1 /* Collect URLs from HTML source.
2 Copyright (C) 1998, 2000 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. */
33 #include "html-parse.h"
41 enum tag_category { TC_LINK, TC_SPEC };
43 /* Here we try to categorize the known tags. Each tag has its ID and
44 cetegory. Category TC_LINK means that one or more of its
45 attributes contain links that should be retrieved. TC_SPEC means
46 that the tag is specific in some way, and has to be handled
50 enum tag_category category;
55 { "applet", TC_LINK },
61 { "bgsound", TC_LINK },
71 { "iframe", TC_LINK },
82 #define TAG_OVERLAY 15
83 { "overlay", TC_LINK },
85 { "script", TC_LINK },
95 /* Flags for specific url-attr pairs handled through TC_LINK: */
97 /* This tag points to an external document not necessary for rendering this
98 document (i.e. it's not an inlined image, stylesheet, etc.). */
102 /* For tags handled by TC_LINK: attributes that contain URLs to
106 const char *attr_name;
108 } url_tag_attr_map[] = {
109 { TAG_A, "href", AF_EXTERNAL },
110 { TAG_APPLET, "code", 0 },
111 { TAG_AREA, "href", AF_EXTERNAL },
112 { TAG_BGSOUND, "src", 0 },
113 { TAG_BODY, "background", 0 },
114 { TAG_EMBED, "src", 0 },
115 { TAG_FIG, "src", 0 },
116 { TAG_FRAME, "src", 0 },
117 { TAG_IFRAME, "src", 0 },
118 { TAG_IMG, "href", 0 },
119 { TAG_IMG, "lowsrc", 0 },
120 { TAG_IMG, "src", 0 },
121 { TAG_INPUT, "src", 0 },
122 { TAG_LAYER, "src", 0 },
123 { TAG_OVERLAY, "src", 0 },
124 { TAG_SCRIPT, "src", 0 },
125 { TAG_TABLE, "background", 0 },
126 { TAG_TD, "background", 0 },
127 { TAG_TH, "background", 0 }
130 /* The lists of interesting tags and attributes are built dynamically,
131 from the information above. However, some places in the code refer
132 to the attributes not mentioned here. We add them manually. */
133 static const char *additional_attributes[] = {
134 "rel", /* for TAG_LINK */
135 "http-equiv", /* for TAG_META */
136 "name", /* for TAG_META */
137 "content" /* for TAG_META */
140 static const char **interesting_tags;
141 static const char **interesting_attributes;
144 init_interesting (void)
146 /* Init the variables interesting_tags and interesting_attributes
147 that are used by the HTML parser to know which tags and
148 attributes we're interested in. We initialize this only once,
149 for performance reasons.
151 Here we also make sure that what we put in interesting_tags
152 matches the user's preferences as specified through --ignore-tags
153 and --follow-tags. */
157 int size = ARRAY_SIZE (known_tags);
158 interesting_tags = (const char **)xmalloc ((size + 1) * sizeof (char *));
160 for (i = 0; i < size; i++)
162 const char *name = known_tags[i].name;
164 /* Normally here we could say:
165 interesting_tags[i] = name;
166 But we need to respect the settings of --ignore-tags and
167 --follow-tags, so the code gets a bit hairier. */
171 /* --ignore-tags was specified. Do not match these
172 specific tags. --ignore-tags takes precedence over
173 --follow-tags, so we process --ignore first and fall
174 through if there's no match. */
176 for (j = 0; opt.ignore_tags[j] != NULL; j++)
177 /* Loop through all the tags this user doesn't care about. */
178 if (strcasecmp(opt.ignore_tags[j], name) == EQ)
189 /* --follow-tags was specified. Only match these specific tags, so
190 continue back to top of for if we don't match one of them. */
192 for (j = 0; opt.follow_tags[j] != NULL; j++)
193 /* Loop through all the tags this user cares about. */
194 if (strcasecmp(opt.follow_tags[j], name) == EQ)
200 continue; /* wasn't one of the explicitly desired tags */
203 /* If we get to here, --follow-tags isn't being used or the
204 tag is among the ones that are followed, and --ignore-tags,
205 if specified, didn't include this tag, so it's an
206 "interesting" one. */
207 interesting_tags[ind++] = name;
209 interesting_tags[ind] = NULL;
212 /* The same for attributes, except we loop through url_tag_attr_map.
213 Here we also need to make sure that the list of attributes is
214 unique, and to include the attributes from additional_attributes. */
217 const char **att = xmalloc ((ARRAY_SIZE (additional_attributes) + 1)
219 /* First copy the "additional" attributes. */
220 for (i = 0; i < ARRAY_SIZE (additional_attributes); i++)
221 att[i] = additional_attributes[i];
224 for (i = 0; i < ARRAY_SIZE (url_tag_attr_map); i++)
227 const char *look_for = url_tag_attr_map[i].attr_name;
228 for (j = 0; j < ind - 1; j++)
229 if (!strcmp (att[j], look_for))
236 att = xrealloc (att, (ind + 2) * sizeof (*att));
237 att[ind++] = look_for;
241 interesting_attributes = att;
246 find_tag (const char *tag_name)
250 /* This is linear search; if the number of tags grow, we can switch
253 for (i = 0; i < ARRAY_SIZE (known_tags); i++)
255 int cmp = strcasecmp (known_tags[i].name, tag_name);
256 /* known_tags are sorted alphabetically, so we can
266 /* Find the value of attribute named NAME in the taginfo TAG. If the
267 attribute is not present, return NULL. If ATTRID is non-NULL, the
268 exact identity of the attribute will be returned. */
270 find_attr (struct taginfo *tag, const char *name, int *attrid)
273 for (i = 0; i < tag->nattrs; i++)
274 if (!strcasecmp (tag->attrs[i].name, name))
278 return tag->attrs[i].value;
283 struct collect_urls_closure {
284 char *text; /* HTML text. */
285 char *base; /* Base URI of the document, possibly
286 changed through <base href=...>. */
287 struct urlpos *head, *tail; /* List of URLs */
288 const char *parent_base; /* Base of the current document. */
289 const char *document_file; /* File name of this document. */
290 int dash_p_leaf_HTML; /* Whether -p is specified, and this
291 document is the "leaf" node of the
293 int nofollow; /* whether NOFOLLOW was specified in a
294 <meta name=robots> tag. */
297 /* Resolve LINK_URI and append it to closure->tail. TAG and ATTRID
298 are the necessary context to store the position and size. */
301 handle_link (struct collect_urls_closure *closure, const char *link_uri,
302 struct taginfo *tag, int attrid)
304 int link_has_scheme = url_has_scheme (link_uri);
305 struct urlpos *newel;
306 const char *base = closure->base ? closure->base : closure->parent_base;
311 DEBUGP (("%s: no base, merge will use \"%s\".\n",
312 closure->document_file, link_uri));
314 if (!link_has_scheme)
316 /* We have no base, and the link does not have a host
317 attached to it. Nothing we can do. */
318 /* #### Should we print a warning here? Wget 1.5.x used to. */
322 url = url_parse (link_uri, NULL);
325 DEBUGP (("%s: link \"%s\" doesn't parse.\n",
326 closure->document_file, link_uri));
332 /* Merge BASE with LINK_URI, but also make sure the result is
333 canonicalized, i.e. that "../" have been resolved.
334 (parse_url will do that for us.) */
336 char *complete_uri = uri_merge (base, link_uri);
338 DEBUGP (("%s: merge(\"%s\", \"%s\") -> %s\n",
339 closure->document_file, base, link_uri, complete_uri));
341 url = url_parse (complete_uri, NULL);
344 DEBUGP (("%s: merged link \"%s\" doesn't parse.\n",
345 closure->document_file, complete_uri));
346 xfree (complete_uri);
349 xfree (complete_uri);
352 newel = (struct urlpos *)xmalloc (sizeof (struct urlpos));
354 memset (newel, 0, sizeof (*newel));
357 newel->pos = tag->attrs[attrid].value_raw_beginning - closure->text;
358 newel->size = tag->attrs[attrid].value_raw_size;
360 /* A URL is relative if the host is not named, and the name does not
362 if (!link_has_scheme && *link_uri != '/')
363 newel->link_relative_p = 1;
364 else if (link_has_scheme)
365 newel->link_complete_p = 1;
369 closure->tail->next = newel;
370 closure->tail = newel;
373 closure->tail = closure->head = newel;
376 /* Examine name and attributes of TAG and take appropriate action.
377 What will be done depends on TAG's category and attribute values.
378 Tags of TC_LINK category have attributes that contain links to
379 follow; tags of TC_SPEC category need to be handled specially.
381 #### It would be nice to split this into several functions. */
384 collect_tags_mapper (struct taginfo *tag, void *arg)
386 struct collect_urls_closure *closure = (struct collect_urls_closure *)arg;
387 int tagid = find_tag (tag->name);
388 assert (tagid != -1);
390 switch (known_tags[tagid].category)
395 int size = ARRAY_SIZE (url_tag_attr_map);
396 for (i = 0; i < size; i++)
397 if (url_tag_attr_map[i].tagid == tagid)
399 /* We've found the index of url_tag_attr_map where the
400 attributes of our tags begin. Now, look for every one of
401 them, and handle it. */
402 /* Need to process the attributes in the order they appear in
403 the tag, as this is required if we convert links. */
405 for (id = 0; id < tag->nattrs; id++)
407 /* This nested loop may seem inefficient (O(n^2)), but it's
408 not, since the number of attributes (n) we loop over is
409 extremely small. In the worst case of IMG with all its
410 possible attributes, n^2 will be only 9. */
411 for (i = first; (i < size && url_tag_attr_map[i].tagid == tagid);
415 if (closure->dash_p_leaf_HTML
416 && (url_tag_attr_map[i].flags & AF_EXTERNAL))
417 /* If we're at a -p leaf node, we don't want to retrieve
418 links to references we know are external to this document,
419 such as <a href=...>. */
422 if (!strcasecmp (tag->attrs[id].name,
423 url_tag_attr_map[i].attr_name))
425 attr_value = tag->attrs[id].value;
427 handle_link (closure, attr_value, tag, id);
438 char *newbase = find_attr (tag, "href", NULL);
442 xfree (closure->base);
443 if (closure->parent_base)
444 closure->base = uri_merge (closure->parent_base, newbase);
446 closure->base = xstrdup (newbase);
452 char *rel = find_attr (tag, "rel", NULL);
453 char *href = find_attr (tag, "href", &id);
456 /* In the normal case, all <link href=...> tags are
459 In the special case of when -p is active, however,
460 and we're at a leaf node (relative to the -l
461 max. depth) in the HTML document tree, the only
462 <LINK> tag we'll follow is a <LINK REL=
463 "stylesheet">, as it'll be necessary for displaying
464 this document properly. We won't follow other
465 <LINK> tags, like <LINK REL="home">, for instance,
466 as they refer to external documents. */
467 if (!closure->dash_p_leaf_HTML
468 || (rel && !strcasecmp (rel, "stylesheet")))
469 handle_link (closure, href, tag, id);
474 /* Some pages use a META tag to specify that the page be
475 refreshed by a new page after a given number of seconds.
476 The general format for this is:
478 <meta http-equiv=Refresh content="NUMBER; URL=index2.html">
480 So we just need to skip past the "NUMBER; URL=" garbage
481 to get to the URL. */
484 char *name = find_attr (tag, "name", NULL);
485 char *http_equiv = find_attr (tag, "http-equiv", &id);
486 if (http_equiv && !strcasecmp (http_equiv, "refresh"))
488 char *refresh = find_attr (tag, "content", NULL);
497 if (!(TOUPPER (*p) == 'U'
498 && TOUPPER (*(p + 1)) == 'R'
499 && TOUPPER (*(p + 2)) == 'L'
505 offset = p - refresh;
506 tag->attrs[id].value_raw_beginning += offset;
507 tag->attrs[id].value_raw_size -= offset;
508 handle_link (closure, p, tag, id);
510 else if (name && !strcasecmp (name, "robots"))
512 /* Handle stuff like:
513 <meta name="robots" content="index,nofollow"> */
514 char *content = find_attr (tag, "content", NULL);
517 if (!strcasecmp (content, "none"))
518 closure->nofollow = 1;
523 /* Find the next occurrence of ',' or the end of
525 char *end = strchr (content, ',');
529 end = content + strlen (content);
530 if (!strncasecmp (content, "nofollow", end - content))
531 closure->nofollow = 1;
539 /* Category is TC_SPEC, but tag name is unhandled. This
547 /* Analyze HTML tags FILE and construct a list of URLs referenced from
548 it. It merges relative links in FILE with THIS_URL. It is aware
549 of <base href=...> and does the right thing.
551 If dash_p_leaf_HTML is non-zero, only the elements needed to render
552 FILE ("non-external" links) will be returned. */
554 get_urls_html (const char *file, const char *this_url, int dash_p_leaf_HTML,
555 int *meta_disallow_follow)
557 struct file_memory *fm;
558 struct collect_urls_closure closure;
561 fm = read_file (file);
564 logprintf (LOG_NOTQUIET, "%s: %s\n", file, strerror (errno));
567 DEBUGP (("Loaded %s (size %ld).\n", file, fm->length));
569 closure.text = fm->content;
570 closure.head = closure.tail = NULL;
572 closure.parent_base = this_url ? this_url : opt.base_href;
573 closure.document_file = file;
574 closure.dash_p_leaf_HTML = dash_p_leaf_HTML;
575 closure.nofollow = 0;
577 if (!interesting_tags)
580 map_html_tags (fm->content, fm->length, interesting_tags,
581 interesting_attributes, collect_tags_mapper, &closure);
583 DEBUGP (("no-follow in %s: %d\n", file, closure.nofollow));
584 if (meta_disallow_follow)
585 *meta_disallow_follow = closure.nofollow;
587 FREE_MAYBE (closure.base);
593 cleanup_html_url (void)
595 FREE_MAYBE (interesting_tags);
596 FREE_MAYBE (interesting_attributes);