1 /* Handling of recursive HTTP retrieving.
2 Copyright (C) 1996, 1997, 1998, 1999, 2000, 2001, 2002, 2003,
3 2004, 2005, 2006, 2007, 2008 Free Software Foundation, Inc.
5 This file is part of GNU Wget.
7 GNU Wget is free software; you can redistribute it and/or modify
8 it under the terms of the GNU General Public License as published by
9 the Free Software Foundation; either version 3 of the License, or
10 (at your option) any later version.
12 GNU Wget is distributed in the hope that it will be useful,
13 but WITHOUT ANY WARRANTY; without even the implied warranty of
14 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 GNU General Public License for more details.
17 You should have received a copy of the GNU General Public License
18 along with Wget. If not, see <http://www.gnu.org/licenses/>.
20 Additional permission under GNU GPL version 3 section 7
22 If you modify this program, or any covered work, by linking or
23 combining it with the OpenSSL project's OpenSSL library (or a
24 modified version of that library), containing parts covered by the
25 terms of the OpenSSL or SSLeay licenses, the Free Software Foundation
26 grants you additional permission to convey the resulting work.
27 Corresponding Source for a non-source form of such a combination
28 shall include the source code for the parts of OpenSSL used as well
29 as that of the covered work. */
38 #endif /* HAVE_UNISTD_H */
55 /* Functions for maintaining the URL queue. */
57 struct queue_element {
58 const char *url; /* the URL to download */
59 const char *referer; /* the referring document */
60 int depth; /* the depth */
61 bool html_allowed; /* whether the document is allowed to
62 be treated as HTML. */
63 bool css_allowed; /* whether the document is allowed to
65 struct queue_element *next; /* next element in queue */
69 struct queue_element *head;
70 struct queue_element *tail;
74 /* Create a URL queue. */
76 static struct url_queue *
79 struct url_queue *queue = xnew0 (struct url_queue);
83 /* Delete a URL queue. */
86 url_queue_delete (struct url_queue *queue)
91 /* Enqueue a URL in the queue. The queue is FIFO: the items will be
92 retrieved ("dequeued") from the queue in the order they were placed
96 url_enqueue (struct url_queue *queue,
97 const char *url, const char *referer, int depth,
98 bool html_allowed, bool css_allowed)
100 struct queue_element *qel = xnew (struct queue_element);
102 qel->referer = referer;
104 qel->html_allowed = html_allowed;
105 qel->css_allowed = css_allowed;
109 if (queue->count > queue->maxcount)
110 queue->maxcount = queue->count;
112 DEBUGP (("Enqueuing %s at depth %d\n", url, depth));
113 DEBUGP (("Queue count %d, maxcount %d.\n", queue->count, queue->maxcount));
116 queue->tail->next = qel;
120 queue->head = queue->tail;
123 /* Take a URL out of the queue. Return true if this operation
124 succeeded, or false if the queue is empty. */
127 url_dequeue (struct url_queue *queue,
128 const char **url, const char **referer, int *depth,
129 bool *html_allowed, bool *css_allowed)
131 struct queue_element *qel = queue->head;
136 queue->head = queue->head->next;
141 *referer = qel->referer;
143 *html_allowed = qel->html_allowed;
144 *css_allowed = qel->css_allowed;
148 DEBUGP (("Dequeuing %s at depth %d\n", qel->url, qel->depth));
149 DEBUGP (("Queue count %d, maxcount %d.\n", queue->count, queue->maxcount));
155 static bool download_child_p (const struct urlpos *, struct url *, int,
156 struct url *, struct hash_table *);
157 static bool descend_redirect_p (const char *, struct url *, int,
158 struct url *, struct hash_table *);
161 /* Retrieve a part of the web beginning with START_URL. This used to
162 be called "recursive retrieval", because the old function was
163 recursive and implemented depth-first search. retrieve_tree on the
164 other hand implements breadth-search traversal of the tree, which
165 results in much nicer ordering of downloads.
167 The algorithm this function uses is simple:
169 1. put START_URL in the queue.
170 2. while there are URLs in the queue:
172 3. get next URL from the queue.
174 5. if the URL is HTML and its depth does not exceed maximum depth,
175 get the list of URLs embedded therein.
176 6. for each of those URLs do the following:
178 7. if the URL is not one of those downloaded before, and if it
179 satisfies the criteria specified by the various command-line
180 options, add it to the queue. */
183 retrieve_tree (struct url *start_url_parsed)
185 uerr_t status = RETROK;
187 /* The queue of URLs we need to load. */
188 struct url_queue *queue;
190 /* The URLs we do not wish to enqueue, because they are already in
191 the queue, but haven't been downloaded yet. */
192 struct hash_table *blacklist;
194 queue = url_queue_new ();
195 blacklist = make_string_hash_table (0);
197 /* Enqueue the starting URL. Use start_url_parsed->url rather than
198 just URL so we enqueue the canonical form of the URL. */
199 url_enqueue (queue, xstrdup (start_url_parsed->url), NULL, 0, true, false);
200 string_set_add (blacklist, start_url_parsed->url);
204 bool descend = false;
205 char *url, *referer, *file = NULL;
207 bool html_allowed, css_allowed;
209 bool dash_p_leaf_HTML = false;
211 if (opt.quota && total_downloaded_bytes > opt.quota)
213 if (status == FWRITEERR)
216 /* Get the next URL from the queue... */
218 if (!url_dequeue (queue,
219 (const char **)&url, (const char **)&referer,
220 &depth, &html_allowed, &css_allowed))
223 /* ...and download it. Note that this download is in most cases
224 unconditional, as download_child_p already makes sure a file
225 doesn't get enqueued twice -- and yet this check is here, and
226 not in download_child_p. This is so that if you run `wget -r
227 URL1 URL2', and a random URL is encountered once under URL1
228 and again under URL2, but at a different (possibly smaller)
229 depth, we want the URL's children to be taken into account
231 if (dl_url_file_map && hash_table_contains (dl_url_file_map, url))
233 file = xstrdup (hash_table_get (dl_url_file_map, url));
235 DEBUGP (("Already downloaded \"%s\", reusing it from \"%s\".\n",
238 /* this sucks, needs to be combined! */
240 && downloaded_html_set
241 && string_set_contains (downloaded_html_set, file))
247 && downloaded_css_set
248 && string_set_contains (downloaded_css_set, file))
257 char *redirected = NULL;
258 struct url *url_parsed = url_parse (url, &url_err);
262 char *error = url_error (url, url_err);
263 logprintf (LOG_NOTQUIET, "%s: %s.\n", url, error);
269 status = retrieve_url (url_parsed, url, &file, &redirected,
270 referer, &dt, false);
273 if (html_allowed && file && status == RETROK
274 && (dt & RETROKF) && (dt & TEXTHTML))
280 /* a little different, css_allowed can override content type
281 lots of web servers serve css with an incorrect content type
283 if (file && status == RETROK
285 ((dt & TEXTCSS) || css_allowed))
293 /* We have been redirected, possibly to another host, or
294 different path, or wherever. Check whether we really
295 want to follow it. */
298 if (!descend_redirect_p (redirected, url_parsed, depth,
299 start_url_parsed, blacklist))
302 /* Make sure that the old pre-redirect form gets
304 string_set_add (blacklist, url);
310 url_free(url_parsed);
315 visited_url (url, referer);
319 && depth >= opt.reclevel && opt.reclevel != INFINITE_RECURSION)
321 if (opt.page_requisites
322 && (depth == opt.reclevel || depth == opt.reclevel + 1))
324 /* When -p is specified, we are allowed to exceed the
325 maximum depth, but only for the "inline" links,
326 i.e. those that are needed to display the page.
327 Originally this could exceed the depth at most by
328 one, but we allow one more level so that the leaf
329 pages that contain frames can be loaded
331 dash_p_leaf_HTML = true;
335 /* Either -p wasn't specified or it was and we've
336 already spent the two extra (pseudo-)levels that it
337 affords us, so we need to bail out. */
338 DEBUGP (("Not descending further; at depth %d, max. %d.\n",
339 depth, opt.reclevel));
344 /* If the downloaded document was HTML or CSS, parse it and enqueue the
345 links it contains. */
349 bool meta_disallow_follow = false;
350 struct urlpos *children
351 = is_css ? get_urls_css_file (file, url) :
352 get_urls_html (file, url, &meta_disallow_follow);
354 if (opt.use_robots && meta_disallow_follow)
356 free_urlpos (children);
362 struct urlpos *child = children;
363 struct url *url_parsed = url_parsed = url_parse (url, NULL);
364 char *referer_url = url;
365 bool strip_auth = (url_parsed != NULL
366 && url_parsed->user != NULL);
367 assert (url_parsed != NULL);
369 /* Strip auth info if present */
371 referer_url = url_string (url_parsed, URL_AUTH_HIDE);
373 for (; child; child = child->next)
375 if (child->ignore_when_downloading)
377 if (dash_p_leaf_HTML && !child->link_inline_p)
379 if (download_child_p (child, url_parsed, depth, start_url_parsed,
382 url_enqueue (queue, xstrdup (child->url->url),
383 xstrdup (referer_url), depth + 1,
384 child->link_expect_html,
385 child->link_expect_css);
386 /* We blacklist the URL we have enqueued, because we
387 don't want to enqueue (and hence download) the
389 string_set_add (blacklist, child->url->url);
395 url_free (url_parsed);
396 free_urlpos (children);
402 || opt.spider /* opt.recursive is implicitely true */
403 || !acceptable (file)))
405 /* Either --delete-after was specified, or we loaded this
406 (otherwise unneeded because of --spider or rejected by -R)
407 HTML file just to harvest its hyperlinks -- in either case,
408 delete the local file. */
409 DEBUGP (("Removing file due to %s in recursive_retrieve():\n",
410 opt.delete_after ? "--delete-after" :
411 (opt.spider ? "--spider" :
412 "recursive rejection criteria")));
413 logprintf (LOG_VERBOSE,
414 (opt.delete_after || opt.spider
415 ? _("Removing %s.\n")
416 : _("Removing %s since it should be rejected.\n")),
419 logprintf (LOG_NOTQUIET, "unlink: %s\n", strerror (errno));
420 logputs (LOG_VERBOSE, "\n");
421 register_delete_file (file);
425 xfree_null (referer);
429 /* If anything is left of the queue due to a premature exit, free it
435 while (url_dequeue (queue,
436 (const char **)&d1, (const char **)&d2, &d3, &d4, &d5))
442 url_queue_delete (queue);
444 string_set_free (blacklist);
446 if (opt.quota && total_downloaded_bytes > opt.quota)
448 else if (status == FWRITEERR)
454 /* Based on the context provided by retrieve_tree, decide whether a
455 URL is to be descended to. This is only ever called from
456 retrieve_tree, but is in a separate function for clarity.
458 The most expensive checks (such as those for robots) are memoized
459 by storing these URLs to BLACKLIST. This may or may not help. It
460 will help if those URLs are encountered many times. */
463 download_child_p (const struct urlpos *upos, struct url *parent, int depth,
464 struct url *start_url_parsed, struct hash_table *blacklist)
466 struct url *u = upos->url;
467 const char *url = u->url;
468 bool u_scheme_like_http;
470 DEBUGP (("Deciding whether to enqueue \"%s\".\n", url));
472 if (string_set_contains (blacklist, url))
476 char *referrer = url_string (parent, URL_AUTH_HIDE_PASSWD);
477 DEBUGP (("download_child_p: parent->url is: %s\n", quote (parent->url)));
478 visited_url (url, referrer);
481 DEBUGP (("Already on the black list.\n"));
485 /* Several things to check for:
486 1. if scheme is not http, and we don't load it
487 2. check for relative links (if relative_only is set)
489 4. check for no-parent
490 5. check for excludes && includes
492 7. check for same host (if spanhost is unset), with possible
493 gethostbyname baggage
494 8. check for robots.txt
496 Addendum: If the URL is FTP, and it is to be loaded, only the
497 domain and suffix settings are "stronger".
499 Note that .html files will get loaded regardless of suffix rules
500 (but that is remedied later with unlink) unless the depth equals
503 More time- and memory- consuming tests should be put later on
506 /* Determine whether URL under consideration has a HTTP-like scheme. */
507 u_scheme_like_http = schemes_are_similar_p (u->scheme, SCHEME_HTTP);
509 /* 1. Schemes other than HTTP are normally not recursed into. */
510 if (!u_scheme_like_http && !(u->scheme == SCHEME_FTP && opt.follow_ftp))
512 DEBUGP (("Not following non-HTTP schemes.\n"));
516 /* 2. If it is an absolute link and they are not followed, throw it
518 if (u_scheme_like_http)
519 if (opt.relative_only && !upos->link_relative_p)
521 DEBUGP (("It doesn't really look like a relative link.\n"));
525 /* 3. If its domain is not to be accepted/looked-up, chuck it
527 if (!accept_domain (u))
529 DEBUGP (("The domain was not accepted.\n"));
533 /* 4. Check for parent directory.
535 If we descended to a different host or changed the scheme, ignore
536 opt.no_parent. Also ignore it for documents needed to display
537 the parent page when in -p mode. */
539 && schemes_are_similar_p (u->scheme, start_url_parsed->scheme)
540 && 0 == strcasecmp (u->host, start_url_parsed->host)
541 && u->port == start_url_parsed->port
542 && !(opt.page_requisites && upos->link_inline_p))
544 if (!subdir_p (start_url_parsed->dir, u->dir))
546 DEBUGP (("Going to \"%s\" would escape \"%s\" with no_parent on.\n",
547 u->dir, start_url_parsed->dir));
552 /* 5. If the file does not match the acceptance list, or is on the
553 rejection list, chuck it out. The same goes for the directory
554 exclusion and inclusion lists. */
555 if (opt.includes || opt.excludes)
557 if (!accdir (u->dir))
559 DEBUGP (("%s (%s) is excluded/not-included.\n", url, u->dir));
564 /* 6. Check for acceptance/rejection rules. We ignore these rules
565 for directories (no file name to match) and for non-leaf HTMLs,
566 which can lead to other files that do need to be downloaded. (-p
567 automatically implies non-leaf because with -p we can, if
568 necesary, overstep the maximum depth to get the page requisites.) */
569 if (u->file[0] != '\0'
570 && !(has_html_suffix_p (u->file)
571 /* The exception only applies to non-leaf HTMLs (but -p
572 always implies non-leaf because we can overstep the
573 maximum depth to get the requisites): */
575 opt.reclevel == INFINITE_RECURSION
577 || depth < opt.reclevel - 1
578 /* -p, which implies non-leaf (see above) */
579 || opt.page_requisites)))
581 if (!acceptable (u->file))
583 DEBUGP (("%s (%s) does not match acc/rej rules.\n",
590 if (schemes_are_similar_p (u->scheme, parent->scheme))
591 if (!opt.spanhost && 0 != strcasecmp (parent->host, u->host))
593 DEBUGP (("This is not the same hostname as the parent's (%s and %s).\n",
594 u->host, parent->host));
599 if (opt.use_robots && u_scheme_like_http)
601 struct robot_specs *specs = res_get_specs (u->host, u->port);
605 if (res_retrieve_file (url, &rfile))
607 specs = res_parse_from_file (rfile);
609 /* Delete the robots.txt file if we chose to either delete the
610 files after downloading or we're just running a spider. */
611 if (opt.delete_after || opt.spider)
613 logprintf (LOG_VERBOSE, "Removing %s.\n", rfile);
615 logprintf (LOG_NOTQUIET, "unlink: %s\n",
623 /* If we cannot get real specs, at least produce
624 dummy ones so that we can register them and stop
625 trying to retrieve them. */
626 specs = res_parse ("", 0);
628 res_register_specs (u->host, u->port, specs);
631 /* Now that we have (or don't have) robots.txt specs, we can
632 check what they say. */
633 if (!res_match_path (specs, u->path))
635 DEBUGP (("Not following %s because robots.txt forbids it.\n", url));
636 string_set_add (blacklist, url);
641 /* The URL has passed all the tests. It can be placed in the
643 DEBUGP (("Decided to load it.\n"));
648 DEBUGP (("Decided NOT to load it.\n"));
653 /* This function determines whether we will consider downloading the
654 children of a URL whose download resulted in a redirection,
655 possibly to another host, etc. It is needed very rarely, and thus
656 it is merely a simple-minded wrapper around download_child_p. */
659 descend_redirect_p (const char *redirected, struct url *orig_parsed, int depth,
660 struct url *start_url_parsed, struct hash_table *blacklist)
662 struct url *new_parsed;
666 assert (orig_parsed != NULL);
668 new_parsed = url_parse (redirected, NULL);
669 assert (new_parsed != NULL);
671 upos = xnew0 (struct urlpos);
672 upos->url = new_parsed;
674 success = download_child_p (upos, orig_parsed, depth,
675 start_url_parsed, blacklist);
677 url_free (new_parsed);
681 DEBUGP (("Redirection \"%s\" failed the test.\n", redirected));
686 /* vim:set sts=2 sw=2 cino+={s: */