xref: /dokuwiki/inc/indexer.php (revision 159765765d0a79d18bc59d46362365c552218f68)
1<?php
2/**
3 * Common DokuWiki functions
4 *
5 * @license    GPL 2 (http://www.gnu.org/licenses/gpl.html)
6 * @author     Andreas Gohr <andi@splitbrain.org>
7 */
8
9if(!defined('DOKU_INC')) die('meh.');
10require_once(DOKU_INC.'inc/io.php');
11require_once(DOKU_INC.'inc/utf8.php');
12require_once(DOKU_INC.'inc/parserutils.php');
13
14// set the minimum token length to use in the index (note, this doesn't apply to numeric tokens)
15if (!defined('IDX_MINWORDLENGTH')) define('IDX_MINWORDLENGTH',3);
16
17// Asian characters are handled as words. The following regexp defines the
18// Unicode-Ranges for Asian characters
19// Ranges taken from http://en.wikipedia.org/wiki/Unicode_block
20// I'm no language expert. If you think some ranges are wrongly chosen or
21// a range is missing, please contact me
22define('IDX_ASIAN1','[\x{0E00}-\x{0E7F}]'); // Thai
23define('IDX_ASIAN2','['.
24                   '\x{2E80}-\x{3040}'.  // CJK -> Hangul
25                   '\x{309D}-\x{30A0}'.
26                   '\x{30FD}-\x{31EF}\x{3200}-\x{D7AF}'.
27                   '\x{F900}-\x{FAFF}'.  // CJK Compatibility Ideographs
28                   '\x{FE30}-\x{FE4F}'.  // CJK Compatibility Forms
29                   ']');
30define('IDX_ASIAN3','['.                // Hiragana/Katakana (can be two characters)
31                   '\x{3042}\x{3044}\x{3046}\x{3048}'.
32                   '\x{304A}-\x{3062}\x{3064}-\x{3082}'.
33                   '\x{3084}\x{3086}\x{3088}-\x{308D}'.
34                   '\x{308F}-\x{3094}'.
35                   '\x{30A2}\x{30A4}\x{30A6}\x{30A8}'.
36                   '\x{30AA}-\x{30C2}\x{30C4}-\x{30E2}'.
37                   '\x{30E4}\x{30E6}\x{30E8}-\x{30ED}'.
38                   '\x{30EF}-\x{30F4}\x{30F7}-\x{30FA}'.
39                   ']['.
40                   '\x{3041}\x{3043}\x{3045}\x{3047}\x{3049}'.
41                   '\x{3063}\x{3083}\x{3085}\x{3087}\x{308E}\x{3095}-\x{309C}'.
42                   '\x{30A1}\x{30A3}\x{30A5}\x{30A7}\x{30A9}'.
43                   '\x{30C3}\x{30E3}\x{30E5}\x{30E7}\x{30EE}\x{30F5}\x{30F6}\x{30FB}\x{30FC}'.
44                   '\x{31F0}-\x{31FF}'.
45                   ']?');
46define('IDX_ASIAN', '(?:'.IDX_ASIAN1.'|'.IDX_ASIAN2.'|'.IDX_ASIAN3.')');
47
48/**
49 * Measure the length of a string.
50 * Differs from strlen in handling of asian characters.
51 *
52 * @author Tom N Harris <tnharris@whoopdedo.org>
53 */
54function wordlen($w){
55    $l = strlen($w);
56    // If left alone, all chinese "words" will get put into w3.idx
57    // So the "length" of a "word" is faked
58    if(preg_match('/'.IDX_ASIAN2.'/u',$w))
59        $l += ord($w) - 0xE1;  // Lead bytes from 0xE2-0xEF
60    return $l;
61}
62
63/**
64 * Write a list of strings to an index file.
65 *
66 * @author Tom N Harris <tnharris@whoopdedo.org>
67 */
68function idx_saveIndex($pre, $wlen, &$idx){
69    global $conf;
70    $fn = $conf['indexdir'].'/'.$pre.$wlen;
71    $fh = @fopen($fn.'.tmp','w');
72    if(!$fh) return false;
73    foreach ($idx as $line) {
74        fwrite($fh,$line);
75    }
76    fclose($fh);
77    if($conf['fperm']) chmod($fn.'.tmp', $conf['fperm']);
78    io_rename($fn.'.tmp', $fn.'.idx');
79    return true;
80}
81
82/**
83 * Read the list of words in an index (if it exists).
84 *
85 * @author Tom N Harris <tnharris@whoopdedo.org>
86 */
87function idx_getIndex($pre, $wlen){
88    global $conf;
89    $fn = $conf['indexdir'].'/'.$pre.$wlen.'.idx';
90    if(!@file_exists($fn)) return array();
91    return file($fn);
92}
93
94/**
95 * Create an empty index file if it doesn't exist yet.
96 *
97 * FIXME: This function isn't currently used. It will probably be removed soon.
98 *
99 * @author Tom N Harris <tnharris@whoopdedo.org>
100 */
101function idx_touchIndex($pre, $wlen){
102    global $conf;
103    $fn = $conf['indexdir'].'/'.$pre.$wlen.'.idx';
104    if(!@file_exists($fn)){
105        touch($fn);
106        if($conf['fperm']) chmod($fn, $conf['fperm']);
107    }
108}
109
110/**
111 * Read a line ending with \n.
112 * Returns false on EOF.
113 *
114 * @author Tom N Harris <tnharris@whoopdedo.org>
115 */
116function _freadline($fh) {
117    if (feof($fh)) return false;
118    $ln = '';
119    while (($buf = fgets($fh,4096)) !== false) {
120        $ln .= $buf;
121        if (substr($buf,-1) == "\n") break;
122    }
123    if ($ln === '') return false;
124    if (substr($ln,-1) != "\n") $ln .= "\n";
125    return $ln;
126}
127
128/**
129 * Write a line to an index file.
130 *
131 * @author Tom N Harris <tnharris@whoopdedo.org>
132 */
133function idx_saveIndexLine($pre, $wlen, $idx, $line){
134    global $conf;
135    if(substr($line,-1) != "\n") $line .= "\n";
136    $fn = $conf['indexdir'].'/'.$pre.$wlen;
137    $fh = @fopen($fn.'.tmp','w');
138    if(!$fh) return false;
139    $ih = @fopen($fn.'.idx','r');
140    if ($ih) {
141        $ln = -1;
142        while (($curline = _freadline($ih)) !== false) {
143            if (++$ln == $idx) {
144                fwrite($fh, $line);
145            } else {
146                fwrite($fh, $curline);
147            }
148        }
149        if ($idx > $ln) {
150            fwrite($fh,$line);
151        }
152        fclose($ih);
153    } else {
154        fwrite($fh,$line);
155    }
156    fclose($fh);
157    if($conf['fperm']) chmod($fn.'.tmp', $conf['fperm']);
158    io_rename($fn.'.tmp', $fn.'.idx');
159    return true;
160}
161
162/**
163 * Read a single line from an index (if it exists).
164 *
165 * @author Tom N Harris <tnharris@whoopdedo.org>
166 */
167function idx_getIndexLine($pre, $wlen, $idx){
168    global $conf;
169    $fn = $conf['indexdir'].'/'.$pre.$wlen.'.idx';
170    if(!@file_exists($fn)) return '';
171    $fh = @fopen($fn,'r');
172    if(!$fh) return '';
173    $ln = -1;
174    while (($line = _freadline($fh)) !== false) {
175        if (++$ln == $idx) break;
176    }
177    fclose($fh);
178    return "$line";
179}
180
181/**
182 * Split a page into words
183 *
184 * Returns an array of word counts, false if an error occurred.
185 * Array is keyed on the word length, then the word index.
186 *
187 * @author Andreas Gohr <andi@splitbrain.org>
188 * @author Christopher Smith <chris@jalakai.co.uk>
189 */
190function idx_getPageWords($page){
191    global $conf;
192    $swfile   = DOKU_INC.'inc/lang/'.$conf['lang'].'/stopwords.txt';
193    if(@file_exists($swfile)){
194        $stopwords = file($swfile);
195    }else{
196        $stopwords = array();
197    }
198
199    $body = '';
200    $data = array($page, $body);
201    $evt = new Doku_Event('INDEXER_PAGE_ADD', $data);
202    if ($evt->advise_before()) $data[1] .= rawWiki($page);
203    $evt->advise_after();
204    unset($evt);
205
206    list($page,$body) = $data;
207
208    $body   = strtr($body, "\r\n\t", '   ');
209    $tokens = explode(' ', $body);
210    $tokens = array_count_values($tokens);   // count the frequency of each token
211
212    // ensure the deaccented or romanised page names of internal links are added to the token array
213    // (this is necessary for the backlink function -- there maybe a better way!)
214    if ($conf['deaccent']) {
215      $links = p_get_metadata($page,'relation references');
216
217      if (!empty($links)) {
218        $tmp = join(' ',array_keys($links));                // make a single string
219        $tmp = strtr($tmp, ':', ' ');                       // replace namespace separator with a space
220        $link_tokens = array_unique(explode(' ', $tmp));    // break into tokens
221
222        foreach ($link_tokens as $link_token) {
223          if (isset($tokens[$link_token])) continue;
224          $tokens[$link_token] = 1;
225        }
226      }
227    }
228
229    $words = array();
230    foreach ($tokens as $word => $count) {
231        $arr = idx_tokenizer($word,$stopwords);
232        $arr = array_count_values($arr);
233        foreach ($arr as $w => $c) {
234            $l = wordlen($w);
235            if(isset($words[$l])){
236                $words[$l][$w] = $c * $count + (isset($words[$l][$w]) ? $words[$l][$w] : 0);
237            }else{
238                $words[$l] = array($w => $c * $count);
239            }
240        }
241    }
242
243    // arrive here with $words = array(wordlen => array(word => frequency))
244
245    $index = array(); //resulting index
246    foreach (array_keys($words) as $wlen){
247        $word_idx = idx_getIndex('w',$wlen);
248        foreach ($words[$wlen] as $word => $freq) {
249            $wid = array_search("$word\n",$word_idx);
250            if(!is_int($wid)){
251                $wid = count($word_idx);
252                $word_idx[] = "$word\n";
253            }
254            if(!isset($index[$wlen]))
255                $index[$wlen] = array();
256            $index[$wlen][$wid] = $freq;
257        }
258
259        // save back word index
260        if(!idx_saveIndex('w',$wlen,$word_idx)){
261            trigger_error("Failed to write word index", E_USER_ERROR);
262            return false;
263        }
264    }
265
266    return $index;
267}
268
269/**
270 * Adds/updates the search for the given page
271 *
272 * This is the core function of the indexer which does most
273 * of the work. This function needs to be called with proper
274 * locking!
275 *
276 * @author Andreas Gohr <andi@splitbrain.org>
277 */
278function idx_addPage($page){
279    global $conf;
280
281    // load known documents
282    $page_idx = idx_getIndex('page','');
283
284    // get page id (this is the linenumber in page.idx)
285    $pid = array_search("$page\n",$page_idx);
286    if(!is_int($pid)){
287        $page_idx[] = "$page\n";
288        $pid = count($page_idx)-1;
289        // page was new - write back
290        if (!idx_saveIndex('page','',$page_idx)){
291            trigger_error("Failed to write page index", E_USER_ERROR);
292            return false;
293        }
294    }
295
296    $pagewords = array();
297    // get word usage in page
298    $words = idx_getPageWords($page);
299    if($words === false) return false;
300
301    if(!empty($words)) {
302        foreach(array_keys($words) as $wlen){
303            $index = idx_getIndex('i',$wlen);
304            foreach($words[$wlen] as $wid => $freq){
305                if($wid<count($index)){
306                    $index[$wid] = idx_updateIndexLine($index[$wid],$pid,$freq);
307                }else{
308                    // New words **should** have been added in increasing order
309                    // starting with the first unassigned index.
310                    // If someone can show how this isn't true, then I'll need to sort
311                    // or do something special.
312                    $index[$wid] = idx_updateIndexLine('',$pid,$freq);
313                }
314                $pagewords[] = "$wlen*$wid";
315            }
316            // save back word index
317            if(!idx_saveIndex('i',$wlen,$index)){
318                trigger_error("Failed to write index", E_USER_ERROR);
319                return false;
320            }
321        }
322    }
323
324    // Remove obsolete index entries
325    $pageword_idx = trim(idx_getIndexLine('pageword','',$pid));
326    if ($pageword_idx !== '') {
327        $oldwords = explode(':',$pageword_idx);
328        $delwords = array_diff($oldwords, $pagewords);
329        $upwords = array();
330        foreach ($delwords as $word) {
331            if($word=='') continue;
332            list($wlen,$wid) = explode('*',$word);
333            $wid = (int)$wid;
334            $upwords[$wlen][] = $wid;
335        }
336        foreach ($upwords as $wlen => $widx) {
337            $index = idx_getIndex('i',$wlen);
338            foreach ($widx as $wid) {
339                $index[$wid] = idx_updateIndexLine($index[$wid],$pid,0);
340            }
341            idx_saveIndex('i',$wlen,$index);
342        }
343    }
344    // Save the reverse index
345    $pageword_idx = join(':',$pagewords)."\n";
346    if(!idx_saveIndexLine('pageword','',$pid,$pageword_idx)){
347        trigger_error("Failed to write word index", E_USER_ERROR);
348        return false;
349    }
350
351    return true;
352}
353
354/**
355 * Write a new index line to the filehandle
356 *
357 * This function writes an line for the index file to the
358 * given filehandle. It removes the given document from
359 * the given line and readds it when $count is >0.
360 *
361 * @deprecated - see idx_updateIndexLine
362 * @author Andreas Gohr <andi@splitbrain.org>
363 */
364function idx_writeIndexLine($fh,$line,$pid,$count){
365    fwrite($fh,idx_updateIndexLine($line,$pid,$count));
366}
367
368/**
369 * Modify an index line with new information
370 *
371 * This returns a line of the index. It removes the
372 * given document from the line and readds it if
373 * $count is >0.
374 *
375 * @author Tom N Harris <tnharris@whoopdedo.org>
376 * @author Andreas Gohr <andi@splitbrain.org>
377 */
378function idx_updateIndexLine($line,$pid,$count){
379    $line = trim($line);
380    $updated = array();
381    if($line != ''){
382        $parts = explode(':',$line);
383        // remove doc from given line
384        foreach($parts as $part){
385            if($part == '') continue;
386            list($doc,$cnt) = explode('*',$part);
387            if($doc != $pid){
388                $updated[] = $part;
389            }
390        }
391    }
392
393    // add doc
394    if ($count){
395        $updated[] = "$pid*$count";
396    }
397
398    return join(':',$updated)."\n";
399}
400
401/**
402 * Get the word lengths that have been indexed.
403 *
404 * Reads the index directory and returns an array of lengths
405 * that there are indices for.
406 *
407 * @author Tom N Harris <tnharris@whoopdedo.org>
408 */
409function idx_indexLengths(&$filter){
410    global $conf;
411    $dir = @opendir($conf['indexdir']);
412    if($dir===false)
413        return array();
414    $idx = array();
415    if(is_array($filter)){
416        while (($f = readdir($dir)) !== false) {
417            if (substr($f,0,1) == 'i' && substr($f,-4) == '.idx'){
418                $i = substr($f,1,-4);
419                if (is_numeric($i) && isset($filter[(int)$i]))
420                    $idx[] = (int)$i;
421            }
422        }
423    }else{
424        // Exact match first.
425        if(@file_exists($conf['indexdir']."/i$filter.idx"))
426            $idx[] = $filter;
427        while (($f = readdir($dir)) !== false) {
428            if (substr($f,0,1) == 'i' && substr($f,-4) == '.idx'){
429                $i = substr($f,1,-4);
430                if (is_numeric($i) && $i > $filter)
431                    $idx[] = (int)$i;
432            }
433        }
434    }
435    closedir($dir);
436    return $idx;
437}
438
439/**
440 * Find the the index number of each search term.
441 *
442 * This will group together words that appear in the same index.
443 * So it should perform better, because it only opens each index once.
444 * Actually, it's not that great. (in my experience) Probably because of the disk cache.
445 * And the sorted function does more work, making it slightly slower in some cases.
446 *
447 * @param array    $words   The query terms. Words should only contain valid characters,
448 *                          with a '*' at either the beginning or end of the word (or both)
449 * @param arrayref $result  Set to word => array("length*id" ...), use this to merge the
450 *                          index locations with the appropriate query term.
451 * @return array            Set to length => array(id ...)
452 *
453 * @author Tom N Harris <tnharris@whoopdedo.org>
454 */
455function idx_getIndexWordsSorted($words,&$result){
456    // parse and sort tokens
457    $tokens = array();
458    $tokenlength = array();
459    $tokenwild = array();
460    foreach($words as $word){
461        $result[$word] = array();
462        $wild = 0;
463        $xword = $word;
464        $wlen = wordlen($word);
465
466        // check for wildcards
467        if(substr($xword,0,1) == '*'){
468            $xword = substr($xword,1);
469            $wild |= 1;
470            $wlen -= 1;
471        }
472        if(substr($xword,-1,1) == '*'){
473            $xword = substr($xword,0,-1);
474            $wild |= 2;
475            $wlen -= 1;
476        }
477        if ($wlen < IDX_MINWORDLENGTH && $wild == 0 && !is_numeric($xword)) continue;
478        if(!isset($tokens[$xword])){
479            $tokenlength[$wlen][] = $xword;
480        }
481        if($wild){
482            $ptn = preg_quote($xword,'/');
483            if(($wild&1) == 0) $ptn = '^'.$ptn;
484            if(($wild&2) == 0) $ptn = $ptn.'$';
485            $tokens[$xword][] = array($word, '/'.$ptn.'/');
486            if(!isset($tokenwild[$xword])) $tokenwild[$xword] = $wlen;
487        }else
488            $tokens[$xword][] = array($word, null);
489    }
490    asort($tokenwild);
491    // $tokens = array( base word => array( [ query word , grep pattern ] ... ) ... )
492    // $tokenlength = array( base word length => base word ... )
493    // $tokenwild = array( base word => base word length ... )
494
495    $length_filter = empty($tokenwild) ? $tokenlength : min(array_keys($tokenlength));
496    $indexes_known = idx_indexLengths($length_filter);
497    if(!empty($tokenwild)) sort($indexes_known);
498    // get word IDs
499    $wids = array();
500    foreach($indexes_known as $ixlen){
501        $word_idx = idx_getIndex('w',$ixlen);
502        // handle exact search
503        if(isset($tokenlength[$ixlen])){
504            foreach($tokenlength[$ixlen] as $xword){
505                $wid = array_search("$xword\n",$word_idx);
506                if(is_int($wid)){
507                    $wids[$ixlen][] = $wid;
508                    foreach($tokens[$xword] as $w)
509                        $result[$w[0]][] = "$ixlen*$wid";
510                }
511            }
512        }
513        // handle wildcard search
514        foreach($tokenwild as $xword => $wlen){
515            if($wlen >= $ixlen) break;
516            foreach($tokens[$xword] as $w){
517                if(is_null($w[1])) continue;
518                foreach(array_keys(preg_grep($w[1],$word_idx)) as $wid){
519                    $wids[$ixlen][] = $wid;
520                    $result[$w[0]][] = "$ixlen*$wid";
521                }
522            }
523        }
524    }
525  return $wids;
526}
527
528/**
529 * Lookup words in index
530 *
531 * Takes an array of word and will return a list of matching
532 * documents for each one.
533 *
534 * Important: No ACL checking is done here! All results are
535 *            returned, regardless of permissions
536 *
537 * @author Andreas Gohr <andi@splitbrain.org>
538 */
539function idx_lookup($words){
540    global $conf;
541
542    $result = array();
543
544    $wids = idx_getIndexWordsSorted($words, $result);
545    if(empty($wids)) return array();
546
547    // load known words and documents
548    $page_idx = idx_getIndex('page','');
549
550    $docs = array();                          // hold docs found
551    foreach(array_keys($wids) as $wlen){
552        $wids[$wlen] = array_unique($wids[$wlen]);
553        $index = idx_getIndex('i',$wlen);
554        foreach($wids[$wlen] as $ixid){
555            if($ixid < count($index))
556                $docs["$wlen*$ixid"] = idx_parseIndexLine($page_idx,$index[$ixid]);
557        }
558    }
559
560    // merge found pages into final result array
561    $final = array();
562    foreach(array_keys($result) as $word){
563        $final[$word] = array();
564        foreach($result[$word] as $wid){
565            $hits = &$docs[$wid];
566            foreach ($hits as $hitkey => $hitcnt) {
567                $final[$word][$hitkey] = $hitcnt + $final[$word][$hitkey];
568            }
569        }
570    }
571    return $final;
572}
573
574/**
575 * Returns a list of documents and counts from a index line
576 *
577 * It omits docs with a count of 0 and pages that no longer
578 * exist.
579 *
580 * @param  array  $page_idx The list of known pages
581 * @param  string $line     A line from the main index
582 * @author Andreas Gohr <andi@splitbrain.org>
583 */
584function idx_parseIndexLine(&$page_idx,$line){
585    $result = array();
586
587    $line = trim($line);
588    if($line == '') return $result;
589
590    $parts = explode(':',$line);
591    foreach($parts as $part){
592        if($part == '') continue;
593        list($doc,$cnt) = explode('*',$part);
594        if(!$cnt) continue;
595        $doc = trim($page_idx[$doc]);
596        if(!$doc) continue;
597        // make sure the document still exists
598        if(!page_exists($doc,'',false)) continue;
599
600        $result[$doc] = $cnt;
601    }
602    return $result;
603}
604
605/**
606 * Tokenizes a string into an array of search words
607 *
608 * Uses the same algorithm as idx_getPageWords()
609 *
610 * @param string   $string     the query as given by the user
611 * @param arrayref $stopwords  array of stopwords
612 * @param boolean  $wc         are wildcards allowed?
613 */
614function idx_tokenizer($string,&$stopwords,$wc=false){
615    $words = array();
616    $wc = ($wc) ? '' : $wc = '\*';
617
618    if(preg_match('/[^0-9A-Za-z]/u', $string)){
619        // handle asian chars as single words (may fail on older PHP version)
620        $asia = @preg_replace('/('.IDX_ASIAN.')/u',' \1 ',$string);
621        if(!is_null($asia)) $string = $asia; //recover from regexp failure
622
623        $arr = explode(' ', utf8_stripspecials($string,' ','\._\-:'.$wc));
624        foreach ($arr as $w) {
625            if (!is_numeric($w) && strlen($w) < IDX_MINWORDLENGTH) continue;
626            $w = utf8_strtolower($w);
627            if($stopwords && is_int(array_search("$w\n",$stopwords))) continue;
628            $words[] = $w;
629        }
630    }else{
631        $w = $string;
632        if (!is_numeric($w) && strlen($w) < IDX_MINWORDLENGTH) return $words;
633        $w = strtolower($w);
634        if(is_int(array_search("$w\n",$stopwords))) return $words;
635        $words[] = $w;
636    }
637
638    return $words;
639}
640
641/**
642 * Create a pagewords index from the existing index.
643 *
644 * @author Tom N Harris <tnharris@whoopdedo.org>
645 */
646function idx_upgradePageWords(){
647    global $conf;
648    $page_idx = idx_getIndex('page','');
649    if (empty($page_idx)) return;
650    $pagewords = array();
651    for ($n=0;$n<count($page_idx);$n++) $pagewords[] = array();
652    unset($page_idx);
653
654    $n=0;
655    foreach (idx_indexLengths($n) as $wlen) {
656        $lines = idx_getIndex('i',$wlen);
657        for ($wid=0;$wid<count($lines);$wid++) {
658            $wkey = "$wlen*$wid";
659            foreach (explode(':',trim($lines[$wid])) as $part) {
660                if($part == '') continue;
661                list($doc,$cnt) = explode('*',$part);
662                $pagewords[(int)$doc][] = $wkey;
663            }
664        }
665    }
666
667    $fn = $conf['indexdir'].'/pageword';
668    $fh = @fopen($fn.'.tmp','w');
669    if (!$fh){
670        trigger_error("Failed to write word index", E_USER_ERROR);
671        return false;
672    }
673    foreach ($pagewords as $line){
674        fwrite($fh, join(':',$line)."\n");
675    }
676    fclose($fh);
677    if($conf['fperm']) chmod($fn.'.tmp', $conf['fperm']);
678    io_rename($fn.'.tmp', $fn.'.idx');
679    return true;
680}
681
682//Setup VIM: ex: et ts=4 enc=utf-8 :
683