xref: /dokuwiki/inc/fulltext.php (revision c2a6d81662045023bdf1617b6b49f71c274d55ca) 
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 * The fulltext search
13 *
14 * Returns a list of matching documents for the given query
15 *
16 * refactored into ft_pageSearch(), _ft_pageSearch() and trigger_event()
17 *
18 */
19function ft_pageSearch($query,&$highlight){
20
21    $data['query'] = $query;
22    $data['highlight'] =& $highlight;
23
24    return trigger_event('SEARCH_QUERY_FULLPAGE', $data, '_ft_pageSearch');
25}
26
27/**
28 * Returns a list of matching documents for the given query
29 *
30 * @author Andreas Gohr <andi@splitbrain.org>
31 * @author Kazutaka Miyasaka <kazmiya@gmail.com>
32 */
33function _ft_pageSearch(&$data) {
34    // parse the given query
35    $q = ft_queryParser($data['query']);
36    $data['highlight'] = $q['highlight'];
37
38    if (empty($q['parsed_ary'])) return array();
39
40    // lookup all words found in the query
41    $lookup = idx_lookup($q['words']);
42
43    // get all pages in this dokuwiki site (!: includes nonexistent pages)
44    $pages_all = array();
45    foreach (idx_getIndex('page', '') as $id) {
46        $pages_all[trim($id)] = 0; // base: 0 hit
47    }
48
49    // process the query
50    $stack = array();
51    foreach ($q['parsed_ary'] as $token) {
52        switch (substr($token, 0, 3)) {
53            case 'W+:':
54            case 'W-:':
55            case 'W_:': // word
56                $word    = substr($token, 3);
57                $stack[] = (array) $lookup[$word];
58                break;
59            case 'P+:':
60            case 'P-:': // phrase
61                $phrase = substr($token, 3);
62                // since phrases are always parsed as ((W1)(W2)...(P)),
63                // the end($stack) always points the pages that contain
64                // all words in this phrase
65                $pages  = end($stack);
66                $pages_matched = array();
67                foreach(array_keys($pages) as $id){
68                    $text = utf8_strtolower(rawWiki($id));
69                    if (strpos($text, $phrase) !== false) {
70                        $pages_matched[$id] = 0; // phrase: always 0 hit
71                    }
72                }
73                $stack[] = $pages_matched;
74                break;
75            case 'N+:':
76            case 'N-:': // namespace
77                $ns = substr($token, 3);
78                $pages_matched = array();
79                foreach (array_keys($pages_all) as $id) {
80                    if (strpos($id, $ns) === 0) {
81                        $pages_matched[$id] = 0; // namespace: always 0 hit
82                    }
83                }
84                $stack[] = $pages_matched;
85                break;
86            case 'AND': // and operation
87                list($pages1, $pages2) = array_splice($stack, -2);
88                $stack[] = ft_resultCombine(array($pages1, $pages2));
89                break;
90            case 'OR':  // or operation
91                list($pages1, $pages2) = array_splice($stack, -2);
92                $stack[] = ft_resultUnite(array($pages1, $pages2));
93                break;
94            case 'NOT': // not operation (unary)
95                $pages   = array_pop($stack);
96                $stack[] = ft_resultComplement(array($pages_all, $pages));
97                break;
98        }
99    }
100    $docs = array_pop($stack);
101
102    if (empty($docs)) return array();
103
104    // check: settings, acls, existence
105    foreach (array_keys($docs) as $id) {
106        if (isHiddenPage($id) || auth_quickaclcheck($id) < AUTH_READ || !page_exists($id, '', false)) {
107            unset($docs[$id]);
108        }
109    }
110
111    // sort docs by count
112    arsort($docs);
113
114    return $docs;
115}
116
117/**
118 * Returns the backlinks for a given page
119 *
120 * Does a quick lookup with the fulltext index, then
121 * evaluates the instructions of the found pages
122 */
123function ft_backlinks($id){
124    global $conf;
125    $swfile   = DOKU_INC.'inc/lang/'.$conf['lang'].'/stopwords.txt';
126    $stopwords = @file_exists($swfile) ? file($swfile) : array();
127
128    $result = array();
129
130    // quick lookup of the pagename
131    $page    = noNS($id);
132    $matches = idx_lookup(idx_tokenizer($page,$stopwords));  // pagename may contain specials (_ or .)
133    $docs    = array_keys(ft_resultCombine(array_values($matches)));
134    $docs    = array_filter($docs,'isVisiblePage'); // discard hidden pages
135    if(!count($docs)) return $result;
136
137    // check metadata for matching links
138    foreach($docs as $match){
139        // metadata relation reference links are already resolved
140        $links = p_get_metadata($match,'relation references');
141        if (isset($links[$id])) $result[] = $match;
142    }
143
144    if(!count($result)) return $result;
145
146    // check ACL permissions
147    foreach(array_keys($result) as $idx){
148        if(auth_quickaclcheck($result[$idx]) < AUTH_READ){
149            unset($result[$idx]);
150        }
151    }
152
153    sort($result);
154    return $result;
155}
156
157/**
158 * Returns the pages that use a given media file
159 *
160 * Does a quick lookup with the fulltext index, then
161 * evaluates the instructions of the found pages
162 *
163 * Aborts after $max found results
164 */
165function ft_mediause($id,$max){
166    global $conf;
167    $swfile   = DOKU_INC.'inc/lang/'.$conf['lang'].'/stopwords.txt';
168    $stopwords = @file_exists($swfile) ? file($swfile) : array();
169
170    if(!$max) $max = 1; // need to find at least one
171
172    $result = array();
173
174    // quick lookup of the mediafile
175    $media   = noNS($id);
176    $matches = idx_lookup(idx_tokenizer($media,$stopwords));
177    $docs    = array_keys(ft_resultCombine(array_values($matches)));
178    if(!count($docs)) return $result;
179
180    // go through all found pages
181    $found = 0;
182    $pcre  = preg_quote($media,'/');
183    foreach($docs as $doc){
184        $ns = getNS($doc);
185        preg_match_all('/\{\{([^|}]*'.$pcre.'[^|}]*)(|[^}]+)?\}\}/i',rawWiki($doc),$matches);
186        foreach($matches[1] as $img){
187            $img = trim($img);
188            if(preg_match('/^https?:\/\//i',$img)) continue; // skip external images
189                list($img) = explode('?',$img);                  // remove any parameters
190            resolve_mediaid($ns,$img,$exists);               // resolve the possibly relative img
191
192            if($img == $id){                                 // we have a match
193                $result[] = $doc;
194                $found++;
195                break;
196            }
197        }
198        if($found >= $max) break;
199    }
200
201    sort($result);
202    return $result;
203}
204
205
206
207/**
208 * Quicksearch for pagenames
209 *
210 * By default it only matches the pagename and ignores the
211 * namespace. This can be changed with the second parameter
212 *
213 * refactored into ft_pageLookup(), _ft_pageLookup() and trigger_event()
214 *
215 * @author Andreas Gohr <andi@splitbrain.org>
216 */
217function ft_pageLookup($id,$pageonly=true){
218    $data = array('id' => $id, 'pageonly' => $pageonly);
219    return trigger_event('SEARCH_QUERY_PAGELOOKUP',$data,'_ft_pageLookup');
220}
221
222function _ft_pageLookup(&$data){
223    // split out original parameterrs
224    $id = $data['id'];
225    $pageonly = $data['pageonly'];
226
227    global $conf;
228    $id    = preg_quote($id,'/');
229    $pages = file($conf['indexdir'].'/page.idx');
230    if($id) $pages = array_values(preg_grep('/'.$id.'/',$pages));
231
232    $cnt = count($pages);
233    for($i=0; $i<$cnt; $i++){
234        if($pageonly){
235            if(!preg_match('/'.$id.'/',noNS($pages[$i]))){
236                unset($pages[$i]);
237                continue;
238            }
239        }
240        if(!page_exists($pages[$i])){
241            unset($pages[$i]);
242            continue;
243        }
244    }
245
246    $pages = array_filter($pages,'isVisiblePage'); // discard hidden pages
247    if(!count($pages)) return array();
248
249    // check ACL permissions
250    foreach(array_keys($pages) as $idx){
251        if(auth_quickaclcheck(trim($pages[$idx])) < AUTH_READ){
252            unset($pages[$idx]);
253        }
254    }
255
256    $pages = array_map('trim',$pages);
257    usort($pages,'ft_pagesorter');
258    return $pages;
259}
260
261/**
262 * Sort pages based on their namespace level first, then on their string
263 * values. This makes higher hierarchy pages rank higher than lower hierarchy
264 * pages.
265 */
266function ft_pagesorter($a, $b){
267    $ac = count(explode(':',$a));
268    $bc = count(explode(':',$b));
269    if($ac < $bc){
270        return -1;
271    }elseif($ac > $bc){
272        return 1;
273    }
274    return strcmp ($a,$b);
275}
276
277/**
278 * Creates a snippet extract
279 *
280 * @author Andreas Gohr <andi@splitbrain.org>
281 * @triggers FULLTEXT_SNIPPET_CREATE
282 */
283function ft_snippet($id,$highlight){
284    $text = rawWiki($id);
285    $evdata = array(
286            'id'        => $id,
287            'text'      => &$text,
288            'highlight' => &$highlight,
289            'snippet'   => '',
290            );
291
292    $evt = new Doku_Event('FULLTEXT_SNIPPET_CREATE',$evdata);
293    if ($evt->advise_before()) {
294        $match = array();
295        $snippets = array();
296        $utf8_offset = $offset = $end = 0;
297        $len = utf8_strlen($text);
298
299        // build a regexp from the phrases to highlight
300        $re1 = '('.join('|',array_map('preg_quote_cb',array_filter((array) $highlight))).')';
301        $re2 = "$re1.{0,75}(?!\\1)$re1";
302        $re3 = "$re1.{0,45}(?!\\1)$re1.{0,45}(?!\\1)(?!\\2)$re1";
303
304        for ($cnt=4; $cnt--;) {
305            if (0) {
306            } else if (preg_match('/'.$re3.'/iu',$text,$match,PREG_OFFSET_CAPTURE,$offset)) {
307            } else if (preg_match('/'.$re2.'/iu',$text,$match,PREG_OFFSET_CAPTURE,$offset)) {
308            } else if (preg_match('/'.$re1.'/iu',$text,$match,PREG_OFFSET_CAPTURE,$offset)) {
309            } else {
310                break;
311            }
312
313            list($str,$idx) = $match[0];
314
315            // convert $idx (a byte offset) into a utf8 character offset
316            $utf8_idx = utf8_strlen(substr($text,0,$idx));
317            $utf8_len = utf8_strlen($str);
318
319            // establish context, 100 bytes surrounding the match string
320            // first look to see if we can go 100 either side,
321            // then drop to 50 adding any excess if the other side can't go to 50,
322            $pre = min($utf8_idx-$utf8_offset,100);
323            $post = min($len-$utf8_idx-$utf8_len,100);
324
325            if ($pre>50 && $post>50) {
326                $pre = $post = 50;
327            } else if ($pre>50) {
328                $pre = min($pre,100-$post);
329            } else if ($post>50) {
330                $post = min($post, 100-$pre);
331            } else {
332                // both are less than 50, means the context is the whole string
333                // make it so and break out of this loop - there is no need for the
334                // complex snippet calculations
335                $snippets = array($text);
336                break;
337            }
338
339            // establish context start and end points, try to append to previous
340            // context if possible
341            $start = $utf8_idx - $pre;
342            $append = ($start < $end) ? $end : false;  // still the end of the previous context snippet
343            $end = $utf8_idx + $utf8_len + $post;      // now set it to the end of this context
344
345            if ($append) {
346                $snippets[count($snippets)-1] .= utf8_substr($text,$append,$end-$append);
347            } else {
348                $snippets[] = utf8_substr($text,$start,$end-$start);
349            }
350
351            // set $offset for next match attempt
352            //   substract strlen to avoid splitting a potential search success,
353            //   this is an approximation as the search pattern may match strings
354            //   of varying length and it will fail if the context snippet
355            //   boundary breaks a matching string longer than the current match
356            $utf8_offset = $utf8_idx + $post;
357            $offset = $idx + strlen(utf8_substr($text,$utf8_idx,$post));
358            $offset = utf8_correctIdx($text,$offset);
359        }
360
361        $m = "\1";
362        $snippets = preg_replace('/'.$re1.'/iu',$m.'$1'.$m,$snippets);
363        $snippet = preg_replace('/'.$m.'([^'.$m.']*?)'.$m.'/iu','<strong class="search_hit">$1</strong>',hsc(join('... ',$snippets)));
364
365        $evdata['snippet'] = $snippet;
366    }
367    $evt->advise_after();
368    unset($evt);
369
370    return $evdata['snippet'];
371}
372
373/**
374 * Combine found documents and sum up their scores
375 *
376 * This function is used to combine searched words with a logical
377 * AND. Only documents available in all arrays are returned.
378 *
379 * based upon PEAR's PHP_Compat function for array_intersect_key()
380 *
381 * @param array $args An array of page arrays
382 */
383function ft_resultCombine($args){
384    $array_count = count($args);
385    if($array_count == 1){
386        return $args[0];
387    }
388
389    $result = array();
390    if ($array_count > 1) {
391        foreach ($args[0] as $key => $value) {
392            $result[$key] = $value;
393            for ($i = 1; $i !== $array_count; $i++) {
394                if (!isset($args[$i][$key])) {
395                    unset($result[$key]);
396                    break;
397                }
398                $result[$key] += $args[$i][$key];
399            }
400        }
401    }
402    return $result;
403}
404
405/**
406 * Unites found documents and sum up their scores
407 *
408 * based upon ft_resultCombine() function
409 *
410 * @param array $args An array of page arrays
411 * @author Kazutaka Miyasaka <kazmiya@gmail.com>
412 */
413function ft_resultUnite($args) {
414    $array_count = count($args);
415    if ($array_count === 1) {
416        return $args[0];
417    }
418
419    $result = $args[0];
420    for ($i = 1; $i !== $array_count; $i++) {
421        foreach (array_keys($args[$i]) as $id) {
422            $result[$id] += $args[$i][$id];
423        }
424    }
425    return $result;
426}
427
428/**
429 * Computes the difference of documents using page id for comparison
430 *
431 * nearly identical to PHP5's array_diff_key()
432 *
433 * @param array $args An array of page arrays
434 * @author Kazutaka Miyasaka <kazmiya@gmail.com>
435 */
436function ft_resultComplement($args) {
437    $array_count = count($args);
438    if ($array_count === 1) {
439        return $args[0];
440    }
441
442    $result = $args[0];
443    foreach (array_keys($result) as $id) {
444        for ($i = 1; $i !== $array_count; $i++) {
445            if (isset($args[$i][$id])) unset($result[$id]);
446        }
447    }
448    return $result;
449}
450
451/**
452 * Parses a search query and builds an array of search formulas
453 *
454 * @author Andreas Gohr <andi@splitbrain.org>
455 * @author Kazutaka Miyasaka <kazmiya@gmail.com>
456 */
457function ft_queryParser($query){
458    global $conf;
459    $swfile    = DOKU_INC.'inc/lang/'.$conf['lang'].'/stopwords.txt';
460    $stopwords = @file_exists($swfile) ? file($swfile) : array();
461
462    /**
463     * parse a search query and transform it into intermediate representation
464     *
465     * in a search query, you can use the following expressions:
466     *
467     *   words:
468     *     include
469     *     -exclude
470     *   phrases:
471     *     "phrase to be included"
472     *     -"phrase you want to exclude"
473     *   namespaces:
474     *     @include:namespace (or ns:include:namespace)
475     *     ^exclude:namespace (or -ns:exclude:namespace)
476     *   groups:
477     *     ()
478     *     -()
479     *   operators:
480     *     and ('and' is the default operator: you can always omit this)
481     *     or  (or pipe symbol '|', lower precedence than 'and')
482     *
483     * e.g. a query [ aa "bb cc" @dd:ee ] means "search pages which contain
484     *      a word 'aa', a phrase 'bb cc' and are within a namespace 'dd:ee'".
485     *      this query is equivalent to [ -(-aa or -"bb cc" or -ns:dd:ee) ]
486     *      as long as you don't mind hit counts.
487     *
488     * intermediate representation consists of the following parts:
489     *
490     *   ( )           - group
491     *   AND           - logical and
492     *   OR            - logical or
493     *   NOT           - logical not
494     *   W+:, W-:, W_: - word      (underscore: no need to highlight)
495     *   P+:, P-:      - phrase    (minus sign: logically in NOT group)
496     *   N+:, N-:      - namespace
497     */
498    $parsed_query = '';
499    $parens_level = 0;
500    $terms = preg_split('/(-?".*?")/u', utf8_strtolower($query), -1, PREG_SPLIT_DELIM_CAPTURE | PREG_SPLIT_NO_EMPTY);
501
502    foreach ($terms as $term) {
503        $parsed = '';
504        if (preg_match('/^(-?)"(.+)"$/u', $term, $matches)) {
505            // phrase-include and phrase-exclude
506            $not = $matches[1] ? 'NOT' : '';
507            $parsed = $not.ft_termParser($matches[2], $stopwords, false, true);
508        } else {
509            // fix incomplete phrase
510            $term = str_replace('"', ' ', $term);
511
512            // fix parentheses
513            $term = str_replace(')'  , ' ) ', $term);
514            $term = str_replace('('  , ' ( ', $term);
515            $term = str_replace('- (', ' -(', $term);
516
517            // treat pipe symbols as 'OR' operators
518            $term = str_replace('|', ' or ', $term);
519
520            // treat ideographic spaces (U+3000) as search term separators
521            // FIXME: some more separators?
522            $term = preg_replace('/[ \x{3000}]+/u', ' ',  $term);
523            $term = trim($term);
524            if ($term === '') continue;
525
526            $tokens = explode(' ', $term);
527            foreach ($tokens as $token) {
528                if ($token === '(') {
529                    // parenthesis-include-open
530                    $parsed .= '(';
531                    ++$parens_level;
532                } elseif ($token === '-(') {
533                    // parenthesis-exclude-open
534                    $parsed .= 'NOT(';
535                    ++$parens_level;
536                } elseif ($token === ')') {
537                    // parenthesis-any-close
538                    if ($parens_level === 0) continue;
539                    $parsed .= ')';
540                    $parens_level--;
541                } elseif ($token === 'and') {
542                    // logical-and (do nothing)
543                } elseif ($token === 'or') {
544                    // logical-or
545                    $parsed .= 'OR';
546                } elseif (preg_match('/^(?:\^|-ns:)(.+)$/u', $token, $matches)) {
547                    // namespace-exclude
548                    $parsed .= 'NOT(N+:'.$matches[1].')';
549                } elseif (preg_match('/^(?:@|ns:)(.+)$/u', $token, $matches)) {
550                    // namespace-include
551                    $parsed .= '(N+:'.$matches[1].')';
552                } elseif (preg_match('/^-(.+)$/', $token, $matches)) {
553                    // word-exclude
554                    $parsed .= 'NOT('.ft_termParser($matches[1], $stopwords).')';
555                } else {
556                    // word-include
557                    $parsed .= ft_termParser($token, $stopwords);
558                }
559            }
560        }
561        $parsed_query .= $parsed;
562    }
563
564    // cleanup (very sensitive)
565    $parsed_query .= str_repeat(')', $parens_level);
566    do {
567        $parsed_query_old = $parsed_query;
568        $parsed_query = preg_replace('/(NOT)?\(\)/u', '', $parsed_query);
569    } while ($parsed_query !== $parsed_query_old);
570    $parsed_query = preg_replace('/(NOT|OR)+\)/u', ')'      , $parsed_query);
571    $parsed_query = preg_replace('/(OR)+/u'      , 'OR'     , $parsed_query);
572    $parsed_query = preg_replace('/\(OR/u'       , '('      , $parsed_query);
573    $parsed_query = preg_replace('/^OR|OR$/u'    , ''       , $parsed_query);
574    $parsed_query = preg_replace('/\)(NOT)?\(/u' , ')AND$1(', $parsed_query);
575
576    // adjustment: make highlightings right
577    $parens_level     = 0;
578    $notgrp_levels    = array();
579    $parsed_query_new = '';
580    $tokens = preg_split('/(NOT\(|[()])/u', $parsed_query, -1, PREG_SPLIT_DELIM_CAPTURE | PREG_SPLIT_NO_EMPTY);
581    foreach ($tokens as $token) {
582        if ($token === 'NOT(') {
583            $notgrp_levels[] = ++$parens_level;
584        } elseif ($token === '(') {
585            ++$parens_level;
586        } elseif ($token === ')') {
587            if ($parens_level-- === end($notgrp_levels)) array_pop($notgrp_levels);
588        } elseif (count($notgrp_levels) % 2 === 1) {
589            // turn highlight-flag off if terms are logically in "NOT" group
590            $token = preg_replace('/([WPN])\+\:/u', '$1-:', $token);
591        }
592        $parsed_query_new .= $token;
593    }
594    $parsed_query = $parsed_query_new;
595
596    /**
597     * convert infix notation string into postfix (Reverse Polish notation) array
598     * by Shunting-yard algorithm
599     *
600     * see: http://en.wikipedia.org/wiki/Reverse_Polish_notation
601     * see: http://en.wikipedia.org/wiki/Shunting-yard_algorithm
602     */
603    $parsed_ary     = array();
604    $ope_stack      = array();
605    $ope_precedence = array(')' => 1, 'OR' => 2, 'AND' => 3, 'NOT' => 4, '(' => 5);
606    $ope_regex      = '/([()]|OR|AND|NOT)/u';
607
608    $tokens = preg_split($ope_regex, $parsed_query, -1, PREG_SPLIT_DELIM_CAPTURE | PREG_SPLIT_NO_EMPTY);
609    foreach ($tokens as $token) {
610        if (preg_match($ope_regex, $token)) {
611            // operator
612            $last_ope = end($ope_stack);
613            while ($ope_precedence[$token] <= $ope_precedence[$last_ope] && $last_ope != '(') {
614                $parsed_ary[] = array_pop($ope_stack);
615                $last_ope = end($ope_stack);
616            }
617            if ($token == ')') {
618                array_pop($ope_stack); // this array_pop always deletes '('
619            } else {
620                $ope_stack[] = $token;
621            }
622        } else {
623            // operand
624            $token_decoded = str_replace(array('OP', 'CP'), array('(', ')'), $token);
625            $parsed_ary[] = $token_decoded;
626        }
627    }
628    $parsed_ary = array_values(array_merge($parsed_ary, array_reverse($ope_stack)));
629
630    // cleanup: each double "NOT" in RPN array actually does nothing
631    $parsed_ary_count = count($parsed_ary);
632    for ($i = 1; $i < $parsed_ary_count; ++$i) {
633        if ($parsed_ary[$i] === 'NOT' && $parsed_ary[$i - 1] === 'NOT') {
634            unset($parsed_ary[$i], $parsed_ary[$i - 1]);
635        }
636    }
637    $parsed_ary = array_values($parsed_ary);
638
639    // build return value
640    $q = array();
641    $q['query']      = $query;
642    $q['parsed_str'] = $parsed_query;
643    $q['parsed_ary'] = $parsed_ary;
644
645    foreach ($q['parsed_ary'] as $token) {
646        if ($token[2] !== ':') continue;
647        $body = substr($token, 3);
648
649        switch (substr($token, 0, 3)) {
650            case 'N+:':
651                     $q['ns'][]        = $body; // for backward compatibility
652                     break;
653            case 'N-:':
654                     $q['notns'][]     = $body; // for backward compatibility
655                     break;
656            case 'W_:':
657                     $q['words'][]     = $body;
658                     break;
659            case 'W-:':
660                     $q['words'][]     = $body;
661                     $q['not'][]       = $body; // for backward compatibility
662                     break;
663            case 'W+:':
664                     $q['words'][]     = $body;
665                     $q['highlight'][] = str_replace('*', '', $body);
666                     $q['and'][]       = $body; // for backward compatibility
667                     break;
668            case 'P-:':
669                     $q['phrases'][]   = $body;
670                     break;
671            case 'P+:':
672                     $q['phrases'][]   = $body;
673                     $q['highlight'][] = str_replace('*', '', $body);
674                     break;
675        }
676    }
677    foreach (array('words', 'phrases', 'highlight', 'ns', 'notns', 'and', 'not') as $key) {
678        $q[$key] = empty($q[$key]) ? array() : array_values(array_unique($q[$key]));
679    }
680
681    return $q;
682}
683
684/**
685 * Transforms given search term into intermediate representation
686 *
687 * This function is used in ft_queryParser() and not for general purpose use.
688 *
689 * @author Kazutaka Miyasaka <kazmiya@gmail.com>
690 */
691function ft_termParser($term, &$stopwords, $consider_asian = true, $phrase_mode = false) {
692    $parsed = '';
693    if ($consider_asian) {
694        // successive asian characters need to be searched as a phrase
695        $words = preg_split('/('.IDX_ASIAN.'+)/u', $term, -1, PREG_SPLIT_DELIM_CAPTURE | PREG_SPLIT_NO_EMPTY);
696        foreach ($words as $word) {
697            if (preg_match('/'.IDX_ASIAN.'/u', $word)) $phrase_mode = true;
698            $parsed .= ft_termParser($word, $stopwords, false, $phrase_mode);
699        }
700    } else {
701        $term_noparen = str_replace(array('(', ')'), ' ', $term);
702        $words = idx_tokenizer($term_noparen, $stopwords, true);
703
704        // W_: no need to highlight
705        if (empty($words)) {
706            $parsed = '()'; // important: do not remove
707        } elseif ($words[0] === $term) {
708            $parsed = '(W+:'.$words[0].')';
709        } elseif ($phrase_mode) {
710            $term_encoded = str_replace(array('(', ')'), array('OP', 'CP'), $term);
711            $parsed = '((W_:'.implode(')(W_:', $words).')(P+:'.$term_encoded.'))';
712        } else {
713            $parsed = '((W+:'.implode(')(W+:', $words).'))';
714        }
715    }
716    return $parsed;
717}
718
719//Setup VIM: ex: et ts=4 enc=utf-8 :
720