1 /* Collect URLs from HTML source.
2 Copyright (C) 1998, 1999, 2000, 2001, 2002, 2003, 2004, 2005, 2006,
3 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. */
39 #include "html-parse.h"
44 #include "recur.h" /* declaration of get_urls_html */
49 typedef void (*tag_handler_t) (int, struct taginfo *, struct map_context *);
51 #define DECLARE_TAG_HANDLER(fun) \
52 static void fun (int, struct taginfo *, struct map_context *)
54 DECLARE_TAG_HANDLER (tag_find_urls);
55 DECLARE_TAG_HANDLER (tag_handle_base);
56 DECLARE_TAG_HANDLER (tag_handle_form);
57 DECLARE_TAG_HANDLER (tag_handle_link);
58 DECLARE_TAG_HANDLER (tag_handle_meta);
85 /* The list of known tags and functions used for handling them. Most
86 tags are simply harvested for URLs. */
87 static struct known_tag {
90 tag_handler_t handler;
92 { TAG_A, "a", tag_find_urls },
93 { TAG_APPLET, "applet", tag_find_urls },
94 { TAG_AREA, "area", tag_find_urls },
95 { TAG_BASE, "base", tag_handle_base },
96 { TAG_BGSOUND, "bgsound", tag_find_urls },
97 { TAG_BODY, "body", tag_find_urls },
98 { TAG_EMBED, "embed", tag_find_urls },
99 { TAG_FIG, "fig", tag_find_urls },
100 { TAG_FORM, "form", tag_handle_form },
101 { TAG_FRAME, "frame", tag_find_urls },
102 { TAG_IFRAME, "iframe", tag_find_urls },
103 { TAG_IMG, "img", tag_find_urls },
104 { TAG_INPUT, "input", tag_find_urls },
105 { TAG_LAYER, "layer", tag_find_urls },
106 { TAG_LINK, "link", tag_handle_link },
107 { TAG_META, "meta", tag_handle_meta },
108 { TAG_OBJECT, "object", tag_find_urls },
109 { TAG_OVERLAY, "overlay", tag_find_urls },
110 { TAG_SCRIPT, "script", tag_find_urls },
111 { TAG_TABLE, "table", tag_find_urls },
112 { TAG_TD, "td", tag_find_urls },
113 { TAG_TH, "th", tag_find_urls }
116 /* tag_url_attributes documents which attributes of which tags contain
117 URLs to harvest. It is used by tag_find_urls. */
119 /* Defines for the FLAGS. */
121 /* The link is "inline", i.e. needs to be retrieved for this document
122 to be correctly rendered. Inline links include inlined images,
123 stylesheets, children frames, etc. */
124 #define ATTR_INLINE 1
126 /* The link is expected to yield HTML contents. It's important not to
127 try to follow HTML obtained by following e.g. <img src="...">
128 regardless of content-type. Doing this causes infinite loops for
129 "images" that return non-404 error pages with links to the same
133 /* For tags handled by tag_find_urls: attributes that contain URLs to
137 const char *attr_name;
139 } tag_url_attributes[] = {
140 { TAG_A, "href", ATTR_HTML },
141 { TAG_APPLET, "code", ATTR_INLINE },
142 { TAG_AREA, "href", ATTR_HTML },
143 { TAG_BGSOUND, "src", ATTR_INLINE },
144 { TAG_BODY, "background", ATTR_INLINE },
145 { TAG_EMBED, "href", ATTR_HTML },
146 { TAG_EMBED, "src", ATTR_INLINE | ATTR_HTML },
147 { TAG_FIG, "src", ATTR_INLINE },
148 { TAG_FRAME, "src", ATTR_INLINE | ATTR_HTML },
149 { TAG_IFRAME, "src", ATTR_INLINE | ATTR_HTML },
150 { TAG_IMG, "href", ATTR_INLINE },
151 { TAG_IMG, "lowsrc", ATTR_INLINE },
152 { TAG_IMG, "src", ATTR_INLINE },
153 { TAG_INPUT, "src", ATTR_INLINE },
154 { TAG_LAYER, "src", ATTR_INLINE | ATTR_HTML },
155 { TAG_OBJECT, "data", ATTR_INLINE },
156 { TAG_OVERLAY, "src", ATTR_INLINE | ATTR_HTML },
157 { TAG_SCRIPT, "src", ATTR_INLINE },
158 { TAG_TABLE, "background", ATTR_INLINE },
159 { TAG_TD, "background", ATTR_INLINE },
160 { TAG_TH, "background", ATTR_INLINE }
163 /* The lists of interesting tags and attributes are built dynamically,
164 from the information above. However, some places in the code refer
165 to the attributes not mentioned here. We add them manually. */
166 static const char *additional_attributes[] = {
167 "rel", /* used by tag_handle_link */
168 "http-equiv", /* used by tag_handle_meta */
169 "name", /* used by tag_handle_meta */
170 "content", /* used by tag_handle_meta */
171 "action" /* used by tag_handle_form */
174 static struct hash_table *interesting_tags;
175 static struct hash_table *interesting_attributes;
178 init_interesting (void)
180 /* Init the variables interesting_tags and interesting_attributes
181 that are used by the HTML parser to know which tags and
182 attributes we're interested in. We initialize this only once,
183 for performance reasons.
185 Here we also make sure that what we put in interesting_tags
186 matches the user's preferences as specified through --ignore-tags
187 and --follow-tags. */
190 interesting_tags = make_nocase_string_hash_table (countof (known_tags));
192 /* First, add all the tags we know hot to handle, mapped to their
193 respective entries in known_tags. */
194 for (i = 0; i < countof (known_tags); i++)
195 hash_table_put (interesting_tags, known_tags[i].name, known_tags + i);
197 /* Then remove the tags ignored through --ignore-tags. */
201 for (ignored = opt.ignore_tags; *ignored; ignored++)
202 hash_table_remove (interesting_tags, *ignored);
205 /* If --follow-tags is specified, use only those tags. */
208 /* Create a new table intersecting --follow-tags and known_tags,
209 and use it as interesting_tags. */
210 struct hash_table *intersect = make_nocase_string_hash_table (0);
212 for (followed = opt.follow_tags; *followed; followed++)
214 struct known_tag *t = hash_table_get (interesting_tags, *followed);
216 continue; /* ignore unknown --follow-tags entries. */
217 hash_table_put (intersect, *followed, t);
219 hash_table_destroy (interesting_tags);
220 interesting_tags = intersect;
223 /* Add the attributes we care about. */
224 interesting_attributes = make_nocase_string_hash_table (10);
225 for (i = 0; i < countof (additional_attributes); i++)
226 hash_table_put (interesting_attributes, additional_attributes[i], "1");
227 for (i = 0; i < countof (tag_url_attributes); i++)
228 hash_table_put (interesting_attributes,
229 tag_url_attributes[i].attr_name, "1");
232 /* Find the value of attribute named NAME in the taginfo TAG. If the
233 attribute is not present, return NULL. If ATTRIND is non-NULL, the
234 index of the attribute in TAG will be stored there. */
237 find_attr (struct taginfo *tag, const char *name, int *attrind)
240 for (i = 0; i < tag->nattrs; i++)
241 if (!strcasecmp (tag->attrs[i].name, name))
245 return tag->attrs[i].value;
251 char *text; /* HTML text. */
252 char *base; /* Base URI of the document, possibly
253 changed through <base href=...>. */
254 const char *parent_base; /* Base of the current document. */
255 const char *document_file; /* File name of this document. */
256 bool nofollow; /* whether NOFOLLOW was specified in a
257 <meta name=robots> tag. */
259 struct urlpos *head, *tail; /* List of URLs that is being
263 /* Append LINK_URI to the urlpos structure that is being built.
265 LINK_URI will be merged with the current document base. TAG and
266 ATTRIND are the necessary context to store the position and
269 static struct urlpos *
270 append_url (const char *link_uri,
271 struct taginfo *tag, int attrind, struct map_context *ctx)
273 int link_has_scheme = url_has_scheme (link_uri);
274 struct urlpos *newel;
275 const char *base = ctx->base ? ctx->base : ctx->parent_base;
280 DEBUGP (("%s: no base, merge will use \"%s\".\n",
281 ctx->document_file, link_uri));
283 if (!link_has_scheme)
285 /* Base URL is unavailable, and the link does not have a
286 location attached to it -- we have to give up. Since
287 this can only happen when using `--force-html -i', print
289 logprintf (LOG_NOTQUIET,
290 _("%s: Cannot resolve incomplete link %s.\n"),
291 ctx->document_file, link_uri);
295 set_ugly_no_encode (true);
296 url = url_parse (link_uri, NULL);
297 set_ugly_no_encode (false);
300 DEBUGP (("%s: link \"%s\" doesn't parse.\n",
301 ctx->document_file, link_uri));
307 /* Merge BASE with LINK_URI, but also make sure the result is
308 canonicalized, i.e. that "../" have been resolved.
309 (parse_url will do that for us.) */
311 char *complete_uri = uri_merge (base, link_uri);
313 DEBUGP (("%s: merge(\"%s\", \"%s\") -> %s\n",
314 ctx->document_file, base, link_uri, complete_uri));
316 set_ugly_no_encode (true);
317 url = url_parse (complete_uri, NULL);
318 set_ugly_no_encode (false);
321 DEBUGP (("%s: merged link \"%s\" doesn't parse.\n",
322 ctx->document_file, complete_uri));
323 xfree (complete_uri);
326 xfree (complete_uri);
329 DEBUGP (("appending \"%s\" to urlpos.\n", url->url));
331 newel = xnew0 (struct urlpos);
333 newel->pos = tag->attrs[attrind].value_raw_beginning - ctx->text;
334 newel->size = tag->attrs[attrind].value_raw_size;
336 /* A URL is relative if the host is not named, and the name does not
338 if (!link_has_scheme && *link_uri != '/')
339 newel->link_relative_p = 1;
340 else if (link_has_scheme)
341 newel->link_complete_p = 1;
345 ctx->tail->next = newel;
349 ctx->tail = ctx->head = newel;
354 /* All the tag_* functions are called from collect_tags_mapper, as
355 specified by KNOWN_TAGS. */
357 /* Default tag handler: collect URLs from attributes specified for
358 this tag by tag_url_attributes. */
361 tag_find_urls (int tagid, struct taginfo *tag, struct map_context *ctx)
367 for (i = 0; i < countof (tag_url_attributes); i++)
368 if (tag_url_attributes[i].tagid == tagid)
370 /* We've found the index of tag_url_attributes where the
371 attributes of our tag begin. */
375 assert (first != -1);
377 /* Loop over the "interesting" attributes of this tag. In this
378 example, it will loop over "src" and "lowsrc".
380 <img src="foo.png" lowsrc="bar.png">
382 This has to be done in the outer loop so that the attributes are
383 processed in the same order in which they appear in the page.
384 This is required when converting links. */
386 for (attrind = 0; attrind < tag->nattrs; attrind++)
388 /* Find whether TAG/ATTRIND is a combination that contains a
390 char *link = tag->attrs[attrind].value;
391 const size_t size = countof (tag_url_attributes);
393 /* If you're cringing at the inefficiency of the nested loops,
394 remember that they both iterate over a very small number of
395 items. The worst-case inner loop is for the IMG tag, which
396 has three attributes. */
397 for (i = first; i < size && tag_url_attributes[i].tagid == tagid; i++)
399 if (0 == strcasecmp (tag->attrs[attrind].name,
400 tag_url_attributes[i].attr_name))
402 struct urlpos *up = append_url (link, tag, attrind, ctx);
405 int flags = tag_url_attributes[i].flags;
406 if (flags & ATTR_INLINE)
407 up->link_inline_p = 1;
408 if (flags & ATTR_HTML)
409 up->link_expect_html = 1;
416 /* Handle the BASE tag, for <base href=...>. */
419 tag_handle_base (int tagid, struct taginfo *tag, struct map_context *ctx)
421 struct urlpos *base_urlpos;
423 char *newbase = find_attr (tag, "href", &attrind);
427 base_urlpos = append_url (newbase, tag, attrind, ctx);
430 base_urlpos->ignore_when_downloading = 1;
431 base_urlpos->link_base_p = 1;
435 if (ctx->parent_base)
436 ctx->base = uri_merge (ctx->parent_base, newbase);
438 ctx->base = xstrdup (newbase);
441 /* Mark the URL found in <form action=...> for conversion. */
444 tag_handle_form (int tagid, struct taginfo *tag, struct map_context *ctx)
447 char *action = find_attr (tag, "action", &attrind);
450 struct urlpos *up = append_url (action, tag, attrind, ctx);
452 up->ignore_when_downloading = 1;
456 /* Handle the LINK tag. It requires special handling because how its
457 links will be followed in -p mode depends on the REL attribute. */
460 tag_handle_link (int tagid, struct taginfo *tag, struct map_context *ctx)
463 char *href = find_attr (tag, "href", &attrind);
465 /* All <link href="..."> link references are external, except those
466 known not to be, such as style sheet and shortcut icon:
468 <link rel="stylesheet" href="...">
469 <link rel="shortcut icon" href="...">
473 struct urlpos *up = append_url (href, tag, attrind, ctx);
476 char *rel = find_attr (tag, "rel", NULL);
478 && (0 == strcasecmp (rel, "stylesheet")
479 || 0 == strcasecmp (rel, "shortcut icon")))
480 up->link_inline_p = 1;
482 /* The external ones usually point to HTML pages, such as
483 <link rel="next" href="..."> */
484 up->link_expect_html = 1;
489 /* Handle the META tag. This requires special handling because of the
490 refresh feature and because of robot exclusion. */
493 tag_handle_meta (int tagid, struct taginfo *tag, struct map_context *ctx)
495 char *name = find_attr (tag, "name", NULL);
496 char *http_equiv = find_attr (tag, "http-equiv", NULL);
498 if (http_equiv && 0 == strcasecmp (http_equiv, "refresh"))
500 /* Some pages use a META tag to specify that the page be
501 refreshed by a new page after a given number of seconds. The
502 general format for this is:
504 <meta http-equiv=Refresh content="NUMBER; URL=index2.html">
506 So we just need to skip past the "NUMBER; URL=" garbage to
509 struct urlpos *entry;
514 char *refresh = find_attr (tag, "content", &attrind);
518 for (p = refresh; c_isdigit (*p); p++)
519 timeout = 10 * timeout + *p - '0';
523 while (c_isspace (*p))
525 if (!( c_toupper (*p) == 'U'
526 && c_toupper (*(p + 1)) == 'R'
527 && c_toupper (*(p + 2)) == 'L'
531 while (c_isspace (*p))
534 entry = append_url (p, tag, attrind, ctx);
537 entry->link_refresh_p = 1;
538 entry->refresh_timeout = timeout;
539 entry->link_expect_html = 1;
542 else if (http_equiv && 0 == strcasecmp (http_equiv, "content-type"))
544 /* Handle stuff like:
545 <meta http-equiv="Content-Type" content="text/html; charset=CHARSET"> */
548 char *content = find_attr (tag, "content", NULL);
552 mcharset = parse_charset (content);
556 /*logprintf (LOG_VERBOSE, "Meta tag charset : %s\n", quote (mcharset));*/
558 set_current_charset (mcharset);
561 else if (name && 0 == strcasecmp (name, "robots"))
563 /* Handle stuff like:
564 <meta name="robots" content="index,nofollow"> */
565 char *content = find_attr (tag, "content", NULL);
568 if (!strcasecmp (content, "none"))
569 ctx->nofollow = true;
574 /* Find the next occurrence of ',' or the end of
576 char *end = strchr (content, ',');
580 end = content + strlen (content);
581 if (!strncasecmp (content, "nofollow", end - content))
582 ctx->nofollow = true;
589 /* Dispatch the tag handler appropriate for the tag we're mapping
590 over. See known_tags[] for definition of tag handlers. */
593 collect_tags_mapper (struct taginfo *tag, void *arg)
595 struct map_context *ctx = (struct map_context *)arg;
597 /* Find the tag in our table of tags. This must not fail because
598 map_html_tags only returns tags found in interesting_tags. */
599 struct known_tag *t = hash_table_get (interesting_tags, tag->name);
602 t->handler (t->tagid, tag, ctx);
605 /* Analyze HTML tags FILE and construct a list of URLs referenced from
606 it. It merges relative links in FILE with URL. It is aware of
607 <base href=...> and does the right thing. */
610 get_urls_html (const char *file, const char *url, bool *meta_disallow_follow)
612 struct file_memory *fm;
613 struct map_context ctx;
617 fm = read_file (file);
620 logprintf (LOG_NOTQUIET, "%s: %s\n", file, strerror (errno));
623 DEBUGP (("Loaded %s (size %s).\n", file, number_to_static_string (fm->length)));
625 ctx.text = fm->content;
626 ctx.head = ctx.tail = NULL;
628 ctx.parent_base = url ? url : opt.base_href;
629 ctx.document_file = file;
630 ctx.nofollow = false;
632 if (!interesting_tags)
635 /* Specify MHT_TRIM_VALUES because of buggy HTML generators that
636 generate <a href=" foo"> instead of <a href="foo"> (browsers
637 ignore spaces as well.) If you really mean space, use &32; or
638 %20. MHT_TRIM_VALUES also causes squashing of embedded newlines,
639 e.g. in <img src="foo.[newline]html">. Such newlines are also
640 ignored by IE and Mozilla and are presumably introduced by
641 writing HTML with editors that force word wrap. */
642 flags = MHT_TRIM_VALUES;
643 if (opt.strict_comments)
644 flags |= MHT_STRICT_COMMENTS;
646 map_html_tags (fm->content, fm->length, collect_tags_mapper, &ctx, flags,
647 interesting_tags, interesting_attributes);
649 DEBUGP (("no-follow in %s: %d\n", file, ctx.nofollow));
650 if (meta_disallow_follow)
651 *meta_disallow_follow = ctx.nofollow;
653 xfree_null (ctx.base);
658 /* This doesn't really have anything to do with HTML, but it's similar
659 to get_urls_html, so we put it here. */
662 get_urls_file (const char *file)
664 struct file_memory *fm;
665 struct urlpos *head, *tail;
666 const char *text, *text_end;
669 fm = read_file (file);
672 logprintf (LOG_NOTQUIET, "%s: %s\n", file, strerror (errno));
675 DEBUGP (("Loaded %s (size %s).\n", file, number_to_static_string (fm->length)));
679 text_end = fm->content + fm->length;
680 while (text < text_end)
684 struct urlpos *entry;
687 const char *line_beg = text;
688 const char *line_end = memchr (text, '\n', text_end - text);
695 /* Strip whitespace from the beginning and end of line. */
696 while (line_beg < line_end && c_isspace (*line_beg))
698 while (line_end > line_beg && c_isspace (*(line_end - 1)))
701 if (line_beg == line_end)
704 /* The URL is in the [line_beg, line_end) region. */
706 /* We must copy the URL to a zero-terminated string, and we
707 can't use alloca because we're in a loop. *sigh*. */
708 url_text = strdupdelim (line_beg, line_end);
712 /* Merge opt.base_href with URL. */
713 char *merged = uri_merge (opt.base_href, url_text);
718 set_ugly_no_encode (true);
719 url = url_parse (url_text, &up_error_code);
720 set_ugly_no_encode (false);
723 logprintf (LOG_NOTQUIET, _("%s: Invalid URL %s: %s\n"),
724 file, url_text, url_error (up_error_code));
730 entry = xnew0 (struct urlpos);
744 cleanup_html_url (void)
746 /* Destroy the hash tables. The hash table keys and values are not
747 allocated by this code, so we don't need to free them here. */
748 if (interesting_tags)
749 hash_table_destroy (interesting_tags);
750 if (interesting_attributes)
751 hash_table_destroy (interesting_attributes);