httpenum-depth.c

harvest是一个下载html网页得机器人

#ifdef DONT_RETRY_FAILS
		mark_failed (up);
#endif

		return (NULL);
	}
	if (up->md5 && md5_in_db (up->md5)) {	/* Have we been here? */
		Debug (42, 1, ("Already Visited MD5: %s\n", up->url));
		return (NULL);
	}
	/* Remember that we've been here before */
	if (up->md5)
		mark_retrieved (up);

	/* Are we dealing with HTML or SOIF, if not we can't get href links */
	if (up->http_mime_hdr == NULL) {
		fprintf (stdout, "%s\t%s\n", up->url, up->md5);
		fflush (stdout);
		return (NULL);
	}

	/* If we were given an HTML document, then use HTMLurls to summarise it */
	if (strstr (up->http_mime_hdr, "text/html") != NULL) {
		/* Extract the HREF's */
		if (pipe (pipefds) < 0) {
			log_errno ("pipe");
			return (NULL);
		}
		if ((pid = fork ()) < 0) {
			log_errno ("fork");
			return (NULL);
		}
		if (pid == 0) {	/* child: HTMLurls */
			char *argv[64];
#if 0 /* kjl/7mar2002 */
			char *enum_url = NULL;

			enum_url = (char *) xmalloc (strlen (up->url) + 20);
			sprintf (enum_url, "ENUMERATOR_URL=%s", up->url);
			putenv (enum_url);
#endif

			close (pipefds[0]);	/* child wont read from pipe */
			dup2 (pipefds[1], 1);	/* stdout -> write:pipe */
			close (pipefds[1]);	/* close pipe, its now stdout */

			argv[0] = "HTMLurls";
			argv[1] = "--base-url";
			argv[2] = up->url;
			argv[3] = up->filename;
			argv[4] = NULL;

			execvp (argv[0], argv);
			sprintf (buf, "execvp: %s", argv[0]);
			log_errno (buf);
			_exit (1);
		}
		close (pipefds[1]);	/* parent wont write */
		if ((fp = fdopen (pipefds[0], "r")) == NULL) {
			log_errno ("fdopen");
			return (NULL);
		}
		/*
		 *  For each HREF pointer, convert it to a URL, and add it to
		 *  the list of URLs to return.
		 */
		Tail = &head;
		while (fgets (buf, BUFSIZ, fp) != NULL) {
			if ((s = strrchr (buf, '\n')) != NULL)
				*s = '\0';	/* strip newline */
			Debug (42, 1, ("httpenum-depth: Input: %s\n", buf));

			/* Make a link in the list for this URL */
			l = (list_t *) xmalloc (sizeof (list_t));
			l->ptr = (void *) xstrdup (buf);
			l->next = (list_t *) NULL;
			*Tail = l;
			Tail = &(l->next);

			nurls++;
		}
		fclose (fp);
		close (pipefds[0]);

		/* Wait for HTMLurls to finish and decide on whether to index the page
		 * according to its exit code
		 */
		if ((err = waitpid (pid, &status, 0)) != pid) {
			Debug (42, 1,
			       ("WARNING: waiting for child %d got %d...\n",
				pid, err));
		}

		/* If we had an exit code of 0 then index it, otherwise don't */
		if (WEXITSTATUS (status) == 0) {
			fprintf (stdout, "%s\t%s\n", up->url, up->md5);
			fflush (stdout);
		}
	} else if (strstr (up->http_mime_hdr, "text/x-soif") != NULL) {
		/* If its a SOIF document then get it from the database */
		char *refs = urldb_getrefs (up->url);
		char *ptr;
		char *abs_ref;

		Log ("%s\tNot Modified\n", up->url);

		/* Flush the URL */
		fprintf (stdout, "%s\t%s\n", up->url, up->md5);
		fflush (stdout);

		Tail = &head;
		for (ptr = strtok (refs, "\n"); ptr; ptr = strtok (NULL, "\n")) {
			abs_ref = url_parse_relative (ptr, up->url);
			if (url_is_allowed (abs_ref)) {
				/* Make an absolute link in the list for this URL */
				l = (list_t *) xmalloc (sizeof (list_t));
				l->ptr = (void *) abs_ref;
				l->next = (list_t *) NULL;
				*Tail = l;
				Tail = &(l->next);

				nurls++;
			} else {
				xfree (abs_ref);
			}
		}
		xfree (refs);
	} else {
		fprintf (stdout, "%s\t%s\n", up->url, up->md5);
		fflush (stdout);
		return (NULL);
	}

	Debug (42, 1, ("Adding %d URLs from %s to workload\n", nurls, up->url));
	return (head);
}

/*
 *  process_url() - Retrieves the given URL (HTTP only), computes an MD5,
 *  and extracts the list of HREF pointers within the HTML document.
 *  If any of the HREF pointers are HTTP and the same host as the given
 *  URL up, then it calls process_url() recursively.
 *  up is closed on exit.
 */
static void
process_url (up, depth)
URL *up;
int depth;
{
	list_t *head = NULL;
	list_t *l = NULL;
	list_t *next_l = NULL;
	char *url = NULL;
	URL *tup = NULL;

	if (max_depth > 0 && depth > (max_depth + 1)) {
		if (not_visited)
			fprintf (not_visited, "[DEPTH] %s\n", up->url);
		Debug (42, 1, ("Maximum Depth of %d Reached: %s\n",
			       max_depth, up->url));
		url_close (up);
		return;
	}
	Debug (42, 1, ("Processing: [%2d] %s\n", depth - 1, up->url));
	if (depth - 1 < 100)
		depth_hist[depth - 1]++;

	if ((head = http_enum (up)) == NULL) {
		url_close (up);
		return;
	}

	url_close (up);

	/*
	 *  Now, for each URL in the list, call process_url() if
	 *  the URL is an HTTP url and it is on the same host
	 */
	for (l = head; l; l = next_l) {
		next_l = l->next;
		url = (char *) l->ptr;
		if ((url != NULL) && ((tup = url_open (url)) != NULL)) {
			if ((tup->type != URL_HTTP)) {
				Debug (42, 1, ("Examining: %s\n", tup->url));
				/* filter_selection() checks access_mask */
				if (!filter_selection (tup)) {
					int re_depth;
					/* Print URL with bogus MD5 to enumerate; */
					/* flush to keep pipe moving */
					/* URL <tab> MD5 */
					re_depth = max_depth - depth + 2;
					if (re_depth < 1)
						re_depth = 1;
					fprintf (stdout, "%s\tDepth=%d\n",
						 tup->url, re_depth);
					fflush (stdout);
					Debug (42, 1,
					       ("HTTPENUM Re-enumeration: %s\tDepth=%d\n",
						tup->url, re_depth));
				}
				url_close (tup);
			} else {
				process_url (tup, depth + 1);
			}
		}

		/* Free List Entry */
		xfree (l->ptr);
		l->ptr = (void *) NULL;
		l->next = (list_t *) NULL;
		xfree (l);
		l = (list_t *) NULL;
	}
}

/* ---------------------------------------------------------------------- */

/*
 *  initialize() - Basic init routines
 */
static void
initialize ()
{
	char *s;
	extern int liburl_conform_rfc1738;
	FILE *logfp = NULL;

#ifdef USE_HOST_CACHE
	host_cache_init ();
#endif

	cur_depth = max_depth = url_max = host_max = 0;
	if ((s = getenv ("HARVEST_URL_MAX")) != NULL)
		url_max = atoi (s);
	if ((s = getenv ("HARVEST_HOST_MAX")) != NULL)
		host_max = atoi (s);
	if ((s = getenv ("HARVEST_DEPTH_MAX")) != NULL)
		max_depth = atoi (s);
	if ((s = getenv ("HARVEST_DEPTH_CUR")) != NULL)
		cur_depth = atoi (s);
	if (url_max < 1)
		url_max = 250;	/* hard-coded maximum */
	if (host_max < 1)
		host_max = 1;	/* hard-coded maximum */
	if (max_depth < 1)
		max_depth = 0;	/* hard-coded maximum */
	host_filterfile = getenv ("HARVEST_HOST_FILTER");
	url_filterfile = getenv ("HARVEST_URL_FILTER");
	access_types = getenv ("HARVEST_ACCESS_TYPES");

	if ((s = getenv ("HARVEST_GATHERER_LOGFILE")) != (char *) NULL)
		logfp = fopen (s, "a+");
	if (logfp == (FILE *) NULL)
		logfp = stderr;
	init_log3 ("httpenum-depth", logfp, stderr);
	init_url ();
	liburl_conform_rfc1738 = 1;
	filter_initialize ();
	Debug (42, 5, ("access_mask: %#02X\n", access_mask));

	/* Open GDBM databases to keep track of where we've been */
	urldb_filename = xstrdup (tempnam (NULL, "Hurl"));
	urldbf = gdbm_open (urldb_filename, 0, GDBM_NEWDB, 0644, NULL);
	if (urldbf == NULL) {
		log_errno (urldb_filename);
		fatal ("gdbm_open: %s: %s", urldb_filename,
		       gdbm_strerror (gdbm_errno));
	}
	hostdb_filename = xstrdup (tempnam (NULL, "Hhost"));
	hostdbf = gdbm_open (hostdb_filename, 0, GDBM_NEWDB, 0644, NULL);
	if (hostdbf == NULL) {
		log_errno (hostdb_filename);
		fatal ("gdbm_open: %s: %s", hostdb_filename,
		       gdbm_strerror (gdbm_errno));
	}
	md5db_filename = xstrdup (tempnam (NULL, "Hmd5"));
	md5dbf = gdbm_open (md5db_filename, 0, GDBM_NEWDB, 0644, NULL);
	if (md5dbf == NULL) {
		log_errno (md5db_filename);
		fatal ("gdbm_open: %s: %s", md5db_filename,
		       gdbm_strerror (gdbm_errno));
	}
	/* open not-visited file */
	if ((s = getenv ("HARVEST_NOT_VISITED_LOG")) != NULL)
		not_visited = fopen (s, "a+");
	if (not_visited)
		setbuf (not_visited, NULL);
}

/* Die gracefully */
static void
sigdie (x)
int x;
{
	int i;

#ifdef USE_HOST_CACHE
	dump_host_cache (42, 9);
#endif
	if (urldbf != NULL)
		gdbm_close (urldbf);
	if (hostdbf != NULL)
		gdbm_close (hostdbf);
	if (md5dbf != NULL)
		gdbm_close (md5dbf);
	if (not_visited)
		fclose (not_visited);

	/* (void) unlink (urldb_filename); */
	crremove (urldb_filename);
	xfree (urldb_filename);
	/* (void) unlink (hostdb_filename); */
	crremove (hostdb_filename);
	xfree (hostdb_filename);
	/* (void) unlink (md5db_filename); */
	crremove (md5db_filename);
	xfree (md5db_filename);

	for (i = 0; i < 100; i++) {
		if (i > max_depth && depth_hist[i] == 0)
			break;
		Log ("Found %8d objects at depth %d\n", depth_hist[i], i);
	}

	Debug (42, 1, ("httpenum-depth: exiting (signal %d)\n", x));
	exit (0);
}

/* ---------------------------------------------------------------------- */

static void
usage ()
{
	fprintf (stderr, "Usage: httpenum-depth http-URL\n");
	exit (1);
}

int
main (argc, argv)
int argc;
char **argv;
{
	URL *up = NULL;
	int depth = 0;

	debug_init ();		/* from $HARVEST_DEBUG */
	for (argc--, argv++; argc > 0 && **argv == '-'; argc--, argv++) {
		if (!strncmp (*argv, "-D", 2)) {
			debug_flag (*argv);
		}
	}

	if (argc != 1)
		usage ();

	for (depth = 0; depth < 100; depth++)
		depth_hist[depth] = 0;

	signal (SIGTERM, sigdie);	/* Die gracefully */
	signal (SIGINT, sigdie);
	signal (SIGPIPE, sigdie);	/* Quickly clean up on broken pipe */

	initialize ();		/* Initialize */

	Debug (42, 1, ("httpenum-depth: Starting...\n"));

	/* Grab the RootNode URL from the command line */
	if ((up = url_open (*argv)) == NULL || up->type != URL_HTTP) {
		usage ();
	}
	/* Mark the RootNode */
	tree_root = xstrdup (up->url);
	printf ("%s\n", up->url);	/* Print tree root */

	process_url (up, cur_depth + 1);	/* Do the Enumeration recursively */
	finish_url ();
	sigdie (0);
}