xref: /dokuwiki/inc/fulltext.php (revision 16905344219a6293705b71cd526fad3ba07b04eb)
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 $result;
136    require_once(DOKU_INC.'inc/parserutils.php');
137
138    // check metadata for matching links
139    foreach($docs as $match){
140        // metadata relation reference links are already resolved
141        $links = p_get_metadata($match,'relation references');
142        if (isset($links[$id])) $result[] = $match;
143    }
144
145    if(!count($result)) return $result;
146
147    // check ACL permissions
148    foreach(array_keys($result) as $idx){
149        if(auth_quickaclcheck($result[$idx]) < AUTH_READ){
150            unset($result[$idx]);
151        }
152    }
153
154    sort($result);
155    return $result;
156}
157
158/**
159 * Returns the pages that use a given media file
160 *
161 * Does a quick lookup with the fulltext index, then
162 * evaluates the instructions of the found pages
163 *
164 * Aborts after $max found results
165 */
166function ft_mediause($id,$max){
167    global $conf;
168    $swfile   = DOKU_INC.'inc/lang/'.$conf['lang'].'/stopwords.txt';
169    $stopwords = @file_exists($swfile) ? file($swfile) : array();
170
171    if(!$max) $max = 1; // need to find at least one
172
173    $result = array();
174
175    // quick lookup of the mediafile
176    $media   = noNS($id);
177    $matches = idx_lookup(idx_tokenizer($media,$stopwords));
178    $docs    = array_keys(ft_resultCombine(array_values($matches)));
179    if(!count($docs)) return $result;
180
181    // go through all found pages
182    $found = 0;
183    $pcre  = preg_quote($media,'/');
184    foreach($docs as $doc){
185        $ns = getNS($doc);
186        preg_match_all('/\{\{([^|}]*'.$pcre.'[^|}]*)(|[^}]+)?\}\}/i',rawWiki($doc),$matches);
187        foreach($matches[1] as $img){
188            $img = trim($img);
189            if(preg_match('/^https?:\/\//i',$img)) continue; // skip external images
190                list($img) = explode('?',$img);                  // remove any parameters
191            resolve_mediaid($ns,$img,$exists);               // resolve the possibly relative img
192
193            if($img == $id){                                 // we have a match
194                $result[] = $doc;
195                $found++;
196                break;
197            }
198        }
199        if($found >= $max) break;
200    }
201
202    sort($result);
203    return $result;
204}
205
206
207
208/**
209 * Quicksearch for pagenames
210 *
211 * By default it only matches the pagename and ignores the
212 * namespace. This can be changed with the second parameter
213 *
214 * refactored into ft_pageLookup(), _ft_pageLookup() and trigger_event()
215 *
216 * @author Andreas Gohr <andi@splitbrain.org>
217 */
218function ft_pageLookup($id,$pageonly=true){
219    $data = array('id' => $id, 'pageonly' => $pageonly);
220    return trigger_event('SEARCH_QUERY_PAGELOOKUP',$data,'_ft_pageLookup');
221}
222
223function _ft_pageLookup(&$data){
224    // split out original parameterrs
225    $id = $data['id'];
226    $pageonly = $data['pageonly'];
227
228    global $conf;
229    $id    = preg_quote($id,'/');
230    $pages = file($conf['indexdir'].'/page.idx');
231    if($id) $pages = array_values(preg_grep('/'.$id.'/',$pages));
232
233    $cnt = count($pages);
234    for($i=0; $i<$cnt; $i++){
235        if($pageonly){
236            if(!preg_match('/'.$id.'/',noNS($pages[$i]))){
237                unset($pages[$i]);
238                continue;
239            }
240        }
241        if(!page_exists($pages[$i])){
242            unset($pages[$i]);
243            continue;
244        }
245    }
246
247    $pages = array_filter($pages,'isVisiblePage'); // discard hidden pages
248    if(!count($pages)) return array();
249
250    // check ACL permissions
251    foreach(array_keys($pages) as $idx){
252        if(auth_quickaclcheck(trim($pages[$idx])) < AUTH_READ){
253            unset($pages[$idx]);
254        }
255    }
256
257    $pages = array_map('trim',$pages);
258    usort($pages,'ft_pagesorter');
259    return $pages;
260}
261
262/**
263 * Sort pages based on their namespace level first, then on their string
264 * values. This makes higher hierarchy pages rank higher than lower hierarchy
265 * pages.
266 */
267function ft_pagesorter($a, $b){
268    $ac = count(explode(':',$a));
269    $bc = count(explode(':',$b));
270    if($ac < $bc){
271        return -1;
272    }elseif($ac > $bc){
273        return 1;
274    }
275    return strcmp ($a,$b);
276}
277
278/**
279 * Creates a snippet extract
280 *
281 * @author Andreas Gohr <andi@splitbrain.org>
282 * @triggers FULLTEXT_SNIPPET_CREATE
283 */
284function ft_snippet($id,$highlight){
285    $text = rawWiki($id);
286    $evdata = array(
287            'id'        => $id,
288            'text'      => &$text,
289            'highlight' => &$highlight,
290            'snippet'   => '',
291            );
292
293    $evt = new Doku_Event('FULLTEXT_SNIPPET_CREATE',$evdata);
294    if ($evt->advise_before()) {
295        $match = array();
296        $snippets = array();
297        $utf8_offset = $offset = $end = 0;
298        $len = utf8_strlen($text);
299
300        // build a regexp from the phrases to highlight
301        $re1 = '('.join('|',array_map('preg_quote_cb',array_filter((array) $highlight))).')';
302        $re2 = "$re1.{0,75}(?!\\1)$re1";
303        $re3 = "$re1.{0,45}(?!\\1)$re1.{0,45}(?!\\1)(?!\\2)$re1";
304
305        for ($cnt=4; $cnt--;) {
306            if (0) {
307            } else if (preg_match('/'.$re3.'/iu',$text,$match,PREG_OFFSET_CAPTURE,$offset)) {
308            } else if (preg_match('/'.$re2.'/iu',$text,$match,PREG_OFFSET_CAPTURE,$offset)) {
309            } else if (preg_match('/'.$re1.'/iu',$text,$match,PREG_OFFSET_CAPTURE,$offset)) {
310            } else {
311                break;
312            }
313
314            list($str,$idx) = $match[0];
315
316            // convert $idx (a byte offset) into a utf8 character offset
317            $utf8_idx = utf8_strlen(substr($text,0,$idx));
318            $utf8_len = utf8_strlen($str);
319
320            // establish context, 100 bytes surrounding the match string
321            // first look to see if we can go 100 either side,
322            // then drop to 50 adding any excess if the other side can't go to 50,
323            $pre = min($utf8_idx-$utf8_offset,100);
324            $post = min($len-$utf8_idx-$utf8_len,100);
325
326            if ($pre>50 && $post>50) {
327                $pre = $post = 50;
328            } else if ($pre>50) {
329                $pre = min($pre,100-$post);
330            } else if ($post>50) {
331                $post = min($post, 100-$pre);
332            } else {
333                // both are less than 50, means the context is the whole string
334                // make it so and break out of this loop - there is no need for the
335                // complex snippet calculations
336                $snippets = array($text);
337                break;
338            }
339
340            // establish context start and end points, try to append to previous
341            // context if possible
342            $start = $utf8_idx - $pre;
343            $append = ($start < $end) ? $end : false;  // still the end of the previous context snippet
344            $end = $utf8_idx + $utf8_len + $post;      // now set it to the end of this context
345
346            if ($append) {
347                $snippets[count($snippets)-1] .= utf8_substr($text,$append,$end-$append);
348            } else {
349                $snippets[] = utf8_substr($text,$start,$end-$start);
350            }
351
352            // set $offset for next match attempt
353            //   substract strlen to avoid splitting a potential search success,
354            //   this is an approximation as the search pattern may match strings
355            //   of varying length and it will fail if the context snippet
356            //   boundary breaks a matching string longer than the current match
357            $utf8_offset = $utf8_idx + $post;
358            $offset = $idx + strlen(utf8_substr($text,$utf8_idx,$post));
359            $offset = utf8_correctIdx($text,$offset);
360        }
361
362        $m = "\1";
363        $snippets = preg_replace('/'.$re1.'/iu',$m.'$1'.$m,$snippets);
364        $snippet = preg_replace('/'.$m.'([^'.$m.']*?)'.$m.'/iu','<strong class="search_hit">$1</strong>',hsc(join('... ',$snippets)));
365
366        $evdata['snippet'] = $snippet;
367    }
368    $evt->advise_after();
369    unset($evt);
370
371    return $evdata['snippet'];
372}
373
374/**
375 * Combine found documents and sum up their scores
376 *
377 * This function is used to combine searched words with a logical
378 * AND. Only documents available in all arrays are returned.
379 *
380 * based upon PEAR's PHP_Compat function for array_intersect_key()
381 *
382 * @param array $args An array of page arrays
383 */
384function ft_resultCombine($args){
385    $array_count = count($args);
386    if($array_count == 1){
387        return $args[0];
388    }
389
390    $result = array();
391    if ($array_count > 1) {
392        foreach ($args[0] as $key => $value) {
393            $result[$key] = $value;
394            for ($i = 1; $i !== $array_count; $i++) {
395                if (!isset($args[$i][$key])) {
396                    unset($result[$key]);
397                    break;
398                }
399                $result[$key] += $args[$i][$key];
400            }
401        }
402    }
403    return $result;
404}
405
406/**
407 * Unites found documents and sum up their scores
408 *
409 * based upon ft_resultCombine() function
410 *
411 * @param array $args An array of page arrays
412 * @author Kazutaka Miyasaka <kazmiya@gmail.com>
413 */
414function ft_resultUnite($args) {
415    $array_count = count($args);
416    if ($array_count === 1) {
417        return $args[0];
418    }
419
420    $result = $args[0];
421    for ($i = 1; $i !== $array_count; $i++) {
422        foreach (array_keys($args[$i]) as $id) {
423            $result[$id] += $args[$i][$id];
424        }
425    }
426    return $result;
427}
428
429/**
430 * Computes the difference of documents using page id for comparison
431 *
432 * nearly identical to PHP5's array_diff_key()
433 *
434 * @param array $args An array of page arrays
435 * @author Kazutaka Miyasaka <kazmiya@gmail.com>
436 */
437function ft_resultComplement($args) {
438    $array_count = count($args);
439    if ($array_count === 1) {
440        return $args[0];
441    }
442
443    $result = $args[0];
444    foreach (array_keys($result) as $id) {
445        for ($i = 1; $i !== $array_count; $i++) {
446            if (isset($args[$i][$id])) unset($result[$id]);
447        }
448    }
449    return $result;
450}
451
452/**
453 * Parses a search query and builds an array of search formulas
454 *
455 * @author Andreas Gohr <andi@splitbrain.org>
456 * @author Kazutaka Miyasaka <kazmiya@gmail.com>
457 */
458function ft_queryParser($query){
459    global $conf;
460    $swfile    = DOKU_INC.'inc/lang/'.$conf['lang'].'/stopwords.txt';
461    $stopwords = @file_exists($swfile) ? file($swfile) : array();
462
463    /**
464     * parse a search query and transform it into intermediate representation
465     *
466     * in a search query, you can use the following expressions:
467     *
468     *   words:
469     *     include
470     *     -exclude
471     *   phrases:
472     *     "phrase to be included"
473     *     -"phrase you want to exclude"
474     *   namespaces:
475     *     @include:namespace (or ns:include:namespace)
476     *     ^exclude:namespace (or -ns:exclude:namespace)
477     *   groups:
478     *     ()
479     *     -()
480     *   operators:
481     *     and ('and' is the default operator: you can always omit this)
482     *     or  (or pipe symbol '|', lower precedence than 'and')
483     *
484     * e.g. a query [ aa "bb cc" @dd:ee ] means "search pages which contain
485     *      a word 'aa', a phrase 'bb cc' and are within a namespace 'dd:ee'".
486     *      this query is equivalent to [ -(-aa or -"bb cc" or -ns:dd:ee) ]
487     *      as long as you don't mind hit counts.
488     *
489     * intermediate representation consists of the following parts:
490     *
491     *   ( )           - group
492     *   AND           - logical and
493     *   OR            - logical or
494     *   NOT           - logical not
495     *   W+:, W-:, W_: - word      (underscore: no need to highlight)
496     *   P+:, P-:      - phrase    (minus sign: logically in NOT group)
497     *   N+:, N-:      - namespace
498     */
499    $parsed_query = '';
500    $parens_level = 0;
501    $terms = preg_split('/(-?".*?")/u', utf8_strtolower($query), -1, PREG_SPLIT_DELIM_CAPTURE | PREG_SPLIT_NO_EMPTY);
502
503    foreach ($terms as $term) {
504        $parsed = '';
505        if (preg_match('/^(-?)"(.+)"$/u', $term, $matches)) {
506            // phrase-include and phrase-exclude
507            $not = $matches[1] ? 'NOT' : '';
508            $parsed = $not.ft_termParser($matches[2], $stopwords, false, true);
509        } else {
510            // fix incomplete phrase
511            $term = str_replace('"', ' ', $term);
512
513            // fix parentheses
514            $term = str_replace(')'  , ' ) ', $term);
515            $term = str_replace('('  , ' ( ', $term);
516            $term = str_replace('- (', ' -(', $term);
517
518            // treat pipe symbols as 'OR' operators
519            $term = str_replace('|', ' or ', $term);
520
521            // treat ideographic spaces (U+3000) as search term separators
522            // FIXME: some more separators?
523            $term = preg_replace('/[ \x{3000}]+/u', ' ',  $term);
524            $term = trim($term);
525            if ($term === '') continue;
526
527            $tokens = explode(' ', $term);
528            foreach ($tokens as $token) {
529                if ($token === '(') {
530                    // parenthesis-include-open
531                    $parsed .= '(';
532                    ++$parens_level;
533                } elseif ($token === '-(') {
534                    // parenthesis-exclude-open
535                    $parsed .= 'NOT(';
536                    ++$parens_level;
537                } elseif ($token === ')') {
538                    // parenthesis-any-close
539                    if ($parens_level === 0) continue;
540                    $parsed .= ')';
541                    $parens_level--;
542                } elseif ($token === 'and') {
543                    // logical-and (do nothing)
544                } elseif ($token === 'or') {
545                    // logical-or
546                    $parsed .= 'OR';
547                } elseif (preg_match('/^(?:\^|-ns:)(.+)$/u', $token, $matches)) {
548                    // namespace-exclude
549                    $parsed .= 'NOT(N+:'.$matches[1].')';
550                } elseif (preg_match('/^(?:@|ns:)(.+)$/u', $token, $matches)) {
551                    // namespace-include
552                    $parsed .= '(N+:'.$matches[1].')';
553                } elseif (preg_match('/^-(.+)$/', $token, $matches)) {
554                    // word-exclude
555                    $parsed .= 'NOT('.ft_termParser($matches[1], $stopwords).')';
556                } else {
557                    // word-include
558                    $parsed .= ft_termParser($token, $stopwords);
559                }
560            }
561        }
562        $parsed_query .= $parsed;
563    }
564
565    // cleanup (very sensitive)
566    $parsed_query .= str_repeat(')', $parens_level);
567    do {
568        $parsed_query_old = $parsed_query;
569        $parsed_query = preg_replace('/(NOT)?\(\)/u', '', $parsed_query);
570    } while ($parsed_query !== $parsed_query_old);
571    $parsed_query = preg_replace('/(NOT|OR)+\)/u', ')'      , $parsed_query);
572    $parsed_query = preg_replace('/(OR)+/u'      , 'OR'     , $parsed_query);
573    $parsed_query = preg_replace('/\(OR/u'       , '('      , $parsed_query);
574    $parsed_query = preg_replace('/^OR|OR$/u'    , ''       , $parsed_query);
575    $parsed_query = preg_replace('/\)(NOT)?\(/u' , ')AND$1(', $parsed_query);
576
577    // adjustment: make highlightings right
578    $parens_level     = 0;
579    $notgrp_levels    = array();
580    $parsed_query_new = '';
581    $tokens = preg_split('/(NOT\(|[()])/u', $parsed_query, -1, PREG_SPLIT_DELIM_CAPTURE | PREG_SPLIT_NO_EMPTY);
582    foreach ($tokens as $token) {
583        if ($token === 'NOT(') {
584            $notgrp_levels[] = ++$parens_level;
585        } elseif ($token === '(') {
586            ++$parens_level;
587        } elseif ($token === ')') {
588            if ($parens_level-- === end($notgrp_levels)) array_pop($notgrp_levels);
589        } elseif (count($notgrp_levels) % 2 === 1) {
590            // turn highlight-flag off if terms are logically in "NOT" group
591            $token = preg_replace('/([WPN])\+\:/u', '$1-:', $token);
592        }
593        $parsed_query_new .= $token;
594    }
595    $parsed_query = $parsed_query_new;
596
597    /**
598     * convert infix notation string into postfix (Reverse Polish notation) array
599     * by Shunting-yard algorithm
600     *
601     * see: http://en.wikipedia.org/wiki/Reverse_Polish_notation
602     * see: http://en.wikipedia.org/wiki/Shunting-yard_algorithm
603     */
604    $parsed_ary     = array();
605    $ope_stack      = array();
606    $ope_precedence = array(')' => 1, 'OR' => 2, 'AND' => 3, 'NOT' => 4, '(' => 5);
607    $ope_regex      = '/([()]|OR|AND|NOT)/u';
608
609    $tokens = preg_split($ope_regex, $parsed_query, -1, PREG_SPLIT_DELIM_CAPTURE | PREG_SPLIT_NO_EMPTY);
610    foreach ($tokens as $token) {
611        if (preg_match($ope_regex, $token)) {
612            // operator
613            $last_ope = end($ope_stack);
614            while ($ope_precedence[$token] <= $ope_precedence[$last_ope] && $last_ope != '(') {
615                $parsed_ary[] = array_pop($ope_stack);
616                $last_ope = end($ope_stack);
617            }
618            if ($token == ')') {
619                array_pop($ope_stack); // this array_pop always deletes '('
620            } else {
621                $ope_stack[] = $token;
622            }
623        } else {
624            // operand
625            $token_decoded = str_replace(array('OP', 'CP'), array('(', ')'), $token);
626            $parsed_ary[] = $token_decoded;
627        }
628    }
629    $parsed_ary = array_values(array_merge($parsed_ary, array_reverse($ope_stack)));
630
631    // cleanup: each double "NOT" in RPN array actually does nothing
632    $parsed_ary_count = count($parsed_ary);
633    for ($i = 1; $i < $parsed_ary_count; ++$i) {
634        if ($parsed_ary[$i] === 'NOT' && $parsed_ary[$i - 1] === 'NOT') {
635            unset($parsed_ary[$i], $parsed_ary[$i - 1]);
636        }
637    }
638    $parsed_ary = array_values($parsed_ary);
639
640    // build return value
641    $q = array();
642    $q['query']      = $query;
643    $q['parsed_str'] = $parsed_query;
644    $q['parsed_ary'] = $parsed_ary;
645
646    foreach ($q['parsed_ary'] as $token) {
647        if ($token[2] !== ':') continue;
648        $body = substr($token, 3);
649
650        switch (substr($token, 0, 3)) {
651            case 'N+:':
652                     $q['ns'][]        = $body; // for backward compatibility
653                     break;
654            case 'N-:':
655                     $q['notns'][]     = $body; // for backward compatibility
656                     break;
657            case 'W_:':
658                     $q['words'][]     = $body;
659                     break;
660            case 'W-:':
661                     $q['words'][]     = $body;
662                     $q['not'][]       = $body; // for backward compatibility
663                     break;
664            case 'W+:':
665                     $q['words'][]     = $body;
666                     $q['highlight'][] = str_replace('*', '', $body);
667                     $q['and'][]       = $body; // for backward compatibility
668                     break;
669            case 'P-:':
670                     $q['phrases'][]   = $body;
671                     break;
672            case 'P+:':
673                     $q['phrases'][]   = $body;
674                     $q['highlight'][] = str_replace('*', '', $body);
675                     break;
676        }
677    }
678    foreach (array('words', 'phrases', 'highlight', 'ns', 'notns', 'and', 'not') as $key) {
679        $q[$key] = empty($q[$key]) ? array() : array_values(array_unique($q[$key]));
680    }
681
682    return $q;
683}
684
685/**
686 * Transforms given search term into intermediate representation
687 *
688 * This function is used in ft_queryParser() and not for general purpose use.
689 *
690 * @author Kazutaka Miyasaka <kazmiya@gmail.com>
691 */
692function ft_termParser($term, &$stopwords, $consider_asian = true, $phrase_mode = false) {
693    $parsed = '';
694    if ($consider_asian) {
695        // successive asian characters need to be searched as a phrase
696        $words = preg_split('/('.IDX_ASIAN.'+)/u', $term, -1, PREG_SPLIT_DELIM_CAPTURE | PREG_SPLIT_NO_EMPTY);
697        foreach ($words as $word) {
698            if (preg_match('/'.IDX_ASIAN.'/u', $word)) $phrase_mode = true;
699            $parsed .= ft_termParser($word, $stopwords, false, $phrase_mode);
700        }
701    } else {
702        $term_noparen = str_replace(array('(', ')'), ' ', $term);
703        $words = idx_tokenizer($term_noparen, $stopwords, true);
704
705        // W_: no need to highlight
706        if (empty($words)) {
707            $parsed = '()'; // important: do not remove
708        } elseif ($words[0] === $term) {
709            $parsed = '(W+:'.$words[0].')';
710        } elseif ($phrase_mode) {
711            $term_encoded = str_replace(array('(', ')'), array('OP', 'CP'), $term);
712            $parsed = '((W_:'.implode(')(W_:', $words).')(P+:'.$term_encoded.'))';
713        } else {
714            $parsed = '((W+:'.implode(')(W+:', $words).'))';
715        }
716    }
717    return $parsed;
718}
719
720//Setup VIM: ex: et ts=4 enc=utf-8 :
721