inc/fulltext.php

   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
   9 use dokuwiki\Extension\Event;
  10
  11 /**
  12  * create snippets for the first few results only
  13  */
  14 if(!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  * @param string     $sort
  26  * @param int|string $after  only show results with mtime after this date, accepts timestap or strtotime arguments
  27  * @param int|string $before only show results with mtime before this date, accepts timestap or strtotime arguments
  28  *
  29  * @return array
  30  */
  31 function ft_pageSearch($query,&$highlight, $sort = null, $after = null, $before = null){
  32
  33     if ($sort === null) {
  34         $sort = 'hits';
  35     }
  36     $data = [
  37         'query' => $query,
  38         'sort' => $sort,
  39         'after' => $after,
  40         'before' => $before
  41     ];
  42     $data['highlight'] =& $highlight;
  43
  44     return Event::createAndTrigger('SEARCH_QUERY_FULLPAGE', $data, '_ft_pageSearch');
  45 }
  46
  47 /**
  48  * Returns a list of matching documents for the given query
  49  *
  50  * @author Andreas Gohr <andi@splitbrain.org>
  51  * @author Kazutaka Miyasaka <kazmiya@gmail.com>
  52  *
  53  * @param array $data event data
  54  * @return array matching documents
  55  */
  56 function _ft_pageSearch(&$data) {
  57     $Indexer = idx_get_indexer();
  58
  59     // parse the given query
  60     $q = ft_queryParser($Indexer, $data['query']);
  61     $data['highlight'] = $q['highlight'];
  62
  63     if (empty($q['parsed_ary'])) return array();
  64
  65     // lookup all words found in the query
  66     $lookup = $Indexer->lookup($q['words']);
  67
  68     // get all pages in this dokuwiki site (!: includes nonexistent pages)
  69     $pages_all = array();
  70     foreach ($Indexer->getPages() as $id) {
  71         $pages_all[$id] = 0; // base: 0 hit
  72     }
  73
  74     // process the query
  75     $stack = array();
  76     foreach ($q['parsed_ary'] as $token) {
  77         switch (substr($token, 0, 3)) {
  78             case 'W+:':
  79             case 'W-:':
  80             case 'W_:': // word
  81                 $word    = substr($token, 3);
  82                 $stack[] = (array) $lookup[$word];
  83                 break;
  84             case 'P+:':
  85             case 'P-:': // phrase
  86                 $phrase = substr($token, 3);
  87                 // since phrases are always parsed as ((W1)(W2)...(P)),
  88                 // the end($stack) always points the pages that contain
  89                 // all words in this phrase
  90                 $pages  = end($stack);
  91                 $pages_matched = array();
  92                 foreach(array_keys($pages) as $id){
  93                     $evdata = array(
  94                         'id' => $id,
  95                         'phrase' => $phrase,
  96                         'text' => rawWiki($id)
  97                     );
  98                     $evt = new Event('FULLTEXT_PHRASE_MATCH',$evdata);
  99                     if ($evt->advise_before() && $evt->result !== true) {
 100                         $text = utf8_strtolower($evdata['text']);
 101                         if (strpos($text, $phrase) !== false) {
 102                             $evt->result = true;
 103                         }
 104                     }
 105                     $evt->advise_after();
 106                     if ($evt->result === true) {
 107                         $pages_matched[$id] = 0; // phrase: always 0 hit
 108                     }
 109                 }
 110                 $stack[] = $pages_matched;
 111                 break;
 112             case 'N+:':
 113             case 'N-:': // namespace
 114                 $ns = cleanID(substr($token, 3)) . ':';
 115                 $pages_matched = array();
 116                 foreach (array_keys($pages_all) as $id) {
 117                     if (strpos($id, $ns) === 0) {
 118                         $pages_matched[$id] = 0; // namespace: always 0 hit
 119                     }
 120                 }
 121                 $stack[] = $pages_matched;
 122                 break;
 123             case 'AND': // and operation
 124                 list($pages1, $pages2) = array_splice($stack, -2);
 125                 $stack[] = ft_resultCombine(array($pages1, $pages2));
 126                 break;
 127             case 'OR':  // or operation
 128                 list($pages1, $pages2) = array_splice($stack, -2);
 129                 $stack[] = ft_resultUnite(array($pages1, $pages2));
 130                 break;
 131             case 'NOT': // not operation (unary)
 132                 $pages   = array_pop($stack);
 133                 $stack[] = ft_resultComplement(array($pages_all, $pages));
 134                 break;
 135         }
 136     }
 137     $docs = array_pop($stack);
 138
 139     if (empty($docs)) return array();
 140
 141     // check: settings, acls, existence
 142     foreach (array_keys($docs) as $id) {
 143         if (isHiddenPage($id) || auth_quickaclcheck($id) < AUTH_READ || !page_exists($id, '', false)) {
 144             unset($docs[$id]);
 145         }
 146     }
 147
 148     $docs = _ft_filterResultsByTime($docs, $data['after'], $data['before']);
 149
 150     if ($data['sort'] === 'mtime') {
 151         uksort($docs, 'ft_pagemtimesorter');
 152     } else {
 153         // sort docs by count
 154         arsort($docs);
 155     }
 156
 157     return $docs;
 158 }
 159
 160 /**
 161  * Returns the backlinks for a given page
 162  *
 163  * Uses the metadata index.
 164  *
 165  * @param string $id           The id for which links shall be returned
 166  * @param bool   $ignore_perms Ignore the fact that pages are hidden or read-protected
 167  * @return array The pages that contain links to the given page
 168  */
 169 function ft_backlinks($id, $ignore_perms = false){
 170     $result = idx_get_indexer()->lookupKey('relation_references', $id);
 171
 172     if(!count($result)) return $result;
 173
 174     // check ACL permissions
 175     foreach(array_keys($result) as $idx){
 176         if(($ignore_perms !== true && (
 177                 isHiddenPage($result[$idx]) || auth_quickaclcheck($result[$idx]) < AUTH_READ
 178             )) || !page_exists($result[$idx], '', false)){
 179             unset($result[$idx]);
 180         }
 181     }
 182
 183     sort($result);
 184     return $result;
 185 }
 186
 187 /**
 188  * Returns the pages that use a given media file
 189  *
 190  * Uses the relation media metadata property and the metadata index.
 191  *
 192  * Note that before 2013-07-31 the second parameter was the maximum number of results and
 193  * permissions were ignored. That's why the parameter is now checked to be explicitely set
 194  * to true (with type bool) in order to be compatible with older uses of the function.
 195  *
 196  * @param string $id           The media id to look for
 197  * @param bool   $ignore_perms Ignore hidden pages and acls (optional, default: false)
 198  * @return array A list of pages that use the given media file
 199  */
 200 function ft_mediause($id, $ignore_perms = false){
 201     $result = idx_get_indexer()->lookupKey('relation_media', $id);
 202
 203     if(!count($result)) return $result;
 204
 205     // check ACL permissions
 206     foreach(array_keys($result) as $idx){
 207         if(($ignore_perms !== true && (
 208                     isHiddenPage($result[$idx]) || auth_quickaclcheck($result[$idx]) < AUTH_READ
 209                 )) || !page_exists($result[$idx], '', false)){
 210             unset($result[$idx]);
 211         }
 212     }
 213
 214     sort($result);
 215     return $result;
 216 }
 217
 218
 219 /**
 220  * Quicksearch for pagenames
 221  *
 222  * By default it only matches the pagename and ignores the
 223  * namespace. This can be changed with the second parameter.
 224  * The third parameter allows to search in titles as well.
 225  *
 226  * The function always returns titles as well
 227  *
 228  * @triggers SEARCH_QUERY_PAGELOOKUP
 229  * @author   Andreas Gohr <andi@splitbrain.org>
 230  * @author   Adrian Lang <lang@cosmocode.de>
 231  *
 232  * @param string     $id       page id
 233  * @param bool       $in_ns    match against namespace as well?
 234  * @param bool       $in_title search in title?
 235  * @param int|string $after    only show results with mtime after this date, accepts timestap or strtotime arguments
 236  * @param int|string $before   only show results with mtime before this date, accepts timestap or strtotime arguments
 237  *
 238  * @return string[]
 239  */
 240 function ft_pageLookup($id, $in_ns=false, $in_title=false, $after = null, $before = null){
 241     $data = [
 242         'id' => $id,
 243         'in_ns' => $in_ns,
 244         'in_title' => $in_title,
 245         'after' => $after,
 246         'before' => $before
 247     ];
 248     $data['has_titles'] = true; // for plugin backward compatibility check
 249     return Event::createAndTrigger('SEARCH_QUERY_PAGELOOKUP', $data, '_ft_pageLookup');
 250 }
 251
 252 /**
 253  * Returns list of pages as array(pageid => First Heading)
 254  *
 255  * @param array &$data event data
 256  * @return string[]
 257  */
 258 function _ft_pageLookup(&$data){
 259     // split out original parameters
 260     $id = $data['id'];
 261     $Indexer = idx_get_indexer();
 262     $parsedQuery = ft_queryParser($Indexer, $id);
 263     if (count($parsedQuery['ns']) > 0) {
 264         $ns = cleanID($parsedQuery['ns'][0]) . ':';
 265         $id = implode(' ', $parsedQuery['highlight']);
 266     }
 267
 268     $in_ns    = $data['in_ns'];
 269     $in_title = $data['in_title'];
 270     $cleaned = cleanID($id);
 271
 272     $Indexer = idx_get_indexer();
 273     $page_idx = $Indexer->getPages();
 274
 275     $pages = array();
 276     if ($id !== '' && $cleaned !== '') {
 277         foreach ($page_idx as $p_id) {
 278             if ((strpos($in_ns ? $p_id : noNSorNS($p_id), $cleaned) !== false)) {
 279                 if (!isset($pages[$p_id]))
 280                     $pages[$p_id] = p_get_first_heading($p_id, METADATA_DONT_RENDER);
 281             }
 282         }
 283         if ($in_title) {
 284             foreach ($Indexer->lookupKey('title', $id, '_ft_pageLookupTitleCompare') as $p_id) {
 285                 if (!isset($pages[$p_id]))
 286                     $pages[$p_id] = p_get_first_heading($p_id, METADATA_DONT_RENDER);
 287             }
 288         }
 289     }
 290
 291     if (isset($ns)) {
 292         foreach (array_keys($pages) as $p_id) {
 293             if (strpos($p_id, $ns) !== 0) {
 294                 unset($pages[$p_id]);
 295             }
 296         }
 297     }
 298
 299     // discard hidden pages
 300     // discard nonexistent pages
 301     // check ACL permissions
 302     foreach(array_keys($pages) as $idx){
 303         if(!isVisiblePage($idx) || !page_exists($idx) ||
 304            auth_quickaclcheck($idx) < AUTH_READ) {
 305             unset($pages[$idx]);
 306         }
 307     }
 308
 309     $pages = _ft_filterResultsByTime($pages, $data['after'], $data['before']);
 310
 311     uksort($pages,'ft_pagesorter');
 312     return $pages;
 313 }
 314
 315
 316 /**
 317  * @param array      $results search results in the form pageid => value
 318  * @param int|string $after   only returns results with mtime after this date, accepts timestap or strtotime arguments
 319  * @param int|string $before  only returns results with mtime after this date, accepts timestap or strtotime arguments
 320  *
 321  * @return array
 322  */
 323 function _ft_filterResultsByTime(array $results, $after, $before) {
 324     if ($after || $before) {
 325         $after = is_int($after) ? $after : strtotime($after);
 326         $before = is_int($before) ? $before : strtotime($before);
 327
 328         foreach ($results as $id => $value) {
 329             $mTime = filemtime(wikiFN($id));
 330             if ($after && $after > $mTime) {
 331                 unset($results[$id]);
 332                 continue;
 333             }
 334             if ($before && $before < $mTime) {
 335                 unset($results[$id]);
 336             }
 337         }
 338     }
 339
 340     return $results;
 341 }
 342
 343 /**
 344  * Tiny helper function for comparing the searched title with the title
 345  * from the search index. This function is a wrapper around stripos with
 346  * adapted argument order and return value.
 347  *
 348  * @param string $search searched title
 349  * @param string $title  title from index
 350  * @return bool
 351  */
 352 function _ft_pageLookupTitleCompare($search, $title) {
 353     return stripos($title, $search) !== false;
 354 }
 355
 356 /**
 357  * Sort pages based on their namespace level first, then on their string
 358  * values. This makes higher hierarchy pages rank higher than lower hierarchy
 359  * pages.
 360  *
 361  * @param string $a
 362  * @param string $b
 363  * @return int Returns < 0 if $a is less than $b; > 0 if $a is greater than $b, and 0 if they are equal.
 364  */
 365 function ft_pagesorter($a, $b){
 366     $ac = count(explode(':',$a));
 367     $bc = count(explode(':',$b));
 368     if($ac < $bc){
 369         return -1;
 370     }elseif($ac > $bc){
 371         return 1;
 372     }
 373     return strcmp ($a,$b);
 374 }
 375
 376 /**
 377  * Sort pages by their mtime, from newest to oldest
 378  *
 379  * @param string $a
 380  * @param string $b
 381  *
 382  * @return int Returns < 0 if $a is newer than $b, > 0 if $b is newer than $a and 0 if they are of the same age
 383  */
 384 function ft_pagemtimesorter($a, $b) {
 385     $mtimeA = filemtime(wikiFN($a));
 386     $mtimeB = filemtime(wikiFN($b));
 387     return $mtimeB - $mtimeA;
 388 }
 389
 390 /**
 391  * Creates a snippet extract
 392  *
 393  * @author Andreas Gohr <andi@splitbrain.org>
 394  * @triggers FULLTEXT_SNIPPET_CREATE
 395  *
 396  * @param string $id page id
 397  * @param array $highlight
 398  * @return mixed
 399  */
 400 function ft_snippet($id,$highlight){
 401     $text = rawWiki($id);
 402     $text = str_replace("\xC2\xAD",'',$text); // remove soft-hyphens
 403     $evdata = array(
 404             'id'        => $id,
 405             'text'      => &$text,
 406             'highlight' => &$highlight,
 407             'snippet'   => '',
 408             );
 409
 410     $evt = new Event('FULLTEXT_SNIPPET_CREATE',$evdata);
 411     if ($evt->advise_before()) {
 412         $match = array();
 413         $snippets = array();
 414         $utf8_offset = $offset = $end = 0;
 415         $len = utf8_strlen($text);
 416
 417         // build a regexp from the phrases to highlight
 418         $re1 = '(' .
 419             join(
 420                 '|',
 421                 array_map(
 422                     'ft_snippet_re_preprocess',
 423                     array_map(
 424                         'preg_quote_cb',
 425                         array_filter((array) $highlight)
 426                     )
 427                 )
 428             ) .
 429             ')';
 430         $re2 = "$re1.{0,75}(?!\\1)$re1";
 431         $re3 = "$re1.{0,45}(?!\\1)$re1.{0,45}(?!\\1)(?!\\2)$re1";
 432
 433         for ($cnt=4; $cnt--;) {
 434             if (0) {
 435             } else if (preg_match('/'.$re3.'/iu',$text,$match,PREG_OFFSET_CAPTURE,$offset)) {
 436             } else if (preg_match('/'.$re2.'/iu',$text,$match,PREG_OFFSET_CAPTURE,$offset)) {
 437             } else if (preg_match('/'.$re1.'/iu',$text,$match,PREG_OFFSET_CAPTURE,$offset)) {
 438             } else {
 439                 break;
 440             }
 441
 442             list($str,$idx) = $match[0];
 443
 444             // convert $idx (a byte offset) into a utf8 character offset
 445             $utf8_idx = utf8_strlen(substr($text,0,$idx));
 446             $utf8_len = utf8_strlen($str);
 447
 448             // establish context, 100 bytes surrounding the match string
 449             // first look to see if we can go 100 either side,
 450             // then drop to 50 adding any excess if the other side can't go to 50,
 451             $pre = min($utf8_idx-$utf8_offset,100);
 452             $post = min($len-$utf8_idx-$utf8_len,100);
 453
 454             if ($pre>50 && $post>50) {
 455                 $pre = $post = 50;
 456             } else if ($pre>50) {
 457                 $pre = min($pre,100-$post);
 458             } else if ($post>50) {
 459                 $post = min($post, 100-$pre);
 460             } else if ($offset == 0) {
 461                 // both are less than 50, means the context is the whole string
 462                 // make it so and break out of this loop - there is no need for the
 463                 // complex snippet calculations
 464                 $snippets = array($text);
 465                 break;
 466             }
 467
 468             // establish context start and end points, try to append to previous
 469             // context if possible
 470             $start = $utf8_idx - $pre;
 471             $append = ($start < $end) ? $end : false;  // still the end of the previous context snippet
 472             $end = $utf8_idx + $utf8_len + $post;      // now set it to the end of this context
 473
 474             if ($append) {
 475                 $snippets[count($snippets)-1] .= utf8_substr($text,$append,$end-$append);
 476             } else {
 477                 $snippets[] = utf8_substr($text,$start,$end-$start);
 478             }
 479
 480             // set $offset for next match attempt
 481             // continue matching after the current match
 482             // if the current match is not the longest possible match starting at the current offset
 483             // this prevents further matching of this snippet but for possible matches of length
 484             // smaller than match length + context (at least 50 characters) this match is part of the context
 485             $utf8_offset = $utf8_idx + $utf8_len;
 486             $offset = $idx + strlen(utf8_substr($text,$utf8_idx,$utf8_len));
 487             $offset = utf8_correctIdx($text,$offset);
 488         }
 489
 490         $m = "\1";
 491         $snippets = preg_replace('/'.$re1.'/iu',$m.'$1'.$m,$snippets);
 492         $snippet = preg_replace(
 493             '/' . $m . '([^' . $m . ']*?)' . $m . '/iu',
 494             '<strong class="search_hit">$1</strong>',
 495             hsc(join('... ', $snippets))
 496         );
 497
 498         $evdata['snippet'] = $snippet;
 499     }
 500     $evt->advise_after();
 501     unset($evt);
 502
 503     return $evdata['snippet'];
 504 }
 505
 506 /**
 507  * Wraps a search term in regex boundary checks.
 508  *
 509  * @param string $term
 510  * @return string
 511  */
 512 function ft_snippet_re_preprocess($term) {
 513     // do not process asian terms where word boundaries are not explicit
 514     if(preg_match('/'.IDX_ASIAN.'/u',$term)){
 515         return $term;
 516     }
 517
 518     if (UTF8_PROPERTYSUPPORT) {
 519         // unicode word boundaries
 520         // see http://stackoverflow.com/a/2449017/172068
 521         $BL = '(?<!\pL)';
 522         $BR = '(?!\pL)';
 523     } else {
 524         // not as correct as above, but at least won't break
 525         $BL = '\b';
 526         $BR = '\b';
 527     }
 528
 529     if(substr($term,0,2) == '\\*'){
 530         $term = substr($term,2);
 531     }else{
 532         $term = $BL.$term;
 533     }
 534
 535     if(substr($term,-2,2) == '\\*'){
 536         $term = substr($term,0,-2);
 537     }else{
 538         $term = $term.$BR;
 539     }
 540
 541     if($term == $BL || $term == $BR || $term == $BL.$BR) $term = '';
 542     return $term;
 543 }
 544
 545 /**
 546  * Combine found documents and sum up their scores
 547  *
 548  * This function is used to combine searched words with a logical
 549  * AND. Only documents available in all arrays are returned.
 550  *
 551  * based upon PEAR's PHP_Compat function for array_intersect_key()
 552  *
 553  * @param array $args An array of page arrays
 554  * @return array
 555  */
 556 function ft_resultCombine($args){
 557     $array_count = count($args);
 558     if($array_count == 1){
 559         return $args[0];
 560     }
 561
 562     $result = array();
 563     if ($array_count > 1) {
 564         foreach ($args[0] as $key => $value) {
 565             $result[$key] = $value;
 566             for ($i = 1; $i !== $array_count; $i++) {
 567                 if (!isset($args[$i][$key])) {
 568                     unset($result[$key]);
 569                     break;
 570                 }
 571                 $result[$key] += $args[$i][$key];
 572             }
 573         }
 574     }
 575     return $result;
 576 }
 577
 578 /**
 579  * Unites found documents and sum up their scores
 580  *
 581  * based upon ft_resultCombine() function
 582  *
 583  * @param array $args An array of page arrays
 584  * @return array
 585  *
 586  * @author Kazutaka Miyasaka <kazmiya@gmail.com>
 587  */
 588 function ft_resultUnite($args) {
 589     $array_count = count($args);
 590     if ($array_count === 1) {
 591         return $args[0];
 592     }
 593
 594     $result = $args[0];
 595     for ($i = 1; $i !== $array_count; $i++) {
 596         foreach (array_keys($args[$i]) as $id) {
 597             $result[$id] += $args[$i][$id];
 598         }
 599     }
 600     return $result;
 601 }
 602
 603 /**
 604  * Computes the difference of documents using page id for comparison
 605  *
 606  * nearly identical to PHP5's array_diff_key()
 607  *
 608  * @param array $args An array of page arrays
 609  * @return array
 610  *
 611  * @author Kazutaka Miyasaka <kazmiya@gmail.com>
 612  */
 613 function ft_resultComplement($args) {
 614     $array_count = count($args);
 615     if ($array_count === 1) {
 616         return $args[0];
 617     }
 618
 619     $result = $args[0];
 620     foreach (array_keys($result) as $id) {
 621         for ($i = 1; $i !== $array_count; $i++) {
 622             if (isset($args[$i][$id])) unset($result[$id]);
 623         }
 624     }
 625     return $result;
 626 }
 627
 628 /**
 629  * Parses a search query and builds an array of search formulas
 630  *
 631  * @author Andreas Gohr <andi@splitbrain.org>
 632  * @author Kazutaka Miyasaka <kazmiya@gmail.com>
 633  *
 634  * @param Doku_Indexer $Indexer
 635  * @param string $query search query
 636  * @return array of search formulas
 637  */
 638 function ft_queryParser($Indexer, $query){
 639     /**
 640      * parse a search query and transform it into intermediate representation
 641      *
 642      * in a search query, you can use the following expressions:
 643      *
 644      *   words:
 645      *     include
 646      *     -exclude
 647      *   phrases:
 648      *     "phrase to be included"
 649      *     -"phrase you want to exclude"
 650      *   namespaces:
 651      *     @include:namespace (or ns:include:namespace)
 652      *     ^exclude:namespace (or -ns:exclude:namespace)
 653      *   groups:
 654      *     ()
 655      *     -()
 656      *   operators:
 657      *     and ('and' is the default operator: you can always omit this)
 658      *     or  (or pipe symbol '|', lower precedence than 'and')
 659      *
 660      * e.g. a query [ aa "bb cc" @dd:ee ] means "search pages which contain
 661      *      a word 'aa', a phrase 'bb cc' and are within a namespace 'dd:ee'".
 662      *      this query is equivalent to [ -(-aa or -"bb cc" or -ns:dd:ee) ]
 663      *      as long as you don't mind hit counts.
 664      *
 665      * intermediate representation consists of the following parts:
 666      *
 667      *   ( )           - group
 668      *   AND           - logical and
 669      *   OR            - logical or
 670      *   NOT           - logical not
 671      *   W+:, W-:, W_: - word      (underscore: no need to highlight)
 672      *   P+:, P-:      - phrase    (minus sign: logically in NOT group)
 673      *   N+:, N-:      - namespace
 674      */
 675     $parsed_query = '';
 676     $parens_level = 0;
 677     $terms = preg_split('/(-?".*?")/u', utf8_strtolower($query), -1, PREG_SPLIT_DELIM_CAPTURE | PREG_SPLIT_NO_EMPTY);
 678
 679     foreach ($terms as $term) {
 680         $parsed = '';
 681         if (preg_match('/^(-?)"(.+)"$/u', $term, $matches)) {
 682             // phrase-include and phrase-exclude
 683             $not = $matches[1] ? 'NOT' : '';
 684             $parsed = $not.ft_termParser($Indexer, $matches[2], false, true);
 685         } else {
 686             // fix incomplete phrase
 687             $term = str_replace('"', ' ', $term);
 688
 689             // fix parentheses
 690             $term = str_replace(')'  , ' ) ', $term);
 691             $term = str_replace('('  , ' ( ', $term);
 692             $term = str_replace('- (', ' -(', $term);
 693
 694             // treat pipe symbols as 'OR' operators
 695             $term = str_replace('|', ' or ', $term);
 696
 697             // treat ideographic spaces (U+3000) as search term separators
 698             // FIXME: some more separators?
 699             $term = preg_replace('/[ \x{3000}]+/u', ' ',  $term);
 700             $term = trim($term);
 701             if ($term === '') continue;
 702
 703             $tokens = explode(' ', $term);
 704             foreach ($tokens as $token) {
 705                 if ($token === '(') {
 706                     // parenthesis-include-open
 707                     $parsed .= '(';
 708                     ++$parens_level;
 709                 } elseif ($token === '-(') {
 710                     // parenthesis-exclude-open
 711                     $parsed .= 'NOT(';
 712                     ++$parens_level;
 713                 } elseif ($token === ')') {
 714                     // parenthesis-any-close
 715                     if ($parens_level === 0) continue;
 716                     $parsed .= ')';
 717                     $parens_level--;
 718                 } elseif ($token === 'and') {
 719                     // logical-and (do nothing)
 720                 } elseif ($token === 'or') {
 721                     // logical-or
 722                     $parsed .= 'OR';
 723                 } elseif (preg_match('/^(?:\^|-ns:)(.+)$/u', $token, $matches)) {
 724                     // namespace-exclude
 725                     $parsed .= 'NOT(N+:'.$matches[1].')';
 726                 } elseif (preg_match('/^(?:@|ns:)(.+)$/u', $token, $matches)) {
 727                     // namespace-include
 728                     $parsed .= '(N+:'.$matches[1].')';
 729                 } elseif (preg_match('/^-(.+)$/', $token, $matches)) {
 730                     // word-exclude
 731                     $parsed .= 'NOT('.ft_termParser($Indexer, $matches[1]).')';
 732                 } else {
 733                     // word-include
 734                     $parsed .= ft_termParser($Indexer, $token);
 735                 }
 736             }
 737         }
 738         $parsed_query .= $parsed;
 739     }
 740
 741     // cleanup (very sensitive)
 742     $parsed_query .= str_repeat(')', $parens_level);
 743     do {
 744         $parsed_query_old = $parsed_query;
 745         $parsed_query = preg_replace('/(NOT)?\(\)/u', '', $parsed_query);
 746     } while ($parsed_query !== $parsed_query_old);
 747     $parsed_query = preg_replace('/(NOT|OR)+\)/u', ')'      , $parsed_query);
 748     $parsed_query = preg_replace('/(OR)+/u'      , 'OR'     , $parsed_query);
 749     $parsed_query = preg_replace('/\(OR/u'       , '('      , $parsed_query);
 750     $parsed_query = preg_replace('/^OR|OR$/u'    , ''       , $parsed_query);
 751     $parsed_query = preg_replace('/\)(NOT)?\(/u' , ')AND$1(', $parsed_query);
 752
 753     // adjustment: make highlightings right
 754     $parens_level     = 0;
 755     $notgrp_levels    = array();
 756     $parsed_query_new = '';
 757     $tokens = preg_split('/(NOT\(|[()])/u', $parsed_query, -1, PREG_SPLIT_DELIM_CAPTURE | PREG_SPLIT_NO_EMPTY);
 758     foreach ($tokens as $token) {
 759         if ($token === 'NOT(') {
 760             $notgrp_levels[] = ++$parens_level;
 761         } elseif ($token === '(') {
 762             ++$parens_level;
 763         } elseif ($token === ')') {
 764             if ($parens_level-- === end($notgrp_levels)) array_pop($notgrp_levels);
 765         } elseif (count($notgrp_levels) % 2 === 1) {
 766             // turn highlight-flag off if terms are logically in "NOT" group
 767             $token = preg_replace('/([WPN])\+\:/u', '$1-:', $token);
 768         }
 769         $parsed_query_new .= $token;
 770     }
 771     $parsed_query = $parsed_query_new;
 772
 773     /**
 774      * convert infix notation string into postfix (Reverse Polish notation) array
 775      * by Shunting-yard algorithm
 776      *
 777      * see: http://en.wikipedia.org/wiki/Reverse_Polish_notation
 778      * see: http://en.wikipedia.org/wiki/Shunting-yard_algorithm
 779      */
 780     $parsed_ary     = array();
 781     $ope_stack      = array();
 782     $ope_precedence = array(')' => 1, 'OR' => 2, 'AND' => 3, 'NOT' => 4, '(' => 5);
 783     $ope_regex      = '/([()]|OR|AND|NOT)/u';
 784
 785     $tokens = preg_split($ope_regex, $parsed_query, -1, PREG_SPLIT_DELIM_CAPTURE | PREG_SPLIT_NO_EMPTY);
 786     foreach ($tokens as $token) {
 787         if (preg_match($ope_regex, $token)) {
 788             // operator
 789             $last_ope = end($ope_stack);
 790             while ($last_ope !== false && $ope_precedence[$token] <= $ope_precedence[$last_ope] && $last_ope != '(') {
 791                 $parsed_ary[] = array_pop($ope_stack);
 792                 $last_ope = end($ope_stack);
 793             }
 794             if ($token == ')') {
 795                 array_pop($ope_stack); // this array_pop always deletes '('
 796             } else {
 797                 $ope_stack[] = $token;
 798             }
 799         } else {
 800             // operand
 801             $token_decoded = str_replace(array('OP', 'CP'), array('(', ')'), $token);
 802             $parsed_ary[] = $token_decoded;
 803         }
 804     }
 805     $parsed_ary = array_values(array_merge($parsed_ary, array_reverse($ope_stack)));
 806
 807     // cleanup: each double "NOT" in RPN array actually does nothing
 808     $parsed_ary_count = count($parsed_ary);
 809     for ($i = 1; $i < $parsed_ary_count; ++$i) {
 810         if ($parsed_ary[$i] === 'NOT' && $parsed_ary[$i - 1] === 'NOT') {
 811             unset($parsed_ary[$i], $parsed_ary[$i - 1]);
 812         }
 813     }
 814     $parsed_ary = array_values($parsed_ary);
 815
 816     // build return value
 817     $q = array();
 818     $q['query']      = $query;
 819     $q['parsed_str'] = $parsed_query;
 820     $q['parsed_ary'] = $parsed_ary;
 821
 822     foreach ($q['parsed_ary'] as $token) {
 823         if ($token[2] !== ':') continue;
 824         $body = substr($token, 3);
 825
 826         switch (substr($token, 0, 3)) {
 827             case 'N+:':
 828                      $q['ns'][]        = $body; // for backward compatibility
 829                      break;
 830             case 'N-:':
 831                      $q['notns'][]     = $body; // for backward compatibility
 832                      break;
 833             case 'W_:':
 834                      $q['words'][]     = $body;
 835                      break;
 836             case 'W-:':
 837                      $q['words'][]     = $body;
 838                      $q['not'][]       = $body; // for backward compatibility
 839                      break;
 840             case 'W+:':
 841                      $q['words'][]     = $body;
 842                      $q['highlight'][] = $body;
 843                      $q['and'][]       = $body; // for backward compatibility
 844                      break;
 845             case 'P-:':
 846                      $q['phrases'][]   = $body;
 847                      break;
 848             case 'P+:':
 849                      $q['phrases'][]   = $body;
 850                      $q['highlight'][] = $body;
 851                      break;
 852         }
 853     }
 854     foreach (array('words', 'phrases', 'highlight', 'ns', 'notns', 'and', 'not') as $key) {
 855         $q[$key] = empty($q[$key]) ? array() : array_values(array_unique($q[$key]));
 856     }
 857
 858     return $q;
 859 }
 860
 861 /**
 862  * Transforms given search term into intermediate representation
 863  *
 864  * This function is used in ft_queryParser() and not for general purpose use.
 865  *
 866  * @author Kazutaka Miyasaka <kazmiya@gmail.com>
 867  *
 868  * @param Doku_Indexer $Indexer
 869  * @param string       $term
 870  * @param bool         $consider_asian
 871  * @param bool         $phrase_mode
 872  * @return string
 873  */
 874 function ft_termParser($Indexer, $term, $consider_asian = true, $phrase_mode = false) {
 875     $parsed = '';
 876     if ($consider_asian) {
 877         // successive asian characters need to be searched as a phrase
 878         $words = preg_split('/('.IDX_ASIAN.'+)/u', $term, -1, PREG_SPLIT_DELIM_CAPTURE | PREG_SPLIT_NO_EMPTY);
 879         foreach ($words as $word) {
 880             $phrase_mode = $phrase_mode ? true : preg_match('/'.IDX_ASIAN.'/u', $word);
 881             $parsed .= ft_termParser($Indexer, $word, false, $phrase_mode);
 882         }
 883     } else {
 884         $term_noparen = str_replace(array('(', ')'), ' ', $term);
 885         $words = $Indexer->tokenizer($term_noparen, true);
 886
 887         // W_: no need to highlight
 888         if (empty($words)) {
 889             $parsed = '()'; // important: do not remove
 890         } elseif ($words[0] === $term) {
 891             $parsed = '(W+:'.$words[0].')';
 892         } elseif ($phrase_mode) {
 893             $term_encoded = str_replace(array('(', ')'), array('OP', 'CP'), $term);
 894             $parsed = '((W_:'.implode(')(W_:', $words).')(P+:'.$term_encoded.'))';
 895         } else {
 896             $parsed = '((W+:'.implode(')(W+:', $words).'))';
 897         }
 898     }
 899     return $parsed;
 900 }
 901
 902 /**
 903  * Recreate a search query string based on parsed parts, doesn't support negated phrases and `OR` searches
 904  *
 905  * @param array $and
 906  * @param array $not
 907  * @param array $phrases
 908  * @param array $ns
 909  * @param array $notns
 910  *
 911  * @return string
 912  */
 913 function ft_queryUnparser_simple(array $and, array $not, array $phrases, array $ns, array $notns) {
 914     $query = implode(' ', $and);
 915     if (!empty($not)) {
 916         $query .= ' -' . implode(' -', $not);
 917     }
 918
 919     if (!empty($phrases)) {
 920         $query .= ' "' . implode('" "', $phrases) . '"';
 921     }
 922
 923     if (!empty($ns)) {
 924         $query .= ' @' . implode(' @', $ns);
 925     }
 926
 927     if (!empty($notns)) {
 928         $query .= ' ^' . implode(' ^', $notns);
 929     }
 930
 931     return $query;
 932 }
 933
 934 //Setup VIM: ex: et ts=4 :