xref: /dokuwiki/lib/exe/indexer.php (revision 2ba9f431b90f359b7fb95b25d2b8267157fd7996) 
1<?php
2/**
3 * DokuWiki indexer
4 *
5 * @license    GPL 2 (http://www.gnu.org/licenses/gpl.html)
6 * @author     Andreas Gohr <andi@splitbrain.org>
7 */
8if(!defined('DOKU_INC')) define('DOKU_INC',dirname(__FILE__).'/../../');
9define('DOKU_DISABLE_GZIP_OUTPUT',1);
10require_once(DOKU_INC.'inc/init.php');
11session_write_close();  //close session
12if(!defined('NL')) define('NL',"\n");
13
14// Version tag used to force rebuild on upgrade
15define('INDEXER_VERSION', 2);
16
17// keep running after browser closes connection
18@ignore_user_abort(true);
19
20// check if user abort worked, if yes send output early
21$defer = !@ignore_user_abort() || $conf['broken_iua'];
22if(!$defer){
23    sendGIF(); // send gif
24}
25
26$ID = cleanID($_REQUEST['id']);
27
28// Catch any possible output (e.g. errors)
29if(!isset($_REQUEST['debug'])) ob_start();
30
31// run one of the jobs
32$tmp = array(); // No event data
33$evt = new Doku_Event('INDEXER_TASKS_RUN', $tmp);
34if ($evt->advise_before()) {
35  runIndexer() or
36  metaUpdate() or
37  runSitemapper() or
38  sendDigest() or
39  runTrimRecentChanges() or
40  runTrimRecentChanges(true) or
41  $evt->advise_after();
42}
43if($defer) sendGIF();
44
45if(!isset($_REQUEST['debug'])) ob_end_clean();
46exit;
47
48// --------------------------------------------------------------------
49
50/**
51 * Trims the recent changes cache (or imports the old changelog) as needed.
52 *
53 * @param media_changes If the media changelog shall be trimmed instead of
54 * the page changelog
55 *
56 * @author Ben Coburn <btcoburn@silicodon.net>
57 */
58function runTrimRecentChanges($media_changes = false) {
59    global $conf;
60
61    $fn = ($media_changes ? $conf['media_changelog'] : $conf['changelog']);
62
63    // Trim the Recent Changes
64    // Trims the recent changes cache to the last $conf['changes_days'] recent
65    // changes or $conf['recent'] items, which ever is larger.
66    // The trimming is only done once a day.
67    if (@file_exists($fn) &&
68        (@filemtime($fn.'.trimmed')+86400)<time() &&
69        !@file_exists($fn.'_tmp')) {
70            @touch($fn.'.trimmed');
71            io_lock($fn);
72            $lines = file($fn);
73            if (count($lines)<=$conf['recent']) {
74                // nothing to trim
75                io_unlock($fn);
76                return false;
77            }
78
79            io_saveFile($fn.'_tmp', '');          // presave tmp as 2nd lock
80            $trim_time = time() - $conf['recent_days']*86400;
81            $out_lines = array();
82
83            for ($i=0; $i<count($lines); $i++) {
84              $log = parseChangelogLine($lines[$i]);
85              if ($log === false) continue;                      // discard junk
86              if ($log['date'] < $trim_time) {
87                $old_lines[$log['date'].".$i"] = $lines[$i];     // keep old lines for now (append .$i to prevent key collisions)
88              } else {
89                $out_lines[$log['date'].".$i"] = $lines[$i];     // definitely keep these lines
90              }
91            }
92
93            if (count($lines)==count($out_lines)) {
94              // nothing to trim
95              @unlink($fn.'_tmp');
96              io_unlock($fn);
97              return false;
98            }
99
100            // sort the final result, it shouldn't be necessary,
101            //   however the extra robustness in making the changelog cache self-correcting is worth it
102            ksort($out_lines);
103            $extra = $conf['recent'] - count($out_lines);        // do we need extra lines do bring us up to minimum
104            if ($extra > 0) {
105              ksort($old_lines);
106              $out_lines = array_merge(array_slice($old_lines,-$extra),$out_lines);
107            }
108
109            // save trimmed changelog
110            io_saveFile($fn.'_tmp', implode('', $out_lines));
111            @unlink($fn);
112            if (!rename($fn.'_tmp', $fn)) {
113                // rename failed so try another way...
114                io_unlock($fn);
115                io_saveFile($fn, implode('', $out_lines));
116                @unlink($fn.'_tmp');
117            } else {
118                io_unlock($fn);
119            }
120            return true;
121    }
122
123    // nothing done
124    return false;
125}
126
127/**
128 * Runs the indexer for the current page
129 *
130 * @author Andreas Gohr <andi@splitbrain.org>
131 */
132function runIndexer(){
133    global $ID;
134    global $conf;
135    print "runIndexer(): started".NL;
136
137    if(!$ID) return false;
138
139    // check if indexing needed
140    $idxtag = metaFN($ID,'.indexed');
141    if(@file_exists($idxtag)){
142        if(io_readFile($idxtag) >= INDEXER_VERSION){
143            $last = @filemtime($idxtag);
144            if($last > @filemtime(wikiFN($ID))){
145                print "runIndexer(): index for $ID up to date".NL;
146                return false;
147            }
148        }
149    }
150
151    // try to aquire a lock
152    $lock = $conf['lockdir'].'/_indexer.lock';
153    while(!@mkdir($lock,$conf['dmode'])){
154        usleep(50);
155        if(time()-@filemtime($lock) > 60*5){
156            // looks like a stale lock - remove it
157            @rmdir($lock);
158            print "runIndexer(): stale lock removed".NL;
159        }else{
160            print "runIndexer(): indexer locked".NL;
161            return false;
162        }
163    }
164    if($conf['dperm']) chmod($lock, $conf['dperm']);
165
166    // do the work
167    idx_addPage($ID);
168
169    // we're finished - save and free lock
170    io_saveFile(metaFN($ID,'.indexed'),INDEXER_VERSION);
171    @rmdir($lock);
172    print "runIndexer(): finished".NL;
173    return true;
174}
175
176/**
177 * Will render the metadata for the page if not exists yet
178 *
179 * This makes sure pages which are created from outside DokuWiki will
180 * gain their data when viewed for the first time.
181 */
182function metaUpdate(){
183    global $ID;
184    print "metaUpdate(): started".NL;
185
186    if(!$ID) return false;
187    $file = metaFN($ID, '.meta');
188    echo "meta file: $file".NL;
189
190    // rendering needed?
191    if (@file_exists($file)) return false;
192    if (!@file_exists(wikiFN($ID))) return false;
193
194    global $conf;
195
196    // gather some additional info from changelog
197    $info = io_grep($conf['changelog'],
198                    '/^(\d+)\t(\d+\.\d+\.\d+\.\d+)\t'.preg_quote($ID,'/').'\t([^\t]+)\t([^\t\n]+)/',
199                    0,true);
200
201    $meta = array();
202    if(!empty($info)){
203        $meta['date']['created'] = $info[0][1];
204        foreach($info as $item){
205            if($item[4] != '*'){
206                $meta['date']['modified'] = $item[1];
207                if($item[3]){
208                    $meta['contributor'][$item[3]] = $item[3];
209                }
210            }
211        }
212    }
213
214    $meta = p_render_metadata($ID, $meta);
215    io_saveFile($file, serialize($meta));
216
217    echo "metaUpdate(): finished".NL;
218    return true;
219}
220
221/**
222 * Builds a Google Sitemap of all public pages known to the indexer
223 *
224 * The map is placed in the root directory named sitemap.xml.gz - This
225 * file needs to be writable!
226 *
227 * @author Andreas Gohr
228 * @link   https://www.google.com/webmasters/sitemaps/docs/en/about.html
229 */
230function runSitemapper(){
231    global $conf;
232    print "runSitemapper(): started".NL;
233    if(!$conf['sitemap']) return false;
234
235    if($conf['compression'] == 'bz2' || $conf['compression'] == 'gz'){
236        $sitemap = 'sitemap.xml.gz';
237    }else{
238        $sitemap = 'sitemap.xml';
239    }
240    print "runSitemapper(): using $sitemap".NL;
241
242    if(@file_exists(DOKU_INC.$sitemap)){
243        if(!is_writable(DOKU_INC.$sitemap)) return false;
244    }else{
245        if(!is_writable(DOKU_INC)) return false;
246    }
247
248    if(@filesize(DOKU_INC.$sitemap) &&
249       @filemtime(DOKU_INC.$sitemap) > (time()-($conf['sitemap']*60*60*24))){
250       print 'runSitemapper(): Sitemap up to date'.NL;
251       return false;
252    }
253
254    $pages = idx_getIndex('page', '');
255    print 'runSitemapper(): creating sitemap using '.count($pages).' pages'.NL;
256
257    // build the sitemap
258    ob_start();
259    print '<?xml version="1.0" encoding="UTF-8"?>'.NL;
260    print '<urlset xmlns="http://www.sitemaps.org/schemas/sitemap/0.9">'.NL;
261    foreach($pages as $id){
262        $id = trim($id);
263        $file = wikiFN($id);
264
265        //skip hidden, non existing and restricted files
266        if(isHiddenPage($id)) continue;
267        $date = @filemtime($file);
268        if(!$date) continue;
269        if(auth_aclcheck($id,'','') < AUTH_READ) continue;
270
271        print '  <url>'.NL;
272        print '    <loc>'.wl($id,'',true).'</loc>'.NL;
273        print '    <lastmod>'.date_iso8601($date).'</lastmod>'.NL;
274        print '  </url>'.NL;
275    }
276    print '</urlset>'.NL;
277    $data = ob_get_contents();
278    ob_end_clean();
279
280    //save the new sitemap
281    io_saveFile(DOKU_INC.$sitemap,$data);
282
283    //ping search engines...
284    $http = new DokuHTTPClient();
285    $http->timeout = 8;
286
287    //ping google
288    print 'runSitemapper(): pinging google'.NL;
289    $url  = 'http://www.google.com/webmasters/sitemaps/ping?sitemap=';
290    $url .= urlencode(DOKU_URL.$sitemap);
291    $resp = $http->get($url);
292    if($http->error) print 'runSitemapper(): '.$http->error.NL;
293    print 'runSitemapper(): '.preg_replace('/[\n\r]/',' ',strip_tags($resp)).NL;
294
295    //ping yahoo
296    print 'runSitemapper(): pinging yahoo'.NL;
297    $url  = 'http://search.yahooapis.com/SiteExplorerService/V1/updateNotification?appid=dokuwiki&url=';
298    $url .= urlencode(DOKU_URL.$sitemap);
299    $resp = $http->get($url);
300    if($http->error) print 'runSitemapper(): '.$http->error.NL;
301    print 'runSitemapper(): '.preg_replace('/[\n\r]/',' ',strip_tags($resp)).NL;
302
303    //ping microsoft
304    print 'runSitemapper(): pinging microsoft'.NL;
305    $url  = 'http://www.bing.com/webmaster/ping.aspx?siteMap=';
306    $url .= urlencode(DOKU_URL.$sitemap);
307    $resp = $http->get($url);
308    if($http->error) print 'runSitemapper(): '.$http->error.NL;
309    print 'runSitemapper(): '.preg_replace('/[\n\r]/',' ',strip_tags($resp)).NL;
310
311    print 'runSitemapper(): finished'.NL;
312    return true;
313}
314
315/**
316 * Send digest and list mails for all subscriptions which are in effect for the
317 * current page
318 *
319 * @author Adrian Lang <lang@cosmocode.de>
320 */
321function sendDigest() {
322    echo 'sendDigest(): start'.NL;
323    global $ID;
324    global $conf;
325    if (!$conf['subscribers']) {
326        return;
327    }
328    $subscriptions = subscription_find($ID, array('style' => '(digest|list)',
329                                                  'escaped' => true));
330    global $auth;
331    global $lang;
332    global $conf;
333    global $USERINFO;
334
335    // remember current user info
336    $olduinfo = $USERINFO;
337    $olduser  = $_SERVER['REMOTE_USER'];
338
339    foreach($subscriptions as $id => $users) {
340        if (!subscription_lock($id)) {
341            continue;
342        }
343        foreach($users as $data) {
344            list($user, $style, $lastupdate) = $data;
345            $lastupdate = (int) $lastupdate;
346            if ($lastupdate + $conf['subscribe_time'] > time()) {
347                // Less than the configured time period passed since last
348                // update.
349                continue;
350            }
351
352            // Work as the user to make sure ACLs apply correctly
353            $USERINFO = $auth->getUserData($user);
354            $_SERVER['REMOTE_USER'] = $user;
355            if ($USERINFO === false) {
356                continue;
357            }
358
359            if (substr($id, -1, 1) === ':') {
360                // The subscription target is a namespace
361                $changes = getRecentsSince($lastupdate, null, getNS($id));
362            } else {
363                if(auth_quickaclcheck($id) < AUTH_READ) continue;
364
365                $meta = p_get_metadata($id);
366                $changes = array($meta['last_change']);
367            }
368
369            // Filter out pages only changed in small and own edits
370            $change_ids = array();
371            foreach($changes as $rev) {
372                $n = 0;
373                while (!is_null($rev) && $rev['date'] >= $lastupdate &&
374                       ($_SERVER['REMOTE_USER'] === $rev['user'] ||
375                        $rev['type'] === DOKU_CHANGE_TYPE_MINOR_EDIT)) {
376                    $rev = getRevisions($rev['id'], $n++, 1);
377                    $rev = (count($rev) > 0) ? $rev[0] : null;
378                }
379
380                if (!is_null($rev) && $rev['date'] >= $lastupdate) {
381                    // Some change was not a minor one and not by myself
382                    $change_ids[] = $rev['id'];
383                }
384            }
385
386            if ($style === 'digest') {
387                foreach($change_ids as $change_id) {
388                    subscription_send_digest($USERINFO['mail'], $change_id,
389                                             $lastupdate);
390                }
391            } elseif ($style === 'list') {
392                subscription_send_list($USERINFO['mail'], $change_ids, $id);
393            }
394            // TODO: Handle duplicate subscriptions.
395
396            // Update notification time.
397            subscription_set($user, $id, $style, time(), true);
398        }
399        subscription_unlock($id);
400    }
401
402    // restore current user info
403    $USERINFO = $olduinfo;
404    $_SERVER['REMOTE_USER'] = $olduser;
405}
406
407/**
408 * Formats a timestamp as ISO 8601 date
409 *
410 * @author <ungu at terong dot com>
411 * @link http://www.php.net/manual/en/function.date.php#54072
412 */
413function date_iso8601($int_date) {
414   //$int_date: current date in UNIX timestamp
415   $date_mod = date('Y-m-d\TH:i:s', $int_date);
416   $pre_timezone = date('O', $int_date);
417   $time_zone = substr($pre_timezone, 0, 3).":".substr($pre_timezone, 3, 2);
418   $date_mod .= $time_zone;
419   return $date_mod;
420}
421
422/**
423 * Just send a 1x1 pixel blank gif to the browser
424 *
425 * @author Andreas Gohr <andi@splitbrain.org>
426 * @author Harry Fuecks <fuecks@gmail.com>
427 */
428function sendGIF(){
429    if(isset($_REQUEST['debug'])){
430        header('Content-Type: text/plain');
431        return;
432    }
433    $img = base64_decode('R0lGODlhAQABAIAAAAAAAP///yH5BAEAAAEALAAAAAABAAEAAAIBTAA7');
434    header('Content-Type: image/gif');
435    header('Content-Length: '.strlen($img));
436    header('Connection: Close');
437    print $img;
438    flush();
439    // Browser should drop connection after this
440    // Thinks it's got the whole image
441}
442
443//Setup VIM: ex: et ts=4 enc=utf-8 :
444// No trailing PHP closing tag - no output please!
445// See Note at http://www.php.net/manual/en/language.basic-syntax.instruction-separation.php
446