xref: /dokuwiki/inc/fulltext.php (revision f078bb0088870b4b68b348d546afa30a80a07e87)
1<?php
2/**
3 * DokuWiki fulltextsearch functions using the index
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.');
10
11/**
12 * create snippets for the first few results only
13 */
14if(!defined('FT_SNIPPET_NUMBER')) define('FT_SNIPPET_NUMBER',15);
15
16/**
17 * The fulltext search
18 *
19 * Returns a list of matching documents for the given query
20 *
21 * refactored into ft_pageSearch(), _ft_pageSearch() and trigger_event()
22 *
23 */
24function ft_pageSearch($query,&$highlight){
25
26    $data['query'] = $query;
27    $data['highlight'] =& $highlight;
28
29    return trigger_event('SEARCH_QUERY_FULLPAGE', $data, '_ft_pageSearch');
30}
31
32/**
33 * Returns a list of matching documents for the given query
34 *
35 * @author Andreas Gohr <andi@splitbrain.org>
36 * @author Kazutaka Miyasaka <kazmiya@gmail.com>
37 */
38function _ft_pageSearch(&$data) {
39    $Indexer = idx_get_indexer();
40
41    // parse the given query
42    $q = ft_queryParser($Indexer, $data['query']);
43    $data['highlight'] = $q['highlight'];
44
45    if (empty($q['parsed_ary'])) return array();
46
47    // lookup all words found in the query
48    $lookup = $Indexer->lookup($q['words']);
49
50    // get all pages in this dokuwiki site (!: includes nonexistent pages)
51    $pages_all = array();
52    foreach ($Indexer->getPages() as $id) {
53        $pages_all[$id] = 0; // base: 0 hit
54    }
55
56    // process the query
57    $stack = array();
58    foreach ($q['parsed_ary'] as $token) {
59        switch (substr($token, 0, 3)) {
60            case 'W+:':
61            case 'W-:':
62            case 'W_:': // word
63                $word    = substr($token, 3);
64                $stack[] = (array) $lookup[$word];
65                break;
66            case 'P+:':
67            case 'P-:': // phrase
68                $phrase = substr($token, 3);
69                // since phrases are always parsed as ((W1)(W2)...(P)),
70                // the end($stack) always points the pages that contain
71                // all words in this phrase
72                $pages  = end($stack);
73                $pages_matched = array();
74                foreach(array_keys($pages) as $id){
75                    $text = utf8_strtolower(rawWiki($id));
76                    if (strpos($text, $phrase) !== false) {
77                        $pages_matched[$id] = 0; // phrase: always 0 hit
78                    }
79                }
80                $stack[] = $pages_matched;
81                break;
82            case 'N+:':
83            case 'N-:': // namespace
84                $ns = substr($token, 3);
85                $pages_matched = array();
86                foreach (array_keys($pages_all) as $id) {
87                    if (strpos($id, $ns) === 0) {
88                        $pages_matched[$id] = 0; // namespace: always 0 hit
89                    }
90                }
91                $stack[] = $pages_matched;
92                break;
93            case 'AND': // and operation
94                list($pages1, $pages2) = array_splice($stack, -2);
95                $stack[] = ft_resultCombine(array($pages1, $pages2));
96                break;
97            case 'OR':  // or operation
98                list($pages1, $pages2) = array_splice($stack, -2);
99                $stack[] = ft_resultUnite(array($pages1, $pages2));
100                break;
101            case 'NOT': // not operation (unary)
102                $pages   = array_pop($stack);
103                $stack[] = ft_resultComplement(array($pages_all, $pages));
104                break;
105        }
106    }
107    $docs = array_pop($stack);
108
109    if (empty($docs)) return array();
110
111    // check: settings, acls, existence
112    foreach (array_keys($docs) as $id) {
113        if (isHiddenPage($id) || auth_quickaclcheck($id) < AUTH_READ || !page_exists($id, '', false)) {
114            unset($docs[$id]);
115        }
116    }
117
118    // sort docs by count
119    arsort($docs);
120
121    return $docs;
122}
123
124/**
125 * Returns the backlinks for a given page
126 *
127 * Uses the metadata index.
128 */
129function ft_backlinks($id){
130    $result = array();
131
132    $result = idx_get_indexer()->lookupKey('relation_references', $id);
133
134    if(!count($result)) return $result;
135
136    // check ACL permissions
137    foreach(array_keys($result) as $idx){
138        if(auth_quickaclcheck($result[$idx]) < AUTH_READ){
139            unset($result[$idx]);
140        }
141    }
142
143    sort($result);
144    return $result;
145}
146
147/**
148 * Returns the pages that use a given media file
149 *
150 * Does a quick lookup with the fulltext index, then
151 * evaluates the instructions of the found pages
152 *
153 * Aborts after $max found results
154 */
155function ft_mediause($id,$max){
156    if(!$max) $max = 1; // need to find at least one
157
158    $result = array();
159
160    // quick lookup of the mediafile
161    // FIXME use metadata key lookup
162    $media   = noNS($id);
163    $matches = idx_lookup(idx_tokenizer($media));
164    $docs    = array_keys(ft_resultCombine(array_values($matches)));
165    if(!count($docs)) return $result;
166
167    // go through all found pages
168    $found = 0;
169    $pcre  = preg_quote($media,'/');
170    foreach($docs as $doc){
171        $ns = getNS($doc);
172        preg_match_all('/\{\{([^|}]*'.$pcre.'[^|}]*)(|[^}]+)?\}\}/i',rawWiki($doc),$matches);
173        foreach($matches[1] as $img){
174            $img = trim($img);
175            if(preg_match('/^https?:\/\//i',$img)) continue; // skip external images
176                list($img) = explode('?',$img);                  // remove any parameters
177            resolve_mediaid($ns,$img,$exists);               // resolve the possibly relative img
178
179            if($img == $id){                                 // we have a match
180                $result[] = $doc;
181                $found++;
182                break;
183            }
184        }
185        if($found >= $max) break;
186    }
187
188    sort($result);
189    return $result;
190}
191
192
193
194/**
195 * Quicksearch for pagenames
196 *
197 * By default it only matches the pagename and ignores the
198 * namespace. This can be changed with the second parameter.
199 * The third parameter allows to search in titles as well.
200 *
201 * The function always returns titles as well
202 *
203 * @triggers SEARCH_QUERY_PAGELOOKUP
204 * @author Andreas Gohr <andi@splitbrain.org>
205 * @author Adrian Lang <lang@cosmocode.de>
206 */
207function ft_pageLookup($id, $in_ns=false, $in_title=false){
208    $data = compact('id', 'in_ns', 'in_title');
209    $data['has_titles'] = true; // for plugin backward compatibility check
210    return trigger_event('SEARCH_QUERY_PAGELOOKUP', $data, '_ft_pageLookup');
211}
212
213function _ft_pageLookup(&$data){
214    // split out original parameters
215    $id = $data['id'];
216    if (preg_match('/(?:^| )@(\w+)/', $id, $matches)) {
217        $ns = cleanID($matches[1]) . ':';
218        $id = str_replace($matches[0], '', $id);
219    }
220
221    $in_ns    = $data['in_ns'];
222    $in_title = $data['in_title'];
223    $cleaned = cleanID($id);
224
225    $Indexer = idx_get_indexer();
226    $page_idx = $Indexer->getPages();
227
228    $pages = array();
229    if ($id !== '' && $cleaned !== '') {
230        foreach ($page_idx as $p_id) {
231            if ((strpos($in_ns ? $p_id : noNSorNS($p_id), $cleaned) !== false)) {
232                if (!isset($pages[$p_id]))
233                    $pages[$p_id] = p_get_first_heading($p_id, false);
234            }
235        }
236        if ($in_title) {
237            foreach ($Indexer->lookupKey('title', "*$id*") as $p_id) {
238                if (!isset($pages[$p_id]))
239                    $pages[$p_id] = p_get_first_heading($p_id, false);
240            }
241        }
242    }
243    if (isset($ns)) {
244        foreach ($page_idx as $p_id) {
245            if (strpos($p_id, $ns) === 0) {
246                if (!isset($pages[$p_id]))
247                    $pages[$p_id] = p_get_first_heading($p_id, false);
248            }
249        }
250    }
251
252    // discard hidden pages
253    // discard nonexistent pages
254    // check ACL permissions
255    foreach(array_keys($pages) as $idx){
256        if(!isVisiblePage($idx) || !page_exists($idx) ||
257           auth_quickaclcheck($idx) < AUTH_READ) {
258            unset($pages[$idx]);
259        }
260    }
261
262    uksort($pages,'ft_pagesorter');
263    return $pages;
264}
265
266/**
267 * Sort pages based on their namespace level first, then on their string
268 * values. This makes higher hierarchy pages rank higher than lower hierarchy
269 * pages.
270 */
271function ft_pagesorter($a, $b){
272    $ac = count(explode(':',$a));
273    $bc = count(explode(':',$b));
274    if($ac < $bc){
275        return -1;
276    }elseif($ac > $bc){
277        return 1;
278    }
279    return strcmp ($a,$b);
280}
281
282/**
283 * Creates a snippet extract
284 *
285 * @author Andreas Gohr <andi@splitbrain.org>
286 * @triggers FULLTEXT_SNIPPET_CREATE
287 */
288function ft_snippet($id,$highlight){
289    $text = rawWiki($id);
290    $evdata = array(
291            'id'        => $id,
292            'text'      => &$text,
293            'highlight' => &$highlight,
294            'snippet'   => '',
295            );
296
297    $evt = new Doku_Event('FULLTEXT_SNIPPET_CREATE',$evdata);
298    if ($evt->advise_before()) {
299        $match = array();
300        $snippets = array();
301        $utf8_offset = $offset = $end = 0;
302        $len = utf8_strlen($text);
303
304        // build a regexp from the phrases to highlight
305        $re1 = '('.join('|',array_map('ft_snippet_re_preprocess', array_map('preg_quote_cb',array_filter((array) $highlight)))).')';
306        $re2 = "$re1.{0,75}(?!\\1)$re1";
307        $re3 = "$re1.{0,45}(?!\\1)$re1.{0,45}(?!\\1)(?!\\2)$re1";
308
309        for ($cnt=4; $cnt--;) {
310            if (0) {
311            } else if (preg_match('/'.$re3.'/iu',$text,$match,PREG_OFFSET_CAPTURE,$offset)) {
312            } else if (preg_match('/'.$re2.'/iu',$text,$match,PREG_OFFSET_CAPTURE,$offset)) {
313            } else if (preg_match('/'.$re1.'/iu',$text,$match,PREG_OFFSET_CAPTURE,$offset)) {
314            } else {
315                break;
316            }
317
318            list($str,$idx) = $match[0];
319
320            // convert $idx (a byte offset) into a utf8 character offset
321            $utf8_idx = utf8_strlen(substr($text,0,$idx));
322            $utf8_len = utf8_strlen($str);
323
324            // establish context, 100 bytes surrounding the match string
325            // first look to see if we can go 100 either side,
326            // then drop to 50 adding any excess if the other side can't go to 50,
327            $pre = min($utf8_idx-$utf8_offset,100);
328            $post = min($len-$utf8_idx-$utf8_len,100);
329
330            if ($pre>50 && $post>50) {
331                $pre = $post = 50;
332            } else if ($pre>50) {
333                $pre = min($pre,100-$post);
334            } else if ($post>50) {
335                $post = min($post, 100-$pre);
336            } else {
337                // both are less than 50, means the context is the whole string
338                // make it so and break out of this loop - there is no need for the
339                // complex snippet calculations
340                $snippets = array($text);
341                break;
342            }
343
344            // establish context start and end points, try to append to previous
345            // context if possible
346            $start = $utf8_idx - $pre;
347            $append = ($start < $end) ? $end : false;  // still the end of the previous context snippet
348            $end = $utf8_idx + $utf8_len + $post;      // now set it to the end of this context
349
350            if ($append) {
351                $snippets[count($snippets)-1] .= utf8_substr($text,$append,$end-$append);
352            } else {
353                $snippets[] = utf8_substr($text,$start,$end-$start);
354            }
355
356            // set $offset for next match attempt
357            //   substract strlen to avoid splitting a potential search success,
358            //   this is an approximation as the search pattern may match strings
359            //   of varying length and it will fail if the context snippet
360            //   boundary breaks a matching string longer than the current match
361            $utf8_offset = $utf8_idx + $post;
362            $offset = $idx + strlen(utf8_substr($text,$utf8_idx,$post));
363            $offset = utf8_correctIdx($text,$offset);
364        }
365
366        $m = "\1";
367        $snippets = preg_replace('/'.$re1.'/iu',$m.'$1'.$m,$snippets);
368        $snippet = preg_replace('/'.$m.'([^'.$m.']*?)'.$m.'/iu','<strong class="search_hit">$1</strong>',hsc(join('... ',$snippets)));
369
370        $evdata['snippet'] = $snippet;
371    }
372    $evt->advise_after();
373    unset($evt);
374
375    return $evdata['snippet'];
376}
377
378/**
379 * Wraps a search term in regex boundary checks.
380 */
381function ft_snippet_re_preprocess($term) {
382    if(substr($term,0,2) == '\\*'){
383        $term = substr($term,2);
384    }else{
385        $term = '\b'.$term;
386    }
387
388    if(substr($term,-2,2) == '\\*'){
389        $term = substr($term,0,-2);
390    }else{
391        $term = $term.'\b';
392    }
393    return $term;
394}
395
396/**
397 * Combine found documents and sum up their scores
398 *
399 * This function is used to combine searched words with a logical
400 * AND. Only documents available in all arrays are returned.
401 *
402 * based upon PEAR's PHP_Compat function for array_intersect_key()
403 *
404 * @param array $args An array of page arrays
405 */
406function ft_resultCombine($args){
407    $array_count = count($args);
408    if($array_count == 1){
409        return $args[0];
410    }
411
412    $result = array();
413    if ($array_count > 1) {
414        foreach ($args[0] as $key => $value) {
415            $result[$key] = $value;
416            for ($i = 1; $i !== $array_count; $i++) {
417                if (!isset($args[$i][$key])) {
418                    unset($result[$key]);
419                    break;
420                }
421                $result[$key] += $args[$i][$key];
422            }
423        }
424    }
425    return $result;
426}
427
428/**
429 * Unites found documents and sum up their scores
430 *
431 * based upon ft_resultCombine() function
432 *
433 * @param array $args An array of page arrays
434 * @author Kazutaka Miyasaka <kazmiya@gmail.com>
435 */
436function ft_resultUnite($args) {
437    $array_count = count($args);
438    if ($array_count === 1) {
439        return $args[0];
440    }
441
442    $result = $args[0];
443    for ($i = 1; $i !== $array_count; $i++) {
444        foreach (array_keys($args[$i]) as $id) {
445            $result[$id] += $args[$i][$id];
446        }
447    }
448    return $result;
449}
450
451/**
452 * Computes the difference of documents using page id for comparison
453 *
454 * nearly identical to PHP5's array_diff_key()
455 *
456 * @param array $args An array of page arrays
457 * @author Kazutaka Miyasaka <kazmiya@gmail.com>
458 */
459function ft_resultComplement($args) {
460    $array_count = count($args);
461    if ($array_count === 1) {
462        return $args[0];
463    }
464
465    $result = $args[0];
466    foreach (array_keys($result) as $id) {
467        for ($i = 1; $i !== $array_count; $i++) {
468            if (isset($args[$i][$id])) unset($result[$id]);
469        }
470    }
471    return $result;
472}
473
474/**
475 * Parses a search query and builds an array of search formulas
476 *
477 * @author Andreas Gohr <andi@splitbrain.org>
478 * @author Kazutaka Miyasaka <kazmiya@gmail.com>
479 */
480function ft_queryParser($Indexer, $query){
481    /**
482     * parse a search query and transform it into intermediate representation
483     *
484     * in a search query, you can use the following expressions:
485     *
486     *   words:
487     *     include
488     *     -exclude
489     *   phrases:
490     *     "phrase to be included"
491     *     -"phrase you want to exclude"
492     *   namespaces:
493     *     @include:namespace (or ns:include:namespace)
494     *     ^exclude:namespace (or -ns:exclude:namespace)
495     *   groups:
496     *     ()
497     *     -()
498     *   operators:
499     *     and ('and' is the default operator: you can always omit this)
500     *     or  (or pipe symbol '|', lower precedence than 'and')
501     *
502     * e.g. a query [ aa "bb cc" @dd:ee ] means "search pages which contain
503     *      a word 'aa', a phrase 'bb cc' and are within a namespace 'dd:ee'".
504     *      this query is equivalent to [ -(-aa or -"bb cc" or -ns:dd:ee) ]
505     *      as long as you don't mind hit counts.
506     *
507     * intermediate representation consists of the following parts:
508     *
509     *   ( )           - group
510     *   AND           - logical and
511     *   OR            - logical or
512     *   NOT           - logical not
513     *   W+:, W-:, W_: - word      (underscore: no need to highlight)
514     *   P+:, P-:      - phrase    (minus sign: logically in NOT group)
515     *   N+:, N-:      - namespace
516     */
517    $parsed_query = '';
518    $parens_level = 0;
519    $terms = preg_split('/(-?".*?")/u', utf8_strtolower($query), -1, PREG_SPLIT_DELIM_CAPTURE | PREG_SPLIT_NO_EMPTY);
520
521    foreach ($terms as $term) {
522        $parsed = '';
523        if (preg_match('/^(-?)"(.+)"$/u', $term, $matches)) {
524            // phrase-include and phrase-exclude
525            $not = $matches[1] ? 'NOT' : '';
526            $parsed = $not.ft_termParser($Indexer, $matches[2], false, true);
527        } else {
528            // fix incomplete phrase
529            $term = str_replace('"', ' ', $term);
530
531            // fix parentheses
532            $term = str_replace(')'  , ' ) ', $term);
533            $term = str_replace('('  , ' ( ', $term);
534            $term = str_replace('- (', ' -(', $term);
535
536            // treat pipe symbols as 'OR' operators
537            $term = str_replace('|', ' or ', $term);
538
539            // treat ideographic spaces (U+3000) as search term separators
540            // FIXME: some more separators?
541            $term = preg_replace('/[ \x{3000}]+/u', ' ',  $term);
542            $term = trim($term);
543            if ($term === '') continue;
544
545            $tokens = explode(' ', $term);
546            foreach ($tokens as $token) {
547                if ($token === '(') {
548                    // parenthesis-include-open
549                    $parsed .= '(';
550                    ++$parens_level;
551                } elseif ($token === '-(') {
552                    // parenthesis-exclude-open
553                    $parsed .= 'NOT(';
554                    ++$parens_level;
555                } elseif ($token === ')') {
556                    // parenthesis-any-close
557                    if ($parens_level === 0) continue;
558                    $parsed .= ')';
559                    $parens_level--;
560                } elseif ($token === 'and') {
561                    // logical-and (do nothing)
562                } elseif ($token === 'or') {
563                    // logical-or
564                    $parsed .= 'OR';
565                } elseif (preg_match('/^(?:\^|-ns:)(.+)$/u', $token, $matches)) {
566                    // namespace-exclude
567                    $parsed .= 'NOT(N+:'.$matches[1].')';
568                } elseif (preg_match('/^(?:@|ns:)(.+)$/u', $token, $matches)) {
569                    // namespace-include
570                    $parsed .= '(N+:'.$matches[1].')';
571                } elseif (preg_match('/^-(.+)$/', $token, $matches)) {
572                    // word-exclude
573                    $parsed .= 'NOT('.ft_termParser($Indexer, $matches[1]).')';
574                } else {
575                    // word-include
576                    $parsed .= ft_termParser($Indexer, $token);
577                }
578            }
579        }
580        $parsed_query .= $parsed;
581    }
582
583    // cleanup (very sensitive)
584    $parsed_query .= str_repeat(')', $parens_level);
585    do {
586        $parsed_query_old = $parsed_query;
587        $parsed_query = preg_replace('/(NOT)?\(\)/u', '', $parsed_query);
588    } while ($parsed_query !== $parsed_query_old);
589    $parsed_query = preg_replace('/(NOT|OR)+\)/u', ')'      , $parsed_query);
590    $parsed_query = preg_replace('/(OR)+/u'      , 'OR'     , $parsed_query);
591    $parsed_query = preg_replace('/\(OR/u'       , '('      , $parsed_query);
592    $parsed_query = preg_replace('/^OR|OR$/u'    , ''       , $parsed_query);
593    $parsed_query = preg_replace('/\)(NOT)?\(/u' , ')AND$1(', $parsed_query);
594
595    // adjustment: make highlightings right
596    $parens_level     = 0;
597    $notgrp_levels    = array();
598    $parsed_query_new = '';
599    $tokens = preg_split('/(NOT\(|[()])/u', $parsed_query, -1, PREG_SPLIT_DELIM_CAPTURE | PREG_SPLIT_NO_EMPTY);
600    foreach ($tokens as $token) {
601        if ($token === 'NOT(') {
602            $notgrp_levels[] = ++$parens_level;
603        } elseif ($token === '(') {
604            ++$parens_level;
605        } elseif ($token === ')') {
606            if ($parens_level-- === end($notgrp_levels)) array_pop($notgrp_levels);
607        } elseif (count($notgrp_levels) % 2 === 1) {
608            // turn highlight-flag off if terms are logically in "NOT" group
609            $token = preg_replace('/([WPN])\+\:/u', '$1-:', $token);
610        }
611        $parsed_query_new .= $token;
612    }
613    $parsed_query = $parsed_query_new;
614
615    /**
616     * convert infix notation string into postfix (Reverse Polish notation) array
617     * by Shunting-yard algorithm
618     *
619     * see: http://en.wikipedia.org/wiki/Reverse_Polish_notation
620     * see: http://en.wikipedia.org/wiki/Shunting-yard_algorithm
621     */
622    $parsed_ary     = array();
623    $ope_stack      = array();
624    $ope_precedence = array(')' => 1, 'OR' => 2, 'AND' => 3, 'NOT' => 4, '(' => 5);
625    $ope_regex      = '/([()]|OR|AND|NOT)/u';
626
627    $tokens = preg_split($ope_regex, $parsed_query, -1, PREG_SPLIT_DELIM_CAPTURE | PREG_SPLIT_NO_EMPTY);
628    foreach ($tokens as $token) {
629        if (preg_match($ope_regex, $token)) {
630            // operator
631            $last_ope = end($ope_stack);
632            while ($ope_precedence[$token] <= $ope_precedence[$last_ope] && $last_ope != '(') {
633                $parsed_ary[] = array_pop($ope_stack);
634                $last_ope = end($ope_stack);
635            }
636            if ($token == ')') {
637                array_pop($ope_stack); // this array_pop always deletes '('
638            } else {
639                $ope_stack[] = $token;
640            }
641        } else {
642            // operand
643            $token_decoded = str_replace(array('OP', 'CP'), array('(', ')'), $token);
644            $parsed_ary[] = $token_decoded;
645        }
646    }
647    $parsed_ary = array_values(array_merge($parsed_ary, array_reverse($ope_stack)));
648
649    // cleanup: each double "NOT" in RPN array actually does nothing
650    $parsed_ary_count = count($parsed_ary);
651    for ($i = 1; $i < $parsed_ary_count; ++$i) {
652        if ($parsed_ary[$i] === 'NOT' && $parsed_ary[$i - 1] === 'NOT') {
653            unset($parsed_ary[$i], $parsed_ary[$i - 1]);
654        }
655    }
656    $parsed_ary = array_values($parsed_ary);
657
658    // build return value
659    $q = array();
660    $q['query']      = $query;
661    $q['parsed_str'] = $parsed_query;
662    $q['parsed_ary'] = $parsed_ary;
663
664    foreach ($q['parsed_ary'] as $token) {
665        if ($token[2] !== ':') continue;
666        $body = substr($token, 3);
667
668        switch (substr($token, 0, 3)) {
669            case 'N+:':
670                     $q['ns'][]        = $body; // for backward compatibility
671                     break;
672            case 'N-:':
673                     $q['notns'][]     = $body; // for backward compatibility
674                     break;
675            case 'W_:':
676                     $q['words'][]     = $body;
677                     break;
678            case 'W-:':
679                     $q['words'][]     = $body;
680                     $q['not'][]       = $body; // for backward compatibility
681                     break;
682            case 'W+:':
683                     $q['words'][]     = $body;
684                     $q['highlight'][] = $body;
685                     $q['and'][]       = $body; // for backward compatibility
686                     break;
687            case 'P-:':
688                     $q['phrases'][]   = $body;
689                     break;
690            case 'P+:':
691                     $q['phrases'][]   = $body;
692                     $q['highlight'][] = $body;
693                     break;
694        }
695    }
696    foreach (array('words', 'phrases', 'highlight', 'ns', 'notns', 'and', 'not') as $key) {
697        $q[$key] = empty($q[$key]) ? array() : array_values(array_unique($q[$key]));
698    }
699
700    return $q;
701}
702
703/**
704 * Transforms given search term into intermediate representation
705 *
706 * This function is used in ft_queryParser() and not for general purpose use.
707 *
708 * @author Kazutaka Miyasaka <kazmiya@gmail.com>
709 */
710function ft_termParser($Indexer, $term, $consider_asian = true, $phrase_mode = false) {
711    $parsed = '';
712    if ($consider_asian) {
713        // successive asian characters need to be searched as a phrase
714        $words = preg_split('/('.IDX_ASIAN.'+)/u', $term, -1, PREG_SPLIT_DELIM_CAPTURE | PREG_SPLIT_NO_EMPTY);
715        foreach ($words as $word) {
716            if (preg_match('/'.IDX_ASIAN.'/u', $word)) $phrase_mode = true;
717            $parsed .= ft_termParser($Indexer, $word, false, $phrase_mode);
718        }
719    } else {
720        $term_noparen = str_replace(array('(', ')'), ' ', $term);
721        $words = $Indexer->tokenizer($term_noparen, true);
722
723        // W_: no need to highlight
724        if (empty($words)) {
725            $parsed = '()'; // important: do not remove
726        } elseif ($words[0] === $term) {
727            $parsed = '(W+:'.$words[0].')';
728        } elseif ($phrase_mode) {
729            $term_encoded = str_replace(array('(', ')'), array('OP', 'CP'), $term);
730            $parsed = '((W_:'.implode(')(W_:', $words).')(P+:'.$term_encoded.'))';
731        } else {
732            $parsed = '((W+:'.implode(')(W+:', $words).'))';
733        }
734    }
735    return $parsed;
736}
737
738//Setup VIM: ex: et ts=4 enc=utf-8 :
739