res.c

wget (command line browser) source code

      SKIP_SPACE (p);

      value_b = p;
      while (!EOL (p))
	++p;
      value_e = p;

      /* Finally, we have a syntactically valid line. */
      if (FIELD_IS ("user-agent"))
	{
	  /* We have to support several cases:

	     --previous records--

	     User-Agent: foo
	     User-Agent: Wget
	     User-Agent: bar
	     ... matching record ...

	     User-Agent: baz
	     User-Agent: qux
	     ... non-matching record ...

	     User-Agent: *
	     ... matching record, but will be pruned later ...

	     We have to respect `User-Agent' at the beginning of each
	     new record simply because we don't know if we're going to
	     encounter "Wget" among the agents or not.  Hence,
	     match_user_agent is called when record_count != 0.

	     But if record_count is 0, we have to keep calling it
	     until it matches, and if that happens, we must not call
	     it any more, until the next record.  Hence the other part
	     of the condition.  */
	  if (record_count != 0 || user_agent_applies == 0)
	    match_user_agent (value_b, value_e - value_b,
			      &user_agent_applies, &user_agent_exact);
	  if (user_agent_exact)
	    found_exact = 1;
	  record_count = 0;
	}
      else if (FIELD_IS ("allow"))
	{
	  if (user_agent_applies)
	    {
	      add_path (specs, value_b, value_e, 1, user_agent_exact);
	    }
	  ++record_count;
	}
      else if (FIELD_IS ("disallow"))
	{
	  if (user_agent_applies)
	    {
	      int allowed = 0;
	      if (value_b == value_e)
		/* Empty "disallow" line means everything is
		   *allowed*!  */
		allowed = 1;
	      add_path (specs, value_b, value_e, allowed, user_agent_exact);
	    }
	  ++record_count;
	}
      else
	{
	  DEBUGP (("Ignoring unknown field at line %d", line_count));
	  goto next;
	}

    next:
      p = lineend_real;
      ++line_count;
    }

  if (found_exact)
    {
      /* We've encountered an exactly matching user-agent.  Throw out
	 all the stuff with user-agent: *.  */
      prune_non_exact (specs);
    }
  else if (specs->size > specs->count)
    {
      /* add_path normally over-allocates specs->paths.  Reallocate it
	 to the correct size in order to conserve some memory.  */
      specs->paths = xrealloc (specs->paths,
			       specs->count * sizeof (struct path_info));
      specs->size = specs->count;
    }

  return specs;
}

/* The same like res_parse, but first map the FILENAME into memory,
   and then parse it.  */

struct robot_specs *
res_parse_from_file (const char *filename)
{
  struct robot_specs *specs;
  struct file_memory *fm = read_file (filename);
  if (!fm)
    {
      logprintf (LOG_NOTQUIET, "Cannot open %s: %s",
		 filename, strerror (errno));
      return NULL;
    }
  specs = res_parse (fm->content, fm->length);
  read_file_free (fm);
  return specs;
}

static void
free_specs (struct robot_specs *specs)
{
  int i;
  for (i = 0; i < specs->count; i++)
    xfree (specs->paths[i].path);
  FREE_MAYBE (specs->paths);
  xfree (specs);
}

/* Matching of a path according to the specs. */

/* If C is '%' and (ptr[1], ptr[2]) form a hexadecimal number, and if
   that number is not a numerical representation of '/', decode C and
   advance the pointer.  */

#define DECODE_MAYBE(c, ptr) do {				\
  if (c == '%' && ISXDIGIT (ptr[1]) && ISXDIGIT (ptr[2]))	\
    {								\
      char decoded = X2DIGITS_TO_NUM (ptr[1], ptr[2]);		\
      if (decoded != '/')					\
	{							\
	  c = decoded;						\
	  ptr += 2;						\
	}							\
    }								\
} while (0)

/* The inner matching engine: return non-zero if RECORD_PATH matches
   URL_PATH.  The rules for matching are described at
   <http://www.robotstxt.org/wc/norobots-rfc.txt>, section 3.2.2.  */

static int
matches (const char *record_path, const char *url_path)
{
  const char *rp = record_path;
  const char *up = url_path;

  for (; ; ++rp, ++up)
    {
      char rc = *rp;
      char uc = *up;
      if (!rc)
	return 1;
      if (!uc)
	return 0;
      DECODE_MAYBE(rc, rp);
      DECODE_MAYBE(uc, up);
      if (rc != uc)
	return 0;
    }
}

/* Iterate through all paths in SPECS.  For the first one that
   matches, return its allow/reject status.  If none matches,
   retrieval is by default allowed.  */

int
res_match_path (const struct robot_specs *specs, const char *path)
{
  int i;
  if (!specs)
    return 1;
  for (i = 0; i < specs->count; i++)
    if (matches (specs->paths[i].path, path))
      {
	int allowedp = specs->paths[i].allowedp;
	DEBUGP (("%s path %s because of rule `%s'.\n",
		 allowedp ? "Allowing" : "Rejecting",
		 path, specs->paths[i].path));
	return allowedp;
      }
  return 1;
}

/* Registering the specs. */

static struct hash_table *registered_specs;

/* Stolen from cookies.c. */
#define SET_HOSTPORT(host, port, result) do {		\
  int HP_len = strlen (host);				\
  result = alloca (HP_len + 1 + numdigit (port) + 1);	\
  memcpy (result, host, HP_len);			\
  result[HP_len] = ':';					\
  number_to_string (result + HP_len + 1, port);		\
} while (0)

/* Register RES specs that below to server on HOST:PORT.  They will
   later be retrievable using res_get_specs.  */

void
res_register_specs (const char *host, int port, struct robot_specs *specs)
{
  struct robot_specs *old;
  char *hp, *hp_old;
  SET_HOSTPORT (host, port, hp);

  if (!registered_specs)
    registered_specs = make_nocase_string_hash_table (0);

  /* Required to shut up the compiler. */
  old    = NULL;
  hp_old = NULL;

  if (hash_table_get_pair (registered_specs, hp, hp_old, old))
    {
      if (old)
	free_specs (old);
      hash_table_put (registered_specs, hp_old, specs);
    }
  else
    {
      hash_table_put (registered_specs, xstrdup (hp), specs);
    }
}

/* Get the specs that belong to HOST:PORT. */

struct robot_specs *
res_get_specs (const char *host, int port)
{
  char *hp;
  SET_HOSTPORT (host, port, hp);
  if (!registered_specs)
    return NULL;
  return hash_table_get (registered_specs, hp);
}

/* Loading the robots file.  */

#define RES_SPECS_LOCATION "/robots.txt"

/* Retrieve the robots.txt from the server root of the server that
   serves URL.  The file will be named according to the currently
   active rules, and the file name will be returned in *file.

   Return non-zero if robots were retrieved OK, zero otherwise.  */

int
res_retrieve_file (const char *url, char **file)
{
  uerr_t err;
  char *robots_url = uri_merge (url, RES_SPECS_LOCATION);

  logputs (LOG_VERBOSE, _("Loading robots.txt; please ignore errors.\n"));
  *file = NULL;
  err = retrieve_url (robots_url, file, NULL, NULL, NULL);
  xfree (robots_url);

  if (err != RETROK && *file != NULL)
    {
      /* If the file is not retrieved correctly, but retrieve_url
	 allocated the file name, deallocate is here so that the
	 caller doesn't have to worry about it.  */
      xfree (*file);
      *file = NULL;
    }
  return err == RETROK;
}

static int
cleanup_hash_table_mapper (void *key, void *value, void *arg_ignored)
{
  xfree (key);
  free_specs (value);
  return 0;
}

void
res_cleanup (void)
{
  if (registered_specs)
    {
      hash_table_map (registered_specs, cleanup_hash_table_mapper, NULL);
      hash_table_destroy (registered_specs);
      registered_specs = NULL;
    }
}