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 if(!defined('DOKU_INC')) die('meh.');
  10 require_once(DOKU_INC.'inc/indexer.php');
  11
  12
  13 /**
  14  * The fulltext search
  15  *
  16  * Returns a list of matching documents for the given query
  17  *
  18  * refactored into ft_pageSearch(), _ft_pageSearch() and trigger_event()
  19  *
  20  */
  21 function ft_pageSearch($query,&$highlight){
  22
  23   $data['query'] = $query;
  24   $data['highlight'] =& $highlight;
  25
  26   return trigger_event('SEARCH_QUERY_FULLPAGE', $data, '_ft_pageSearch');
  27 }
  28
  29 /**
  30  * Returns a list of matching documents for the given query
  31  *
  32  * @author Andreas Gohr <andi@splitbrain.org>
  33  * @author Kazutaka Miyasaka <kazmiya@gmail.com>
  34  */
  35 function _ft_pageSearch(&$data) {
  36     // parse the given query
  37     $q = ft_queryParser($data['query']);
  38     $data['highlight'] = $q['highlight'];
  39
  40     if (empty($q['parsed_ary'])) return array();
  41
  42     // lookup all words found in the query
  43     $lookup = idx_lookup($q['words']);
  44
  45     // get all pages in this dokuwiki site (!: includes nonexistent pages)
  46     $pages_all = array();
  47     foreach (idx_getIndex('page', '') as $id) {
  48         $pages_all[trim($id)] = 0; // base: 0 hit
  49     }
  50
  51     // process the query
  52     $stack = array();
  53     foreach ($q['parsed_ary'] as $token) {
  54         switch (substr($token, 0, 3)) {
  55             case 'W+:':
  56             case 'W-:':
  57             case 'W_:': // word
  58                 $word    = substr($token, 3);
  59                 $stack[] = (array) $lookup[$word];
  60                 break;
  61             case 'P+:':
  62             case 'P-:': // phrase
  63                 $phrase = substr($token, 3);
  64                 // since phrases are always parsed as ((W1)(W2)...(P)),
  65                 // the end($stack) always points the pages that contain
  66                 // all words in this phrase
  67                 $pages  = end($stack);
  68                 $pages_matched = array();
  69                 foreach(array_keys($pages) as $id){
  70                     $text = utf8_strtolower(rawWiki($id));
  71                     if (strpos($text, $phrase) !== false) {
  72                         $pages_matched[$id] = 0; // phrase: always 0 hit
  73                     }
  74                 }
  75                 $stack[] = $pages_matched;
  76                 break;
  77             case 'N+:':
  78             case 'N-:': // namespace
  79                 $ns = substr($token, 3);
  80                 $pages_matched = array();
  81                 foreach (array_keys($pages_all) as $id) {
  82                     if (strpos($id, $ns) === 0) {
  83                         $pages_matched[$id] = 0; // namespace: always 0 hit
  84                     }
  85                 }
  86                 $stack[] = $pages_matched;
  87                 break;
  88             case 'AND': // and operation
  89                 list($pages1, $pages2) = array_splice($stack, -2);
  90                 $stack[] = ft_resultCombine(array($pages1, $pages2));
  91                 break;
  92             case 'OR':  // or operation
  93                 list($pages1, $pages2) = array_splice($stack, -2);
  94                 $stack[] = ft_resultUnite(array($pages1, $pages2));
  95                 break;
  96             case 'NOT': // not operation (unary)
  97                 $pages   = array_pop($stack);
  98                 $stack[] = ft_resultComplement(array($pages_all, $pages));
  99                 break;
 100         }
 101     }
 102     $docs = array_pop($stack);
 103
 104     if (empty($docs)) return array();
 105
 106     // check: settings, acls, existence
 107     foreach (array_keys($docs) as $id) {
 108         if (isHiddenPage($id) || auth_quickaclcheck($id) < AUTH_READ || !page_exists($id, '', false)) {
 109             unset($docs[$id]);
 110         }
 111     }
 112
 113     // sort docs by count
 114     arsort($docs);
 115
 116     return $docs;
 117 }
 118
 119 /**
 120  * Returns the backlinks for a given page
 121  *
 122  * Does a quick lookup with the fulltext index, then
 123  * evaluates the instructions of the found pages
 124  */
 125 function ft_backlinks($id){
 126     global $conf;
 127     $swfile   = DOKU_INC.'inc/lang/'.$conf['lang'].'/stopwords.txt';
 128     $stopwords = @file_exists($swfile) ? file($swfile) : array();
 129
 130     $result = array();
 131
 132     // quick lookup of the pagename
 133     $page    = noNS($id);
 134     $matches = idx_lookup(idx_tokenizer($page,$stopwords));  // pagename may contain specials (_ or .)
 135     $docs    = array_keys(ft_resultCombine(array_values($matches)));
 136     $docs    = array_filter($docs,'isVisiblePage'); // discard hidden pages
 137     if(!count($docs)) return $result;
 138     require_once(DOKU_INC.'inc/parserutils.php');
 139
 140     // check metadata for matching links
 141     foreach($docs as $match){
 142         // metadata relation reference links are already resolved
 143         $links = p_get_metadata($match,'relation references');
 144         if (isset($links[$id])) $result[] = $match;
 145     }
 146
 147     if(!count($result)) return $result;
 148
 149     // check ACL permissions
 150     foreach(array_keys($result) as $idx){
 151         if(auth_quickaclcheck($result[$idx]) < AUTH_READ){
 152             unset($result[$idx]);
 153         }
 154     }
 155
 156     sort($result);
 157     return $result;
 158 }
 159
 160 /**
 161  * Returns the pages that use a given media file
 162  *
 163  * Does a quick lookup with the fulltext index, then
 164  * evaluates the instructions of the found pages
 165  *
 166  * Aborts after $max found results
 167  */
 168 function ft_mediause($id,$max){
 169     global $conf;
 170     $swfile   = DOKU_INC.'inc/lang/'.$conf['lang'].'/stopwords.txt';
 171     $stopwords = @file_exists($swfile) ? file($swfile) : array();
 172
 173     if(!$max) $max = 1; // need to find at least one
 174
 175     $result = array();
 176
 177     // quick lookup of the mediafile
 178     $media   = noNS($id);
 179     $matches = idx_lookup(idx_tokenizer($media,$stopwords));
 180     $docs    = array_keys(ft_resultCombine(array_values($matches)));
 181     if(!count($docs)) return $result;
 182
 183     // go through all found pages
 184     $found = 0;
 185     $pcre  = preg_quote($media,'/');
 186     foreach($docs as $doc){
 187         $ns = getNS($doc);
 188         preg_match_all('/\{\{([^|}]*'.$pcre.'[^|}]*)(|[^}]+)?\}\}/i',rawWiki($doc),$matches);
 189         foreach($matches[1] as $img){
 190             $img = trim($img);
 191             if(preg_match('/^https?:\/\//i',$img)) continue; // skip external images
 192             list($img) = explode('?',$img);                  // remove any parameters
 193             resolve_mediaid($ns,$img,$exists);               // resolve the possibly relative img
 194
 195             if($img == $id){                                 // we have a match
 196                 $result[] = $doc;
 197                 $found++;
 198                 break;
 199             }
 200         }
 201         if($found >= $max) break;
 202     }
 203
 204     sort($result);
 205     return $result;
 206 }
 207
 208
 209
 210 /**
 211  * Quicksearch for pagenames
 212  *
 213  * By default it only matches the pagename and ignores the
 214  * namespace. This can be changed with the second parameter
 215  *
 216  * refactored into ft_pageLookup(), _ft_pageLookup() and trigger_event()
 217  *
 218  * @author Andreas Gohr <andi@splitbrain.org>
 219  */
 220 function ft_pageLookup($id,$pageonly=true){
 221     $data = array('id' => $id, 'pageonly' => $pageonly);
 222     return trigger_event('SEARCH_QUERY_PAGELOOKUP',$data,'_ft_pageLookup');
 223 }
 224
 225 function _ft_pageLookup(&$data){
 226     // split out original parameterrs
 227     $id = $data['id'];
 228     $pageonly = $data['pageonly'];
 229
 230     global $conf;
 231     $id    = preg_quote($id,'/');
 232     $pages = file($conf['indexdir'].'/page.idx');
 233     if($id) $pages = array_values(preg_grep('/'.$id.'/',$pages));
 234
 235     $cnt = count($pages);
 236     for($i=0; $i<$cnt; $i++){
 237         if($pageonly){
 238             if(!preg_match('/'.$id.'/',noNS($pages[$i]))){
 239                 unset($pages[$i]);
 240                 continue;
 241             }
 242         }
 243         if(!page_exists($pages[$i])){
 244             unset($pages[$i]);
 245             continue;
 246         }
 247     }
 248
 249     $pages = array_filter($pages,'isVisiblePage'); // discard hidden pages
 250     if(!count($pages)) return array();
 251
 252     // check ACL permissions
 253     foreach(array_keys($pages) as $idx){
 254         if(auth_quickaclcheck(trim($pages[$idx])) < AUTH_READ){
 255             unset($pages[$idx]);
 256         }
 257     }
 258
 259     $pages = array_map('trim',$pages);
 260     usort($pages,'ft_pagesorter');
 261     return $pages;
 262 }
 263
 264 /**
 265  * Sort pages based on their namespace level first, then on their string
 266  * values. This makes higher hierarchy pages rank higher than lower hierarchy
 267  * pages.
 268  */
 269 function ft_pagesorter($a, $b){
 270     $ac = count(explode(':',$a));
 271     $bc = count(explode(':',$b));
 272     if($ac < $bc){
 273         return -1;
 274     }elseif($ac > $bc){
 275         return 1;
 276     }
 277     return strcmp ($a,$b);
 278 }
 279
 280 /**
 281  * Creates a snippet extract
 282  *
 283  * @author Andreas Gohr <andi@splitbrain.org>
 284  * @triggers FULLTEXT_SNIPPET_CREATE
 285  */
 286 function ft_snippet($id,$highlight){
 287     $text = rawWiki($id);
 288     $evdata = array(
 289                 'id'        => $id,
 290                 'text'      => &$text,
 291                 'highlight' => &$highlight,
 292                 'snippet'   => '',
 293               );
 294
 295     $evt = new Doku_Event('FULLTEXT_SNIPPET_CREATE',$evdata);
 296     if ($evt->advise_before()) {
 297         $match = array();
 298         $snippets = array();
 299         $utf8_offset = $offset = $end = 0;
 300         $len = utf8_strlen($text);
 301
 302         // build a regexp from the phrases to highlight
 303         $re1 = '('.join('|',array_map('preg_quote_cb',array_filter((array) $highlight))).')';
 304         $re2 = "$re1.{0,75}(?!\\1)$re1";
 305         $re3 = "$re1.{0,45}(?!\\1)$re1.{0,45}(?!\\1)(?!\\2)$re1";
 306
 307         for ($cnt=4; $cnt--;) {
 308           if (0) {
 309           } else if (preg_match('/'.$re3.'/iu',$text,$match,PREG_OFFSET_CAPTURE,$offset)) {
 310           } else if (preg_match('/'.$re2.'/iu',$text,$match,PREG_OFFSET_CAPTURE,$offset)) {
 311           } else if (preg_match('/'.$re1.'/iu',$text,$match,PREG_OFFSET_CAPTURE,$offset)) {
 312           } else {
 313             break;
 314           }
 315
 316           list($str,$idx) = $match[0];
 317
 318           // convert $idx (a byte offset) into a utf8 character offset
 319           $utf8_idx = utf8_strlen(substr($text,0,$idx));
 320           $utf8_len = utf8_strlen($str);
 321
 322           // establish context, 100 bytes surrounding the match string
 323           // first look to see if we can go 100 either side,
 324           // then drop to 50 adding any excess if the other side can't go to 50,
 325           $pre = min($utf8_idx-$utf8_offset,100);
 326           $post = min($len-$utf8_idx-$utf8_len,100);
 327
 328           if ($pre>50 && $post>50) {
 329             $pre = $post = 50;
 330           } else if ($pre>50) {
 331             $pre = min($pre,100-$post);
 332           } else if ($post>50) {
 333             $post = min($post, 100-$pre);
 334           } else {
 335             // both are less than 50, means the context is the whole string
 336             // make it so and break out of this loop - there is no need for the
 337             // complex snippet calculations
 338             $snippets = array($text);
 339             break;
 340           }
 341
 342           // establish context start and end points, try to append to previous
 343           // context if possible
 344           $start = $utf8_idx - $pre;
 345           $append = ($start < $end) ? $end : false;  // still the end of the previous context snippet
 346           $end = $utf8_idx + $utf8_len + $post;      // now set it to the end of this context
 347
 348           if ($append) {
 349             $snippets[count($snippets)-1] .= utf8_substr($text,$append,$end-$append);
 350           } else {
 351             $snippets[] = utf8_substr($text,$start,$end-$start);
 352           }
 353
 354           // set $offset for next match attempt
 355           //   substract strlen to avoid splitting a potential search success,
 356           //   this is an approximation as the search pattern may match strings
 357           //   of varying length and it will fail if the context snippet
 358           //   boundary breaks a matching string longer than the current match
 359           $utf8_offset = $utf8_idx + $post;
 360           $offset = $idx + strlen(utf8_substr($text,$utf8_idx,$post));
 361           $offset = utf8_correctIdx($text,$offset);
 362         }
 363
 364         $m = "\1";
 365         $snippets = preg_replace('/'.$re1.'/iu',$m.'$1'.$m,$snippets);
 366         $snippet = preg_replace('/'.$m.'([^'.$m.']*?)'.$m.'/iu','<strong class="search_hit">$1</strong>',hsc(join('... ',$snippets)));
 367
 368         $evdata['snippet'] = $snippet;
 369     }
 370     $evt->advise_after();
 371     unset($evt);
 372
 373     return $evdata['snippet'];
 374 }
 375
 376 /**
 377  * Combine found documents and sum up their scores
 378  *
 379  * This function is used to combine searched words with a logical
 380  * AND. Only documents available in all arrays are returned.
 381  *
 382  * based upon PEAR's PHP_Compat function for array_intersect_key()
 383  *
 384  * @param array $args An array of page arrays
 385  */
 386 function ft_resultCombine($args){
 387     $array_count = count($args);
 388     if($array_count == 1){
 389         return $args[0];
 390     }
 391
 392     $result = array();
 393     if ($array_count > 1) {
 394       foreach ($args[0] as $key => $value) {
 395         $result[$key] = $value;
 396         for ($i = 1; $i !== $array_count; $i++) {
 397             if (!isset($args[$i][$key])) {
 398                 unset($result[$key]);
 399                 break;
 400             }
 401             $result[$key] += $args[$i][$key];
 402         }
 403       }
 404     }
 405     return $result;
 406 }
 407
 408 /**
 409  * Unites found documents and sum up their scores
 410  *
 411  * based upon ft_resultCombine() function
 412  *
 413  * @param array $args An array of page arrays
 414  * @author Kazutaka Miyasaka <kazmiya@gmail.com>
 415  */
 416 function ft_resultUnite($args) {
 417     $array_count = count($args);
 418     if ($array_count === 1) {
 419         return $args[0];
 420     }
 421
 422     $result = $args[0];
 423     for ($i = 1; $i !== $array_count; $i++) {
 424         foreach (array_keys($args[$i]) as $id) {
 425             $result[$id] += $args[$i][$id];
 426         }
 427     }
 428     return $result;
 429 }
 430
 431 /**
 432  * Computes the difference of documents using page id for comparison
 433  *
 434  * nearly identical to PHP5's array_diff_key()
 435  *
 436  * @param array $args An array of page arrays
 437  * @author Kazutaka Miyasaka <kazmiya@gmail.com>
 438  */
 439 function ft_resultComplement($args) {
 440     $array_count = count($args);
 441     if ($array_count === 1) {
 442         return $args[0];
 443     }
 444
 445     $result = $args[0];
 446     foreach (array_keys($result) as $id) {
 447         for ($i = 1; $i !== $array_count; $i++) {
 448             if (isset($args[$i][$id])) unset($result[$id]);
 449         }
 450     }
 451     return $result;
 452 }
 453
 454 /**
 455  * Parses a search query and builds an array of search formulas
 456  *
 457  * @author Andreas Gohr <andi@splitbrain.org>
 458  * @author Kazutaka Miyasaka <kazmiya@gmail.com>
 459  */
 460 function ft_queryParser($query){
 461     global $conf;
 462     $swfile    = DOKU_INC.'inc/lang/'.$conf['lang'].'/stopwords.txt';
 463     $stopwords = @file_exists($swfile) ? file($swfile) : array();
 464
 465     /**
 466      * parse a search query and transform it into intermediate representation
 467      *
 468      * in a search query, you can use the following expressions:
 469      *
 470      *   words:
 471      *     include
 472      *     -exclude
 473      *   phrases:
 474      *     "phrase to be included"
 475      *     -"phrase you want to exclude"
 476      *   namespaces:
 477      *     @include:namespace (or ns:include:namespace)
 478      *     ^exclude:namespace (or -ns:exclude:namespace)
 479      *   groups:
 480      *     ()
 481      *     -()
 482      *   operators:
 483      *     and ('and' is the default operator: you can always omit this)
 484      *     or  (or pipe symbol '|', lower precedence than 'and')
 485      *
 486      * e.g. a query [ aa "bb cc" @dd:ee ] means "search pages which contain
 487      *      a word 'aa', a phrase 'bb cc' and are within a namespace 'dd:ee'".
 488      *      this query is equivalent to [ -(-aa or -"bb cc" or -ns:dd:ee) ]
 489      *      as long as you don't mind hit counts.
 490      *
 491      * intermediate representation consists of the following parts:
 492      *
 493      *   ( )           - group
 494      *   AND           - logical and
 495      *   OR            - logical or
 496      *   NOT           - logical not
 497      *   W+:, W-:, W_: - word      (underscore: no need to highlight)
 498      *   P+:, P-:      - phrase    (minus sign: logically in NOT group)
 499      *   N+:, N-:      - namespace
 500      */
 501     $parsed_query = '';
 502     $parens_level = 0;
 503     $terms = preg_split('/(-?".*?")/u', utf8_strtolower($query), -1, PREG_SPLIT_DELIM_CAPTURE | PREG_SPLIT_NO_EMPTY);
 504
 505     foreach ($terms as $term) {
 506         $parsed = '';
 507         if (preg_match('/^(-?)"(.+)"$/u', $term, $matches)) {
 508             // phrase-include and phrase-exclude
 509             $not = $matches[1] ? 'NOT' : '';
 510             $parsed = $not.ft_termParser($matches[2], $stopwords, false, true);
 511         } else {
 512             // fix incomplete phrase
 513             $term = str_replace('"', ' ', $term);
 514
 515             // fix parentheses
 516             $term = str_replace(')'  , ' ) ', $term);
 517             $term = str_replace('('  , ' ( ', $term);
 518             $term = str_replace('- (', ' -(', $term);
 519
 520             // treat pipe symbols as 'OR' operators
 521             $term = str_replace('|', ' or ', $term);
 522
 523             // treat ideographic spaces (U+3000) as search term separators
 524             // FIXME: some more separators?
 525             $term = preg_replace('/[ \x{3000}]+/u', ' ',  $term);
 526             $term = trim($term);
 527             if ($term === '') continue;
 528
 529             $tokens = explode(' ', $term);
 530             foreach ($tokens as $token) {
 531                 if ($token === '(') {
 532                     // parenthesis-include-open
 533                     $parsed .= '(';
 534                     ++$parens_level;
 535                 } elseif ($token === '-(') {
 536                     // parenthesis-exclude-open
 537                     $parsed .= 'NOT(';
 538                     ++$parens_level;
 539                 } elseif ($token === ')') {
 540                     // parenthesis-any-close
 541                     if ($parens_level === 0) continue;
 542                     $parsed .= ')';
 543                     $parens_level--;
 544                 } elseif ($token === 'and') {
 545                     // logical-and (do nothing)
 546                 } elseif ($token === 'or') {
 547                     // logical-or
 548                     $parsed .= 'OR';
 549                 } elseif (preg_match('/^(?:\^|-ns:)(.+)$/u', $token, $matches)) {
 550                     // namespace-exclude
 551                     $parsed .= 'NOT(N+:'.$matches[1].')';
 552                 } elseif (preg_match('/^(?:@|ns:)(.+)$/u', $token, $matches)) {
 553                     // namespace-include
 554                     $parsed .= '(N+:'.$matches[1].')';
 555                 } elseif (preg_match('/^-(.+)$/', $token, $matches)) {
 556                     // word-exclude
 557                     $parsed .= 'NOT('.ft_termParser($matches[1], $stopwords).')';
 558                 } else {
 559                     // word-include
 560                     $parsed .= ft_termParser($token, $stopwords);
 561                 }
 562             }
 563         }
 564         $parsed_query .= $parsed;
 565     }
 566
 567     // cleanup (very sensitive)
 568     $parsed_query .= str_repeat(')', $parens_level);
 569     do {
 570         $parsed_query_old = $parsed_query;
 571         $parsed_query = preg_replace('/(NOT)?\(\)/u', '', $parsed_query);
 572     } while ($parsed_query !== $parsed_query_old);
 573     $parsed_query = preg_replace('/(NOT|OR)+\)/u', ')'      , $parsed_query);
 574     $parsed_query = preg_replace('/(OR)+/u'      , 'OR'     , $parsed_query);
 575     $parsed_query = preg_replace('/\(OR/u'       , '('      , $parsed_query);
 576     $parsed_query = preg_replace('/^OR|OR$/u'    , ''       , $parsed_query);
 577     $parsed_query = preg_replace('/\)(NOT)?\(/u' , ')AND$1(', $parsed_query);
 578
 579     // adjustment: make highlightings right
 580     $parens_level     = 0;
 581     $notgrp_levels    = array();
 582     $parsed_query_new = '';
 583     $tokens = preg_split('/(NOT\(|[()])/u', $parsed_query, -1, PREG_SPLIT_DELIM_CAPTURE | PREG_SPLIT_NO_EMPTY);
 584     foreach ($tokens as $token) {
 585         if ($token === 'NOT(') {
 586             $notgrp_levels[] = ++$parens_level;
 587         } elseif ($token === '(') {
 588             ++$parens_level;
 589         } elseif ($token === ')') {
 590             if ($parens_level-- === end($notgrp_levels)) array_pop($notgrp_levels);
 591         } elseif (count($notgrp_levels) % 2 === 1) {
 592             // turn highlight-flag off if terms are logically in "NOT" group
 593             $token = preg_replace('/([WPN])\+\:/u', '$1-:', $token);
 594         }
 595         $parsed_query_new .= $token;
 596     }
 597     $parsed_query = $parsed_query_new;
 598
 599     /**
 600      * convert infix notation string into postfix (Reverse Polish notation) array
 601      * by Shunting-yard algorithm
 602      *
 603      * see: http://en.wikipedia.org/wiki/Reverse_Polish_notation
 604      * see: http://en.wikipedia.org/wiki/Shunting-yard_algorithm
 605      */
 606     $parsed_ary     = array();
 607     $ope_stack      = array();
 608     $ope_precedence = array(')' => 1, 'OR' => 2, 'AND' => 3, 'NOT' => 4, '(' => 5);
 609     $ope_regex      = '/([()]|OR|AND|NOT)/u';
 610
 611     $tokens = preg_split($ope_regex, $parsed_query, -1, PREG_SPLIT_DELIM_CAPTURE | PREG_SPLIT_NO_EMPTY);
 612     foreach ($tokens as $token) {
 613         if (preg_match($ope_regex, $token)) {
 614             // operator
 615             $last_ope = end($ope_stack);
 616             while ($ope_precedence[$token] <= $ope_precedence[$last_ope] && $last_ope != '(') {
 617                 $parsed_ary[] = array_pop($ope_stack);
 618                 $last_ope = end($ope_stack);
 619             }
 620             if ($token == ')') {
 621                 array_pop($ope_stack); // this array_pop always deletes '('
 622             } else {
 623                 $ope_stack[] = $token;
 624             }
 625         } else {
 626             // operand
 627             $token_decoded = str_replace(array('OP', 'CP'), array('(', ')'), $token);
 628             $parsed_ary[] = $token_decoded;
 629         }
 630     }
 631     $parsed_ary = array_values(array_merge($parsed_ary, array_reverse($ope_stack)));
 632
 633     // cleanup: each double "NOT" in RPN array actually does nothing
 634     $parsed_ary_count = count($parsed_ary);
 635     for ($i = 1; $i < $parsed_ary_count; ++$i) {
 636         if ($parsed_ary[$i] === 'NOT' && $parsed_ary[$i - 1] === 'NOT') {
 637             unset($parsed_ary[$i], $parsed_ary[$i - 1]);
 638         }
 639     }
 640     $parsed_ary = array_values($parsed_ary);
 641
 642     // build return value
 643     $q = array();
 644     $q['query']      = $query;
 645     $q['parsed_str'] = $parsed_query;
 646     $q['parsed_ary'] = $parsed_ary;
 647
 648     foreach ($q['parsed_ary'] as $token) {
 649         if ($token[2] !== ':') continue;
 650         $body = substr($token, 3);
 651
 652         switch (substr($token, 0, 3)) {
 653             case 'N+:':
 654                 $q['ns'][]        = $body; // for backward compatibility
 655                 break;
 656             case 'N-:':
 657                 $q['notns'][]     = $body; // for backward compatibility
 658                 break;
 659             case 'W_:':
 660                 $q['words'][]     = $body;
 661                 break;
 662             case 'W-:':
 663                 $q['words'][]     = $body;
 664                 $q['not'][]       = $body; // for backward compatibility
 665                 break;
 666             case 'W+:':
 667                 $q['words'][]     = $body;
 668                 $q['highlight'][] = str_replace('*', '', $body);
 669                 $q['and'][]       = $body; // for backward compatibility
 670                 break;
 671             case 'P-:':
 672                 $q['phrases'][]   = $body;
 673                 break;
 674             case 'P+:':
 675                 $q['phrases'][]   = $body;
 676                 $q['highlight'][] = str_replace('*', '', $body);
 677                 break;
 678         }
 679     }
 680     foreach (array('words', 'phrases', 'highlight', 'ns', 'notns', 'and', 'not') as $key) {
 681         $q[$key] = empty($q[$key]) ? array() : array_values(array_unique($q[$key]));
 682     }
 683
 684     return $q;
 685 }
 686
 687 /**
 688  * Transforms given search term into intermediate representation
 689  *
 690  * This function is used in ft_queryParser() and not for general purpose use.
 691  *
 692  * @author Kazutaka Miyasaka <kazmiya@gmail.com>
 693  */
 694 function ft_termParser($term, &$stopwords, $consider_asian = true, $phrase_mode = false) {
 695     $parsed = '';
 696     if ($consider_asian) {
 697         // successive asian characters need to be searched as a phrase
 698         $words = preg_split('/('.IDX_ASIAN.'+)/u', $term, -1, PREG_SPLIT_DELIM_CAPTURE | PREG_SPLIT_NO_EMPTY);
 699         foreach ($words as $word) {
 700             if (preg_match('/'.IDX_ASIAN.'/u', $word)) $phrase_mode = true;
 701             $parsed .= ft_termParser($word, $stopwords, false, $phrase_mode);
 702         }
 703     } else {
 704         $term_noparen = str_replace(array('(', ')'), ' ', $term);
 705         $words = idx_tokenizer($term_noparen, $stopwords, true);
 706
 707         // W_: no need to highlight
 708         if (empty($words)) {
 709             $parsed = '()'; // important: do not remove
 710         } elseif ($words[0] === $term) {
 711             $parsed = '(W+:'.$words[0].')';
 712         } elseif ($phrase_mode) {
 713             $term_encoded = str_replace(array('(', ')'), array('OP', 'CP'), $term);
 714             $parsed = '((W_:'.implode(')(W_:', $words).')(P+:'.$term_encoded.'))';
 715         } else {
 716             $parsed = '((W+:'.implode(')(W+:', $words).'))';
 717         }
 718     }
 719     return $parsed;
 720 }
 721
 722 //Setup VIM: ex: et ts=4 enc=utf-8 :