xref: /dokuwiki/inc/fulltext.php (revision b005809cbc9d5e96a687d5d0445452fb317981fb)
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 * @param string $query
24 * @param array $highlight
25 * @return array
26 */
27function ft_pageSearch($query,&$highlight, $sort = 'hits'){
28
29    $data = array();
30    $data['query'] = $query;
31    $data['highlight'] =& $highlight;
32    $data['sort'] = $sort;
33
34    return trigger_event('SEARCH_QUERY_FULLPAGE', $data, '_ft_pageSearch');
35}
36
37/**
38 * Returns a list of matching documents for the given query
39 *
40 * @author Andreas Gohr <andi@splitbrain.org>
41 * @author Kazutaka Miyasaka <kazmiya@gmail.com>
42 *
43 * @param array $data event data
44 * @return array matching documents
45 */
46function _ft_pageSearch(&$data) {
47    $Indexer = idx_get_indexer();
48
49    // parse the given query
50    $q = ft_queryParser($Indexer, $data['query']);
51    $data['highlight'] = $q['highlight'];
52
53    if (empty($q['parsed_ary'])) return array();
54
55    // lookup all words found in the query
56    $lookup = $Indexer->lookup($q['words']);
57
58    // get all pages in this dokuwiki site (!: includes nonexistent pages)
59    $pages_all = array();
60    foreach ($Indexer->getPages() as $id) {
61        $pages_all[$id] = 0; // base: 0 hit
62    }
63
64    // process the query
65    $stack = array();
66    foreach ($q['parsed_ary'] as $token) {
67        switch (substr($token, 0, 3)) {
68            case 'W+:':
69            case 'W-:':
70            case 'W_:': // word
71                $word    = substr($token, 3);
72                $stack[] = (array) $lookup[$word];
73                break;
74            case 'P+:':
75            case 'P-:': // phrase
76                $phrase = substr($token, 3);
77                // since phrases are always parsed as ((W1)(W2)...(P)),
78                // the end($stack) always points the pages that contain
79                // all words in this phrase
80                $pages  = end($stack);
81                $pages_matched = array();
82                foreach(array_keys($pages) as $id){
83                    $evdata = array(
84                        'id' => $id,
85                        'phrase' => $phrase,
86                        'text' => rawWiki($id)
87                    );
88                    $evt = new Doku_Event('FULLTEXT_PHRASE_MATCH',$evdata);
89                    if ($evt->advise_before() && $evt->result !== true) {
90                        $text = utf8_strtolower($evdata['text']);
91                        if (strpos($text, $phrase) !== false) {
92                            $evt->result = true;
93                        }
94                    }
95                    $evt->advise_after();
96                    if ($evt->result === true) {
97                        $pages_matched[$id] = 0; // phrase: always 0 hit
98                    }
99                }
100                $stack[] = $pages_matched;
101                break;
102            case 'N+:':
103            case 'N-:': // namespace
104                $ns = cleanID(substr($token, 3)) . ':';
105                $pages_matched = array();
106                foreach (array_keys($pages_all) as $id) {
107                    if (strpos($id, $ns) === 0) {
108                        $pages_matched[$id] = 0; // namespace: always 0 hit
109                    }
110                }
111                $stack[] = $pages_matched;
112                break;
113            case 'AND': // and operation
114                list($pages1, $pages2) = array_splice($stack, -2);
115                $stack[] = ft_resultCombine(array($pages1, $pages2));
116                break;
117            case 'OR':  // or operation
118                list($pages1, $pages2) = array_splice($stack, -2);
119                $stack[] = ft_resultUnite(array($pages1, $pages2));
120                break;
121            case 'NOT': // not operation (unary)
122                $pages   = array_pop($stack);
123                $stack[] = ft_resultComplement(array($pages_all, $pages));
124                break;
125        }
126    }
127    $docs = array_pop($stack);
128
129    if (empty($docs)) return array();
130
131    // check: settings, acls, existence
132    foreach (array_keys($docs) as $id) {
133        if (isHiddenPage($id) || auth_quickaclcheck($id) < AUTH_READ || !page_exists($id, '', false)) {
134            unset($docs[$id]);
135        }
136    }
137
138    $docs = _ft_filterResultsByTime($docs);
139
140    if ($data['sort'] === 'mtime') {
141        uksort($docs, 'ft_pagemtimesorter');
142    } else {
143        // sort docs by count
144        arsort($docs);
145    }
146
147    return $docs;
148}
149
150/**
151 * Returns the backlinks for a given page
152 *
153 * Uses the metadata index.
154 *
155 * @param string $id           The id for which links shall be returned
156 * @param bool   $ignore_perms Ignore the fact that pages are hidden or read-protected
157 * @return array The pages that contain links to the given page
158 */
159function ft_backlinks($id, $ignore_perms = false){
160    $result = idx_get_indexer()->lookupKey('relation_references', $id);
161
162    if(!count($result)) return $result;
163
164    // check ACL permissions
165    foreach(array_keys($result) as $idx){
166        if(($ignore_perms !== true && (
167                isHiddenPage($result[$idx]) || auth_quickaclcheck($result[$idx]) < AUTH_READ
168            )) || !page_exists($result[$idx], '', false)){
169            unset($result[$idx]);
170        }
171    }
172
173    sort($result);
174    return $result;
175}
176
177/**
178 * Returns the pages that use a given media file
179 *
180 * Uses the relation media metadata property and the metadata index.
181 *
182 * Note that before 2013-07-31 the second parameter was the maximum number of results and
183 * permissions were ignored. That's why the parameter is now checked to be explicitely set
184 * to true (with type bool) in order to be compatible with older uses of the function.
185 *
186 * @param string $id           The media id to look for
187 * @param bool   $ignore_perms Ignore hidden pages and acls (optional, default: false)
188 * @return array A list of pages that use the given media file
189 */
190function ft_mediause($id, $ignore_perms = false){
191    $result = idx_get_indexer()->lookupKey('relation_media', $id);
192
193    if(!count($result)) return $result;
194
195    // check ACL permissions
196    foreach(array_keys($result) as $idx){
197        if(($ignore_perms !== true && (
198                    isHiddenPage($result[$idx]) || auth_quickaclcheck($result[$idx]) < AUTH_READ
199                )) || !page_exists($result[$idx], '', false)){
200            unset($result[$idx]);
201        }
202    }
203
204    sort($result);
205    return $result;
206}
207
208
209
210/**
211 * Quicksearch for pagenames
212 *
213 * By default it only matches the pagename and ignores the
214 * namespace. This can be changed with the second parameter.
215 * The third parameter allows to search in titles as well.
216 *
217 * The function always returns titles as well
218 *
219 * @triggers SEARCH_QUERY_PAGELOOKUP
220 * @author Andreas Gohr <andi@splitbrain.org>
221 * @author Adrian Lang <lang@cosmocode.de>
222 *
223 * @param string $id        page id
224 * @param bool   $in_ns     match against namespace as well?
225 * @param bool   $in_title  search in title?
226 * @return string[]
227 */
228function ft_pageLookup($id, $in_ns=false, $in_title=false){
229    $data = compact('id', 'in_ns', 'in_title');
230    $data['has_titles'] = true; // for plugin backward compatibility check
231    return trigger_event('SEARCH_QUERY_PAGELOOKUP', $data, '_ft_pageLookup');
232}
233
234/**
235 * Returns list of pages as array(pageid => First Heading)
236 *
237 * @param array &$data event data
238 * @return string[]
239 */
240function _ft_pageLookup(&$data){
241    // split out original parameters
242    $id = $data['id'];
243    $Indexer = idx_get_indexer();
244    $parsedQuery = ft_queryParser($Indexer, $id);
245    if (count($parsedQuery['ns']) > 0) {
246        $ns = cleanID($parsedQuery['ns'][0]) . ':';
247        $id = implode(' ', $parsedQuery['highlight']);
248    }
249
250    $in_ns    = $data['in_ns'];
251    $in_title = $data['in_title'];
252    $cleaned = cleanID($id);
253
254    $Indexer = idx_get_indexer();
255    $page_idx = $Indexer->getPages();
256
257    $pages = array();
258    if ($id !== '' && $cleaned !== '') {
259        foreach ($page_idx as $p_id) {
260            if ((strpos($in_ns ? $p_id : noNSorNS($p_id), $cleaned) !== false)) {
261                if (!isset($pages[$p_id]))
262                    $pages[$p_id] = p_get_first_heading($p_id, METADATA_DONT_RENDER);
263            }
264        }
265        if ($in_title) {
266            foreach ($Indexer->lookupKey('title', $id, '_ft_pageLookupTitleCompare') as $p_id) {
267                if (!isset($pages[$p_id]))
268                    $pages[$p_id] = p_get_first_heading($p_id, METADATA_DONT_RENDER);
269            }
270        }
271    }
272
273    if (isset($ns)) {
274        foreach (array_keys($pages) as $p_id) {
275            if (strpos($p_id, $ns) !== 0) {
276                unset($pages[$p_id]);
277            }
278        }
279    }
280
281    // discard hidden pages
282    // discard nonexistent pages
283    // check ACL permissions
284    foreach(array_keys($pages) as $idx){
285        if(!isVisiblePage($idx) || !page_exists($idx) ||
286           auth_quickaclcheck($idx) < AUTH_READ) {
287            unset($pages[$idx]);
288        }
289    }
290
291    $pages = _ft_filterResultsByTime($pages);
292
293    uksort($pages,'ft_pagesorter');
294    return $pages;
295}
296
297
298/**
299 * @param array $results search results in the form pageid => value
300 *
301 * @return array
302 */
303function _ft_filterResultsByTime(array $results) {
304    global $INPUT;
305    if ($INPUT->has('after') || $INPUT->has('before')) {
306        $after = $INPUT->str('after');
307        $after = is_int($after) ? $after : strtotime($after);
308
309        $before = $INPUT->str('before');
310        $before = is_int($before) ? $before : strtotime($before);
311
312        foreach ($results as $id => $value) {
313            $mTime = filemtime(wikiFN($id));
314            if ($after && $after > $mTime) {
315                unset($results[$id]);
316                continue;
317            }
318            if ($before && $before < $mTime) {
319                unset($results[$id]);
320            }
321        }
322    }
323
324    return $results;
325}
326
327/**
328 * Tiny helper function for comparing the searched title with the title
329 * from the search index. This function is a wrapper around stripos with
330 * adapted argument order and return value.
331 *
332 * @param string $search searched title
333 * @param string $title  title from index
334 * @return bool
335 */
336function _ft_pageLookupTitleCompare($search, $title) {
337    return stripos($title, $search) !== false;
338}
339
340/**
341 * Sort pages based on their namespace level first, then on their string
342 * values. This makes higher hierarchy pages rank higher than lower hierarchy
343 * pages.
344 *
345 * @param string $a
346 * @param string $b
347 * @return int Returns < 0 if $a is less than $b; > 0 if $a is greater than $b, and 0 if they are equal.
348 */
349function ft_pagesorter($a, $b){
350    $ac = count(explode(':',$a));
351    $bc = count(explode(':',$b));
352    if($ac < $bc){
353        return -1;
354    }elseif($ac > $bc){
355        return 1;
356    }
357    return strcmp ($a,$b);
358}
359
360/**
361 * Sort pages by their mtime, from newest to oldest
362 *
363 * @param string $a
364 * @param string $b
365 *
366 * @return int Returns < 0 if $a is newer than $b, > 0 if $b is newer than $a and 0 if they are of the same age
367 */
368function ft_pagemtimesorter($a, $b) {
369    $mtimeA = filemtime(wikiFN($a));
370    $mtimeB = filemtime(wikiFN($b));
371    return $mtimeB - $mtimeA;
372}
373
374/**
375 * Creates a snippet extract
376 *
377 * @author Andreas Gohr <andi@splitbrain.org>
378 * @triggers FULLTEXT_SNIPPET_CREATE
379 *
380 * @param string $id page id
381 * @param array $highlight
382 * @return mixed
383 */
384function ft_snippet($id,$highlight){
385    $text = rawWiki($id);
386    $text = str_replace("\xC2\xAD",'',$text); // remove soft-hyphens
387    $evdata = array(
388            'id'        => $id,
389            'text'      => &$text,
390            'highlight' => &$highlight,
391            'snippet'   => '',
392            );
393
394    $evt = new Doku_Event('FULLTEXT_SNIPPET_CREATE',$evdata);
395    if ($evt->advise_before()) {
396        $match = array();
397        $snippets = array();
398        $utf8_offset = $offset = $end = 0;
399        $len = utf8_strlen($text);
400
401        // build a regexp from the phrases to highlight
402        $re1 = '('.join('|',array_map('ft_snippet_re_preprocess', array_map('preg_quote_cb',array_filter((array) $highlight)))).')';
403        $re2 = "$re1.{0,75}(?!\\1)$re1";
404        $re3 = "$re1.{0,45}(?!\\1)$re1.{0,45}(?!\\1)(?!\\2)$re1";
405
406        for ($cnt=4; $cnt--;) {
407            if (0) {
408            } else if (preg_match('/'.$re3.'/iu',$text,$match,PREG_OFFSET_CAPTURE,$offset)) {
409            } else if (preg_match('/'.$re2.'/iu',$text,$match,PREG_OFFSET_CAPTURE,$offset)) {
410            } else if (preg_match('/'.$re1.'/iu',$text,$match,PREG_OFFSET_CAPTURE,$offset)) {
411            } else {
412                break;
413            }
414
415            list($str,$idx) = $match[0];
416
417            // convert $idx (a byte offset) into a utf8 character offset
418            $utf8_idx = utf8_strlen(substr($text,0,$idx));
419            $utf8_len = utf8_strlen($str);
420
421            // establish context, 100 bytes surrounding the match string
422            // first look to see if we can go 100 either side,
423            // then drop to 50 adding any excess if the other side can't go to 50,
424            $pre = min($utf8_idx-$utf8_offset,100);
425            $post = min($len-$utf8_idx-$utf8_len,100);
426
427            if ($pre>50 && $post>50) {
428                $pre = $post = 50;
429            } else if ($pre>50) {
430                $pre = min($pre,100-$post);
431            } else if ($post>50) {
432                $post = min($post, 100-$pre);
433            } else if ($offset == 0) {
434                // both are less than 50, means the context is the whole string
435                // make it so and break out of this loop - there is no need for the
436                // complex snippet calculations
437                $snippets = array($text);
438                break;
439            }
440
441            // establish context start and end points, try to append to previous
442            // context if possible
443            $start = $utf8_idx - $pre;
444            $append = ($start < $end) ? $end : false;  // still the end of the previous context snippet
445            $end = $utf8_idx + $utf8_len + $post;      // now set it to the end of this context
446
447            if ($append) {
448                $snippets[count($snippets)-1] .= utf8_substr($text,$append,$end-$append);
449            } else {
450                $snippets[] = utf8_substr($text,$start,$end-$start);
451            }
452
453            // set $offset for next match attempt
454            // continue matching after the current match
455            // if the current match is not the longest possible match starting at the current offset
456            // this prevents further matching of this snippet but for possible matches of length
457            // smaller than match length + context (at least 50 characters) this match is part of the context
458            $utf8_offset = $utf8_idx + $utf8_len;
459            $offset = $idx + strlen(utf8_substr($text,$utf8_idx,$utf8_len));
460            $offset = utf8_correctIdx($text,$offset);
461        }
462
463        $m = "\1";
464        $snippets = preg_replace('/'.$re1.'/iu',$m.'$1'.$m,$snippets);
465        $snippet = preg_replace('/'.$m.'([^'.$m.']*?)'.$m.'/iu','<strong class="search_hit">$1</strong>',hsc(join('... ',$snippets)));
466
467        $evdata['snippet'] = $snippet;
468    }
469    $evt->advise_after();
470    unset($evt);
471
472    return $evdata['snippet'];
473}
474
475/**
476 * Wraps a search term in regex boundary checks.
477 *
478 * @param string $term
479 * @return string
480 */
481function ft_snippet_re_preprocess($term) {
482    // do not process asian terms where word boundaries are not explicit
483    if(preg_match('/'.IDX_ASIAN.'/u',$term)){
484        return $term;
485    }
486
487    if (UTF8_PROPERTYSUPPORT) {
488        // unicode word boundaries
489        // see http://stackoverflow.com/a/2449017/172068
490        $BL = '(?<!\pL)';
491        $BR = '(?!\pL)';
492    } else {
493        // not as correct as above, but at least won't break
494        $BL = '\b';
495        $BR = '\b';
496    }
497
498    if(substr($term,0,2) == '\\*'){
499        $term = substr($term,2);
500    }else{
501        $term = $BL.$term;
502    }
503
504    if(substr($term,-2,2) == '\\*'){
505        $term = substr($term,0,-2);
506    }else{
507        $term = $term.$BR;
508    }
509
510    if($term == $BL || $term == $BR || $term == $BL.$BR) $term = '';
511    return $term;
512}
513
514/**
515 * Combine found documents and sum up their scores
516 *
517 * This function is used to combine searched words with a logical
518 * AND. Only documents available in all arrays are returned.
519 *
520 * based upon PEAR's PHP_Compat function for array_intersect_key()
521 *
522 * @param array $args An array of page arrays
523 * @return array
524 */
525function ft_resultCombine($args){
526    $array_count = count($args);
527    if($array_count == 1){
528        return $args[0];
529    }
530
531    $result = array();
532    if ($array_count > 1) {
533        foreach ($args[0] as $key => $value) {
534            $result[$key] = $value;
535            for ($i = 1; $i !== $array_count; $i++) {
536                if (!isset($args[$i][$key])) {
537                    unset($result[$key]);
538                    break;
539                }
540                $result[$key] += $args[$i][$key];
541            }
542        }
543    }
544    return $result;
545}
546
547/**
548 * Unites found documents and sum up their scores
549 *
550 * based upon ft_resultCombine() function
551 *
552 * @param array $args An array of page arrays
553 * @return array
554 *
555 * @author Kazutaka Miyasaka <kazmiya@gmail.com>
556 */
557function ft_resultUnite($args) {
558    $array_count = count($args);
559    if ($array_count === 1) {
560        return $args[0];
561    }
562
563    $result = $args[0];
564    for ($i = 1; $i !== $array_count; $i++) {
565        foreach (array_keys($args[$i]) as $id) {
566            $result[$id] += $args[$i][$id];
567        }
568    }
569    return $result;
570}
571
572/**
573 * Computes the difference of documents using page id for comparison
574 *
575 * nearly identical to PHP5's array_diff_key()
576 *
577 * @param array $args An array of page arrays
578 * @return array
579 *
580 * @author Kazutaka Miyasaka <kazmiya@gmail.com>
581 */
582function ft_resultComplement($args) {
583    $array_count = count($args);
584    if ($array_count === 1) {
585        return $args[0];
586    }
587
588    $result = $args[0];
589    foreach (array_keys($result) as $id) {
590        for ($i = 1; $i !== $array_count; $i++) {
591            if (isset($args[$i][$id])) unset($result[$id]);
592        }
593    }
594    return $result;
595}
596
597/**
598 * Parses a search query and builds an array of search formulas
599 *
600 * @author Andreas Gohr <andi@splitbrain.org>
601 * @author Kazutaka Miyasaka <kazmiya@gmail.com>
602 *
603 * @param Doku_Indexer $Indexer
604 * @param string $query search query
605 * @return array of search formulas
606 */
607function ft_queryParser($Indexer, $query){
608    /**
609     * parse a search query and transform it into intermediate representation
610     *
611     * in a search query, you can use the following expressions:
612     *
613     *   words:
614     *     include
615     *     -exclude
616     *   phrases:
617     *     "phrase to be included"
618     *     -"phrase you want to exclude"
619     *   namespaces:
620     *     @include:namespace (or ns:include:namespace)
621     *     ^exclude:namespace (or -ns:exclude:namespace)
622     *   groups:
623     *     ()
624     *     -()
625     *   operators:
626     *     and ('and' is the default operator: you can always omit this)
627     *     or  (or pipe symbol '|', lower precedence than 'and')
628     *
629     * e.g. a query [ aa "bb cc" @dd:ee ] means "search pages which contain
630     *      a word 'aa', a phrase 'bb cc' and are within a namespace 'dd:ee'".
631     *      this query is equivalent to [ -(-aa or -"bb cc" or -ns:dd:ee) ]
632     *      as long as you don't mind hit counts.
633     *
634     * intermediate representation consists of the following parts:
635     *
636     *   ( )           - group
637     *   AND           - logical and
638     *   OR            - logical or
639     *   NOT           - logical not
640     *   W+:, W-:, W_: - word      (underscore: no need to highlight)
641     *   P+:, P-:      - phrase    (minus sign: logically in NOT group)
642     *   N+:, N-:      - namespace
643     */
644    $parsed_query = '';
645    $parens_level = 0;
646    $terms = preg_split('/(-?".*?")/u', utf8_strtolower($query), -1, PREG_SPLIT_DELIM_CAPTURE | PREG_SPLIT_NO_EMPTY);
647
648    foreach ($terms as $term) {
649        $parsed = '';
650        if (preg_match('/^(-?)"(.+)"$/u', $term, $matches)) {
651            // phrase-include and phrase-exclude
652            $not = $matches[1] ? 'NOT' : '';
653            $parsed = $not.ft_termParser($Indexer, $matches[2], false, true);
654        } else {
655            // fix incomplete phrase
656            $term = str_replace('"', ' ', $term);
657
658            // fix parentheses
659            $term = str_replace(')'  , ' ) ', $term);
660            $term = str_replace('('  , ' ( ', $term);
661            $term = str_replace('- (', ' -(', $term);
662
663            // treat pipe symbols as 'OR' operators
664            $term = str_replace('|', ' or ', $term);
665
666            // treat ideographic spaces (U+3000) as search term separators
667            // FIXME: some more separators?
668            $term = preg_replace('/[ \x{3000}]+/u', ' ',  $term);
669            $term = trim($term);
670            if ($term === '') continue;
671
672            $tokens = explode(' ', $term);
673            foreach ($tokens as $token) {
674                if ($token === '(') {
675                    // parenthesis-include-open
676                    $parsed .= '(';
677                    ++$parens_level;
678                } elseif ($token === '-(') {
679                    // parenthesis-exclude-open
680                    $parsed .= 'NOT(';
681                    ++$parens_level;
682                } elseif ($token === ')') {
683                    // parenthesis-any-close
684                    if ($parens_level === 0) continue;
685                    $parsed .= ')';
686                    $parens_level--;
687                } elseif ($token === 'and') {
688                    // logical-and (do nothing)
689                } elseif ($token === 'or') {
690                    // logical-or
691                    $parsed .= 'OR';
692                } elseif (preg_match('/^(?:\^|-ns:)(.+)$/u', $token, $matches)) {
693                    // namespace-exclude
694                    $parsed .= 'NOT(N+:'.$matches[1].')';
695                } elseif (preg_match('/^(?:@|ns:)(.+)$/u', $token, $matches)) {
696                    // namespace-include
697                    $parsed .= '(N+:'.$matches[1].')';
698                } elseif (preg_match('/^-(.+)$/', $token, $matches)) {
699                    // word-exclude
700                    $parsed .= 'NOT('.ft_termParser($Indexer, $matches[1]).')';
701                } else {
702                    // word-include
703                    $parsed .= ft_termParser($Indexer, $token);
704                }
705            }
706        }
707        $parsed_query .= $parsed;
708    }
709
710    // cleanup (very sensitive)
711    $parsed_query .= str_repeat(')', $parens_level);
712    do {
713        $parsed_query_old = $parsed_query;
714        $parsed_query = preg_replace('/(NOT)?\(\)/u', '', $parsed_query);
715    } while ($parsed_query !== $parsed_query_old);
716    $parsed_query = preg_replace('/(NOT|OR)+\)/u', ')'      , $parsed_query);
717    $parsed_query = preg_replace('/(OR)+/u'      , 'OR'     , $parsed_query);
718    $parsed_query = preg_replace('/\(OR/u'       , '('      , $parsed_query);
719    $parsed_query = preg_replace('/^OR|OR$/u'    , ''       , $parsed_query);
720    $parsed_query = preg_replace('/\)(NOT)?\(/u' , ')AND$1(', $parsed_query);
721
722    // adjustment: make highlightings right
723    $parens_level     = 0;
724    $notgrp_levels    = array();
725    $parsed_query_new = '';
726    $tokens = preg_split('/(NOT\(|[()])/u', $parsed_query, -1, PREG_SPLIT_DELIM_CAPTURE | PREG_SPLIT_NO_EMPTY);
727    foreach ($tokens as $token) {
728        if ($token === 'NOT(') {
729            $notgrp_levels[] = ++$parens_level;
730        } elseif ($token === '(') {
731            ++$parens_level;
732        } elseif ($token === ')') {
733            if ($parens_level-- === end($notgrp_levels)) array_pop($notgrp_levels);
734        } elseif (count($notgrp_levels) % 2 === 1) {
735            // turn highlight-flag off if terms are logically in "NOT" group
736            $token = preg_replace('/([WPN])\+\:/u', '$1-:', $token);
737        }
738        $parsed_query_new .= $token;
739    }
740    $parsed_query = $parsed_query_new;
741
742    /**
743     * convert infix notation string into postfix (Reverse Polish notation) array
744     * by Shunting-yard algorithm
745     *
746     * see: http://en.wikipedia.org/wiki/Reverse_Polish_notation
747     * see: http://en.wikipedia.org/wiki/Shunting-yard_algorithm
748     */
749    $parsed_ary     = array();
750    $ope_stack      = array();
751    $ope_precedence = array(')' => 1, 'OR' => 2, 'AND' => 3, 'NOT' => 4, '(' => 5);
752    $ope_regex      = '/([()]|OR|AND|NOT)/u';
753
754    $tokens = preg_split($ope_regex, $parsed_query, -1, PREG_SPLIT_DELIM_CAPTURE | PREG_SPLIT_NO_EMPTY);
755    foreach ($tokens as $token) {
756        if (preg_match($ope_regex, $token)) {
757            // operator
758            $last_ope = end($ope_stack);
759            while ($last_ope !== false && $ope_precedence[$token] <= $ope_precedence[$last_ope] && $last_ope != '(') {
760                $parsed_ary[] = array_pop($ope_stack);
761                $last_ope = end($ope_stack);
762            }
763            if ($token == ')') {
764                array_pop($ope_stack); // this array_pop always deletes '('
765            } else {
766                $ope_stack[] = $token;
767            }
768        } else {
769            // operand
770            $token_decoded = str_replace(array('OP', 'CP'), array('(', ')'), $token);
771            $parsed_ary[] = $token_decoded;
772        }
773    }
774    $parsed_ary = array_values(array_merge($parsed_ary, array_reverse($ope_stack)));
775
776    // cleanup: each double "NOT" in RPN array actually does nothing
777    $parsed_ary_count = count($parsed_ary);
778    for ($i = 1; $i < $parsed_ary_count; ++$i) {
779        if ($parsed_ary[$i] === 'NOT' && $parsed_ary[$i - 1] === 'NOT') {
780            unset($parsed_ary[$i], $parsed_ary[$i - 1]);
781        }
782    }
783    $parsed_ary = array_values($parsed_ary);
784
785    // build return value
786    $q = array();
787    $q['query']      = $query;
788    $q['parsed_str'] = $parsed_query;
789    $q['parsed_ary'] = $parsed_ary;
790
791    foreach ($q['parsed_ary'] as $token) {
792        if ($token[2] !== ':') continue;
793        $body = substr($token, 3);
794
795        switch (substr($token, 0, 3)) {
796            case 'N+:':
797                     $q['ns'][]        = $body; // for backward compatibility
798                     break;
799            case 'N-:':
800                     $q['notns'][]     = $body; // for backward compatibility
801                     break;
802            case 'W_:':
803                     $q['words'][]     = $body;
804                     break;
805            case 'W-:':
806                     $q['words'][]     = $body;
807                     $q['not'][]       = $body; // for backward compatibility
808                     break;
809            case 'W+:':
810                     $q['words'][]     = $body;
811                     $q['highlight'][] = $body;
812                     $q['and'][]       = $body; // for backward compatibility
813                     break;
814            case 'P-:':
815                     $q['phrases'][]   = $body;
816                     break;
817            case 'P+:':
818                     $q['phrases'][]   = $body;
819                     $q['highlight'][] = $body;
820                     break;
821        }
822    }
823    foreach (array('words', 'phrases', 'highlight', 'ns', 'notns', 'and', 'not') as $key) {
824        $q[$key] = empty($q[$key]) ? array() : array_values(array_unique($q[$key]));
825    }
826
827    return $q;
828}
829
830/**
831 * Transforms given search term into intermediate representation
832 *
833 * This function is used in ft_queryParser() and not for general purpose use.
834 *
835 * @author Kazutaka Miyasaka <kazmiya@gmail.com>
836 *
837 * @param Doku_Indexer $Indexer
838 * @param string       $term
839 * @param bool         $consider_asian
840 * @param bool         $phrase_mode
841 * @return string
842 */
843function ft_termParser($Indexer, $term, $consider_asian = true, $phrase_mode = false) {
844    $parsed = '';
845    if ($consider_asian) {
846        // successive asian characters need to be searched as a phrase
847        $words = preg_split('/('.IDX_ASIAN.'+)/u', $term, -1, PREG_SPLIT_DELIM_CAPTURE | PREG_SPLIT_NO_EMPTY);
848        foreach ($words as $word) {
849            $phrase_mode = $phrase_mode ? true : preg_match('/'.IDX_ASIAN.'/u', $word);
850            $parsed .= ft_termParser($Indexer, $word, false, $phrase_mode);
851        }
852    } else {
853        $term_noparen = str_replace(array('(', ')'), ' ', $term);
854        $words = $Indexer->tokenizer($term_noparen, true);
855
856        // W_: no need to highlight
857        if (empty($words)) {
858            $parsed = '()'; // important: do not remove
859        } elseif ($words[0] === $term) {
860            $parsed = '(W+:'.$words[0].')';
861        } elseif ($phrase_mode) {
862            $term_encoded = str_replace(array('(', ')'), array('OP', 'CP'), $term);
863            $parsed = '((W_:'.implode(')(W_:', $words).')(P+:'.$term_encoded.'))';
864        } else {
865            $parsed = '((W+:'.implode(')(W+:', $words).'))';
866        }
867    }
868    return $parsed;
869}
870
871/**
872 * Recreate a search query string based on parsed parts, doesn't support negated phrases and `OR` searches
873 *
874 * @param array $and
875 * @param array $not
876 * @param array $phrases
877 * @param array $ns
878 * @param array $notns
879 *
880 * @return string
881 */
882function ft_queryUnparser_simple(array $and, array $not, array $phrases, array $ns, array $notns) {
883    $query = implode(' ', $and);
884    if (!empty($not)) {
885        $query .= ' -' . implode(' -', $not);
886    }
887
888    if (!empty($phrases)) {
889        $query .= ' "' . implode('" "', $phrases) . '"';
890    }
891
892    if (!empty($ns)) {
893        $query .= ' @' . implode(' @', $ns);
894    }
895
896    if (!empty($notns)) {
897        $query .= ' ^' . implode(' ^', $notns);
898    }
899
900    return $query;
901}
902
903//Setup VIM: ex: et ts=4 :
904