1 /* Collect URLs from HTML source.
2 Copyright (C) 1998, 2000, 2001, 2002, 2003 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. */
43 #include "html-parse.h"
55 typedef void (*tag_handler_t) PARAMS ((int, struct taginfo *,
56 struct map_context *));
58 #define DECLARE_TAG_HANDLER(fun) \
59 static void fun PARAMS ((int, struct taginfo *, struct map_context *))
61 DECLARE_TAG_HANDLER (tag_find_urls);
62 DECLARE_TAG_HANDLER (tag_handle_base);
63 DECLARE_TAG_HANDLER (tag_handle_form);
64 DECLARE_TAG_HANDLER (tag_handle_link);
65 DECLARE_TAG_HANDLER (tag_handle_meta);
91 /* The list of known tags and functions used for handling them. Most
92 tags are simply harvested for URLs. */
93 static struct known_tag {
96 tag_handler_t handler;
98 { TAG_A, "a", tag_find_urls },
99 { TAG_APPLET, "applet", tag_find_urls },
100 { TAG_AREA, "area", tag_find_urls },
101 { TAG_BASE, "base", tag_handle_base },
102 { TAG_BGSOUND, "bgsound", tag_find_urls },
103 { TAG_BODY, "body", tag_find_urls },
104 { TAG_EMBED, "embed", tag_find_urls },
105 { TAG_FIG, "fig", tag_find_urls },
106 { TAG_FORM, "form", tag_handle_form },
107 { TAG_FRAME, "frame", tag_find_urls },
108 { TAG_IFRAME, "iframe", tag_find_urls },
109 { TAG_IMG, "img", tag_find_urls },
110 { TAG_INPUT, "input", tag_find_urls },
111 { TAG_LAYER, "layer", tag_find_urls },
112 { TAG_LINK, "link", tag_handle_link },
113 { TAG_META, "meta", tag_handle_meta },
114 { TAG_OVERLAY, "overlay", tag_find_urls },
115 { TAG_SCRIPT, "script", tag_find_urls },
116 { TAG_TABLE, "table", tag_find_urls },
117 { TAG_TD, "td", tag_find_urls },
118 { TAG_TH, "th", tag_find_urls }
121 /* tag_url_attributes documents which attributes of which tags contain
122 URLs to harvest. It is used by tag_find_urls. */
124 /* Defines for the FLAGS. */
126 /* The link is "inline", i.e. needs to be retrieved for this document
127 to be correctly rendered. Inline links include inlined images,
128 stylesheets, children frames, etc. */
129 #define ATTR_INLINE 1
131 /* The link is expected to yield HTML contents. It's important not to
132 try to follow HTML obtained by following e.g. <img src="...">
133 regardless of content-type. Doing this causes infinite loops for
134 "images" that return non-404 error pages with links to the same
138 /* For tags handled by tag_find_urls: attributes that contain URLs to
142 const char *attr_name;
144 } tag_url_attributes[] = {
145 { TAG_A, "href", ATTR_HTML },
146 { TAG_APPLET, "code", ATTR_INLINE },
147 { TAG_AREA, "href", ATTR_HTML },
148 { TAG_BGSOUND, "src", ATTR_INLINE },
149 { TAG_BODY, "background", ATTR_INLINE },
150 { TAG_EMBED, "href", ATTR_HTML },
151 { TAG_EMBED, "src", ATTR_INLINE | ATTR_HTML },
152 { TAG_FIG, "src", ATTR_INLINE },
153 { TAG_FRAME, "src", ATTR_INLINE | ATTR_HTML },
154 { TAG_IFRAME, "src", ATTR_INLINE | ATTR_HTML },
155 { TAG_IMG, "href", ATTR_INLINE },
156 { TAG_IMG, "lowsrc", ATTR_INLINE },
157 { TAG_IMG, "src", ATTR_INLINE },
158 { TAG_INPUT, "src", ATTR_INLINE },
159 { TAG_LAYER, "src", ATTR_INLINE | ATTR_HTML },
160 { TAG_OVERLAY, "src", ATTR_INLINE | ATTR_HTML },
161 { TAG_SCRIPT, "src", ATTR_INLINE },
162 { TAG_TABLE, "background", ATTR_INLINE },
163 { TAG_TD, "background", ATTR_INLINE },
164 { TAG_TH, "background", ATTR_INLINE }
167 /* The lists of interesting tags and attributes are built dynamically,
168 from the information above. However, some places in the code refer
169 to the attributes not mentioned here. We add them manually. */
170 static const char *additional_attributes[] = {
171 "rel", /* used by tag_handle_link */
172 "http-equiv", /* used by tag_handle_meta */
173 "name", /* used by tag_handle_meta */
174 "content", /* used by tag_handle_meta */
175 "action" /* used by tag_handle_form */
178 struct hash_table *interesting_tags;
179 struct hash_table *interesting_attributes;
182 init_interesting (void)
184 /* Init the variables interesting_tags and interesting_attributes
185 that are used by the HTML parser to know which tags and
186 attributes we're interested in. We initialize this only once,
187 for performance reasons.
189 Here we also make sure that what we put in interesting_tags
190 matches the user's preferences as specified through --ignore-tags
191 and --follow-tags. */
194 interesting_tags = make_nocase_string_hash_table (countof (known_tags));
196 /* First, add all the tags we know hot to handle, mapped to their
197 respective entries in known_tags. */
198 for (i = 0; i < countof (known_tags); i++)
199 hash_table_put (interesting_tags, known_tags[i].name, known_tags + i);
201 /* Then remove the tags ignored through --ignore-tags. */
205 for (ignored = opt.ignore_tags; *ignored; ignored++)
206 hash_table_remove (interesting_tags, *ignored);
209 /* If --follow-tags is specified, use only those tags. */
212 /* Create a new table intersecting --follow-tags and known_tags,
213 and use it as interesting_tags. */
214 struct hash_table *intersect = make_nocase_string_hash_table (0);
216 for (followed = opt.follow_tags; *followed; followed++)
218 struct known_tag *t = hash_table_get (interesting_tags, *followed);
220 continue; /* ignore unknown --follow-tags entries. */
221 hash_table_put (intersect, *followed, t);
223 hash_table_destroy (interesting_tags);
224 interesting_tags = intersect;
227 /* Add the attributes we care about. */
228 interesting_attributes = make_nocase_string_hash_table (10);
229 for (i = 0; i < countof (additional_attributes); i++)
230 hash_table_put (interesting_attributes, additional_attributes[i], "1");
231 for (i = 0; i < countof (tag_url_attributes); i++)
232 hash_table_put (interesting_attributes,
233 tag_url_attributes[i].attr_name, "1");
236 /* Find the value of attribute named NAME in the taginfo TAG. If the
237 attribute is not present, return NULL. If ATTRIND is non-NULL, the
238 index of the attribute in TAG will be stored there. */
241 find_attr (struct taginfo *tag, const char *name, int *attrind)
244 for (i = 0; i < tag->nattrs; i++)
245 if (!strcasecmp (tag->attrs[i].name, name))
249 return tag->attrs[i].value;
255 char *text; /* HTML text. */
256 char *base; /* Base URI of the document, possibly
257 changed through <base href=...>. */
258 const char *parent_base; /* Base of the current document. */
259 const char *document_file; /* File name of this document. */
260 int nofollow; /* whether NOFOLLOW was specified in a
261 <meta name=robots> tag. */
263 struct urlpos *head, *tail; /* List of URLs that is being
267 /* Append LINK_URI to the urlpos structure that is being built.
269 LINK_URI will be merged with the current document base. TAG and
270 ATTRIND are the necessary context to store the position and
273 static struct urlpos *
274 append_url (const char *link_uri,
275 struct taginfo *tag, int attrind, struct map_context *ctx)
277 int link_has_scheme = url_has_scheme (link_uri);
278 struct urlpos *newel;
279 const char *base = ctx->base ? ctx->base : ctx->parent_base;
284 DEBUGP (("%s: no base, merge will use \"%s\".\n",
285 ctx->document_file, link_uri));
287 if (!link_has_scheme)
289 /* Base URL is unavailable, and the link does not have a
290 location attached to it -- we have to give up. Since
291 this can only happen when using `--force-html -i', print
293 logprintf (LOG_NOTQUIET,
294 _("%s: Cannot resolve incomplete link %s.\n"),
295 ctx->document_file, link_uri);
299 url = url_parse (link_uri, NULL);
302 DEBUGP (("%s: link \"%s\" doesn't parse.\n",
303 ctx->document_file, link_uri));
309 /* Merge BASE with LINK_URI, but also make sure the result is
310 canonicalized, i.e. that "../" have been resolved.
311 (parse_url will do that for us.) */
313 char *complete_uri = uri_merge (base, link_uri);
315 DEBUGP (("%s: merge(\"%s\", \"%s\") -> %s\n",
316 ctx->document_file, base, link_uri, complete_uri));
318 url = url_parse (complete_uri, NULL);
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)
366 for (i = 0; i < countof (tag_url_attributes); i++)
367 if (tag_url_attributes[i].tagid == tagid)
369 /* We've found the index of tag_url_attributes where the
370 attributes of our tag begin. */
374 assert (first != -1);
376 /* Loop over the "interesting" attributes of this tag. In this
377 example, it will loop over "src" and "lowsrc".
379 <img src="foo.png" lowsrc="bar.png">
381 This has to be done in the outer loop so that the attributes are
382 processed in the same order in which they appear in the page.
383 This is required when converting links. */
385 for (attrind = 0; attrind < tag->nattrs; attrind++)
387 /* Find whether TAG/ATTRIND is a combination that contains a
389 char *link = tag->attrs[attrind].value;
390 const int size = countof (tag_url_attributes);
392 /* If you're cringing at the inefficiency of the nested loops,
393 remember that they both iterate over a very small number of
394 items. The worst-case inner loop is for the IMG tag, which
395 has three attributes. */
396 for (i = first; i < size && tag_url_attributes[i].tagid == tagid; i++)
398 if (0 == strcasecmp (tag->attrs[attrind].name,
399 tag_url_attributes[i].attr_name))
401 struct urlpos *up = append_url (link, tag, attrind, ctx);
404 int flags = tag_url_attributes[i].flags;
405 if (flags & ATTR_INLINE)
406 up->link_inline_p = 1;
407 if (flags & ATTR_HTML)
408 up->link_expect_html = 1;
415 /* Handle the BASE tag, for <base href=...>. */
418 tag_handle_base (int tagid, struct taginfo *tag, struct map_context *ctx)
420 struct urlpos *base_urlpos;
422 char *newbase = find_attr (tag, "href", &attrind);
426 base_urlpos = append_url (newbase, tag, attrind, ctx);
429 base_urlpos->ignore_when_downloading = 1;
430 base_urlpos->link_base_p = 1;
434 if (ctx->parent_base)
435 ctx->base = uri_merge (ctx->parent_base, newbase);
437 ctx->base = xstrdup (newbase);
440 /* Mark the URL found in <form action=...> for conversion. */
443 tag_handle_form (int tagid, struct taginfo *tag, struct map_context *ctx)
446 char *action = find_attr (tag, "action", &attrind);
449 struct urlpos *up = append_url (action, tag, attrind, ctx);
451 up->ignore_when_downloading = 1;
455 /* Handle the LINK tag. It requires special handling because how its
456 links will be followed in -p mode depends on the REL attribute. */
459 tag_handle_link (int tagid, struct taginfo *tag, struct map_context *ctx)
462 char *href = find_attr (tag, "href", &attrind);
464 /* All <link href="..."> link references are external, except those
465 known not to be, such as style sheet and shortcut icon:
467 <link rel="stylesheet" href="...">
468 <link rel="shortcut icon" href="...">
472 struct urlpos *up = append_url (href, tag, attrind, ctx);
475 char *rel = find_attr (tag, "rel", NULL);
477 && (0 == strcasecmp (rel, "stylesheet")
478 || 0 == strcasecmp (rel, "shortcut icon")))
479 up->link_inline_p = 1;
484 /* Handle the META tag. This requires special handling because of the
485 refresh feature and because of robot exclusion. */
488 tag_handle_meta (int tagid, struct taginfo *tag, struct map_context *ctx)
490 char *name = find_attr (tag, "name", NULL);
491 char *http_equiv = find_attr (tag, "http-equiv", NULL);
493 if (http_equiv && 0 == strcasecmp (http_equiv, "refresh"))
495 /* Some pages use a META tag to specify that the page be
496 refreshed by a new page after a given number of seconds. The
497 general format for this is:
499 <meta http-equiv=Refresh content="NUMBER; URL=index2.html">
501 So we just need to skip past the "NUMBER; URL=" garbage to
504 struct urlpos *entry;
509 char *refresh = find_attr (tag, "content", &attrind);
513 for (p = refresh; ISDIGIT (*p); p++)
514 timeout = 10 * timeout + *p - '0';
520 if (!( TOUPPER (*p) == 'U'
521 && TOUPPER (*(p + 1)) == 'R'
522 && TOUPPER (*(p + 2)) == 'L'
529 entry = append_url (p, tag, attrind, ctx);
532 entry->link_refresh_p = 1;
533 entry->refresh_timeout = timeout;
534 entry->link_expect_html = 1;
537 else if (name && 0 == strcasecmp (name, "robots"))
539 /* Handle stuff like:
540 <meta name="robots" content="index,nofollow"> */
541 char *content = find_attr (tag, "content", NULL);
544 if (!strcasecmp (content, "none"))
550 /* Find the next occurrence of ',' or the end of
552 char *end = strchr (content, ',');
556 end = content + strlen (content);
557 if (!strncasecmp (content, "nofollow", end - content))
565 /* Dispatch the tag handler appropriate for the tag we're mapping
566 over. See known_tags[] for definition of tag handlers. */
569 collect_tags_mapper (struct taginfo *tag, void *arg)
571 struct map_context *ctx = (struct map_context *)arg;
573 /* Find the tag in our table of tags. This must not fail because
574 map_html_tags only returns tags found in interesting_tags. */
575 struct known_tag *t = hash_table_get (interesting_tags, tag->name);
578 t->handler (t->tagid, tag, ctx);
581 /* Analyze HTML tags FILE and construct a list of URLs referenced from
582 it. It merges relative links in FILE with URL. It is aware of
583 <base href=...> and does the right thing. */
586 get_urls_html (const char *file, const char *url, int *meta_disallow_follow)
588 struct file_memory *fm;
589 struct map_context ctx;
593 fm = read_file (file);
596 logprintf (LOG_NOTQUIET, "%s: %s\n", file, strerror (errno));
599 DEBUGP (("Loaded %s (size %ld).\n", file, fm->length));
601 ctx.text = fm->content;
602 ctx.head = ctx.tail = NULL;
604 ctx.parent_base = url ? url : opt.base_href;
605 ctx.document_file = file;
608 if (!interesting_tags)
611 /* Specify MHT_TRIM_VALUES because of buggy HTML generators that
612 generate <a href=" foo"> instead of <a href="foo"> (Netscape
613 ignores spaces as well.) If you really mean space, use &32; or
615 flags = MHT_TRIM_VALUES;
616 if (opt.strict_comments)
617 flags |= MHT_STRICT_COMMENTS;
619 map_html_tags (fm->content, fm->length, collect_tags_mapper, &ctx, flags,
620 interesting_tags, interesting_attributes);
622 DEBUGP (("no-follow in %s: %d\n", file, ctx.nofollow));
623 if (meta_disallow_follow)
624 *meta_disallow_follow = ctx.nofollow;
626 xfree_null (ctx.base);
631 /* This doesn't really have anything to do with HTML, but it's similar
632 to get_urls_html, so we put it here. */
635 get_urls_file (const char *file)
637 struct file_memory *fm;
638 struct urlpos *head, *tail;
639 const char *text, *text_end;
642 fm = read_file (file);
645 logprintf (LOG_NOTQUIET, "%s: %s\n", file, strerror (errno));
648 DEBUGP (("Loaded %s (size %ld).\n", file, fm->length));
652 text_end = fm->content + fm->length;
653 while (text < text_end)
657 struct urlpos *entry;
660 const char *line_beg = text;
661 const char *line_end = memchr (text, '\n', text_end - text);
668 /* Strip whitespace from the beginning and end of line. */
669 while (line_beg < line_end && ISSPACE (*line_beg))
671 while (line_end > line_beg && ISSPACE (*(line_end - 1)))
674 if (line_beg == line_end)
677 /* The URL is in the [line_beg, line_end) region. */
679 /* We must copy the URL to a zero-terminated string, and we
680 can't use alloca because we're in a loop. *sigh*. */
681 url_text = strdupdelim (line_beg, line_end);
685 /* Merge opt.base_href with URL. */
686 char *merged = uri_merge (opt.base_href, url_text);
691 url = url_parse (url_text, &up_error_code);
694 logprintf (LOG_NOTQUIET, "%s: Invalid URL %s: %s\n",
695 file, url_text, url_error (up_error_code));
701 entry = xnew0 (struct urlpos);
716 cleanup_html_url (void)
718 /* Destroy the hash tables. The hash table keys and values are not
719 allocated by this code, so we don't need to free them here. */
720 if (interesting_tags)
721 hash_table_destroy (interesting_tags);
722 if (interesting_attributes)
723 hash_table_destroy (interesting_attributes);