xref: /dokuwiki/lib/exe/indexer.php (revision c275533eba3389cf4c186d995947aedfff356177) 
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 $conf;
135    print "runIndexer(): started".NL;
136
137    // Move index files (if needed)
138    // Uses the importoldindex plugin to upgrade the index automatically.
139    // FIXME: Remove this from runIndexer when it is no longer needed.
140    if (@file_exists($conf['cachedir'].'/page.idx') &&
141        (!@file_exists($conf['indexdir'].'/page.idx') ||
142         !filesize($conf['indexdir'].'/page.idx'))  &&
143        !@file_exists($conf['indexdir'].'/index_importing')) {
144        echo "trigger TEMPORARY_INDEX_UPGRADE_EVENT\n";
145        $tmp = array(); // no event data
146        trigger_event('TEMPORARY_INDEX_UPGRADE_EVENT', $tmp);
147    }
148
149    if(!$ID) return false;
150
151    // check if indexing needed
152    $idxtag = metaFN($ID,'.indexed');
153    if(@file_exists($idxtag)){
154        if(io_readFile($idxtag) >= INDEXER_VERSION){
155            $last = @filemtime($idxtag);
156            if($last > @filemtime(wikiFN($ID))){
157                print "runIndexer(): index for $ID up to date".NL;
158                return false;
159            }
160        }
161    }
162
163    // try to aquire a lock
164    $lock = $conf['lockdir'].'/_indexer.lock';
165    while(!@mkdir($lock,$conf['dmode'])){
166        usleep(50);
167        if(time()-@filemtime($lock) > 60*5){
168            // looks like a stale lock - remove it
169            @rmdir($lock);
170            print "runIndexer(): stale lock removed".NL;
171        }else{
172            print "runIndexer(): indexer locked".NL;
173            return false;
174        }
175    }
176    if($conf['dperm']) chmod($lock, $conf['dperm']);
177
178    // upgrade to version 2
179    if (!@file_exists($conf['indexdir'].'/pageword.idx'))
180        idx_upgradePageWords();
181
182    // do the work
183    idx_addPage($ID);
184
185    // we're finished - save and free lock
186    io_saveFile(metaFN($ID,'.indexed'),INDEXER_VERSION);
187    @rmdir($lock);
188    print "runIndexer(): finished".NL;
189    return true;
190}
191
192/**
193 * Will render the metadata for the page if not exists yet
194 *
195 * This makes sure pages which are created from outside DokuWiki will
196 * gain their data when viewed for the first time.
197 */
198function metaUpdate(){
199    global $ID;
200    print "metaUpdate(): started".NL;
201
202    if(!$ID) return false;
203    $file = metaFN($ID, '.meta');
204    echo "meta file: $file".NL;
205
206    // rendering needed?
207    if (@file_exists($file)) return false;
208    if (!@file_exists(wikiFN($ID))) return false;
209
210    global $conf;
211
212    // gather some additional info from changelog
213    $info = io_grep($conf['changelog'],
214                    '/^(\d+)\t(\d+\.\d+\.\d+\.\d+)\t'.preg_quote($ID,'/').'\t([^\t]+)\t([^\t\n]+)/',
215                    0,true);
216
217    $meta = array();
218    if(!empty($info)){
219        $meta['date']['created'] = $info[0][1];
220        foreach($info as $item){
221            if($item[4] != '*'){
222                $meta['date']['modified'] = $item[1];
223                if($item[3]){
224                    $meta['contributor'][$item[3]] = $item[3];
225                }
226            }
227        }
228    }
229
230    $meta = p_render_metadata($ID, $meta);
231    io_saveFile($file, serialize($meta));
232
233    echo "metaUpdate(): finished".NL;
234    return true;
235}
236
237/**
238 * Builds a Google Sitemap of all public pages known to the indexer
239 *
240 * The map is placed in the root directory named sitemap.xml.gz - This
241 * file needs to be writable!
242 *
243 * @author Andreas Gohr
244 * @link   https://www.google.com/webmasters/sitemaps/docs/en/about.html
245 */
246function runSitemapper(){
247    global $conf;
248    print "runSitemapper(): started".NL;
249    if(!$conf['sitemap']) return false;
250
251    if($conf['compression'] == 'bz2' || $conf['compression'] == 'gz'){
252        $sitemap = 'sitemap.xml.gz';
253    }else{
254        $sitemap = 'sitemap.xml';
255    }
256    print "runSitemapper(): using $sitemap".NL;
257
258    if(@file_exists(DOKU_INC.$sitemap)){
259        if(!is_writable(DOKU_INC.$sitemap)) return false;
260    }else{
261        if(!is_writable(DOKU_INC)) return false;
262    }
263
264    if(@filesize(DOKU_INC.$sitemap) &&
265       @filemtime(DOKU_INC.$sitemap) > (time()-($conf['sitemap']*60*60*24))){
266       print 'runSitemapper(): Sitemap up to date'.NL;
267       return false;
268    }
269
270    $pages = file($conf['indexdir'].'/page.idx');
271    print 'runSitemapper(): creating sitemap using '.count($pages).' pages'.NL;
272
273    // build the sitemap
274    ob_start();
275    print '<?xml version="1.0" encoding="UTF-8"?>'.NL;
276    print '<urlset xmlns="http://www.sitemaps.org/schemas/sitemap/0.9">'.NL;
277    foreach($pages as $id){
278        $id = trim($id);
279        $file = wikiFN($id);
280
281        //skip hidden, non existing and restricted files
282        if(isHiddenPage($id)) continue;
283        $date = @filemtime($file);
284        if(!$date) continue;
285        if(auth_aclcheck($id,'','') < AUTH_READ) continue;
286
287        print '  <url>'.NL;
288        print '    <loc>'.wl($id,'',true).'</loc>'.NL;
289        print '    <lastmod>'.date_iso8601($date).'</lastmod>'.NL;
290        print '  </url>'.NL;
291    }
292    print '</urlset>'.NL;
293    $data = ob_get_contents();
294    ob_end_clean();
295
296    //save the new sitemap
297    io_saveFile(DOKU_INC.$sitemap,$data);
298
299    //ping search engines...
300    $http = new DokuHTTPClient();
301    $http->timeout = 8;
302
303    //ping google
304    print 'runSitemapper(): pinging google'.NL;
305    $url  = 'http://www.google.com/webmasters/sitemaps/ping?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    //ping yahoo
312    print 'runSitemapper(): pinging yahoo'.NL;
313    $url  = 'http://search.yahooapis.com/SiteExplorerService/V1/updateNotification?appid=dokuwiki&url=';
314    $url .= urlencode(DOKU_URL.$sitemap);
315    $resp = $http->get($url);
316    if($http->error) print 'runSitemapper(): '.$http->error.NL;
317    print 'runSitemapper(): '.preg_replace('/[\n\r]/',' ',strip_tags($resp)).NL;
318
319    //ping microsoft
320    print 'runSitemapper(): pinging microsoft'.NL;
321    $url  = 'http://www.bing.com/webmaster/ping.aspx?siteMap=';
322    $url .= urlencode(DOKU_URL.$sitemap);
323    $resp = $http->get($url);
324    if($http->error) print 'runSitemapper(): '.$http->error.NL;
325    print 'runSitemapper(): '.preg_replace('/[\n\r]/',' ',strip_tags($resp)).NL;
326
327    print 'runSitemapper(): finished'.NL;
328    return true;
329}
330
331/**
332 * Send digest and list mails for all subscriptions which are in effect for the
333 * current page
334 *
335 * @author Adrian Lang <lang@cosmocode.de>
336 */
337function sendDigest() {
338    echo 'sendDigest(): start'.NL;
339    global $ID;
340    global $conf;
341    if (!$conf['subscribers']) {
342        return;
343    }
344    $subscriptions = subscription_find($ID, array('style' => '(digest|list)',
345                                                  'escaped' => true));
346    global $auth;
347    global $lang;
348    global $conf;
349    global $USERINFO;
350
351    // remember current user info
352    $olduinfo = $USERINFO;
353    $olduser  = $_SERVER['REMOTE_USER'];
354
355    foreach($subscriptions as $id => $users) {
356        if (!subscription_lock($id)) {
357            continue;
358        }
359        foreach($users as $data) {
360            list($user, $style, $lastupdate) = $data;
361            $lastupdate = (int) $lastupdate;
362            if ($lastupdate + $conf['subscribe_time'] > time()) {
363                // Less than the configured time period passed since last
364                // update.
365                continue;
366            }
367
368            // Work as the user to make sure ACLs apply correctly
369            $USERINFO = $auth->getUserData($user);
370            $_SERVER['REMOTE_USER'] = $user;
371            if ($USERINFO === false) {
372                continue;
373            }
374
375            if (substr($id, -1, 1) === ':') {
376                // The subscription target is a namespace
377                $changes = getRecentsSince($lastupdate, null, getNS($id));
378                if (count($changes) === 0) {
379                    continue;
380                }
381                if ($style === 'digest') {
382                    foreach($changes as $change) {
383                        subscription_send_digest($USERINFO['mail'], $change,
384                                                 $lastupdate);
385                    }
386                } elseif ($style === 'list') {
387                    subscription_send_list($USERINFO['mail'], $changes, $id);
388                }
389                // TODO: Handle duplicate subscriptions.
390            } else {
391                if(auth_quickaclcheck($id) < AUTH_READ) continue;
392
393                $meta = p_get_metadata($id);
394                $rev = $meta['last_change']['date'];
395                if ($rev < $lastupdate) {
396                    // There is no new revision.
397                    continue;
398                }
399                subscription_send_digest($USERINFO['mail'], $meta['last_change'],
400                                         $lastupdate);
401            }
402            // Update notification time.
403            subscription_set($user, $id, $style, time(), true);
404        }
405        subscription_unlock($id);
406    }
407
408    // restore current user info
409    $USERINFO = $olduinfo;
410    $_SERVER['REMOTE_USER'] = $olduser;
411}
412
413/**
414 * Formats a timestamp as ISO 8601 date
415 *
416 * @author <ungu at terong dot com>
417 * @link http://www.php.net/manual/en/function.date.php#54072
418 */
419function date_iso8601($int_date) {
420   //$int_date: current date in UNIX timestamp
421   $date_mod = date('Y-m-d\TH:i:s', $int_date);
422   $pre_timezone = date('O', $int_date);
423   $time_zone = substr($pre_timezone, 0, 3).":".substr($pre_timezone, 3, 2);
424   $date_mod .= $time_zone;
425   return $date_mod;
426}
427
428/**
429 * Just send a 1x1 pixel blank gif to the browser
430 *
431 * @author Andreas Gohr <andi@splitbrain.org>
432 * @author Harry Fuecks <fuecks@gmail.com>
433 */
434function sendGIF(){
435    if(isset($_REQUEST['debug'])){
436        header('Content-Type: text/plain');
437        return;
438    }
439    $img = base64_decode('R0lGODlhAQABAIAAAAAAAP///yH5BAEAAAEALAAAAAABAAEAAAIBTAA7');
440    header('Content-Type: image/gif');
441    header('Content-Length: '.strlen($img));
442    header('Connection: Close');
443    print $img;
444    flush();
445    // Browser should drop connection after this
446    // Thinks it's got the whole image
447}
448
449//Setup VIM: ex: et ts=4 enc=utf-8 :
450// No trailing PHP closing tag - no output please!
451// See Note at http://www.php.net/manual/en/language.basic-syntax.instruction-separation.php
452