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 },
96 /* Flags for specific url-attr pairs handled through TC_LINK: */
98 /* This tag points to an external document not necessary for rendering this
99 document (i.e. it's not an inlined image, stylesheet, etc.). */
100 #define AF_EXTERNAL 1
103 /* For tags handled by TC_LINK: attributes that contain URLs to
107 const char *attr_name;
109 } url_tag_attr_map[] = {
110 { TAG_A, "href", AF_EXTERNAL },
111 { TAG_APPLET, "code", 0 },
112 { TAG_AREA, "href", AF_EXTERNAL },
113 { TAG_BGSOUND, "src", 0 },
114 { TAG_BODY, "background", 0 },
115 { TAG_EMBED, "src", 0 },
116 { TAG_FIG, "src", 0 },
117 { TAG_FRAME, "src", 0 },
118 { TAG_IFRAME, "src", 0 },
119 { TAG_IMG, "href", 0 },
120 { TAG_IMG, "lowsrc", 0 },
121 { TAG_IMG, "src", 0 },
122 { TAG_INPUT, "src", 0 },
123 { TAG_LAYER, "src", 0 },
124 { TAG_OVERLAY, "src", 0 },
125 { TAG_SCRIPT, "src", 0 },
126 { TAG_TABLE, "background", 0 },
127 { TAG_TD, "background", 0 },
128 { TAG_TH, "background", 0 }
131 /* The lists of interesting tags and attributes are built dynamically,
132 from the information above. However, some places in the code refer
133 to the attributes not mentioned here. We add them manually. */
134 static const char *additional_attributes[] = {
135 "rel", /* for TAG_LINK */
136 "http-equiv", /* for TAG_META */
137 "name", /* for TAG_META */
138 "content" /* for TAG_META */
141 static const char **interesting_tags;
142 static const char **interesting_attributes;
145 init_interesting (void)
147 /* Init the variables interesting_tags and interesting_attributes
148 that are used by the HTML parser to know which tags and
149 attributes we're interested in. We initialize this only once,
150 for performance reasons.
152 Here we also make sure that what we put in interesting_tags
153 matches the user's preferences as specified through --ignore-tags
154 and --follow-tags. */
158 int size = ARRAY_SIZE (known_tags);
159 interesting_tags = (const char **)xmalloc ((size + 1) * sizeof (char *));
161 for (i = 0; i < size; i++)
163 const char *name = known_tags[i].name;
165 /* Normally here we could say:
166 interesting_tags[i] = name;
167 But we need to respect the settings of --ignore-tags and
168 --follow-tags, so the code gets a bit harier. */
172 /* --ignore-tags was specified. Do not match these
173 specific tags. --ignore-tags takes precedence over
174 --follow-tags, so we process --ignore first and fall
175 through if there's no match. */
177 for (j = 0; opt.ignore_tags[j] != NULL; j++)
178 /* Loop through all the tags this user doesn't care
180 if (strcasecmp(opt.ignore_tags[j], name) == EQ)
191 /* --follow-tags was specified. Only match these specific
192 tags, so return FALSE if we don't match one of them. */
194 for (j = 0; opt.follow_tags[j] != NULL; j++)
195 /* Loop through all the tags this user cares about. */
196 if (strcasecmp(opt.follow_tags[j], name) == EQ)
202 continue; /* wasn't one of the explicitly
206 /* If we get to here, --follow-tags isn't being used or the
207 tag is among the ones that are follwed, and --ignore-tags,
208 if specified, didn't include this tag, so it's an
209 "interesting" one. */
210 interesting_tags[ind++] = name;
212 interesting_tags[ind] = NULL;
215 /* The same for attributes, except we loop through url_tag_attr_map.
216 Here we also need to make sure that the list of attributes is
217 unique, and to include the attributes from additional_attributes. */
220 const char **att = xmalloc ((ARRAY_SIZE (additional_attributes) + 1)
222 /* First copy the "additional" attributes. */
223 for (i = 0; i < ARRAY_SIZE (additional_attributes); i++)
224 att[i] = additional_attributes[i];
227 for (i = 0; i < ARRAY_SIZE (url_tag_attr_map); i++)
230 const char *look_for = url_tag_attr_map[i].attr_name;
231 for (j = 0; j < ind - 1; j++)
232 if (!strcmp (att[j], look_for))
239 att = xrealloc (att, (ind + 2) * sizeof (*att));
240 att[ind++] = look_for;
244 interesting_attributes = att;
249 find_tag (const char *tag_name)
253 /* This is linear search; if the number of tags grow, we can switch
256 for (i = 0; i < ARRAY_SIZE (known_tags); i++)
258 int cmp = strcasecmp (known_tags[i].name, tag_name);
259 /* known_tags are sorted alphabetically, so we can
269 /* Find the value of attribute named NAME in the taginfo TAG. If the
270 attribute is not present, return NULL. If ATTRID is non-NULL, the
271 exact identity of the attribute will be returned. */
273 find_attr (struct taginfo *tag, const char *name, int *attrid)
276 for (i = 0; i < tag->nattrs; i++)
277 if (!strcasecmp (tag->attrs[i].name, name))
281 return tag->attrs[i].value;
286 struct collect_urls_closure {
287 char *text; /* HTML text. */
288 char *base; /* Base URI of the document, possibly
289 changed through <base href=...>. */
290 urlpos *head, *tail; /* List of URLs */
291 const char *parent_base; /* Base of the current document. */
292 const char *document_file; /* File name of this document. */
293 int dash_p_leaf_HTML; /* Whether -p is specified, and this
294 document is the "leaf" node of the
296 int nofollow; /* whether NOFOLLOW was specified in a
297 <meta name=robots> tag. */
300 /* Resolve LINK_URI and append it to closure->tail. TAG and ATTRID
301 are the necessary context to store the position and size. */
304 handle_link (struct collect_urls_closure *closure, const char *link_uri,
305 struct taginfo *tag, int attrid)
307 int no_proto = !has_proto (link_uri);
310 const char *base = closure->base ? closure->base : closure->parent_base;
313 char *fragment = strrchr (link_uri, '#');
317 /* Nullify the fragment identifier, i.e. everything after the
318 last occurrence of `#', inclusive. This copying is
319 relatively inefficient, but it doesn't matter because
320 fragment identifiers don't come up all that often. */
321 int hashlen = fragment - link_uri;
322 char *p = alloca (hashlen + 1);
323 memcpy (p, link_uri, hashlen);
332 /* We have no base, and the link does not have a protocol or
333 a host attached to it. Nothing we can do. */
334 /* #### Should we print a warning here? Wget 1.5.x used to. */
338 complete_uri = xstrdup (link_uri);
341 complete_uri = url_concat (base, link_uri);
343 DEBUGP (("%s: merge(\"%s\", \"%s\") -> %s\n",
344 closure->document_file, base ? base : "(null)",
345 link_uri, complete_uri));
347 newel = (urlpos *)xmalloc (sizeof (urlpos));
349 memset (newel, 0, sizeof (*newel));
351 newel->url = complete_uri;
352 newel->pos = tag->attrs[attrid].value_raw_beginning - closure->text;
353 newel->size = tag->attrs[attrid].value_raw_size;
355 /* A URL is relative if the host and protocol are not named, and the
356 name does not start with `/'. */
357 if (no_proto && *link_uri != '/')
358 newel->link_relative_p = 1;
360 newel->link_complete_p = 1;
364 closure->tail->next = newel;
365 closure->tail = newel;
368 closure->tail = closure->head = newel;
371 /* #### Document what this does.
372 #### It would be nice to split this into several functions. */
375 collect_tags_mapper (struct taginfo *tag, void *arg)
377 struct collect_urls_closure *closure = (struct collect_urls_closure *)arg;
378 int tagid = find_tag (tag->name);
379 assert (tagid != -1);
381 switch (known_tags[tagid].category)
386 int size = ARRAY_SIZE (url_tag_attr_map);
387 for (i = 0; i < size; i++)
388 if (url_tag_attr_map[i].tagid == tagid)
390 /* We've found the index of url_tag_attr_map where the
391 attributes of our tags begin. Now, look for every one of
392 them, and handle it. */
393 for (; (i < size && url_tag_attr_map[i].tagid == tagid); i++)
397 if (closure->dash_p_leaf_HTML
398 && (url_tag_attr_map[i].flags & AF_EXTERNAL))
399 /* If we're at a -p leaf node, we don't want to retrieve
400 links to references we know are external to this document,
401 such as <a href=...>. */
404 /* This find_attr() buried in a loop may seem inefficient
405 (O(n^2)), but it's not, since the number of attributes
406 (n) we loop over is extremely small. In the worst case
407 of IMG with all its possible attributes, n^2 will be
409 attr_value = find_attr (tag, url_tag_attr_map[i].attr_name, &id);
411 handle_link (closure, attr_value, tag, id);
420 char *newbase = find_attr (tag, "href", NULL);
424 xfree (closure->base);
425 if (closure->parent_base)
426 closure->base = url_concat (closure->parent_base, newbase);
428 closure->base = xstrdup (newbase);
434 char *rel = find_attr (tag, "rel", NULL);
435 char *href = find_attr (tag, "href", &id);
438 /* In the normal case, all <link href=...> tags are
441 In the special case of when -p is active, however,
442 and we're at a leaf node (relative to the -l
443 max. depth) in the HTML document tree, the only
444 <LINK> tag we'll follow is a <LINK REL=
445 "stylesheet">, as it'll be necessary for displaying
446 this document properly. We won't follow other
447 <LINK> tags, like <LINK REL="home">, for instance,
448 as they refer to external documents. */
449 if (!closure->dash_p_leaf_HTML
450 || (rel && !strcasecmp (rel, "stylesheet")))
451 handle_link (closure, href, tag, id);
456 /* Some pages use a META tag to specify that the page be
457 refreshed by a new page after a given number of seconds.
458 The general format for this is:
460 <meta http-equiv=Refresh content="NUMBER; URL=index2.html">
462 So we just need to skip past the "NUMBER; URL=" garbage
463 to get to the URL. */
466 char *name = find_attr (tag, "name", NULL);
467 char *http_equiv = find_attr (tag, "http-equiv", &id);
468 if (http_equiv && !strcasecmp (http_equiv, "refresh"))
470 char *refresh = find_attr (tag, "content", NULL);
479 if (!(TOUPPER (*p) == 'U'
480 && TOUPPER (*(p + 1)) == 'R'
481 && TOUPPER (*(p + 2)) == 'L'
487 offset = p - refresh;
488 tag->attrs[id].value_raw_beginning += offset;
489 tag->attrs[id].value_raw_size -= offset;
490 handle_link (closure, p, tag, id);
492 else if (name && !strcasecmp (name, "robots"))
494 /* Handle stuff like:
495 <meta name="robots" content="index,nofollow"> */
496 char *content = find_attr (tag, "content", NULL);
499 if (!strcasecmp (content, "none"))
500 closure->nofollow = 1;
505 /* Find the next occurrence of ',' or the end of
507 char *end = strchr (content, ',');
511 end = content + strlen (content);
512 if (!strncasecmp (content, "nofollow", end - content))
513 closure->nofollow = 1;
521 /* Category is TC_SPEC, but tag name is unhandled. This
529 /* Scan FILE, retrieving links to HTML documents from it. Each link is
531 Similar to get_urls_file, but for HTML files. FILE is scanned as
532 an HTML document. get_urls_html() constructs the URLs from the
535 If SILENT is non-zero, do not barf on baseless relative links. */
537 get_urls_html (const char *file, const char *this_url, int dash_p_leaf_HTML,
538 int *meta_disallow_follow)
540 struct file_memory *fm;
541 struct collect_urls_closure closure;
544 fm = read_file (file);
547 logprintf (LOG_NOTQUIET, "%s: %s\n", file, strerror (errno));
550 DEBUGP (("Loaded %s (size %ld).\n", file, fm->length));
552 closure.text = fm->content;
553 closure.head = closure.tail = NULL;
555 closure.parent_base = this_url ? this_url : opt.base_href;
556 closure.document_file = file;
557 closure.dash_p_leaf_HTML = dash_p_leaf_HTML;
558 closure.nofollow = 0;
560 if (!interesting_tags)
563 map_html_tags (fm->content, fm->length, interesting_tags,
564 interesting_attributes, collect_tags_mapper, &closure);
566 DEBUGP (("no-follow in %s: %d\n", file, closure.nofollow));
567 if (meta_disallow_follow)
568 *meta_disallow_follow = closure.nofollow;
570 FREE_MAYBE (closure.base);
576 cleanup_html_url (void)
578 FREE_MAYBE (interesting_tags);
579 FREE_MAYBE (interesting_attributes);