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 struct iri *iri; /* sXXXav */
64 bool css_allowed; /* whether the document is allowed to
66 struct queue_element *next; /* next element in queue */
70 struct queue_element *head;
71 struct queue_element *tail;
75 /* Create a URL queue. */
77 static struct url_queue *
80 struct url_queue *queue = xnew0 (struct url_queue);
84 /* Delete a URL queue. */
87 url_queue_delete (struct url_queue *queue)
92 /* Enqueue a URL in the queue. The queue is FIFO: the items will be
93 retrieved ("dequeued") from the queue in the order they were placed
97 url_enqueue (struct url_queue *queue, struct iri *i,
98 const char *url, const char *referer, int depth,
99 bool html_allowed, bool css_allowed)
101 struct queue_element *qel = xnew (struct queue_element);
104 qel->referer = referer;
106 qel->html_allowed = html_allowed;
107 qel->css_allowed = css_allowed;
111 if (queue->count > queue->maxcount)
112 queue->maxcount = queue->count;
114 DEBUGP (("Enqueuing %s at depth %d\n", url, depth));
115 DEBUGP (("Queue count %d, maxcount %d.\n", queue->count, queue->maxcount));
118 DEBUGP (("[IRI Enqueuing %s with %s\n", quote_n (0, url),
119 i->uri_encoding ? quote_n (1, i->uri_encoding) : "None"));
122 queue->tail->next = qel;
126 queue->head = queue->tail;
129 /* Take a URL out of the queue. Return true if this operation
130 succeeded, or false if the queue is empty. */
133 url_dequeue (struct url_queue *queue, struct iri **i,
134 const char **url, const char **referer, int *depth,
135 bool *html_allowed, bool *css_allowed)
137 struct queue_element *qel = queue->head;
142 queue->head = queue->head->next;
148 *referer = qel->referer;
150 *html_allowed = qel->html_allowed;
151 *css_allowed = qel->css_allowed;
155 DEBUGP (("Dequeuing %s at depth %d\n", qel->url, qel->depth));
156 DEBUGP (("Queue count %d, maxcount %d.\n", queue->count, queue->maxcount));
162 static bool download_child_p (const struct urlpos *, struct url *, int,
163 struct url *, struct hash_table *, struct iri *);
164 static bool descend_redirect_p (const char *, struct url *, int,
165 struct url *, struct hash_table *, struct iri *);
168 /* Retrieve a part of the web beginning with START_URL. This used to
169 be called "recursive retrieval", because the old function was
170 recursive and implemented depth-first search. retrieve_tree on the
171 other hand implements breadth-search traversal of the tree, which
172 results in much nicer ordering of downloads.
174 The algorithm this function uses is simple:
176 1. put START_URL in the queue.
177 2. while there are URLs in the queue:
179 3. get next URL from the queue.
181 5. if the URL is HTML and its depth does not exceed maximum depth,
182 get the list of URLs embedded therein.
183 6. for each of those URLs do the following:
185 7. if the URL is not one of those downloaded before, and if it
186 satisfies the criteria specified by the various command-line
187 options, add it to the queue. */
190 retrieve_tree (struct url *start_url_parsed, struct iri *pi)
192 uerr_t status = RETROK;
194 /* The queue of URLs we need to load. */
195 struct url_queue *queue;
197 /* The URLs we do not wish to enqueue, because they are already in
198 the queue, but haven't been downloaded yet. */
199 struct hash_table *blacklist;
202 struct iri *i = iri_new ();
204 #define COPYSTR(x) (x) ? xstrdup(x) : NULL;
205 /* Duplicate pi struct if not NULL */
208 i->uri_encoding = COPYSTR (pi->uri_encoding);
209 i->content_encoding = COPYSTR (pi->content_encoding);
210 i->utf8_encode = pi->utf8_encode;
213 set_uri_encoding (i, opt.locale, true);
216 queue = url_queue_new ();
217 blacklist = make_string_hash_table (0);
219 /* Enqueue the starting URL. Use start_url_parsed->url rather than
220 just URL so we enqueue the canonical form of the URL. */
221 url_enqueue (queue, i, xstrdup (start_url_parsed->url), NULL, 0, true,
223 string_set_add (blacklist, start_url_parsed->url);
227 bool descend = false;
228 char *url, *referer, *file = NULL;
230 bool html_allowed, css_allowed;
232 bool dash_p_leaf_HTML = false;
234 if (opt.quota && total_downloaded_bytes > opt.quota)
236 if (status == FWRITEERR)
239 /* Get the next URL from the queue... */
241 if (!url_dequeue (queue, (struct iri **) &i,
242 (const char **)&url, (const char **)&referer,
243 &depth, &html_allowed, &css_allowed))
246 /* ...and download it. Note that this download is in most cases
247 unconditional, as download_child_p already makes sure a file
248 doesn't get enqueued twice -- and yet this check is here, and
249 not in download_child_p. This is so that if you run `wget -r
250 URL1 URL2', and a random URL is encountered once under URL1
251 and again under URL2, but at a different (possibly smaller)
252 depth, we want the URL's children to be taken into account
254 if (dl_url_file_map && hash_table_contains (dl_url_file_map, url))
256 file = xstrdup (hash_table_get (dl_url_file_map, url));
258 DEBUGP (("Already downloaded \"%s\", reusing it from \"%s\".\n",
261 /* this sucks, needs to be combined! */
263 && downloaded_html_set
264 && string_set_contains (downloaded_html_set, file))
270 && downloaded_css_set
271 && string_set_contains (downloaded_css_set, file))
280 char *redirected = NULL;
281 struct url *url_parsed = url_parse (url, &url_err, i, true);
283 status = retrieve_url (url_parsed, url, &file, &redirected, referer,
286 if (html_allowed && file && status == RETROK
287 && (dt & RETROKF) && (dt & TEXTHTML))
293 /* a little different, css_allowed can override content type
294 lots of web servers serve css with an incorrect content type
296 if (file && status == RETROK
298 ((dt & TEXTCSS) || css_allowed))
306 /* We have been redirected, possibly to another host, or
307 different path, or wherever. Check whether we really
308 want to follow it. */
311 if (!descend_redirect_p (redirected, url_parsed, depth,
312 start_url_parsed, blacklist, i))
315 /* Make sure that the old pre-redirect form gets
317 string_set_add (blacklist, url);
323 url_free(url_parsed);
328 visited_url (url, referer);
332 && depth >= opt.reclevel && opt.reclevel != INFINITE_RECURSION)
334 if (opt.page_requisites
335 && (depth == opt.reclevel || depth == opt.reclevel + 1))
337 /* When -p is specified, we are allowed to exceed the
338 maximum depth, but only for the "inline" links,
339 i.e. those that are needed to display the page.
340 Originally this could exceed the depth at most by
341 one, but we allow one more level so that the leaf
342 pages that contain frames can be loaded
344 dash_p_leaf_HTML = true;
348 /* Either -p wasn't specified or it was and we've
349 already spent the two extra (pseudo-)levels that it
350 affords us, so we need to bail out. */
351 DEBUGP (("Not descending further; at depth %d, max. %d.\n",
352 depth, opt.reclevel));
357 /* If the downloaded document was HTML or CSS, parse it and enqueue the
358 links it contains. */
362 bool meta_disallow_follow = false;
363 struct urlpos *children
364 = is_css ? get_urls_css_file (file, url) :
365 get_urls_html (file, url, &meta_disallow_follow, i);
367 if (opt.use_robots && meta_disallow_follow)
369 free_urlpos (children);
375 struct urlpos *child = children;
376 struct url *url_parsed = url_parse (url, NULL, i, true);
378 char *referer_url = url;
379 bool strip_auth = (url_parsed != NULL
380 && url_parsed->user != NULL);
381 assert (url_parsed != NULL);
383 /* Strip auth info if present */
385 referer_url = url_string (url_parsed, URL_AUTH_HIDE);
387 for (; child; child = child->next)
389 if (child->ignore_when_downloading)
391 if (dash_p_leaf_HTML && !child->link_inline_p)
393 if (download_child_p (child, url_parsed, depth, start_url_parsed,
397 set_uri_encoding (ci, i->content_encoding, false);
398 url_enqueue (queue, ci, xstrdup (child->url->url),
399 xstrdup (referer_url), depth + 1,
400 child->link_expect_html,
401 child->link_expect_css);
402 /* We blacklist the URL we have enqueued, because we
403 don't want to enqueue (and hence download) the
405 string_set_add (blacklist, child->url->url);
411 url_free (url_parsed);
412 free_urlpos (children);
418 || opt.spider /* opt.recursive is implicitely true */
419 || !acceptable (file)))
421 /* Either --delete-after was specified, or we loaded this
422 (otherwise unneeded because of --spider or rejected by -R)
423 HTML file just to harvest its hyperlinks -- in either case,
424 delete the local file. */
425 DEBUGP (("Removing file due to %s in recursive_retrieve():\n",
426 opt.delete_after ? "--delete-after" :
427 (opt.spider ? "--spider" :
428 "recursive rejection criteria")));
429 logprintf (LOG_VERBOSE,
430 (opt.delete_after || opt.spider
431 ? _("Removing %s.\n")
432 : _("Removing %s since it should be rejected.\n")),
435 logprintf (LOG_NOTQUIET, "unlink: %s\n", strerror (errno));
436 logputs (LOG_VERBOSE, "\n");
437 register_delete_file (file);
441 xfree_null (referer);
446 /* If anything is left of the queue due to a premature exit, free it
453 while (url_dequeue (queue, (struct iri **)&d6,
454 (const char **)&d1, (const char **)&d2, &d3, &d4, &d5))
461 url_queue_delete (queue);
463 string_set_free (blacklist);
465 if (opt.quota && total_downloaded_bytes > opt.quota)
467 else if (status == FWRITEERR)
473 /* Based on the context provided by retrieve_tree, decide whether a
474 URL is to be descended to. This is only ever called from
475 retrieve_tree, but is in a separate function for clarity.
477 The most expensive checks (such as those for robots) are memoized
478 by storing these URLs to BLACKLIST. This may or may not help. It
479 will help if those URLs are encountered many times. */
482 download_child_p (const struct urlpos *upos, struct url *parent, int depth,
483 struct url *start_url_parsed, struct hash_table *blacklist,
486 struct url *u = upos->url;
487 const char *url = u->url;
488 bool u_scheme_like_http;
490 DEBUGP (("Deciding whether to enqueue \"%s\".\n", url));
492 if (string_set_contains (blacklist, url))
496 char *referrer = url_string (parent, URL_AUTH_HIDE_PASSWD);
497 DEBUGP (("download_child_p: parent->url is: %s\n", quote (parent->url)));
498 visited_url (url, referrer);
501 DEBUGP (("Already on the black list.\n"));
505 /* Several things to check for:
506 1. if scheme is not http, and we don't load it
507 2. check for relative links (if relative_only is set)
509 4. check for no-parent
510 5. check for excludes && includes
512 7. check for same host (if spanhost is unset), with possible
513 gethostbyname baggage
514 8. check for robots.txt
516 Addendum: If the URL is FTP, and it is to be loaded, only the
517 domain and suffix settings are "stronger".
519 Note that .html files will get loaded regardless of suffix rules
520 (but that is remedied later with unlink) unless the depth equals
523 More time- and memory- consuming tests should be put later on
526 /* Determine whether URL under consideration has a HTTP-like scheme. */
527 u_scheme_like_http = schemes_are_similar_p (u->scheme, SCHEME_HTTP);
529 /* 1. Schemes other than HTTP are normally not recursed into. */
530 if (!u_scheme_like_http && !(u->scheme == SCHEME_FTP && opt.follow_ftp))
532 DEBUGP (("Not following non-HTTP schemes.\n"));
536 /* 2. If it is an absolute link and they are not followed, throw it
538 if (u_scheme_like_http)
539 if (opt.relative_only && !upos->link_relative_p)
541 DEBUGP (("It doesn't really look like a relative link.\n"));
545 /* 3. If its domain is not to be accepted/looked-up, chuck it
547 if (!accept_domain (u))
549 DEBUGP (("The domain was not accepted.\n"));
553 /* 4. Check for parent directory.
555 If we descended to a different host or changed the scheme, ignore
556 opt.no_parent. Also ignore it for documents needed to display
557 the parent page when in -p mode. */
559 && schemes_are_similar_p (u->scheme, start_url_parsed->scheme)
560 && 0 == strcasecmp (u->host, start_url_parsed->host)
561 && u->port == start_url_parsed->port
562 && !(opt.page_requisites && upos->link_inline_p))
564 if (!subdir_p (start_url_parsed->dir, u->dir))
566 DEBUGP (("Going to \"%s\" would escape \"%s\" with no_parent on.\n",
567 u->dir, start_url_parsed->dir));
572 /* 5. If the file does not match the acceptance list, or is on the
573 rejection list, chuck it out. The same goes for the directory
574 exclusion and inclusion lists. */
575 if (opt.includes || opt.excludes)
577 if (!accdir (u->dir))
579 DEBUGP (("%s (%s) is excluded/not-included.\n", url, u->dir));
584 /* 6. Check for acceptance/rejection rules. We ignore these rules
585 for directories (no file name to match) and for non-leaf HTMLs,
586 which can lead to other files that do need to be downloaded. (-p
587 automatically implies non-leaf because with -p we can, if
588 necesary, overstep the maximum depth to get the page requisites.) */
589 if (u->file[0] != '\0'
590 && !(has_html_suffix_p (u->file)
591 /* The exception only applies to non-leaf HTMLs (but -p
592 always implies non-leaf because we can overstep the
593 maximum depth to get the requisites): */
595 opt.reclevel == INFINITE_RECURSION
597 || depth < opt.reclevel - 1
598 /* -p, which implies non-leaf (see above) */
599 || opt.page_requisites)))
601 if (!acceptable (u->file))
603 DEBUGP (("%s (%s) does not match acc/rej rules.\n",
610 if (schemes_are_similar_p (u->scheme, parent->scheme))
611 if (!opt.spanhost && 0 != strcasecmp (parent->host, u->host))
613 DEBUGP (("This is not the same hostname as the parent's (%s and %s).\n",
614 u->host, parent->host));
619 if (opt.use_robots && u_scheme_like_http)
621 struct robot_specs *specs = res_get_specs (u->host, u->port);
625 if (res_retrieve_file (url, &rfile, iri))
627 specs = res_parse_from_file (rfile);
629 /* Delete the robots.txt file if we chose to either delete the
630 files after downloading or we're just running a spider. */
631 if (opt.delete_after || opt.spider)
633 logprintf (LOG_VERBOSE, "Removing %s.\n", rfile);
635 logprintf (LOG_NOTQUIET, "unlink: %s\n",
643 /* If we cannot get real specs, at least produce
644 dummy ones so that we can register them and stop
645 trying to retrieve them. */
646 specs = res_parse ("", 0);
648 res_register_specs (u->host, u->port, specs);
651 /* Now that we have (or don't have) robots.txt specs, we can
652 check what they say. */
653 if (!res_match_path (specs, u->path))
655 DEBUGP (("Not following %s because robots.txt forbids it.\n", url));
656 string_set_add (blacklist, url);
661 /* The URL has passed all the tests. It can be placed in the
663 DEBUGP (("Decided to load it.\n"));
668 DEBUGP (("Decided NOT to load it.\n"));
673 /* This function determines whether we will consider downloading the
674 children of a URL whose download resulted in a redirection,
675 possibly to another host, etc. It is needed very rarely, and thus
676 it is merely a simple-minded wrapper around download_child_p. */
679 descend_redirect_p (const char *redirected, struct url *orig_parsed, int depth,
680 struct url *start_url_parsed, struct hash_table *blacklist,
683 struct url *new_parsed;
687 assert (orig_parsed != NULL);
689 new_parsed = url_parse (redirected, NULL, NULL, false);
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 (new_parsed);
702 DEBUGP (("Redirection \"%s\" failed the test.\n", redirected));
707 /* vim:set sts=2 sw=2 cino+={s: */