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 LARGE_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, int 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 1 if this operation succeeded,
124 or 0 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 int download_child_p (const struct urlpos *, struct url *, int,
155 struct url *, struct hash_table *);
156 static int 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, 1);
209 string_set_add (blacklist, start_url_parsed->url);
214 char *url, *referer, *file = NULL;
215 int depth, html_allowed;
216 int dash_p_leaf_HTML = 0;
218 if (opt.quota && total_downloaded_bytes > opt.quota)
220 if (status == FWRITEERR)
223 /* Get the next URL from the queue... */
225 if (!url_dequeue (queue,
226 (const char **)&url, (const char **)&referer,
227 &depth, &html_allowed))
230 /* ...and download it. Note that this download is in most cases
231 unconditional, as download_child_p already makes sure a file
232 doesn't get enqueued twice -- and yet this check is here, and
233 not in download_child_p. This is so that if you run `wget -r
234 URL1 URL2', and a random URL is encountered once under URL1
235 and again under URL2, but at a different (possibly smaller)
236 depth, we want the URL's children to be taken into account
238 if (dl_url_file_map && hash_table_contains (dl_url_file_map, url))
240 file = xstrdup (hash_table_get (dl_url_file_map, url));
242 DEBUGP (("Already downloaded \"%s\", reusing it from \"%s\".\n",
246 && downloaded_html_set
247 && string_set_contains (downloaded_html_set, file))
253 char *redirected = NULL;
254 int oldrec = opt.recursive;
257 status = retrieve_url (url, &file, &redirected, referer, &dt);
258 opt.recursive = oldrec;
260 if (html_allowed && file && status == RETROK
261 && (dt & RETROKF) && (dt & TEXTHTML))
266 /* We have been redirected, possibly to another host, or
267 different path, or wherever. Check whether we really
268 want to follow it. */
271 if (!descend_redirect_p (redirected, url, depth,
272 start_url_parsed, blacklist))
275 /* Make sure that the old pre-redirect form gets
277 string_set_add (blacklist, url);
286 && depth >= opt.reclevel && opt.reclevel != INFINITE_RECURSION)
288 if (opt.page_requisites
289 && (depth == opt.reclevel || depth == opt.reclevel + 1))
291 /* When -p is specified, we are allowed to exceed the
292 maximum depth, but only for the "inline" links,
293 i.e. those that are needed to display the page.
294 Originally this could exceed the depth at most by
295 one, but we allow one more level so that the leaf
296 pages that contain frames can be loaded
298 dash_p_leaf_HTML = 1;
302 /* Either -p wasn't specified or it was and we've
303 already spent the two extra (pseudo-)levels that it
304 affords us, so we need to bail out. */
305 DEBUGP (("Not descending further; at depth %d, max. %d.\n",
306 depth, opt.reclevel));
311 /* If the downloaded document was HTML, parse it and enqueue the
312 links it contains. */
316 int meta_disallow_follow = 0;
317 struct urlpos *children
318 = get_urls_html (file, url, &meta_disallow_follow);
320 if (opt.use_robots && meta_disallow_follow)
322 free_urlpos (children);
328 struct urlpos *child = children;
329 struct url *url_parsed = url_parsed = url_parse (url, NULL);
330 assert (url_parsed != NULL);
332 for (; child; child = child->next)
334 if (child->ignore_when_downloading)
336 if (dash_p_leaf_HTML && !child->link_inline_p)
338 if (download_child_p (child, url_parsed, depth, start_url_parsed,
341 url_enqueue (queue, xstrdup (child->url->url),
342 xstrdup (url), depth + 1,
343 child->link_expect_html);
344 /* We blacklist the URL we have enqueued, because we
345 don't want to enqueue (and hence download) the
347 string_set_add (blacklist, child->url->url);
351 url_free (url_parsed);
352 free_urlpos (children);
356 if (opt.delete_after || (file && !acceptable (file)))
358 /* Either --delete-after was specified, or we loaded this
359 otherwise rejected (e.g. by -R) HTML file just so we
360 could harvest its hyperlinks -- in either case, delete
362 DEBUGP (("Removing file due to %s in recursive_retrieve():\n",
363 opt.delete_after ? "--delete-after" :
364 "recursive rejection criteria"));
365 logprintf (LOG_VERBOSE,
367 ? _("Removing %s.\n")
368 : _("Removing %s since it should be rejected.\n")),
371 logprintf (LOG_NOTQUIET, "unlink: %s\n", strerror (errno));
372 register_delete_file (file);
376 xfree_null (referer);
380 /* If anything is left of the queue due to a premature exit, free it
385 while (url_dequeue (queue,
386 (const char **)&d1, (const char **)&d2, &d3, &d4))
392 url_queue_delete (queue);
394 if (start_url_parsed)
395 url_free (start_url_parsed);
396 string_set_free (blacklist);
398 if (opt.quota && total_downloaded_bytes > opt.quota)
400 else if (status == FWRITEERR)
406 /* Based on the context provided by retrieve_tree, decide whether a
407 URL is to be descended to. This is only ever called from
408 retrieve_tree, but is in a separate function for clarity.
410 The most expensive checks (such as those for robots) are memoized
411 by storing these URLs to BLACKLIST. This may or may not help. It
412 will help if those URLs are encountered many times. */
415 download_child_p (const struct urlpos *upos, struct url *parent, int depth,
416 struct url *start_url_parsed, struct hash_table *blacklist)
418 struct url *u = upos->url;
419 const char *url = u->url;
420 int u_scheme_like_http;
422 DEBUGP (("Deciding whether to enqueue \"%s\".\n", url));
424 if (string_set_contains (blacklist, url))
426 DEBUGP (("Already on the black list.\n"));
430 /* Several things to check for:
431 1. if scheme is not http, and we don't load it
432 2. check for relative links (if relative_only is set)
434 4. check for no-parent
435 5. check for excludes && includes
437 7. check for same host (if spanhost is unset), with possible
438 gethostbyname baggage
439 8. check for robots.txt
441 Addendum: If the URL is FTP, and it is to be loaded, only the
442 domain and suffix settings are "stronger".
444 Note that .html files will get loaded regardless of suffix rules
445 (but that is remedied later with unlink) unless the depth equals
448 More time- and memory- consuming tests should be put later on
451 /* Determine whether URL under consideration has a HTTP-like scheme. */
452 u_scheme_like_http = schemes_are_similar_p (u->scheme, SCHEME_HTTP);
454 /* 1. Schemes other than HTTP are normally not recursed into. */
455 if (!u_scheme_like_http && !(u->scheme == SCHEME_FTP && opt.follow_ftp))
457 DEBUGP (("Not following non-HTTP schemes.\n"));
461 /* 2. If it is an absolute link and they are not followed, throw it
463 if (u_scheme_like_http)
464 if (opt.relative_only && !upos->link_relative_p)
466 DEBUGP (("It doesn't really look like a relative link.\n"));
470 /* 3. If its domain is not to be accepted/looked-up, chuck it
472 if (!accept_domain (u))
474 DEBUGP (("The domain was not accepted.\n"));
478 /* 4. Check for parent directory.
480 If we descended to a different host or changed the scheme, ignore
481 opt.no_parent. Also ignore it for documents needed to display
482 the parent page when in -p mode. */
484 && schemes_are_similar_p (u->scheme, start_url_parsed->scheme)
485 && 0 == strcasecmp (u->host, start_url_parsed->host)
486 && u->port == start_url_parsed->port
487 && !(opt.page_requisites && upos->link_inline_p))
489 if (!frontcmp (start_url_parsed->dir, u->dir))
491 DEBUGP (("Going to \"%s\" would escape \"%s\" with no_parent on.\n",
492 u->dir, start_url_parsed->dir));
497 /* 5. If the file does not match the acceptance list, or is on the
498 rejection list, chuck it out. The same goes for the directory
499 exclusion and inclusion lists. */
500 if (opt.includes || opt.excludes)
502 if (!accdir (u->dir, ALLABS))
504 DEBUGP (("%s (%s) is excluded/not-included.\n", url, u->dir));
509 /* 6. Check for acceptance/rejection rules. We ignore these rules
510 for directories (no file name to match) and for non-leaf HTMLs,
511 which can lead to other files that do need to be downloaded. (-p
512 automatically implies non-leaf because with -p we can, if
513 necesary, overstep the maximum depth to get the page requisites.) */
514 if (u->file[0] != '\0'
515 && !(has_html_suffix_p (u->file)
516 /* The exception only applies to non-leaf HTMLs (but -p
517 always implies non-leaf because we can overstep the
518 maximum depth to get the requisites): */
520 opt.reclevel == INFINITE_RECURSION
522 || depth < opt.reclevel - 1
523 /* -p, which implies non-leaf (see above) */
524 || opt.page_requisites)))
526 if (!acceptable (u->file))
528 DEBUGP (("%s (%s) does not match acc/rej rules.\n",
535 if (schemes_are_similar_p (u->scheme, parent->scheme))
536 if (!opt.spanhost && 0 != strcasecmp (parent->host, u->host))
538 DEBUGP (("This is not the same hostname as the parent's (%s and %s).\n",
539 u->host, parent->host));
544 if (opt.use_robots && u_scheme_like_http)
546 struct robot_specs *specs = res_get_specs (u->host, u->port);
550 if (res_retrieve_file (url, &rfile))
552 specs = res_parse_from_file (rfile);
557 /* If we cannot get real specs, at least produce
558 dummy ones so that we can register them and stop
559 trying to retrieve them. */
560 specs = res_parse ("", 0);
562 res_register_specs (u->host, u->port, specs);
565 /* Now that we have (or don't have) robots.txt specs, we can
566 check what they say. */
567 if (!res_match_path (specs, u->path))
569 DEBUGP (("Not following %s because robots.txt forbids it.\n", url));
570 string_set_add (blacklist, url);
575 /* The URL has passed all the tests. It can be placed in the
577 DEBUGP (("Decided to load it.\n"));
582 DEBUGP (("Decided NOT to load it.\n"));
587 /* This function determines whether we will consider downloading the
588 children of a URL whose download resulted in a redirection,
589 possibly to another host, etc. It is needed very rarely, and thus
590 it is merely a simple-minded wrapper around download_child_p. */
593 descend_redirect_p (const char *redirected, const char *original, int depth,
594 struct url *start_url_parsed, struct hash_table *blacklist)
596 struct url *orig_parsed, *new_parsed;
600 orig_parsed = url_parse (original, NULL);
601 assert (orig_parsed != NULL);
603 new_parsed = url_parse (redirected, NULL);
604 assert (new_parsed != NULL);
606 upos = xnew0 (struct urlpos);
607 upos->url = new_parsed;
609 success = download_child_p (upos, orig_parsed, depth,
610 start_url_parsed, blacklist);
612 url_free (orig_parsed);
613 url_free (new_parsed);
617 DEBUGP (("Redirection \"%s\" failed the test.\n", redirected));