xref: /dokuwiki/inc/indexer.php (revision b5f365b32820bb760e8c3729affce219f696d1ba)
1<?php
2/**
3 * Functions to create the fulltext search index
4 *
5 * @license    GPL 2 (http://www.gnu.org/licenses/gpl.html)
6 * @author     Andreas Gohr <andi@splitbrain.org>
7 * @author     Tom N Harris <tnharris@whoopdedo.org>
8 */
9
10if(!defined('DOKU_INC')) die('meh.');
11
12// Version tag used to force rebuild on upgrade
13define('INDEXER_VERSION', 8);
14
15// set the minimum token length to use in the index (note, this doesn't apply to numeric tokens)
16if (!defined('IDX_MINWORDLENGTH')) define('IDX_MINWORDLENGTH',2);
17
18// Asian characters are handled as words. The following regexp defines the
19// Unicode-Ranges for Asian characters
20// Ranges taken from http://en.wikipedia.org/wiki/Unicode_block
21// I'm no language expert. If you think some ranges are wrongly chosen or
22// a range is missing, please contact me
23define('IDX_ASIAN1','[\x{0E00}-\x{0E7F}]'); // Thai
24define('IDX_ASIAN2','['.
25                   '\x{2E80}-\x{3040}'.  // CJK -> Hangul
26                   '\x{309D}-\x{30A0}'.
27                   '\x{30FD}-\x{31EF}\x{3200}-\x{D7AF}'.
28                   '\x{F900}-\x{FAFF}'.  // CJK Compatibility Ideographs
29                   '\x{FE30}-\x{FE4F}'.  // CJK Compatibility Forms
30                   "\xF0\xA0\x80\x80-\xF0\xAA\x9B\x9F". // CJK Extension B
31                   "\xF0\xAA\x9C\x80-\xF0\xAB\x9C\xBF". // CJK Extension C
32                   "\xF0\xAB\x9D\x80-\xF0\xAB\xA0\x9F". // CJK Extension D
33                   "\xF0\xAF\xA0\x80-\xF0\xAF\xAB\xBF". // CJK Compatibility Supplement
34                   ']');
35define('IDX_ASIAN3','['.                // Hiragana/Katakana (can be two characters)
36                   '\x{3042}\x{3044}\x{3046}\x{3048}'.
37                   '\x{304A}-\x{3062}\x{3064}-\x{3082}'.
38                   '\x{3084}\x{3086}\x{3088}-\x{308D}'.
39                   '\x{308F}-\x{3094}'.
40                   '\x{30A2}\x{30A4}\x{30A6}\x{30A8}'.
41                   '\x{30AA}-\x{30C2}\x{30C4}-\x{30E2}'.
42                   '\x{30E4}\x{30E6}\x{30E8}-\x{30ED}'.
43                   '\x{30EF}-\x{30F4}\x{30F7}-\x{30FA}'.
44                   ']['.
45                   '\x{3041}\x{3043}\x{3045}\x{3047}\x{3049}'.
46                   '\x{3063}\x{3083}\x{3085}\x{3087}\x{308E}\x{3095}-\x{309C}'.
47                   '\x{30A1}\x{30A3}\x{30A5}\x{30A7}\x{30A9}'.
48                   '\x{30C3}\x{30E3}\x{30E5}\x{30E7}\x{30EE}\x{30F5}\x{30F6}\x{30FB}\x{30FC}'.
49                   '\x{31F0}-\x{31FF}'.
50                   ']?');
51define('IDX_ASIAN', '(?:'.IDX_ASIAN1.'|'.IDX_ASIAN2.'|'.IDX_ASIAN3.')');
52
53/**
54 * Version of the indexer taking into consideration the external tokenizer.
55 * The indexer is only compatible with data written by the same version.
56 *
57 * @triggers INDEXER_VERSION_GET
58 * Plugins that modify what gets indexed should hook this event and
59 * add their version info to the event data like so:
60 *     $data[$plugin_name] = $plugin_version;
61 *
62 * @author Tom N Harris <tnharris@whoopdedo.org>
63 * @author Michael Hamann <michael@content-space.de>
64 */
65function idx_get_version(){
66    static $indexer_version = null;
67    if ($indexer_version == null) {
68        $version = INDEXER_VERSION;
69
70        // DokuWiki version is included for the convenience of plugins
71        $data = array('dokuwiki'=>$version);
72        trigger_event('INDEXER_VERSION_GET', $data, null, false);
73        unset($data['dokuwiki']); // this needs to be first
74        ksort($data);
75        foreach ($data as $plugin=>$vers)
76            $version .= '+'.$plugin.'='.$vers;
77        $indexer_version = $version;
78    }
79    return $indexer_version;
80}
81
82/**
83 * Measure the length of a string.
84 * Differs from strlen in handling of asian characters.
85 *
86 * @author Tom N Harris <tnharris@whoopdedo.org>
87 */
88function wordlen($w){
89    $l = strlen($w);
90    // If left alone, all chinese "words" will get put into w3.idx
91    // So the "length" of a "word" is faked
92    if(preg_match_all('/[\xE2-\xEF]/',$w,$leadbytes)) {
93        foreach($leadbytes[0] as $b)
94            $l += ord($b) - 0xE1;
95    }
96    return $l;
97}
98
99/**
100 * Class that encapsulates operations on the indexer database.
101 *
102 * @author Tom N Harris <tnharris@whoopdedo.org>
103 */
104class Doku_Indexer {
105    /**
106     * @var array $pidCache Cache for getPID()
107     */
108    protected $pidCache = array();
109
110    /**
111     * Adds the contents of a page to the fulltext index
112     *
113     * The added text replaces previous words for the same page.
114     * An empty value erases the page.
115     *
116     * @param string    $page   a page name
117     * @param string    $text   the body of the page
118     * @return boolean          the function completed successfully
119     * @author Tom N Harris <tnharris@whoopdedo.org>
120     * @author Andreas Gohr <andi@splitbrain.org>
121     */
122    public function addPageWords($page, $text) {
123        if (!$this->lock())
124            return "locked";
125
126        // load known documents
127        $pid = $this->getPIDNoLock($page);
128        if ($pid === false) {
129            $this->unlock();
130            return false;
131        }
132
133        $pagewords = array();
134        // get word usage in page
135        $words = $this->getPageWords($text);
136        if ($words === false) {
137            $this->unlock();
138            return false;
139        }
140
141        if (!empty($words)) {
142            foreach (array_keys($words) as $wlen) {
143                $index = $this->getIndex('i', $wlen);
144                foreach ($words[$wlen] as $wid => $freq) {
145                    $idx = ($wid<count($index)) ? $index[$wid] : '';
146                    $index[$wid] = $this->updateTuple($idx, $pid, $freq);
147                    $pagewords[] = "$wlen*$wid";
148                }
149                if (!$this->saveIndex('i', $wlen, $index)) {
150                    $this->unlock();
151                    return false;
152                }
153            }
154        }
155
156        // Remove obsolete index entries
157        $pageword_idx = $this->getIndexKey('pageword', '', $pid);
158        if ($pageword_idx !== '') {
159            $oldwords = explode(':',$pageword_idx);
160            $delwords = array_diff($oldwords, $pagewords);
161            $upwords = array();
162            foreach ($delwords as $word) {
163                if ($word != '') {
164                    list($wlen,$wid) = explode('*', $word);
165                    $wid = (int)$wid;
166                    $upwords[$wlen][] = $wid;
167                }
168            }
169            foreach ($upwords as $wlen => $widx) {
170                $index = $this->getIndex('i', $wlen);
171                foreach ($widx as $wid) {
172                    $index[$wid] = $this->updateTuple($index[$wid], $pid, 0);
173                }
174                $this->saveIndex('i', $wlen, $index);
175            }
176        }
177        // Save the reverse index
178        $pageword_idx = join(':', $pagewords);
179        if (!$this->saveIndexKey('pageword', '', $pid, $pageword_idx)) {
180            $this->unlock();
181            return false;
182        }
183
184        $this->unlock();
185        return true;
186    }
187
188    /**
189     * Split the words in a page and add them to the index.
190     *
191     * @param string    $text   content of the page
192     * @return array            list of word IDs and number of times used
193     * @author Andreas Gohr <andi@splitbrain.org>
194     * @author Christopher Smith <chris@jalakai.co.uk>
195     * @author Tom N Harris <tnharris@whoopdedo.org>
196     */
197    protected function getPageWords($text) {
198
199        $tokens = $this->tokenizer($text);
200        $tokens = array_count_values($tokens);  // count the frequency of each token
201
202        $words = array();
203        foreach ($tokens as $w=>$c) {
204            $l = wordlen($w);
205            if (isset($words[$l])){
206                $words[$l][$w] = $c + (isset($words[$l][$w]) ? $words[$l][$w] : 0);
207            }else{
208                $words[$l] = array($w => $c);
209            }
210        }
211
212        // arrive here with $words = array(wordlen => array(word => frequency))
213        $word_idx_modified = false;
214        $index = array();   //resulting index
215        foreach (array_keys($words) as $wlen) {
216            $word_idx = $this->getIndex('w', $wlen);
217            foreach ($words[$wlen] as $word => $freq) {
218                $word = (string)$word;
219                $wid = array_search($word, $word_idx, true);
220                if ($wid === false) {
221                    $wid = count($word_idx);
222                    $word_idx[] = $word;
223                    $word_idx_modified = true;
224                }
225                if (!isset($index[$wlen]))
226                    $index[$wlen] = array();
227                $index[$wlen][$wid] = $freq;
228            }
229            // save back the word index
230            if ($word_idx_modified && !$this->saveIndex('w', $wlen, $word_idx))
231                return false;
232        }
233
234        return $index;
235    }
236
237    /**
238     * Add/update keys to/of the metadata index.
239     *
240     * Adding new keys does not remove other keys for the page.
241     * An empty value will erase the key.
242     * The $key parameter can be an array to add multiple keys. $value will
243     * not be used if $key is an array.
244     *
245     * @param string    $page   a page name
246     * @param mixed     $key    a key string or array of key=>value pairs
247     * @param mixed     $value  the value or list of values
248     * @return boolean          the function completed successfully
249     * @author Tom N Harris <tnharris@whoopdedo.org>
250     * @author Michael Hamann <michael@content-space.de>
251     */
252    public function addMetaKeys($page, $key, $value=null) {
253        if (!is_array($key)) {
254            $key = array($key => $value);
255        } elseif (!is_null($value)) {
256            // $key is array, but $value is not null
257            trigger_error("array passed to addMetaKeys but value is not null", E_USER_WARNING);
258        }
259
260        if (!$this->lock())
261            return "locked";
262
263        // load known documents
264        $pid = $this->getPIDNoLock($page);
265        if ($pid === false) {
266            $this->unlock();
267            return false;
268        }
269
270        // Special handling for titles so the index file is simpler
271        if (array_key_exists('title', $key)) {
272            $value = $key['title'];
273            if (is_array($value))
274                $value = $value[0];
275            $this->saveIndexKey('title', '', $pid, $value);
276            unset($key['title']);
277        }
278
279        foreach ($key as $name => $values) {
280            $metaname = idx_cleanName($name);
281            $this->addIndexKey('metadata', '', $metaname);
282            $metaidx = $this->getIndex($metaname.'_i', '');
283            $metawords = $this->getIndex($metaname.'_w', '');
284            $addwords = false;
285
286            if (!is_array($values)) $values = array($values);
287
288            $val_idx = $this->getIndexKey($metaname.'_p', '', $pid);
289            if ($val_idx != '') {
290                $val_idx = explode(':', $val_idx);
291                // -1 means remove, 0 keep, 1 add
292                $val_idx = array_combine($val_idx, array_fill(0, count($val_idx), -1));
293            } else {
294                $val_idx = array();
295            }
296
297            foreach ($values as $val) {
298                $val = (string)$val;
299                if ($val !== "") {
300                    $id = array_search($val, $metawords, true);
301                    if ($id === false) {
302                        $id = count($metawords);
303                        $metawords[$id] = $val;
304                        $addwords = true;
305                    }
306                    // test if value is already in the index
307                    if (isset($val_idx[$id]) && $val_idx[$id] <= 0)
308                        $val_idx[$id] = 0;
309                    else // else add it
310                        $val_idx[$id] = 1;
311                }
312            }
313
314            if ($addwords)
315                $this->saveIndex($metaname.'_w', '', $metawords);
316            $vals_changed = false;
317            foreach ($val_idx as $id => $action) {
318                if ($action == -1) {
319                    $metaidx[$id] = $this->updateTuple($metaidx[$id], $pid, 0);
320                    $vals_changed = true;
321                    unset($val_idx[$id]);
322                } elseif ($action == 1) {
323                    $metaidx[$id] = $this->updateTuple($metaidx[$id], $pid, 1);
324                    $vals_changed = true;
325                }
326            }
327
328            if ($vals_changed) {
329                $this->saveIndex($metaname.'_i', '', $metaidx);
330                $val_idx = implode(':', array_keys($val_idx));
331                $this->saveIndexKey($metaname.'_p', '', $pid, $val_idx);
332            }
333
334            unset($metaidx);
335            unset($metawords);
336        }
337
338        $this->unlock();
339        return true;
340    }
341
342    /**
343     * Rename a page in the search index without changing the indexed content. This function doesn't check if the
344     * old or new name exists in the filesystem. It returns an error if the old page isn't in the page list of the
345     * indexer and it deletes all previously indexed content of the new page.
346     *
347     * @param string $oldpage The old page name
348     * @param string $newpage The new page name
349     * @return string|bool If the page was successfully renamed, can be a message in the case of an error
350     */
351    public function renamePage($oldpage, $newpage) {
352        if (!$this->lock()) return 'locked';
353
354        $pages = $this->getPages();
355
356        $id = array_search($oldpage, $pages, true);
357        if ($id === false) {
358            $this->unlock();
359            return 'page is not in index';
360        }
361
362        $new_id = array_search($newpage, $pages, true);
363        if ($new_id !== false) {
364            // make sure the page is not in the index anymore
365            if ($this->deletePageNoLock($newpage) !== true) {
366                return false;
367            }
368
369            $pages[$new_id] = 'deleted:'.time().rand(0, 9999);
370        }
371
372        $pages[$id] = $newpage;
373
374        // update index
375        if (!$this->saveIndex('page', '', $pages)) {
376            $this->unlock();
377            return false;
378        }
379
380        // reset the pid cache
381        $this->pidCache = array();
382
383        $this->unlock();
384        return true;
385    }
386
387    /**
388     * Renames a meta value in the index. This doesn't change the meta value in the pages, it assumes that all pages
389     * will be updated.
390     *
391     * @param string $key       The metadata key of which a value shall be changed
392     * @param string $oldvalue  The old value that shall be renamed
393     * @param string $newvalue  The new value to which the old value shall be renamed, can exist (then values will be merged)
394     * @return bool|string      If renaming the value has been successful, false or error message on error.
395     */
396    public function renameMetaValue($key, $oldvalue, $newvalue) {
397        if (!$this->lock()) return 'locked';
398
399        // change the relation references index
400        $metavalues = $this->getIndex($key, '_w');
401        $oldid = array_search($oldvalue, $metavalues, true);
402        if ($oldid !== false) {
403            $newid = array_search($newvalue, $metavalues, true);
404            if ($newid !== false) {
405                // free memory
406                unset ($metavalues);
407
408                // okay, now we have two entries for the same value. we need to merge them.
409                $indexline = $this->getIndexKey($key.'_i', '', $oldid);
410                if ($indexline != '') {
411                    $newindexline = $this->getIndexKey($key.'_i', '', $newid);
412                    $pagekeys     = $this->getIndex($key.'_p', '');
413                    $parts = explode(':', $indexline);
414                    foreach ($parts as $part) {
415                        list($id, $count) = explode('*', $part);
416                        $newindexline =  $this->updateTuple($newindexline, $id, $count);
417
418                        $keyline = explode(':', $pagekeys[$id]);
419                        // remove old meta value
420                        $keyline = array_diff($keyline, array($oldid));
421                        // add new meta value when not already present
422                        if (!in_array($newid, $keyline)) {
423                            array_push($keyline, $newid);
424                        }
425                        $pagekeys[$id] = implode(':', $keyline);
426                    }
427                    $this->saveIndex($key.'_p', '', $pagekeys);
428                    unset($pagekeys);
429                    $this->saveIndexKey($key.'_i', '', $oldid, '');
430                    $this->saveIndexKey($key.'_i', '', $newid, $newindexline);
431                }
432            } else {
433                $metavalues[$oldid] = $newvalue;
434                if (!$this->saveIndex($key.'_w', '', $metavalues)) {
435                    $this->unlock();
436                    return false;
437                }
438            }
439        }
440
441        $this->unlock();
442        return true;
443    }
444
445    /**
446     * Remove a page from the index
447     *
448     * Erases entries in all known indexes.
449     *
450     * @param string    $page   a page name
451     * @return boolean          the function completed successfully
452     * @author Tom N Harris <tnharris@whoopdedo.org>
453     */
454    public function deletePage($page) {
455        if (!$this->lock())
456            return "locked";
457
458        $result = $this->deletePageNoLock($page);
459
460        $this->unlock();
461
462        return $result;
463    }
464
465    /**
466     * Remove a page from the index without locking the index, only use this function if the index is already locked
467     *
468     * Erases entries in all known indexes.
469     *
470     * @param string    $page   a page name
471     * @return boolean          the function completed successfully
472     * @author Tom N Harris <tnharris@whoopdedo.org>
473     */
474    protected function deletePageNoLock($page) {
475        // load known documents
476        $pid = $this->getPIDNoLock($page);
477        if ($pid === false) {
478            return false;
479        }
480
481        // Remove obsolete index entries
482        $pageword_idx = $this->getIndexKey('pageword', '', $pid);
483        if ($pageword_idx !== '') {
484            $delwords = explode(':',$pageword_idx);
485            $upwords = array();
486            foreach ($delwords as $word) {
487                if ($word != '') {
488                    list($wlen,$wid) = explode('*', $word);
489                    $wid = (int)$wid;
490                    $upwords[$wlen][] = $wid;
491                }
492            }
493            foreach ($upwords as $wlen => $widx) {
494                $index = $this->getIndex('i', $wlen);
495                foreach ($widx as $wid) {
496                    $index[$wid] = $this->updateTuple($index[$wid], $pid, 0);
497                }
498                $this->saveIndex('i', $wlen, $index);
499            }
500        }
501        // Save the reverse index
502        if (!$this->saveIndexKey('pageword', '', $pid, "")) {
503            return false;
504        }
505
506        $this->saveIndexKey('title', '', $pid, "");
507        $keyidx = $this->getIndex('metadata', '');
508        foreach ($keyidx as $metaname) {
509            $val_idx = explode(':', $this->getIndexKey($metaname.'_p', '', $pid));
510            $meta_idx = $this->getIndex($metaname.'_i', '');
511            foreach ($val_idx as $id) {
512                if ($id === '') continue;
513                $meta_idx[$id] = $this->updateTuple($meta_idx[$id], $pid, 0);
514            }
515            $this->saveIndex($metaname.'_i', '', $meta_idx);
516            $this->saveIndexKey($metaname.'_p', '', $pid, '');
517        }
518
519        return true;
520    }
521
522    /**
523     * Clear the whole index
524     *
525     * @return bool If the index has been cleared successfully
526     */
527    public function clear() {
528        global $conf;
529
530        if (!$this->lock()) return false;
531
532        @unlink($conf['indexdir'].'/page.idx');
533        @unlink($conf['indexdir'].'/title.idx');
534        @unlink($conf['indexdir'].'/pageword.idx');
535        @unlink($conf['indexdir'].'/metadata.idx');
536        $dir = @opendir($conf['indexdir']);
537        if($dir!==false){
538            while(($f = readdir($dir)) !== false){
539                if(substr($f,-4)=='.idx' &&
540                    (substr($f,0,1)=='i' || substr($f,0,1)=='w'
541                        || substr($f,-6)=='_w.idx' || substr($f,-6)=='_i.idx' || substr($f,-6)=='_p.idx'))
542                    @unlink($conf['indexdir']."/$f");
543            }
544        }
545        @unlink($conf['indexdir'].'/lengths.idx');
546
547        // clear the pid cache
548        $this->pidCache = array();
549
550        $this->unlock();
551        return true;
552    }
553
554    /**
555     * Split the text into words for fulltext search
556     *
557     * TODO: does this also need &$stopwords ?
558     *
559     * @triggers INDEXER_TEXT_PREPARE
560     * This event allows plugins to modify the text before it gets tokenized.
561     * Plugins intercepting this event should also intercept INDEX_VERSION_GET
562     *
563     * @param string    $text   plain text
564     * @param boolean   $wc     are wildcards allowed?
565     * @return array            list of words in the text
566     * @author Tom N Harris <tnharris@whoopdedo.org>
567     * @author Andreas Gohr <andi@splitbrain.org>
568     */
569    public function tokenizer($text, $wc=false) {
570        $wc = ($wc) ? '' : '\*';
571        $stopwords =& idx_get_stopwords();
572
573        // prepare the text to be tokenized
574        $evt = new Doku_Event('INDEXER_TEXT_PREPARE', $text);
575        if ($evt->advise_before(true)) {
576            if (preg_match('/[^0-9A-Za-z ]/u', $text)) {
577                // handle asian chars as single words (may fail on older PHP version)
578                $asia = @preg_replace('/('.IDX_ASIAN.')/u', ' \1 ', $text);
579                if (!is_null($asia)) $text = $asia; // recover from regexp falure
580            }
581        }
582        $evt->advise_after();
583        unset($evt);
584
585        $text = strtr($text,
586                       array(
587                           "\r" => ' ',
588                           "\n" => ' ',
589                           "\t" => ' ',
590                           "\xC2\xAD" => '', //soft-hyphen
591                       )
592                     );
593        if (preg_match('/[^0-9A-Za-z ]/u', $text))
594            $text = utf8_stripspecials($text, ' ', '\._\-:'.$wc);
595
596        $wordlist = explode(' ', $text);
597        foreach ($wordlist as $i => $word) {
598            $wordlist[$i] = (preg_match('/[^0-9A-Za-z]/u', $word)) ?
599                utf8_strtolower($word) : strtolower($word);
600        }
601
602        foreach ($wordlist as $i => $word) {
603            if ((!is_numeric($word) && strlen($word) < IDX_MINWORDLENGTH)
604              || array_search($word, $stopwords, true) !== false)
605                unset($wordlist[$i]);
606        }
607        return array_values($wordlist);
608    }
609
610    /**
611     * Get the numeric PID of a page
612     *
613     * @param string $page The page to get the PID for
614     * @return bool|int The page id on success, false on error
615     */
616    public function getPID($page) {
617        // return PID without locking when it is in the cache
618        if (isset($this->pidCache[$page])) return $this->pidCache[$page];
619
620        if (!$this->lock())
621            return false;
622
623        // load known documents
624        $pid = $this->getPIDNoLock($page);
625        if ($pid === false) {
626            $this->unlock();
627            return false;
628        }
629
630        $this->unlock();
631        return $pid;
632    }
633
634    /**
635     * Get the numeric PID of a page without locking the index.
636     * Only use this function when the index is already locked.
637     *
638     * @param string $page The page to get the PID for
639     * @return bool|int The page id on success, false on error
640     */
641    protected function getPIDNoLock($page) {
642        // avoid expensive addIndexKey operation for the most recently requested pages by using a cache
643        if (isset($this->pidCache[$page])) return $this->pidCache[$page];
644        $pid = $this->addIndexKey('page', '', $page);
645        // limit cache to 10 entries by discarding the oldest element as in DokuWiki usually only the most recently
646        // added item will be requested again
647        if (count($this->pidCache) > 10) array_shift($this->pidCache);
648        $this->pidCache[$page] = $pid;
649        return $pid;
650    }
651
652    /**
653     * Get the page id of a numeric PID
654     *
655     * @param int $pid The PID to get the page id for
656     * @return string The page id
657     */
658    public function getPageFromPID($pid) {
659        return $this->getIndexKey('page', '', $pid);
660    }
661
662    /**
663     * Find pages in the fulltext index containing the words,
664     *
665     * The search words must be pre-tokenized, meaning only letters and
666     * numbers with an optional wildcard
667     *
668     * The returned array will have the original tokens as key. The values
669     * in the returned list is an array with the page names as keys and the
670     * number of times that token appears on the page as value.
671     *
672     * @param array  $tokens list of words to search for
673     * @return array         list of page names with usage counts
674     * @author Tom N Harris <tnharris@whoopdedo.org>
675     * @author Andreas Gohr <andi@splitbrain.org>
676     */
677    public function lookup(&$tokens) {
678        $result = array();
679        $wids = $this->getIndexWords($tokens, $result);
680        if (empty($wids)) return array();
681        // load known words and documents
682        $page_idx = $this->getIndex('page', '');
683        $docs = array();
684        foreach (array_keys($wids) as $wlen) {
685            $wids[$wlen] = array_unique($wids[$wlen]);
686            $index = $this->getIndex('i', $wlen);
687            foreach($wids[$wlen] as $ixid) {
688                if ($ixid < count($index))
689                    $docs["$wlen*$ixid"] = $this->parseTuples($page_idx, $index[$ixid]);
690            }
691        }
692        // merge found pages into final result array
693        $final = array();
694        foreach ($result as $word => $res) {
695            $final[$word] = array();
696            foreach ($res as $wid) {
697                // handle the case when ($ixid < count($index)) has been false
698                // and thus $docs[$wid] hasn't been set.
699                if (!isset($docs[$wid])) continue;
700                $hits = &$docs[$wid];
701                foreach ($hits as $hitkey => $hitcnt) {
702                    // make sure the document still exists
703                    if (!page_exists($hitkey, '', false)) continue;
704                    if (!isset($final[$word][$hitkey]))
705                        $final[$word][$hitkey] = $hitcnt;
706                    else
707                        $final[$word][$hitkey] += $hitcnt;
708                }
709            }
710        }
711        return $final;
712    }
713
714    /**
715     * Find pages containing a metadata key.
716     *
717     * The metadata values are compared as case-sensitive strings. Pass a
718     * callback function that returns true or false to use a different
719     * comparison function. The function will be called with the $value being
720     * searched for as the first argument, and the word in the index as the
721     * second argument. The function preg_match can be used directly if the
722     * values are regexes.
723     *
724     * @param string    $key    name of the metadata key to look for
725     * @param string    $value  search term to look for, must be a string or array of strings
726     * @param callback  $func   comparison function
727     * @return array            lists with page names, keys are query values if $value is array
728     * @author Tom N Harris <tnharris@whoopdedo.org>
729     * @author Michael Hamann <michael@content-space.de>
730     */
731    public function lookupKey($key, &$value, $func=null) {
732        if (!is_array($value))
733            $value_array = array($value);
734        else
735            $value_array =& $value;
736
737        // the matching ids for the provided value(s)
738        $value_ids = array();
739
740        $metaname = idx_cleanName($key);
741
742        // get all words in order to search the matching ids
743        if ($key == 'title') {
744            $words = $this->getIndex('title', '');
745        } else {
746            $words = $this->getIndex($metaname.'_w', '');
747        }
748
749        if (!is_null($func)) {
750            foreach ($value_array as $val) {
751                foreach ($words as $i => $word) {
752                    if (call_user_func_array($func, array($val, $word)))
753                        $value_ids[$i][] = $val;
754                }
755            }
756        } else {
757            foreach ($value_array as $val) {
758                $xval = $val;
759                $caret = '^';
760                $dollar = '$';
761                // check for wildcards
762                if (substr($xval, 0, 1) == '*') {
763                    $xval = substr($xval, 1);
764                    $caret = '';
765                }
766                if (substr($xval, -1, 1) == '*') {
767                    $xval = substr($xval, 0, -1);
768                    $dollar = '';
769                }
770                if (!$caret || !$dollar) {
771                    $re = $caret.preg_quote($xval, '/').$dollar;
772                    foreach(array_keys(preg_grep('/'.$re.'/', $words)) as $i)
773                        $value_ids[$i][] = $val;
774                } else {
775                    if (($i = array_search($val, $words, true)) !== false)
776                        $value_ids[$i][] = $val;
777                }
778            }
779        }
780
781        unset($words); // free the used memory
782
783        // initialize the result so it won't be null
784        $result = array();
785        foreach ($value_array as $val) {
786            $result[$val] = array();
787        }
788
789        $page_idx = $this->getIndex('page', '');
790
791        // Special handling for titles
792        if ($key == 'title') {
793            foreach ($value_ids as $pid => $val_list) {
794                $page = $page_idx[$pid];
795                foreach ($val_list as $val) {
796                    $result[$val][] = $page;
797                }
798            }
799        } else {
800            // load all lines and pages so the used lines can be taken and matched with the pages
801            $lines = $this->getIndex($metaname.'_i', '');
802
803            foreach ($value_ids as $value_id => $val_list) {
804                // parse the tuples of the form page_id*1:page2_id*1 and so on, return value
805                // is an array with page_id => 1, page2_id => 1 etc. so take the keys only
806                $pages = array_keys($this->parseTuples($page_idx, $lines[$value_id]));
807                foreach ($val_list as $val) {
808                    $result[$val] = array_merge($result[$val], $pages);
809                }
810            }
811        }
812        if (!is_array($value)) $result = $result[$value];
813        return $result;
814    }
815
816    /**
817     * Find the index ID of each search term.
818     *
819     * The query terms should only contain valid characters, with a '*' at
820     * either the beginning or end of the word (or both).
821     * The $result parameter can be used to merge the index locations with
822     * the appropriate query term.
823     *
824     * @param array  $words  The query terms.
825     * @param array  $result Set to word => array("length*id" ...)
826     * @return array         Set to length => array(id ...)
827     * @author Tom N Harris <tnharris@whoopdedo.org>
828     */
829    protected function getIndexWords(&$words, &$result) {
830        $tokens = array();
831        $tokenlength = array();
832        $tokenwild = array();
833        foreach ($words as $word) {
834            $result[$word] = array();
835            $caret = '^';
836            $dollar = '$';
837            $xword = $word;
838            $wlen = wordlen($word);
839
840            // check for wildcards
841            if (substr($xword, 0, 1) == '*') {
842                $xword = substr($xword, 1);
843                $caret = '';
844                $wlen -= 1;
845            }
846            if (substr($xword, -1, 1) == '*') {
847                $xword = substr($xword, 0, -1);
848                $dollar = '';
849                $wlen -= 1;
850            }
851            if ($wlen < IDX_MINWORDLENGTH && $caret && $dollar && !is_numeric($xword))
852                continue;
853            if (!isset($tokens[$xword]))
854                $tokenlength[$wlen][] = $xword;
855            if (!$caret || !$dollar) {
856                $re = $caret.preg_quote($xword, '/').$dollar;
857                $tokens[$xword][] = array($word, '/'.$re.'/');
858                if (!isset($tokenwild[$xword]))
859                    $tokenwild[$xword] = $wlen;
860            } else {
861                $tokens[$xword][] = array($word, null);
862            }
863        }
864        asort($tokenwild);
865        // $tokens = array( base word => array( [ query term , regexp ] ... ) ... )
866        // $tokenlength = array( base word length => base word ... )
867        // $tokenwild = array( base word => base word length ... )
868        $length_filter = empty($tokenwild) ? $tokenlength : min(array_keys($tokenlength));
869        $indexes_known = $this->indexLengths($length_filter);
870        if (!empty($tokenwild)) sort($indexes_known);
871        // get word IDs
872        $wids = array();
873        foreach ($indexes_known as $ixlen) {
874            $word_idx = $this->getIndex('w', $ixlen);
875            // handle exact search
876            if (isset($tokenlength[$ixlen])) {
877                foreach ($tokenlength[$ixlen] as $xword) {
878                    $wid = array_search($xword, $word_idx, true);
879                    if ($wid !== false) {
880                        $wids[$ixlen][] = $wid;
881                        foreach ($tokens[$xword] as $w)
882                            $result[$w[0]][] = "$ixlen*$wid";
883                    }
884                }
885            }
886            // handle wildcard search
887            foreach ($tokenwild as $xword => $wlen) {
888                if ($wlen >= $ixlen) break;
889                foreach ($tokens[$xword] as $w) {
890                    if (is_null($w[1])) continue;
891                    foreach(array_keys(preg_grep($w[1], $word_idx)) as $wid) {
892                        $wids[$ixlen][] = $wid;
893                        $result[$w[0]][] = "$ixlen*$wid";
894                    }
895                }
896            }
897        }
898        return $wids;
899    }
900
901    /**
902     * Return a list of all pages
903     * Warning: pages may not exist!
904     *
905     * @param string    $key    list only pages containing the metadata key (optional)
906     * @return array            list of page names
907     * @author Tom N Harris <tnharris@whoopdedo.org>
908     */
909    public function getPages($key=null) {
910        $page_idx = $this->getIndex('page', '');
911        if (is_null($key)) return $page_idx;
912
913        $metaname = idx_cleanName($key);
914
915        // Special handling for titles
916        if ($key == 'title') {
917            $title_idx = $this->getIndex('title', '');
918            array_splice($page_idx, count($title_idx));
919            foreach ($title_idx as $i => $title)
920                if ($title === "") unset($page_idx[$i]);
921            return array_values($page_idx);
922        }
923
924        $pages = array();
925        $lines = $this->getIndex($metaname.'_i', '');
926        foreach ($lines as $line) {
927            $pages = array_merge($pages, $this->parseTuples($page_idx, $line));
928        }
929        return array_keys($pages);
930    }
931
932    /**
933     * Return a list of words sorted by number of times used
934     *
935     * @param int       $min    bottom frequency threshold
936     * @param int       $max    upper frequency limit. No limit if $max<$min
937     * @param int       $minlen minimum length of words to count
938     * @param string    $key    metadata key to list. Uses the fulltext index if not given
939     * @return array            list of words as the keys and frequency as values
940     * @author Tom N Harris <tnharris@whoopdedo.org>
941     */
942    public function histogram($min=1, $max=0, $minlen=3, $key=null) {
943        if ($min < 1)
944            $min = 1;
945        if ($max < $min)
946            $max = 0;
947
948        $result = array();
949
950        if ($key == 'title') {
951            $index = $this->getIndex('title', '');
952            $index = array_count_values($index);
953            foreach ($index as $val => $cnt) {
954                if ($cnt >= $min && (!$max || $cnt <= $max) && strlen($val) >= $minlen)
955                    $result[$val] = $cnt;
956            }
957        }
958        elseif (!is_null($key)) {
959            $metaname = idx_cleanName($key);
960            $index = $this->getIndex($metaname.'_i', '');
961            $val_idx = array();
962            foreach ($index as $wid => $line) {
963                $freq = $this->countTuples($line);
964                if ($freq >= $min && (!$max || $freq <= $max))
965                    $val_idx[$wid] = $freq;
966            }
967            if (!empty($val_idx)) {
968                $words = $this->getIndex($metaname.'_w', '');
969                foreach ($val_idx as $wid => $freq) {
970                    if (strlen($words[$wid]) >= $minlen)
971                        $result[$words[$wid]] = $freq;
972                }
973            }
974        }
975        else {
976            $lengths = idx_listIndexLengths();
977            foreach ($lengths as $length) {
978                if ($length < $minlen) continue;
979                $index = $this->getIndex('i', $length);
980                $words = null;
981                foreach ($index as $wid => $line) {
982                    $freq = $this->countTuples($line);
983                    if ($freq >= $min && (!$max || $freq <= $max)) {
984                        if ($words === null)
985                            $words = $this->getIndex('w', $length);
986                        $result[$words[$wid]] = $freq;
987                    }
988                }
989            }
990        }
991
992        arsort($result);
993        return $result;
994    }
995
996    /**
997     * Lock the indexer.
998     *
999     * @author Tom N Harris <tnharris@whoopdedo.org>
1000     */
1001    protected function lock() {
1002        global $conf;
1003        $status = true;
1004        $run = 0;
1005        $lock = $conf['lockdir'].'/_indexer.lock';
1006        while (!@mkdir($lock, $conf['dmode'])) {
1007            usleep(50);
1008            if(is_dir($lock) && time()-@filemtime($lock) > 60*5){
1009                // looks like a stale lock - remove it
1010                if (!@rmdir($lock)) {
1011                    $status = "removing the stale lock failed";
1012                    return false;
1013                } else {
1014                    $status = "stale lock removed";
1015                }
1016            }elseif($run++ == 1000){
1017                // we waited 5 seconds for that lock
1018                return false;
1019            }
1020        }
1021        if (!empty($conf['dperm'])) {
1022            chmod($lock, $conf['dperm']);
1023        }
1024        return $status;
1025    }
1026
1027    /**
1028     * Release the indexer lock.
1029     *
1030     * @author Tom N Harris <tnharris@whoopdedo.org>
1031     */
1032    protected function unlock() {
1033        global $conf;
1034        @rmdir($conf['lockdir'].'/_indexer.lock');
1035        return true;
1036    }
1037
1038    /**
1039     * Retrieve the entire index.
1040     *
1041     * The $suffix argument is for an index that is split into
1042     * multiple parts. Different index files should use different
1043     * base names.
1044     *
1045     * @param string    $idx    name of the index
1046     * @param string    $suffix subpart identifier
1047     * @return array            list of lines without CR or LF
1048     * @author Tom N Harris <tnharris@whoopdedo.org>
1049     */
1050    protected function getIndex($idx, $suffix) {
1051        global $conf;
1052        $fn = $conf['indexdir'].'/'.$idx.$suffix.'.idx';
1053        if (!@file_exists($fn)) return array();
1054        return file($fn, FILE_IGNORE_NEW_LINES);
1055    }
1056
1057    /**
1058     * Replace the contents of the index with an array.
1059     *
1060     * @param string    $idx    name of the index
1061     * @param string    $suffix subpart identifier
1062     * @param array     $lines  list of lines without LF
1063     * @return bool             If saving succeeded
1064     * @author Tom N Harris <tnharris@whoopdedo.org>
1065     */
1066    protected function saveIndex($idx, $suffix, &$lines) {
1067        global $conf;
1068        $fn = $conf['indexdir'].'/'.$idx.$suffix;
1069        $fh = @fopen($fn.'.tmp', 'w');
1070        if (!$fh) return false;
1071        fwrite($fh, join("\n", $lines));
1072        if (!empty($lines))
1073            fwrite($fh, "\n");
1074        fclose($fh);
1075        if (isset($conf['fperm']))
1076            chmod($fn.'.tmp', $conf['fperm']);
1077        io_rename($fn.'.tmp', $fn.'.idx');
1078        return true;
1079    }
1080
1081    /**
1082     * Retrieve a line from the index.
1083     *
1084     * @param string    $idx    name of the index
1085     * @param string    $suffix subpart identifier
1086     * @param int       $id     the line number
1087     * @return string           a line with trailing whitespace removed
1088     * @author Tom N Harris <tnharris@whoopdedo.org>
1089     */
1090    protected function getIndexKey($idx, $suffix, $id) {
1091        global $conf;
1092        $fn = $conf['indexdir'].'/'.$idx.$suffix.'.idx';
1093        if (!@file_exists($fn)) return '';
1094        $fh = @fopen($fn, 'r');
1095        if (!$fh) return '';
1096        $ln = -1;
1097        while (($line = fgets($fh)) !== false) {
1098            if (++$ln == $id) break;
1099        }
1100        fclose($fh);
1101        return rtrim((string)$line);
1102    }
1103
1104    /**
1105     * Write a line into the index.
1106     *
1107     * @param string    $idx    name of the index
1108     * @param string    $suffix subpart identifier
1109     * @param int       $id     the line number
1110     * @param string    $line   line to write
1111     * @return bool             If saving succeeded
1112     * @author Tom N Harris <tnharris@whoopdedo.org>
1113     */
1114    protected function saveIndexKey($idx, $suffix, $id, $line) {
1115        global $conf;
1116        if (substr($line, -1) != "\n")
1117            $line .= "\n";
1118        $fn = $conf['indexdir'].'/'.$idx.$suffix;
1119        $fh = @fopen($fn.'.tmp', 'w');
1120        if (!$fh) return false;
1121        $ih = @fopen($fn.'.idx', 'r');
1122        if ($ih) {
1123            $ln = -1;
1124            while (($curline = fgets($ih)) !== false) {
1125                fwrite($fh, (++$ln == $id) ? $line : $curline);
1126            }
1127            if ($id > $ln) {
1128                while ($id > ++$ln)
1129                    fwrite($fh, "\n");
1130                fwrite($fh, $line);
1131            }
1132            fclose($ih);
1133        } else {
1134            $ln = -1;
1135            while ($id > ++$ln)
1136                fwrite($fh, "\n");
1137            fwrite($fh, $line);
1138        }
1139        fclose($fh);
1140        if (isset($conf['fperm']))
1141            chmod($fn.'.tmp', $conf['fperm']);
1142        io_rename($fn.'.tmp', $fn.'.idx');
1143        return true;
1144    }
1145
1146    /**
1147     * Retrieve or insert a value in the index.
1148     *
1149     * @param string    $idx    name of the index
1150     * @param string    $suffix subpart identifier
1151     * @param string    $value  line to find in the index
1152     * @return int|bool          line number of the value in the index or false if writing the index failed
1153     * @author Tom N Harris <tnharris@whoopdedo.org>
1154     */
1155    protected function addIndexKey($idx, $suffix, $value) {
1156        $index = $this->getIndex($idx, $suffix);
1157        $id = array_search($value, $index, true);
1158        if ($id === false) {
1159            $id = count($index);
1160            $index[$id] = $value;
1161            if (!$this->saveIndex($idx, $suffix, $index)) {
1162                trigger_error("Failed to write $idx index", E_USER_ERROR);
1163                return false;
1164            }
1165        }
1166        return $id;
1167    }
1168
1169    /**
1170     * Get the list of lengths indexed in the wiki.
1171     *
1172     * Read the index directory or a cache file and returns
1173     * a sorted array of lengths of the words used in the wiki.
1174     *
1175     * @author YoBoY <yoboy.leguesh@gmail.com>
1176     */
1177    protected function listIndexLengths() {
1178        return idx_listIndexLengths();
1179    }
1180
1181    /**
1182     * Get the word lengths that have been indexed.
1183     *
1184     * Reads the index directory and returns an array of lengths
1185     * that there are indices for.
1186     *
1187     * @author YoBoY <yoboy.leguesh@gmail.com>
1188     */
1189    protected function indexLengths($filter) {
1190        global $conf;
1191        $idx = array();
1192        if (is_array($filter)) {
1193            // testing if index files exist only
1194            $path = $conf['indexdir']."/i";
1195            foreach ($filter as $key => $value) {
1196                if (@file_exists($path.$key.'.idx'))
1197                    $idx[] = $key;
1198            }
1199        } else {
1200            $lengths = idx_listIndexLengths();
1201            foreach ($lengths as $key => $length) {
1202                // keep all the values equal or superior
1203                if ((int)$length >= (int)$filter)
1204                    $idx[] = $length;
1205            }
1206        }
1207        return $idx;
1208    }
1209
1210    /**
1211     * Insert or replace a tuple in a line.
1212     *
1213     * @author Tom N Harris <tnharris@whoopdedo.org>
1214     */
1215    protected function updateTuple($line, $id, $count) {
1216        $newLine = $line;
1217        if ($newLine !== '')
1218            $newLine = preg_replace('/(^|:)'.preg_quote($id,'/').'\*\d*/', '', $newLine);
1219        $newLine = trim($newLine, ':');
1220        if ($count) {
1221            if (strlen($newLine) > 0)
1222                return "$id*$count:".$newLine;
1223            else
1224                return "$id*$count".$newLine;
1225        }
1226        return $newLine;
1227    }
1228
1229    /**
1230     * Split a line into an array of tuples.
1231     *
1232     * @author Tom N Harris <tnharris@whoopdedo.org>
1233     * @author Andreas Gohr <andi@splitbrain.org>
1234     */
1235    protected function parseTuples(&$keys, $line) {
1236        $result = array();
1237        if ($line == '') return $result;
1238        $parts = explode(':', $line);
1239        foreach ($parts as $tuple) {
1240            if ($tuple === '') continue;
1241            list($key, $cnt) = explode('*', $tuple);
1242            if (!$cnt) continue;
1243            $key = $keys[$key];
1244            if (!$key) continue;
1245            $result[$key] = $cnt;
1246        }
1247        return $result;
1248    }
1249
1250    /**
1251     * Sum the counts in a list of tuples.
1252     *
1253     * @author Tom N Harris <tnharris@whoopdedo.org>
1254     */
1255    protected function countTuples($line) {
1256        $freq = 0;
1257        $parts = explode(':', $line);
1258        foreach ($parts as $tuple) {
1259            if ($tuple === '') continue;
1260            list($pid, $cnt) = explode('*', $tuple);
1261            $freq += (int)$cnt;
1262        }
1263        return $freq;
1264    }
1265}
1266
1267/**
1268 * Create an instance of the indexer.
1269 *
1270 * @return Doku_Indexer               a Doku_Indexer
1271 * @author Tom N Harris <tnharris@whoopdedo.org>
1272 */
1273function idx_get_indexer() {
1274    static $Indexer;
1275    if (!isset($Indexer)) {
1276        $Indexer = new Doku_Indexer();
1277    }
1278    return $Indexer;
1279}
1280
1281/**
1282 * Returns words that will be ignored.
1283 *
1284 * @return array                list of stop words
1285 * @author Tom N Harris <tnharris@whoopdedo.org>
1286 */
1287function & idx_get_stopwords() {
1288    static $stopwords = null;
1289    if (is_null($stopwords)) {
1290        global $conf;
1291        $swfile = DOKU_INC.'inc/lang/'.$conf['lang'].'/stopwords.txt';
1292        if(@file_exists($swfile)){
1293            $stopwords = file($swfile, FILE_IGNORE_NEW_LINES);
1294        }else{
1295            $stopwords = array();
1296        }
1297    }
1298    return $stopwords;
1299}
1300
1301/**
1302 * Adds/updates the search index for the given page
1303 *
1304 * Locking is handled internally.
1305 *
1306 * @param string        $page   name of the page to index
1307 * @param boolean       $verbose    print status messages
1308 * @param boolean       $force  force reindexing even when the index is up to date
1309 * @return boolean              the function completed successfully
1310 * @author Tom N Harris <tnharris@whoopdedo.org>
1311 */
1312function idx_addPage($page, $verbose=false, $force=false) {
1313    $idxtag = metaFN($page,'.indexed');
1314    // check if page was deleted but is still in the index
1315    if (!page_exists($page)) {
1316        if (!@file_exists($idxtag)) {
1317            if ($verbose) print("Indexer: $page does not exist, ignoring".DOKU_LF);
1318            return false;
1319        }
1320        $Indexer = idx_get_indexer();
1321        $result = $Indexer->deletePage($page);
1322        if ($result === "locked") {
1323            if ($verbose) print("Indexer: locked".DOKU_LF);
1324            return false;
1325        }
1326        @unlink($idxtag);
1327        return $result;
1328    }
1329
1330    // check if indexing needed
1331    if(!$force && @file_exists($idxtag)){
1332        if(trim(io_readFile($idxtag)) == idx_get_version()){
1333            $last = @filemtime($idxtag);
1334            if($last > @filemtime(wikiFN($page))){
1335                if ($verbose) print("Indexer: index for $page up to date".DOKU_LF);
1336                return false;
1337            }
1338        }
1339    }
1340
1341    $indexenabled = p_get_metadata($page, 'internal index', METADATA_RENDER_UNLIMITED);
1342    if ($indexenabled === false) {
1343        $result = false;
1344        if (@file_exists($idxtag)) {
1345            $Indexer = idx_get_indexer();
1346            $result = $Indexer->deletePage($page);
1347            if ($result === "locked") {
1348                if ($verbose) print("Indexer: locked".DOKU_LF);
1349                return false;
1350            }
1351            @unlink($idxtag);
1352        }
1353        if ($verbose) print("Indexer: index disabled for $page".DOKU_LF);
1354        return $result;
1355    }
1356
1357    $Indexer = idx_get_indexer();
1358    $pid = $Indexer->getPID($page);
1359    if ($pid === false) {
1360        if ($verbose) print("Indexer: getting the PID failed for $page".DOKU_LF);
1361        return false;
1362    }
1363    $body = '';
1364    $metadata = array();
1365    $metadata['title'] = p_get_metadata($page, 'title', METADATA_RENDER_UNLIMITED);
1366    if (($references = p_get_metadata($page, 'relation references', METADATA_RENDER_UNLIMITED)) !== null)
1367        $metadata['relation_references'] = array_keys($references);
1368    else
1369        $metadata['relation_references'] = array();
1370
1371    if (($media = p_get_metadata($page, 'relation media', METADATA_RENDER_UNLIMITED)) !== null)
1372        $metadata['relation_media'] = array_keys($media);
1373    else
1374        $metadata['relation_media'] = array();
1375
1376    $data = compact('page', 'body', 'metadata', 'pid');
1377    $evt = new Doku_Event('INDEXER_PAGE_ADD', $data);
1378    if ($evt->advise_before()) $data['body'] = $data['body'] . " " . rawWiki($page);
1379    $evt->advise_after();
1380    unset($evt);
1381    extract($data);
1382
1383    $result = $Indexer->addPageWords($page, $body);
1384    if ($result === "locked") {
1385        if ($verbose) print("Indexer: locked".DOKU_LF);
1386        return false;
1387    }
1388
1389    if ($result) {
1390        $result = $Indexer->addMetaKeys($page, $metadata);
1391        if ($result === "locked") {
1392            if ($verbose) print("Indexer: locked".DOKU_LF);
1393            return false;
1394        }
1395    }
1396
1397    if ($result)
1398        io_saveFile(metaFN($page,'.indexed'), idx_get_version());
1399    if ($verbose) {
1400        print("Indexer: finished".DOKU_LF);
1401        return true;
1402    }
1403    return $result;
1404}
1405
1406/**
1407 * Find tokens in the fulltext index
1408 *
1409 * Takes an array of words and will return a list of matching
1410 * pages for each one.
1411 *
1412 * Important: No ACL checking is done here! All results are
1413 *            returned, regardless of permissions
1414 *
1415 * @param array      $words  list of words to search for
1416 * @return array             list of pages found, associated with the search terms
1417 */
1418function idx_lookup(&$words) {
1419    $Indexer = idx_get_indexer();
1420    return $Indexer->lookup($words);
1421}
1422
1423/**
1424 * Split a string into tokens
1425 *
1426 */
1427function idx_tokenizer($string, $wc=false) {
1428    $Indexer = idx_get_indexer();
1429    return $Indexer->tokenizer($string, $wc);
1430}
1431
1432/* For compatibility */
1433
1434/**
1435 * Read the list of words in an index (if it exists).
1436 *
1437 * @author Tom N Harris <tnharris@whoopdedo.org>
1438 */
1439function idx_getIndex($idx, $suffix) {
1440    global $conf;
1441    $fn = $conf['indexdir'].'/'.$idx.$suffix.'.idx';
1442    if (!@file_exists($fn)) return array();
1443    return file($fn);
1444}
1445
1446/**
1447 * Get the list of lengths indexed in the wiki.
1448 *
1449 * Read the index directory or a cache file and returns
1450 * a sorted array of lengths of the words used in the wiki.
1451 *
1452 * @author YoBoY <yoboy.leguesh@gmail.com>
1453 */
1454function idx_listIndexLengths() {
1455    global $conf;
1456    // testing what we have to do, create a cache file or not.
1457    if ($conf['readdircache'] == 0) {
1458        $docache = false;
1459    } else {
1460        clearstatcache();
1461        if (@file_exists($conf['indexdir'].'/lengths.idx')
1462        && (time() < @filemtime($conf['indexdir'].'/lengths.idx') + $conf['readdircache'])) {
1463            if (($lengths = @file($conf['indexdir'].'/lengths.idx', FILE_IGNORE_NEW_LINES | FILE_SKIP_EMPTY_LINES)) !== false) {
1464                $idx = array();
1465                foreach ($lengths as $length) {
1466                    $idx[] = (int)$length;
1467                }
1468                return $idx;
1469            }
1470        }
1471        $docache = true;
1472    }
1473
1474    if ($conf['readdircache'] == 0 || $docache) {
1475        $dir = @opendir($conf['indexdir']);
1476        if ($dir === false)
1477            return array();
1478        $idx = array();
1479        while (($f = readdir($dir)) !== false) {
1480            if (substr($f, 0, 1) == 'i' && substr($f, -4) == '.idx') {
1481                $i = substr($f, 1, -4);
1482                if (is_numeric($i))
1483                    $idx[] = (int)$i;
1484            }
1485        }
1486        closedir($dir);
1487        sort($idx);
1488        // save this in a file
1489        if ($docache) {
1490            $handle = @fopen($conf['indexdir'].'/lengths.idx', 'w');
1491            @fwrite($handle, implode("\n", $idx));
1492            @fclose($handle);
1493        }
1494        return $idx;
1495    }
1496
1497    return array();
1498}
1499
1500/**
1501 * Get the word lengths that have been indexed.
1502 *
1503 * Reads the index directory and returns an array of lengths
1504 * that there are indices for.
1505 *
1506 * @author YoBoY <yoboy.leguesh@gmail.com>
1507 */
1508function idx_indexLengths($filter) {
1509    global $conf;
1510    $idx = array();
1511    if (is_array($filter)) {
1512        // testing if index files exist only
1513        $path = $conf['indexdir']."/i";
1514        foreach ($filter as $key => $value) {
1515            if (@file_exists($path.$key.'.idx'))
1516                $idx[] = $key;
1517        }
1518    } else {
1519        $lengths = idx_listIndexLengths();
1520        foreach ($lengths as $key => $length) {
1521            // keep all the values equal or superior
1522            if ((int)$length >= (int)$filter)
1523                $idx[] = $length;
1524        }
1525    }
1526    return $idx;
1527}
1528
1529/**
1530 * Clean a name of a key for use as a file name.
1531 *
1532 * Romanizes non-latin characters, then strips away anything that's
1533 * not a letter, number, or underscore.
1534 *
1535 * @author Tom N Harris <tnharris@whoopdedo.org>
1536 */
1537function idx_cleanName($name) {
1538    $name = utf8_romanize(trim((string)$name));
1539    $name = preg_replace('#[ \./\\:-]+#', '_', $name);
1540    $name = preg_replace('/[^A-Za-z0-9_]/', '', $name);
1541    return strtolower($name);
1542}
1543
1544//Setup VIM: ex: et ts=4 :
1545