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 */
56 /* Functions for maintaining the URL queue. */
58 struct queue_element {
59 const char *url; /* the URL to download */
60 const char *referer; /* the referring document */
61 int depth; /* the depth */
62 bool html_allowed; /* whether the document is allowed to
63 be treated as HTML. */
64 struct iri *iri; /* sXXXav */
65 bool css_allowed; /* whether the document is allowed to
67 struct queue_element *next; /* next element in queue */
71 struct queue_element *head;
72 struct queue_element *tail;
76 /* Create a URL queue. */
78 static struct url_queue *
81 struct url_queue *queue = xnew0 (struct url_queue);
85 /* Delete a URL queue. */
88 url_queue_delete (struct url_queue *queue)
93 /* Enqueue a URL in the queue. The queue is FIFO: the items will be
94 retrieved ("dequeued") from the queue in the order they were placed
98 url_enqueue (struct url_queue *queue, struct iri *i,
99 const char *url, const char *referer, int depth,
100 bool html_allowed, bool css_allowed)
102 struct queue_element *qel = xnew (struct queue_element);
105 qel->referer = referer;
107 qel->html_allowed = html_allowed;
108 qel->css_allowed = css_allowed;
112 if (queue->count > queue->maxcount)
113 queue->maxcount = queue->count;
115 DEBUGP (("Enqueuing %s at depth %d\n", url, depth));
116 DEBUGP (("Queue count %d, maxcount %d.\n", queue->count, queue->maxcount));
119 DEBUGP (("[IRI Enqueuing %s with %s\n", quote (url),
120 quote (i->uri_encoding)));
123 queue->tail->next = qel;
127 queue->head = queue->tail;
130 /* Take a URL out of the queue. Return true if this operation
131 succeeded, or false if the queue is empty. */
134 url_dequeue (struct url_queue *queue, struct iri **i,
135 const char **url, const char **referer, int *depth,
136 bool *html_allowed, bool *css_allowed)
138 struct queue_element *qel = queue->head;
143 queue->head = queue->head->next;
149 *referer = qel->referer;
151 *html_allowed = qel->html_allowed;
152 *css_allowed = qel->css_allowed;
156 DEBUGP (("Dequeuing %s at depth %d\n", qel->url, qel->depth));
157 DEBUGP (("Queue count %d, maxcount %d.\n", queue->count, queue->maxcount));
163 static bool download_child_p (const struct urlpos *, struct url *, int,
164 struct url *, struct hash_table *, struct iri *);
165 static bool descend_redirect_p (const char *, const char *, int,
166 struct url *, struct hash_table *, struct iri *);
169 /* Retrieve a part of the web beginning with START_URL. This used to
170 be called "recursive retrieval", because the old function was
171 recursive and implemented depth-first search. retrieve_tree on the
172 other hand implements breadth-search traversal of the tree, which
173 results in much nicer ordering of downloads.
175 The algorithm this function uses is simple:
177 1. put START_URL in the queue.
178 2. while there are URLs in the queue:
180 3. get next URL from the queue.
182 5. if the URL is HTML and its depth does not exceed maximum depth,
183 get the list of URLs embedded therein.
184 6. for each of those URLs do the following:
186 7. if the URL is not one of those downloaded before, and if it
187 satisfies the criteria specified by the various command-line
188 options, add it to the queue. */
191 retrieve_tree (const char *start_url)
193 uerr_t status = RETROK;
195 /* The queue of URLs we need to load. */
196 struct url_queue *queue;
198 /* The URLs we do not wish to enqueue, because they are already in
199 the queue, but haven't been downloaded yet. */
200 struct hash_table *blacklist;
203 struct url *start_url_parsed;
204 struct iri *i = iri_new ();
205 set_uri_encoding (i, opt.locale);
207 start_url_parsed = url_parse (start_url, &up_error_code, i);
208 if (!start_url_parsed)
210 logprintf (LOG_NOTQUIET, "%s: %s.\n", start_url,
211 url_error (up_error_code));
215 queue = url_queue_new ();
216 blacklist = make_string_hash_table (0);
218 /* Enqueue the starting URL. Use start_url_parsed->url rather than
219 just URL so we enqueue the canonical form of the URL. */
220 url_enqueue (queue, i, xstrdup (start_url_parsed->url), NULL, 0, true,
222 string_set_add (blacklist, start_url_parsed->url);
226 bool descend = false;
227 char *url, *referer, *file = NULL;
229 bool html_allowed, css_allowed;
231 bool dash_p_leaf_HTML = false;
233 if (opt.quota && total_downloaded_bytes > opt.quota)
235 if (status == FWRITEERR)
238 /* Get the next URL from the queue... */
240 if (!url_dequeue (queue, (struct iri **) &i,
241 (const char **)&url, (const char **)&referer,
242 &depth, &html_allowed, &css_allowed))
245 /* ...and download it. Note that this download is in most cases
246 unconditional, as download_child_p already makes sure a file
247 doesn't get enqueued twice -- and yet this check is here, and
248 not in download_child_p. This is so that if you run `wget -r
249 URL1 URL2', and a random URL is encountered once under URL1
250 and again under URL2, but at a different (possibly smaller)
251 depth, we want the URL's children to be taken into account
253 if (dl_url_file_map && hash_table_contains (dl_url_file_map, url))
255 file = xstrdup (hash_table_get (dl_url_file_map, url));
257 DEBUGP (("Already downloaded \"%s\", reusing it from \"%s\".\n",
260 /* this sucks, needs to be combined! */
262 && downloaded_html_set
263 && string_set_contains (downloaded_html_set, file))
269 && downloaded_css_set
270 && string_set_contains (downloaded_css_set, file))
279 char *redirected = NULL;
281 status = retrieve_url (url, &file, &redirected, referer, &dt,
284 if (html_allowed && file && status == RETROK
285 && (dt & RETROKF) && (dt & TEXTHTML))
291 /* a little different, css_allowed can override content type
292 lots of web servers serve css with an incorrect content type
294 if (file && status == RETROK
296 ((dt & TEXTCSS) || css_allowed))
304 /* We have been redirected, possibly to another host, or
305 different path, or wherever. Check whether we really
306 want to follow it. */
309 if (!descend_redirect_p (redirected, url, depth,
310 start_url_parsed, blacklist, i))
313 /* Make sure that the old pre-redirect form gets
315 string_set_add (blacklist, url);
325 visited_url (url, referer);
329 && depth >= opt.reclevel && opt.reclevel != INFINITE_RECURSION)
331 if (opt.page_requisites
332 && (depth == opt.reclevel || depth == opt.reclevel + 1))
334 /* When -p is specified, we are allowed to exceed the
335 maximum depth, but only for the "inline" links,
336 i.e. those that are needed to display the page.
337 Originally this could exceed the depth at most by
338 one, but we allow one more level so that the leaf
339 pages that contain frames can be loaded
341 dash_p_leaf_HTML = true;
345 /* Either -p wasn't specified or it was and we've
346 already spent the two extra (pseudo-)levels that it
347 affords us, so we need to bail out. */
348 DEBUGP (("Not descending further; at depth %d, max. %d.\n",
349 depth, opt.reclevel));
354 /* If the downloaded document was HTML or CSS, parse it and enqueue the
355 links it contains. */
359 bool meta_disallow_follow = false;
360 struct urlpos *children
361 = is_css ? get_urls_css_file (file, url) :
362 get_urls_html (file, url, &meta_disallow_follow, i);
364 if (opt.use_robots && meta_disallow_follow)
366 free_urlpos (children);
372 struct urlpos *child = children;
373 struct url *url_parsed = url_parse (url, NULL, i);
375 char *referer_url = url;
376 bool strip_auth = (url_parsed != NULL
377 && url_parsed->user != NULL);
378 assert (url_parsed != NULL);
380 /* Strip auth info if present */
382 referer_url = url_string (url_parsed, URL_AUTH_HIDE);
384 for (; child; child = child->next)
386 if (child->ignore_when_downloading)
388 if (dash_p_leaf_HTML && !child->link_inline_p)
390 if (download_child_p (child, url_parsed, depth, start_url_parsed,
394 set_uri_encoding (ci, i->content_encoding);
395 url_enqueue (queue, ci, xstrdup (child->url->url),
396 xstrdup (referer_url), depth + 1,
397 child->link_expect_html,
398 child->link_expect_css);
399 /* We blacklist the URL we have enqueued, because we
400 don't want to enqueue (and hence download) the
402 string_set_add (blacklist, child->url->url);
408 url_free (url_parsed);
409 free_urlpos (children);
415 || opt.spider /* opt.recursive is implicitely true */
416 || !acceptable (file)))
418 /* Either --delete-after was specified, or we loaded this
419 (otherwise unneeded because of --spider or rejected by -R)
420 HTML file just to harvest its hyperlinks -- in either case,
421 delete the local file. */
422 DEBUGP (("Removing file due to %s in recursive_retrieve():\n",
423 opt.delete_after ? "--delete-after" :
424 (opt.spider ? "--spider" :
425 "recursive rejection criteria")));
426 logprintf (LOG_VERBOSE,
427 (opt.delete_after || opt.spider
428 ? _("Removing %s.\n")
429 : _("Removing %s since it should be rejected.\n")),
432 logprintf (LOG_NOTQUIET, "unlink: %s\n", strerror (errno));
433 logputs (LOG_VERBOSE, "\n");
434 register_delete_file (file);
438 xfree_null (referer);
443 /* If anything is left of the queue due to a premature exit, free it
450 while (url_dequeue (queue, (struct iri **)&d6,
451 (const char **)&d1, (const char **)&d2, &d3, &d4, &d5))
458 url_queue_delete (queue);
460 if (start_url_parsed)
461 url_free (start_url_parsed);
462 string_set_free (blacklist);
464 if (opt.quota && total_downloaded_bytes > opt.quota)
466 else if (status == FWRITEERR)
472 /* Based on the context provided by retrieve_tree, decide whether a
473 URL is to be descended to. This is only ever called from
474 retrieve_tree, but is in a separate function for clarity.
476 The most expensive checks (such as those for robots) are memoized
477 by storing these URLs to BLACKLIST. This may or may not help. It
478 will help if those URLs are encountered many times. */
481 download_child_p (const struct urlpos *upos, struct url *parent, int depth,
482 struct url *start_url_parsed, struct hash_table *blacklist,
485 struct url *u = upos->url;
486 const char *url = u->url;
487 bool u_scheme_like_http;
489 DEBUGP (("Deciding whether to enqueue \"%s\".\n", url));
491 if (string_set_contains (blacklist, url))
495 char *referrer = url_string (parent, URL_AUTH_HIDE_PASSWD);
496 DEBUGP (("download_child_p: parent->url is: %s\n", quote (parent->url)));
497 visited_url (url, referrer);
500 DEBUGP (("Already on the black list.\n"));
504 /* Several things to check for:
505 1. if scheme is not http, and we don't load it
506 2. check for relative links (if relative_only is set)
508 4. check for no-parent
509 5. check for excludes && includes
511 7. check for same host (if spanhost is unset), with possible
512 gethostbyname baggage
513 8. check for robots.txt
515 Addendum: If the URL is FTP, and it is to be loaded, only the
516 domain and suffix settings are "stronger".
518 Note that .html files will get loaded regardless of suffix rules
519 (but that is remedied later with unlink) unless the depth equals
522 More time- and memory- consuming tests should be put later on
525 /* Determine whether URL under consideration has a HTTP-like scheme. */
526 u_scheme_like_http = schemes_are_similar_p (u->scheme, SCHEME_HTTP);
528 /* 1. Schemes other than HTTP are normally not recursed into. */
529 if (!u_scheme_like_http && !(u->scheme == SCHEME_FTP && opt.follow_ftp))
531 DEBUGP (("Not following non-HTTP schemes.\n"));
535 /* 2. If it is an absolute link and they are not followed, throw it
537 if (u_scheme_like_http)
538 if (opt.relative_only && !upos->link_relative_p)
540 DEBUGP (("It doesn't really look like a relative link.\n"));
544 /* 3. If its domain is not to be accepted/looked-up, chuck it
546 if (!accept_domain (u))
548 DEBUGP (("The domain was not accepted.\n"));
552 /* 4. Check for parent directory.
554 If we descended to a different host or changed the scheme, ignore
555 opt.no_parent. Also ignore it for documents needed to display
556 the parent page when in -p mode. */
558 && schemes_are_similar_p (u->scheme, start_url_parsed->scheme)
559 && 0 == strcasecmp (u->host, start_url_parsed->host)
560 && u->port == start_url_parsed->port
561 && !(opt.page_requisites && upos->link_inline_p))
563 if (!subdir_p (start_url_parsed->dir, u->dir))
565 DEBUGP (("Going to \"%s\" would escape \"%s\" with no_parent on.\n",
566 u->dir, start_url_parsed->dir));
571 /* 5. If the file does not match the acceptance list, or is on the
572 rejection list, chuck it out. The same goes for the directory
573 exclusion and inclusion lists. */
574 if (opt.includes || opt.excludes)
576 if (!accdir (u->dir))
578 DEBUGP (("%s (%s) is excluded/not-included.\n", url, u->dir));
583 /* 6. Check for acceptance/rejection rules. We ignore these rules
584 for directories (no file name to match) and for non-leaf HTMLs,
585 which can lead to other files that do need to be downloaded. (-p
586 automatically implies non-leaf because with -p we can, if
587 necesary, overstep the maximum depth to get the page requisites.) */
588 if (u->file[0] != '\0'
589 && !(has_html_suffix_p (u->file)
590 /* The exception only applies to non-leaf HTMLs (but -p
591 always implies non-leaf because we can overstep the
592 maximum depth to get the requisites): */
594 opt.reclevel == INFINITE_RECURSION
596 || depth < opt.reclevel - 1
597 /* -p, which implies non-leaf (see above) */
598 || opt.page_requisites)))
600 if (!acceptable (u->file))
602 DEBUGP (("%s (%s) does not match acc/rej rules.\n",
609 if (schemes_are_similar_p (u->scheme, parent->scheme))
610 if (!opt.spanhost && 0 != strcasecmp (parent->host, u->host))
612 DEBUGP (("This is not the same hostname as the parent's (%s and %s).\n",
613 u->host, parent->host));
618 if (opt.use_robots && u_scheme_like_http)
620 struct robot_specs *specs = res_get_specs (u->host, u->port);
624 if (res_retrieve_file (url, &rfile, iri))
626 specs = res_parse_from_file (rfile);
628 /* Delete the robots.txt file if we chose to either delete the
629 files after downloading or we're just running a spider. */
630 if (opt.delete_after || opt.spider)
632 logprintf (LOG_VERBOSE, "Removing %s.\n", rfile);
634 logprintf (LOG_NOTQUIET, "unlink: %s\n",
642 /* If we cannot get real specs, at least produce
643 dummy ones so that we can register them and stop
644 trying to retrieve them. */
645 specs = res_parse ("", 0);
647 res_register_specs (u->host, u->port, specs);
650 /* Now that we have (or don't have) robots.txt specs, we can
651 check what they say. */
652 if (!res_match_path (specs, u->path))
654 DEBUGP (("Not following %s because robots.txt forbids it.\n", url));
655 string_set_add (blacklist, url);
660 /* The URL has passed all the tests. It can be placed in the
662 DEBUGP (("Decided to load it.\n"));
667 DEBUGP (("Decided NOT to load it.\n"));
672 /* This function determines whether we will consider downloading the
673 children of a URL whose download resulted in a redirection,
674 possibly to another host, etc. It is needed very rarely, and thus
675 it is merely a simple-minded wrapper around download_child_p. */
678 descend_redirect_p (const char *redirected, const char *original, int depth,
679 struct url *start_url_parsed, struct hash_table *blacklist,
682 struct url *orig_parsed, *new_parsed;
686 orig_parsed = url_parse (original, NULL, NULL);
687 assert (orig_parsed != NULL);
689 new_parsed = url_parse (redirected, NULL, NULL);
690 assert (new_parsed != NULL);
692 upos = xnew0 (struct urlpos);
693 upos->url = new_parsed;
695 success = download_child_p (upos, orig_parsed, depth,
696 start_url_parsed, blacklist, iri);
698 url_free (orig_parsed);
699 url_free (new_parsed);
703 DEBUGP (("Redirection \"%s\" failed the test.\n", redirected));
708 /* vim:set sts=2 sw=2 cino+={s: */