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 (url),
119 i->uri_encoding ? quote (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 *, const char *, 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 (const char *start_url, 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 url *start_url_parsed;
203 struct iri *i = iri_new ();
205 #define COPYSTR(x) (x) ? xstrdup(x) : NULL;
206 /* Duplicate pi struct if not NULL */
209 i->uri_encoding = COPYSTR (pi->uri_encoding);
210 i->content_encoding = COPYSTR (pi->content_encoding);
211 i->utf8_encode = pi->utf8_encode;
214 set_uri_encoding (i, opt.locale, true);
217 start_url_parsed = url_parse (start_url, &up_error_code, i);
218 if (!start_url_parsed)
220 logprintf (LOG_NOTQUIET, "%s: %s.\n", start_url,
221 url_error (up_error_code));
225 queue = url_queue_new ();
226 blacklist = make_string_hash_table (0);
228 /* Enqueue the starting URL. Use start_url_parsed->url rather than
229 just URL so we enqueue the canonical form of the URL. */
230 url_enqueue (queue, i, xstrdup (start_url_parsed->url), NULL, 0, true,
232 string_set_add (blacklist, start_url_parsed->url);
236 bool descend = false;
237 char *url, *referer, *file = NULL;
239 bool html_allowed, css_allowed;
241 bool dash_p_leaf_HTML = false;
243 if (opt.quota && total_downloaded_bytes > opt.quota)
245 if (status == FWRITEERR)
248 /* Get the next URL from the queue... */
250 if (!url_dequeue (queue, (struct iri **) &i,
251 (const char **)&url, (const char **)&referer,
252 &depth, &html_allowed, &css_allowed))
255 /* ...and download it. Note that this download is in most cases
256 unconditional, as download_child_p already makes sure a file
257 doesn't get enqueued twice -- and yet this check is here, and
258 not in download_child_p. This is so that if you run `wget -r
259 URL1 URL2', and a random URL is encountered once under URL1
260 and again under URL2, but at a different (possibly smaller)
261 depth, we want the URL's children to be taken into account
263 if (dl_url_file_map && hash_table_contains (dl_url_file_map, url))
265 file = xstrdup (hash_table_get (dl_url_file_map, url));
267 DEBUGP (("Already downloaded \"%s\", reusing it from \"%s\".\n",
270 /* this sucks, needs to be combined! */
272 && downloaded_html_set
273 && string_set_contains (downloaded_html_set, file))
279 && downloaded_css_set
280 && string_set_contains (downloaded_css_set, file))
289 char *redirected = NULL;
291 status = retrieve_url (url, &file, &redirected, referer, &dt,
294 if (html_allowed && file && status == RETROK
295 && (dt & RETROKF) && (dt & TEXTHTML))
301 /* a little different, css_allowed can override content type
302 lots of web servers serve css with an incorrect content type
304 if (file && status == RETROK
306 ((dt & TEXTCSS) || css_allowed))
314 /* We have been redirected, possibly to another host, or
315 different path, or wherever. Check whether we really
316 want to follow it. */
319 if (!descend_redirect_p (redirected, url, depth,
320 start_url_parsed, blacklist, i))
323 /* Make sure that the old pre-redirect form gets
325 string_set_add (blacklist, url);
335 visited_url (url, referer);
339 && depth >= opt.reclevel && opt.reclevel != INFINITE_RECURSION)
341 if (opt.page_requisites
342 && (depth == opt.reclevel || depth == opt.reclevel + 1))
344 /* When -p is specified, we are allowed to exceed the
345 maximum depth, but only for the "inline" links,
346 i.e. those that are needed to display the page.
347 Originally this could exceed the depth at most by
348 one, but we allow one more level so that the leaf
349 pages that contain frames can be loaded
351 dash_p_leaf_HTML = true;
355 /* Either -p wasn't specified or it was and we've
356 already spent the two extra (pseudo-)levels that it
357 affords us, so we need to bail out. */
358 DEBUGP (("Not descending further; at depth %d, max. %d.\n",
359 depth, opt.reclevel));
364 /* If the downloaded document was HTML or CSS, parse it and enqueue the
365 links it contains. */
369 bool meta_disallow_follow = false;
370 struct urlpos *children
371 = is_css ? get_urls_css_file (file, url) :
372 get_urls_html (file, url, &meta_disallow_follow, i);
374 if (opt.use_robots && meta_disallow_follow)
376 free_urlpos (children);
382 struct urlpos *child = children;
383 struct url *url_parsed = url_parse (url, NULL, i);
385 char *referer_url = url;
386 bool strip_auth = (url_parsed != NULL
387 && url_parsed->user != NULL);
388 assert (url_parsed != NULL);
390 /* Strip auth info if present */
392 referer_url = url_string (url_parsed, URL_AUTH_HIDE);
394 for (; child; child = child->next)
396 if (child->ignore_when_downloading)
398 if (dash_p_leaf_HTML && !child->link_inline_p)
400 if (download_child_p (child, url_parsed, depth, start_url_parsed,
404 set_uri_encoding (ci, i->content_encoding, false);
405 url_enqueue (queue, ci, xstrdup (child->url->url),
406 xstrdup (referer_url), depth + 1,
407 child->link_expect_html,
408 child->link_expect_css);
409 /* We blacklist the URL we have enqueued, because we
410 don't want to enqueue (and hence download) the
412 string_set_add (blacklist, child->url->url);
418 url_free (url_parsed);
419 free_urlpos (children);
425 || opt.spider /* opt.recursive is implicitely true */
426 || !acceptable (file)))
428 /* Either --delete-after was specified, or we loaded this
429 (otherwise unneeded because of --spider or rejected by -R)
430 HTML file just to harvest its hyperlinks -- in either case,
431 delete the local file. */
432 DEBUGP (("Removing file due to %s in recursive_retrieve():\n",
433 opt.delete_after ? "--delete-after" :
434 (opt.spider ? "--spider" :
435 "recursive rejection criteria")));
436 logprintf (LOG_VERBOSE,
437 (opt.delete_after || opt.spider
438 ? _("Removing %s.\n")
439 : _("Removing %s since it should be rejected.\n")),
442 logprintf (LOG_NOTQUIET, "unlink: %s\n", strerror (errno));
443 logputs (LOG_VERBOSE, "\n");
444 register_delete_file (file);
448 xfree_null (referer);
453 /* If anything is left of the queue due to a premature exit, free it
460 while (url_dequeue (queue, (struct iri **)&d6,
461 (const char **)&d1, (const char **)&d2, &d3, &d4, &d5))
468 url_queue_delete (queue);
470 if (start_url_parsed)
471 url_free (start_url_parsed);
472 string_set_free (blacklist);
474 if (opt.quota && total_downloaded_bytes > opt.quota)
476 else if (status == FWRITEERR)
482 /* Based on the context provided by retrieve_tree, decide whether a
483 URL is to be descended to. This is only ever called from
484 retrieve_tree, but is in a separate function for clarity.
486 The most expensive checks (such as those for robots) are memoized
487 by storing these URLs to BLACKLIST. This may or may not help. It
488 will help if those URLs are encountered many times. */
491 download_child_p (const struct urlpos *upos, struct url *parent, int depth,
492 struct url *start_url_parsed, struct hash_table *blacklist,
495 struct url *u = upos->url;
496 const char *url = u->url;
497 bool u_scheme_like_http;
499 DEBUGP (("Deciding whether to enqueue \"%s\".\n", url));
501 if (string_set_contains (blacklist, url))
505 char *referrer = url_string (parent, URL_AUTH_HIDE_PASSWD);
506 DEBUGP (("download_child_p: parent->url is: %s\n", quote (parent->url)));
507 visited_url (url, referrer);
510 DEBUGP (("Already on the black list.\n"));
514 /* Several things to check for:
515 1. if scheme is not http, and we don't load it
516 2. check for relative links (if relative_only is set)
518 4. check for no-parent
519 5. check for excludes && includes
521 7. check for same host (if spanhost is unset), with possible
522 gethostbyname baggage
523 8. check for robots.txt
525 Addendum: If the URL is FTP, and it is to be loaded, only the
526 domain and suffix settings are "stronger".
528 Note that .html files will get loaded regardless of suffix rules
529 (but that is remedied later with unlink) unless the depth equals
532 More time- and memory- consuming tests should be put later on
535 /* Determine whether URL under consideration has a HTTP-like scheme. */
536 u_scheme_like_http = schemes_are_similar_p (u->scheme, SCHEME_HTTP);
538 /* 1. Schemes other than HTTP are normally not recursed into. */
539 if (!u_scheme_like_http && !(u->scheme == SCHEME_FTP && opt.follow_ftp))
541 DEBUGP (("Not following non-HTTP schemes.\n"));
545 /* 2. If it is an absolute link and they are not followed, throw it
547 if (u_scheme_like_http)
548 if (opt.relative_only && !upos->link_relative_p)
550 DEBUGP (("It doesn't really look like a relative link.\n"));
554 /* 3. If its domain is not to be accepted/looked-up, chuck it
556 if (!accept_domain (u))
558 DEBUGP (("The domain was not accepted.\n"));
562 /* 4. Check for parent directory.
564 If we descended to a different host or changed the scheme, ignore
565 opt.no_parent. Also ignore it for documents needed to display
566 the parent page when in -p mode. */
568 && schemes_are_similar_p (u->scheme, start_url_parsed->scheme)
569 && 0 == strcasecmp (u->host, start_url_parsed->host)
570 && u->port == start_url_parsed->port
571 && !(opt.page_requisites && upos->link_inline_p))
573 if (!subdir_p (start_url_parsed->dir, u->dir))
575 DEBUGP (("Going to \"%s\" would escape \"%s\" with no_parent on.\n",
576 u->dir, start_url_parsed->dir));
581 /* 5. If the file does not match the acceptance list, or is on the
582 rejection list, chuck it out. The same goes for the directory
583 exclusion and inclusion lists. */
584 if (opt.includes || opt.excludes)
586 if (!accdir (u->dir))
588 DEBUGP (("%s (%s) is excluded/not-included.\n", url, u->dir));
593 /* 6. Check for acceptance/rejection rules. We ignore these rules
594 for directories (no file name to match) and for non-leaf HTMLs,
595 which can lead to other files that do need to be downloaded. (-p
596 automatically implies non-leaf because with -p we can, if
597 necesary, overstep the maximum depth to get the page requisites.) */
598 if (u->file[0] != '\0'
599 && !(has_html_suffix_p (u->file)
600 /* The exception only applies to non-leaf HTMLs (but -p
601 always implies non-leaf because we can overstep the
602 maximum depth to get the requisites): */
604 opt.reclevel == INFINITE_RECURSION
606 || depth < opt.reclevel - 1
607 /* -p, which implies non-leaf (see above) */
608 || opt.page_requisites)))
610 if (!acceptable (u->file))
612 DEBUGP (("%s (%s) does not match acc/rej rules.\n",
619 if (schemes_are_similar_p (u->scheme, parent->scheme))
620 if (!opt.spanhost && 0 != strcasecmp (parent->host, u->host))
622 DEBUGP (("This is not the same hostname as the parent's (%s and %s).\n",
623 u->host, parent->host));
628 if (opt.use_robots && u_scheme_like_http)
630 struct robot_specs *specs = res_get_specs (u->host, u->port);
634 if (res_retrieve_file (url, &rfile, iri))
636 specs = res_parse_from_file (rfile);
638 /* Delete the robots.txt file if we chose to either delete the
639 files after downloading or we're just running a spider. */
640 if (opt.delete_after || opt.spider)
642 logprintf (LOG_VERBOSE, "Removing %s.\n", rfile);
644 logprintf (LOG_NOTQUIET, "unlink: %s\n",
652 /* If we cannot get real specs, at least produce
653 dummy ones so that we can register them and stop
654 trying to retrieve them. */
655 specs = res_parse ("", 0);
657 res_register_specs (u->host, u->port, specs);
660 /* Now that we have (or don't have) robots.txt specs, we can
661 check what they say. */
662 if (!res_match_path (specs, u->path))
664 DEBUGP (("Not following %s because robots.txt forbids it.\n", url));
665 string_set_add (blacklist, url);
670 /* The URL has passed all the tests. It can be placed in the
672 DEBUGP (("Decided to load it.\n"));
677 DEBUGP (("Decided NOT to load it.\n"));
682 /* This function determines whether we will consider downloading the
683 children of a URL whose download resulted in a redirection,
684 possibly to another host, etc. It is needed very rarely, and thus
685 it is merely a simple-minded wrapper around download_child_p. */
688 descend_redirect_p (const char *redirected, const char *original, int depth,
689 struct url *start_url_parsed, struct hash_table *blacklist,
692 struct url *orig_parsed, *new_parsed;
696 orig_parsed = url_parse (original, NULL, NULL);
697 assert (orig_parsed != NULL);
699 new_parsed = url_parse (redirected, NULL, NULL);
700 assert (new_parsed != NULL);
702 upos = xnew0 (struct urlpos);
703 upos->url = new_parsed;
705 success = download_child_p (upos, orig_parsed, depth,
706 start_url_parsed, blacklist, iri);
708 url_free (orig_parsed);
709 url_free (new_parsed);
713 DEBUGP (("Redirection \"%s\" failed the test.\n", redirected));
718 /* vim:set sts=2 sw=2 cino+={s: */