xref: /dokuwiki/lib/exe/indexer.php (revision c06cd517522c06ce1095926770bef7a66cd2714e) 
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');
11require_once(DOKU_INC.'inc/auth.php');
12require_once(DOKU_INC.'inc/events.php');
13session_write_close();  //close session
14if(!defined('NL')) define('NL',"\n");
15
16// Version tag used to force rebuild on upgrade
17define('INDEXER_VERSION', 2);
18
19// keep running after browser closes connection
20@ignore_user_abort(true);
21
22// check if user abort worked, if yes send output early
23if(@ignore_user_abort() && !$conf['broken_iua']){
24    sendGIF(); // send gif
25    $defer = false;
26}else{
27    $defer = true;
28}
29
30$ID = cleanID($_REQUEST['id']);
31
32// Catch any possible output (e.g. errors)
33if(!$_REQUEST['debug']) ob_start();
34
35// run one of the jobs
36$tmp = array();
37$evt = new Doku_Event('INDEXER_TASKS_RUN', $tmp);
38if ($evt->advise_before()) {
39  runIndexer() or
40  metaUpdate() or
41  runSitemapper() or
42  runTrimRecentChanges() or
43  $evt->advise_after();
44}
45if($defer) sendGIF();
46
47if(!$_REQUEST['debug']) ob_end_clean();
48exit;
49
50// --------------------------------------------------------------------
51
52/**
53 * Trims the recent changes cache (or imports the old changelog) as needed.
54 *
55 * @author Ben Coburn <btcoburn@silicodon.net>
56 */
57function runTrimRecentChanges() {
58    global $conf;
59
60    // Import old changelog (if needed)
61    // Uses the imporoldchangelog plugin to upgrade the changelog automaticaly.
62    // FIXME: Remove this from runTrimRecentChanges when it is no longer needed.
63    if (isset($conf['changelog_old']) &&
64        @file_exists($conf['changelog_old']) && !@file_exists($conf['changelog']) &&
65        !@file_exists($conf['changelog'].'_importing') && !@file_exists($conf['changelog'].'_tmp')) {
66            $tmp = array(); // no event data
67            trigger_event('TEMPORARY_CHANGELOG_UPGRADE_EVENT', $tmp);
68            return true;
69    }
70
71    // Trim the Recent Changes
72    // Trims the recent changes cache to the last $conf['changes_days'] recent
73    // changes or $conf['recent'] items, which ever is larger.
74    // The trimming is only done once a day.
75    if (@file_exists($conf['changelog']) &&
76        (filectime($conf['changelog'])+86400)<time() &&
77        !@file_exists($conf['changelog'].'_tmp')) {
78            io_lock($conf['changelog']);
79            $lines = file($conf['changelog']);
80            if (count($lines)<=$conf['recent']) {
81                // nothing to trim
82                io_unlock($conf['changelog']);
83                return false;
84            }
85
86            io_saveFile($conf['changelog'].'_tmp', '');          // presave tmp as 2nd lock
87            $trim_time = time() - $conf['recent_days']*86400;
88            $out_lines = array();
89
90            for ($i=0; $i<count($lines); $i++) {
91              $log = parseChangelogLine($lines[$i]);
92              if ($log === false) continue;                      // discard junk
93              if ($log['date'] < $trim_time) {
94                $old_lines[$log['date'].".$i"] = $lines[$i];     // keep old lines for now (append .$i to prevent key collisions)
95              } else {
96                $out_lines[$log['date'].".$i"] = $lines[$i];     // definitely keep these lines
97              }
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($conf['changelog'].'_tmp', implode('', $out_lines));
111            @unlink($conf['changelog']);
112            if (!rename($conf['changelog'].'_tmp', $conf['changelog'])) {
113                // rename failed so try another way...
114                io_unlock($conf['changelog']);
115                io_saveFile($conf['changelog'], implode('', $out_lines));
116                @unlink($conf['changelog'].'_tmp');
117            } else {
118                io_unlock($conf['changelog']);
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, $conf['dperm']);
177
178    require_once(DOKU_INC.'inc/indexer.php');
179
180    // upgrade to version 2
181    if (!@file_exists($conf['indexdir'].'/pageword.idx'))
182        idx_upgradePageWords();
183
184    // do the work
185    idx_addPage($ID);
186
187    // we're finished - save and free lock
188    io_saveFile(metaFN($ID,'.indexed'),INDEXER_VERSION);
189    @rmdir($lock);
190    print "runIndexer(): finished".NL;
191    return true;
192}
193
194/**
195 * Will render the metadata for the page if not exists yet
196 *
197 * This makes sure pages which are created from outside DokuWiki will
198 * gain their data when viewed for the first time.
199 */
200function metaUpdate(){
201    global $ID;
202    print "metaUpdate(): started".NL;
203
204    if(!$ID) return false;
205    $file = metaFN($ID, '.meta');
206    echo "meta file: $file".NL;
207
208    // rendering needed?
209    if (@file_exists($file)) return false;
210    if (!@file_exists(wikiFN($ID))) return false;
211
212    require_once(DOKU_INC.'inc/common.php');
213    require_once(DOKU_INC.'inc/parserutils.php');
214    global $conf;
215
216
217    // gather some additional info from changelog
218    $info = io_grep($conf['changelog'],
219                    '/^(\d+)\t(\d+\.\d+\.\d+\.\d+)\t'.preg_quote($ID,'/').'\t([^\t]+)\t([^\t\n]+)/',
220                    0,true);
221
222    $meta = array();
223    if(!empty($info)){
224        $meta['date']['created'] = $info[0][1];
225        foreach($info as $item){
226            if($item[4] != '*'){
227                $meta['date']['modified'] = $item[1];
228                if($item[3]){
229                    $meta['contributor'][$item[3]] = $item[3];
230                }
231            }
232        }
233    }
234
235    $meta = p_render_metadata($ID, $meta);
236    io_saveFile($file, serialize($meta));
237
238    echo "metaUpdate(): finished".NL;
239    return true;
240}
241
242/**
243 * Builds a Google Sitemap of all public pages known to the indexer
244 *
245 * The map is placed in the root directory named sitemap.xml.gz - This
246 * file needs to be writable!
247 *
248 * @author Andreas Gohr
249 * @link   https://www.google.com/webmasters/sitemaps/docs/en/about.html
250 */
251function runSitemapper(){
252    global $conf;
253    print "runSitemapper(): started".NL;
254    if(!$conf['sitemap']) return false;
255
256    if($conf['compression'] == 'bz2' || $conf['compression'] == 'gz'){
257        $sitemap = 'sitemap.xml.gz';
258    }else{
259        $sitemap = 'sitemap.xml';
260    }
261    print "runSitemapper(): using $sitemap".NL;
262
263    if(@file_exists(DOKU_INC.$sitemap)){
264        if(!is_writable(DOKU_INC.$sitemap)) return false;
265    }else{
266        if(!is_writable(DOKU_INC)) return false;
267    }
268
269    if(@filesize(DOKU_INC.$sitemap) &&
270       @filemtime(DOKU_INC.$sitemap) > (time()-($conf['sitemap']*60*60*24))){
271       print 'runSitemapper(): Sitemap up to date'.NL;
272       return false;
273    }
274
275    $pages = file($conf['indexdir'].'/page.idx');
276    print 'runSitemapper(): creating sitemap using '.count($pages).' pages'.NL;
277
278    // build the sitemap
279    ob_start();
280    print '<?xml version="1.0" encoding="UTF-8"?>'.NL;
281    print '<urlset xmlns="http://www.sitemaps.org/schemas/sitemap/0.9">'.NL;
282    foreach($pages as $id){
283        $id = trim($id);
284        $file = wikiFN($id);
285
286        //skip hidden, non existing and restricted files
287        if(isHiddenPage($id)) continue;
288        $date = @filemtime($file);
289        if(!$date) continue;
290        if(auth_aclcheck($id,'','') < AUTH_READ) continue;
291
292        print '  <url>'.NL;
293        print '    <loc>'.wl($id,'',true).'</loc>'.NL;
294        print '    <lastmod>'.date_iso8601($date).'</lastmod>'.NL;
295        print '  </url>'.NL;
296    }
297    print '</urlset>'.NL;
298    $data = ob_get_contents();
299    ob_end_clean();
300
301    //save the new sitemap
302    io_saveFile(DOKU_INC.$sitemap,$data);
303
304    //ping search engines...
305    $http = new DokuHTTPClient();
306    $http->timeout = 8;
307
308    //ping google
309    print 'runSitemapper(): pinging google'.NL;
310    $url  = 'http://www.google.com/webmasters/sitemaps/ping?sitemap=';
311    $url .= urlencode(DOKU_URL.$sitemap);
312    $resp = $http->get($url);
313    if($http->error) print 'runSitemapper(): '.$http->error.NL;
314    print 'runSitemapper(): '.preg_replace('/[\n\r]/',' ',strip_tags($resp)).NL;
315
316    //ping yahoo
317    print 'runSitemapper(): pinging yahoo'.NL;
318    $url  = 'http://search.yahooapis.com/SiteExplorerService/V1/updateNotification?appid=dokuwiki&url=';
319    $url .= urlencode(DOKU_URL.$sitemap);
320    $resp = $http->get($url);
321    if($http->error) print 'runSitemapper(): '.$http->error.NL;
322    print 'runSitemapper(): '.preg_replace('/[\n\r]/',' ',strip_tags($resp)).NL;
323
324    //ping microsoft
325    print 'runSitemapper(): pinging microsoft'.NL;
326    $url  = 'http://search.live.com/ping?sitemap=';
327    $url .= urlencode(DOKU_URL.$sitemap);
328    $resp = $http->get($url);
329    if($http->error) print 'runSitemapper(): '.$http->error.NL;
330    print 'runSitemapper(): '.preg_replace('/[\n\r]/',' ',strip_tags($resp)).NL;
331
332    print 'runSitemapper(): finished'.NL;
333    return true;
334}
335
336/**
337 * Formats a timestamp as ISO 8601 date
338 *
339 * @author <ungu at terong dot com>
340 * @link http://www.php.net/manual/en/function.date.php#54072
341 */
342function date_iso8601($int_date) {
343   //$int_date: current date in UNIX timestamp
344   $date_mod = date('Y-m-d\TH:i:s', $int_date);
345   $pre_timezone = date('O', $int_date);
346   $time_zone = substr($pre_timezone, 0, 3).":".substr($pre_timezone, 3, 2);
347   $date_mod .= $time_zone;
348   return $date_mod;
349}
350
351/**
352 * Just send a 1x1 pixel blank gif to the browser
353 *
354 * @author Andreas Gohr <andi@splitbrain.org>
355 * @author Harry Fuecks <fuecks@gmail.com>
356 */
357function sendGIF(){
358    if($_REQUEST['debug']){
359        header('Content-Type: text/plain');
360        return;
361    }
362    $img = base64_decode('R0lGODlhAQABAIAAAAAAAP///yH5BAEAAAEALAAAAAABAAEAAAIBTAA7');
363    header('Content-Type: image/gif');
364    header('Content-Length: '.strlen($img));
365    header('Connection: Close');
366    print $img;
367    flush();
368    // Browser should drop connection after this
369    // Thinks it's got the whole image
370}
371
372//Setup VIM: ex: et ts=4 enc=utf-8 :
373// No trailing PHP closing tag - no output please!
374// See Note at http://www.php.net/manual/en/language.basic-syntax.instruction-separation.php
375