type == CMARK_NODE_PARAGRAPH ||
-          block_type == CMARK_NODE_HEADING);
+          block_type == CMARK_NODE_HEADING ||
+          block_type == CMARK_NODE_TABLE_CELL);
 }
 
 static void add_line(cmark_node *node, cmark_chunk *ch, cmark_parser *parser) {
diff --git a/src/cmark.h b/src/cmark.h
index 6322b5094..4076176a2 100644
--- a/src/cmark.h
+++ b/src/cmark.h
@@ -46,8 +46,13 @@ typedef enum {
   CMARK_NODE_HEADING,
   CMARK_NODE_THEMATIC_BREAK,
 
+  /* blocks with no syntax rules in the current specification */
+  CMARK_NODE_TABLE,
+  CMARK_NODE_TABLE_ROW,
+  CMARK_NODE_TABLE_CELL,
+
   CMARK_NODE_FIRST_BLOCK = CMARK_NODE_DOCUMENT,
-  CMARK_NODE_LAST_BLOCK = CMARK_NODE_THEMATIC_BREAK,
+  CMARK_NODE_LAST_BLOCK = CMARK_NODE_TABLE_CELL,
 
   /* Inline */
   CMARK_NODE_TEXT,
@@ -460,6 +465,11 @@ CMARK_EXPORT int cmark_node_get_end_line(cmark_node *node);
  */
 CMARK_EXPORT int cmark_node_get_end_column(cmark_node *node);
 
+CMARK_EXPORT int cmark_node_get_n_table_columns(cmark_node *node);
+CMARK_EXPORT int cmark_node_set_n_table_columns(cmark_node *node, int n_columns);
+CMARK_EXPORT int cmark_node_is_table_header(cmark_node *node);
+CMARK_EXPORT int cmark_node_set_is_table_header(cmark_node *node, int is_table_header);
+
 /**
  * ## Tree Manipulation
  */
diff --git a/src/commonmark.c b/src/commonmark.c
index f1589f576..486dc441c 100644
--- a/src/commonmark.c
+++ b/src/commonmark.c
@@ -333,6 +333,33 @@ static int S_render_node(cmark_renderer *renderer, cmark_node *node,
     }
     break;
 
+  case CMARK_NODE_TABLE:
+    BLANKLINE();
+    break;
+
+  case CMARK_NODE_TABLE_ROW:
+    if (entering) {
+      CR();
+      LIT("|");
+    }
+    break;
+  case CMARK_NODE_TABLE_CELL:
+    if (entering) {
+    } else {
+      LIT(" |");
+      if (node->parent->as.table_row.is_header && !node->next) {
+        int i;
+        int n_cols = node->parent->parent->as.table.n_columns;
+        CR();
+        LIT("|");
+        for (i = 0; i < n_cols; i++) {
+          LIT(" --- |");
+        }
+        CR();
+      }
+    }
+    break;
+
   case CMARK_NODE_TEXT:
     OUT(cmark_node_get_literal(node), allow_wrap, NORMAL);
     break;
diff --git a/src/html.c b/src/html.c
index a680e4a50..f8f51026c 100644
--- a/src/html.c
+++ b/src/html.c
@@ -27,6 +27,8 @@ static CMARK_INLINE void cr(cmark_strbuf *html) {
 struct render_state {
   cmark_strbuf *html;
   cmark_node *plain;
+  bool need_closing_table_body;
+  bool in_table_header;
 };
 
 static void S_render_sourcepos(cmark_node *node, cmark_strbuf *html,
@@ -217,6 +219,65 @@ static int S_render_node(cmark_node *node, cmark_event_type ev_type,
     }
     break;
 
+  case CMARK_NODE_TABLE:
+    if (entering) {
+      cr(html);
+      cmark_strbuf_puts(html, "<table");
+      S_render_sourcepos(node, html, options);
+      cmark_strbuf_putc(html, '>');
+      state->need_closing_table_body = false;
+    } else {
+      if (state->need_closing_table_body)
+        cmark_strbuf_puts(html, "</tbody>");
+      state->need_closing_table_body = false;
+      cmark_strbuf_puts(html, "</table>");
+    }
+    break;
+
+  case CMARK_NODE_TABLE_ROW:
+   if (entering) {
+     cr(html);
+     if (node->as.table_row.is_header) {
+       state->in_table_header = true;
+       cmark_strbuf_puts(html, "<thead>");
+       cr(html);
+     }
+     cmark_strbuf_puts(html, "<tr");
+     S_render_sourcepos(node, html, options);
+     cmark_strbuf_putc(html, '>');
+   } else {
+     cr(html);
+     cmark_strbuf_puts(html, "</tr>");
+     if (node->as.table_row.is_header) {
+       cr(html);
+       cmark_strbuf_puts(html, "</thead>");
+       cr(html);
+       cmark_strbuf_puts(html, "<tbody>");
+       state->need_closing_table_body = true;
+       state->in_table_header = false;
+     }
+   }
+   break;
+
+  case CMARK_NODE_TABLE_CELL:
+   if (entering) {
+     cr(html);
+     if (state->in_table_header) {
+       cmark_strbuf_puts(html, "<th");
+     } else {
+       cmark_strbuf_puts(html, "<td");
+     }
+     S_render_sourcepos(node, html, options);
+     cmark_strbuf_putc(html, '>');
+   } else {
+     if (state->in_table_header) {
+       cmark_strbuf_puts(html, "</th>");
+     } else {
+       cmark_strbuf_puts(html, "</td>");
+     }
+   }
+   break;
+
   case CMARK_NODE_TEXT:
     escape_html(html, node->as.literal.data, node->as.literal.len);
     break;
@@ -327,7 +388,7 @@ char *cmark_render_html(cmark_node *root, int options) {
   cmark_strbuf html = CMARK_BUF_INIT(cmark_node_mem(root));
   cmark_event_type ev_type;
   cmark_node *cur;
-  struct render_state state = {&html, NULL};
+  struct render_state state = {&html, NULL, false, false};
   cmark_iter *iter = cmark_iter_new(root);
 
   while ((ev_type = cmark_iter_next(iter)) != CMARK_EVENT_DONE) {
diff --git a/src/latex.c b/src/latex.c
index 7c3decda0..306a508de 100644
--- a/src/latex.c
+++ b/src/latex.c
@@ -332,6 +332,44 @@ static int S_render_node(cmark_renderer *renderer, cmark_node *node,
     }
     break;
 
+  case CMARK_NODE_TABLE:
+    if (entering) {
+      int i, n_cols;
+      CR();
+      LIT("\\begin{table}");
+      CR();
+      LIT("\\begin{tabular}{");
+
+      n_cols = node->as.table.n_columns;
+      for (i = 0; i < n_cols; i++) {
+        LIT("l");
+      }
+      LIT("}");
+      CR();
+    } else {
+      LIT("\\end{tabular}");
+      CR();
+      LIT("\\end{table}");
+      CR();
+    }
+    break;
+
+  case CMARK_NODE_TABLE_ROW:
+    if (!entering) {
+      CR();
+    }
+    break;
+
+  case CMARK_NODE_TABLE_CELL:
+    if (!entering) {
+      if (node->next) {
+        LIT(" & ");
+      } else {
+        LIT(" \\\\");
+      }
+    }
+    break;
+
   case CMARK_NODE_TEXT:
     OUT(cmark_node_get_literal(node), allow_wrap, NORMAL);
     break;
diff --git a/src/man.c b/src/man.c
index 1c76f68bb..a2691b063 100644
--- a/src/man.c
+++ b/src/man.c
@@ -173,6 +173,40 @@ static int S_render_node(cmark_renderer *renderer, cmark_node *node,
     }
     break;
 
+  case CMARK_NODE_TABLE:
+    if (entering) {
+      int i, n_cols;
+      CR();
+      LIT(".TS");
+      CR();
+      LIT("tab(@);");
+      CR();
+
+      n_cols = node->as.table.n_columns;
+
+      for (i = 0; i < n_cols; i++) {
+        LIT("c");
+      }
+
+      if (n_cols) {
+        LIT(".");
+        CR();
+      }
+    } else {
+      LIT(".TE");
+      CR();
+    }
+    break;
+  case CMARK_NODE_TABLE_ROW:
+    if (!entering) {
+      CR();
+    }
+    break;
+  case CMARK_NODE_TABLE_CELL:
+    if (!entering && node->next) {
+      LIT("@");
+    }
+    break;
   case CMARK_NODE_TEXT:
     OUT(cmark_node_get_literal(node), allow_wrap, NORMAL);
     break;
diff --git a/src/node.c b/src/node.c
index 7cfa7147e..7743ae19b 100644
--- a/src/node.c
+++ b/src/node.c
@@ -67,6 +67,17 @@ static bool S_can_contain(cmark_node *node, cmark_node *child) {
   case CMARK_NODE_IMAGE:
   case CMARK_NODE_CUSTOM_INLINE:
     return S_is_inline(child);
+  case CMARK_NODE_TABLE:
+    return child->type == CMARK_NODE_TABLE_ROW;
+  case CMARK_NODE_TABLE_ROW:
+    return child->type == CMARK_NODE_TABLE_CELL;
+  case CMARK_NODE_TABLE_CELL:
+    return child->type == CMARK_NODE_TEXT ||
+           child->type == CMARK_NODE_CODE ||
+           child->type == CMARK_NODE_EMPH ||
+           child->type == CMARK_NODE_STRONG ||
+           child->type == CMARK_NODE_LINK ||
+           child->type == CMARK_NODE_IMAGE;
 
   default:
     break;
@@ -213,6 +224,15 @@ const char *cmark_node_get_type_string(cmark_node *node) {
     return "html_block";
   case CMARK_NODE_CUSTOM_BLOCK:
     return "custom_block";
+  case CMARK_NODE_TABLE:
+    return "table";
+  case CMARK_NODE_TABLE_ROW:
+    if (node->as.table_row.is_header)
+      return "table_header";
+    else
+      return "table_row";
+  case CMARK_NODE_TABLE_CELL:
+    return "table_cell";
   case CMARK_NODE_PARAGRAPH:
     return "paragraph";
   case CMARK_NODE_HEADING:
@@ -739,6 +759,68 @@ int cmark_node_get_end_column(cmark_node *node) {
   return node->end_column;
 }
 
+int cmark_node_get_n_table_columns(cmark_node *node) {
+  if (node == NULL) {
+    return -1;
+  }
+
+  switch (node->type) {
+  case CMARK_NODE_TABLE:
+    return node->as.table.n_columns;
+  default:
+    break;
+  }
+
+  return -1;
+}
+
+int cmark_node_set_n_table_columns(cmark_node *node, int n_columns) {
+  if (node == NULL) {
+    return 0;
+  }
+
+  switch (node->type) {
+    case CMARK_NODE_TABLE:
+      node->as.table.n_columns = n_columns;
+      return 1;
+    default:
+      break;
+  }
+
+  return 0;
+}
+
+int cmark_node_is_table_header(cmark_node *node) {
+  if (node == NULL) {
+    return 0;
+  }
+
+  switch (node->type) {
+  case CMARK_NODE_TABLE_ROW:
+    return node->as.table_row.is_header;
+  default:
+    break;
+  }
+
+  return 1;
+}
+
+int cmark_node_set_is_table_header(cmark_node *node, int is_table_header) {
+  if (node == NULL) {
+    return 0;
+  }
+
+  switch (node->type) {
+    case CMARK_NODE_TABLE_ROW:
+      node->as.table_row.is_header = is_table_header;
+      return 1;
+    default:
+      break;
+  }
+
+  return 0;
+}
+
 // Unlink a node without adjusting its next, prev, and parent pointers.
 static void S_node_unlink(cmark_node *node) {
   if (node == NULL) {
diff --git a/src/node.h b/src/node.h
index 93932aae0..cbb0e551b 100644
--- a/src/node.h
+++ b/src/node.h
@@ -47,6 +47,14 @@ typedef struct {
   cmark_chunk on_exit;
 } cmark_custom;
 
+typedef struct {
+  int n_columns;
+} cmark_table;
+
+typedef struct {
+  bool is_header;
+} cmark_table_row;
+
 enum cmark_node__internal_flags {
   CMARK_NODE__OPEN = (1 << 0),
   CMARK_NODE__LAST_LINE_BLANK = (1 << 1),
@@ -80,6 +88,8 @@ struct cmark_node {
     cmark_heading heading;
     cmark_link link;
     cmark_custom custom;
+    cmark_table table;
+    cmark_table_row table_row;
     int html_block_type;
   } as;
 };

From 634833ca5edb015685bd656e73935f46974187ca Mon Sep 17 00:00:00 2001
From: Mathieu Duponchelle <mathieu.duponchelle@opencreed.com>
Date: Sat, 27 Feb 2016 02:41:24 +0100
Subject: [PATCH 12/20] Extensions: implement an example "core extension"

---
 CMakeLists.txt               |   1 +
 Makefile                     |  14 +++
 extensions/CMakeLists.txt    |  32 +++++
 extensions/core-extensions.c | 237 +++++++++++++++++++++++++++++++++++
 extensions/ext_scanners.h    |  20 +++
 extensions/ext_scanners.re   |  65 ++++++++++
 6 files changed, 369 insertions(+)
 create mode 100644 extensions/CMakeLists.txt
 create mode 100644 extensions/core-extensions.c
 create mode 100644 extensions/ext_scanners.h
 create mode 100644 extensions/ext_scanners.re

diff --git a/CMakeLists.txt b/CMakeLists.txt
index 2ab6a7257..b9abddc55 100755
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -16,6 +16,7 @@ set(PROJECT_VERSION ${PROJECT_VERSION_MAJOR}.${PROJECT_VERSION_MINOR}.${PROJECT_
 option(CMARK_TESTS "Build cmark tests and enable testing" ON)
 
 add_subdirectory(src)
+add_subdirectory(extensions)
 if(CMARK_TESTS)
   add_subdirectory(api_test)
 endif()
diff --git a/Makefile b/Makefile
index 039089e19..5bb1fd2db 100644
--- a/Makefile
+++ b/Makefile
@@ -1,4 +1,5 @@
 SRCDIR=src
+EXTDIR=extensions
 DATADIR=data
 BUILDDIR?=build
 GENERATOR?=Unix Makefiles
@@ -115,6 +116,19 @@ $(SRCDIR)/scanners.c: $(SRCDIR)/scanners.re
 		--encoding-policy substitute -o $@ $<
 	clang-format -style llvm -i $@
 
+# We include scanners.c in the repository, so this shouldn't
+# normally need to be generated.
+$(EXTDIR)/ext_scanners.c: $(EXTDIR)/ext_scanners.re
+	@case "$$(re2c -v)" in \
+	    *\ 0.13.*|*\ 0.14|*\ 0.14.1) \
+		echo "re2c >= 0.14.2 is required"; \
+		false; \
+		;; \
+	esac
+	re2c --case-insensitive -b -i --no-generation-date -8 \
+		--encoding-policy substitute -o $@ $<
+	clang-format -style llvm -i $@
+
 # We include entities.inc in the repository, so normally this
 # doesn't need to be regenerated:
 $(SRCDIR)/entities.inc: tools/make_entities_inc.py
diff --git a/extensions/CMakeLists.txt b/extensions/CMakeLists.txt
new file mode 100644
index 000000000..85d9e4450
--- /dev/null
+++ b/extensions/CMakeLists.txt
@@ -0,0 +1,32 @@
+cmake_minimum_required(VERSION 2.8)
+set(LIBRARY "cmarkextensions")
+set(LIBRARY_SOURCES
+    ${PROJECT_SOURCE_DIR}/src/buffer.c
+    ${PROJECT_SOURCE_DIR}/src/cmark_ctype.c
+    core-extensions.c
+    ext_scanners.c
+    ext_scanners.h
+   )
+
+include_directories(
+  ${PROJECT_SOURCE_DIR}/src
+  ${PROJECT_BINARY_DIR}/src
+)
+
+# We make LIB_INSTALL_DIR configurable rather than
+# hard-coding lib, because on some OSes different locations
+# are used for different architectures (e.g. /usr/lib64 on
+# 64-bit Fedora).
+if(NOT LIB_INSTALL_DIR)
+  set(LIB_INSTALL_DIR "lib" CACHE STRING
+  "Set the installation directory for libraries." FORCE)
+endif(NOT LIB_INSTALL_DIR)
+
+include_directories(. ${CMAKE_CURRENT_BINARY_DIR})
+
+set(CMAKE_C_FLAGS_PROFILE "${CMAKE_C_FLAGS_RELEASE} -pg")
+set(CMAKE_LINKER_PROFILE "${CMAKE_LINKER_FLAGS_RELEASE} -pg")
+
+add_library(${LIBRARY} SHARED ${LIBRARY_SOURCES})
+
+target_link_libraries(cmarkextensions libcmark)
diff --git a/extensions/core-extensions.c b/extensions/core-extensions.c
new file mode 100644
index 000000000..6b650dd4b
--- /dev/null
+++ b/extensions/core-extensions.c
@@ -0,0 +1,237 @@
+#include <stdio.h>
+#include <string.h>
+
+#include <cmark.h>
+#include <cmark_extension_api.h>
+
+#include "buffer.h"
+#include "ext_scanners.h"
+
+typedef struct {
+  int n_columns;
+  cmark_llist *cells;
+} table_row;
+
+static void free_table_cell(void *data) {
+  cmark_strbuf_free((cmark_strbuf *) data);
+  free(data);
+}
+
+static void free_table_row(table_row *row) {
+
+  if (!row)
+    return;
+
+  cmark_llist_free_full(row->cells, (cmark_free_func) free_table_cell);
+
+  free(row);
+}
+
+static cmark_strbuf *unescape_pipes(unsigned char *string, bufsize_t len)
+{
+  cmark_strbuf *res = (cmark_strbuf *)malloc(sizeof(cmark_strbuf));
+  bufsize_t r, w;
+
+  cmark_strbuf_init(res, len + 1);
+  cmark_strbuf_put(res, string, len);
+  cmark_strbuf_putc(res, '\0');
+
+  for (r = 0, w = 0; r < len; ++r) {
+    if (res->ptr[r] == '\\' && res->ptr[r + 1] == '|')
+      r++;
+
+    res->ptr[w++] = res->ptr[r];
+  }
+
+  cmark_strbuf_truncate(res, w);
+
+  return res;
+}
+
+static table_row *row_from_string(unsigned char *string, int len) {
+  table_row *row = NULL;
+  bufsize_t cell_matched = 0;
+  bufsize_t cell_offset = 0;
+
+  row = malloc(sizeof(table_row));
+  row->n_columns = 0;
+  row->cells = NULL;
+
+  do {
+    cell_matched = scan_table_cell(string, len, cell_offset);
+    if (cell_matched) {
+      cmark_strbuf *cell_buf = unescape_pipes(string + cell_offset + 1,
+          cell_matched - 1);
+      row->n_columns += 1;
+      row->cells = cmark_llist_append(row->cells, cell_buf);
+    }
+    cell_offset += cell_matched;
+  } while (cell_matched);
+
+  cell_matched = scan_table_row_end(string, len, cell_offset);
+  cell_offset += cell_matched;
+
+  if (!cell_matched || cell_offset != len) {
+    free_table_row(row);
+    row = NULL;
+  }
+
+  return row;
+}
+
+static cmark_node *try_opening_table_header(cmark_syntax_extension *self,
+                                            cmark_parser * parser,
+                                            cmark_node   * parent_container,
+                                            unsigned char   * input,
+                                            int len) {
+  bufsize_t matched = scan_table_start(input, len, cmark_parser_get_first_nonspace(parser));
+  cmark_node *table_header;
+  table_row *header_row = NULL;
+  table_row *marker_row = NULL;
+  const char *parent_string;
+
+  if (!matched)
+    goto done;
+
+  parent_string = cmark_node_get_string_content(parent_container);
+
+  header_row = row_from_string((unsigned char *) parent_string, strlen(parent_string));
+
+  if (!header_row) {
+    goto done;
+  }
+
+  marker_row = row_from_string(input + cmark_parser_get_first_nonspace(parser),
+      len - cmark_parser_get_first_nonspace(parser));
+
+  assert(marker_row);
+
+  if (header_row->n_columns != marker_row->n_columns) {
+    goto done;
+  }
+
+  if (!cmark_node_set_type(parent_container, CMARK_NODE_TABLE)) {
+    goto done;
+  }
+
+  cmark_node_set_syntax_extension(parent_container, self);
+  cmark_node_set_n_table_columns(parent_container, header_row->n_columns);
+
+  table_header = cmark_parser_add_child(parser, parent_container,
+      CMARK_NODE_TABLE_ROW, cmark_parser_get_offset(parser));
+  cmark_node_set_syntax_extension(table_header, self);
+  cmark_node_set_is_table_header(table_header, true);
+
+  {
+    cmark_llist *tmp;
+
+    for (tmp = header_row->cells; tmp; tmp = tmp->next) {
+      cmark_strbuf *cell_buf = (cmark_strbuf *) tmp->data;
+      cmark_node *header_cell = cmark_parser_add_child(parser, table_header,
+          CMARK_NODE_TABLE_CELL, cmark_parser_get_offset(parser));
+      cmark_node_set_string_content(header_cell, (char *) cell_buf->ptr);
+      cmark_node_set_syntax_extension(header_cell, self);
+    }
+  }
+
+  cmark_parser_advance_offset(parser, input,
+                   strlen(input) - 1 - cmark_parser_get_offset(parser),
+                   false);
+done:
+  free_table_row(header_row);
+  free_table_row(marker_row);
+  return parent_container;
+}
+
+static cmark_node *try_opening_table_row(cmark_syntax_extension *self,
+                                         cmark_parser * parser,
+                                         cmark_node   * parent_container,
+                                         unsigned char   * input,
+                                         int len) {
+  cmark_node *table_row_block;
+  table_row *row;
+
+  if (cmark_parser_is_blank(parser))
+    return NULL;
+
+  table_row_block = cmark_parser_add_child(parser, parent_container,
+      CMARK_NODE_TABLE_ROW, cmark_parser_get_offset(parser));
+
+  cmark_node_set_syntax_extension(table_row_block, self);
+
+  /* We don't advance the offset here */
+
+  row = row_from_string(input + cmark_parser_get_first_nonspace(parser),
+      len - cmark_parser_get_first_nonspace(parser));
+
+  {
+    cmark_llist *tmp;
+
+    for (tmp = row->cells; tmp; tmp = tmp->next) {
+      cmark_strbuf *cell_buf = (cmark_strbuf *) tmp->data;
+      cmark_node *cell = cmark_parser_add_child(parser, table_row_block,
+          CMARK_NODE_TABLE_CELL, cmark_parser_get_offset(parser));
+      cmark_node_set_string_content(cell, (char *) cell_buf->ptr);
+      cmark_node_set_syntax_extension(cell, self);
+    }
+  }
+
+  free_table_row(row);
+
+  cmark_parser_advance_offset(parser, input,
+                   len - 1 - cmark_parser_get_offset(parser),
+                   false);
+
+  return table_row_block;
+}
+
+static cmark_node *try_opening_table_block(cmark_syntax_extension * syntax_extension,
+                                           int               indented,
+                                           cmark_parser    * parser,
+                                           cmark_node      * parent_container,
+                                           unsigned char      * input,
+                                           int len) {
+  cmark_node_type parent_type = cmark_node_get_type(parent_container);
+
+  if (!indented && parent_type == CMARK_NODE_PARAGRAPH) {
+    return try_opening_table_header(syntax_extension, parser, parent_container, input, len);
+  } else if (!indented && parent_type == CMARK_NODE_TABLE) {
+    return try_opening_table_row(syntax_extension, parser, parent_container, input, len);
+  }
+
+  return NULL;
+}
+
+static int table_matches(cmark_syntax_extension *self,
+                          cmark_parser * parser,
+                          unsigned char   * input,
+                          int len,
+                          cmark_node   * parent_container) {
+  int res = 0;
+
+  if (cmark_node_get_type(parent_container) == CMARK_NODE_TABLE) {
+    table_row *new_row = row_from_string(input + cmark_parser_get_first_nonspace(parser),
+        len - cmark_parser_get_first_nonspace(parser));
+    if (new_row) {
+        if (new_row->n_columns == cmark_node_get_n_table_columns(parent_container))
+          res = 1;
+    }
+    free_table_row(new_row);
+  }
+
+  return res;
+}
+
+static cmark_syntax_extension *register_table_syntax_extension(void) {
+  cmark_syntax_extension *ext = cmark_syntax_extension_new("piped-tables");
+
+  cmark_syntax_extension_set_match_block_func(ext, table_matches);
+  cmark_syntax_extension_set_open_block_func(ext, try_opening_table_block);
+
+  return ext;
+}
+
+int init_libcmarkextensions(cmark_plugin *plugin) {
+  cmark_plugin_register_syntax_extension(plugin, register_table_syntax_extension());
+  return 1;
+}
diff --git a/extensions/ext_scanners.h b/extensions/ext_scanners.h
new file mode 100644
index 000000000..c96b18490
--- /dev/null
+++ b/extensions/ext_scanners.h
@@ -0,0 +1,20 @@
+#include "cmark.h"
+#include "chunk.h"
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+bufsize_t _ext_scan_at(bufsize_t (*scanner)(const unsigned char *), unsigned char *ptr,
+                       int len, bufsize_t offset);
+bufsize_t _scan_table_start(const unsigned char *p);
+bufsize_t _scan_table_cell(const unsigned char *p);
+bufsize_t _scan_table_row_end(const unsigned char *p);
+
+#define scan_table_start(c, l, n) _ext_scan_at(&_scan_table_start, c, l, n)
+#define scan_table_cell(c, l, n) _ext_scan_at(&_scan_table_cell, c, l, n)
+#define scan_table_row_end(c, l, n) _ext_scan_at(&_scan_table_row_end, c, l, n)
+
+#ifdef __cplusplus
+}
+#endif
diff --git a/extensions/ext_scanners.re b/extensions/ext_scanners.re
new file mode 100644
index 000000000..7ad561f51
--- /dev/null
+++ b/extensions/ext_scanners.re
@@ -0,0 +1,65 @@
+#include <stdlib.h>
+#include "ext_scanners.h"
+
+bufsize_t _ext_scan_at(bufsize_t (*scanner)(const unsigned char *), unsigned char *ptr, int len, bufsize_t offset)
+{
+	bufsize_t res;
+
+        if (ptr == NULL || offset > len) {
+          return 0;
+        } else {
+	  unsigned char lim = ptr[len];
+
+	  ptr[len] = '\0';
+	  res = scanner(ptr + offset);
+	  ptr[len] = lim;
+        }
+
+	return res;
+}
+
+/*!re2c
+  re2c:define:YYCTYPE  = "unsigned char";
+  re2c:define:YYCURSOR = p;
+  re2c:define:YYMARKER = marker;
+  re2c:define:YYCTXMARKER = marker;
+  re2c:yyfill:enable = 0;
+
+  spacechar = [ \t\v\f];
+  newline = [\r]?[\n];
+
+  escaped_char = [\\][|!"#$%&'()*+,./:;<=>?@[\\\]^_`{}~-];
+
+  table_marker = [|](spacechar*[-]+spacechar*);
+  table_cell = [|](escaped_char|[^|\r\n])+;
+*/
+
+bufsize_t _scan_table_cell(const unsigned char *p)
+{
+  const unsigned char *marker = NULL;
+  const unsigned char *start = p;
+/*!re2c
+  table_cell { return (bufsize_t)(p - start); }
+  .? { return 0; }
+*/
+}
+
+bufsize_t _scan_table_row_end(const unsigned char *p)
+{
+  const unsigned char *marker = NULL;
+  const unsigned char *start = p;
+/*!re2c
+  [|]newline { return (bufsize_t)(p - start); }
+  .? { return 0; }
+*/
+}
+
+bufsize_t _scan_table_start(const unsigned char *p)
+{
+  const unsigned char *marker = NULL;
+  const unsigned char *start = p;
+/*!re2c
+  (table_marker)+ [|]newline { return (bufsize_t)(p - start); }
+  .? { return 0; }
+*/
+}

From 228be72ef294f35788cd18d9cbbf19653a2eb7bb Mon Sep 17 00:00:00 2001
From: Mathieu Duponchelle <mathieu.duponchelle@opencreed.com>
Date: Sat, 27 Feb 2016 02:42:23 +0100
Subject: [PATCH 13/20] Check in extension scanners separately

---
 extensions/core-extensions.c |  17 +-
 extensions/ext_scanners.c    | 585 +++++++++++++++++++++++++++++++++++
 2 files changed, 594 insertions(+), 8 deletions(-)
 create mode 100644 extensions/ext_scanners.c

diff --git a/extensions/core-extensions.c b/extensions/core-extensions.c
index 6b650dd4b..9b904ed86 100644
--- a/extensions/core-extensions.c
+++ b/extensions/core-extensions.c
@@ -4,6 +4,7 @@
 #include <cmark.h>
 #include <cmark_extension_api.h>
 
+#include "parser.h"
 #include "buffer.h"
 #include "ext_scanners.h"
 
@@ -27,12 +28,12 @@ static void free_table_row(table_row *row) {
   free(row);
 }
 
-static cmark_strbuf *unescape_pipes(unsigned char *string, bufsize_t len)
+static cmark_strbuf *unescape_pipes(cmark_mem *mem, unsigned char *string, bufsize_t len)
 {
   cmark_strbuf *res = (cmark_strbuf *)malloc(sizeof(cmark_strbuf));
   bufsize_t r, w;
 
-  cmark_strbuf_init(res, len + 1);
+  cmark_strbuf_init(mem, res, len + 1);
   cmark_strbuf_put(res, string, len);
   cmark_strbuf_putc(res, '\0');
 
@@ -48,7 +49,7 @@ static cmark_strbuf *unescape_pipes(unsigned char *string, bufsize_t len)
   return res;
 }
 
-static table_row *row_from_string(unsigned char *string, int len) {
+static table_row *row_from_string(cmark_mem *mem, unsigned char *string, int len) {
   table_row *row = NULL;
   bufsize_t cell_matched = 0;
   bufsize_t cell_offset = 0;
@@ -60,7 +61,7 @@ static table_row *row_from_string(unsigned char *string, int len) {
   do {
     cell_matched = scan_table_cell(string, len, cell_offset);
     if (cell_matched) {
-      cmark_strbuf *cell_buf = unescape_pipes(string + cell_offset + 1,
+      cmark_strbuf *cell_buf = unescape_pipes(mem, string + cell_offset + 1,
           cell_matched - 1);
       row->n_columns += 1;
       row->cells = cmark_llist_append(row->cells, cell_buf);
@@ -95,13 +96,13 @@ static cmark_node *try_opening_table_header(cmark_syntax_extension *self,
 
   parent_string = cmark_node_get_string_content(parent_container);
 
-  header_row = row_from_string((unsigned char *) parent_string, strlen(parent_string));
+  header_row = row_from_string(parser->mem, (unsigned char *) parent_string, strlen(parent_string));
 
   if (!header_row) {
     goto done;
   }
 
-  marker_row = row_from_string(input + cmark_parser_get_first_nonspace(parser),
+  marker_row = row_from_string(parser->mem, input + cmark_parser_get_first_nonspace(parser),
       len - cmark_parser_get_first_nonspace(parser));
 
   assert(marker_row);
@@ -161,7 +162,7 @@ static cmark_node *try_opening_table_row(cmark_syntax_extension *self,
 
   /* We don't advance the offset here */
 
-  row = row_from_string(input + cmark_parser_get_first_nonspace(parser),
+  row = row_from_string(parser->mem, input + cmark_parser_get_first_nonspace(parser),
       len - cmark_parser_get_first_nonspace(parser));
 
   {
@@ -210,7 +211,7 @@ static int table_matches(cmark_syntax_extension *self,
   int res = 0;
 
   if (cmark_node_get_type(parent_container) == CMARK_NODE_TABLE) {
-    table_row *new_row = row_from_string(input + cmark_parser_get_first_nonspace(parser),
+    table_row *new_row = row_from_string(parser->mem, input + cmark_parser_get_first_nonspace(parser),
         len - cmark_parser_get_first_nonspace(parser));
     if (new_row) {
         if (new_row->n_columns == cmark_node_get_n_table_columns(parent_container))
diff --git a/extensions/ext_scanners.c b/extensions/ext_scanners.c
new file mode 100644
index 000000000..78df8d242
--- /dev/null
+++ b/extensions/ext_scanners.c
@@ -0,0 +1,585 @@
+/* Generated by re2c 0.16 */
+#include <stdlib.h>
+#include "ext_scanners.h"
+
+bufsize_t _ext_scan_at(bufsize_t (*scanner)(const unsigned char *),
+                       unsigned char *ptr, int len, bufsize_t offset) {
+  bufsize_t res;
+
+  if (ptr == NULL || offset > len) {
+    return 0;
+  } else {
+    unsigned char lim = ptr[len];
+
+    ptr[len] = '\0';
+    res = scanner(ptr + offset);
+    ptr[len] = lim;
+  }
+
+  return res;
+}
+
+bufsize_t _scan_table_cell(const unsigned char *p) {
+  const unsigned char *marker = NULL;
+  const unsigned char *start = p;
+
+  {
+    unsigned char yych;
+    unsigned int yyaccept = 0;
+    static const unsigned char yybm[] = {
+        64, 64, 64,  64, 64, 64, 64, 64, 64, 64, 0,  64, 64, 0,  64, 64, 64, 64,
+        64, 64, 64,  64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64,
+        64, 64, 64,  64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64,
+        64, 64, 64,  64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64,
+        64, 64, 64,  64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64,
+        64, 64, 128, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64,
+        64, 64, 64,  64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 0,  64,
+        64, 64, 0,   0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,
+        0,  0,  0,   0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,
+        0,  0,  0,   0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,
+        0,  0,  0,   0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,
+        0,  0,  0,   0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,
+        0,  0,  0,   0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,
+        0,  0,  0,   0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,
+        0,  0,  0,   0,
+    };
+    yych = *(marker = p);
+    if (yych <= 0xDF) {
+      if (yych <= '{') {
+        if (yych != '\n')
+          goto yy3;
+      } else {
+        if (yych <= '|')
+          goto yy4;
+        if (yych <= 0x7F)
+          goto yy3;
+        if (yych >= 0xC2)
+          goto yy5;
+      }
+    } else {
+      if (yych <= 0xEF) {
+        if (yych <= 0xE0)
+          goto yy7;
+        if (yych == 0xED)
+          goto yy9;
+        goto yy8;
+      } else {
+        if (yych <= 0xF0)
+          goto yy10;
+        if (yych <= 0xF3)
+          goto yy11;
+        if (yych <= 0xF4)
+          goto yy12;
+      }
+    }
+  yy2 : { return 0; }
+  yy3:
+    yych = *++p;
+    goto yy2;
+  yy4:
+    yyaccept = 0;
+    yych = *(marker = ++p);
+    if (yych <= '{') {
+      if (yych <= '\n') {
+        if (yych <= '\t')
+          goto yy14;
+        goto yy2;
+      } else {
+        if (yych == '\r')
+          goto yy2;
+        goto yy14;
+      }
+    } else {
+      if (yych <= 0x7F) {
+        if (yych <= '|')
+          goto yy2;
+        goto yy14;
+      } else {
+        if (yych <= 0xC1)
+          goto yy2;
+        if (yych <= 0xF4)
+          goto yy14;
+        goto yy2;
+      }
+    }
+  yy5:
+    yych = *++p;
+    if (yych <= 0x7F)
+      goto yy6;
+    if (yych <= 0xBF)
+      goto yy3;
+  yy6:
+    p = marker;
+    if (yyaccept == 0) {
+      goto yy2;
+    } else {
+      goto yy15;
+    }
+  yy7:
+    yych = *++p;
+    if (yych <= 0x9F)
+      goto yy6;
+    if (yych <= 0xBF)
+      goto yy5;
+    goto yy6;
+  yy8:
+    yych = *++p;
+    if (yych <= 0x7F)
+      goto yy6;
+    if (yych <= 0xBF)
+      goto yy5;
+    goto yy6;
+  yy9:
+    yych = *++p;
+    if (yych <= 0x7F)
+      goto yy6;
+    if (yych <= 0x9F)
+      goto yy5;
+    goto yy6;
+  yy10:
+    yych = *++p;
+    if (yych <= 0x8F)
+      goto yy6;
+    if (yych <= 0xBF)
+      goto yy8;
+    goto yy6;
+  yy11:
+    yych = *++p;
+    if (yych <= 0x7F)
+      goto yy6;
+    if (yych <= 0xBF)
+      goto yy8;
+    goto yy6;
+  yy12:
+    yych = *++p;
+    if (yych <= 0x7F)
+      goto yy6;
+    if (yych <= 0x8F)
+      goto yy8;
+    goto yy6;
+  yy13:
+    yyaccept = 1;
+    marker = ++p;
+    yych = *p;
+  yy14:
+    if (yybm[0 + yych] & 64) {
+      goto yy13;
+    }
+    if (yych <= 0xEC) {
+      if (yych <= 0xC1) {
+        if (yych <= '\r')
+          goto yy15;
+        if (yych <= '\\')
+          goto yy16;
+      } else {
+        if (yych <= 0xDF)
+          goto yy18;
+        if (yych <= 0xE0)
+          goto yy19;
+        goto yy20;
+      }
+    } else {
+      if (yych <= 0xF0) {
+        if (yych <= 0xED)
+          goto yy21;
+        if (yych <= 0xEF)
+          goto yy20;
+        goto yy22;
+      } else {
+        if (yych <= 0xF3)
+          goto yy23;
+        if (yych <= 0xF4)
+          goto yy24;
+      }
+    }
+  yy15 : { return (bufsize_t)(p - start); }
+  yy16:
+    yyaccept = 1;
+    marker = ++p;
+    yych = *p;
+    if (yybm[0 + yych] & 128) {
+      goto yy16;
+    }
+    if (yych <= 0xDF) {
+      if (yych <= '\f') {
+        if (yych == '\n')
+          goto yy15;
+        goto yy13;
+      } else {
+        if (yych <= '\r')
+          goto yy15;
+        if (yych <= 0x7F)
+          goto yy13;
+        if (yych <= 0xC1)
+          goto yy15;
+      }
+    } else {
+      if (yych <= 0xEF) {
+        if (yych <= 0xE0)
+          goto yy19;
+        if (yych == 0xED)
+          goto yy21;
+        goto yy20;
+      } else {
+        if (yych <= 0xF0)
+          goto yy22;
+        if (yych <= 0xF3)
+          goto yy23;
+        if (yych <= 0xF4)
+          goto yy24;
+        goto yy15;
+      }
+    }
+  yy18:
+    ++p;
+    yych = *p;
+    if (yych <= 0x7F)
+      goto yy6;
+    if (yych <= 0xBF)
+      goto yy13;
+    goto yy6;
+  yy19:
+    ++p;
+    yych = *p;
+    if (yych <= 0x9F)
+      goto yy6;
+    if (yych <= 0xBF)
+      goto yy18;
+    goto yy6;
+  yy20:
+    ++p;
+    yych = *p;
+    if (yych <= 0x7F)
+      goto yy6;
+    if (yych <= 0xBF)
+      goto yy18;
+    goto yy6;
+  yy21:
+    ++p;
+    yych = *p;
+    if (yych <= 0x7F)
+      goto yy6;
+    if (yych <= 0x9F)
+      goto yy18;
+    goto yy6;
+  yy22:
+    ++p;
+    yych = *p;
+    if (yych <= 0x8F)
+      goto yy6;
+    if (yych <= 0xBF)
+      goto yy20;
+    goto yy6;
+  yy23:
+    ++p;
+    yych = *p;
+    if (yych <= 0x7F)
+      goto yy6;
+    if (yych <= 0xBF)
+      goto yy20;
+    goto yy6;
+  yy24:
+    ++p;
+    yych = *p;
+    if (yych <= 0x7F)
+      goto yy6;
+    if (yych <= 0x8F)
+      goto yy20;
+    goto yy6;
+  }
+}
+
+bufsize_t _scan_table_row_end(const unsigned char *p) {
+  const unsigned char *marker = NULL;
+  const unsigned char *start = p;
+
+  {
+    unsigned char yych;
+    yych = *(marker = p);
+    if (yych <= 0xDF) {
+      if (yych <= '{') {
+        if (yych != '\n')
+          goto yy28;
+      } else {
+        if (yych <= '|')
+          goto yy29;
+        if (yych <= 0x7F)
+          goto yy28;
+        if (yych >= 0xC2)
+          goto yy30;
+      }
+    } else {
+      if (yych <= 0xEF) {
+        if (yych <= 0xE0)
+          goto yy32;
+        if (yych == 0xED)
+          goto yy34;
+        goto yy33;
+      } else {
+        if (yych <= 0xF0)
+          goto yy35;
+        if (yych <= 0xF3)
+          goto yy36;
+        if (yych <= 0xF4)
+          goto yy37;
+      }
+    }
+  yy27 : { return 0; }
+  yy28:
+    yych = *++p;
+    goto yy27;
+  yy29:
+    yych = *(marker = ++p);
+    if (yych == '\n')
+      goto yy38;
+    if (yych == '\r')
+      goto yy40;
+    goto yy27;
+  yy30:
+    yych = *++p;
+    if (yych <= 0x7F)
+      goto yy31;
+    if (yych <= 0xBF)
+      goto yy28;
+  yy31:
+    p = marker;
+    goto yy27;
+  yy32:
+    yych = *++p;
+    if (yych <= 0x9F)
+      goto yy31;
+    if (yych <= 0xBF)
+      goto yy30;
+    goto yy31;
+  yy33:
+    yych = *++p;
+    if (yych <= 0x7F)
+      goto yy31;
+    if (yych <= 0xBF)
+      goto yy30;
+    goto yy31;
+  yy34:
+    yych = *++p;
+    if (yych <= 0x7F)
+      goto yy31;
+    if (yych <= 0x9F)
+      goto yy30;
+    goto yy31;
+  yy35:
+    yych = *++p;
+    if (yych <= 0x8F)
+      goto yy31;
+    if (yych <= 0xBF)
+      goto yy33;
+    goto yy31;
+  yy36:
+    yych = *++p;
+    if (yych <= 0x7F)
+      goto yy31;
+    if (yych <= 0xBF)
+      goto yy33;
+    goto yy31;
+  yy37:
+    yych = *++p;
+    if (yych <= 0x7F)
+      goto yy31;
+    if (yych <= 0x8F)
+      goto yy33;
+    goto yy31;
+  yy38:
+    ++p;
+    { return (bufsize_t)(p - start); }
+  yy40:
+    ++p;
+    if ((yych = *p) == '\n')
+      goto yy38;
+    goto yy31;
+  }
+}
+
+bufsize_t _scan_table_start(const unsigned char *p) {
+  const unsigned char *marker = NULL;
+  const unsigned char *start = p;
+
+  {
+    unsigned char yych;
+    static const unsigned char yybm[] = {
+        0, 0,   0, 0, 0, 0, 0, 0, 0, 64, 0,  64, 64, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+        0, 0,   0, 0, 0, 0, 0, 0, 0, 0,  64, 0,  0,  0, 0, 0, 0, 0, 0, 0, 0, 0,
+        0, 128, 0, 0, 0, 0, 0, 0, 0, 0,  0,  0,  0,  0, 0, 0, 0, 0, 0, 0, 0, 0,
+        0, 0,   0, 0, 0, 0, 0, 0, 0, 0,  0,  0,  0,  0, 0, 0, 0, 0, 0, 0, 0, 0,
+        0, 0,   0, 0, 0, 0, 0, 0, 0, 0,  0,  0,  0,  0, 0, 0, 0, 0, 0, 0, 0, 0,
+        0, 0,   0, 0, 0, 0, 0, 0, 0, 0,  0,  0,  0,  0, 0, 0, 0, 0, 0, 0, 0, 0,
+        0, 0,   0, 0, 0, 0, 0, 0, 0, 0,  0,  0,  0,  0, 0, 0, 0, 0, 0, 0, 0, 0,
+        0, 0,   0, 0, 0, 0, 0, 0, 0, 0,  0,  0,  0,  0, 0, 0, 0, 0, 0, 0, 0, 0,
+        0, 0,   0, 0, 0, 0, 0, 0, 0, 0,  0,  0,  0,  0, 0, 0, 0, 0, 0, 0, 0, 0,
+        0, 0,   0, 0, 0, 0, 0, 0, 0, 0,  0,  0,  0,  0, 0, 0, 0, 0, 0, 0, 0, 0,
+        0, 0,   0, 0, 0, 0, 0, 0, 0, 0,  0,  0,  0,  0, 0, 0, 0, 0, 0, 0, 0, 0,
+        0, 0,   0, 0, 0, 0, 0, 0, 0, 0,  0,  0,  0,  0,
+    };
+    yych = *(marker = p);
+    if (yych <= 0xDF) {
+      if (yych <= '{') {
+        if (yych != '\n')
+          goto yy44;
+      } else {
+        if (yych <= '|')
+          goto yy45;
+        if (yych <= 0x7F)
+          goto yy44;
+        if (yych >= 0xC2)
+          goto yy46;
+      }
+    } else {
+      if (yych <= 0xEF) {
+        if (yych <= 0xE0)
+          goto yy48;
+        if (yych == 0xED)
+          goto yy50;
+        goto yy49;
+      } else {
+        if (yych <= 0xF0)
+          goto yy51;
+        if (yych <= 0xF3)
+          goto yy52;
+        if (yych <= 0xF4)
+          goto yy53;
+      }
+    }
+  yy43 : { return 0; }
+  yy44:
+    yych = *++p;
+    goto yy43;
+  yy45:
+    yych = *(marker = ++p);
+    if (yybm[0 + yych] & 64) {
+      goto yy54;
+    }
+    if (yych == '-')
+      goto yy56;
+    goto yy43;
+  yy46:
+    yych = *++p;
+    if (yych <= 0x7F)
+      goto yy47;
+    if (yych <= 0xBF)
+      goto yy44;
+  yy47:
+    p = marker;
+    goto yy43;
+  yy48:
+    yych = *++p;
+    if (yych <= 0x9F)
+      goto yy47;
+    if (yych <= 0xBF)
+      goto yy46;
+    goto yy47;
+  yy49:
+    yych = *++p;
+    if (yych <= 0x7F)
+      goto yy47;
+    if (yych <= 0xBF)
+      goto yy46;
+    goto yy47;
+  yy50:
+    yych = *++p;
+    if (yych <= 0x7F)
+      goto yy47;
+    if (yych <= 0x9F)
+      goto yy46;
+    goto yy47;
+  yy51:
+    yych = *++p;
+    if (yych <= 0x8F)
+      goto yy47;
+    if (yych <= 0xBF)
+      goto yy49;
+    goto yy47;
+  yy52:
+    yych = *++p;
+    if (yych <= 0x7F)
+      goto yy47;
+    if (yych <= 0xBF)
+      goto yy49;
+    goto yy47;
+  yy53:
+    yych = *++p;
+    if (yych <= 0x7F)
+      goto yy47;
+    if (yych <= 0x8F)
+      goto yy49;
+    goto yy47;
+  yy54:
+    ++p;
+    yych = *p;
+    if (yybm[0 + yych] & 64) {
+      goto yy54;
+    }
+    if (yych != '-')
+      goto yy47;
+  yy56:
+    ++p;
+    yych = *p;
+    if (yybm[0 + yych] & 128) {
+      goto yy56;
+    }
+    if (yych <= '\f') {
+      if (yych == '\t')
+        goto yy58;
+      if (yych <= '\n')
+        goto yy47;
+    } else {
+      if (yych <= ' ') {
+        if (yych <= 0x1F)
+          goto yy47;
+      } else {
+        if (yych == '|')
+          goto yy60;
+        goto yy47;
+      }
+    }
+  yy58:
+    ++p;
+    yych = *p;
+    if (yych <= '\f') {
+      if (yych == '\t')
+        goto yy58;
+      if (yych <= '\n')
+        goto yy47;
+      goto yy58;
+    } else {
+      if (yych <= ' ') {
+        if (yych <= 0x1F)
+          goto yy47;
+        goto yy58;
+      } else {
+        if (yych != '|')
+          goto yy47;
+      }
+    }
+  yy60:
+    ++p;
+    yych = *p;
+    if (yybm[0 + yych] & 64) {
+      goto yy54;
+    }
+    if (yych <= '\r') {
+      if (yych <= 0x08)
+        goto yy47;
+      if (yych >= '\v')
+        goto yy63;
+    } else {
+      if (yych == '-')
+        goto yy56;
+      goto yy47;
+    }
+  yy61:
+    ++p;
+    { return (bufsize_t)(p - start); }
+  yy63:
+    ++p;
+    if ((yych = *p) == '\n')
+      goto yy61;
+    goto yy47;
+  }
+}

From c174e30acc9a49969406e1ab3dfcf247efe4a938 Mon Sep 17 00:00:00 2001
From: Mathieu Duponchelle <mathieu.duponchelle@opencreed.com>
Date: Fri, 4 Mar 2016 20:49:01 +0100
Subject: [PATCH 14/20] inlines: Expose inline parser

---
 src/cmark_extension_api.h |  96 +++++++++++++++++++++++++++
 src/inlines.c             | 132 ++++++++++++++++++++++++++++++++++----
 2 files changed, 216 insertions(+), 12 deletions(-)

diff --git a/src/cmark_extension_api.h b/src/cmark_extension_api.h
index e74f5676b..fb3264e91 100644
--- a/src/cmark_extension_api.h
+++ b/src/cmark_extension_api.h
@@ -63,6 +63,21 @@ typedef struct cmark_plugin cmark_plugin;
  */
 typedef struct cmark_syntax_extension cmark_syntax_extension;
 
+typedef struct subject cmark_inline_parser;
+
+/** Exposed raw for now */
+
+typedef struct delimiter {
+  struct delimiter *previous;
+  struct delimiter *next;
+  cmark_node *inl_text;
+  int position;
+  unsigned char delim_char;
+  int can_open;
+  int can_close;
+  int active;
+} delimiter;
+
 /**
  * ### Plugin API.
  *
@@ -359,6 +374,87 @@ CMARK_EXPORT cmark_syntax_extension *cmark_node_get_syntax_extension(cmark_node
 CMARK_EXPORT int cmark_node_set_syntax_extension(cmark_node *node,
                                                   cmark_syntax_extension *extension);
 
+/**
+ * ## Inline syntax extension helpers
+ *
+ * The inline parsing process is described in detail at
+ * <http://spec.commonmark.org/0.24/#phase-2-inline-structure>
+ */
+
+/** Should return 'true' if the predicate matches 'c', 'false' otherwise
+ */
+typedef int (*cmark_inline_predicate)(int c);
+
+/** Advance the current inline parsing offset */
+CMARK_EXPORT
+void cmark_inline_parser_advance_offset(cmark_inline_parser *parser);
+
+/** Get the current inline parsing offset */
+CMARK_EXPORT
+int cmark_inline_parser_get_offset(cmark_inline_parser *parser);
+
+/** Get the character located at the current inline parsing offset
+ */
+CMARK_EXPORT
+unsigned char cmark_inline_parser_peek_char(cmark_inline_parser *parser);
+
+/** Get the character located 'pos' bytes in the current line.
+ */
+CMARK_EXPORT
+unsigned char cmark_inline_parser_peek_at(cmark_inline_parser *parser, int pos);
+
+/** Whether the inline parser has reached the end of the current line
+ */
+CMARK_EXPORT
+int cmark_inline_parser_is_eof(cmark_inline_parser *parser);
+
+/** Get the characters located after the current inline parsing offset
+ * while 'pred' matches. Free after usage.
+ */
+CMARK_EXPORT
+char *cmark_inline_parser_take_while(cmark_inline_parser *parser, cmark_inline_predicate pred);
+
+/** Push a delimiter on the delimiter stack.
+ * See <<http://spec.commonmark.org/0.24/#phase-2-inline-structure> for
+ * more information on the parameters
+ */
+CMARK_EXPORT
+void cmark_inline_parser_push_delimiter(cmark_inline_parser *parser,
+                                  unsigned char c,
+                                  int can_open,
+                                  int can_close,
+                                  cmark_node *inl_text);
+
+/** Remove 'delim' from the delimiter stack
+ */
+CMARK_EXPORT
+void cmark_inline_parser_remove_delimiter(cmark_inline_parser *parser, delimiter *delim);
+
+CMARK_EXPORT
+delimiter *cmark_inline_parser_get_last_delimiter(cmark_inline_parser *parser);
+
+/** Convenience function to scan a given delimiter.
+ *
+ * 'left_flanking' and 'right_flanking' will be set to true if they
+ * respectively precede and follow a non-space, non-punctuation
+ * character.
+ *
+ * Additionally, 'punct_before' and 'punct_after' will respectively be set
+ * if the preceding or following character is a punctuation character.
+ *
+ * Note that 'left_flanking' and 'right_flanking' can both be 'true'.
+ *
+ * Returns the number of delimiters encountered, in the limit
+ * of 'max_delims', and advances the inline parsing offset.
+ */
+CMARK_EXPORT
+int cmark_inline_parser_scan_delimiters(cmark_inline_parser *parser,
+                                  int max_delims,
+                                  unsigned char c,
+                                  int *left_flanking,
+                                  int *right_flanking,
+                                  int *punct_before,
+                                  int *punct_after);
 #ifdef __cplusplus
 }
 #endif
diff --git a/src/inlines.c b/src/inlines.c
index 8f18e6c96..999a9cfca 100644
--- a/src/inlines.c
+++ b/src/inlines.c
@@ -30,18 +30,7 @@ static const char *RIGHTSINGLEQUOTE = "\xE2\x80\x99";
 #define make_emph(mem) make_simple(mem, CMARK_NODE_EMPH)
 #define make_strong(mem) make_simple(mem, CMARK_NODE_STRONG)
 
-typedef struct delimiter {
-  struct delimiter *previous;
-  struct delimiter *next;
-  cmark_node *inl_text;
-  bufsize_t position;
-  unsigned char delim_char;
-  bool can_open;
-  bool can_close;
-  bool active;
-} delimiter;
-
-typedef struct {
+typedef struct subject{
   cmark_mem *mem;
   cmark_chunk input;
   bufsize_t pos;
@@ -1127,3 +1116,122 @@ bufsize_t cmark_parse_reference_inline(cmark_mem *mem, cmark_strbuf *input,
   cmark_reference_create(refmap, &lab, &url, &title);
   return subj.pos;
 }
+
+unsigned char cmark_inline_parser_peek_char(cmark_inline_parser *parser) {
+  return peek_char(parser);
+}
+
+unsigned char cmark_inline_parser_peek_at(cmark_inline_parser *parser, bufsize_t pos) {
+  return peek_at(parser, pos);
+}
+
+int cmark_inline_parser_is_eof(cmark_inline_parser *parser) {
+  return is_eof(parser);
+}
+
+static char *
+my_strndup (const char *s, size_t n)
+{
+  char *result;
+  size_t len = strlen (s);
+
+  if (n < len)
+    len = n;
+
+  result = (char *) malloc (len + 1);
+  if (!result)
+    return 0;
+
+  result[len] = '\0';
+  return (char *) memcpy (result, s, len);
+}
+
+char *cmark_inline_parser_take_while(cmark_inline_parser *parser, cmark_inline_predicate pred) {
+  unsigned char c;
+  bufsize_t startpos = parser->pos;
+  bufsize_t len = 0;
+
+  while ((c = peek_char(parser)) && (*pred)(c)) {
+    advance(parser);
+    len++;
+  }
+
+  return my_strndup((const char *) parser->input.data + startpos, len);
+}
+
+void cmark_inline_parser_push_delimiter(cmark_inline_parser *parser,
+                                  unsigned char c,
+                                  int can_open,
+                                  int can_close,
+                                  cmark_node *inl_text) {
+  push_delimiter(parser, c, can_open, can_close, inl_text);
+}
+
+void cmark_inline_parser_remove_delimiter(cmark_inline_parser *parser, delimiter *delim) {
+  remove_delimiter(parser, delim);
+}
+
+int cmark_inline_parser_scan_delimiters(cmark_inline_parser *parser,
+                                  int max_delims,
+                                  unsigned char c,
+                                  int *left_flanking,
+                                  int *right_flanking,
+                                  int *punct_before,
+                                  int *punct_after) {
+  int numdelims = 0;
+  bufsize_t before_char_pos;
+  int32_t after_char = 0;
+  int32_t before_char = 0;
+  int len;
+  bool space_before, space_after;
+
+  if (parser->pos == 0) {
+    before_char = 10;
+  } else {
+    before_char_pos = parser->pos - 1;
+    // walk back to the beginning of the UTF_8 sequence:
+    while (peek_at(parser, before_char_pos) >> 6 == 2 && before_char_pos > 0) {
+      before_char_pos -= 1;
+    }
+    len = cmark_utf8proc_iterate(parser->input.data + before_char_pos,
+                                 parser->pos - before_char_pos, &before_char);
+    if (len == -1) {
+      before_char = 10;
+    }
+  }
+
+  while (peek_char(parser) == c && numdelims <= max_delims) {
+    numdelims++;
+    advance(parser);
+  }
+
+  len = cmark_utf8proc_iterate(parser->input.data + parser->pos,
+                               parser->input.len - parser->pos, &after_char);
+  if (len == -1) {
+    after_char = 10;
+  }
+
+  *punct_before = cmark_utf8proc_is_punctuation(before_char);
+  *punct_after = cmark_utf8proc_is_punctuation(after_char);
+  space_before = cmark_utf8proc_is_space(before_char);
+  space_after = cmark_utf8proc_is_space(after_char);
+
+  *left_flanking = numdelims > 0 && !cmark_utf8proc_is_space(after_char) &&
+                  !(*punct_after && !space_before && !*punct_before);
+  *right_flanking = numdelims > 0 && !cmark_utf8proc_is_space(before_char) &&
+                  !(*punct_before && !space_after && !*punct_after);
+
+  return numdelims;
+}
+
+void cmark_inline_parser_advance_offset(cmark_inline_parser *parser) {
+  advance(parser);
+}
+
+int cmark_inline_parser_get_offset(cmark_inline_parser *parser) {
+  return parser->pos;
+}
+
+delimiter *cmark_inline_parser_get_last_delimiter(cmark_inline_parser *parser) {
+  return parser->last_delim;
+}

From 315f477305e75a284bcd6f2ed9b565fc1d790253 Mon Sep 17 00:00:00 2001
From: Mathieu Duponchelle <mathieu.duponchelle@opencreed.com>
Date: Sat, 5 Mar 2016 02:27:57 +0100
Subject: [PATCH 15/20] inline extension draft

---
 src/blocks.c              |  37 +++++++--
 src/cmark_extension_api.h |  67 ++++++++++++++++
 src/inlines.c             | 160 ++++++++++++++++++++++++++++----------
 src/inlines.h             |   7 +-
 src/parser.h              |   1 +
 src/syntax_extension.c    |  16 ++++
 src/syntax_extension.h    |   9 ++-
 7 files changed, 245 insertions(+), 52 deletions(-)

diff --git a/src/blocks.c b/src/blocks.c
index d0e958575..562658d69 100644
--- a/src/blocks.c
+++ b/src/blocks.c
@@ -84,6 +84,11 @@ static cmark_node *make_document(cmark_mem *mem) {
 int cmark_parser_attach_syntax_extension(cmark_parser *parser,
                                       cmark_syntax_extension *extension) {
   parser->syntax_extensions = cmark_llist_append(parser->syntax_extensions, extension);
+  if (extension->match_inline && extension->insert_inline_from_delim) {
+    parser->inline_syntax_extensions = cmark_llist_append(
+        parser->inline_syntax_extensions, extension);
+  }
+
   return 1;
 }
 
@@ -111,6 +116,7 @@ cmark_parser *cmark_parser_new_with_mem(int options, cmark_mem *mem) {
   parser->options = options;
   parser->last_buffer_ended_with_cr = false;
   parser->syntax_extensions = NULL;
+  parser->inline_syntax_extensions = NULL;
 
   return parser;
 }
@@ -126,6 +132,7 @@ void cmark_parser_free(cmark_parser *parser) {
   cmark_strbuf_free(&parser->linebuf);
   cmark_reference_map_free(parser->refmap);
   cmark_llist_free(parser->syntax_extensions);
+  cmark_llist_free(parser->inline_syntax_extensions);
   mem->free(parser);
 }
 
@@ -392,23 +399,43 @@ static cmark_node *add_child(cmark_parser *parser, cmark_node *parent,
   return child;
 }
 
+static void manage_extensions_special_characters(cmark_parser *parser, bool add) {
+  cmark_llist *tmp_ext;
+
+  for (tmp_ext = parser->inline_syntax_extensions; tmp_ext; tmp_ext=tmp_ext->next) {
+    cmark_syntax_extension *ext = (cmark_syntax_extension *) tmp_ext->data;
+    cmark_llist *tmp_char;
+    for (tmp_char = ext->special_inline_chars; tmp_char; tmp_char=tmp_char->next) {
+      unsigned char c = (unsigned char) (unsigned long) tmp_char->data;
+      if (add)
+        cmark_inlines_add_special_character(c);
+      else
+        cmark_inlines_remove_special_character(c);
+    }
+  }
+}
+
 // Walk through node and all children, recursively, parsing
 // string content into inline content where appropriate.
-static void process_inlines(cmark_mem *mem, cmark_node *root, cmark_reference_map *refmap,
+static void process_inlines(cmark_parser *parser, cmark_reference_map *refmap,
                             int options) {
-  cmark_iter *iter = cmark_iter_new(root);
+  cmark_iter *iter = cmark_iter_new(parser->root);
   cmark_node *cur;
   cmark_event_type ev_type;
 
+  manage_extensions_special_characters(parser, true);
+
   while ((ev_type = cmark_iter_next(iter)) != CMARK_EVENT_DONE) {
     cur = cmark_iter_get_node(iter);
     if (ev_type == CMARK_EVENT_ENTER) {
-      if (contains_inlines(S_type(cur))) {
-        cmark_parse_inlines(mem, cur, refmap, options);
+      if (contains_inlines(cur->type)) {
+        cmark_parse_inlines(parser, cur, refmap, options);
       }
     }
   }
 
+  manage_extensions_special_characters(parser, false);
+
   cmark_iter_free(iter);
 }
 
@@ -487,7 +514,7 @@ static cmark_node *finalize_document(cmark_parser *parser) {
   }
 
   finalize(parser, parser->root);
-  process_inlines(parser->mem, parser->root, parser->refmap, parser->options);
+  process_inlines(parser, parser->refmap, parser->options);
 
   return parser->root;
 }
diff --git a/src/cmark_extension_api.h b/src/cmark_extension_api.h
index fb3264e91..636df7a12 100644
--- a/src/cmark_extension_api.h
+++ b/src/cmark_extension_api.h
@@ -60,6 +60,43 @@ typedef struct cmark_plugin cmark_plugin;
  * new block with cmark_parser_make_block and cmark_parser_add_child.
  * If no function was provided is NULL, the extension will have
  * no effect at all on the final block structure of the AST.
+ *
+ * #### Inline parsing phase hooks
+ *
+ * For each character provided by the extension through
+ * 'cmark_syntax_extension_set_special_inline_chars',
+ * the function provided by the extension through
+ * 'cmark_syntax_extension_set_match_inline_func'
+ * will get called, it is the responsibility of the extension
+ * to scan the characters located at the current inline parsing offset
+ * with the cmark_inline_parser API.
+ *
+ * Depending on the type of the extension, it can either:
+ *
+ * * Scan forward, determine that the syntax matches and return
+ *   a newly-created inline node with the appropriate type.
+ *   This is the technique that would be used if inline code
+ *   (with backticks) was implemented as an extension.
+ * * Scan only the character(s) that its syntax rules require
+ *   for opening and closing nodes, push a delimiter on the
+ *   delimiter stack, and return a simple text node with its
+ *   contents set to the character(s) consumed.
+ *   This is the technique that would be used if emphasis
+ *   inlines were implemented as an extension.
+ *
+ * When an extension has pushed delimiters on the stack,
+ * the function provided through
+ * 'cmark_syntax_extension_set_inline_from_delim_func'
+ * will get called in a latter phase,
+ * when the inline parser has matched opener and closer delimiters
+ * created by the extension together.
+ *
+ * It is then the responsibility of the extension to modify
+ * and populate the opener inline text node, and to remove
+ * the necessary delimiters from the delimiter stack.
+ *
+ * Finally, the extension should return NULL if its scan didn't
+ * match its syntax rules.
  */
 typedef struct cmark_syntax_extension cmark_syntax_extension;
 
@@ -153,6 +190,18 @@ typedef cmark_node * (*cmark_open_block_func) (cmark_syntax_extension *extension
                                        unsigned char *input,
                                        int len);
 
+typedef cmark_node *(*cmark_match_inline_func)(cmark_syntax_extension *extension,
+                                       cmark_parser *parser,
+                                       cmark_node *parent,
+                                       unsigned char character,
+                                       cmark_inline_parser *inline_parser);
+
+typedef delimiter *(*cmark_inline_from_delim_func)(cmark_syntax_extension *extension,
+                                           cmark_parser *parser,
+                                           cmark_inline_parser *inline_parser,
+                                           delimiter *opener,
+                                           delimiter *closer);
+
 /** Should return 'true' if 'input' can be contained in 'container',
  *  'false' otherwise.
  */
@@ -184,6 +233,24 @@ CMARK_EXPORT
 void cmark_syntax_extension_set_match_block_func(cmark_syntax_extension *extension,
                                                  cmark_match_block_func func);
 
+/** See the documentation for 'cmark_syntax_extension'
+ */
+CMARK_EXPORT
+void cmark_syntax_extension_set_match_inline_func(cmark_syntax_extension *extension,
+                                                  cmark_match_inline_func func);
+
+/** See the documentation for 'cmark_syntax_extension'
+ */
+CMARK_EXPORT
+void cmark_syntax_extension_set_inline_from_delim_func(cmark_syntax_extension *extension,
+                                                       cmark_inline_from_delim_func func);
+
+/** See the documentation for 'cmark_syntax_extension'
+ */
+CMARK_EXPORT
+void cmark_syntax_extension_set_special_inline_chars(cmark_syntax_extension *extension,
+                                                     cmark_llist *special_chars);
+
 /** Return the index of the line currently being parsed, starting with 1.
  */
 CMARK_EXPORT
diff --git a/src/inlines.c b/src/inlines.c
index 999a9cfca..72f4cfd75 100644
--- a/src/inlines.c
+++ b/src/inlines.c
@@ -12,6 +12,7 @@
 #include "utf8.h"
 #include "scanners.h"
 #include "inlines.h"
+#include "syntax_extension.h"
 
 static const char *EMDASH = "\xE2\x80\x94";
 static const char *ENDASH = "\xE2\x80\x93";
@@ -45,7 +46,7 @@ static CMARK_INLINE bool S_is_line_end_char(char c) {
 static delimiter *S_insert_emph(subject *subj, delimiter *opener,
                                 delimiter *closer);
 
-static int parse_inline(subject *subj, cmark_node *parent, int options);
+static int parse_inline(cmark_parser *parser, subject *subj, cmark_node *parent, int options);
 
 static void subject_from_buf(cmark_mem *mem, subject *e, cmark_strbuf *buffer,
                              cmark_reference_map *refmap);
@@ -434,7 +435,40 @@ static cmark_node *handle_period(subject *subj, bool smart) {
   }
 }
 
-static void process_emphasis(subject *subj, delimiter *stack_bottom) {
+static void add_extensions_openers_bottom(cmark_parser *parser,
+    delimiter **openers_bottom, delimiter *stack_bottom) {
+  cmark_llist *tmp_ext;
+
+  for (tmp_ext = parser->inline_syntax_extensions; tmp_ext; tmp_ext=tmp_ext->next) {
+    cmark_syntax_extension *ext = (cmark_syntax_extension *) tmp_ext->data;
+    cmark_llist *tmp_char;
+    for (tmp_char = ext->special_inline_chars; tmp_char; tmp_char=tmp_char->next) {
+      unsigned char c = (unsigned char) (unsigned long) tmp_char->data;
+
+      openers_bottom[c] = stack_bottom;
+    }
+  }
+}
+
+static cmark_syntax_extension *get_extension_for_special_char(cmark_parser *parser, unsigned char c) {
+  cmark_llist *tmp_ext;
+
+  for (tmp_ext = parser->inline_syntax_extensions; tmp_ext; tmp_ext=tmp_ext->next) {
+    cmark_syntax_extension *ext = (cmark_syntax_extension *) tmp_ext->data;
+    cmark_llist *tmp_char;
+    for (tmp_char = ext->special_inline_chars; tmp_char; tmp_char=tmp_char->next) {
+      unsigned char tmp_c = (unsigned char) (unsigned long) tmp_char->data;
+
+      if (tmp_c == c) {
+        return ext;
+      }
+    }
+  }
+
+  return NULL;
+}
+
+static void process_emphasis(cmark_parser *parser, subject *subj, delimiter *stack_bottom) {
   delimiter *closer = subj->last_delim;
   delimiter *opener;
   delimiter *old_closer;
@@ -446,6 +480,7 @@ static void process_emphasis(subject *subj, delimiter *stack_bottom) {
   openers_bottom['_'] = stack_bottom;
   openers_bottom['\''] = stack_bottom;
   openers_bottom['"'] = stack_bottom;
+  add_extensions_openers_bottom(parser, openers_bottom, stack_bottom);
 
   // move back to first relevant delim.
   while (closer != NULL && closer->previous != stack_bottom) {
@@ -454,8 +489,10 @@ static void process_emphasis(subject *subj, delimiter *stack_bottom) {
 
   // now move forward, looking for closers, and handling each
   while (closer != NULL) {
+    cmark_syntax_extension *extension = get_extension_for_special_char(parser, closer->delim_char);
     if (closer->can_close &&
-        (closer->delim_char == '*' || closer->delim_char == '_' ||
+        (extension != NULL ||
+         closer->delim_char == '*' || closer->delim_char == '_' ||
          closer->delim_char == '"' || closer->delim_char == '\'')) {
       // Now look backwards for first matching opener:
       opener = closer->previous;
@@ -469,7 +506,13 @@ static void process_emphasis(subject *subj, delimiter *stack_bottom) {
         opener = opener->previous;
       }
       old_closer = closer;
-      if (closer->delim_char == '*' || closer->delim_char == '_') {
+
+      if (extension) {
+        if (opener_found)
+          closer = extension->insert_inline_from_delim(extension, parser, subj, opener, closer);
+        else
+          closer = closer->next;
+      } else if (closer->delim_char == '*' || closer->delim_char == '_') {
         if (opener_found) {
           closer = S_insert_emph(subj, opener, closer);
         } else {
@@ -737,9 +780,7 @@ static int link_label(subject *subj, cmark_chunk *raw_label) {
   subj->pos = startpos; // rewind
   return 0;
 }
-
-// Return a link, an image, or a literal close bracket.
-static cmark_node *handle_close_bracket(subject *subj) {
+static cmark_node *handle_close_bracket(cmark_parser *parser, subject *subj) {
   bufsize_t initial_pos;
   bufsize_t starturl, endurl, starttitle, endtitle, endall;
   bufsize_t n;
@@ -864,7 +905,7 @@ static cmark_node *handle_close_bracket(subject *subj) {
   // Free the bracket [:
   cmark_node_free(opener->inl_text);
 
-  process_emphasis(subj, opener);
+  process_emphasis(parser, subj, opener);
 
   // Now, if we have a link, we also want to deactivate earlier link
   // delimiters. (This code can be removed if we decide to allow links
@@ -908,36 +949,36 @@ static cmark_node *handle_newline(subject *subj) {
   }
 }
 
-static bufsize_t subject_find_special_char(subject *subj, int options) {
-  // "\r\n\\`&_*[]<!"
-  static const int8_t SPECIAL_CHARS[256] = {
-      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
-      0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0,
-      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
-      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 0, 1,
-      1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
-      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
-      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
-      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
-      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
-      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
-      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0};
-
-  // " ' . -
-  static const char SMART_PUNCT_CHARS[] = {
-      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
-      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 1, 1, 0,
-      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
-      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
-      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
-      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
-      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
-      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
-      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
-      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
-      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
-  };
+// "\r\n\\`&_*[]<!"
+static int8_t SPECIAL_CHARS[256] = {
+    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+    0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0,
+    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 0, 1,
+    1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0};
+
+// " ' . -
+static char SMART_PUNCT_CHARS[] = {
+    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 1, 1, 0,
+    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+};
 
+static bufsize_t subject_find_special_char(subject *subj, int options) {
   bufsize_t n = subj->pos + 1;
 
   while (n < subj->input.len) {
@@ -951,9 +992,36 @@ static bufsize_t subject_find_special_char(subject *subj, int options) {
   return subj->input.len;
 }
 
+void cmark_inlines_add_special_character(unsigned char c) {
+  SPECIAL_CHARS[c] = 1;
+}
+
+void cmark_inlines_remove_special_character(unsigned char c) {
+  SPECIAL_CHARS[c] = 0;
+}
+
+static cmark_node *try_extensions(cmark_parser *parser,
+                                  cmark_node *parent,
+                                  unsigned char c,
+                                  subject *subj) {
+  cmark_node *res = NULL;
+  cmark_llist *tmp;
+
+  for (tmp = parser->inline_syntax_extensions; tmp; tmp = tmp->next) {
+    cmark_syntax_extension *ext = (cmark_syntax_extension *) tmp->data;
+
+    res = ext->match_inline(ext, parser, parent, c, subj);
+
+    if (res)
+      break;
+  }
+
+  return res;
+}
+
 // Parse an inline, advancing subject, and add it as a child of parent.
 // Return 0 if no inline can be parsed, 1 otherwise.
-static int parse_inline(subject *subj, cmark_node *parent, int options) {
+static int parse_inline(cmark_parser *parser, subject *subj, cmark_node *parent, int options) {
   cmark_node *new_inl = NULL;
   cmark_chunk contents;
   unsigned char c;
@@ -997,7 +1065,7 @@ static int parse_inline(subject *subj, cmark_node *parent, int options) {
     push_delimiter(subj, '[', true, false, new_inl);
     break;
   case ']':
-    new_inl = handle_close_bracket(subj);
+    new_inl = handle_close_bracket(parser, subj);
     break;
   case '!':
     advance(subj);
@@ -1010,6 +1078,10 @@ static int parse_inline(subject *subj, cmark_node *parent, int options) {
     }
     break;
   default:
+    new_inl = try_extensions(parser, parent, c, subj);
+    if (new_inl != NULL)
+      break;
+
     endpos = subject_find_special_char(subj, options);
     contents = cmark_chunk_dup(&subj->input, subj->pos, endpos - subj->pos);
     subj->pos = endpos;
@@ -1029,16 +1101,18 @@ static int parse_inline(subject *subj, cmark_node *parent, int options) {
 }
 
 // Parse inlines from parent's string_content, adding as children of parent.
-extern void cmark_parse_inlines(cmark_mem *mem, cmark_node *parent, cmark_reference_map *refmap,
+extern void cmark_parse_inlines(cmark_parser *parser,
+                                cmark_node *parent,
+                                cmark_reference_map *refmap,
                                 int options) {
   subject subj;
-  subject_from_buf(mem, &subj, &parent->content, refmap);
+  subject_from_buf(parser->mem, &subj, &parent->content, refmap);
   cmark_chunk_rtrim(&subj.input);
 
-  while (!is_eof(&subj) && parse_inline(&subj, parent, options))
+  while (!is_eof(&subj) && parse_inline(parser, &subj, parent, options))
     ;
 
-  process_emphasis(&subj, NULL);
+  process_emphasis(parser, &subj, NULL);
 }
 
 // Parse zero or more space characters, including at most one newline.
diff --git a/src/inlines.h b/src/inlines.h
index cbe783032..586b53fa7 100644
--- a/src/inlines.h
+++ b/src/inlines.h
@@ -8,12 +8,17 @@ extern "C" {
 cmark_chunk cmark_clean_url(cmark_mem *mem, cmark_chunk *url);
 cmark_chunk cmark_clean_title(cmark_mem *mem, cmark_chunk *title);
 
-void cmark_parse_inlines(cmark_mem *mem, cmark_node *parent, cmark_reference_map *refmap,
+void cmark_parse_inlines(cmark_parser *parser,
+                         cmark_node *parent,
+                         cmark_reference_map *refmap,
                          int options);
 
 bufsize_t cmark_parse_reference_inline(cmark_mem *mem, cmark_strbuf *input,
                                        cmark_reference_map *refmap);
 
+void cmark_inlines_add_special_character(unsigned char c);
+void cmark_inlines_remove_special_character(unsigned char c);
+
 #ifdef __cplusplus
 }
 #endif
diff --git a/src/parser.h b/src/parser.h
index 1129f3bae..247423a76 100644
--- a/src/parser.h
+++ b/src/parser.h
@@ -46,6 +46,7 @@ struct cmark_parser {
   int options;
   bool last_buffer_ended_with_cr;
   cmark_llist *syntax_extensions;
+  cmark_llist *inline_syntax_extensions;
 };
 
 #ifdef __cplusplus
diff --git a/src/syntax_extension.c b/src/syntax_extension.c
index 11a92c020..bad984adb 100644
--- a/src/syntax_extension.c
+++ b/src/syntax_extension.c
@@ -5,6 +5,7 @@
 #include "buffer.h"
 
 void cmark_syntax_extension_free(cmark_syntax_extension *extension) {
+  cmark_llist_free(extension->special_inline_chars);
   free(extension->name);
   free(extension);
 }
@@ -25,3 +26,18 @@ void cmark_syntax_extension_set_match_block_func(cmark_syntax_extension *extensi
                                                  cmark_match_block_func func) {
   extension->last_block_matches = func;
 }
+
+void cmark_syntax_extension_set_match_inline_func(cmark_syntax_extension *extension,
+                                                  cmark_match_inline_func func) {
+  extension->match_inline = func;
+}
+
+void cmark_syntax_extension_set_inline_from_delim_func(cmark_syntax_extension *extension,
+                                                       cmark_inline_from_delim_func func) {
+  extension->insert_inline_from_delim = func;
+}
+
+void cmark_syntax_extension_set_special_inline_chars(cmark_syntax_extension *extension,
+                                                     cmark_llist *special_chars) {
+  extension->special_inline_chars = special_chars;
+}
diff --git a/src/syntax_extension.h b/src/syntax_extension.h
index 5565b4523..f9af1f12c 100644
--- a/src/syntax_extension.h
+++ b/src/syntax_extension.h
@@ -5,9 +5,12 @@
 #include "cmark_extension_api.h"
 
 struct cmark_syntax_extension {
-  cmark_match_block_func   last_block_matches;
-  cmark_open_block_func    try_opening_block;
-  char                   * name;
+  cmark_match_block_func          last_block_matches;
+  cmark_open_block_func           try_opening_block;
+  cmark_match_inline_func         match_inline;
+  cmark_inline_from_delim_func    insert_inline_from_delim;
+  cmark_llist                   * special_inline_chars;
+  char                          * name;
 };
 
 #endif

From 5f64582e14b529d28647474ba616a1d7ac6445bd Mon Sep 17 00:00:00 2001
From: Mathieu Duponchelle <mathieu.duponchelle@opencreed.com>
Date: Sat, 5 Mar 2016 03:44:25 +0100
Subject: [PATCH 16/20] Define a strikethrough inline node type.

---
 src/cmark.h      |  5 ++++-
 src/commonmark.c |  4 ++++
 src/html.c       |  8 ++++++++
 src/latex.c      |  9 +++++++++
 src/man.c        | 23 +++++++++++++++++++++++
 src/node.c       |  6 +++++-
 6 files changed, 53 insertions(+), 2 deletions(-)

diff --git a/src/cmark.h b/src/cmark.h
index 4076176a2..6616ee593 100644
--- a/src/cmark.h
+++ b/src/cmark.h
@@ -66,8 +66,11 @@ typedef enum {
   CMARK_NODE_LINK,
   CMARK_NODE_IMAGE,
 
+  /* inlines with no syntax rules in the current specification */
+  CMARK_NODE_STRIKETHROUGH,
+
   CMARK_NODE_FIRST_INLINE = CMARK_NODE_TEXT,
-  CMARK_NODE_LAST_INLINE = CMARK_NODE_IMAGE,
+  CMARK_NODE_LAST_INLINE = CMARK_NODE_STRIKETHROUGH,
 } cmark_node_type;
 
 /* For backwards compatibility: */
diff --git a/src/commonmark.c b/src/commonmark.c
index 486dc441c..f5d3af1b3 100644
--- a/src/commonmark.c
+++ b/src/commonmark.c
@@ -483,6 +483,10 @@ static int S_render_node(cmark_renderer *renderer, cmark_node *node,
     }
     break;
 
+  case CMARK_NODE_STRIKETHROUGH:
+    OUT(cmark_node_get_string_content(node), false, LITERAL);
+    break;
+
   default:
     assert(false);
     break;
diff --git a/src/html.c b/src/html.c
index f8f51026c..f710c2428 100644
--- a/src/html.c
+++ b/src/html.c
@@ -374,6 +374,14 @@ static int S_render_node(cmark_node *node, cmark_event_type ev_type,
     }
     break;
 
+  case CMARK_NODE_STRIKETHROUGH:
+    if (entering) {
+      cmark_strbuf_puts(html, "<del>");
+    } else {
+      cmark_strbuf_puts(html, "</del>");
+    }
+  break;
+
   default:
     assert(false);
     break;
diff --git a/src/latex.c b/src/latex.c
index 306a508de..f85e46689 100644
--- a/src/latex.c
+++ b/src/latex.c
@@ -463,6 +463,15 @@ static int S_render_node(cmark_renderer *renderer, cmark_node *node,
     }
     break;
 
+  case CMARK_NODE_STRIKETHROUGH:
+    /* requires \usepackage{ulem} */
+    if (entering) {
+      LIT("\\sout{");
+    } else {
+      LIT("}");
+    }
+    break;
+
   default:
     assert(false);
     break;
diff --git a/src/man.c b/src/man.c
index a2691b063..7242927bd 100644
--- a/src/man.c
+++ b/src/man.c
@@ -82,6 +82,19 @@ static int S_render_node(cmark_renderer *renderer, cmark_node *node,
 
   switch (node->type) {
   case CMARK_NODE_DOCUMENT:
+    if (entering) {
+      /* Define a strikethrough macro */
+      /* Commenting out because this makes tests fail
+      LIT(".de ST");
+      CR();
+      LIT(".nr ww \\w'\\\\$1'");
+      CR();
+      LIT("\\Z@\\v'-.25m'\\l'\\\\n[ww]u'@\\\\$1");
+      CR();
+      LIT("..");
+      CR();
+      */
+    }
     break;
 
   case CMARK_NODE_BLOCK_QUOTE:
@@ -273,6 +286,16 @@ static int S_render_node(cmark_renderer *renderer, cmark_node *node,
     }
     break;
 
+  case CMARK_NODE_STRIKETHROUGH:
+    if (entering) {
+      CR();
+      LIT(".ST \"");
+    } else {
+      LIT("\"");
+      CR();
+    }
+    break;
+
   default:
     assert(false);
     break;
diff --git a/src/node.c b/src/node.c
index 7743ae19b..0e03210d0 100644
--- a/src/node.c
+++ b/src/node.c
@@ -65,6 +65,7 @@ static bool S_can_contain(cmark_node *node, cmark_node *child) {
   case CMARK_NODE_STRONG:
   case CMARK_NODE_LINK:
   case CMARK_NODE_IMAGE:
+  case CMARK_NODE_STRIKETHROUGH:
   case CMARK_NODE_CUSTOM_INLINE:
     return S_is_inline(child);
   case CMARK_NODE_TABLE:
@@ -77,7 +78,8 @@ static bool S_can_contain(cmark_node *node, cmark_node *child) {
            child->type == CMARK_NODE_EMPH ||
            child->type == CMARK_NODE_STRONG ||
            child->type == CMARK_NODE_LINK ||
-           child->type == CMARK_NODE_IMAGE;
+           child->type == CMARK_NODE_IMAGE ||
+           child->type == CMARK_NODE_STRIKETHROUGH;
 
   default:
     break;
@@ -259,6 +261,8 @@ const char *cmark_node_get_type_string(cmark_node *node) {
     return "link";
   case CMARK_NODE_IMAGE:
     return "image";
+  case CMARK_NODE_STRIKETHROUGH:
+    return "strikethrough";
   }
 
   return "<unknown>";

From 7fbb2e65fdd50e8c98107fa59b50822eda448e37 Mon Sep 17 00:00:00 2001
From: Mathieu Duponchelle <mathieu.duponchelle@opencreed.com>
Date: Fri, 4 Mar 2016 20:54:03 +0100
Subject: [PATCH 17/20] Example inline extension: strikethrough text

---
 extensions/core-extensions.c | 87 ++++++++++++++++++++++++++++++++++++
 1 file changed, 87 insertions(+)

diff --git a/extensions/core-extensions.c b/extensions/core-extensions.c
index 9b904ed86..ad9715636 100644
--- a/extensions/core-extensions.c
+++ b/extensions/core-extensions.c
@@ -232,7 +232,94 @@ static cmark_syntax_extension *register_table_syntax_extension(void) {
   return ext;
 }
 
+static cmark_node *strikethrough_match(cmark_syntax_extension *self,
+                                       cmark_parser *parser,
+                                       cmark_node *parent,
+                                       unsigned char character,
+                                       cmark_inline_parser *inline_parser)
+{
+  cmark_node *res = NULL;
+  int left_flanking, right_flanking, punct_before, punct_after;
+  int num_delims;
+
+  /* Exit early */
+  if (character != '~')
+    return NULL;
+
+  num_delims = cmark_inline_parser_scan_delimiters(inline_parser, 1, '~',
+      &left_flanking, &right_flanking, &punct_before, &punct_after);
+
+  if (num_delims > 0) { /* Should not be needed */
+    int can_open, can_close;
+
+    res = cmark_node_new_with_mem(CMARK_NODE_TEXT, parser->mem);
+    cmark_node_set_literal(res, "~");
+
+    can_open = left_flanking;
+    can_close = right_flanking;
+    if (can_open || can_close)
+      cmark_inline_parser_push_delimiter(inline_parser, character, can_open, can_close, res);
+  }
+
+  return res;
+}
+
+static delimiter *strikethrough_insert(cmark_syntax_extension *self,
+                                       cmark_parser *parser,
+                                       cmark_inline_parser *inline_parser,
+                                       delimiter *opener,
+                                       delimiter *closer)
+{
+  cmark_node *strikethrough;
+  cmark_node *tmp, *next;
+  delimiter *delim, *tmp_delim;
+  delimiter *res = closer->next;
+
+  strikethrough = opener->inl_text;
+
+  if (!cmark_node_set_type(strikethrough, CMARK_NODE_STRIKETHROUGH))
+    goto done;
+
+  cmark_node_set_string_content(strikethrough, "~");
+  tmp = cmark_node_next(opener->inl_text);
+
+  while (tmp) {
+    if (tmp == closer->inl_text)
+      break;
+    next = cmark_node_next(tmp);
+    cmark_node_append_child(strikethrough, tmp);
+    tmp = next;
+  }
+
+  cmark_node_free(closer->inl_text);
+
+  delim = closer;
+  while (delim != NULL && delim != opener) {
+    tmp_delim = delim->previous;
+    cmark_inline_parser_remove_delimiter(inline_parser, delim);
+    delim = tmp_delim;
+  }
+
+  cmark_inline_parser_remove_delimiter(inline_parser, opener);
+
+done:
+  return res;
+}
+
+static cmark_syntax_extension *create_strikethrough_extension(void) {
+  cmark_syntax_extension *ext = cmark_syntax_extension_new("tilde_strikethrough");
+  cmark_llist *special_chars = NULL;
+
+  cmark_syntax_extension_set_match_inline_func(ext, strikethrough_match);
+  cmark_syntax_extension_set_inline_from_delim_func(ext, strikethrough_insert);
+  special_chars = cmark_llist_append(special_chars, (void *) '~');
+  cmark_syntax_extension_set_special_inline_chars(ext, special_chars);
+
+  return ext;
+}
+
 int init_libcmarkextensions(cmark_plugin *plugin) {
   cmark_plugin_register_syntax_extension(plugin, register_table_syntax_extension());
+  cmark_plugin_register_syntax_extension(plugin, create_strikethrough_extension());
   return 1;
 }

From 44b2d76a4762b02ffbd6b18be95004beae394ed7 Mon Sep 17 00:00:00 2001
From: Mathieu Duponchelle <mathieu.duponchelle@opencreed.com>
Date: Sun, 6 Mar 2016 10:17:19 +0100
Subject: [PATCH 18/20] cmark_parser: improve lifecycle management.

This slightly breaks API as finish will now return NULL
after it has been called once, but I believe that's for the
best.

This fixes potential leaks of root when a parser was fed but
not finished, and makes reusing a parser instance multiple
times possible.
---
 src/blocks.c | 69 ++++++++++++++++++++++++++++++++++++----------------
 1 file changed, 48 insertions(+), 21 deletions(-)

diff --git a/src/blocks.c b/src/blocks.c
index 562658d69..14cee3ba9 100644
--- a/src/blocks.c
+++ b/src/blocks.c
@@ -92,32 +92,46 @@ int cmark_parser_attach_syntax_extension(cmark_parser *parser,
   return 1;
 }
 
-cmark_parser *cmark_parser_new_with_mem(int options, cmark_mem *mem) {
-  cmark_parser *parser = (cmark_parser *)mem->calloc(1, sizeof(cmark_parser));
-  parser->mem = mem;
+static void cmark_parser_dispose(cmark_parser *parser) {
+  if (parser->root)
+    cmark_node_free(parser->root);
+
+  if (parser->refmap)
+    cmark_reference_map_free(parser->refmap);
+}
+
+static void cmark_parser_reset(cmark_parser *parser) {
+  cmark_llist *saved_exts = parser->syntax_extensions;
+  cmark_llist *saved_inline_exts = parser->inline_syntax_extensions;
+  int saved_options = parser->options;
+  cmark_mem *saved_mem = parser->mem;
+
+  cmark_parser_dispose(parser);
 
-  cmark_node *document = make_document(mem);
+  memset(parser, 0, sizeof(cmark_parser));
+  parser->mem = saved_mem;
 
-  cmark_strbuf_init(mem, &parser->curline, 256);
-  cmark_strbuf_init(mem, &parser->linebuf, 0);
+  cmark_strbuf_init(parser->mem, &parser->curline, 256);
+  cmark_strbuf_init(parser->mem, &parser->linebuf, 0);
 
-  parser->refmap = cmark_reference_map_new(mem);
+  cmark_node *document = make_document(parser->mem);
+
+  parser->refmap = cmark_reference_map_new(parser->mem);
   parser->root = document;
   parser->current = document;
-  parser->line_number = 0;
-  parser->offset = 0;
-  parser->column = 0;
-  parser->first_nonspace = 0;
-  parser->first_nonspace_column = 0;
-  parser->indent = 0;
-  parser->blank = false;
-  parser->partially_consumed_tab = false;
-  parser->last_line_length = 0;
-  parser->options = options;
+
   parser->last_buffer_ended_with_cr = false;
-  parser->syntax_extensions = NULL;
-  parser->inline_syntax_extensions = NULL;
 
+  parser->syntax_extensions = saved_exts;
+  parser->inline_syntax_extensions = saved_inline_exts;
+  parser->options = saved_options;
+}
+
+cmark_parser *cmark_parser_new_with_mem(int options, cmark_mem *mem) {
+  cmark_parser *parser = (cmark_parser *)mem->calloc(1, sizeof(cmark_parser));
+  parser->mem = mem;
+  parser->options = options;
+  cmark_parser_reset(parser);
   return parser;
 }
 
@@ -128,9 +142,9 @@ cmark_parser *cmark_parser_new(int options) {
 
 void cmark_parser_free(cmark_parser *parser) {
   cmark_mem *mem = parser->mem;
+  cmark_parser_dispose(parser);
   cmark_strbuf_free(&parser->curline);
   cmark_strbuf_free(&parser->linebuf);
-  cmark_reference_map_free(parser->refmap);
   cmark_llist_free(parser->syntax_extensions);
   cmark_llist_free(parser->inline_syntax_extensions);
   mem->free(parser);
@@ -1248,6 +1262,12 @@ static void S_process_line(cmark_parser *parser, const unsigned char *buffer,
 }
 
 cmark_node *cmark_parser_finish(cmark_parser *parser) {
+  cmark_node *res;
+
+  /* Parser was already finished once */
+  if (parser->root == NULL)
+    return NULL;
+
   if (parser->linebuf.size) {
     S_process_line(parser, parser->linebuf.ptr, parser->linebuf.size);
     cmark_strbuf_clear(&parser->linebuf);
@@ -1260,13 +1280,20 @@ cmark_node *cmark_parser_finish(cmark_parser *parser) {
   }
 
   cmark_strbuf_free(&parser->curline);
+  cmark_strbuf_free(&parser->linebuf);
 
 #if CMARK_DEBUG_NODES
   if (cmark_node_check(parser->root, stderr)) {
     abort();
   }
 #endif
-  return parser->root;
+
+  res = parser->root;
+  parser->root = NULL;
+
+  cmark_parser_reset(parser);
+
+  return res;
 }
 
 int cmark_parser_get_line_number(cmark_parser *parser) {

From cf19502bdae4967f5117c9d90860def8896caeb8 Mon Sep 17 00:00:00 2001
From: Mathieu Duponchelle <mathieu.duponchelle@opencreed.com>
Date: Sun, 3 Apr 2016 00:59:11 +0200
Subject: [PATCH 19/20] syntax-extension: add "priv" field.

+ And pointer to a free function as well.
---
 src/cmark_extension_api.h | 11 +++++++++++
 src/syntax_extension.c    | 11 +++++++++++
 src/syntax_extension.h    |  2 ++
 3 files changed, 24 insertions(+)

diff --git a/src/cmark_extension_api.h b/src/cmark_extension_api.h
index 636df7a12..db6f4cd01 100644
--- a/src/cmark_extension_api.h
+++ b/src/cmark_extension_api.h
@@ -97,6 +97,10 @@ typedef struct cmark_plugin cmark_plugin;
  *
  * Finally, the extension should return NULL if its scan didn't
  * match its syntax rules.
+ *
+ * The extension can store whatever private data it might need
+ * with 'cmark_syntax_extension_set_private',
+ * and optionally define a free function for this data.
  */
 typedef struct cmark_syntax_extension cmark_syntax_extension;
 
@@ -251,6 +255,13 @@ CMARK_EXPORT
 void cmark_syntax_extension_set_special_inline_chars(cmark_syntax_extension *extension,
                                                      cmark_llist *special_chars);
 
+/** See the documentation for 'cmark_syntax_extension'
+ */
+CMARK_EXPORT
+void cmark_syntax_extension_set_private(cmark_syntax_extension *extension,
+                                        void *priv,
+                                        cmark_free_func free_func);
+
 /** Return the index of the line currently being parsed, starting with 1.
  */
 CMARK_EXPORT
diff --git a/src/syntax_extension.c b/src/syntax_extension.c
index bad984adb..d8c4459d9 100644
--- a/src/syntax_extension.c
+++ b/src/syntax_extension.c
@@ -5,6 +5,10 @@
 #include "buffer.h"
 
 void cmark_syntax_extension_free(cmark_syntax_extension *extension) {
+  if (extension->free_function && extension->priv) {
+    extension->free_function(extension->priv);
+  }
+
   cmark_llist_free(extension->special_inline_chars);
   free(extension->name);
   free(extension);
@@ -41,3 +45,10 @@ void cmark_syntax_extension_set_special_inline_chars(cmark_syntax_extension *ext
                                                      cmark_llist *special_chars) {
   extension->special_inline_chars = special_chars;
 }
+
+void cmark_syntax_extension_set_private(cmark_syntax_extension *extension,
+                                        void *priv,
+                                        cmark_free_func free_func) {
+  extension->priv = priv;
+  extension->free_function = free_func;
+}
diff --git a/src/syntax_extension.h b/src/syntax_extension.h
index f9af1f12c..f46a7d2fe 100644
--- a/src/syntax_extension.h
+++ b/src/syntax_extension.h
@@ -11,6 +11,8 @@ struct cmark_syntax_extension {
   cmark_inline_from_delim_func    insert_inline_from_delim;
   cmark_llist                   * special_inline_chars;
   char                          * name;
+  void                          * priv;
+  cmark_free_func                 free_function;
 };
 
 #endif

From 389216b027fa2c27834a119a863ff9e1a99d77a6 Mon Sep 17 00:00:00 2001
From: Mathieu Duponchelle <mathieu.duponchelle@opencreed.com>
Date: Sun, 3 Apr 2016 00:59:45 +0200
Subject: [PATCH 20/20] cmark_parser: implement and expose reentrant feed
 function.

This can be useful for transclusion extensions for example,
http://talk.commonmark.org/t/transclusion-or-including-sub-documents-for-reuse/270
---
 src/blocks.c              | 22 +++++++++++++++++++++-
 src/cmark_extension_api.h |  4 ++++
 2 files changed, 25 insertions(+), 1 deletion(-)

diff --git a/src/blocks.c b/src/blocks.c
index 14cee3ba9..637e1536c 100644
--- a/src/blocks.c
+++ b/src/blocks.c
@@ -567,6 +567,19 @@ void cmark_parser_feed(cmark_parser *parser, const char *buffer, size_t len) {
   S_parser_feed(parser, (const unsigned char *)buffer, len, false);
 }
 
+void cmark_parser_feed_reentrant(cmark_parser *parser, const char *buffer, size_t len) {
+  cmark_strbuf saved_linebuf;
+
+  cmark_strbuf_init(parser->mem, &saved_linebuf, 0);
+  cmark_strbuf_puts(&saved_linebuf, cmark_strbuf_cstr(&parser->linebuf));
+  cmark_strbuf_clear(&parser->linebuf);
+
+  S_parser_feed(parser, (const unsigned char *)buffer, len, true);
+
+  cmark_strbuf_sets(&parser->linebuf, cmark_strbuf_cstr(&saved_linebuf));
+  cmark_strbuf_free(&saved_linebuf);
+}
+
 static void S_parser_feed(cmark_parser *parser, const unsigned char *buffer,
                           size_t len, bool eof) {
   const unsigned char *end = buffer + len;
@@ -1214,6 +1227,9 @@ static void S_process_line(cmark_parser *parser, const unsigned char *buffer,
   bool all_matched = true;
   cmark_node *container;
   cmark_chunk input;
+  cmark_node *current;
+
+  cmark_strbuf_clear(&parser->curline);
 
   if (parser->options & CMARK_OPT_VALIDATE_UTF8)
     cmark_utf8proc_check(&parser->curline, buffer, bytes);
@@ -1245,9 +1261,13 @@ static void S_process_line(cmark_parser *parser, const unsigned char *buffer,
   if (parser->blank && S_last_line_blank(container))
     break_out_of_lists(parser, &container);
 
+  current = parser->current;
+
   open_new_blocks(parser, &container, &input, all_matched);
 
-  add_text_to_container(parser, container, last_matched_container, &input);
+  /* parser->current might have changed if feed_reentrant was called */
+  if (current == parser->current)
+    add_text_to_container(parser, container, last_matched_container, &input);
 
 finished:
   parser->last_line_length = input.len;
diff --git a/src/cmark_extension_api.h b/src/cmark_extension_api.h
index db6f4cd01..18e08142c 100644
--- a/src/cmark_extension_api.h
+++ b/src/cmark_extension_api.h
@@ -416,6 +416,10 @@ void cmark_parser_advance_offset(cmark_parser *parser,
                                  int count,
                                  int columns);
 
+
+CMARK_EXPORT
+void cmark_parser_feed_reentrant(cmark_parser *parser, const char *buffer, size_t len);
+
 /** Attach the syntax 'extension' to the 'parser', to provide extra syntax
  *  rules.
  *  See the documentation for cmark_syntax_extension for more information.