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  */use dokuwiki\Extension\Event;
   8
   9 /**
  10  * create snippets for the first few results only
  11  */
  12 if(!defined('FT_SNIPPET_NUMBER')) define('FT_SNIPPET_NUMBER',15);
  13
  14 /**
  15  * The fulltext search
  16  *
  17  * Returns a list of matching documents for the given query
  18  *
  19  * refactored into ft_pageSearch(), _ft_pageSearch() and trigger_event()
  20  *
  21  * @param string     $query
  22  * @param array      $highlight
  23  * @param string     $sort
  24  * @param int|string $after  only show results with mtime after this date, accepts timestap or strtotime arguments
  25  * @param int|string $before only show results with mtime before this date, accepts timestap or strtotime arguments
  26  *
  27  * @return array
  28  */
  29 function ft_pageSearch($query,&$highlight, $sort = null, $after = null, $before = null){
  30
  31     if ($sort === null) {
  32         $sort = 'hits';
  33     }
  34     $data = [
  35         'query' => $query,
  36         'sort' => $sort,
  37         'after' => $after,
  38         'before' => $before
  39     ];
  40     $data['highlight'] =& $highlight;
  41
  42     return Event::createAndTrigger('SEARCH_QUERY_FULLPAGE', $data, '_ft_pageSearch');
  43 }
  44
  45 /**
  46  * Returns a list of matching documents for the given query
  47  *
  48  * @author Andreas Gohr <andi@splitbrain.org>
  49  * @author Kazutaka Miyasaka <kazmiya@gmail.com>
  50  *
  51  * @param array $data event data
  52  * @return array matching documents
  53  */
  54 function _ft_pageSearch(&$data) {
  55     $Indexer = idx_get_indexer();
  56
  57     // parse the given query
  58     $q = ft_queryParser($Indexer, $data['query']);
  59     $data['highlight'] = $q['highlight'];
  60
  61     if (empty($q['parsed_ary'])) return array();
  62
  63     // lookup all words found in the query
  64     $lookup = $Indexer->lookup($q['words']);
  65
  66     // get all pages in this dokuwiki site (!: includes nonexistent pages)
  67     $pages_all = array();
  68     foreach ($Indexer->getPages() as $id) {
  69         $pages_all[$id] = 0; // base: 0 hit
  70     }
  71
  72     // process the query
  73     $stack = array();
  74     foreach ($q['parsed_ary'] as $token) {
  75         switch (substr($token, 0, 3)) {
  76             case 'W+:':
  77             case 'W-:':
  78             case 'W_:': // word
  79                 $word    = substr($token, 3);
  80                 $stack[] = (array) $lookup[$word];
  81                 break;
  82             case 'P+:':
  83             case 'P-:': // phrase
  84                 $phrase = substr($token, 3);
  85                 // since phrases are always parsed as ((W1)(W2)...(P)),
  86                 // the end($stack) always points the pages that contain
  87                 // all words in this phrase
  88                 $pages  = end($stack);
  89                 $pages_matched = array();
  90                 foreach(array_keys($pages) as $id){
  91                     $evdata = array(
  92                         'id' => $id,
  93                         'phrase' => $phrase,
  94                         'text' => rawWiki($id)
  95                     );
  96                     $evt = new Event('FULLTEXT_PHRASE_MATCH',$evdata);
  97                     if ($evt->advise_before() && $evt->result !== true) {
  98                         $text = \dokuwiki\Utf8\PhpString::strtolower($evdata['text']);
  99                         if (strpos($text, $phrase) !== false) {
 100                             $evt->result = true;
 101                         }
 102                     }
 103                     $evt->advise_after();
 104                     if ($evt->result === true) {
 105                         $pages_matched[$id] = 0; // phrase: always 0 hit
 106                     }
 107                 }
 108                 $stack[] = $pages_matched;
 109                 break;
 110             case 'N+:':
 111             case 'N-:': // namespace
 112                 $ns = cleanID(substr($token, 3)) . ':';
 113                 $pages_matched = array();
 114                 foreach (array_keys($pages_all) as $id) {
 115                     if (strpos($id, $ns) === 0) {
 116                         $pages_matched[$id] = 0; // namespace: always 0 hit
 117                     }
 118                 }
 119                 $stack[] = $pages_matched;
 120                 break;
 121             case 'AND': // and operation
 122                 list($pages1, $pages2) = array_splice($stack, -2);
 123                 $stack[] = ft_resultCombine(array($pages1, $pages2));
 124                 break;
 125             case 'OR':  // or operation
 126                 list($pages1, $pages2) = array_splice($stack, -2);
 127                 $stack[] = ft_resultUnite(array($pages1, $pages2));
 128                 break;
 129             case 'NOT': // not operation (unary)
 130                 $pages   = array_pop($stack);
 131                 $stack[] = ft_resultComplement(array($pages_all, $pages));
 132                 break;
 133         }
 134     }
 135     $docs = array_pop($stack);
 136
 137     if (empty($docs)) return array();
 138
 139     // check: settings, acls, existence
 140     foreach (array_keys($docs) as $id) {
 141         if (isHiddenPage($id) || auth_quickaclcheck($id) < AUTH_READ || !page_exists($id, '', false)) {
 142             unset($docs[$id]);
 143         }
 144     }
 145
 146     $docs = _ft_filterResultsByTime($docs, $data['after'], $data['before']);
 147
 148     if ($data['sort'] === 'mtime') {
 149         uksort($docs, 'ft_pagemtimesorter');
 150     } else {
 151         // sort docs by count
 152         arsort($docs);
 153     }
 154
 155     return $docs;
 156 }
 157
 158 /**
 159  * Returns the backlinks for a given page
 160  *
 161  * Uses the metadata index.
 162  *
 163  * @param string $id           The id for which links shall be returned
 164  * @param bool   $ignore_perms Ignore the fact that pages are hidden or read-protected
 165  * @return array The pages that contain links to the given page
 166  */
 167 function ft_backlinks($id, $ignore_perms = false){
 168     $result = idx_get_indexer()->lookupKey('relation_references', $id);
 169
 170     if(!count($result)) return $result;
 171
 172     // check ACL permissions
 173     foreach(array_keys($result) as $idx){
 174         if(($ignore_perms !== true && (
 175                 isHiddenPage($result[$idx]) || auth_quickaclcheck($result[$idx]) < AUTH_READ
 176             )) || !page_exists($result[$idx], '', false)){
 177             unset($result[$idx]);
 178         }
 179     }
 180
 181     sort($result);
 182     return $result;
 183 }
 184
 185 /**
 186  * Returns the pages that use a given media file
 187  *
 188  * Uses the relation media metadata property and the metadata index.
 189  *
 190  * Note that before 2013-07-31 the second parameter was the maximum number of results and
 191  * permissions were ignored. That's why the parameter is now checked to be explicitely set
 192  * to true (with type bool) in order to be compatible with older uses of the function.
 193  *
 194  * @param string $id           The media id to look for
 195  * @param bool   $ignore_perms Ignore hidden pages and acls (optional, default: false)
 196  * @return array A list of pages that use the given media file
 197  */
 198 function ft_mediause($id, $ignore_perms = false){
 199     $result = idx_get_indexer()->lookupKey('relation_media', $id);
 200
 201     if(!count($result)) return $result;
 202
 203     // check ACL permissions
 204     foreach(array_keys($result) as $idx){
 205         if(($ignore_perms !== true && (
 206                     isHiddenPage($result[$idx]) || auth_quickaclcheck($result[$idx]) < AUTH_READ
 207                 )) || !page_exists($result[$idx], '', false)){
 208             unset($result[$idx]);
 209         }
 210     }
 211
 212     sort($result);
 213     return $result;
 214 }
 215
 216
 217 /**
 218  * Quicksearch for pagenames
 219  *
 220  * By default it only matches the pagename and ignores the
 221  * namespace. This can be changed with the second parameter.
 222  * The third parameter allows to search in titles as well.
 223  *
 224  * The function always returns titles as well
 225  *
 226  * @triggers SEARCH_QUERY_PAGELOOKUP
 227  * @author   Andreas Gohr <andi@splitbrain.org>
 228  * @author   Adrian Lang <lang@cosmocode.de>
 229  *
 230  * @param string     $id       page id
 231  * @param bool       $in_ns    match against namespace as well?
 232  * @param bool       $in_title search in title?
 233  * @param int|string $after    only show results with mtime after this date, accepts timestap or strtotime arguments
 234  * @param int|string $before   only show results with mtime before this date, accepts timestap or strtotime arguments
 235  *
 236  * @return string[]
 237  */
 238 function ft_pageLookup($id, $in_ns=false, $in_title=false, $after = null, $before = null){
 239     $data = [
 240         'id' => $id,
 241         'in_ns' => $in_ns,
 242         'in_title' => $in_title,
 243         'after' => $after,
 244         'before' => $before
 245     ];
 246     $data['has_titles'] = true; // for plugin backward compatibility check
 247     return Event::createAndTrigger('SEARCH_QUERY_PAGELOOKUP', $data, '_ft_pageLookup');
 248 }
 249
 250 /**
 251  * Returns list of pages as array(pageid => First Heading)
 252  *
 253  * @param array &$data event data
 254  * @return string[]
 255  */
 256 function _ft_pageLookup(&$data){
 257     // split out original parameters
 258     $id = $data['id'];
 259     $Indexer = idx_get_indexer();
 260     $parsedQuery = ft_queryParser($Indexer, $id);
 261     if (count($parsedQuery['ns']) > 0) {
 262         $ns = cleanID($parsedQuery['ns'][0]) . ':';
 263         $id = implode(' ', $parsedQuery['highlight']);
 264     }
 265
 266     $in_ns    = $data['in_ns'];
 267     $in_title = $data['in_title'];
 268     $cleaned = cleanID($id);
 269
 270     $Indexer = idx_get_indexer();
 271     $page_idx = $Indexer->getPages();
 272
 273     $pages = array();
 274     if ($id !== '' && $cleaned !== '') {
 275         foreach ($page_idx as $p_id) {
 276             if ((strpos($in_ns ? $p_id : noNSorNS($p_id), $cleaned) !== false)) {
 277                 if (!isset($pages[$p_id]))
 278                     $pages[$p_id] = p_get_first_heading($p_id, METADATA_DONT_RENDER);
 279             }
 280         }
 281         if ($in_title) {
 282             foreach ($Indexer->lookupKey('title', $id, '_ft_pageLookupTitleCompare') as $p_id) {
 283                 if (!isset($pages[$p_id]))
 284                     $pages[$p_id] = p_get_first_heading($p_id, METADATA_DONT_RENDER);
 285             }
 286         }
 287     }
 288
 289     if (isset($ns)) {
 290         foreach (array_keys($pages) as $p_id) {
 291             if (strpos($p_id, $ns) !== 0) {
 292                 unset($pages[$p_id]);
 293             }
 294         }
 295     }
 296
 297     // discard hidden pages
 298     // discard nonexistent pages
 299     // check ACL permissions
 300     foreach(array_keys($pages) as $idx){
 301         if(!isVisiblePage($idx) || !page_exists($idx) ||
 302            auth_quickaclcheck($idx) < AUTH_READ) {
 303             unset($pages[$idx]);
 304         }
 305     }
 306
 307     $pages = _ft_filterResultsByTime($pages, $data['after'], $data['before']);
 308
 309     uksort($pages,'ft_pagesorter');
 310     return $pages;
 311 }
 312
 313
 314 /**
 315  * @param array      $results search results in the form pageid => value
 316  * @param int|string $after   only returns results with mtime after this date, accepts timestap or strtotime arguments
 317  * @param int|string $before  only returns results with mtime after this date, accepts timestap or strtotime arguments
 318  *
 319  * @return array
 320  */
 321 function _ft_filterResultsByTime(array $results, $after, $before) {
 322     if ($after || $before) {
 323         $after = is_int($after) ? $after : strtotime($after);
 324         $before = is_int($before) ? $before : strtotime($before);
 325
 326         foreach ($results as $id => $value) {
 327             $mTime = filemtime(wikiFN($id));
 328             if ($after && $after > $mTime) {
 329                 unset($results[$id]);
 330                 continue;
 331             }
 332             if ($before && $before < $mTime) {
 333                 unset($results[$id]);
 334             }
 335         }
 336     }
 337
 338     return $results;
 339 }
 340
 341 /**
 342  * Tiny helper function for comparing the searched title with the title
 343  * from the search index. This function is a wrapper around stripos with
 344  * adapted argument order and return value.
 345  *
 346  * @param string $search searched title
 347  * @param string $title  title from index
 348  * @return bool
 349  */
 350 function _ft_pageLookupTitleCompare($search, $title) {
 351     return stripos($title, $search) !== false;
 352 }
 353
 354 /**
 355  * Sort pages based on their namespace level first, then on their string
 356  * values. This makes higher hierarchy pages rank higher than lower hierarchy
 357  * pages.
 358  *
 359  * @param string $a
 360  * @param string $b
 361  * @return int Returns < 0 if $a is less than $b; > 0 if $a is greater than $b, and 0 if they are equal.
 362  */
 363 function ft_pagesorter($a, $b){
 364     $ac = count(explode(':',$a));
 365     $bc = count(explode(':',$b));
 366     if($ac < $bc){
 367         return -1;
 368     }elseif($ac > $bc){
 369         return 1;
 370     }
 371     return strcmp ($a,$b);
 372 }
 373
 374 /**
 375  * Sort pages by their mtime, from newest to oldest
 376  *
 377  * @param string $a
 378  * @param string $b
 379  *
 380  * @return int Returns < 0 if $a is newer than $b, > 0 if $b is newer than $a and 0 if they are of the same age
 381  */
 382 function ft_pagemtimesorter($a, $b) {
 383     $mtimeA = filemtime(wikiFN($a));
 384     $mtimeB = filemtime(wikiFN($b));
 385     return $mtimeB - $mtimeA;
 386 }
 387
 388 /**
 389  * Creates a snippet extract
 390  *
 391  * @author Andreas Gohr <andi@splitbrain.org>
 392  * @triggers FULLTEXT_SNIPPET_CREATE
 393  *
 394  * @param string $id page id
 395  * @param array $highlight
 396  * @return mixed
 397  */
 398 function ft_snippet($id,$highlight){
 399     $text = rawWiki($id);
 400     $text = str_replace("\xC2\xAD",'',$text); // remove soft-hyphens
 401     $evdata = array(
 402             'id'        => $id,
 403             'text'      => &$text,
 404             'highlight' => &$highlight,
 405             'snippet'   => '',
 406             );
 407
 408     $evt = new Event('FULLTEXT_SNIPPET_CREATE',$evdata);
 409     if ($evt->advise_before()) {
 410         $match = array();
 411         $snippets = array();
 412         $utf8_offset = $offset = $end = 0;
 413         $len = \dokuwiki\Utf8\PhpString::strlen($text);
 414
 415         // build a regexp from the phrases to highlight
 416         $re1 = '(' .
 417             join(
 418                 '|',
 419                 array_map(
 420                     'ft_snippet_re_preprocess',
 421                     array_map(
 422                         'preg_quote_cb',
 423                         array_filter((array) $highlight)
 424                     )
 425                 )
 426             ) .
 427             ')';
 428         $re2 = "$re1.{0,75}(?!\\1)$re1";
 429         $re3 = "$re1.{0,45}(?!\\1)$re1.{0,45}(?!\\1)(?!\\2)$re1";
 430
 431         for ($cnt=4; $cnt--;) {
 432             if (0) {
 433             } else if (preg_match('/'.$re3.'/iu',$text,$match,PREG_OFFSET_CAPTURE,$offset)) {
 434             } else if (preg_match('/'.$re2.'/iu',$text,$match,PREG_OFFSET_CAPTURE,$offset)) {
 435             } else if (preg_match('/'.$re1.'/iu',$text,$match,PREG_OFFSET_CAPTURE,$offset)) {
 436             } else {
 437                 break;
 438             }
 439
 440             list($str,$idx) = $match[0];
 441
 442             // convert $idx (a byte offset) into a utf8 character offset
 443             $utf8_idx = \dokuwiki\Utf8\PhpString::strlen(substr($text,0,$idx));
 444             $utf8_len = \dokuwiki\Utf8\PhpString::strlen($str);
 445
 446             // establish context, 100 bytes surrounding the match string
 447             // first look to see if we can go 100 either side,
 448             // then drop to 50 adding any excess if the other side can't go to 50,
 449             $pre = min($utf8_idx-$utf8_offset,100);
 450             $post = min($len-$utf8_idx-$utf8_len,100);
 451
 452             if ($pre>50 && $post>50) {
 453                 $pre = $post = 50;
 454             } else if ($pre>50) {
 455                 $pre = min($pre,100-$post);
 456             } else if ($post>50) {
 457                 $post = min($post, 100-$pre);
 458             } else if ($offset == 0) {
 459                 // both are less than 50, means the context is the whole string
 460                 // make it so and break out of this loop - there is no need for the
 461                 // complex snippet calculations
 462                 $snippets = array($text);
 463                 break;
 464             }
 465
 466             // establish context start and end points, try to append to previous
 467             // context if possible
 468             $start = $utf8_idx - $pre;
 469             $append = ($start < $end) ? $end : false;  // still the end of the previous context snippet
 470             $end = $utf8_idx + $utf8_len + $post;      // now set it to the end of this context
 471
 472             if ($append) {
 473                 $snippets[count($snippets)-1] .= \dokuwiki\Utf8\PhpString::substr($text,$append,$end-$append);
 474             } else {
 475                 $snippets[] = \dokuwiki\Utf8\PhpString::substr($text,$start,$end-$start);
 476             }
 477
 478             // set $offset for next match attempt
 479             // continue matching after the current match
 480             // if the current match is not the longest possible match starting at the current offset
 481             // this prevents further matching of this snippet but for possible matches of length
 482             // smaller than match length + context (at least 50 characters) this match is part of the context
 483             $utf8_offset = $utf8_idx + $utf8_len;
 484             $offset = $idx + strlen(\dokuwiki\Utf8\PhpString::substr($text,$utf8_idx,$utf8_len));
 485             $offset = \dokuwiki\Utf8\Clean::correctIdx($text,$offset);
 486         }
 487
 488         $m = "\1";
 489         $snippets = preg_replace('/'.$re1.'/iu',$m.'$1'.$m,$snippets);
 490         $snippet = preg_replace(
 491             '/' . $m . '([^' . $m . ']*?)' . $m . '/iu',
 492             '<strong class="search_hit">$1</strong>',
 493             hsc(join('... ', $snippets))
 494         );
 495
 496         $evdata['snippet'] = $snippet;
 497     }
 498     $evt->advise_after();
 499     unset($evt);
 500
 501     return $evdata['snippet'];
 502 }
 503
 504 /**
 505  * Wraps a search term in regex boundary checks.
 506  *
 507  * @param string $term
 508  * @return string
 509  */
 510 function ft_snippet_re_preprocess($term) {
 511     // do not process asian terms where word boundaries are not explicit
 512     if(preg_match('/'.IDX_ASIAN.'/u',$term)){
 513         return $term;
 514     }
 515
 516     if (UTF8_PROPERTYSUPPORT) {
 517         // unicode word boundaries
 518         // see http://stackoverflow.com/a/2449017/172068
 519         $BL = '(?<!\pL)';
 520         $BR = '(?!\pL)';
 521     } else {
 522         // not as correct as above, but at least won't break
 523         $BL = '\b';
 524         $BR = '\b';
 525     }
 526
 527     if(substr($term,0,2) == '\\*'){
 528         $term = substr($term,2);
 529     }else{
 530         $term = $BL.$term;
 531     }
 532
 533     if(substr($term,-2,2) == '\\*'){
 534         $term = substr($term,0,-2);
 535     }else{
 536         $term = $term.$BR;
 537     }
 538
 539     if($term == $BL || $term == $BR || $term == $BL.$BR) $term = '';
 540     return $term;
 541 }
 542
 543 /**
 544  * Combine found documents and sum up their scores
 545  *
 546  * This function is used to combine searched words with a logical
 547  * AND. Only documents available in all arrays are returned.
 548  *
 549  * based upon PEAR's PHP_Compat function for array_intersect_key()
 550  *
 551  * @param array $args An array of page arrays
 552  * @return array
 553  */
 554 function ft_resultCombine($args){
 555     $array_count = count($args);
 556     if($array_count == 1){
 557         return $args[0];
 558     }
 559
 560     $result = array();
 561     if ($array_count > 1) {
 562         foreach ($args[0] as $key => $value) {
 563             $result[$key] = $value;
 564             for ($i = 1; $i !== $array_count; $i++) {
 565                 if (!isset($args[$i][$key])) {
 566                     unset($result[$key]);
 567                     break;
 568                 }
 569                 $result[$key] += $args[$i][$key];
 570             }
 571         }
 572     }
 573     return $result;
 574 }
 575
 576 /**
 577  * Unites found documents and sum up their scores
 578  *
 579  * based upon ft_resultCombine() function
 580  *
 581  * @param array $args An array of page arrays
 582  * @return array
 583  *
 584  * @author Kazutaka Miyasaka <kazmiya@gmail.com>
 585  */
 586 function ft_resultUnite($args) {
 587     $array_count = count($args);
 588     if ($array_count === 1) {
 589         return $args[0];
 590     }
 591
 592     $result = $args[0];
 593     for ($i = 1; $i !== $array_count; $i++) {
 594         foreach (array_keys($args[$i]) as $id) {
 595             $result[$id] += $args[$i][$id];
 596         }
 597     }
 598     return $result;
 599 }
 600
 601 /**
 602  * Computes the difference of documents using page id for comparison
 603  *
 604  * nearly identical to PHP5's array_diff_key()
 605  *
 606  * @param array $args An array of page arrays
 607  * @return array
 608  *
 609  * @author Kazutaka Miyasaka <kazmiya@gmail.com>
 610  */
 611 function ft_resultComplement($args) {
 612     $array_count = count($args);
 613     if ($array_count === 1) {
 614         return $args[0];
 615     }
 616
 617     $result = $args[0];
 618     foreach (array_keys($result) as $id) {
 619         for ($i = 1; $i !== $array_count; $i++) {
 620             if (isset($args[$i][$id])) unset($result[$id]);
 621         }
 622     }
 623     return $result;
 624 }
 625
 626 /**
 627  * Parses a search query and builds an array of search formulas
 628  *
 629  * @author Andreas Gohr <andi@splitbrain.org>
 630  * @author Kazutaka Miyasaka <kazmiya@gmail.com>
 631  *
 632  * @param Doku_Indexer $Indexer
 633  * @param string $query search query
 634  * @return array of search formulas
 635  */
 636 function ft_queryParser($Indexer, $query){
 637     /**
 638      * parse a search query and transform it into intermediate representation
 639      *
 640      * in a search query, you can use the following expressions:
 641      *
 642      *   words:
 643      *     include
 644      *     -exclude
 645      *   phrases:
 646      *     "phrase to be included"
 647      *     -"phrase you want to exclude"
 648      *   namespaces:
 649      *     @include:namespace (or ns:include:namespace)
 650      *     ^exclude:namespace (or -ns:exclude:namespace)
 651      *   groups:
 652      *     ()
 653      *     -()
 654      *   operators:
 655      *     and ('and' is the default operator: you can always omit this)
 656      *     or  (or pipe symbol '|', lower precedence than 'and')
 657      *
 658      * e.g. a query [ aa "bb cc" @dd:ee ] means "search pages which contain
 659      *      a word 'aa', a phrase 'bb cc' and are within a namespace 'dd:ee'".
 660      *      this query is equivalent to [ -(-aa or -"bb cc" or -ns:dd:ee) ]
 661      *      as long as you don't mind hit counts.
 662      *
 663      * intermediate representation consists of the following parts:
 664      *
 665      *   ( )           - group
 666      *   AND           - logical and
 667      *   OR            - logical or
 668      *   NOT           - logical not
 669      *   W+:, W-:, W_: - word      (underscore: no need to highlight)
 670      *   P+:, P-:      - phrase    (minus sign: logically in NOT group)
 671      *   N+:, N-:      - namespace
 672      */
 673     $parsed_query = '';
 674     $parens_level = 0;
 675     $terms = preg_split('/(-?".*?")/u', \dokuwiki\Utf8\PhpString::strtolower($query), -1, PREG_SPLIT_DELIM_CAPTURE | PREG_SPLIT_NO_EMPTY);
 676
 677     foreach ($terms as $term) {
 678         $parsed = '';
 679         if (preg_match('/^(-?)"(.+)"$/u', $term, $matches)) {
 680             // phrase-include and phrase-exclude
 681             $not = $matches[1] ? 'NOT' : '';
 682             $parsed = $not.ft_termParser($Indexer, $matches[2], false, true);
 683         } else {
 684             // fix incomplete phrase
 685             $term = str_replace('"', ' ', $term);
 686
 687             // fix parentheses
 688             $term = str_replace(')'  , ' ) ', $term);
 689             $term = str_replace('('  , ' ( ', $term);
 690             $term = str_replace('- (', ' -(', $term);
 691
 692             // treat pipe symbols as 'OR' operators
 693             $term = str_replace('|', ' or ', $term);
 694
 695             // treat ideographic spaces (U+3000) as search term separators
 696             // FIXME: some more separators?
 697             $term = preg_replace('/[ \x{3000}]+/u', ' ',  $term);
 698             $term = trim($term);
 699             if ($term === '') continue;
 700
 701             $tokens = explode(' ', $term);
 702             foreach ($tokens as $token) {
 703                 if ($token === '(') {
 704                     // parenthesis-include-open
 705                     $parsed .= '(';
 706                     ++$parens_level;
 707                 } elseif ($token === '-(') {
 708                     // parenthesis-exclude-open
 709                     $parsed .= 'NOT(';
 710                     ++$parens_level;
 711                 } elseif ($token === ')') {
 712                     // parenthesis-any-close
 713                     if ($parens_level === 0) continue;
 714                     $parsed .= ')';
 715                     $parens_level--;
 716                 } elseif ($token === 'and') {
 717                     // logical-and (do nothing)
 718                 } elseif ($token === 'or') {
 719                     // logical-or
 720                     $parsed .= 'OR';
 721                 } elseif (preg_match('/^(?:\^|-ns:)(.+)$/u', $token, $matches)) {
 722                     // namespace-exclude
 723                     $parsed .= 'NOT(N+:'.$matches[1].')';
 724                 } elseif (preg_match('/^(?:@|ns:)(.+)$/u', $token, $matches)) {
 725                     // namespace-include
 726                     $parsed .= '(N+:'.$matches[1].')';
 727                 } elseif (preg_match('/^-(.+)$/', $token, $matches)) {
 728                     // word-exclude
 729                     $parsed .= 'NOT('.ft_termParser($Indexer, $matches[1]).')';
 730                 } else {
 731                     // word-include
 732                     $parsed .= ft_termParser($Indexer, $token);
 733                 }
 734             }
 735         }
 736         $parsed_query .= $parsed;
 737     }
 738
 739     // cleanup (very sensitive)
 740     $parsed_query .= str_repeat(')', $parens_level);
 741     do {
 742         $parsed_query_old = $parsed_query;
 743         $parsed_query = preg_replace('/(NOT)?\(\)/u', '', $parsed_query);
 744     } while ($parsed_query !== $parsed_query_old);
 745     $parsed_query = preg_replace('/(NOT|OR)+\)/u', ')'      , $parsed_query);
 746     $parsed_query = preg_replace('/(OR)+/u'      , 'OR'     , $parsed_query);
 747     $parsed_query = preg_replace('/\(OR/u'       , '('      , $parsed_query);
 748     $parsed_query = preg_replace('/^OR|OR$/u'    , ''       , $parsed_query);
 749     $parsed_query = preg_replace('/\)(NOT)?\(/u' , ')AND$1(', $parsed_query);
 750
 751     // adjustment: make highlightings right
 752     $parens_level     = 0;
 753     $notgrp_levels    = array();
 754     $parsed_query_new = '';
 755     $tokens = preg_split('/(NOT\(|[()])/u', $parsed_query, -1, PREG_SPLIT_DELIM_CAPTURE | PREG_SPLIT_NO_EMPTY);
 756     foreach ($tokens as $token) {
 757         if ($token === 'NOT(') {
 758             $notgrp_levels[] = ++$parens_level;
 759         } elseif ($token === '(') {
 760             ++$parens_level;
 761         } elseif ($token === ')') {
 762             if ($parens_level-- === end($notgrp_levels)) array_pop($notgrp_levels);
 763         } elseif (count($notgrp_levels) % 2 === 1) {
 764             // turn highlight-flag off if terms are logically in "NOT" group
 765             $token = preg_replace('/([WPN])\+\:/u', '$1-:', $token);
 766         }
 767         $parsed_query_new .= $token;
 768     }
 769     $parsed_query = $parsed_query_new;
 770
 771     /**
 772      * convert infix notation string into postfix (Reverse Polish notation) array
 773      * by Shunting-yard algorithm
 774      *
 775      * see: http://en.wikipedia.org/wiki/Reverse_Polish_notation
 776      * see: http://en.wikipedia.org/wiki/Shunting-yard_algorithm
 777      */
 778     $parsed_ary     = array();
 779     $ope_stack      = array();
 780     $ope_precedence = array(')' => 1, 'OR' => 2, 'AND' => 3, 'NOT' => 4, '(' => 5);
 781     $ope_regex      = '/([()]|OR|AND|NOT)/u';
 782
 783     $tokens = preg_split($ope_regex, $parsed_query, -1, PREG_SPLIT_DELIM_CAPTURE | PREG_SPLIT_NO_EMPTY);
 784     foreach ($tokens as $token) {
 785         if (preg_match($ope_regex, $token)) {
 786             // operator
 787             $last_ope = end($ope_stack);
 788             while ($last_ope !== false && $ope_precedence[$token] <= $ope_precedence[$last_ope] && $last_ope != '(') {
 789                 $parsed_ary[] = array_pop($ope_stack);
 790                 $last_ope = end($ope_stack);
 791             }
 792             if ($token == ')') {
 793                 array_pop($ope_stack); // this array_pop always deletes '('
 794             } else {
 795                 $ope_stack[] = $token;
 796             }
 797         } else {
 798             // operand
 799             $token_decoded = str_replace(array('OP', 'CP'), array('(', ')'), $token);
 800             $parsed_ary[] = $token_decoded;
 801         }
 802     }
 803     $parsed_ary = array_values(array_merge($parsed_ary, array_reverse($ope_stack)));
 804
 805     // cleanup: each double "NOT" in RPN array actually does nothing
 806     $parsed_ary_count = count($parsed_ary);
 807     for ($i = 1; $i < $parsed_ary_count; ++$i) {
 808         if ($parsed_ary[$i] === 'NOT' && $parsed_ary[$i - 1] === 'NOT') {
 809             unset($parsed_ary[$i], $parsed_ary[$i - 1]);
 810         }
 811     }
 812     $parsed_ary = array_values($parsed_ary);
 813
 814     // build return value
 815     $q = array();
 816     $q['query']      = $query;
 817     $q['parsed_str'] = $parsed_query;
 818     $q['parsed_ary'] = $parsed_ary;
 819
 820     foreach ($q['parsed_ary'] as $token) {
 821         if ($token[2] !== ':') continue;
 822         $body = substr($token, 3);
 823
 824         switch (substr($token, 0, 3)) {
 825             case 'N+:':
 826                      $q['ns'][]        = $body; // for backward compatibility
 827                      break;
 828             case 'N-:':
 829                      $q['notns'][]     = $body; // for backward compatibility
 830                      break;
 831             case 'W_:':
 832                      $q['words'][]     = $body;
 833                      break;
 834             case 'W-:':
 835                      $q['words'][]     = $body;
 836                      $q['not'][]       = $body; // for backward compatibility
 837                      break;
 838             case 'W+:':
 839                      $q['words'][]     = $body;
 840                      $q['highlight'][] = $body;
 841                      $q['and'][]       = $body; // for backward compatibility
 842                      break;
 843             case 'P-:':
 844                      $q['phrases'][]   = $body;
 845                      break;
 846             case 'P+:':
 847                      $q['phrases'][]   = $body;
 848                      $q['highlight'][] = $body;
 849                      break;
 850         }
 851     }
 852     foreach (array('words', 'phrases', 'highlight', 'ns', 'notns', 'and', 'not') as $key) {
 853         $q[$key] = empty($q[$key]) ? array() : array_values(array_unique($q[$key]));
 854     }
 855
 856     return $q;
 857 }
 858
 859 /**
 860  * Transforms given search term into intermediate representation
 861  *
 862  * This function is used in ft_queryParser() and not for general purpose use.
 863  *
 864  * @author Kazutaka Miyasaka <kazmiya@gmail.com>
 865  *
 866  * @param Doku_Indexer $Indexer
 867  * @param string       $term
 868  * @param bool         $consider_asian
 869  * @param bool         $phrase_mode
 870  * @return string
 871  */
 872 function ft_termParser($Indexer, $term, $consider_asian = true, $phrase_mode = false) {
 873     $parsed = '';
 874     if ($consider_asian) {
 875         // successive asian characters need to be searched as a phrase
 876         $words = preg_split('/('.IDX_ASIAN.'+)/u', $term, -1, PREG_SPLIT_DELIM_CAPTURE | PREG_SPLIT_NO_EMPTY);
 877         foreach ($words as $word) {
 878             $phrase_mode = $phrase_mode ? true : preg_match('/'.IDX_ASIAN.'/u', $word);
 879             $parsed .= ft_termParser($Indexer, $word, false, $phrase_mode);
 880         }
 881     } else {
 882         $term_noparen = str_replace(array('(', ')'), ' ', $term);
 883         $words = $Indexer->tokenizer($term_noparen, true);
 884
 885         // W_: no need to highlight
 886         if (empty($words)) {
 887             $parsed = '()'; // important: do not remove
 888         } elseif ($words[0] === $term) {
 889             $parsed = '(W+:'.$words[0].')';
 890         } elseif ($phrase_mode) {
 891             $term_encoded = str_replace(array('(', ')'), array('OP', 'CP'), $term);
 892             $parsed = '((W_:'.implode(')(W_:', $words).')(P+:'.$term_encoded.'))';
 893         } else {
 894             $parsed = '((W+:'.implode(')(W+:', $words).'))';
 895         }
 896     }
 897     return $parsed;
 898 }
 899
 900 /**
 901  * Recreate a search query string based on parsed parts, doesn't support negated phrases and `OR` searches
 902  *
 903  * @param array $and
 904  * @param array $not
 905  * @param array $phrases
 906  * @param array $ns
 907  * @param array $notns
 908  *
 909  * @return string
 910  */
 911 function ft_queryUnparser_simple(array $and, array $not, array $phrases, array $ns, array $notns) {
 912     $query = implode(' ', $and);
 913     if (!empty($not)) {
 914         $query .= ' -' . implode(' -', $not);
 915     }
 916
 917     if (!empty($phrases)) {
 918         $query .= ' "' . implode('" "', $phrases) . '"';
 919     }
 920
 921     if (!empty($ns)) {
 922         $query .= ' @' . implode(' @', $ns);
 923     }
 924
 925     if (!empty($notns)) {
 926         $query .= ' ^' . implode(' ^', $notns);
 927     }
 928
 929     return $query;
 930 }
 931
 932 //Setup VIM: ex: et ts=4 :