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"
48 typedef void (*tag_handler_t) (int, struct taginfo *, struct map_context *);
50 #define DECLARE_TAG_HANDLER(fun) \
51 static void fun (int, struct taginfo *, struct map_context *)
53 DECLARE_TAG_HANDLER (tag_find_urls);
54 DECLARE_TAG_HANDLER (tag_handle_base);
55 DECLARE_TAG_HANDLER (tag_handle_form);
56 DECLARE_TAG_HANDLER (tag_handle_link);
57 DECLARE_TAG_HANDLER (tag_handle_meta);
84 /* The list of known tags and functions used for handling them. Most
85 tags are simply harvested for URLs. */
86 static struct known_tag {
89 tag_handler_t handler;
91 { TAG_A, "a", tag_find_urls },
92 { TAG_APPLET, "applet", tag_find_urls },
93 { TAG_AREA, "area", tag_find_urls },
94 { TAG_BASE, "base", tag_handle_base },
95 { TAG_BGSOUND, "bgsound", tag_find_urls },
96 { TAG_BODY, "body", tag_find_urls },
97 { TAG_EMBED, "embed", tag_find_urls },
98 { TAG_FIG, "fig", tag_find_urls },
99 { TAG_FORM, "form", tag_handle_form },
100 { TAG_FRAME, "frame", tag_find_urls },
101 { TAG_IFRAME, "iframe", tag_find_urls },
102 { TAG_IMG, "img", tag_find_urls },
103 { TAG_INPUT, "input", tag_find_urls },
104 { TAG_LAYER, "layer", tag_find_urls },
105 { TAG_LINK, "link", tag_handle_link },
106 { TAG_META, "meta", tag_handle_meta },
107 { TAG_OBJECT, "object", tag_find_urls },
108 { TAG_OVERLAY, "overlay", tag_find_urls },
109 { TAG_SCRIPT, "script", tag_find_urls },
110 { TAG_TABLE, "table", tag_find_urls },
111 { TAG_TD, "td", tag_find_urls },
112 { TAG_TH, "th", tag_find_urls }
115 /* tag_url_attributes documents which attributes of which tags contain
116 URLs to harvest. It is used by tag_find_urls. */
118 /* Defines for the FLAGS. */
120 /* The link is "inline", i.e. needs to be retrieved for this document
121 to be correctly rendered. Inline links include inlined images,
122 stylesheets, children frames, etc. */
123 #define ATTR_INLINE 1
125 /* The link is expected to yield HTML contents. It's important not to
126 try to follow HTML obtained by following e.g. <img src="...">
127 regardless of content-type. Doing this causes infinite loops for
128 "images" that return non-404 error pages with links to the same
132 /* For tags handled by tag_find_urls: attributes that contain URLs to
136 const char *attr_name;
138 } tag_url_attributes[] = {
139 { TAG_A, "href", ATTR_HTML },
140 { TAG_APPLET, "code", ATTR_INLINE },
141 { TAG_AREA, "href", ATTR_HTML },
142 { TAG_BGSOUND, "src", ATTR_INLINE },
143 { TAG_BODY, "background", ATTR_INLINE },
144 { TAG_EMBED, "href", ATTR_HTML },
145 { TAG_EMBED, "src", ATTR_INLINE | ATTR_HTML },
146 { TAG_FIG, "src", ATTR_INLINE },
147 { TAG_FRAME, "src", ATTR_INLINE | ATTR_HTML },
148 { TAG_IFRAME, "src", ATTR_INLINE | ATTR_HTML },
149 { TAG_IMG, "href", ATTR_INLINE },
150 { TAG_IMG, "lowsrc", ATTR_INLINE },
151 { TAG_IMG, "src", ATTR_INLINE },
152 { TAG_INPUT, "src", ATTR_INLINE },
153 { TAG_LAYER, "src", ATTR_INLINE | ATTR_HTML },
154 { TAG_OBJECT, "data", ATTR_INLINE },
155 { TAG_OVERLAY, "src", ATTR_INLINE | ATTR_HTML },
156 { TAG_SCRIPT, "src", ATTR_INLINE },
157 { TAG_TABLE, "background", ATTR_INLINE },
158 { TAG_TD, "background", ATTR_INLINE },
159 { TAG_TH, "background", ATTR_INLINE }
162 /* The lists of interesting tags and attributes are built dynamically,
163 from the information above. However, some places in the code refer
164 to the attributes not mentioned here. We add them manually. */
165 static const char *additional_attributes[] = {
166 "rel", /* used by tag_handle_link */
167 "http-equiv", /* used by tag_handle_meta */
168 "name", /* used by tag_handle_meta */
169 "content", /* used by tag_handle_meta */
170 "action", /* used by tag_handle_form */
171 "style" /* used by check_style_attr */
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;
250 /* used for calls to append_url */
251 #define ATTR_POS(tag, attrind, ctx) \
252 (tag->attrs[attrind].value_raw_beginning - ctx->text)
253 #define ATTR_SIZE(tag, attrind) \
254 (tag->attrs[attrind].value_raw_size)
256 /* Append LINK_URI to the urlpos structure that is being built.
258 LINK_URI will be merged with the current document base.
262 append_url (const char *link_uri, int position, int size,
263 struct map_context *ctx)
265 int link_has_scheme = url_has_scheme (link_uri);
266 struct urlpos *newel;
267 const char *base = ctx->base ? ctx->base : ctx->parent_base;
272 DEBUGP (("%s: no base, merge will use \"%s\".\n",
273 ctx->document_file, link_uri));
275 if (!link_has_scheme)
277 /* Base URL is unavailable, and the link does not have a
278 location attached to it -- we have to give up. Since
279 this can only happen when using `--force-html -i', print
281 logprintf (LOG_NOTQUIET,
282 _("%s: Cannot resolve incomplete link %s.\n"),
283 ctx->document_file, link_uri);
287 url = url_parse (link_uri, NULL);
290 DEBUGP (("%s: link \"%s\" doesn't parse.\n",
291 ctx->document_file, link_uri));
297 /* Merge BASE with LINK_URI, but also make sure the result is
298 canonicalized, i.e. that "../" have been resolved.
299 (parse_url will do that for us.) */
301 char *complete_uri = uri_merge (base, link_uri);
303 DEBUGP (("%s: merge(\"%s\", \"%s\") -> %s\n",
304 ctx->document_file, base, link_uri, complete_uri));
306 url = url_parse (complete_uri, NULL);
309 DEBUGP (("%s: merged link \"%s\" doesn't parse.\n",
310 ctx->document_file, complete_uri));
311 xfree (complete_uri);
314 xfree (complete_uri);
317 DEBUGP (("appending \"%s\" to urlpos.\n", url->url));
319 newel = xnew0 (struct urlpos);
321 newel->pos = position;
324 /* A URL is relative if the host is not named, and the name does not
326 if (!link_has_scheme && *link_uri != '/')
327 newel->link_relative_p = 1;
328 else if (link_has_scheme)
329 newel->link_complete_p = 1;
333 ctx->tail->next = newel;
337 ctx->tail = ctx->head = newel;
343 check_style_attr (struct taginfo *tag, struct map_context *ctx)
346 char *style = find_attr (tag, "style", &attrind);
350 /* raw pos and raw size include the quotes, hence the +1 -2 */
351 get_urls_css (ctx, ATTR_POS(tag,attrind,ctx)+1, ATTR_SIZE(tag,attrind)-2);
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, ATTR_POS(tag,attrind,ctx),
403 ATTR_SIZE(tag,attrind), ctx);
406 int flags = tag_url_attributes[i].flags;
407 if (flags & ATTR_INLINE)
408 up->link_inline_p = 1;
409 if (flags & ATTR_HTML)
410 up->link_expect_html = 1;
417 /* Handle the BASE tag, for <base href=...>. */
420 tag_handle_base (int tagid, struct taginfo *tag, struct map_context *ctx)
422 struct urlpos *base_urlpos;
424 char *newbase = find_attr (tag, "href", &attrind);
428 base_urlpos = append_url (newbase, ATTR_POS(tag,attrind,ctx),
429 ATTR_SIZE(tag,attrind), ctx);
432 base_urlpos->ignore_when_downloading = 1;
433 base_urlpos->link_base_p = 1;
437 if (ctx->parent_base)
438 ctx->base = uri_merge (ctx->parent_base, newbase);
440 ctx->base = xstrdup (newbase);
443 /* Mark the URL found in <form action=...> for conversion. */
446 tag_handle_form (int tagid, struct taginfo *tag, struct map_context *ctx)
449 char *action = find_attr (tag, "action", &attrind);
453 struct urlpos *up = append_url (action, ATTR_POS(tag,attrind,ctx),
454 ATTR_SIZE(tag,attrind), ctx);
456 up->ignore_when_downloading = 1;
460 /* Handle the LINK tag. It requires special handling because how its
461 links will be followed in -p mode depends on the REL attribute. */
464 tag_handle_link (int tagid, struct taginfo *tag, struct map_context *ctx)
467 char *href = find_attr (tag, "href", &attrind);
469 /* All <link href="..."> link references are external, except those
470 known not to be, such as style sheet and shortcut icon:
472 <link rel="stylesheet" href="...">
473 <link rel="shortcut icon" href="...">
477 struct urlpos *up = append_url (href, ATTR_POS(tag,attrind,ctx),
478 ATTR_SIZE(tag,attrind), ctx);
481 char *rel = find_attr (tag, "rel", NULL);
484 if (0 == strcasecmp (rel, "stylesheet"))
486 up->link_inline_p = 1;
487 up->link_expect_css = 1;
489 else if (0 == strcasecmp (rel, "shortcut icon"))
491 up->link_inline_p = 1;
495 /* The external ones usually point to HTML pages, such as
496 <link rel="next" href="..."> */
497 up->link_expect_html = 1;
502 /* Handle the META tag. This requires special handling because of the
503 refresh feature and because of robot exclusion. */
506 tag_handle_meta (int tagid, struct taginfo *tag, struct map_context *ctx)
508 char *name = find_attr (tag, "name", NULL);
509 char *http_equiv = find_attr (tag, "http-equiv", NULL);
511 if (http_equiv && 0 == strcasecmp (http_equiv, "refresh"))
513 /* Some pages use a META tag to specify that the page be
514 refreshed by a new page after a given number of seconds. The
515 general format for this is:
517 <meta http-equiv=Refresh content="NUMBER; URL=index2.html">
519 So we just need to skip past the "NUMBER; URL=" garbage to
522 struct urlpos *entry;
527 char *refresh = find_attr (tag, "content", &attrind);
531 for (p = refresh; c_isdigit (*p); p++)
532 timeout = 10 * timeout + *p - '0';
536 while (c_isspace (*p))
538 if (!( c_toupper (*p) == 'U'
539 && c_toupper (*(p + 1)) == 'R'
540 && c_toupper (*(p + 2)) == 'L'
544 while (c_isspace (*p))
547 entry = append_url (p, ATTR_POS(tag,attrind,ctx),
548 ATTR_SIZE(tag,attrind), ctx);
551 entry->link_refresh_p = 1;
552 entry->refresh_timeout = timeout;
553 entry->link_expect_html = 1;
556 else if (name && 0 == strcasecmp (name, "robots"))
558 /* Handle stuff like:
559 <meta name="robots" content="index,nofollow"> */
560 char *content = find_attr (tag, "content", NULL);
563 if (!strcasecmp (content, "none"))
564 ctx->nofollow = true;
569 /* Find the next occurrence of ',' or the end of
571 char *end = strchr (content, ',');
575 end = content + strlen (content);
576 if (!strncasecmp (content, "nofollow", end - content))
577 ctx->nofollow = true;
584 /* Dispatch the tag handler appropriate for the tag we're mapping
585 over. See known_tags[] for definition of tag handlers. */
588 collect_tags_mapper (struct taginfo *tag, void *arg)
590 struct map_context *ctx = (struct map_context *)arg;
592 /* Find the tag in our table of tags. This must not fail because
593 map_html_tags only returns tags found in interesting_tags.
595 I've changed this for now, I'm passing NULL as interesting_tags
596 to map_html_tags. This way we can check all tags for a style
599 struct known_tag *t = hash_table_get (interesting_tags, tag->name);
602 t->handler (t->tagid, tag, ctx);
604 check_style_attr (tag, ctx);
606 if (tag->end_tag_p && (0 == strcasecmp (tag->name, "style")) &&
607 tag->contents_begin && tag->contents_end)
610 get_urls_css (ctx, tag->contents_begin - ctx->text,
611 tag->contents_end - tag->contents_begin);
615 /* Analyze HTML tags FILE and construct a list of URLs referenced from
616 it. It merges relative links in FILE with URL. It is aware of
617 <base href=...> and does the right thing. */
620 get_urls_html (const char *file, const char *url, bool *meta_disallow_follow)
622 struct file_memory *fm;
623 struct map_context ctx;
627 fm = read_file (file);
630 logprintf (LOG_NOTQUIET, "%s: %s\n", file, strerror (errno));
633 DEBUGP (("Loaded %s (size %s).\n", file, number_to_static_string (fm->length)));
635 ctx.text = fm->content;
636 ctx.head = ctx.tail = NULL;
638 ctx.parent_base = url ? url : opt.base_href;
639 ctx.document_file = file;
640 ctx.nofollow = false;
642 if (!interesting_tags)
645 /* Specify MHT_TRIM_VALUES because of buggy HTML generators that
646 generate <a href=" foo"> instead of <a href="foo"> (browsers
647 ignore spaces as well.) If you really mean space, use &32; or
648 %20. MHT_TRIM_VALUES also causes squashing of embedded newlines,
649 e.g. in <img src="foo.[newline]html">. Such newlines are also
650 ignored by IE and Mozilla and are presumably introduced by
651 writing HTML with editors that force word wrap. */
652 flags = MHT_TRIM_VALUES;
653 if (opt.strict_comments)
654 flags |= MHT_STRICT_COMMENTS;
656 /* the NULL here used to be interesting_tags */
657 map_html_tags (fm->content, fm->length, collect_tags_mapper, &ctx, flags,
658 NULL, interesting_attributes);
660 DEBUGP (("no-follow in %s: %d\n", file, ctx.nofollow));
661 if (meta_disallow_follow)
662 *meta_disallow_follow = ctx.nofollow;
664 xfree_null (ctx.base);
669 /* This doesn't really have anything to do with HTML, but it's similar
670 to get_urls_html, so we put it here. */
673 get_urls_file (const char *file)
675 struct file_memory *fm;
676 struct urlpos *head, *tail;
677 const char *text, *text_end;
680 fm = read_file (file);
683 logprintf (LOG_NOTQUIET, "%s: %s\n", file, strerror (errno));
686 DEBUGP (("Loaded %s (size %s).\n", file, number_to_static_string (fm->length)));
690 text_end = fm->content + fm->length;
691 while (text < text_end)
695 struct urlpos *entry;
698 const char *line_beg = text;
699 const char *line_end = memchr (text, '\n', text_end - text);
706 /* Strip whitespace from the beginning and end of line. */
707 while (line_beg < line_end && c_isspace (*line_beg))
709 while (line_end > line_beg && c_isspace (*(line_end - 1)))
712 if (line_beg == line_end)
715 /* The URL is in the [line_beg, line_end) region. */
717 /* We must copy the URL to a zero-terminated string, and we
718 can't use alloca because we're in a loop. *sigh*. */
719 url_text = strdupdelim (line_beg, line_end);
723 /* Merge opt.base_href with URL. */
724 char *merged = uri_merge (opt.base_href, url_text);
729 url = url_parse (url_text, &up_error_code);
732 char *error = url_error (url_text, up_error_code);
733 logprintf (LOG_NOTQUIET, _("%s: Invalid URL %s: %s\n"),
734 file, url_text, error);
741 entry = xnew0 (struct urlpos);
755 cleanup_html_url (void)
757 /* Destroy the hash tables. The hash table keys and values are not
758 allocated by this code, so we don't need to free them here. */
759 if (interesting_tags)
760 hash_table_destroy (interesting_tags);
761 if (interesting_attributes)
762 hash_table_destroy (interesting_attributes);