1 /* Collect URLs from HTML source.
2 Copyright (C) 1998, 2000 Free Software Foundation, Inc.
4 This file is part of Wget.
6 This program 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 This program 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 this program; if not, write to the Free Software
18 Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. */
34 #include "html-parse.h"
42 enum tag_category { TC_LINK, TC_SPEC };
44 /* Here we try to categorize the known tags. Each tag has its ID and
45 cetegory. Category TC_LINK means that one or more of its
46 attributes contain links that should be retrieved. TC_SPEC means
47 that the tag is specific in some way, and has to be handled
51 enum tag_category category;
56 { "applet", TC_LINK },
62 { "bgsound", TC_LINK },
72 { "iframe", TC_LINK },
83 #define TAG_OVERLAY 15
84 { "overlay", TC_LINK },
86 { "script", TC_LINK },
95 /* Flags for specific url-attr pairs handled through TC_LINK: */
98 /* For tags handled by TC_LINK: attributes that contain URLs to
102 const char *attr_name;
104 } url_tag_attr_map[] = {
105 { TAG_A, "href", AF_EXTERNAL },
106 { TAG_APPLET, "code", 0 },
107 { TAG_AREA, "href", AF_EXTERNAL },
108 { TAG_BGSOUND, "src", 0 },
109 { TAG_BODY, "background", 0 },
110 { TAG_EMBED, "src", 0 },
111 { TAG_FIG, "src", 0 },
112 { TAG_FRAME, "src", 0 },
113 { TAG_IFRAME, "src", 0 },
114 { TAG_IMG, "href", 0 },
115 { TAG_IMG, "lowsrc", 0 },
116 { TAG_IMG, "src", 0 },
117 { TAG_INPUT, "src", 0 },
118 { TAG_LAYER, "src", 0 },
119 { TAG_OVERLAY, "src", 0 },
120 { TAG_SCRIPT, "src", 0 },
121 { TAG_TABLE, "background", 0 },
122 { TAG_TD, "background", 0 },
123 { TAG_TH, "background", 0 }
126 /* The lists of interesting tags and attributes are built dynamically,
127 from the information above. However, some places in the code refer
128 to the attributes not mentioned here. We add them manually. */
129 static const char *additional_attributes[] = {
130 "rel", /* for TAG_LINK */
131 "http-equiv", /* for TAG_META */
132 "name", /* for TAG_META */
133 "content" /* for TAG_META */
136 static const char **interesting_tags;
137 static const char **interesting_attributes;
140 init_interesting (void)
142 /* Init the variables interesting_tags and interesting_attributes
143 that are used by the HTML parser to know which tags and
144 attributes we're interested in. We initialize this only once,
145 for performance reasons.
147 Here we also make sure that what we put in interesting_tags
148 matches the user's preferences as specified through --ignore-tags
149 and --follow-tags. */
153 int size = ARRAY_SIZE (known_tags);
154 interesting_tags = (const char **)xmalloc ((size + 1) * sizeof (char *));
156 for (i = 0; i < size; i++)
158 const char *name = known_tags[i].name;
160 /* Normally here we could say:
161 interesting_tags[i] = name;
162 But we need to respect the settings of --ignore-tags and
163 --follow-tags, so the code gets a bit harier. */
167 /* --ignore-tags was specified. Do not match these
168 specific tags. --ignore-tags takes precedence over
169 --follow-tags, so we process --ignore first and fall
170 through if there's no match. */
172 for (j = 0; opt.ignore_tags[j] != NULL; j++)
173 /* Loop through all the tags this user doesn't care
175 if (strcasecmp(opt.ignore_tags[j], name) == EQ)
186 /* --follow-tags was specified. Only match these specific
187 tags, so return FALSE if we don't match one of them. */
189 for (j = 0; opt.follow_tags[j] != NULL; j++)
190 /* Loop through all the tags this user cares about. */
191 if (strcasecmp(opt.follow_tags[j], name) == EQ)
197 continue; /* wasn't one of the explicitly
201 /* If we get to here, --follow-tags isn't being used or the
202 tag is among the ones that are follwed, and --ignore-tags,
203 if specified, didn't include this tag, so it's an
204 "interesting" one. */
205 interesting_tags[ind++] = name;
207 interesting_tags[ind] = NULL;
210 /* The same for attributes, except we loop through url_tag_attr_map.
211 Here we also need to make sure that the list of attributes is
212 unique, and to include the attributes from additional_attributes. */
215 const char **att = xmalloc ((ARRAY_SIZE (additional_attributes) + 1)
217 /* First copy the "additional" attributes. */
218 for (i = 0; i < ARRAY_SIZE (additional_attributes); i++)
219 att[i] = additional_attributes[i];
222 for (i = 0; i < ARRAY_SIZE (url_tag_attr_map); i++)
225 const char *look_for = url_tag_attr_map[i].attr_name;
226 for (j = 0; j < ind - 1; j++)
227 if (!strcmp (att[j], look_for))
234 att = xrealloc (att, (ind + 2) * sizeof (*att));
235 att[ind++] = look_for;
239 interesting_attributes = att;
244 find_tag (const char *tag_name)
248 /* This is linear search; if the number of tags grow, we can switch
251 for (i = 0; i < ARRAY_SIZE (known_tags); i++)
253 int cmp = strcasecmp (known_tags[i].name, tag_name);
254 /* known_tags are sorted alphabetically, so we can
264 /* Find the value of attribute named NAME in the taginfo TAG. If the
265 attribute is not present, return NULL. If ATTRID is non-NULL, the
266 exact identity of the attribute will be returned. */
268 find_attr (struct taginfo *tag, const char *name, int *attrid)
271 for (i = 0; i < tag->nattrs; i++)
272 if (!strcasecmp (tag->attrs[i].name, name))
276 return tag->attrs[i].value;
281 struct collect_urls_closure {
282 char *text; /* HTML text. */
283 char *base; /* Base URI of the document, possibly
284 changed through <base href=...>. */
285 urlpos *head, *tail; /* List of URLs */
286 const char *parent_base; /* Base of the current document. */
287 const char *document_file; /* File name of this document. */
288 int dash_p_leaf_HTML; /* Whether -p is specified, and this
289 document is the "leaf" node of the
291 int nofollow; /* whether NOFOLLOW was specified in a
292 <meta name=robots> tag. */
295 /* Resolve LINK_URI and append it to closure->tail. TAG and ATTRID
296 are the necessary context to store the position and size. */
299 handle_link (struct collect_urls_closure *closure, const char *link_uri,
300 struct taginfo *tag, int attrid)
302 int no_proto = !has_proto (link_uri);
305 const char *base = closure->base ? closure->base : closure->parent_base;
308 char *fragment = strrchr (link_uri, '#');
312 /* Nullify the fragment identifier, i.e. everything after the
313 last occurrence of `#', inclusive. This copying is
314 relatively inefficient, but it doesn't matter because
315 fragment identifiers don't come up all that often. */
316 int hashlen = fragment - link_uri;
317 char *p = alloca (hashlen + 1);
318 memcpy (p, link_uri, hashlen);
327 /* We have no base, and the link does not have a protocol or
328 a host attached to it. Nothing we can do. */
329 /* #### Should we print a warning here? Wget 1.5.x used to. */
333 complete_uri = xstrdup (link_uri);
336 complete_uri = url_concat (base, link_uri);
338 DEBUGP (("%s: merge(\"%s\", \"%s\") -> %s\n",
339 closure->document_file, base ? base : "(null)",
340 link_uri, complete_uri));
342 newel = (urlpos *)xmalloc (sizeof (urlpos));
344 memset (newel, 0, sizeof (*newel));
346 newel->url = complete_uri;
347 newel->pos = tag->attrs[attrid].value_raw_beginning - closure->text;
348 newel->size = tag->attrs[attrid].value_raw_size;
350 /* A URL is relative if the host and protocol are not named, and the
351 name does not start with `/'.
352 #### This logic might need some rethinking. */
353 if (no_proto && *link_uri != '/')
354 newel->flags |= (URELATIVE | UNOPROTO);
356 newel->flags |= UNOPROTO;
360 closure->tail->next = newel;
361 closure->tail = newel;
364 closure->tail = closure->head = newel;
367 /* #### Document what this does.
368 #### It would be nice to split this into several functions. */
371 collect_tags_mapper (struct taginfo *tag, void *arg)
373 struct collect_urls_closure *closure = (struct collect_urls_closure *)arg;
374 int tagid = find_tag (tag->name);
375 assert (tagid != -1);
377 switch (known_tags[tagid].category)
382 int size = ARRAY_SIZE (url_tag_attr_map);
383 for (i = 0; i < size; i++)
384 if (url_tag_attr_map[i].tagid == tagid)
386 /* We've found the index of url_tag_attr_map where the
387 attributes of our tags begin. Now, look for every one of
388 them, and handle it. */
389 for (; (i < size && url_tag_attr_map[i].tagid == tagid); i++)
393 if (closure->dash_p_leaf_HTML
394 && (url_tag_attr_map[i].flags & AF_EXTERNAL))
395 /* If we're at a -p leaf node, we don't want to retrieve
396 links to references we know are external, such as <a
400 /* This find_attr() buried in a loop may seem inefficient
401 (O(n^2)), but it's not, since the number of attributes
402 (n) we loop over is extremely small. In the worst case
403 of IMG with all its possible attributes, n^2 will be
405 attr_value = find_attr (tag, url_tag_attr_map[i].attr_name, &id);
407 handle_link (closure, attr_value, tag, id);
416 char *newbase = find_attr (tag, "href", NULL);
420 free (closure->base);
421 if (closure->parent_base)
422 closure->base = url_concat (closure->parent_base, newbase);
424 closure->base = xstrdup (newbase);
430 char *rel = find_attr (tag, "rel", NULL);
431 char *href = find_attr (tag, "href", &id);
434 /* In the normal case, all <link href=...> tags are
437 In the special case of when -p is active, however,
438 and we're at a leaf node (relative to the -l
439 max. depth) in the HTML document tree, the only
440 <LINK> tag we'll follow is a <LINK REL=
441 "stylesheet">, as it's necessary for displaying
442 this document properly. We won't follow other
443 <LINK> tags, like <LINK REL="home">, for instance,
444 as they refer to external documents. */
445 if (!closure->dash_p_leaf_HTML
446 || (rel && !strcasecmp (rel, "stylesheet")))
447 handle_link (closure, href, tag, id);
452 /* Some pages use a META tag to specify that the page be
453 refreshed by a new page after a given number of seconds.
454 The general format for this is:
456 <meta http-equiv=Refresh content="NUMBER; URL=index2.html">
458 So we just need to skip past the "NUMBER; URL=" garbage
459 to get to the URL. */
462 char *name = find_attr (tag, "name", NULL);
463 char *http_equiv = find_attr (tag, "http-equiv", &id);
464 if (http_equiv && !strcasecmp (http_equiv, "refresh"))
466 char *refresh = find_attr (tag, "content", NULL);
475 if (!(TOUPPER (*p) == 'U'
476 && TOUPPER (*(p + 1)) == 'R'
477 && TOUPPER (*(p + 2)) == 'L'
483 offset = p - refresh;
484 tag->attrs[id].value_raw_beginning += offset;
485 tag->attrs[id].value_raw_size -= offset;
486 handle_link (closure, p, tag, id);
488 else if (name && !strcasecmp (name, "robots"))
490 /* Handle stuff like:
491 <meta name="robots" content="index,nofollow"> */
492 char *content = find_attr (tag, "content", NULL);
495 if (!strcasecmp (content, "none"))
496 closure->nofollow = 1;
501 /* Find the next occurrence of ',' or the end of
503 char *end = strchr (content, ',');
507 end = content + strlen (content);
508 if (!strncasecmp (content, "nofollow", end - content))
509 closure->nofollow = 1;
517 /* Category is TC_SPEC, but tag name is unhandled. This
525 /* Scan FILE, retrieving links to HTML documents from it. Each link is
527 Similar to get_urls_file, but for HTML files. FILE is scanned as
528 an HTML document. get_urls_html() constructs the URLs from the
531 If SILENT is non-zero, do not barf on baseless relative links. */
533 get_urls_html (const char *file, const char *this_url, int dash_p_leaf_HTML,
534 int *meta_disallow_follow)
536 struct file_memory *fm;
537 struct collect_urls_closure closure;
540 fm = read_file (file);
543 logprintf (LOG_NOTQUIET, "%s: %s\n", file, strerror (errno));
546 DEBUGP (("Loaded %s (size %ld).\n", file, fm->length));
548 closure.text = fm->content;
549 closure.head = closure.tail = NULL;
551 closure.parent_base = this_url ? this_url : opt.base_href;
552 closure.document_file = file;
553 closure.dash_p_leaf_HTML = dash_p_leaf_HTML;
554 closure.nofollow = 0;
556 if (!interesting_tags)
559 map_html_tags (fm->content, fm->length, interesting_tags,
560 interesting_attributes, collect_tags_mapper, &closure);
562 DEBUGP (("no-follow in %s: %d\n", file, closure.nofollow));
563 if (meta_disallow_follow)
564 *meta_disallow_follow = closure.nofollow;
566 FREE_MAYBE (closure.base);