1 /* Handling of recursive HTTP retrieving.
2 Copyright (C) 1996, 1997, 1998, 1999, 2000, 2001, 2002, 2003, 2004,
3 2005, 2006, 2007, 2008, 2009, 2010 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. */
53 /* Functions for maintaining the URL queue. */
55 struct queue_element {
56 const char *url; /* the URL to download */
57 const char *referer; /* the referring document */
58 int depth; /* the depth */
59 bool html_allowed; /* whether the document is allowed to
60 be treated as HTML. */
61 struct iri *iri; /* sXXXav */
62 bool css_allowed; /* whether the document is allowed to
64 struct queue_element *next; /* next element in queue */
68 struct queue_element *head;
69 struct queue_element *tail;
73 /* Create a URL queue. */
75 static struct url_queue *
78 struct url_queue *queue = xnew0 (struct url_queue);
82 /* Delete a URL queue. */
85 url_queue_delete (struct url_queue *queue)
90 /* Enqueue a URL in the queue. The queue is FIFO: the items will be
91 retrieved ("dequeued") from the queue in the order they were placed
95 url_enqueue (struct url_queue *queue, struct iri *i,
96 const char *url, const char *referer, int depth,
97 bool html_allowed, bool css_allowed)
99 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",
113 quotearg_n_style (0, escape_quoting_style, url), depth));
114 DEBUGP (("Queue count %d, maxcount %d.\n", queue->count, queue->maxcount));
117 DEBUGP (("[IRI Enqueuing %s with %s\n", quote_n (0, url),
118 i->uri_encoding ? quote_n (1, i->uri_encoding) : "None"));
121 queue->tail->next = qel;
125 queue->head = queue->tail;
128 /* Take a URL out of the queue. Return true if this operation
129 succeeded, or false if the queue is empty. */
132 url_dequeue (struct url_queue *queue, struct iri **i,
133 const char **url, const char **referer, int *depth,
134 bool *html_allowed, bool *css_allowed)
136 struct queue_element *qel = queue->head;
141 queue->head = queue->head->next;
147 *referer = qel->referer;
149 *html_allowed = qel->html_allowed;
150 *css_allowed = qel->css_allowed;
154 DEBUGP (("Dequeuing %s at depth %d\n",
155 quotearg_n_style (0, escape_quoting_style, 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;
201 struct iri *i = iri_new ();
203 #define COPYSTR(x) (x) ? xstrdup(x) : NULL;
204 /* Duplicate pi struct if not NULL */
207 i->uri_encoding = COPYSTR (pi->uri_encoding);
208 i->content_encoding = COPYSTR (pi->content_encoding);
209 i->utf8_encode = pi->utf8_encode;
212 set_uri_encoding (i, opt.locale, true);
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;
280 struct url *url_parsed = url_parse (url, &url_err, i, true);
282 status = retrieve_url (url_parsed, url, &file, &redirected, referer,
283 &dt, false, i, true);
285 if (html_allowed && file && status == RETROK
286 && (dt & RETROKF) && (dt & TEXTHTML))
292 /* a little different, css_allowed can override content type
293 lots of web servers serve css with an incorrect content type
295 if (file && status == RETROK
297 ((dt & TEXTCSS) || css_allowed))
305 /* We have been redirected, possibly to another host, or
306 different path, or wherever. Check whether we really
307 want to follow it. */
310 if (!descend_redirect_p (redirected, url_parsed, depth,
311 start_url_parsed, blacklist, i))
314 /* Make sure that the old pre-redirect form gets
316 string_set_add (blacklist, url);
325 url = xstrdup (url_parsed->url);
327 url_free(url_parsed);
332 visited_url (url, referer);
336 && depth >= opt.reclevel && opt.reclevel != INFINITE_RECURSION)
338 if (opt.page_requisites
339 && (depth == opt.reclevel || depth == opt.reclevel + 1))
341 /* When -p is specified, we are allowed to exceed the
342 maximum depth, but only for the "inline" links,
343 i.e. those that are needed to display the page.
344 Originally this could exceed the depth at most by
345 one, but we allow one more level so that the leaf
346 pages that contain frames can be loaded
348 dash_p_leaf_HTML = true;
352 /* Either -p wasn't specified or it was and we've
353 already spent the two extra (pseudo-)levels that it
354 affords us, so we need to bail out. */
355 DEBUGP (("Not descending further; at depth %d, max. %d.\n",
356 depth, opt.reclevel));
361 /* If the downloaded document was HTML or CSS, parse it and enqueue the
362 links it contains. */
366 bool meta_disallow_follow = false;
367 struct urlpos *children
368 = is_css ? get_urls_css_file (file, url) :
369 get_urls_html (file, url, &meta_disallow_follow, i);
371 if (opt.use_robots && meta_disallow_follow)
373 free_urlpos (children);
379 struct urlpos *child = children;
380 struct url *url_parsed = url_parse (url, NULL, i, true);
382 char *referer_url = url;
383 bool strip_auth = (url_parsed != NULL
384 && url_parsed->user != NULL);
385 assert (url_parsed != NULL);
387 /* Strip auth info if present */
389 referer_url = url_string (url_parsed, URL_AUTH_HIDE);
391 for (; child; child = child->next)
393 if (child->ignore_when_downloading)
395 if (dash_p_leaf_HTML && !child->link_inline_p)
397 if (download_child_p (child, url_parsed, depth, start_url_parsed,
401 set_uri_encoding (ci, i->content_encoding, false);
402 url_enqueue (queue, ci, xstrdup (child->url->url),
403 xstrdup (referer_url), depth + 1,
404 child->link_expect_html,
405 child->link_expect_css);
406 /* We blacklist the URL we have enqueued, because we
407 don't want to enqueue (and hence download) the
409 string_set_add (blacklist, child->url->url);
415 url_free (url_parsed);
416 free_urlpos (children);
422 || opt.spider /* opt.recursive is implicitely true */
423 || !acceptable (file)))
425 /* Either --delete-after was specified, or we loaded this
426 (otherwise unneeded because of --spider or rejected by -R)
427 HTML file just to harvest its hyperlinks -- in either case,
428 delete the local file. */
429 DEBUGP (("Removing file due to %s in recursive_retrieve():\n",
430 opt.delete_after ? "--delete-after" :
431 (opt.spider ? "--spider" :
432 "recursive rejection criteria")));
433 logprintf (LOG_VERBOSE,
434 (opt.delete_after || opt.spider
435 ? _("Removing %s.\n")
436 : _("Removing %s since it should be rejected.\n")),
439 logprintf (LOG_NOTQUIET, "unlink: %s\n", strerror (errno));
440 logputs (LOG_VERBOSE, "\n");
441 register_delete_file (file);
445 xfree_null (referer);
450 /* If anything is left of the queue due to a premature exit, free it
457 while (url_dequeue (queue, (struct iri **)&d6,
458 (const char **)&d1, (const char **)&d2, &d3, &d4, &d5))
465 url_queue_delete (queue);
467 string_set_free (blacklist);
469 if (opt.quota && total_downloaded_bytes > opt.quota)
471 else if (status == FWRITEERR)
477 /* Based on the context provided by retrieve_tree, decide whether a
478 URL is to be descended to. This is only ever called from
479 retrieve_tree, but is in a separate function for clarity.
481 The most expensive checks (such as those for robots) are memoized
482 by storing these URLs to BLACKLIST. This may or may not help. It
483 will help if those URLs are encountered many times. */
486 download_child_p (const struct urlpos *upos, struct url *parent, int depth,
487 struct url *start_url_parsed, struct hash_table *blacklist,
490 struct url *u = upos->url;
491 const char *url = u->url;
492 bool u_scheme_like_http;
494 DEBUGP (("Deciding whether to enqueue \"%s\".\n", url));
496 if (string_set_contains (blacklist, url))
500 char *referrer = url_string (parent, URL_AUTH_HIDE_PASSWD);
501 DEBUGP (("download_child_p: parent->url is: %s\n", quote (parent->url)));
502 visited_url (url, referrer);
505 DEBUGP (("Already on the black list.\n"));
509 /* Several things to check for:
510 1. if scheme is not http, and we don't load it
511 2. check for relative links (if relative_only is set)
513 4. check for no-parent
514 5. check for excludes && includes
516 7. check for same host (if spanhost is unset), with possible
517 gethostbyname baggage
518 8. check for robots.txt
520 Addendum: If the URL is FTP, and it is to be loaded, only the
521 domain and suffix settings are "stronger".
523 Note that .html files will get loaded regardless of suffix rules
524 (but that is remedied later with unlink) unless the depth equals
527 More time- and memory- consuming tests should be put later on
530 /* Determine whether URL under consideration has a HTTP-like scheme. */
531 u_scheme_like_http = schemes_are_similar_p (u->scheme, SCHEME_HTTP);
533 /* 1. Schemes other than HTTP are normally not recursed into. */
534 if (!u_scheme_like_http && !(u->scheme == SCHEME_FTP && opt.follow_ftp))
536 DEBUGP (("Not following non-HTTP schemes.\n"));
540 /* 2. If it is an absolute link and they are not followed, throw it
542 if (u_scheme_like_http)
543 if (opt.relative_only && !upos->link_relative_p)
545 DEBUGP (("It doesn't really look like a relative link.\n"));
549 /* 3. If its domain is not to be accepted/looked-up, chuck it
551 if (!accept_domain (u))
553 DEBUGP (("The domain was not accepted.\n"));
557 /* 4. Check for parent directory.
559 If we descended to a different host or changed the scheme, ignore
560 opt.no_parent. Also ignore it for documents needed to display
561 the parent page when in -p mode. */
563 && schemes_are_similar_p (u->scheme, start_url_parsed->scheme)
564 && 0 == strcasecmp (u->host, start_url_parsed->host)
565 && u->port == start_url_parsed->port
566 && !(opt.page_requisites && upos->link_inline_p))
568 if (!subdir_p (start_url_parsed->dir, u->dir))
570 DEBUGP (("Going to \"%s\" would escape \"%s\" with no_parent on.\n",
571 u->dir, start_url_parsed->dir));
576 /* 5. If the file does not match the acceptance list, or is on the
577 rejection list, chuck it out. The same goes for the directory
578 exclusion and inclusion lists. */
579 if (opt.includes || opt.excludes)
581 if (!accdir (u->dir))
583 DEBUGP (("%s (%s) is excluded/not-included.\n", url, u->dir));
588 /* 6. Check for acceptance/rejection rules. We ignore these rules
589 for directories (no file name to match) and for non-leaf HTMLs,
590 which can lead to other files that do need to be downloaded. (-p
591 automatically implies non-leaf because with -p we can, if
592 necesary, overstep the maximum depth to get the page requisites.) */
593 if (u->file[0] != '\0'
594 && !(has_html_suffix_p (u->file)
595 /* The exception only applies to non-leaf HTMLs (but -p
596 always implies non-leaf because we can overstep the
597 maximum depth to get the requisites): */
599 opt.reclevel == INFINITE_RECURSION
601 || depth < opt.reclevel - 1
602 /* -p, which implies non-leaf (see above) */
603 || opt.page_requisites)))
605 if (!acceptable (u->file))
607 DEBUGP (("%s (%s) does not match acc/rej rules.\n",
614 if (schemes_are_similar_p (u->scheme, parent->scheme))
615 if (!opt.spanhost && 0 != strcasecmp (parent->host, u->host))
617 DEBUGP (("This is not the same hostname as the parent's (%s and %s).\n",
618 u->host, parent->host));
623 if (opt.use_robots && u_scheme_like_http)
625 struct robot_specs *specs = res_get_specs (u->host, u->port);
629 if (res_retrieve_file (url, &rfile, iri))
631 specs = res_parse_from_file (rfile);
633 /* Delete the robots.txt file if we chose to either delete the
634 files after downloading or we're just running a spider. */
635 if (opt.delete_after || opt.spider)
637 logprintf (LOG_VERBOSE, "Removing %s.\n", rfile);
639 logprintf (LOG_NOTQUIET, "unlink: %s\n",
647 /* If we cannot get real specs, at least produce
648 dummy ones so that we can register them and stop
649 trying to retrieve them. */
650 specs = res_parse ("", 0);
652 res_register_specs (u->host, u->port, specs);
655 /* Now that we have (or don't have) robots.txt specs, we can
656 check what they say. */
657 if (!res_match_path (specs, u->path))
659 DEBUGP (("Not following %s because robots.txt forbids it.\n", url));
660 string_set_add (blacklist, url);
665 /* The URL has passed all the tests. It can be placed in the
667 DEBUGP (("Decided to load it.\n"));
672 DEBUGP (("Decided NOT to load it.\n"));
677 /* This function determines whether we will consider downloading the
678 children of a URL whose download resulted in a redirection,
679 possibly to another host, etc. It is needed very rarely, and thus
680 it is merely a simple-minded wrapper around download_child_p. */
683 descend_redirect_p (const char *redirected, struct url *orig_parsed, int depth,
684 struct url *start_url_parsed, struct hash_table *blacklist,
687 struct url *new_parsed;
691 assert (orig_parsed != NULL);
693 new_parsed = url_parse (redirected, NULL, NULL, false);
694 assert (new_parsed != NULL);
696 upos = xnew0 (struct urlpos);
697 upos->url = new_parsed;
699 success = download_child_p (upos, orig_parsed, depth,
700 start_url_parsed, blacklist, iri);
702 url_free (new_parsed);
706 DEBUGP (("Redirection \"%s\" failed the test.\n", redirected));
711 /* vim:set sts=2 sw=2 cino+={s: */