src/inlines.c - master - clang/cmark

src/inlines.c
1 509 строк45 KB
Vijay Misal
Fix reference link title incorrectly kept when followed by garbage
12 авг 2026, 18:48
12 авг 2026, 18:487042d99
#include <stdbool.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>

#include "cmark_ctype.h"
#include "node.h"
#include "parser.h"
#include "references.h"
#include "cmark.h"
#include "houdini.h"
#include "utf8.h"
#include "scanners.h"
#include "inlines.h"

static const char *EMDASH = "\xE2\x80\x94";
static const char *ENDASH = "\xE2\x80\x93";
static const char *ELLIPSES = "\xE2\x80\xA6";
static const char *LEFTDOUBLEQUOTE = "\xE2\x80\x9C";
static const char *RIGHTDOUBLEQUOTE = "\xE2\x80\x9D";
static const char *LEFTSINGLEQUOTE = "\xE2\x80\x98";
static const char *RIGHTSINGLEQUOTE = "\xE2\x80\x99";

// Macros for creating various kinds of simple.
#define make_linebreak(mem) make_simple(mem, CMARK_NODE_LINEBREAK)
#define make_softbreak(mem) make_simple(mem, CMARK_NODE_SOFTBREAK)
#define make_emph(mem) make_simple(mem, CMARK_NODE_EMPH)
#define make_strong(mem) make_simple(mem, CMARK_NODE_STRONG)

#define MAXBACKTICKS 1000

typedef struct delimiter {
  struct delimiter *previous;
  struct delimiter *next;
  cmark_node *inl_text;
  bufsize_t position;
  bufsize_t length;
  unsigned char delim_char;
  bool can_open;
  bool can_close;
} delimiter;

typedef struct bracket {
  struct bracket *previous;
  cmark_node *inl_text;
  bufsize_t position;
  bool image;
  bool active;
  bool bracket_after;
} bracket;

#define FLAG_SKIP_HTML_CDATA        (1u << 0)
#define FLAG_SKIP_HTML_DECLARATION  (1u << 1)
#define FLAG_SKIP_HTML_PI           (1u << 2)
#define FLAG_SKIP_HTML_COMMENT      (1u << 3)

typedef struct {
  cmark_mem *mem;
  cmark_chunk input;
  unsigned flags;
  int line;
  bufsize_t pos;
  int block_offset;
  int column_offset;
  cmark_reference_map *refmap;
  delimiter *last_delim;
  bracket *last_bracket;
  bufsize_t backticks[MAXBACKTICKS + 1];
  bool scanned_for_backticks;
  bool no_link_openers;
} subject;

static inline bool S_is_line_end_char(char c) {
  return (c == '\n' || c == '\r');
}

static delimiter *S_insert_emph(subject *subj, delimiter *opener,
                                delimiter *closer);

static int parse_inline(subject *subj, cmark_node *parent, int options);

static void subject_from_buf(cmark_mem *mem, int line_number, int block_offset, subject *e,
                             cmark_chunk *chunk, cmark_reference_map *refmap);
static bufsize_t subject_find_special_char(subject *subj, int options);

// Create an inline with a literal string value.
static inline cmark_node *make_literal(subject *subj, cmark_node_type t,
                                       int start_column, int end_column) {
  cmark_node *e = (cmark_node *)subj->mem->calloc(1, sizeof(*e));
  e->mem = subj->mem;
  e->type = (uint16_t)t;
  e->start_line = e->end_line = subj->line;
  // columns are 1 based.
  e->start_column = start_column + 1 + subj->column_offset + subj->block_offset;
  e->end_column = end_column + 1 + subj->column_offset + subj->block_offset;
  return e;
}

// Create an inline with no value.
static inline cmark_node *make_simple(cmark_mem *mem, cmark_node_type t) {
  cmark_node *e = (cmark_node *)mem->calloc(1, sizeof(*e));
  e->mem = mem;
  e->type = t;
  return e;
}

static cmark_node *make_str(subject *subj, int sc, int ec, cmark_chunk s) {
  cmark_node *e = make_literal(subj, CMARK_NODE_TEXT, sc, ec);
  e->data = (unsigned char *)subj->mem->realloc(NULL, s.len + 1);
  if (s.data != NULL) {
    memcpy(e->data, s.data, s.len);
  }
  e->data[s.len] = 0;
  e->len = s.len;
  return e;
}

static cmark_node *make_str_from_buf(subject *subj, int sc, int ec,
                                     cmark_strbuf *buf) {
  cmark_node *e = make_literal(subj, CMARK_NODE_TEXT, sc, ec);
  e->len = buf->size;
  e->data = cmark_strbuf_detach(buf);
  return e;
}

// Like make_str, but parses entities.
static cmark_node *make_str_with_entities(subject *subj,
                                          int start_column, int end_column,
                                          cmark_chunk *content) {
  cmark_strbuf unescaped = CMARK_BUF_INIT(subj->mem);

  if (houdini_unescape_html(&unescaped, content->data, content->len)) {
    return make_str_from_buf(subj, start_column, end_column, &unescaped);
  } else {
    return make_str(subj, start_column, end_column, *content);
  }
}

// Like cmark_node_append_child but without costly sanity checks.
// Assumes that child was newly created.
static void append_child(cmark_node *node, cmark_node *child) {
  cmark_node *old_last_child = node->last_child;

  child->next = NULL;
  child->prev = old_last_child;
  child->parent = node;
  node->last_child = child;

  if (old_last_child) {
    old_last_child->next = child;
  } else {
    // Also set first_child if node previously had no children.
    node->first_child = child;
  }
}

// Duplicate a chunk by creating a copy of the buffer not by reusing the
// buffer like cmark_chunk_dup does.
static unsigned char *cmark_strdup(cmark_mem *mem, unsigned char *src) {
  if (src == NULL) {
    return NULL;
  }
  size_t len = strlen((char *)src);
  unsigned char *data = (unsigned char *)mem->realloc(NULL, len + 1);
  memcpy(data, src, len + 1);
  return data;
}

static unsigned char *cmark_clean_autolink(cmark_mem *mem, cmark_chunk *url,
                                           int is_email) {
  cmark_strbuf buf = CMARK_BUF_INIT(mem);

  cmark_chunk_trim(url);

  if (is_email)
    cmark_strbuf_puts(&buf, "mailto:");

  houdini_unescape_html_f(&buf, url->data, url->len);
  return cmark_strbuf_detach(&buf);
}

static inline cmark_node *make_autolink(subject *subj, int start_column,
                                        int end_column, cmark_chunk url,
                                        int is_email) {
  cmark_node *link = make_simple(subj->mem, CMARK_NODE_LINK);
  link->as.link.url = cmark_clean_autolink(subj->mem, &url, is_email);
  link->as.link.title = NULL;
  link->start_line = link->end_line = subj->line;
  link->start_column = start_column + 1;
  link->end_column = end_column + 1;
  append_child(link, make_str_with_entities(subj, start_column + 1, end_column - 1, &url));
  return link;
}

static void subject_from_buf(cmark_mem *mem, int line_number, int block_offset, subject *e,
                             cmark_chunk *chunk, cmark_reference_map *refmap) {
  int i;
  e->mem = mem;
  e->input = *chunk;
  e->flags = 0;
  e->line = line_number;
  e->pos = 0;
  e->block_offset = block_offset;
  e->column_offset = 0;
  e->refmap = refmap;
  e->last_delim = NULL;
  e->last_bracket = NULL;
  for (i = 0; i <= MAXBACKTICKS; i++) {
    e->backticks[i] = 0;
  }
  e->scanned_for_backticks = false;
  e->no_link_openers = true;
}

static inline int isbacktick(int c) { return (c == '`'); }

static inline unsigned char peek_char(subject *subj) {
  // NULL bytes should have been stripped out by now.  If they're
  // present, it's a programming error:
  assert(!(subj->pos < subj->input.len && subj->input.data[subj->pos] == 0));
  return (subj->pos < subj->input.len) ? subj->input.data[subj->pos] : 0;
}

static inline unsigned char peek_at(subject *subj, bufsize_t pos) {
  return subj->input.data[pos];
}

// Return true if there are more characters in the subject.
static inline int is_eof(subject *subj) {
  return (subj->pos >= subj->input.len);
}

// Advance the subject.  Doesn't check for eof.
#define advance(subj) (subj)->pos += 1

static inline bool skip_spaces(subject *subj) {
  bool skipped = false;
  while (peek_char(subj) == ' ' || peek_char(subj) == '\t') {
    advance(subj);
    skipped = true;
  }
  return skipped;
}

static inline bool skip_line_end(subject *subj) {
  bool seen_line_end_char = false;
  if (peek_char(subj) == '\r') {
    advance(subj);
    seen_line_end_char = true;
  }
  if (peek_char(subj) == '\n') {
    advance(subj);
    seen_line_end_char = true;
  }
  return seen_line_end_char || is_eof(subj);
}

// Take characters while a predicate holds, and return a string.
static inline cmark_chunk take_while(subject *subj, int (*f)(int)) {
  unsigned char c;
  bufsize_t startpos = subj->pos;
  bufsize_t len = 0;

  while ((c = peek_char(subj)) && (*f)(c)) {
    advance(subj);
    len++;
  }

  return cmark_chunk_dup(&subj->input, startpos, len);
}

// Return the number of newlines in a given span of text in a subject.  If
// the number is greater than zero, also return the number of characters
// between the last newline and the end of the span in `since_newline`.
static int count_newlines(subject *subj, bufsize_t from, bufsize_t len, int *since_newline) {
  int nls = 0;
  int since_nl = 0;

  while (len--) {
    if (subj->input.data[from++] == '\n') {
      ++nls;
      since_nl = 0;
    } else {
      ++since_nl;
    }
  }

  if (!nls)
    return 0;

  *since_newline = since_nl;
  return nls;
}

// Adjust `node`'s `end_line`, `end_column`, and `subj`'s `line` and
// `column_offset` according to the number of newlines in a just-matched span
// of text in `subj`.
static void adjust_subj_node_newlines(subject *subj, cmark_node *node, int matchlen, int extra, int options) {
  if (!(options & CMARK_OPT_SOURCEPOS)) {
    return;
  }

  int since_newline;
  int newlines = count_newlines(subj, subj->pos - matchlen - extra, matchlen, &since_newline);
  if (newlines) {
    subj->line += newlines;
    node->end_line += newlines;
    node->end_column = since_newline;
    subj->column_offset = -subj->pos + since_newline + extra;
  }
}

// Try to process a backtick code span that began with a
// span of ticks of length openticklength length (already
// parsed).  Return 0 if you don't find matching closing
// backticks, otherwise return the position in the subject
// after the closing backticks.
static bufsize_t scan_to_closing_backticks(subject *subj,
                                           bufsize_t openticklength) {

  bool found = false;
  if (openticklength > MAXBACKTICKS) {
    // we limit backtick string length because of the array subj->backticks:
    return 0;
  }
  if (subj->scanned_for_backticks &&
      subj->backticks[openticklength] <= subj->pos) {
    // return if we already know there's no closer
    return 0;
  }
  while (!found) {
    // read non backticks
    unsigned char c;
    while ((c = peek_char(subj)) && c != '`') {
      advance(subj);
    }
    if (is_eof(subj)) {
      break;
    }
    bufsize_t numticks = 0;
    while (peek_char(subj) == '`') {
      advance(subj);
      numticks++;
    }
    // store position of ender
    if (numticks <= MAXBACKTICKS) {
      subj->backticks[numticks] = subj->pos - numticks;
    }
    if (numticks == openticklength) {
      return (subj->pos);
    }
  }
  // got through whole input without finding closer
  subj->scanned_for_backticks = true;
  return 0;
}

// Destructively modify string, converting newlines to
// spaces, then removing a single leading + trailing space,
// unless the code span consists entirely of space characters.
static void S_normalize_code(cmark_strbuf *s) {
  bufsize_t r, w;
  bool contains_nonspace = false;

  for (r = 0, w = 0; r < s->size; ++r) {
    switch (s->ptr[r]) {
    case '\r':
      if (s->ptr[r + 1] != '\n') {
        s->ptr[w++] = ' ';
      }
      break;
    case '\n':
      s->ptr[w++] = ' ';
      break;
    default:
      s->ptr[w++] = s->ptr[r];
    }
    if (s->ptr[r] != ' ') {
      contains_nonspace = true;
    }
  }

  // begins and ends with space?
  if (contains_nonspace &&
      s->ptr[0] == ' ' && s->ptr[w - 1] == ' ') {
    cmark_strbuf_drop(s, 1);
    cmark_strbuf_truncate(s, w - 2);
  } else {
    cmark_strbuf_truncate(s, w);
  }

}


// Parse backtick code section or raw backticks, return an inline.
// Assumes that the subject has a backtick at the current position.
static cmark_node *handle_backticks(subject *subj, int options) {
  bufsize_t initpos = subj->pos;
  cmark_chunk openticks = take_while(subj, isbacktick);
  bufsize_t startpos = subj->pos;
  bufsize_t endpos = scan_to_closing_backticks(subj, openticks.len);

  if (endpos == 0) {      // not found
    subj->pos = startpos; // rewind
    return make_str(subj, initpos, initpos + openticks.len - 1, openticks);
  } else {
    cmark_strbuf buf = CMARK_BUF_INIT(subj->mem);

    cmark_strbuf_set(&buf, subj->input.data + startpos,
                     endpos - startpos - openticks.len);
    S_normalize_code(&buf);

    cmark_node *node = make_literal(subj, CMARK_NODE_CODE, startpos,
                                    endpos - openticks.len - 1);
    node->len = buf.size;
    node->data = cmark_strbuf_detach(&buf);
    adjust_subj_node_newlines(subj, node, endpos - startpos, openticks.len, options);
    return node;
  }
}


// Scan ***, **, or * and return number scanned, or 0.
// Advances position.
static int scan_delims(subject *subj, unsigned char c, bool *can_open,
                       bool *can_close) {
  int numdelims = 0;
  bufsize_t before_char_pos;
  int32_t after_char = 0;
  int32_t before_char = 0;
  int len;
  bool left_flanking, right_flanking;

  if (subj->pos == 0) {
    before_char = 10;
  } else {
    before_char_pos = subj->pos - 1;
    // walk back to the beginning of the UTF_8 sequence:
    while (peek_at(subj, before_char_pos) >> 6 == 2 && before_char_pos > 0) {
      before_char_pos -= 1;
    }
    len = cmark_utf8proc_iterate(subj->input.data + before_char_pos,
                                 subj->pos - before_char_pos, &before_char);
    if (len == -1) {
      before_char = 10;
    }
  }

  if (c == '\'' || c == '"') {
    numdelims++;
    advance(subj); // limit to 1 delim for quotes
  } else {
    while (peek_char(subj) == c) {
      numdelims++;
      advance(subj);
    }
  }

  len = cmark_utf8proc_iterate(subj->input.data + subj->pos,
                               subj->input.len - subj->pos, &after_char);
  if (len == -1) {
    after_char = 10;
  }
  left_flanking = numdelims > 0 && !cmark_utf8proc_is_space(after_char) &&
                  (!cmark_utf8proc_is_punctuation_or_symbol(after_char) ||
                   cmark_utf8proc_is_space(before_char) ||
                   cmark_utf8proc_is_punctuation_or_symbol(before_char));
  right_flanking = numdelims > 0 && !cmark_utf8proc_is_space(before_char) &&
                   (!cmark_utf8proc_is_punctuation_or_symbol(before_char) ||
                    cmark_utf8proc_is_space(after_char) ||
                    cmark_utf8proc_is_punctuation_or_symbol(after_char));
  if (c == '_') {
    *can_open = left_flanking &&
                (!right_flanking ||
                 cmark_utf8proc_is_punctuation_or_symbol(before_char));
    *can_close = right_flanking &&
                 (!left_flanking ||
                  cmark_utf8proc_is_punctuation_or_symbol(after_char));
  } else if (c == '\'' || c == '"') {
    *can_open = left_flanking &&
         (!right_flanking || before_char == '(' || before_char == '[') &&
         before_char != ']' && before_char != ')';
    *can_close = right_flanking;
  } else {
    *can_open = left_flanking;
    *can_close = right_flanking;
  }
  return numdelims;
}

/*
static void print_delimiters(subject *subj)
{
        delimiter *delim;
        delim = subj->last_delim;
        while (delim != NULL) {
                printf("Item at stack pos %p: %d %d %d next(%p) prev(%p)\n",
                       (void*)delim, delim->delim_char,
                       delim->can_open, delim->can_close,
                       (void*)delim->next, (void*)delim->previous);
                delim = delim->previous;
        }
}
*/

static void remove_delimiter(subject *subj, delimiter *delim) {
  if (delim == NULL)
    return;
  if (delim->next == NULL) {
    // end of list:
    assert(delim == subj->last_delim);
    subj->last_delim = delim->previous;
  } else {
    delim->next->previous = delim->previous;
  }
  if (delim->previous != NULL) {
    delim->previous->next = delim->next;
  }
  subj->mem->free(delim);
}

static void pop_bracket(subject *subj) {
  bracket *b;
  if (subj->last_bracket == NULL)
    return;
  b = subj->last_bracket;
  subj->last_bracket = subj->last_bracket->previous;
  subj->mem->free(b);
}

static void push_delimiter(subject *subj, unsigned char c, bool can_open,
                           bool can_close, cmark_node *inl_text) {
  delimiter *delim = (delimiter *)subj->mem->calloc(1, sizeof(delimiter));
  delim->delim_char = c;
  delim->can_open = can_open;
  delim->can_close = can_close;
  delim->inl_text = inl_text;
  delim->position = subj->pos;
  delim->length = inl_text->len;
  delim->previous = subj->last_delim;
  delim->next = NULL;
  if (delim->previous != NULL) {
    delim->previous->next = delim;
  }
  subj->last_delim = delim;
}

static void push_bracket(subject *subj, bool image, cmark_node *inl_text) {
  bracket *b = (bracket *)subj->mem->calloc(1, sizeof(bracket));
  if (subj->last_bracket != NULL) {
    subj->last_bracket->bracket_after = true;
  }
  b->image = image;
  b->active = true;
  b->inl_text = inl_text;
  b->previous = subj->last_bracket;
  b->position = subj->pos;
  b->bracket_after = false;
  subj->last_bracket = b;
  if (!image) {
    subj->no_link_openers = false;
  }
}

// Assumes the subject has a c at the current position.
static cmark_node *handle_delim(subject *subj, unsigned char c, bool smart) {
  bufsize_t numdelims;
  cmark_node *inl_text;
  bool can_open, can_close;
  cmark_chunk contents;

  numdelims = scan_delims(subj, c, &can_open, &can_close);

  if (c == '\'' && smart) {
    contents = cmark_chunk_literal(RIGHTSINGLEQUOTE);
  } else if (c == '"' && smart) {
    contents =
        cmark_chunk_literal(can_close ? RIGHTDOUBLEQUOTE : LEFTDOUBLEQUOTE);
  } else {
    contents = cmark_chunk_dup(&subj->input, subj->pos - numdelims, numdelims);
  }

  inl_text = make_str(subj, subj->pos - numdelims, subj->pos - 1, contents);

  if ((can_open || can_close) && (!(c == '\'' || c == '"') || smart)) {
    push_delimiter(subj, c, can_open, can_close, inl_text);
  }

  return inl_text;
}

// Assumes we have a hyphen at the current position.
static cmark_node *handle_hyphen(subject *subj, bool smart) {
  int startpos = subj->pos;

  advance(subj);

  if (!smart || peek_char(subj) != '-') {
    return make_str(subj, subj->pos - 1, subj->pos - 1, cmark_chunk_literal("-"));
  }

  while (smart && peek_char(subj) == '-') {
    advance(subj);
  }

  int numhyphens = subj->pos - startpos;
  int en_count = 0;
  int em_count = 0;
  int i;
  cmark_strbuf buf = CMARK_BUF_INIT(subj->mem);

  if (numhyphens % 3 == 0) { // if divisible by 3, use all em dashes
    em_count = numhyphens / 3;
  } else if (numhyphens % 2 == 0) { // if divisible by 2, use all en dashes
    en_count = numhyphens / 2;
  } else if (numhyphens % 3 == 2) { // use one en dash at end
    en_count = 1;
    em_count = (numhyphens - 2) / 3;
  } else { // use two en dashes at the end
    en_count = 2;
    em_count = (numhyphens - 4) / 3;
  }

  for (i = em_count; i > 0; i--) {
    cmark_strbuf_puts(&buf, EMDASH);
  }

  for (i = en_count; i > 0; i--) {
    cmark_strbuf_puts(&buf, ENDASH);
  }

  return make_str_from_buf(subj, startpos, subj->pos - 1, &buf);
}

// Assumes we have a period at the current position.
static cmark_node *handle_period(subject *subj, bool smart) {
  advance(subj);
  if (smart && peek_char(subj) == '.') {
    advance(subj);
    if (peek_char(subj) == '.') {
      advance(subj);
      return make_str(subj, subj->pos - 3, subj->pos - 1, cmark_chunk_literal(ELLIPSES));
    } else {
      return make_str(subj, subj->pos - 2, subj->pos - 1, cmark_chunk_literal(".."));
    }
  } else {
    return make_str(subj, subj->pos - 1, subj->pos - 1, cmark_chunk_literal("."));
  }
}

static void process_emphasis(subject *subj, bufsize_t stack_bottom) {
  delimiter *candidate;
  delimiter *closer = NULL;
  delimiter *opener;
  delimiter *old_closer;
  bool opener_found;
  int openers_bottom_index = 0;
  bufsize_t openers_bottom[15] = {stack_bottom, stack_bottom, stack_bottom,
                                  stack_bottom, stack_bottom, stack_bottom,
                                  stack_bottom, stack_bottom, stack_bottom,
                                  stack_bottom, stack_bottom, stack_bottom,
                                  stack_bottom, stack_bottom, stack_bottom};

  // move back to first relevant delim.
  candidate = subj->last_delim;
  while (candidate != NULL && candidate->position >= stack_bottom) {
    closer = candidate;
    candidate = candidate->previous;
  }

  // now move forward, looking for closers, and handling each
  while (closer != NULL) {
    if (closer->can_close) {
      switch (closer->delim_char) {
      case '"':
        openers_bottom_index = 0;
        break;
      case '\'':
        openers_bottom_index = 1;
        break;
      case '_':
        openers_bottom_index = 2 +
                (closer->can_open ? 3 : 0) + (closer->length % 3);
        break;
      case '*':
        openers_bottom_index = 8 +
                (closer->can_open ? 3 : 0) + (closer->length % 3);
        break;
      default:
        assert(false);
      }

      // Now look backwards for first matching opener:
      opener = closer->previous;
      opener_found = false;
      while (opener != NULL &&
             opener->position >= openers_bottom[openers_bottom_index]) {
        if (opener->can_open && opener->delim_char == closer->delim_char) {
          // interior closer of size 2 can't match opener of size 1
          // or of size 1 can't match 2
          if (!(closer->can_open || opener->can_close) ||
              closer->length % 3 == 0 ||
              (opener->length + closer->length) % 3 != 0) {
            opener_found = true;
            break;
          }
        }
        opener = opener->previous;
      }
      old_closer = closer;
      if (closer->delim_char == '*' || closer->delim_char == '_') {
        if (opener_found) {
          closer = S_insert_emph(subj, opener, closer);
        } else {
          closer = closer->next;
        }
      } else if (closer->delim_char == '\'' || closer->delim_char == '"') {
        if (closer->delim_char == '\'') {
          cmark_node_set_literal(closer->inl_text, RIGHTSINGLEQUOTE);
        } else {
          cmark_node_set_literal(closer->inl_text, RIGHTDOUBLEQUOTE);
        }
        closer = closer->next;
        if (opener_found) {
          if (old_closer->delim_char == '\'') {
            cmark_node_set_literal(opener->inl_text, LEFTSINGLEQUOTE);
          } else {
            cmark_node_set_literal(opener->inl_text, LEFTDOUBLEQUOTE);
          }
          remove_delimiter(subj, opener);
          remove_delimiter(subj, old_closer);
        }
      }
      if (!opener_found) {
        // set lower bound for future searches for openers
        openers_bottom[openers_bottom_index] = old_closer->position;
        if (!old_closer->can_open) {
          // we can remove a closer that can't be an
          // opener, once we've seen there's no
          // matching opener:
          remove_delimiter(subj, old_closer);
        }
      }
    } else {
      closer = closer->next;
    }
  }
  // free all delimiters in list until stack_bottom:
  while (subj->last_delim != NULL &&
         subj->last_delim->position >= stack_bottom) {
    remove_delimiter(subj, subj->last_delim);
  }
}

static delimiter *S_insert_emph(subject *subj, delimiter *opener,
                                delimiter *closer) {
  delimiter *delim, *tmp_delim;
  bufsize_t use_delims;
  cmark_node *opener_inl = opener->inl_text;
  cmark_node *closer_inl = closer->inl_text;
  bufsize_t opener_num_chars = opener_inl->len;
  bufsize_t closer_num_chars = closer_inl->len;
  cmark_node *tmp, *tmpnext, *emph;

  // calculate the actual number of characters used from this closer
  use_delims = (closer_num_chars >= 2 && opener_num_chars >= 2) ? 2 : 1;

  // remove used characters from associated inlines.
  opener_num_chars -= use_delims;
  closer_num_chars -= use_delims;
  opener_inl->len = opener_num_chars;
  opener_inl->data[opener_num_chars] = 0;
  opener_inl->end_column -= use_delims;
  closer_inl->len = closer_num_chars;
  closer_inl->data[closer_num_chars] = 0;
  closer_inl->start_column += use_delims;

  // free delimiters between opener and closer
  delim = closer->previous;
  while (delim != NULL && delim != opener) {
    tmp_delim = delim->previous;
    remove_delimiter(subj, delim);
    delim = tmp_delim;
  }

  // create new emph or strong, and splice it in to our inlines
  // between the opener and closer
  emph = use_delims == 1 ? make_emph(subj->mem) : make_strong(subj->mem);

  tmp = opener_inl->next;
  if (tmp && tmp != closer_inl) {
    emph->first_child = tmp;
    tmp->prev = NULL;

    while (tmp && tmp != closer_inl) {
      tmpnext = tmp->next;
      tmp->parent = emph;
      if (tmpnext == closer_inl) {
        emph->last_child = tmp;
        tmp->next = NULL;
      }
      tmp = tmpnext;
    }
  }

  opener_inl->next = emph;
  closer_inl->prev = emph;
  emph->prev = opener_inl;
  emph->next = closer_inl;
  emph->parent = opener_inl->parent;

  emph->start_line = opener_inl->start_line;
  emph->end_line = closer_inl->end_line;
  emph->start_column = opener_inl->start_column + opener_inl->len;
  emph->end_column = closer_inl->end_column - closer_inl->len;

  // if opener has 0 characters, remove it and its associated inline
  if (opener_num_chars == 0) {
    cmark_node_free(opener_inl);
    remove_delimiter(subj, opener);
  }

  // if closer has 0 characters, remove it and its associated inline
  if (closer_num_chars == 0) {
    // remove empty closer inline
    cmark_node_free(closer_inl);
    // remove closer from list
    tmp_delim = closer->next;
    remove_delimiter(subj, closer);
    closer = tmp_delim;
  }

  return closer;
}

// Parse backslash-escape or just a backslash, returning an inline.
static cmark_node *handle_backslash(subject *subj) {
  advance(subj);
  unsigned char nextchar = peek_char(subj);
  if (cmark_ispunct(
          nextchar)) { // only ascii symbols and newline can be escaped
    advance(subj);
    return make_str(subj, subj->pos - 2, subj->pos - 1, cmark_chunk_dup(&subj->input, subj->pos - 1, 1));
  } else if (!is_eof(subj) && skip_line_end(subj)) {
    return make_linebreak(subj->mem);
  } else {
    return make_str(subj, subj->pos - 1, subj->pos - 1, cmark_chunk_literal("\\"));
  }
}

// Parse an entity or a regular "&" string.
// Assumes the subject has an '&' character at the current position.
static cmark_node *handle_entity(subject *subj) {
  cmark_strbuf ent = CMARK_BUF_INIT(subj->mem);
  bufsize_t len;

  advance(subj);

  len = houdini_unescape_ent(&ent, subj->input.data + subj->pos,
                             subj->input.len - subj->pos);

  if (len <= 0)
    return make_str(subj, subj->pos - 1, subj->pos - 1, cmark_chunk_literal("&"));

  subj->pos += len;
  return make_str_from_buf(subj, subj->pos - 1 - len, subj->pos - 1, &ent);
}

// Clean a URL: remove surrounding whitespace, and remove \ that escape
// punctuation.
unsigned char *cmark_clean_url(cmark_mem *mem, cmark_chunk *url) {
  cmark_strbuf buf = CMARK_BUF_INIT(mem);

  cmark_chunk_trim(url);

  houdini_unescape_html_f(&buf, url->data, url->len);

  cmark_strbuf_unescape(&buf);
  return cmark_strbuf_detach(&buf);
}

unsigned char *cmark_clean_title(cmark_mem *mem, cmark_chunk *title) {
  cmark_strbuf buf = CMARK_BUF_INIT(mem);
  unsigned char first, last;

  if (title->len == 0) {
    return NULL;
  }

  first = title->data[0];
  last = title->data[title->len - 1];

  // remove surrounding quotes if any:
  if ((first == '\'' && last == '\'') || (first == '(' && last == ')') ||
      (first == '"' && last == '"')) {
    houdini_unescape_html_f(&buf, title->data + 1, title->len - 2);
  } else {
    houdini_unescape_html_f(&buf, title->data, title->len);
  }

  cmark_strbuf_unescape(&buf);
  return cmark_strbuf_detach(&buf);
}

// Parse an autolink or HTML tag.
// Assumes the subject has a '<' character at the current position.
static cmark_node *handle_pointy_brace(subject *subj, int options) {
  bufsize_t matchlen = 0;
  cmark_chunk contents;

  advance(subj); // advance past first <

  // first try to match a URL autolink
  matchlen = scan_autolink_uri(&subj->input, subj->pos);
  if (matchlen > 0) {
    contents = cmark_chunk_dup(&subj->input, subj->pos, matchlen - 1);
    subj->pos += matchlen;

    return make_autolink(subj, subj->pos - 1 - matchlen, subj->pos - 1, contents, 0);
  }

  // next try to match an email autolink
  matchlen = scan_autolink_email(&subj->input, subj->pos);
  if (matchlen > 0) {
    contents = cmark_chunk_dup(&subj->input, subj->pos, matchlen - 1);
    subj->pos += matchlen;

    return make_autolink(subj, subj->pos - 1 - matchlen, subj->pos - 1, contents, 1);
  }

  // finally, try to match an html tag
  if (subj->pos + 2 <= subj->input.len) {
    int c = subj->input.data[subj->pos];
    if (c == '!' && (subj->flags & FLAG_SKIP_HTML_COMMENT) == 0) {
      c = subj->input.data[subj->pos+1];
      if (c == '-' && subj->input.data[subj->pos+2] == '-') {
	if (subj->pos + 3 < subj->input.len &&
            subj->input.data[subj->pos+3] == '>') {
	  matchlen = 4;
	} else if (subj->pos + 4 < subj->input.len &&
                   subj->input.data[subj->pos+3] == '-' &&
                   subj->input.data[subj->pos+4] == '>') {
          matchlen = 5;
        } else {
          matchlen = scan_html_comment(&subj->input, subj->pos + 1);
          if (matchlen > 0) {
            matchlen += 1; // prefix "<"
	  } else { // no match through end of input: set a flag so
		   // we don't reparse looking for -->:
	    subj->flags |= FLAG_SKIP_HTML_COMMENT;
	  }
	}
      } else if (c == '[') {
        if ((subj->flags & FLAG_SKIP_HTML_CDATA) == 0) {
          matchlen = scan_html_cdata(&subj->input, subj->pos + 2);
          if (matchlen > 0) {
            // The regex doesn't require the final "]]>". But if we're not at
            // the end of input, it must come after the match. Otherwise,
            // disable subsequent scans to avoid quadratic behavior.
            matchlen += 5; // prefix "![", suffix "]]>"
            if (subj->pos + matchlen > subj->input.len) {
              subj->flags |= FLAG_SKIP_HTML_CDATA;
              matchlen = 0;
            }
          }
        }
      } else if ((subj->flags & FLAG_SKIP_HTML_DECLARATION) == 0) {
        matchlen = scan_html_declaration(&subj->input, subj->pos + 1);
        if (matchlen > 0) {
          matchlen += 2; // prefix "!", suffix ">"
          if (subj->pos + matchlen > subj->input.len) {
            subj->flags |= FLAG_SKIP_HTML_DECLARATION;
            matchlen = 0;
          }
        }
      }
    } else if (c == '?') {
      if ((subj->flags & FLAG_SKIP_HTML_PI) == 0) {
        // Note that we allow an empty match.
        matchlen = scan_html_pi(&subj->input, subj->pos + 1);
        matchlen += 3; // prefix "?", suffix "?>"
        if (subj->pos + matchlen > subj->input.len) {
          subj->flags |= FLAG_SKIP_HTML_PI;
          matchlen = 0;
        }
      }
    } else {
      matchlen = scan_html_tag(&subj->input, subj->pos);
    }
  }
  if (matchlen > 0) {
    const unsigned char *src = subj->input.data + subj->pos - 1;
    bufsize_t len = matchlen + 1;
    subj->pos += matchlen;
    cmark_node *node = make_literal(subj, CMARK_NODE_HTML_INLINE,
                                    subj->pos - matchlen - 1, subj->pos - 1);
    node->data = (unsigned char *)subj->mem->realloc(NULL, len + 1);
    memcpy(node->data, src, len);
    node->data[len] = 0;
    node->len = len;
    adjust_subj_node_newlines(subj, node, matchlen, 1, options);
    return node;
  }

  // if nothing matches, just return the opening <:
  return make_str(subj, subj->pos - 1, subj->pos - 1, cmark_chunk_literal("<"));
}

// Parse a link label.  Returns 1 if successful.
// Note:  unescaped brackets are not allowed in labels.
// The label begins with `[` and ends with the first `]` character
// encountered.  Backticks in labels do not start code spans.
static int link_label(subject *subj, cmark_chunk *raw_label) {
  bufsize_t startpos = subj->pos;
  int length = 0;
  unsigned char c;

  // advance past [
  if (peek_char(subj) == '[') {
    advance(subj);
  } else {
    return 0;
  }

  while ((c = peek_char(subj)) && c != '[' && c != ']') {
    if (c == '\\') {
      advance(subj);
      length++;
      if (cmark_ispunct(peek_char(subj))) {
        advance(subj);
        length++;
      }
    } else {
      advance(subj);
      length++;
    }
    if (length > MAX_LINK_LABEL_LENGTH) {
      goto noMatch;
    }
  }

  if (c == ']') { // match found
    *raw_label =
        cmark_chunk_dup(&subj->input, startpos + 1, subj->pos - (startpos + 1));
    cmark_chunk_trim(raw_label);
    advance(subj); // advance past ]
    return 1;
  }

noMatch:
  subj->pos = startpos; // rewind
  return 0;
}

static bufsize_t manual_scan_link_url_2(cmark_chunk *input, bufsize_t offset,
                                        cmark_chunk *output) {
  bufsize_t i = offset;
  size_t nb_p = 0;

  while (i < input->len) {
    if (input->data[i] == '\\' &&
        i + 1 < input-> len &&
        cmark_ispunct(input->data[i+1]))
      i += 2;
    else if (input->data[i] == '(') {
      ++nb_p;
      ++i;
      if (nb_p > 32)
        return -1;
    } else if (input->data[i] == ')') {
      if (nb_p == 0)
        break;
      --nb_p;
      ++i;
    } else if (cmark_isspace(input->data[i])) {
      if (i == offset) {
        return -1;
      }
      break;
    } else {
      ++i;
    }
  }

  if (i >= input->len || nb_p != 0)
    return -1;

  {
    cmark_chunk result = {input->data + offset, i - offset};
    *output = result;
  }
  return i - offset;
}

static bufsize_t manual_scan_link_url(cmark_chunk *input, bufsize_t offset,
                                      cmark_chunk *output) {
  bufsize_t i = offset;

  if (i < input->len && input->data[i] == '<') {
    ++i;
    while (i < input->len) {
      if (input->data[i] == '>') {
        ++i;
        break;
      } else if (input->data[i] == '\\')
        i += 2;
      else if (input->data[i] == '\n' || input->data[i] == '<')
        return -1;
      else
        ++i;
    }
  } else {
    return manual_scan_link_url_2(input, offset, output);
  }

  if (i >= input->len)
    return -1;

  {
    cmark_chunk result = {input->data + offset + 1, i - 2 - offset};
    *output = result;
  }
  return i - offset;
}

// Return a link, an image, or a literal close bracket.
static cmark_node *handle_close_bracket(subject *subj) {
  bufsize_t initial_pos, after_link_text_pos;
  bufsize_t endurl, starttitle, endtitle, endall;
  bufsize_t sps, n;
  cmark_reference *ref = NULL;
  cmark_chunk url_chunk, title_chunk;
  unsigned char *url, *title;
  bracket *opener;
  cmark_node *inl;
  cmark_chunk raw_label;
  int found_label;
  cmark_node *tmp, *tmpnext;
  bool is_image;

  advance(subj); // advance past ]
  initial_pos = subj->pos;

  // get last [ or ![
  opener = subj->last_bracket;

  if (opener == NULL) {
    return make_str(subj, subj->pos - 1, subj->pos - 1, cmark_chunk_literal("]"));
  }

  // If we got here, we matched a potential link/image text.
  // Now we check to see if it's a link/image.
  is_image = opener->image;

  if (!is_image && subj->no_link_openers) {
    // take delimiter off stack
    pop_bracket(subj);
    return make_str(subj, subj->pos - 1, subj->pos - 1, cmark_chunk_literal("]"));
  }

  after_link_text_pos = subj->pos;

  // First, look for an inline link.
  if (peek_char(subj) == '(' &&
      ((sps = scan_spacechars(&subj->input, subj->pos + 1)) > -1) &&
      ((n = manual_scan_link_url(&subj->input, subj->pos + 1 + sps,
                                 &url_chunk)) > -1)) {

    // try to parse an explicit link:
    endurl = subj->pos + 1 + sps + n;
    starttitle = endurl + scan_spacechars(&subj->input, endurl);

    // ensure there are spaces btw url and title
    endtitle = (starttitle == endurl)
                   ? starttitle
                   : starttitle + scan_link_title(&subj->input, starttitle);

    endall = endtitle + scan_spacechars(&subj->input, endtitle);

    if (peek_at(subj, endall) == ')') {
      subj->pos = endall + 1;

      title_chunk =
          cmark_chunk_dup(&subj->input, starttitle, endtitle - starttitle);
      url = cmark_clean_url(subj->mem, &url_chunk);
      title = cmark_clean_title(subj->mem, &title_chunk);
      cmark_chunk_free(&url_chunk);
      cmark_chunk_free(&title_chunk);
      goto match;

    } else {
      // it could still be a shortcut reference link
      subj->pos = after_link_text_pos;
    }
  }

  // Next, look for a following [link label] that matches in refmap.
  // skip spaces
  raw_label = cmark_chunk_literal("");
  found_label = link_label(subj, &raw_label);
  if (!found_label) {
    // If we have a shortcut reference link, back up
    // to before the spaces we skipped.
    subj->pos = initial_pos;
  }

  if ((!found_label || raw_label.len == 0) && !opener->bracket_after) {
    cmark_chunk_free(&raw_label);
    raw_label = cmark_chunk_dup(&subj->input, opener->position,
                                initial_pos - opener->position - 1);
    found_label = true;
  }

  if (found_label) {
    ref = cmark_reference_lookup(subj->refmap, &raw_label);
    cmark_chunk_free(&raw_label);
  }

  if (ref != NULL) { // found
    url = cmark_strdup(subj->mem, ref->url);
    title = cmark_strdup(subj->mem, ref->title);
    goto match;
  } else {
    goto noMatch;
  }

noMatch:
  // If we fall through to here, it means we didn't match a link:
  pop_bracket(subj); // remove this opener from delimiter list
  subj->pos = initial_pos;
  return make_str(subj, subj->pos - 1, subj->pos - 1, cmark_chunk_literal("]"));

match:
  inl = make_simple(subj->mem, is_image ? CMARK_NODE_IMAGE : CMARK_NODE_LINK);
  inl->as.link.url = url;
  inl->as.link.title = title;
  inl->start_line = inl->end_line = subj->line;
  inl->start_column = opener->inl_text->start_column;
  inl->end_column = subj->pos + subj->column_offset + subj->block_offset;
  cmark_node_insert_before(opener->inl_text, inl);
  // Add link text:
  tmp = opener->inl_text->next;
  while (tmp) {
    tmpnext = tmp->next;
    cmark_node_unlink(tmp);
    append_child(inl, tmp);
    tmp = tmpnext;
  }

  // Free the bracket [:
  cmark_node_free(opener->inl_text);

  process_emphasis(subj, opener->position);
  pop_bracket(subj);

  // Now, if we have a link, we also want to deactivate links until
  // we get a new opener. (This code can be removed if we decide to allow links
  // inside links.)
  if (!is_image) {
    subj->no_link_openers = true;
  }

  return NULL;
}

// Parse a hard or soft linebreak, returning an inline.
// Assumes the subject has a cr or newline at the current position.
static cmark_node *handle_newline(subject *subj) {
  bufsize_t nlpos = subj->pos;
  // skip over cr, crlf, or lf:
  if (peek_at(subj, subj->pos) == '\r') {
    advance(subj);
  }
  if (peek_at(subj, subj->pos) == '\n') {
    advance(subj);
  }
  ++subj->line;
  subj->column_offset = -subj->pos;
  // skip spaces at beginning of line
  skip_spaces(subj);
  if (nlpos > 1 && peek_at(subj, nlpos - 1) == ' ' &&
      peek_at(subj, nlpos - 2) == ' ') {
    return make_linebreak(subj->mem);
  } else {
    return make_softbreak(subj->mem);
  }
}

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,
  };

  bufsize_t n = subj->pos + 1;

  while (n < subj->input.len) {
    if (SPECIAL_CHARS[subj->input.data[n]])
      return n;
    if (options & CMARK_OPT_SMART && SMART_PUNCT_CHARS[subj->input.data[n]])
      return n;
    n++;
  }

  return subj->input.len;
}

// 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) {
  cmark_node *new_inl = NULL;
  cmark_chunk contents;
  unsigned char c;
  bufsize_t startpos, endpos;
  c = peek_char(subj);
  if (c == 0) {
    return 0;
  }
  switch (c) {
  case '\r':
  case '\n':
    new_inl = handle_newline(subj);
    break;
  case '`':
    new_inl = handle_backticks(subj, options);
    break;
  case '\\':
    new_inl = handle_backslash(subj);
    break;
  case '&':
    new_inl = handle_entity(subj);
    break;
  case '<':
    new_inl = handle_pointy_brace(subj, options);
    break;
  case '*':
  case '_':
  case '\'':
  case '"':
    new_inl = handle_delim(subj, c, (options & CMARK_OPT_SMART) != 0);
    break;
  case '-':
    new_inl = handle_hyphen(subj, (options & CMARK_OPT_SMART) != 0);
    break;
  case '.':
    new_inl = handle_period(subj, (options & CMARK_OPT_SMART) != 0);
    break;
  case '[':
    advance(subj);
    new_inl = make_str(subj, subj->pos - 1, subj->pos - 1, cmark_chunk_literal("["));
    push_bracket(subj, false, new_inl);
    break;
  case ']':
    new_inl = handle_close_bracket(subj);
    break;
  case '!':
    advance(subj);
    if (peek_char(subj) == '[') {
      advance(subj);
      new_inl = make_str(subj, subj->pos - 2, subj->pos - 1, cmark_chunk_literal("!["));
      push_bracket(subj, true, new_inl);
    } else {
      new_inl = make_str(subj, subj->pos - 1, subj->pos - 1, cmark_chunk_literal("!"));
    }
    break;
  default:
    endpos = subject_find_special_char(subj, options);
    contents = cmark_chunk_dup(&subj->input, subj->pos, endpos - subj->pos);
    startpos = subj->pos;
    subj->pos = endpos;

    // if we're at a newline, strip trailing spaces.
    if (S_is_line_end_char(peek_char(subj))) {
      cmark_chunk_rtrim(&contents);
    }

    new_inl = make_str(subj, startpos, endpos - 1, contents);
  }
  if (new_inl != NULL) {
    append_child(parent, new_inl);
  }

  return 1;
}

// Parse inlines from parent's string_content, adding as children of parent.
void cmark_parse_inlines(cmark_mem *mem, cmark_node *parent,
                         cmark_reference_map *refmap, int options) {
  int internal_offset = parent->type == CMARK_NODE_HEADING ?
    parent->as.heading.internal_offset : 0;
  subject subj;
  cmark_chunk content = {parent->data, parent->len};
  subject_from_buf(mem, parent->start_line, parent->start_column - 1 + internal_offset, &subj, &content, refmap);
  cmark_chunk_rtrim(&subj.input);

  while (!is_eof(&subj) && parse_inline(&subj, parent, options))
    ;

  process_emphasis(&subj, 0);
  // free bracket and delim stack
  while (subj.last_delim) {
    remove_delimiter(&subj, subj.last_delim);
  }
  while (subj.last_bracket) {
    pop_bracket(&subj);
  }
}

// Parse zero or more space characters, including at most one newline.
static void spnl(subject *subj) {
  skip_spaces(subj);
  if (skip_line_end(subj)) {
    skip_spaces(subj);
  }
}

// Parse reference.  Assumes string begins with '[' character.
// Modify refmap if a reference is encountered.
// Return 0 if no reference found, otherwise position of subject
// after reference is parsed.
bufsize_t cmark_parse_reference_inline(cmark_mem *mem, cmark_chunk *input,
                                       cmark_reference_map *refmap) {
  subject subj;

  cmark_chunk lab;
  cmark_chunk url;
  cmark_chunk title;

  bufsize_t matchlen = 0;
  bufsize_t beforetitle;

  subject_from_buf(mem, -1, 0, &subj, input, NULL);

  // parse label:
  if (!link_label(&subj, &lab) || lab.len == 0)
    return 0;

  // colon:
  if (peek_char(&subj) == ':') {
    advance(&subj);
  } else {
    return 0;
  }

  // parse link url:
  spnl(&subj);
  if ((matchlen = manual_scan_link_url(&subj.input, subj.pos, &url)) > -1) {
    subj.pos += matchlen;
  } else {
    return 0;
  }

  // parse optional link_title
  beforetitle = subj.pos;
  spnl(&subj);
  matchlen = subj.pos == beforetitle ? 0 : scan_link_title(&subj.input, subj.pos);
  if (matchlen) {
    title = cmark_chunk_dup(&subj.input, subj.pos, matchlen);
    subj.pos += matchlen;
  } else {
    subj.pos = beforetitle;
    title = cmark_chunk_literal("");
  }

  // parse final spaces and newline:
  skip_spaces(&subj);
  if (!skip_line_end(&subj)) {
    if (matchlen) { // try rewinding before title
      subj.pos = beforetitle;
      skip_spaces(&subj);
      if (!skip_line_end(&subj)) {
        return 0;
      }
      // the previously scanned title turned out to be invalid (it was
      // followed by trailing garbage instead of the end of the line), so
      // the reference has no title after all.
      title = cmark_chunk_literal("");
    } else {
      return 0;
    }
  }
  // insert reference into refmap
  cmark_reference_create(refmap, &lab, &url, &title);
  return subj.pos;
}