void
register_download (const char *url, const char *file)
{
if (!opt.convert_links)
return;
if (!dl_file_url_map)
dl_file_url_map = make_string_hash_table (0);
hash_table_put (dl_file_url_map, xstrdup (file), xstrdup (url));
if (!dl_url_file_map)
dl_url_file_map = make_string_hash_table (0);
hash_table_put (dl_url_file_map, xstrdup (url), xstrdup (file));
}
downloaded_file_t
downloaded_file (downloaded_file_t mode, const char *file)
{
downloaded_file_t *ptr;
if (mode == CHECK_FOR_FILE)
{
if (!downloaded_files_hash)
return FILE_NOT_ALREADY_DOWNLOADED;
ptr = hash_table_get (downloaded_files_hash, file);
if (!ptr)
return FILE_NOT_ALREADY_DOWNLOADED;
return *ptr;
}
if (!downloaded_files_hash)
downloaded_files_hash = make_string_hash_table (0);
ptr = hash_table_get (downloaded_files_hash, file);
if (ptr)
return *ptr;
ptr = downloaded_mode_to_ptr (mode);
hash_table_put (downloaded_files_hash, xstrdup (file), ptr);
return FILE_NOT_ALREADY_DOWNLOADED;
}
void
register_css (const char *file)
{
if (!downloaded_css_set)
downloaded_css_set = make_string_hash_table (0);
string_set_add (downloaded_css_set, file);
}
int
main (void)
{
struct hash_table *ht = make_string_hash_table (0);
char line[80];
while ((fgets (line, sizeof (line), stdin)))
{
int len = strlen (line);
if (len <= 1)
continue;
line[--len] = '\0';
if (!hash_table_contains (ht, line))
hash_table_put (ht, strdup (line), "here I am!");
#if 1
if (len % 5 == 0)
{
char *line_copy;
if (hash_table_get_pair (ht, line, &line_copy, NULL))
{
hash_table_remove (ht, line);
xfree (line_copy);
}
}
#endif
}
#if 0
print_hash (ht);
#endif
#if 1
printf ("%d %d\n", ht->count, ht->size);
#endif
return 0;
}
void
register_html (const char *url, const char *file)
{
if (!downloaded_html_set)
downloaded_html_set = make_string_hash_table (0);
string_set_add (downloaded_html_set, file);
}
/* Remembers broken links. */
void
nonexisting_url (const char *url)
{
/* Ignore robots.txt URLs */
if (is_robots_txt_url (url))
return;
if (!nonexisting_urls_set)
nonexisting_urls_set = make_string_hash_table (0);
string_set_add (nonexisting_urls_set, url);
}
void
register_html (const char *url, const char *file)
{
if (!downloaded_html_set)
downloaded_html_set = make_string_hash_table (0);
else if (hash_table_contains (downloaded_html_set, file))
return;
/* The set and the list should use the same copy of FILE, but the
slist interface insists on strduping the string it gets. Oh
well. */
string_set_add (downloaded_html_set, file);
downloaded_html_list = slist_prepend (downloaded_html_list, file);
}
static void
write_backup_file (const char *file, downloaded_file_t downloaded_file_return)
{
/* Rather than just writing over the original .html file with the
converted version, save the former to *.orig. Note we only do
this for files we've _successfully_ downloaded, so we don't
clobber .orig files sitting around from previous invocations.
On VMS, use "_orig" instead of ".orig". See "wget.h". */
/* Construct the backup filename as the original name plus ".orig". */
size_t filename_len = strlen (file);
char* filename_plus_orig_suffix;
/* TODO: hack this to work with css files */
if (downloaded_file_return == FILE_DOWNLOADED_AND_HTML_EXTENSION_ADDED)
{
/* Just write "orig" over "html". We need to do it this way
because when we're checking to see if we've downloaded the
file before (to see if we can skip downloading it), we don't
know if it's a text/html file. Therefore we don't know yet
at that stage that -E is going to cause us to tack on
".html", so we need to compare vs. the original URL plus
".orig", not the original URL plus ".html.orig". */
filename_plus_orig_suffix = alloca (filename_len + 1);
strcpy (filename_plus_orig_suffix, file);
strcpy ((filename_plus_orig_suffix + filename_len) - 4, "orig");
}
else /* downloaded_file_return == FILE_DOWNLOADED_NORMALLY */
{
/* Append ".orig" to the name. */
filename_plus_orig_suffix = alloca (filename_len + sizeof (ORIG_SFX));
strcpy (filename_plus_orig_suffix, file);
strcpy (filename_plus_orig_suffix + filename_len, ORIG_SFX);
}
if (!converted_files)
converted_files = make_string_hash_table (0);
/* We can get called twice on the same URL thanks to the
convert_all_links() call in main. If we write the .orig file
each time in such a case, it'll end up containing the first-pass
conversion, not the original file. So, see if we've already been
called on this file. */
if (!string_set_contains (converted_files, file))
{
/* Rename <file> to <file>.orig before former gets written over. */
if (rename (file, filename_plus_orig_suffix) != 0)
logprintf (LOG_NOTQUIET, _("Cannot back up %s as %s: %s\n"),
file, filename_plus_orig_suffix, strerror (errno));
/* Remember that we've already written a .orig backup for this file.
Note that we never free this memory since we need it till the
convert_all_links() call, which is one of the last things the
program does before terminating. BTW, I'm not sure if it would be
safe to just set 'converted_file_ptr->string' to 'file' below,
rather than making a copy of the string... Another note is that I
thought I could just add a field to the urlpos structure saying
that we'd written a .orig file for this URL, but that didn't work,
so I had to make this separate list.
-- Dan Harkless <*****@*****.**>
This [adding a field to the urlpos structure] didn't work
because convert_file() is called from convert_all_links at
the end of the retrieval with a freshly built new urlpos
list.
-- Hrvoje Niksic <*****@*****.**>
*/
string_set_add (converted_files, file);
}
}
uerr_t
retrieve_tree (const char *start_url)
{
uerr_t status = RETROK;
/* The queue of URLs we need to load. */
struct url_queue *queue;
/* The URLs we do not wish to enqueue, because they are already in
the queue, but haven't been downloaded yet. */
struct hash_table *blacklist;
int up_error_code;
struct url *start_url_parsed = url_parse (start_url, &up_error_code);
if (!start_url_parsed)
{
logprintf (LOG_NOTQUIET, "%s: %s.\n", start_url,
url_error (up_error_code));
return URLERROR;
}
queue = url_queue_new ();
blacklist = make_string_hash_table (0);
/* Enqueue the starting URL. Use start_url_parsed->url rather than
just URL so we enqueue the canonical form of the URL. */
url_enqueue (queue, xstrdup (start_url_parsed->url), NULL, 0, 1);
string_set_add (blacklist, start_url_parsed->url);
while (1)
{
int descend = 0;
char *url, *referer, *file = NULL;
int depth, html_allowed;
boolean dash_p_leaf_HTML = FALSE;
if (opt.quota && total_downloaded_bytes > opt.quota)
break;
if (status == FWRITEERR)
break;
/* Get the next URL from the queue... */
if (!url_dequeue (queue,
(const char **)&url, (const char **)&referer,
&depth, &html_allowed))
break;
/* ...and download it. Note that this download is in most cases
unconditional, as download_child_p already makes sure a file
doesn't get enqueued twice -- and yet this check is here, and
not in download_child_p. This is so that if you run `wget -r
URL1 URL2', and a random URL is encountered once under URL1
and again under URL2, but at a different (possibly smaller)
depth, we want the URL's children to be taken into account
the second time. */
if (dl_url_file_map && hash_table_contains (dl_url_file_map, url))
{
file = xstrdup (hash_table_get (dl_url_file_map, url));
DEBUGP (("Already downloaded \"%s\", reusing it from \"%s\".\n",
url, file));
if (html_allowed
&& downloaded_html_set
&& string_set_contains (downloaded_html_set, file))
descend = 1;
}
else
{
int dt = 0;
char *redirected = NULL;
int oldrec = opt.recursive;
opt.recursive = 0;
status = retrieve_url (url, &file, &redirected, referer, &dt);
opt.recursive = oldrec;
if (html_allowed && file && status == RETROK
&& (dt & RETROKF) && (dt & TEXTHTML))
descend = 1;
if (redirected)
{
/* We have been redirected, possibly to another host, or
different path, or wherever. Check whether we really
want to follow it. */
if (descend)
{
if (!descend_redirect_p (redirected, url, depth,
start_url_parsed, blacklist))
descend = 0;
else
/* Make sure that the old pre-redirect form gets
blacklisted. */
string_set_add (blacklist, url);
}
xfree (url);
url = redirected;
}
}
if (descend
&& depth >= opt.reclevel && opt.reclevel != INFINITE_RECURSION)
{
if (opt.page_requisites
&& (depth == opt.reclevel || depth == opt.reclevel + 1))
{
/* When -p is specified, we are allowed to exceed the
maximum depth, but only for the "inline" links,
i.e. those that are needed to display the page.
Originally this could exceed the depth at most by
one, but we allow one more level so that the leaf
pages that contain frames can be loaded
correctly. */
dash_p_leaf_HTML = TRUE;
}
else
{
/* Either -p wasn't specified or it was and we've
already spent the two extra (pseudo-)levels that it
affords us, so we need to bail out. */
DEBUGP (("Not descending further; at depth %d, max. %d.\n",
depth, opt.reclevel));
descend = 0;
}
}
/* If the downloaded document was HTML, parse it and enqueue the
links it contains. */
if (descend)
{
int meta_disallow_follow = 0;
struct urlpos *children
= get_urls_html (file, url, &meta_disallow_follow);
if (opt.use_robots && meta_disallow_follow)
{
free_urlpos (children);
children = NULL;
}
if (children)
{
struct urlpos *child = children;
struct url *url_parsed = url_parsed = url_parse (url, NULL);
assert (url_parsed != NULL);
for (; child; child = child->next)
{
if (child->ignore_when_downloading)
continue;
if (dash_p_leaf_HTML && !child->link_inline_p)
continue;
if (download_child_p (child, url_parsed, depth, start_url_parsed,
blacklist))
{
url_enqueue (queue, xstrdup (child->url->url),
xstrdup (url), depth + 1,
child->link_expect_html);
/* We blacklist the URL we have enqueued, because we
don't want to enqueue (and hence download) the
same URL twice. */
string_set_add (blacklist, child->url->url);
}
}
url_free (url_parsed);
free_urlpos (children);
}
}
if (opt.delete_after || (file && !acceptable (file)))
{
/* Either --delete-after was specified, or we loaded this
otherwise rejected (e.g. by -R) HTML file just so we
could harvest its hyperlinks -- in either case, delete
the local file. */
DEBUGP (("Removing file due to %s in recursive_retrieve():\n",
opt.delete_after ? "--delete-after" :
"recursive rejection criteria"));
logprintf (LOG_VERBOSE,
(opt.delete_after
? _("Removing %s.\n")
: _("Removing %s since it should be rejected.\n")),
file);
if (unlink (file))
logprintf (LOG_NOTQUIET, "unlink: %s\n", strerror (errno));
register_delete_file (file);
}
xfree (url);
FREE_MAYBE (referer);
FREE_MAYBE (file);
}
/* If anything is left of the queue due to a premature exit, free it
now. */
{
char *d1, *d2;
int d3, d4;
while (url_dequeue (queue,
(const char **)&d1, (const char **)&d2, &d3, &d4))
{
xfree (d1);
FREE_MAYBE (d2);
}
}
url_queue_delete (queue);
if (start_url_parsed)
url_free (start_url_parsed);
string_set_free (blacklist);
if (opt.quota && total_downloaded_bytes > opt.quota)
return QUOTEXC;
else if (status == FWRITEERR)
return FWRITEERR;
else
return RETROK;
}
/* The core of recursive retrieving. Endless recursion is avoided by
having all URLs stored to a linked list of URLs, which is checked
before loading any URL. That way no URL can get loaded twice.
The function also supports specification of maximum recursion depth
and a number of other goodies. */
uerr_t
recursive_retrieve (const char *file, const char *this_url)
{
char *constr, *filename, *newloc;
char *canon_this_url = NULL;
int dt, inl, dash_p_leaf_HTML = FALSE;
int meta_disallow_follow;
int this_url_ftp; /* See below the explanation */
uerr_t err;
struct urlinfo *rurl;
urlpos *url_list, *cur_url;
char *rfile; /* For robots */
struct urlinfo *u;
assert (this_url != NULL);
assert (file != NULL);
/* If quota was exceeded earlier, bail out. */
if (downloaded_exceeds_quota ())
return QUOTEXC;
/* Cache the current URL in the list. */
if (first_time)
{
/* These three operations need to be done only once per Wget
run. They should probably be at a different location. */
if (!undesirable_urls)
undesirable_urls = make_string_hash_table (0);
hash_table_clear (undesirable_urls);
string_set_add (undesirable_urls, this_url);
/* Enter this_url to the hash table, in original and "enhanced" form. */
u = newurl ();
err = parseurl (this_url, u, 0);
if (err == URLOK)
{
string_set_add (undesirable_urls, u->url);
if (opt.no_parent)
base_dir = xstrdup (u->dir); /* Set the base dir. */
/* Set the canonical this_url to be sent as referer. This
problem exists only when running the first time. */
canon_this_url = xstrdup (u->url);
}
else
{
DEBUGP (("Double yuck! The *base* URL is broken.\n"));
base_dir = NULL;
}
freeurl (u, 1);
depth = 1;
robots_host = NULL;
forbidden = NULL;
first_time = 0;
}
else
++depth;
if (opt.reclevel != INFINITE_RECURSION && depth > opt.reclevel)
/* We've exceeded the maximum recursion depth specified by the user. */
{
if (opt.page_requisites && depth <= opt.reclevel + 1)
/* When -p is specified, we can do one more partial recursion from the
"leaf nodes" on the HTML document tree. The recursion is partial in
that we won't traverse any <A> or <AREA> tags, nor any <LINK> tags
except for <LINK REL="stylesheet">. */
dash_p_leaf_HTML = TRUE;
else
/* Either -p wasn't specified or it was and we've already gone the one
extra (pseudo-)level that it affords us, so we need to bail out. */
{
DEBUGP (("Recursion depth %d exceeded max. depth %d.\n",
depth, opt.reclevel));
--depth;
return RECLEVELEXC;
}
}
/* Determine whether this_url is an FTP URL. If it is, it means
that the retrieval is done through proxy. In that case, FTP
links will be followed by default and recursion will not be
turned off when following them. */
this_url_ftp = (urlproto (this_url) == URLFTP);
/* Get the URL-s from an HTML file: */
url_list = get_urls_html (file, canon_this_url ? canon_this_url : this_url,
dash_p_leaf_HTML, &meta_disallow_follow);
if (opt.use_robots && meta_disallow_follow)
{
/* The META tag says we are not to follow this file. Respect
that. */
free_urlpos (url_list);
url_list = NULL;
}
/* Decide what to do with each of the URLs. A URL will be loaded if
it meets several requirements, discussed later. */
for (cur_url = url_list; cur_url; cur_url = cur_url->next)
{
/* If quota was exceeded earlier, bail out. */
if (downloaded_exceeds_quota ())
break;
/* Parse the URL for convenient use in other functions, as well
as to get the optimized form. It also checks URL integrity. */
u = newurl ();
if (parseurl (cur_url->url, u, 0) != URLOK)
{
DEBUGP (("Yuck! A bad URL.\n"));
freeurl (u, 1);
continue;
}
if (u->proto == URLFILE)
{
DEBUGP (("Nothing to do with file:// around here.\n"));
freeurl (u, 1);
continue;
}
assert (u->url != NULL);
constr = xstrdup (u->url);
/* Several checkings whether a file is acceptable to load:
1. check if URL is ftp, and we don't load it
2. check for relative links (if relative_only is set)
3. check for domain
4. check for no-parent
5. check for excludes && includes
6. check for suffix
7. check for same host (if spanhost is unset), with possible
gethostbyname baggage
8. check for robots.txt
Addendum: If the URL is FTP, and it is to be loaded, only the
domain and suffix settings are "stronger".
Note that .html and (yuck) .htm will get loaded regardless of
suffix rules (but that is remedied later with unlink) unless
the depth equals the maximum depth.
More time- and memory- consuming tests should be put later on
the list. */
/* inl is set if the URL we are working on (constr) is stored in
undesirable_urls. Using it is crucial to avoid unnecessary
repeated continuous hits to the hash table. */
inl = string_set_contains (undesirable_urls, constr);
/* If it is FTP, and FTP is not followed, chuck it out. */
if (!inl)
if (u->proto == URLFTP && !opt.follow_ftp && !this_url_ftp)
{
DEBUGP (("Uh, it is FTP but i'm not in the mood to follow FTP.\n"));
string_set_add (undesirable_urls, constr);
inl = 1;
}
/* If it is absolute link and they are not followed, chuck it
out. */
if (!inl && u->proto != URLFTP)
if (opt.relative_only && !cur_url->link_relative_p)
{
DEBUGP (("It doesn't really look like a relative link.\n"));
string_set_add (undesirable_urls, constr);
inl = 1;
}
/* If its domain is not to be accepted/looked-up, chuck it out. */
if (!inl)
if (!accept_domain (u))
{
DEBUGP (("I don't like the smell of that domain.\n"));
string_set_add (undesirable_urls, constr);
inl = 1;
}
/* Check for parent directory. */
if (!inl && opt.no_parent
/* If the new URL is FTP and the old was not, ignore
opt.no_parent. */
&& !(!this_url_ftp && u->proto == URLFTP))
{
/* Check for base_dir first. */
if (!(base_dir && frontcmp (base_dir, u->dir)))
{
/* Failing that, check for parent dir. */
struct urlinfo *ut = newurl ();
if (parseurl (this_url, ut, 0) != URLOK)
DEBUGP (("Double yuck! The *base* URL is broken.\n"));
else if (!frontcmp (ut->dir, u->dir))
{
/* Failing that too, kill the URL. */
DEBUGP (("Trying to escape parental guidance with no_parent on.\n"));
string_set_add (undesirable_urls, constr);
inl = 1;
}
freeurl (ut, 1);
}
}
/* If the file does not match the acceptance list, or is on the
rejection list, chuck it out. The same goes for the
directory exclude- and include- lists. */
if (!inl && (opt.includes || opt.excludes))
{
if (!accdir (u->dir, ALLABS))
{
DEBUGP (("%s (%s) is excluded/not-included.\n", constr, u->dir));
string_set_add (undesirable_urls, constr);
inl = 1;
}
}
if (!inl)
{
char *suf = NULL;
/* We check for acceptance/rejection rules only for non-HTML
documents. Since we don't know whether they really are
HTML, it will be deduced from (an OR-ed list):
1) u->file is "" (meaning it is a directory)
2) suffix exists, AND:
a) it is "html", OR
b) it is "htm"
If the file *is* supposed to be HTML, it will *not* be
subject to acc/rej rules, unless a finite maximum depth has
been specified and the current depth is the maximum depth. */
if (!
(!*u->file
|| (((suf = suffix (constr)) != NULL)
&& ((!strcmp (suf, "html") || !strcmp (suf, "htm"))
&& ((opt.reclevel != INFINITE_RECURSION) &&
(depth != opt.reclevel))))))
{
if (!acceptable (u->file))
{
DEBUGP (("%s (%s) does not match acc/rej rules.\n",
constr, u->file));
string_set_add (undesirable_urls, constr);
inl = 1;
}
}
FREE_MAYBE (suf);
}
/* Optimize the URL (which includes possible DNS lookup) only
after all other possibilities have been exhausted. */
if (!inl)
{
if (!opt.simple_check)
opt_url (u);
else
{
char *p;
/* Just lowercase the hostname. */
for (p = u->host; *p; p++)
*p = TOLOWER (*p);
xfree (u->url);
u->url = str_url (u, 0);
}
xfree (constr);
constr = xstrdup (u->url);
string_set_add (undesirable_urls, constr);
if (!inl && !((u->proto == URLFTP) && !this_url_ftp))
if (!opt.spanhost && this_url && !same_host (this_url, constr))
{
DEBUGP (("This is not the same hostname as the parent's.\n"));
string_set_add (undesirable_urls, constr);
inl = 1;
}
}
/* What about robots.txt? */
if (!inl && opt.use_robots && u->proto == URLHTTP)
{
/* Since Wget knows about only one set of robot rules at a
time, /robots.txt must be reloaded whenever a new host is
accessed.
robots_host holds the host the current `forbid' variable
is assigned to. */
if (!robots_host || !same_host (robots_host, u->host))
{
FREE_MAYBE (robots_host);
/* Now make robots_host the new host, no matter what the
result will be. So if there is no /robots.txt on the
site, Wget will not retry getting robots all the
time. */
robots_host = xstrdup (u->host);
free_vec (forbidden);
forbidden = NULL;
err = retrieve_robots (constr, ROBOTS_FILENAME);
if (err == ROBOTSOK)
{
rurl = robots_url (constr, ROBOTS_FILENAME);
rfile = url_filename (rurl);
forbidden = parse_robots (rfile);
freeurl (rurl, 1);
xfree (rfile);
}
}
/* Now that we have (or don't have) robots, we can check for
them. */
if (!robots_match (u, forbidden))
{
DEBUGP (("Stuffing %s because %s forbids it.\n", this_url,
ROBOTS_FILENAME));
string_set_add (undesirable_urls, constr);
inl = 1;
}
}
filename = NULL;
/* If it wasn't chucked out, do something with it. */
if (!inl)
{
DEBUGP (("I've decided to load it -> "));
/* Add it to the list of already-loaded URL-s. */
string_set_add (undesirable_urls, constr);
/* Automatically followed FTPs will *not* be downloaded
recursively. */
if (u->proto == URLFTP)
{
/* Don't you adore side-effects? */
opt.recursive = 0;
}
/* Reset its type. */
dt = 0;
/* Retrieve it. */
retrieve_url (constr, &filename, &newloc,
canon_this_url ? canon_this_url : this_url, &dt);
if (u->proto == URLFTP)
{
/* Restore... */
opt.recursive = 1;
}
if (newloc)
{
xfree (constr);
constr = newloc;
}
/* If there was no error, and the type is text/html, parse
it recursively. */
if (dt & TEXTHTML)
{
if (dt & RETROKF)
recursive_retrieve (filename, constr);
}
else
DEBUGP (("%s is not text/html so we don't chase.\n",
filename ? filename: "(null)"));
if (opt.delete_after || (filename && !acceptable (filename)))
/* Either --delete-after was specified, or we loaded this otherwise
rejected (e.g. by -R) HTML file just so we could harvest its
hyperlinks -- in either case, delete the local file. */
{
DEBUGP (("Removing file due to %s in recursive_retrieve():\n",
opt.delete_after ? "--delete-after" :
"recursive rejection criteria"));
logprintf (LOG_VERBOSE,
(opt.delete_after ? _("Removing %s.\n")
: _("Removing %s since it should be rejected.\n")),
filename);
if (unlink (filename))
logprintf (LOG_NOTQUIET, "unlink: %s\n", strerror (errno));
dt &= ~RETROKF;
}
/* If everything was OK, and links are to be converted, let's
store the local filename. */
if (opt.convert_links && (dt & RETROKF) && (filename != NULL))
{
cur_url->convert = CO_CONVERT_TO_RELATIVE;
cur_url->local_name = xstrdup (filename);
}
}
else
DEBUGP (("%s already in list, so we don't load.\n", constr));
/* Free filename and constr. */
FREE_MAYBE (filename);
FREE_MAYBE (constr);
freeurl (u, 1);
/* Increment the pbuf for the appropriate size. */
}
if (opt.convert_links && !opt.delete_after)
/* This is merely the first pass: the links that have been
successfully downloaded are converted. In the second pass,
convert_all_links() will also convert those links that have NOT
been downloaded to their canonical form. */
convert_links (file, url_list);
/* Free the linked list of URL-s. */
free_urlpos (url_list);
/* Free the canonical this_url. */
FREE_MAYBE (canon_this_url);
/* Decrement the recursion depth. */
--depth;
if (downloaded_exceeds_quota ())
return QUOTEXC;
else
return RETROK;
}