cookies.c

Wget很好的处理了http和ftp的下载,很值得学习的经典代码

  struct weighed_cookie *wc2 = (struct weighed_cookie *)p2;

  /* Subtractions take `wc2' as the first argument becauase we want a
     sort in *decreasing* order of goodness.  */
  int dgdiff = wc2->domain_goodness - wc1->domain_goodness;
  int pgdiff = wc2->path_goodness - wc1->path_goodness;

  /* Sort by domain goodness; if these are the same, sort by path
     goodness.  (The sorting order isn't really specified; maybe it
     should be the other way around.)  */
  return dgdiff ? dgdiff : pgdiff;
}

/* Generate a `Cookie' header for a request that goes to HOST:PORT and
   requests PATH from the server.  The resulting string is allocated
   with `malloc', and the caller is responsible for freeing it.  If no
   cookies pertain to this request, i.e. no cookie header should be
   generated, NULL is returned.  */

char *
cookie_header (struct cookie_jar *jar, const char *host,
               int port, const char *path, bool secflag)
{
  struct cookie **chains;
  int chain_count;

  struct cookie *cookie;
  struct weighed_cookie *outgoing;
  int count, i, ocnt;
  char *result;
  int result_size, pos;
  PREPEND_SLASH (path);         /* see cookie_handle_set_cookie */

  /* First, find the cookie chains whose domains match HOST. */

  /* Allocate room for find_chains_of_host to write to.  The number of
     chains can at most equal the number of subdomains, hence
     1+<number of dots>.  */
  chains = alloca_array (struct cookie *, 1 + count_char (host, '.'));
  chain_count = find_chains_of_host (jar, host, chains);

  /* No cookies for this host. */
  if (!chain_count)
    return NULL;

  cookies_now = time (NULL);

  /* Now extract from the chains those cookies that match our host
     (for domain_exact cookies), port (for cookies with port other
     than PORT_ANY), etc.  See matching_cookie for details.  */

  /* Count the number of matching cookies. */
  count = 0;
  for (i = 0; i < chain_count; i++)
    for (cookie = chains[i]; cookie; cookie = cookie->next)
      if (cookie_matches_url (cookie, host, port, path, secflag, NULL))
        ++count;
  if (!count)
    return NULL;                /* no cookies matched */

  /* Allocate the array. */
  outgoing = alloca_array (struct weighed_cookie, count);

  /* Fill the array with all the matching cookies from the chains that
     match HOST. */
  ocnt = 0;
  for (i = 0; i < chain_count; i++)
    for (cookie = chains[i]; cookie; cookie = cookie->next)
      {
        int pg;
        if (!cookie_matches_url (cookie, host, port, path, secflag, &pg))
          continue;
        outgoing[ocnt].cookie = cookie;
        outgoing[ocnt].domain_goodness = strlen (cookie->domain);
        outgoing[ocnt].path_goodness   = pg;
        ++ocnt;
      }
  assert (ocnt == count);

  /* Eliminate duplicate cookies; that is, those whose name and value
     are the same.  */
  count = eliminate_dups (outgoing, count);

  /* Sort the array so that best-matching domains come first, and
     that, within one domain, best-matching paths come first. */
  qsort (outgoing, count, sizeof (struct weighed_cookie), goodness_comparator);

  /* Count the space the name=value pairs will take. */
  result_size = 0;
  for (i = 0; i < count; i++)
    {
      struct cookie *c = outgoing[i].cookie;
      /* name=value */
      result_size += strlen (c->attr) + 1 + strlen (c->value);
    }

  /* Allocate output buffer:
     name=value pairs -- result_size
     "; " separators  -- (count - 1) * 2
     \0 terminator    -- 1 */
  result_size = result_size + (count - 1) * 2 + 1;
  result = xmalloc (result_size);
  pos = 0;
  for (i = 0; i < count; i++)
    {
      struct cookie *c = outgoing[i].cookie;
      int namlen = strlen (c->attr);
      int vallen = strlen (c->value);

      memcpy (result + pos, c->attr, namlen);
      pos += namlen;
      result[pos++] = '=';
      memcpy (result + pos, c->value, vallen);
      pos += vallen;
      if (i < count - 1)
        {
          result[pos++] = ';';
          result[pos++] = ' ';
        }
    }
  result[pos++] = '\0';
  assert (pos == result_size);
  return result;
}

/* Support for loading and saving cookies.  The format used for
   loading and saving should be the format of the `cookies.txt' file
   used by Netscape and Mozilla, at least the Unix versions.
   (Apparently IE can export cookies in that format as well.)  The
   format goes like this:

       DOMAIN DOMAIN-FLAG PATH SECURE-FLAG TIMESTAMP ATTR-NAME ATTR-VALUE

     DOMAIN      -- cookie domain, optionally followed by :PORT
     DOMAIN-FLAG -- whether all hosts in the domain match
     PATH        -- cookie path
     SECURE-FLAG -- whether cookie requires secure connection
     TIMESTAMP   -- expiry timestamp, number of seconds since epoch
     ATTR-NAME   -- name of the cookie attribute
     ATTR-VALUE  -- value of the cookie attribute (empty if absent)

   The fields are separated by TABs.  All fields are mandatory, except
   for ATTR-VALUE.  The `-FLAG' fields are boolean, their legal values
   being "TRUE" and "FALSE'.  Empty lines, lines consisting of
   whitespace only, and comment lines (beginning with # optionally
   preceded by whitespace) are ignored.

   Example line from cookies.txt (split in two lines for readability):

       .google.com      TRUE    /       FALSE   2147368447      \
       PREF     ID=34bb47565bbcd47b:LD=en:NR=20:TM=985172580:LM=985739012

*/

/* If the region [B, E) ends with :<digits>, parse the number, return
   it, and store new boundary (location of the `:') to DOMAIN_E_PTR.
   If port is not specified, return 0.  */

static int
domain_port (const char *domain_b, const char *domain_e,
             const char **domain_e_ptr)
{
  int port = 0;
  const char *p;
  const char *colon = memchr (domain_b, ':', domain_e - domain_b);
  if (!colon)
    return 0;
  for (p = colon + 1; p < domain_e && ISDIGIT (*p); p++)
    port = 10 * port + (*p - '0');
  if (p < domain_e)
    /* Garbage following port number. */
    return 0;
  *domain_e_ptr = colon;
  return port;
}

#define GET_WORD(p, b, e) do {                  \
  b = p;                                        \
  while (*p && *p != '\t')                      \
    ++p;                                        \
  e = p;                                        \
  if (b == e || !*p)                            \
    goto next;                                  \
  ++p;                                          \
} while (0)

/* Load cookies from FILE.  */

void
cookie_jar_load (struct cookie_jar *jar, const char *file)
{
  char *line;
  FILE *fp = fopen (file, "r");
  if (!fp)
    {
      logprintf (LOG_NOTQUIET, _("Cannot open cookies file `%s': %s\n"),
                 file, strerror (errno));
      return;
    }
  cookies_now = time (NULL);

  for (; ((line = read_whole_line (fp)) != NULL); xfree (line))
    {
      struct cookie *cookie;
      char *p = line;

      double expiry;
      int port;

      char *domain_b  = NULL, *domain_e  = NULL;
      char *domflag_b = NULL, *domflag_e = NULL;
      char *path_b    = NULL, *path_e    = NULL;
      char *secure_b  = NULL, *secure_e  = NULL;
      char *expires_b = NULL, *expires_e = NULL;
      char *name_b    = NULL, *name_e    = NULL;
      char *value_b   = NULL, *value_e   = NULL;

      /* Skip leading white-space. */
      while (*p && ISSPACE (*p))
        ++p;
      /* Ignore empty lines.  */
      if (!*p || *p == '#')
        continue;

      GET_WORD (p, domain_b,  domain_e);
      GET_WORD (p, domflag_b, domflag_e);
      GET_WORD (p, path_b,    path_e);
      GET_WORD (p, secure_b,  secure_e);
      GET_WORD (p, expires_b, expires_e);
      GET_WORD (p, name_b,    name_e);

      /* Don't use GET_WORD for value because it ends with newline,
         not TAB.  */
      value_b = p;
      value_e = p + strlen (p);
      if (value_e > value_b && value_e[-1] == '\n')
        --value_e;
      if (value_e > value_b && value_e[-1] == '\r')
        --value_e;
      /* Empty values are legal (I think), so don't bother checking. */

      cookie = cookie_new ();

      cookie->attr    = strdupdelim (name_b, name_e);
      cookie->value   = strdupdelim (value_b, value_e);
      cookie->path    = strdupdelim (path_b, path_e);
      cookie->secure  = BOUNDED_EQUAL (secure_b, secure_e, "TRUE");

      /* Curl source says, quoting Andre Garcia: "flag: A TRUE/FALSE
         value indicating if all machines within a given domain can
         access the variable.  This value is set automatically by the
         browser, depending on the value set for the domain."  */
      cookie->domain_exact = !BOUNDED_EQUAL (domflag_b, domflag_e, "TRUE");

      /* DOMAIN needs special treatment because we might need to
         extract the port.  */
      port = domain_port (domain_b, domain_e, (const char **)&domain_e);
      if (port)
        cookie->port = port;

      if (*domain_b == '.')
        ++domain_b;             /* remove leading dot internally */
      cookie->domain  = strdupdelim (domain_b, domain_e);

      /* safe default in case EXPIRES field is garbled. */
      expiry = (double)cookies_now - 1;

      /* I don't like changing the line, but it's safe here.  (line is
         malloced.)  */
      *expires_e = '\0';
      sscanf (expires_b, "%lf", &expiry);

      if (expiry == 0)
        {
          /* EXPIRY can be 0 for session cookies saved because the
             user specified `--keep-session-cookies' in the past.
             They remain session cookies, and will be saved only if
             the user has specified `keep-session-cookies' again.  */
        }
      else
        {
          if (expiry < cookies_now)
            goto abort_cookie;  /* ignore stale cookie. */
          cookie->expiry_time = expiry;
          cookie->permanent = 1;
        }

      store_cookie (jar, cookie);

    next:
      continue;

    abort_cookie:
      delete_cookie (cookie);
    }
  fclose (fp);
}

/* Save cookies, in format described above, to FILE. */

void
cookie_jar_save (struct cookie_jar *jar, const char *file)
{
  FILE *fp;
  hash_table_iterator iter;

  DEBUGP (("Saving cookies to %s.\n", file));

  cookies_now = time (NULL);

  fp = fopen (file, "w");
  if (!fp)
    {
      logprintf (LOG_NOTQUIET, _("Cannot open cookies file `%s': %s\n"),
                 file, strerror (errno));
      return;
    }

  fputs ("# HTTP cookie file.\n", fp);
  fprintf (fp, "# Generated by Wget on %s.\n", datetime_str (cookies_now));
  fputs ("# Edit at your own risk.\n\n", fp);

  for (hash_table_iterate (jar->chains, &iter);
       hash_table_iter_next (&iter);
       )
    {
      const char *domain = iter.key;
      struct cookie *cookie = iter.value;
      for (; cookie; cookie = cookie->next)
        {
          if (!cookie->permanent && !opt.keep_session_cookies)
            continue;
          if (cookie_expired_p (cookie))
            continue;
          if (!cookie->domain_exact)
            fputc ('.', fp);
          fputs (domain, fp);
          if (cookie->port != PORT_ANY)
            fprintf (fp, ":%d", cookie->port);
          fprintf (fp, "\t%s\t%s\t%s\t%.0f\t%s\t%s\n",
                   cookie->domain_exact ? "FALSE" : "TRUE",
                   cookie->path, cookie->secure ? "TRUE" : "FALSE",
                   (double)cookie->expiry_time,
                   cookie->attr, cookie->value);
          if (ferror (fp))
            goto out;
        }
    }
 out:
  if (ferror (fp))
    logprintf (LOG_NOTQUIET, _("Error writing to `%s': %s\n"),
               file, strerror (errno));
  if (fclose (fp) < 0)
    logprintf (LOG_NOTQUIET, _("Error closing `%s': %s\n"),
               file, strerror (errno));

  DEBUGP (("Done saving cookies.\n"));
}

/* Clean up cookie-related data. */

void
cookie_jar_delete (struct cookie_jar *jar)
{
  /* Iterate over chains (indexed by domain) and free them. */
  hash_table_iterator iter;
  for (hash_table_iterate (jar->chains, &iter); hash_table_iter_next (&iter); )
    {
      struct cookie *chain = iter.value;
      xfree (iter.key);
      /* Then all cookies in this chain. */
      while (chain)
        {
          struct cookie *next = chain->next;
          delete_cookie (chain);
          chain = next;
        }
    }
  hash_table_destroy (jar->chains);
  xfree (jar);
}

/* Test cases.  Currently this is only tests parse_set_cookies.  To
   use, recompile Wget with -DTEST_COOKIES and call test_cookies()
   from main.  */

#ifdef TEST_COOKIES
void
test_cookies (void)
{
  /* Tests expected to succeed: */
  static struct {
    const char *data;
    const char *results[10];
  } tests_succ[] = {
    { "arg=value", {"arg", "value", NULL} },
    { "arg1=value1;arg2=value2", {"arg1", "value1", "arg2", "value2", NULL} },
    { "arg1=value1; arg2=value2", {"arg1", "value1", "arg2", "value2", NULL} },
    { "arg1=value1;  arg2=value2;", {"arg1", "value1", "arg2", "value2", NULL} },
    { "arg1=value1;  arg2=value2;  ", {"arg1", "value1", "arg2", "value2", NULL} },
    { "arg1=\"value1\"; arg2=\"\"", {"arg1", "value1", "arg2", "", NULL} },
    { "arg=", {"arg", "", NULL} },
    { "arg1=; arg2=", {"arg1", "", "arg2", "", NULL} },
    { "arg1 = ; arg2= ", {"arg1", "", "arg2", "", NULL} },
  };

  /* Tests expected to fail: */
  static char *tests_fail[] = {
    ";",
    "arg=\"unterminated",
    "=empty-name",
    "arg1=;=another-empty-name",
  };
  int i;

  for (i = 0; i < countof (tests_succ); i++)
    {
      int ind;
      const char *data = tests_succ[i].data;
      const char **expected = tests_succ[i].results;
      struct cookie *c;

      c = parse_set_cookie (data, true);
      if (!c)
        {
          printf ("NULL cookie returned for valid data: %s\n", data);
          continue;
        }

      /* Test whether extract_param handles these cases correctly. */
      {
        param_token name, value;
        const char *ptr = data;
        int j = 0;
        while (extract_param (&ptr, &name, &value, ';'))
          {
            char *n = strdupdelim (name.b, name.e);
            char *v = strdupdelim (value.b, value.e);
            if (!expected[j])
              {
                printf ("Too many parameters for '%s'\n", data);
                break;
              }
            if (0 != strcmp (expected[j], n))
              printf ("Invalid name %d for '%s' (expected '%s', got '%s')\n",
                      j / 2 + 1, data, expected[j], n);
            if (0 != strcmp (expected[j + 1], v))
              printf ("Invalid value %d for '%s' (expected '%s', got '%s')\n",
                      j / 2 + 1, data, expected[j + 1], v);
            j += 2;
            free (n);
            free (v);
          }
        if (expected[j])
          printf ("Too few parameters for '%s'\n", data);
      }
    }

  for (i = 0; i < countof (tests_fail); i++)
    {
      struct cookie *c;
      char *data = tests_fail[i];
      c = parse_set_cookie (data, true);
      if (c)
        printf ("Failed to report error on invalid data: %s\n", data);
    }
}
#endif /* TEST_COOKIES */