1 /* Handling of recursive HTTP retrieving.
2 Copyright (C) 1995, 1996, 1997, 2000 Free Software Foundation, Inc.
4 This file is part of Wget.
6 This program is free software; you can redistribute it and/or modify
7 it under the terms of the GNU General Public License as published by
8 the Free Software Foundation; either version 2 of the License, or
9 (at your option) any later version.
11 This program is distributed in the hope that it will be useful,
12 but WITHOUT ANY WARRANTY; without even the implied warranty of
13 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 GNU General Public License for more details.
16 You should have received a copy of the GNU General Public License
17 along with this program; if not, write to the Free Software
18 Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. */
28 #endif /* HAVE_STRING_H */
31 #endif /* HAVE_UNISTD_H */
34 #include <sys/types.h>
50 extern char *version_string;
52 #define ROBOTS_FILENAME "robots.txt"
54 static struct hash_table *dl_file_url_map;
55 static struct hash_table *dl_url_file_map;
57 /* List of HTML files downloaded in this Wget run. Used for link
58 conversion after Wget is done. */
59 static slist *downloaded_html_files;
61 /* List of undesirable-to-load URLs. */
62 static struct hash_table *undesirable_urls;
64 /* List of forbidden locations. */
65 static char **forbidden = NULL;
67 /* Current recursion depth. */
70 /* Base directory we're recursing from (used by no_parent). */
71 static char *base_dir;
73 /* The host name for which we last checked robots. */
74 static char *robots_host;
76 static int first_time = 1;
78 /* Construct the robots URL. */
79 static struct urlinfo *robots_url PARAMS ((const char *, const char *));
80 static uerr_t retrieve_robots PARAMS ((const char *, const char *));
81 static char **parse_robots PARAMS ((const char *));
82 static int robots_match PARAMS ((struct urlinfo *, char **));
85 /* Cleanup the data structures associated with recursive retrieving
86 (the variables above). */
88 recursive_cleanup (void)
92 string_set_free (undesirable_urls);
93 undesirable_urls = NULL;
97 free_keys_and_values (dl_file_url_map);
98 hash_table_destroy (dl_file_url_map);
99 dl_file_url_map = NULL;
103 free_keys_and_values (dl_url_file_map);
104 hash_table_destroy (dl_url_file_map);
105 dl_url_file_map = NULL;
107 undesirable_urls = NULL;
108 free_vec (forbidden);
110 slist_free (downloaded_html_files);
111 downloaded_html_files = NULL;
112 FREE_MAYBE (base_dir);
113 FREE_MAYBE (robots_host);
117 /* Reset FIRST_TIME to 1, so that some action can be taken in
118 recursive_retrieve(). */
120 recursive_reset (void)
125 /* The core of recursive retrieving. Endless recursion is avoided by
126 having all URLs stored to a linked list of URLs, which is checked
127 before loading any URL. That way no URL can get loaded twice.
129 The function also supports specification of maximum recursion depth
130 and a number of other goodies. */
132 recursive_retrieve (const char *file, const char *this_url)
134 char *constr, *filename, *newloc;
135 char *canon_this_url = NULL;
136 int dt, inl, dash_p_leaf_HTML = FALSE;
137 int meta_disallow_follow;
138 int this_url_ftp; /* See below the explanation */
140 struct urlinfo *rurl;
141 urlpos *url_list, *cur_url;
142 char *rfile; /* For robots */
145 assert (this_url != NULL);
146 assert (file != NULL);
147 /* If quota was exceeded earlier, bail out. */
148 if (downloaded_exceeds_quota ())
150 /* Cache the current URL in the list. */
153 /* These three operations need to be done only once per Wget
154 run. They should probably be at a different location. */
155 if (!undesirable_urls)
156 undesirable_urls = make_string_hash_table (0);
158 hash_table_clear (undesirable_urls);
159 string_set_add (undesirable_urls, this_url);
160 /* Enter this_url to the hash table, in original and "enhanced" form. */
162 err = parseurl (this_url, u, 0);
165 string_set_add (undesirable_urls, u->url);
167 base_dir = xstrdup (u->dir); /* Set the base dir. */
168 /* Set the canonical this_url to be sent as referer. This
169 problem exists only when running the first time. */
170 canon_this_url = xstrdup (u->url);
174 DEBUGP (("Double yuck! The *base* URL is broken.\n"));
186 if (opt.reclevel != INFINITE_RECURSION && depth > opt.reclevel)
187 /* We've exceeded the maximum recursion depth specified by the user. */
189 if (opt.page_requisites && depth <= opt.reclevel + 1)
190 /* When -p is specified, we can do one more partial recursion from the
191 "leaf nodes" on the HTML document tree. The recursion is partial in
192 that we won't traverse any <A> or <AREA> tags, nor any <LINK> tags
193 except for <LINK REL="stylesheet">. */
194 dash_p_leaf_HTML = TRUE;
196 /* Either -p wasn't specified or it was and we've already gone the one
197 extra (pseudo-)level that it affords us, so we need to bail out. */
199 DEBUGP (("Recursion depth %d exceeded max. depth %d.\n",
200 depth, opt.reclevel));
206 /* Determine whether this_url is an FTP URL. If it is, it means
207 that the retrieval is done through proxy. In that case, FTP
208 links will be followed by default and recursion will not be
209 turned off when following them. */
210 this_url_ftp = (urlproto (this_url) == URLFTP);
212 /* Get the URL-s from an HTML file: */
213 url_list = get_urls_html (file, canon_this_url ? canon_this_url : this_url,
214 dash_p_leaf_HTML, &meta_disallow_follow);
216 if (opt.use_robots && meta_disallow_follow)
218 /* The META tag says we are not to follow this file. Respect
220 free_urlpos (url_list);
224 /* Decide what to do with each of the URLs. A URL will be loaded if
225 it meets several requirements, discussed later. */
226 for (cur_url = url_list; cur_url; cur_url = cur_url->next)
228 /* If quota was exceeded earlier, bail out. */
229 if (downloaded_exceeds_quota ())
231 /* Parse the URL for convenient use in other functions, as well
232 as to get the optimized form. It also checks URL integrity. */
234 if (parseurl (cur_url->url, u, 0) != URLOK)
236 DEBUGP (("Yuck! A bad URL.\n"));
240 if (u->proto == URLFILE)
242 DEBUGP (("Nothing to do with file:// around here.\n"));
246 assert (u->url != NULL);
247 constr = xstrdup (u->url);
249 /* Several checkings whether a file is acceptable to load:
250 1. check if URL is ftp, and we don't load it
251 2. check for relative links (if relative_only is set)
253 4. check for no-parent
254 5. check for excludes && includes
256 7. check for same host (if spanhost is unset), with possible
257 gethostbyname baggage
258 8. check for robots.txt
260 Addendum: If the URL is FTP, and it is to be loaded, only the
261 domain and suffix settings are "stronger".
263 Note that .html and (yuck) .htm will get loaded regardless of
264 suffix rules (but that is remedied later with unlink) unless
265 the depth equals the maximum depth.
267 More time- and memory- consuming tests should be put later on
270 /* inl is set if the URL we are working on (constr) is stored in
271 undesirable_urls. Using it is crucial to avoid unnecessary
272 repeated continuous hits to the hash table. */
273 inl = string_set_exists (undesirable_urls, constr);
275 /* If it is FTP, and FTP is not followed, chuck it out. */
277 if (u->proto == URLFTP && !opt.follow_ftp && !this_url_ftp)
279 DEBUGP (("Uh, it is FTP but i'm not in the mood to follow FTP.\n"));
280 string_set_add (undesirable_urls, constr);
283 /* If it is absolute link and they are not followed, chuck it
285 if (!inl && u->proto != URLFTP)
286 if (opt.relative_only && !cur_url->link_relative_p)
288 DEBUGP (("It doesn't really look like a relative link.\n"));
289 string_set_add (undesirable_urls, constr);
292 /* If its domain is not to be accepted/looked-up, chuck it out. */
294 if (!accept_domain (u))
296 DEBUGP (("I don't like the smell of that domain.\n"));
297 string_set_add (undesirable_urls, constr);
300 /* Check for parent directory. */
301 if (!inl && opt.no_parent
302 /* If the new URL is FTP and the old was not, ignore
304 && !(!this_url_ftp && u->proto == URLFTP))
306 /* Check for base_dir first. */
307 if (!(base_dir && frontcmp (base_dir, u->dir)))
309 /* Failing that, check for parent dir. */
310 struct urlinfo *ut = newurl ();
311 if (parseurl (this_url, ut, 0) != URLOK)
312 DEBUGP (("Double yuck! The *base* URL is broken.\n"));
313 else if (!frontcmp (ut->dir, u->dir))
315 /* Failing that too, kill the URL. */
316 DEBUGP (("Trying to escape parental guidance with no_parent on.\n"));
317 string_set_add (undesirable_urls, constr);
323 /* If the file does not match the acceptance list, or is on the
324 rejection list, chuck it out. The same goes for the
325 directory exclude- and include- lists. */
326 if (!inl && (opt.includes || opt.excludes))
328 if (!accdir (u->dir, ALLABS))
330 DEBUGP (("%s (%s) is excluded/not-included.\n", constr, u->dir));
331 string_set_add (undesirable_urls, constr);
338 /* We check for acceptance/rejection rules only for non-HTML
339 documents. Since we don't know whether they really are
340 HTML, it will be deduced from (an OR-ed list):
342 1) u->file is "" (meaning it is a directory)
343 2) suffix exists, AND:
347 If the file *is* supposed to be HTML, it will *not* be
348 subject to acc/rej rules, unless a finite maximum depth has
349 been specified and the current depth is the maximum depth. */
352 || (((suf = suffix (constr)) != NULL)
353 && ((!strcmp (suf, "html") || !strcmp (suf, "htm"))
354 && ((opt.reclevel != INFINITE_RECURSION) &&
355 (depth != opt.reclevel))))))
357 if (!acceptable (u->file))
359 DEBUGP (("%s (%s) does not match acc/rej rules.\n",
361 string_set_add (undesirable_urls, constr);
367 /* Optimize the URL (which includes possible DNS lookup) only
368 after all other possibilities have been exhausted. */
371 if (!opt.simple_check)
376 /* Just lowercase the hostname. */
377 for (p = u->host; *p; p++)
380 u->url = str_url (u, 0);
383 constr = xstrdup (u->url);
384 string_set_add (undesirable_urls, constr);
385 if (!inl && !((u->proto == URLFTP) && !this_url_ftp))
386 if (!opt.spanhost && this_url && !same_host (this_url, constr))
388 DEBUGP (("This is not the same hostname as the parent's.\n"));
389 string_set_add (undesirable_urls, constr);
393 /* What about robots.txt? */
394 if (!inl && opt.use_robots && u->proto == URLHTTP)
396 /* Since Wget knows about only one set of robot rules at a
397 time, /robots.txt must be reloaded whenever a new host is
400 robots_host holds the host the current `forbid' variable
402 if (!robots_host || !same_host (robots_host, u->host))
404 FREE_MAYBE (robots_host);
405 /* Now make robots_host the new host, no matter what the
406 result will be. So if there is no /robots.txt on the
407 site, Wget will not retry getting robots all the
409 robots_host = xstrdup (u->host);
410 free_vec (forbidden);
412 err = retrieve_robots (constr, ROBOTS_FILENAME);
415 rurl = robots_url (constr, ROBOTS_FILENAME);
416 rfile = url_filename (rurl);
417 forbidden = parse_robots (rfile);
423 /* Now that we have (or don't have) robots, we can check for
425 if (!robots_match (u, forbidden))
427 DEBUGP (("Stuffing %s because %s forbids it.\n", this_url,
429 string_set_add (undesirable_urls, constr);
435 /* If it wasn't chucked out, do something with it. */
438 DEBUGP (("I've decided to load it -> "));
439 /* Add it to the list of already-loaded URL-s. */
440 string_set_add (undesirable_urls, constr);
441 /* Automatically followed FTPs will *not* be downloaded
443 if (u->proto == URLFTP)
445 /* Don't you adore side-effects? */
448 /* Reset its type. */
451 retrieve_url (constr, &filename, &newloc,
452 canon_this_url ? canon_this_url : this_url, &dt);
453 if (u->proto == URLFTP)
463 /* If there was no error, and the type is text/html, parse
468 recursive_retrieve (filename, constr);
471 DEBUGP (("%s is not text/html so we don't chase.\n",
472 filename ? filename: "(null)"));
474 if (opt.delete_after || (filename && !acceptable (filename)))
475 /* Either --delete-after was specified, or we loaded this otherwise
476 rejected (e.g. by -R) HTML file just so we could harvest its
477 hyperlinks -- in either case, delete the local file. */
479 DEBUGP (("Removing file due to %s in recursive_retrieve():\n",
480 opt.delete_after ? "--delete-after" :
481 "recursive rejection criteria"));
482 logprintf (LOG_VERBOSE,
483 (opt.delete_after ? _("Removing %s.\n")
484 : _("Removing %s since it should be rejected.\n")),
486 if (unlink (filename))
487 logprintf (LOG_NOTQUIET, "unlink: %s\n", strerror (errno));
491 /* If everything was OK, and links are to be converted, let's
492 store the local filename. */
493 if (opt.convert_links && (dt & RETROKF) && (filename != NULL))
495 cur_url->convert = CO_CONVERT_TO_RELATIVE;
496 cur_url->local_name = xstrdup (filename);
500 DEBUGP (("%s already in list, so we don't load.\n", constr));
501 /* Free filename and constr. */
502 FREE_MAYBE (filename);
505 /* Increment the pbuf for the appropriate size. */
507 if (opt.convert_links && !opt.delete_after)
508 /* This is merely the first pass: the links that have been
509 successfully downloaded are converted. In the second pass,
510 convert_all_links() will also convert those links that have NOT
511 been downloaded to their canonical form. */
512 convert_links (file, url_list);
513 /* Free the linked list of URL-s. */
514 free_urlpos (url_list);
515 /* Free the canonical this_url. */
516 FREE_MAYBE (canon_this_url);
517 /* Decrement the recursion depth. */
519 if (downloaded_exceeds_quota ())
526 register_download (const char *url, const char *file)
528 if (!opt.convert_links)
530 if (!dl_file_url_map)
531 dl_file_url_map = make_string_hash_table (0);
532 hash_table_put (dl_file_url_map, xstrdup (file), xstrdup (url));
533 if (!dl_url_file_map)
534 dl_url_file_map = make_string_hash_table (0);
535 hash_table_put (dl_url_file_map, xstrdup (url), xstrdup (file));
539 register_html (const char *url, const char *file)
541 if (!opt.convert_links)
543 downloaded_html_files = slist_prepend (downloaded_html_files, file);
546 /* convert_links() is called from recursive_retrieve() after we're
547 done with an HTML file. This call to convert_links is not complete
548 because it converts only the downloaded files, and Wget cannot know
549 which files will be downloaded afterwards. So, if we have file
552 <a href="/c/something.gif">
554 and /c/something.gif was not downloaded because it exceeded the
555 recursion depth, the reference will *not* be changed.
557 However, later we can encounter /c/something.gif from an "upper"
558 level HTML (let's call it filetwo.html), and it gets downloaded.
560 But now we have a problem because /c/something.gif will be
561 correctly transformed in filetwo.html, but not in fileone.html,
562 since Wget could not have known that /c/something.gif will be
563 downloaded in the future.
565 This is why Wget must, after the whole retrieval, call
566 convert_all_links to go once more through the entire list of
567 retrieved HTMLs, and re-convert them.
569 All the downloaded HTMLs are kept in downloaded_html_files, and downloaded URLs
570 in urls_downloaded. From these two lists information is
573 convert_all_links (void)
577 /* Destructively reverse downloaded_html_files to get it in the right order.
578 recursive_retrieve() used slist_prepend() consistently. */
579 downloaded_html_files = slist_nreverse (downloaded_html_files);
581 for (html = downloaded_html_files; html; html = html->next)
583 urlpos *urls, *cur_url;
586 DEBUGP (("Rescanning %s\n", html->string));
587 /* Determine the URL of the HTML file. get_urls_html will need
589 url = hash_table_get (dl_file_url_map, html->string);
591 DEBUGP (("It should correspond to %s.\n", url));
593 DEBUGP (("I cannot find the corresponding URL.\n"));
594 /* Parse the HTML file... */
595 urls = get_urls_html (html->string, url, FALSE, NULL);
596 /* We don't respect meta_disallow_follow here because, even if
597 the file is not followed, we might still want to convert the
598 links that have been followed from other files. */
599 for (cur_url = urls; cur_url; cur_url = cur_url->next)
603 /* The URL must be in canonical form to be compared. */
604 struct urlinfo *u = newurl ();
605 uerr_t res = parseurl (cur_url->url, u, 0);
611 /* We decide the direction of conversion according to whether
612 a URL was downloaded. Downloaded URLs will be converted
613 ABS2REL, whereas non-downloaded will be converted REL2ABS. */
614 local_name = hash_table_get (dl_url_file_map, u->url);
616 DEBUGP (("%s marked for conversion, local %s\n",
617 u->url, local_name));
618 /* Decide on the conversion direction. */
621 /* We've downloaded this URL. Convert it to relative
622 form. We do this even if the URL already is in
623 relative form, because our directory structure may
624 not be identical to that on the server (think `-nd',
625 `--cut-dirs', etc.) */
626 cur_url->convert = CO_CONVERT_TO_RELATIVE;
627 cur_url->local_name = xstrdup (local_name);
631 /* We haven't downloaded this URL. If it's not already
632 complete (including a full host name), convert it to
633 that form, so it can be reached while browsing this
635 if (!cur_url->link_complete_p)
636 cur_url->convert = CO_CONVERT_TO_COMPLETE;
637 cur_url->local_name = NULL;
641 /* Convert the links in the file. */
642 convert_links (html->string, urls);
648 /* Robots support. */
650 /* Construct the robots URL. */
651 static struct urlinfo *
652 robots_url (const char *url, const char *robots_filename)
654 struct urlinfo *u = newurl ();
657 err = parseurl (url, u, 0);
658 assert (err == URLOK && u->proto == URLHTTP);
662 u->dir = xstrdup ("");
663 u->file = xstrdup (robots_filename);
664 u->url = str_url (u, 0);
668 /* Retrieves the robots_filename from the root server directory, if
669 possible. Returns ROBOTSOK if robots were retrieved OK, and
670 NOROBOTS if robots could not be retrieved for any reason. */
672 retrieve_robots (const char *url, const char *robots_filename)
678 u = robots_url (url, robots_filename);
679 logputs (LOG_VERBOSE, _("Loading robots.txt; please ignore errors.\n"));
680 err = retrieve_url (u->url, NULL, NULL, NULL, &dt);
688 /* Parse the robots_filename and return the disallowed path components
689 in a malloc-ed vector of character pointers.
691 It should be fully compliant with the syntax as described in the
692 file norobots.txt, adopted by the robots mailing list
693 (robots@webcrawler.com). */
695 parse_robots (const char *robots_filename)
699 char *line, *cmd, *str, *p;
700 char *base_version, *version;
702 int wget_matched; /* is the part meant for Wget? */
707 fp = fopen (robots_filename, "rb");
711 /* Kill version number. */
714 STRDUP_ALLOCA (base_version, opt.useragent);
715 STRDUP_ALLOCA (version, opt.useragent);
719 int len = 10 + strlen (version_string);
720 base_version = (char *)alloca (len);
721 sprintf (base_version, "Wget/%s", version_string);
722 version = (char *)alloca (len);
723 sprintf (version, "Wget/%s", version_string);
725 for (p = version; *p; p++)
727 for (p = base_version; *p && *p != '/'; p++)
731 /* Setting this to 1 means that Wget considers itself under
732 restrictions by default, even if the User-Agent field is not
733 present. However, if it finds the user-agent set to anything
734 other than Wget, the rest will be ignored (up to the following
735 User-Agent field). Thus you may have something like:
739 User-Agent: stupid-robot
748 In this case the 1, 2, 5, 6 and 7 disallow lines will be
751 while ((line = read_whole_line (fp)))
753 int len = strlen (line);
754 /* Destroy <CR><LF> if present. */
755 if (len && line[len - 1] == '\n')
757 if (len && line[len - 1] == '\r')
759 /* According to specifications, optional space may be at the
761 DEBUGP (("Line: %s\n", line));
763 for (cmd = line; *cmd && ISSPACE (*cmd); cmd++);
767 DEBUGP (("(chucked out)\n"));
771 for (str = cmd; *str && *str != ':'; str++);
775 DEBUGP (("(chucked out)\n"));
778 /* Zero-terminate the command. */
780 /* Look for the string beginning... */
781 for (; *str && ISSPACE (*str); str++);
782 /* Look for comments or trailing spaces and kill them off. */
783 for (p = str; *p; p++)
784 if (*p && ISSPACE (*p) && ((*(p + 1) == '#') || (*(p + 1) == '\0')))
786 /* We have found either a shell-style comment `<sp>+#' or some
787 trailing spaces. Now rewind to the beginning of the spaces
788 and place '\0' there. */
789 while (p > str && ISSPACE (*p))
797 if (!strcasecmp (cmd, "User-agent"))
800 /* Lowercase the agent string. */
801 for (p = str; *p; p++)
803 /* If the string is `*', it matches. */
804 if (*str == '*' && !*(str + 1))
808 /* If the string contains wildcards, we'll run it through
810 if (has_wildcards_p (str))
812 /* If the string contains '/', compare with the full
813 version. Else, compare it to base_version. */
814 if (strchr (str, '/'))
815 match = !fnmatch (str, version, 0);
817 match = !fnmatch (str, base_version, 0);
819 else /* Substring search */
821 if (strstr (version, str))
827 /* If Wget is not matched, skip all the entries up to the
828 next User-agent field. */
829 wget_matched = match;
831 else if (!wget_matched)
834 DEBUGP (("(chucking out since it is not applicable for Wget)\n"));
837 else if (!strcasecmp (cmd, "Disallow"))
839 /* If "Disallow" is empty, the robot is welcome. */
843 entries = (char **)xmalloc (sizeof (char *));
849 entries = (char **)xrealloc (entries, (num + 2)* sizeof (char *));
850 entries[num] = xstrdup (str);
851 entries[++num] = NULL;
852 /* Strip trailing spaces, according to specifications. */
853 for (i = strlen (str); i >= 0 && ISSPACE (str[i]); i--)
854 if (ISSPACE (str[i]))
860 /* unknown command */
861 DEBUGP (("(chucked out)\n"));
869 /* May the URL url be loaded according to disallowing rules stored in
872 robots_match (struct urlinfo *u, char **fb)
878 DEBUGP (("Matching %s against: ", u->path));
881 DEBUGP (("%s ", *fb));
883 /* If dir is fb, we may not load the file. */
884 if (strncmp (u->path, *fb, l) == 0)
886 DEBUGP (("matched.\n"));
887 return 0; /* Matches, i.e. does not load... */
890 DEBUGP (("not matched.\n"));