/* Collect URLs from HTML source.
- Copyright (C) 1998, 2000 Free Software Foundation, Inc.
+ Copyright (C) 1998, 2000, 2001 Free Software Foundation, Inc.
This file is part of GNU Wget.
char *text; /* HTML text. */
char *base; /* Base URI of the document, possibly
changed through <base href=...>. */
- urlpos *head, *tail; /* List of URLs */
+ struct urlpos *head, *tail; /* List of URLs */
const char *parent_base; /* Base of the current document. */
const char *document_file; /* File name of this document. */
- int dash_p_leaf_HTML; /* Whether -p is specified, and this
- document is the "leaf" node of the
- HTML tree. */
int nofollow; /* whether NOFOLLOW was specified in a
<meta name=robots> tag. */
};
/* Resolve LINK_URI and append it to closure->tail. TAG and ATTRID
are the necessary context to store the position and size. */
-static void
+static struct urlpos *
handle_link (struct collect_urls_closure *closure, const char *link_uri,
struct taginfo *tag, int attrid)
{
- int no_scheme = !url_has_scheme (link_uri);
- urlpos *newel;
-
+ int link_has_scheme = url_has_scheme (link_uri);
+ struct urlpos *newel;
const char *base = closure->base ? closure->base : closure->parent_base;
- char *complete_uri;
-
- char *fragment = strrchr (link_uri, '#');
-
- if (fragment)
- {
- /* Nullify the fragment identifier, i.e. everything after the
- last occurrence of `#', inclusive. This copying is
- relatively inefficient, but it doesn't matter because
- fragment identifiers don't come up all that often. */
- int hashlen = fragment - link_uri;
- char *p = alloca (hashlen + 1);
- memcpy (p, link_uri, hashlen);
- p[hashlen] = '\0';
- link_uri = p;
- }
+ struct url *url;
if (!base)
{
- if (no_scheme)
+ DEBUGP (("%s: no base, merge will use \"%s\".\n",
+ closure->document_file, link_uri));
+
+ if (!link_has_scheme)
{
/* We have no base, and the link does not have a host
attached to it. Nothing we can do. */
/* #### Should we print a warning here? Wget 1.5.x used to. */
- return;
+ return NULL;
+ }
+
+ url = url_parse (link_uri, NULL);
+ if (!url)
+ {
+ DEBUGP (("%s: link \"%s\" doesn't parse.\n",
+ closure->document_file, link_uri));
+ return NULL;
}
- else
- complete_uri = xstrdup (link_uri);
}
else
- complete_uri = uri_merge (base, link_uri);
+ {
+ /* Merge BASE with LINK_URI, but also make sure the result is
+ canonicalized, i.e. that "../" have been resolved.
+ (parse_url will do that for us.) */
+
+ char *complete_uri = uri_merge (base, link_uri);
+
+ DEBUGP (("%s: merge(\"%s\", \"%s\") -> %s\n",
+ closure->document_file, base, link_uri, complete_uri));
- DEBUGP (("%s: merge(\"%s\", \"%s\") -> %s\n",
- closure->document_file, base ? base : "(null)",
- link_uri, complete_uri));
+ url = url_parse (complete_uri, NULL);
+ if (!url)
+ {
+ DEBUGP (("%s: merged link \"%s\" doesn't parse.\n",
+ closure->document_file, complete_uri));
+ xfree (complete_uri);
+ return NULL;
+ }
+ xfree (complete_uri);
+ }
- newel = (urlpos *)xmalloc (sizeof (urlpos));
+ newel = (struct urlpos *)xmalloc (sizeof (struct urlpos));
memset (newel, 0, sizeof (*newel));
newel->next = NULL;
- newel->url = complete_uri;
+ newel->url = url;
newel->pos = tag->attrs[attrid].value_raw_beginning - closure->text;
newel->size = tag->attrs[attrid].value_raw_size;
/* A URL is relative if the host is not named, and the name does not
start with `/'. */
- if (no_scheme && *link_uri != '/')
+ if (!link_has_scheme && *link_uri != '/')
newel->link_relative_p = 1;
- else if (!no_scheme)
+ else if (link_has_scheme)
newel->link_complete_p = 1;
if (closure->tail)
}
else
closure->tail = closure->head = newel;
+
+ return newel;
}
/* Examine name and attributes of TAG and take appropriate action.
for (i = first; (i < size && url_tag_attr_map[i].tagid == tagid);
i++)
{
- char *attr_value;
- if (closure->dash_p_leaf_HTML
- && (url_tag_attr_map[i].flags & AF_EXTERNAL))
- /* If we're at a -p leaf node, we don't want to retrieve
- links to references we know are external to this document,
- such as <a href=...>. */
- continue;
-
- if (!strcasecmp (tag->attrs[id].name,
- url_tag_attr_map[i].attr_name))
+ if (0 == strcasecmp (tag->attrs[id].name,
+ url_tag_attr_map[i].attr_name))
{
- attr_value = tag->attrs[id].value;
+ char *attr_value = tag->attrs[id].value;
if (attr_value)
- handle_link (closure, attr_value, tag, id);
+ {
+ struct urlpos *entry;
+ entry = handle_link (closure, attr_value, tag, id);
+ if (entry != NULL
+ && !(url_tag_attr_map[i].flags & AF_EXTERNAL))
+ entry->link_inline_p = 1;
+ }
}
}
}
{
case TAG_BASE:
{
- char *newbase = find_attr (tag, "href", NULL);
+ struct urlpos *base_urlpos;
+ int id;
+ char *newbase = find_attr (tag, "href", &id);
if (!newbase)
break;
+
+ base_urlpos = handle_link (closure, newbase, tag, id);
+ if (!base_urlpos)
+ break;
+ base_urlpos->ignore_when_downloading = 1;
+ base_urlpos->link_base_p = 1;
+
if (closure->base)
xfree (closure->base);
if (closure->parent_base)
case TAG_LINK:
{
int id;
- char *rel = find_attr (tag, "rel", NULL);
char *href = find_attr (tag, "href", &id);
+
+ /* All <link href="..."> link references are external,
+ except for <link rel="stylesheet" href="...">. */
if (href)
{
- /* In the normal case, all <link href=...> tags are
- fair game.
-
- In the special case of when -p is active, however,
- and we're at a leaf node (relative to the -l
- max. depth) in the HTML document tree, the only
- <LINK> tag we'll follow is a <LINK REL=
- "stylesheet">, as it'll be necessary for displaying
- this document properly. We won't follow other
- <LINK> tags, like <LINK REL="home">, for instance,
- as they refer to external documents. */
- if (!closure->dash_p_leaf_HTML
- || (rel && !strcasecmp (rel, "stylesheet")))
- handle_link (closure, href, tag, id);
+ struct urlpos *entry;
+ entry = handle_link (closure, href, tag, id);
+ if (entry != NULL)
+ {
+ char *rel = find_attr (tag, "rel", NULL);
+ if (rel && 0 == strcasecmp (rel, "stylesheet"))
+ entry->link_inline_p = 1;
+ }
}
}
break;
So we just need to skip past the "NUMBER; URL=" garbage
to get to the URL. */
{
- int id;
char *name = find_attr (tag, "name", NULL);
- char *http_equiv = find_attr (tag, "http-equiv", &id);
+ char *http_equiv = find_attr (tag, "http-equiv", NULL);
if (http_equiv && !strcasecmp (http_equiv, "refresh"))
{
- char *refresh = find_attr (tag, "content", NULL);
- char *p = refresh;
- int offset;
- while (ISDIGIT (*p))
- ++p;
+ struct urlpos *entry;
+
+ int id;
+ char *p, *refresh = find_attr (tag, "content", &id);
+ int timeout = 0;
+
+ for (p = refresh; ISDIGIT (*p); p++)
+ timeout = 10 * timeout + *p - '0';
if (*p++ != ';')
return;
+
while (ISSPACE (*p))
++p;
if (!(TOUPPER (*p) == 'U'
p += 4;
while (ISSPACE (*p))
++p;
- offset = p - refresh;
- tag->attrs[id].value_raw_beginning += offset;
- tag->attrs[id].value_raw_size -= offset;
- handle_link (closure, p, tag, id);
+
+ entry = handle_link (closure, p, tag, id);
+ if (entry)
+ {
+ entry->link_refresh_p = 1;
+ entry->refresh_timeout = timeout;
+ }
}
else if (name && !strcasecmp (name, "robots"))
{
}
/* Analyze HTML tags FILE and construct a list of URLs referenced from
- it. It merges relative links in FILE with THIS_URL. It is aware
- of <base href=...> and does the right thing.
-
- If dash_p_leaf_HTML is non-zero, only the elements needed to render
- FILE ("non-external" links) will be returned. */
-urlpos *
-get_urls_html (const char *file, const char *this_url, int dash_p_leaf_HTML,
- int *meta_disallow_follow)
+ it. It merges relative links in FILE with URL. It is aware of
+ <base href=...> and does the right thing. */
+struct urlpos *
+get_urls_html (const char *file, const char *url, int *meta_disallow_follow)
{
struct file_memory *fm;
struct collect_urls_closure closure;
closure.text = fm->content;
closure.head = closure.tail = NULL;
closure.base = NULL;
- closure.parent_base = this_url ? this_url : opt.base_href;
+ closure.parent_base = url ? url : opt.base_href;
closure.document_file = file;
- closure.dash_p_leaf_HTML = dash_p_leaf_HTML;
closure.nofollow = 0;
if (!interesting_tags)