1 /* Handling of recursive HTTP retrieving.
2 Copyright (C) 1995, 1996, 1997, 2000, 2001 Free Software Foundation, Inc.
4 This file is part of GNU Wget.
6 GNU Wget is free software; you can redistribute it and/or modify
7 it under the terms of the GNU General Public License as published by
8 the Free Software Foundation; either version 2 of the License, or
9 (at your option) any later version.
11 GNU Wget is distributed in the hope that it will be useful,
12 but WITHOUT ANY WARRANTY; without even the implied warranty of
13 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 GNU General Public License for more details.
16 You should have received a copy of the GNU General Public License
17 along with Wget; if not, write to the Free Software
18 Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
20 In addition, as a special exception, the Free Software Foundation
21 gives permission to link the code of its release of Wget with the
22 OpenSSL project's "OpenSSL" library (or with modified versions of it
23 that use the same license as the "OpenSSL" library), and distribute
24 the linked executables. You must obey the GNU General Public License
25 in all respects for all of the code used other than "OpenSSL". If you
26 modify this file, you may extend this exception to your version of the
27 file, but you are not obligated to do so. If you do not wish to do
28 so, delete this exception statement from your version. */
37 #endif /* HAVE_UNISTD_H */
52 extern char *version_string;
53 extern SUM_SIZE_INT total_downloaded_bytes;
55 extern struct hash_table *dl_url_file_map;
56 extern struct hash_table *downloaded_html_set;
58 /* Functions for maintaining the URL queue. */
60 struct queue_element {
61 const char *url; /* the URL to download */
62 const char *referer; /* the referring document */
63 int depth; /* the depth */
64 unsigned int html_allowed :1; /* whether the document is allowed to
65 be treated as HTML. */
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,
99 const char *url, const char *referer, int depth, bool html_allowed)
101 struct queue_element *qel = xnew (struct queue_element);
103 qel->referer = referer;
105 qel->html_allowed = html_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,
131 struct queue_element *qel = queue->head;
136 queue->head = queue->head->next;
141 *referer = qel->referer;
143 *html_allowed = qel->html_allowed;
147 DEBUGP (("Dequeuing %s at depth %d\n", qel->url, qel->depth));
148 DEBUGP (("Queue count %d, maxcount %d.\n", queue->count, queue->maxcount));
154 static bool download_child_p (const struct urlpos *, struct url *, int,
155 struct url *, struct hash_table *);
156 static bool descend_redirect_p (const char *, const char *, int,
157 struct url *, struct hash_table *);
160 /* Retrieve a part of the web beginning with START_URL. This used to
161 be called "recursive retrieval", because the old function was
162 recursive and implemented depth-first search. retrieve_tree on the
163 other hand implements breadth-search traversal of the tree, which
164 results in much nicer ordering of downloads.
166 The algorithm this function uses is simple:
168 1. put START_URL in the queue.
169 2. while there are URLs in the queue:
171 3. get next URL from the queue.
173 5. if the URL is HTML and its depth does not exceed maximum depth,
174 get the list of URLs embedded therein.
175 6. for each of those URLs do the following:
177 7. if the URL is not one of those downloaded before, and if it
178 satisfies the criteria specified by the various command-line
179 options, add it to the queue. */
182 retrieve_tree (const char *start_url)
184 uerr_t status = RETROK;
186 /* The queue of URLs we need to load. */
187 struct url_queue *queue;
189 /* The URLs we do not wish to enqueue, because they are already in
190 the queue, but haven't been downloaded yet. */
191 struct hash_table *blacklist;
194 struct url *start_url_parsed = url_parse (start_url, &up_error_code);
196 if (!start_url_parsed)
198 logprintf (LOG_NOTQUIET, "%s: %s.\n", start_url,
199 url_error (up_error_code));
203 queue = url_queue_new ();
204 blacklist = make_string_hash_table (0);
206 /* Enqueue the starting URL. Use start_url_parsed->url rather than
207 just URL so we enqueue the canonical form of the URL. */
208 url_enqueue (queue, xstrdup (start_url_parsed->url), NULL, 0, true);
209 string_set_add (blacklist, start_url_parsed->url);
213 bool descend = false;
214 char *url, *referer, *file = NULL;
217 bool dash_p_leaf_HTML = false;
219 if (opt.quota && total_downloaded_bytes > opt.quota)
221 if (status == FWRITEERR)
224 /* Get the next URL from the queue... */
226 if (!url_dequeue (queue,
227 (const char **)&url, (const char **)&referer,
228 &depth, &html_allowed))
231 /* ...and download it. Note that this download is in most cases
232 unconditional, as download_child_p already makes sure a file
233 doesn't get enqueued twice -- and yet this check is here, and
234 not in download_child_p. This is so that if you run `wget -r
235 URL1 URL2', and a random URL is encountered once under URL1
236 and again under URL2, but at a different (possibly smaller)
237 depth, we want the URL's children to be taken into account
239 if (dl_url_file_map && hash_table_contains (dl_url_file_map, url))
241 file = xstrdup (hash_table_get (dl_url_file_map, url));
243 DEBUGP (("Already downloaded \"%s\", reusing it from \"%s\".\n",
247 && downloaded_html_set
248 && string_set_contains (downloaded_html_set, file))
254 char *redirected = NULL;
255 bool oldrec = opt.recursive;
257 opt.recursive = false;
258 status = retrieve_url (url, &file, &redirected, referer, &dt);
259 opt.recursive = oldrec;
261 if (html_allowed && file && status == RETROK
262 && (dt & RETROKF) && (dt & TEXTHTML))
267 /* We have been redirected, possibly to another host, or
268 different path, or wherever. Check whether we really
269 want to follow it. */
272 if (!descend_redirect_p (redirected, url, depth,
273 start_url_parsed, blacklist))
276 /* Make sure that the old pre-redirect form gets
278 string_set_add (blacklist, url);
287 && depth >= opt.reclevel && opt.reclevel != INFINITE_RECURSION)
289 if (opt.page_requisites
290 && (depth == opt.reclevel || depth == opt.reclevel + 1))
292 /* When -p is specified, we are allowed to exceed the
293 maximum depth, but only for the "inline" links,
294 i.e. those that are needed to display the page.
295 Originally this could exceed the depth at most by
296 one, but we allow one more level so that the leaf
297 pages that contain frames can be loaded
299 dash_p_leaf_HTML = true;
303 /* Either -p wasn't specified or it was and we've
304 already spent the two extra (pseudo-)levels that it
305 affords us, so we need to bail out. */
306 DEBUGP (("Not descending further; at depth %d, max. %d.\n",
307 depth, opt.reclevel));
312 /* If the downloaded document was HTML, parse it and enqueue the
313 links it contains. */
317 bool meta_disallow_follow = false;
318 struct urlpos *children
319 = get_urls_html (file, url, &meta_disallow_follow);
321 if (opt.use_robots && meta_disallow_follow)
323 free_urlpos (children);
329 struct urlpos *child = children;
330 struct url *url_parsed = url_parsed = url_parse (url, NULL);
331 assert (url_parsed != NULL);
333 for (; child; child = child->next)
335 if (child->ignore_when_downloading)
337 if (dash_p_leaf_HTML && !child->link_inline_p)
339 if (download_child_p (child, url_parsed, depth, start_url_parsed,
342 url_enqueue (queue, xstrdup (child->url->url),
343 xstrdup (url), depth + 1,
344 child->link_expect_html);
345 /* We blacklist the URL we have enqueued, because we
346 don't want to enqueue (and hence download) the
348 string_set_add (blacklist, child->url->url);
352 url_free (url_parsed);
353 free_urlpos (children);
357 if (opt.delete_after || (file && !acceptable (file)))
359 /* Either --delete-after was specified, or we loaded this
360 otherwise rejected (e.g. by -R) HTML file just so we
361 could harvest its hyperlinks -- in either case, delete
363 DEBUGP (("Removing file due to %s in recursive_retrieve():\n",
364 opt.delete_after ? "--delete-after" :
365 "recursive rejection criteria"));
366 logprintf (LOG_VERBOSE,
368 ? _("Removing %s.\n")
369 : _("Removing %s since it should be rejected.\n")),
372 logprintf (LOG_NOTQUIET, "unlink: %s\n", strerror (errno));
373 register_delete_file (file);
377 xfree_null (referer);
381 /* If anything is left of the queue due to a premature exit, free it
387 while (url_dequeue (queue,
388 (const char **)&d1, (const char **)&d2, &d3, &d4))
394 url_queue_delete (queue);
396 if (start_url_parsed)
397 url_free (start_url_parsed);
398 string_set_free (blacklist);
400 if (opt.quota && total_downloaded_bytes > opt.quota)
402 else if (status == FWRITEERR)
408 /* Based on the context provided by retrieve_tree, decide whether a
409 URL is to be descended to. This is only ever called from
410 retrieve_tree, but is in a separate function for clarity.
412 The most expensive checks (such as those for robots) are memoized
413 by storing these URLs to BLACKLIST. This may or may not help. It
414 will help if those URLs are encountered many times. */
417 download_child_p (const struct urlpos *upos, struct url *parent, int depth,
418 struct url *start_url_parsed, struct hash_table *blacklist)
420 struct url *u = upos->url;
421 const char *url = u->url;
422 bool u_scheme_like_http;
424 DEBUGP (("Deciding whether to enqueue \"%s\".\n", url));
426 if (string_set_contains (blacklist, url))
428 DEBUGP (("Already on the black list.\n"));
432 /* Several things to check for:
433 1. if scheme is not http, and we don't load it
434 2. check for relative links (if relative_only is set)
436 4. check for no-parent
437 5. check for excludes && includes
439 7. check for same host (if spanhost is unset), with possible
440 gethostbyname baggage
441 8. check for robots.txt
443 Addendum: If the URL is FTP, and it is to be loaded, only the
444 domain and suffix settings are "stronger".
446 Note that .html files will get loaded regardless of suffix rules
447 (but that is remedied later with unlink) unless the depth equals
450 More time- and memory- consuming tests should be put later on
453 /* Determine whether URL under consideration has a HTTP-like scheme. */
454 u_scheme_like_http = schemes_are_similar_p (u->scheme, SCHEME_HTTP);
456 /* 1. Schemes other than HTTP are normally not recursed into. */
457 if (!u_scheme_like_http && !(u->scheme == SCHEME_FTP && opt.follow_ftp))
459 DEBUGP (("Not following non-HTTP schemes.\n"));
463 /* 2. If it is an absolute link and they are not followed, throw it
465 if (u_scheme_like_http)
466 if (opt.relative_only && !upos->link_relative_p)
468 DEBUGP (("It doesn't really look like a relative link.\n"));
472 /* 3. If its domain is not to be accepted/looked-up, chuck it
474 if (!accept_domain (u))
476 DEBUGP (("The domain was not accepted.\n"));
480 /* 4. Check for parent directory.
482 If we descended to a different host or changed the scheme, ignore
483 opt.no_parent. Also ignore it for documents needed to display
484 the parent page when in -p mode. */
486 && schemes_are_similar_p (u->scheme, start_url_parsed->scheme)
487 && 0 == strcasecmp (u->host, start_url_parsed->host)
488 && u->port == start_url_parsed->port
489 && !(opt.page_requisites && upos->link_inline_p))
491 if (!frontcmp (start_url_parsed->dir, u->dir))
493 DEBUGP (("Going to \"%s\" would escape \"%s\" with no_parent on.\n",
494 u->dir, start_url_parsed->dir));
499 /* 5. If the file does not match the acceptance list, or is on the
500 rejection list, chuck it out. The same goes for the directory
501 exclusion and inclusion lists. */
502 if (opt.includes || opt.excludes)
504 if (!accdir (u->dir, ALLABS))
506 DEBUGP (("%s (%s) is excluded/not-included.\n", url, u->dir));
511 /* 6. Check for acceptance/rejection rules. We ignore these rules
512 for directories (no file name to match) and for non-leaf HTMLs,
513 which can lead to other files that do need to be downloaded. (-p
514 automatically implies non-leaf because with -p we can, if
515 necesary, overstep the maximum depth to get the page requisites.) */
516 if (u->file[0] != '\0'
517 && !(has_html_suffix_p (u->file)
518 /* The exception only applies to non-leaf HTMLs (but -p
519 always implies non-leaf because we can overstep the
520 maximum depth to get the requisites): */
522 opt.reclevel == INFINITE_RECURSION
524 || depth < opt.reclevel - 1
525 /* -p, which implies non-leaf (see above) */
526 || opt.page_requisites)))
528 if (!acceptable (u->file))
530 DEBUGP (("%s (%s) does not match acc/rej rules.\n",
537 if (schemes_are_similar_p (u->scheme, parent->scheme))
538 if (!opt.spanhost && 0 != strcasecmp (parent->host, u->host))
540 DEBUGP (("This is not the same hostname as the parent's (%s and %s).\n",
541 u->host, parent->host));
546 if (opt.use_robots && u_scheme_like_http)
548 struct robot_specs *specs = res_get_specs (u->host, u->port);
552 if (res_retrieve_file (url, &rfile))
554 specs = res_parse_from_file (rfile);
559 /* If we cannot get real specs, at least produce
560 dummy ones so that we can register them and stop
561 trying to retrieve them. */
562 specs = res_parse ("", 0);
564 res_register_specs (u->host, u->port, specs);
567 /* Now that we have (or don't have) robots.txt specs, we can
568 check what they say. */
569 if (!res_match_path (specs, u->path))
571 DEBUGP (("Not following %s because robots.txt forbids it.\n", url));
572 string_set_add (blacklist, url);
577 /* The URL has passed all the tests. It can be placed in the
579 DEBUGP (("Decided to load it.\n"));
584 DEBUGP (("Decided NOT to load it.\n"));
589 /* This function determines whether we will consider downloading the
590 children of a URL whose download resulted in a redirection,
591 possibly to another host, etc. It is needed very rarely, and thus
592 it is merely a simple-minded wrapper around download_child_p. */
595 descend_redirect_p (const char *redirected, const char *original, int depth,
596 struct url *start_url_parsed, struct hash_table *blacklist)
598 struct url *orig_parsed, *new_parsed;
602 orig_parsed = url_parse (original, NULL);
603 assert (orig_parsed != NULL);
605 new_parsed = url_parse (redirected, NULL);
606 assert (new_parsed != NULL);
608 upos = xnew0 (struct urlpos);
609 upos->url = new_parsed;
611 success = download_child_p (upos, orig_parsed, depth,
612 start_url_parsed, blacklist);
614 url_free (orig_parsed);
615 url_free (new_parsed);
619 DEBUGP (("Redirection \"%s\" failed the test.\n", redirected));