xref: /dokuwiki/inc/fulltext.php (revision 159ef5962053a685e348239dfa0b3648ed516962)
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    // parse the given query
40    $q = ft_queryParser($data['query']);
41    $data['highlight'] = $q['highlight'];
42
43    if (empty($q['parsed_ary'])) return array();
44
45    // lookup all words found in the query
46    $lookup = idx_lookup($q['words']);
47
48    // get all pages in this dokuwiki site (!: includes nonexistent pages)
49    $pages_all = array();
50    foreach (idx_getIndex('page', '') as $id) {
51        $pages_all[trim($id)] = 0; // base: 0 hit
52    }
53
54    // process the query
55    $stack = array();
56    foreach ($q['parsed_ary'] as $token) {
57        switch (substr($token, 0, 3)) {
58            case 'W+:':
59            case 'W-:':
60            case 'W_:': // word
61                $word    = substr($token, 3);
62                $stack[] = (array) $lookup[$word];
63                break;
64            case 'P+:':
65            case 'P-:': // phrase
66                $phrase = substr($token, 3);
67                // since phrases are always parsed as ((W1)(W2)...(P)),
68                // the end($stack) always points the pages that contain
69                // all words in this phrase
70                $pages  = end($stack);
71                $pages_matched = array();
72                foreach(array_keys($pages) as $id){
73                    $text = utf8_strtolower(rawWiki($id));
74                    if (strpos($text, $phrase) !== false) {
75                        $pages_matched[$id] = 0; // phrase: always 0 hit
76                    }
77                }
78                $stack[] = $pages_matched;
79                break;
80            case 'N+:':
81            case 'N-:': // namespace
82                $ns = substr($token, 3);
83                $pages_matched = array();
84                foreach (array_keys($pages_all) as $id) {
85                    if (strpos($id, $ns) === 0) {
86                        $pages_matched[$id] = 0; // namespace: always 0 hit
87                    }
88                }
89                $stack[] = $pages_matched;
90                break;
91            case 'AND': // and operation
92                list($pages1, $pages2) = array_splice($stack, -2);
93                $stack[] = ft_resultCombine(array($pages1, $pages2));
94                break;
95            case 'OR':  // or operation
96                list($pages1, $pages2) = array_splice($stack, -2);
97                $stack[] = ft_resultUnite(array($pages1, $pages2));
98                break;
99            case 'NOT': // not operation (unary)
100                $pages   = array_pop($stack);
101                $stack[] = ft_resultComplement(array($pages_all, $pages));
102                break;
103        }
104    }
105    $docs = array_pop($stack);
106
107    if (empty($docs)) return array();
108
109    // check: settings, acls, existence
110    foreach (array_keys($docs) as $id) {
111        if (isHiddenPage($id) || auth_quickaclcheck($id) < AUTH_READ || !page_exists($id, '', false)) {
112            unset($docs[$id]);
113        }
114    }
115
116    // sort docs by count
117    arsort($docs);
118
119    return $docs;
120}
121
122/**
123 * Returns the backlinks for a given page
124 *
125 * Does a quick lookup with the fulltext index, then
126 * evaluates the instructions of the found pages
127 */
128function ft_backlinks($id){
129    global $conf;
130    $swfile   = DOKU_INC.'inc/lang/'.$conf['lang'].'/stopwords.txt';
131    $stopwords = @file_exists($swfile) ? file($swfile) : array();
132
133    $result = array();
134
135    // quick lookup of the pagename
136    $page    = noNS($id);
137    $matches = idx_lookup(idx_tokenizer($page,$stopwords));  // pagename may contain specials (_ or .)
138    $docs    = array_keys(ft_resultCombine(array_values($matches)));
139    $docs    = array_filter($docs,'isVisiblePage'); // discard hidden pages
140    if(!count($docs)) return $result;
141
142    // check metadata for matching links
143    foreach($docs as $match){
144        // metadata relation reference links are already resolved
145        $links = p_get_metadata($match,'relation references');
146        if (isset($links[$id])) $result[] = $match;
147    }
148
149    if(!count($result)) return $result;
150
151    // check ACL permissions
152    foreach(array_keys($result) as $idx){
153        if(auth_quickaclcheck($result[$idx]) < AUTH_READ){
154            unset($result[$idx]);
155        }
156    }
157
158    sort($result);
159    return $result;
160}
161
162/**
163 * Returns the pages that use a given media file
164 *
165 * Does a quick lookup with the fulltext index, then
166 * evaluates the instructions of the found pages
167 *
168 * Aborts after $max found results
169 */
170function ft_mediause($id,$max){
171    global $conf;
172    $swfile   = DOKU_INC.'inc/lang/'.$conf['lang'].'/stopwords.txt';
173    $stopwords = @file_exists($swfile) ? file($swfile) : array();
174
175    if(!$max) $max = 1; // need to find at least one
176
177    $result = array();
178
179    // quick lookup of the mediafile
180    $media   = noNS($id);
181    $matches = idx_lookup(idx_tokenizer($media,$stopwords));
182    $docs    = array_keys(ft_resultCombine(array_values($matches)));
183    if(!count($docs)) return $result;
184
185    // go through all found pages
186    $found = 0;
187    $pcre  = preg_quote($media,'/');
188    foreach($docs as $doc){
189        $ns = getNS($doc);
190        preg_match_all('/\{\{([^|}]*'.$pcre.'[^|}]*)(|[^}]+)?\}\}/i',rawWiki($doc),$matches);
191        foreach($matches[1] as $img){
192            $img = trim($img);
193            if(preg_match('/^https?:\/\//i',$img)) continue; // skip external images
194                list($img) = explode('?',$img);                  // remove any parameters
195            resolve_mediaid($ns,$img,$exists);               // resolve the possibly relative img
196
197            if($img == $id){                                 // we have a match
198                $result[] = $doc;
199                $found++;
200                break;
201            }
202        }
203        if($found >= $max) break;
204    }
205
206    sort($result);
207    return $result;
208}
209
210
211
212/**
213 * Quicksearch for pagenames
214 *
215 * By default it only matches the pagename and ignores the
216 * namespace. This can be changed with the second parameter.
217 * The third parameter allows to search in titles as well.
218 * If the function should search in titles as well, the return array
219 * has the ids as key and the titles as value.
220 *
221 * refactored into ft_pageLookup(), _ft_pageLookup() and trigger_event()
222 *
223 * @author Andreas Gohr <andi@splitbrain.org>
224 */
225function ft_pageLookup($id, $not_in_ns=true, $not_in_title=true){
226    $data = compact('id', 'not_in_ns', 'not_in_title');
227    return trigger_event('SEARCH_QUERY_PAGELOOKUP', $data, '_ft_pageLookup');
228}
229
230function _ft_pageLookup(&$data){
231    // split out original parameters
232    $id = $data['id'];
233    $in_ns = !$data['not_in_ns'];
234    $in_title = !$data['not_in_title'];
235
236    global $conf;
237    $pages  = array_map('rtrim', file($conf['indexdir'].'/page.idx'));
238    $titles = array_map('rtrim', file($conf['indexdir'].'/title.idx'));
239    $pages = array_combine($pages, $titles);
240
241    if($id !== '' && cleanID($id) !== '') {
242        $cleaned = cleanID($id);
243        $matched_pages = array();
244        foreach ($pages as $p_id => $p_title) {
245            if ((strpos($in_ns ? $p_id : noNSorNS($p_id), $cleaned) !== false) ||
246                ($in_title && stripos($p_title, $id) !== false)) {
247                $matched_pages[$p_id] = $p_title;
248            }
249        }
250        $pages = $matched_pages;
251    }
252
253    // discard hidden pages
254    // discard nonexistent pages
255    // check ACL permissions
256    foreach(array_keys($pages) as $idx){
257        if(!isVisiblePage($idx) || !page_exists($idx) ||
258           auth_quickaclcheck($idx) < AUTH_READ) {
259            unset($pages[$idx]);
260        }
261    }
262
263    uasort($pages,'ft_pagesorter');
264    return $in_title ? $pages : array_keys($pages);
265}
266
267/**
268 * Sort pages based on their namespace level first, then on their string
269 * values. This makes higher hierarchy pages rank higher than lower hierarchy
270 * pages.
271 */
272function ft_pagesorter($a, $b){
273    $ac = count(explode(':',$a));
274    $bc = count(explode(':',$b));
275    if($ac < $bc){
276        return -1;
277    }elseif($ac > $bc){
278        return 1;
279    }
280    return strcmp ($a,$b);
281}
282
283/**
284 * Creates a snippet extract
285 *
286 * @author Andreas Gohr <andi@splitbrain.org>
287 * @triggers FULLTEXT_SNIPPET_CREATE
288 */
289function ft_snippet($id,$highlight){
290    $text = rawWiki($id);
291    $evdata = array(
292            'id'        => $id,
293            'text'      => &$text,
294            'highlight' => &$highlight,
295            'snippet'   => '',
296            );
297
298    $evt = new Doku_Event('FULLTEXT_SNIPPET_CREATE',$evdata);
299    if ($evt->advise_before()) {
300        $match = array();
301        $snippets = array();
302        $utf8_offset = $offset = $end = 0;
303        $len = utf8_strlen($text);
304
305        // build a regexp from the phrases to highlight
306        $re1 = '('.join('|',array_map('preg_quote_cb',array_filter((array) $highlight))).')';
307        $re2 = "$re1.{0,75}(?!\\1)$re1";
308        $re3 = "$re1.{0,45}(?!\\1)$re1.{0,45}(?!\\1)(?!\\2)$re1";
309
310        for ($cnt=4; $cnt--;) {
311            if (0) {
312            } else if (preg_match('/'.$re3.'/iu',$text,$match,PREG_OFFSET_CAPTURE,$offset)) {
313            } else if (preg_match('/'.$re2.'/iu',$text,$match,PREG_OFFSET_CAPTURE,$offset)) {
314            } else if (preg_match('/'.$re1.'/iu',$text,$match,PREG_OFFSET_CAPTURE,$offset)) {
315            } else {
316                break;
317            }
318
319            list($str,$idx) = $match[0];
320
321            // convert $idx (a byte offset) into a utf8 character offset
322            $utf8_idx = utf8_strlen(substr($text,0,$idx));
323            $utf8_len = utf8_strlen($str);
324
325            // establish context, 100 bytes surrounding the match string
326            // first look to see if we can go 100 either side,
327            // then drop to 50 adding any excess if the other side can't go to 50,
328            $pre = min($utf8_idx-$utf8_offset,100);
329            $post = min($len-$utf8_idx-$utf8_len,100);
330
331            if ($pre>50 && $post>50) {
332                $pre = $post = 50;
333            } else if ($pre>50) {
334                $pre = min($pre,100-$post);
335            } else if ($post>50) {
336                $post = min($post, 100-$pre);
337            } else {
338                // both are less than 50, means the context is the whole string
339                // make it so and break out of this loop - there is no need for the
340                // complex snippet calculations
341                $snippets = array($text);
342                break;
343            }
344
345            // establish context start and end points, try to append to previous
346            // context if possible
347            $start = $utf8_idx - $pre;
348            $append = ($start < $end) ? $end : false;  // still the end of the previous context snippet
349            $end = $utf8_idx + $utf8_len + $post;      // now set it to the end of this context
350
351            if ($append) {
352                $snippets[count($snippets)-1] .= utf8_substr($text,$append,$end-$append);
353            } else {
354                $snippets[] = utf8_substr($text,$start,$end-$start);
355            }
356
357            // set $offset for next match attempt
358            //   substract strlen to avoid splitting a potential search success,
359            //   this is an approximation as the search pattern may match strings
360            //   of varying length and it will fail if the context snippet
361            //   boundary breaks a matching string longer than the current match
362            $utf8_offset = $utf8_idx + $post;
363            $offset = $idx + strlen(utf8_substr($text,$utf8_idx,$post));
364            $offset = utf8_correctIdx($text,$offset);
365        }
366
367        $m = "\1";
368        $snippets = preg_replace('/'.$re1.'/iu',$m.'$1'.$m,$snippets);
369        $snippet = preg_replace('/'.$m.'([^'.$m.']*?)'.$m.'/iu','<strong class="search_hit">$1</strong>',hsc(join('... ',$snippets)));
370
371        $evdata['snippet'] = $snippet;
372    }
373    $evt->advise_after();
374    unset($evt);
375
376    return $evdata['snippet'];
377}
378
379/**
380 * Combine found documents and sum up their scores
381 *
382 * This function is used to combine searched words with a logical
383 * AND. Only documents available in all arrays are returned.
384 *
385 * based upon PEAR's PHP_Compat function for array_intersect_key()
386 *
387 * @param array $args An array of page arrays
388 */
389function ft_resultCombine($args){
390    $array_count = count($args);
391    if($array_count == 1){
392        return $args[0];
393    }
394
395    $result = array();
396    if ($array_count > 1) {
397        foreach ($args[0] as $key => $value) {
398            $result[$key] = $value;
399            for ($i = 1; $i !== $array_count; $i++) {
400                if (!isset($args[$i][$key])) {
401                    unset($result[$key]);
402                    break;
403                }
404                $result[$key] += $args[$i][$key];
405            }
406        }
407    }
408    return $result;
409}
410
411/**
412 * Unites found documents and sum up their scores
413 *
414 * based upon ft_resultCombine() function
415 *
416 * @param array $args An array of page arrays
417 * @author Kazutaka Miyasaka <kazmiya@gmail.com>
418 */
419function ft_resultUnite($args) {
420    $array_count = count($args);
421    if ($array_count === 1) {
422        return $args[0];
423    }
424
425    $result = $args[0];
426    for ($i = 1; $i !== $array_count; $i++) {
427        foreach (array_keys($args[$i]) as $id) {
428            $result[$id] += $args[$i][$id];
429        }
430    }
431    return $result;
432}
433
434/**
435 * Computes the difference of documents using page id for comparison
436 *
437 * nearly identical to PHP5's array_diff_key()
438 *
439 * @param array $args An array of page arrays
440 * @author Kazutaka Miyasaka <kazmiya@gmail.com>
441 */
442function ft_resultComplement($args) {
443    $array_count = count($args);
444    if ($array_count === 1) {
445        return $args[0];
446    }
447
448    $result = $args[0];
449    foreach (array_keys($result) as $id) {
450        for ($i = 1; $i !== $array_count; $i++) {
451            if (isset($args[$i][$id])) unset($result[$id]);
452        }
453    }
454    return $result;
455}
456
457/**
458 * Parses a search query and builds an array of search formulas
459 *
460 * @author Andreas Gohr <andi@splitbrain.org>
461 * @author Kazutaka Miyasaka <kazmiya@gmail.com>
462 */
463function ft_queryParser($query){
464    global $conf;
465    $swfile    = DOKU_INC.'inc/lang/'.$conf['lang'].'/stopwords.txt';
466    $stopwords = @file_exists($swfile) ? file($swfile) : array();
467
468    /**
469     * parse a search query and transform it into intermediate representation
470     *
471     * in a search query, you can use the following expressions:
472     *
473     *   words:
474     *     include
475     *     -exclude
476     *   phrases:
477     *     "phrase to be included"
478     *     -"phrase you want to exclude"
479     *   namespaces:
480     *     @include:namespace (or ns:include:namespace)
481     *     ^exclude:namespace (or -ns:exclude:namespace)
482     *   groups:
483     *     ()
484     *     -()
485     *   operators:
486     *     and ('and' is the default operator: you can always omit this)
487     *     or  (or pipe symbol '|', lower precedence than 'and')
488     *
489     * e.g. a query [ aa "bb cc" @dd:ee ] means "search pages which contain
490     *      a word 'aa', a phrase 'bb cc' and are within a namespace 'dd:ee'".
491     *      this query is equivalent to [ -(-aa or -"bb cc" or -ns:dd:ee) ]
492     *      as long as you don't mind hit counts.
493     *
494     * intermediate representation consists of the following parts:
495     *
496     *   ( )           - group
497     *   AND           - logical and
498     *   OR            - logical or
499     *   NOT           - logical not
500     *   W+:, W-:, W_: - word      (underscore: no need to highlight)
501     *   P+:, P-:      - phrase    (minus sign: logically in NOT group)
502     *   N+:, N-:      - namespace
503     */
504    $parsed_query = '';
505    $parens_level = 0;
506    $terms = preg_split('/(-?".*?")/u', utf8_strtolower($query), -1, PREG_SPLIT_DELIM_CAPTURE | PREG_SPLIT_NO_EMPTY);
507
508    foreach ($terms as $term) {
509        $parsed = '';
510        if (preg_match('/^(-?)"(.+)"$/u', $term, $matches)) {
511            // phrase-include and phrase-exclude
512            $not = $matches[1] ? 'NOT' : '';
513            $parsed = $not.ft_termParser($matches[2], $stopwords, false, true);
514        } else {
515            // fix incomplete phrase
516            $term = str_replace('"', ' ', $term);
517
518            // fix parentheses
519            $term = str_replace(')'  , ' ) ', $term);
520            $term = str_replace('('  , ' ( ', $term);
521            $term = str_replace('- (', ' -(', $term);
522
523            // treat pipe symbols as 'OR' operators
524            $term = str_replace('|', ' or ', $term);
525
526            // treat ideographic spaces (U+3000) as search term separators
527            // FIXME: some more separators?
528            $term = preg_replace('/[ \x{3000}]+/u', ' ',  $term);
529            $term = trim($term);
530            if ($term === '') continue;
531
532            $tokens = explode(' ', $term);
533            foreach ($tokens as $token) {
534                if ($token === '(') {
535                    // parenthesis-include-open
536                    $parsed .= '(';
537                    ++$parens_level;
538                } elseif ($token === '-(') {
539                    // parenthesis-exclude-open
540                    $parsed .= 'NOT(';
541                    ++$parens_level;
542                } elseif ($token === ')') {
543                    // parenthesis-any-close
544                    if ($parens_level === 0) continue;
545                    $parsed .= ')';
546                    $parens_level--;
547                } elseif ($token === 'and') {
548                    // logical-and (do nothing)
549                } elseif ($token === 'or') {
550                    // logical-or
551                    $parsed .= 'OR';
552                } elseif (preg_match('/^(?:\^|-ns:)(.+)$/u', $token, $matches)) {
553                    // namespace-exclude
554                    $parsed .= 'NOT(N+:'.$matches[1].')';
555                } elseif (preg_match('/^(?:@|ns:)(.+)$/u', $token, $matches)) {
556                    // namespace-include
557                    $parsed .= '(N+:'.$matches[1].')';
558                } elseif (preg_match('/^-(.+)$/', $token, $matches)) {
559                    // word-exclude
560                    $parsed .= 'NOT('.ft_termParser($matches[1], $stopwords).')';
561                } else {
562                    // word-include
563                    $parsed .= ft_termParser($token, $stopwords);
564                }
565            }
566        }
567        $parsed_query .= $parsed;
568    }
569
570    // cleanup (very sensitive)
571    $parsed_query .= str_repeat(')', $parens_level);
572    do {
573        $parsed_query_old = $parsed_query;
574        $parsed_query = preg_replace('/(NOT)?\(\)/u', '', $parsed_query);
575    } while ($parsed_query !== $parsed_query_old);
576    $parsed_query = preg_replace('/(NOT|OR)+\)/u', ')'      , $parsed_query);
577    $parsed_query = preg_replace('/(OR)+/u'      , 'OR'     , $parsed_query);
578    $parsed_query = preg_replace('/\(OR/u'       , '('      , $parsed_query);
579    $parsed_query = preg_replace('/^OR|OR$/u'    , ''       , $parsed_query);
580    $parsed_query = preg_replace('/\)(NOT)?\(/u' , ')AND$1(', $parsed_query);
581
582    // adjustment: make highlightings right
583    $parens_level     = 0;
584    $notgrp_levels    = array();
585    $parsed_query_new = '';
586    $tokens = preg_split('/(NOT\(|[()])/u', $parsed_query, -1, PREG_SPLIT_DELIM_CAPTURE | PREG_SPLIT_NO_EMPTY);
587    foreach ($tokens as $token) {
588        if ($token === 'NOT(') {
589            $notgrp_levels[] = ++$parens_level;
590        } elseif ($token === '(') {
591            ++$parens_level;
592        } elseif ($token === ')') {
593            if ($parens_level-- === end($notgrp_levels)) array_pop($notgrp_levels);
594        } elseif (count($notgrp_levels) % 2 === 1) {
595            // turn highlight-flag off if terms are logically in "NOT" group
596            $token = preg_replace('/([WPN])\+\:/u', '$1-:', $token);
597        }
598        $parsed_query_new .= $token;
599    }
600    $parsed_query = $parsed_query_new;
601
602    /**
603     * convert infix notation string into postfix (Reverse Polish notation) array
604     * by Shunting-yard algorithm
605     *
606     * see: http://en.wikipedia.org/wiki/Reverse_Polish_notation
607     * see: http://en.wikipedia.org/wiki/Shunting-yard_algorithm
608     */
609    $parsed_ary     = array();
610    $ope_stack      = array();
611    $ope_precedence = array(')' => 1, 'OR' => 2, 'AND' => 3, 'NOT' => 4, '(' => 5);
612    $ope_regex      = '/([()]|OR|AND|NOT)/u';
613
614    $tokens = preg_split($ope_regex, $parsed_query, -1, PREG_SPLIT_DELIM_CAPTURE | PREG_SPLIT_NO_EMPTY);
615    foreach ($tokens as $token) {
616        if (preg_match($ope_regex, $token)) {
617            // operator
618            $last_ope = end($ope_stack);
619            while ($ope_precedence[$token] <= $ope_precedence[$last_ope] && $last_ope != '(') {
620                $parsed_ary[] = array_pop($ope_stack);
621                $last_ope = end($ope_stack);
622            }
623            if ($token == ')') {
624                array_pop($ope_stack); // this array_pop always deletes '('
625            } else {
626                $ope_stack[] = $token;
627            }
628        } else {
629            // operand
630            $token_decoded = str_replace(array('OP', 'CP'), array('(', ')'), $token);
631            $parsed_ary[] = $token_decoded;
632        }
633    }
634    $parsed_ary = array_values(array_merge($parsed_ary, array_reverse($ope_stack)));
635
636    // cleanup: each double "NOT" in RPN array actually does nothing
637    $parsed_ary_count = count($parsed_ary);
638    for ($i = 1; $i < $parsed_ary_count; ++$i) {
639        if ($parsed_ary[$i] === 'NOT' && $parsed_ary[$i - 1] === 'NOT') {
640            unset($parsed_ary[$i], $parsed_ary[$i - 1]);
641        }
642    }
643    $parsed_ary = array_values($parsed_ary);
644
645    // build return value
646    $q = array();
647    $q['query']      = $query;
648    $q['parsed_str'] = $parsed_query;
649    $q['parsed_ary'] = $parsed_ary;
650
651    foreach ($q['parsed_ary'] as $token) {
652        if ($token[2] !== ':') continue;
653        $body = substr($token, 3);
654
655        switch (substr($token, 0, 3)) {
656            case 'N+:':
657                     $q['ns'][]        = $body; // for backward compatibility
658                     break;
659            case 'N-:':
660                     $q['notns'][]     = $body; // for backward compatibility
661                     break;
662            case 'W_:':
663                     $q['words'][]     = $body;
664                     break;
665            case 'W-:':
666                     $q['words'][]     = $body;
667                     $q['not'][]       = $body; // for backward compatibility
668                     break;
669            case 'W+:':
670                     $q['words'][]     = $body;
671                     $q['highlight'][] = str_replace('*', '', $body);
672                     $q['and'][]       = $body; // for backward compatibility
673                     break;
674            case 'P-:':
675                     $q['phrases'][]   = $body;
676                     break;
677            case 'P+:':
678                     $q['phrases'][]   = $body;
679                     $q['highlight'][] = str_replace('*', '', $body);
680                     break;
681        }
682    }
683    foreach (array('words', 'phrases', 'highlight', 'ns', 'notns', 'and', 'not') as $key) {
684        $q[$key] = empty($q[$key]) ? array() : array_values(array_unique($q[$key]));
685    }
686
687    return $q;
688}
689
690/**
691 * Transforms given search term into intermediate representation
692 *
693 * This function is used in ft_queryParser() and not for general purpose use.
694 *
695 * @author Kazutaka Miyasaka <kazmiya@gmail.com>
696 */
697function ft_termParser($term, &$stopwords, $consider_asian = true, $phrase_mode = false) {
698    $parsed = '';
699    if ($consider_asian) {
700        // successive asian characters need to be searched as a phrase
701        $words = preg_split('/('.IDX_ASIAN.'+)/u', $term, -1, PREG_SPLIT_DELIM_CAPTURE | PREG_SPLIT_NO_EMPTY);
702        foreach ($words as $word) {
703            if (preg_match('/'.IDX_ASIAN.'/u', $word)) $phrase_mode = true;
704            $parsed .= ft_termParser($word, $stopwords, false, $phrase_mode);
705        }
706    } else {
707        $term_noparen = str_replace(array('(', ')'), ' ', $term);
708        $words = idx_tokenizer($term_noparen, $stopwords, true);
709
710        // W_: no need to highlight
711        if (empty($words)) {
712            $parsed = '()'; // important: do not remove
713        } elseif ($words[0] === $term) {
714            $parsed = '(W+:'.$words[0].')';
715        } elseif ($phrase_mode) {
716            $term_encoded = str_replace(array('(', ')'), array('OP', 'CP'), $term);
717            $parsed = '((W_:'.implode(')(W_:', $words).')(P+:'.$term_encoded.'))';
718        } else {
719            $parsed = '((W+:'.implode(')(W+:', $words).'))';
720        }
721    }
722    return $parsed;
723}
724
725//Setup VIM: ex: et ts=4 enc=utf-8 :
726