[ Index ]

PHP Cross Reference of DokuWiki

title

Body

[close]

/inc/ -> fulltext.php (source)

   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  use dokuwiki\Utf8\Sort;
  11  
  12  /**
  13   * create snippets for the first few results only
  14   */
  15  if(!defined('FT_SNIPPET_NUMBER')) define('FT_SNIPPET_NUMBER',15);
  16  
  17  /**
  18   * The fulltext search
  19   *
  20   * Returns a list of matching documents for the given query
  21   *
  22   * refactored into ft_pageSearch(), _ft_pageSearch() and trigger_event()
  23   *
  24   * @param string     $query
  25   * @param array      $highlight
  26   * @param string     $sort
  27   * @param int|string $after  only show results with mtime after this date, accepts timestap or strtotime arguments
  28   * @param int|string $before only show results with mtime before this date, accepts timestap or strtotime arguments
  29   *
  30   * @return array
  31   */
  32  function ft_pageSearch($query,&$highlight, $sort = null, $after = null, $before = null){
  33  
  34      if ($sort === null) {
  35          $sort = 'hits';
  36      }
  37      $data = [
  38          'query' => $query,
  39          'sort' => $sort,
  40          'after' => $after,
  41          'before' => $before
  42      ];
  43      $data['highlight'] =& $highlight;
  44  
  45      return Event::createAndTrigger('SEARCH_QUERY_FULLPAGE', $data, '_ft_pageSearch');
  46  }
  47  
  48  /**
  49   * Returns a list of matching documents for the given query
  50   *
  51   * @author Andreas Gohr <andi@splitbrain.org>
  52   * @author Kazutaka Miyasaka <kazmiya@gmail.com>
  53   *
  54   * @param array $data event data
  55   * @return array matching documents
  56   */
  57  function _ft_pageSearch(&$data) {
  58      $Indexer = idx_get_indexer();
  59  
  60      // parse the given query
  61      $q = ft_queryParser($Indexer, $data['query']);
  62      $data['highlight'] = $q['highlight'];
  63  
  64      if (empty($q['parsed_ary'])) return array();
  65  
  66      // lookup all words found in the query
  67      $lookup = $Indexer->lookup($q['words']);
  68  
  69      // get all pages in this dokuwiki site (!: includes nonexistent pages)
  70      $pages_all = array();
  71      foreach ($Indexer->getPages() as $id) {
  72          $pages_all[$id] = 0; // base: 0 hit
  73      }
  74  
  75      // process the query
  76      $stack = array();
  77      foreach ($q['parsed_ary'] as $token) {
  78          switch (substr($token, 0, 3)) {
  79              case 'W+:':
  80              case 'W-:':
  81              case 'W_:': // word
  82                  $word    = substr($token, 3);
  83                  if(isset($lookup[$word])) {
  84                      $stack[] = (array)$lookup[$word];
  85                  }
  86                  break;
  87              case 'P+:':
  88              case 'P-:': // phrase
  89                  $phrase = substr($token, 3);
  90                  // since phrases are always parsed as ((W1)(W2)...(P)),
  91                  // the end($stack) always points the pages that contain
  92                  // all words in this phrase
  93                  $pages  = end($stack);
  94                  $pages_matched = array();
  95                  foreach(array_keys($pages) as $id){
  96                      $evdata = array(
  97                          'id' => $id,
  98                          'phrase' => $phrase,
  99                          'text' => rawWiki($id)
 100                      );
 101                      $evt = new Event('FULLTEXT_PHRASE_MATCH',$evdata);
 102                      if ($evt->advise_before() && $evt->result !== true) {
 103                          $text = \dokuwiki\Utf8\PhpString::strtolower($evdata['text']);
 104                          if (strpos($text, $phrase) !== false) {
 105                              $evt->result = true;
 106                          }
 107                      }
 108                      $evt->advise_after();
 109                      if ($evt->result === true) {
 110                          $pages_matched[$id] = 0; // phrase: always 0 hit
 111                      }
 112                  }
 113                  $stack[] = $pages_matched;
 114                  break;
 115              case 'N+:':
 116              case 'N-:': // namespace
 117                  $ns = cleanID(substr($token, 3)) . ':';
 118                  $pages_matched = array();
 119                  foreach (array_keys($pages_all) as $id) {
 120                      if (strpos($id, $ns) === 0) {
 121                          $pages_matched[$id] = 0; // namespace: always 0 hit
 122                      }
 123                  }
 124                  $stack[] = $pages_matched;
 125                  break;
 126              case 'AND': // and operation
 127                  list($pages1, $pages2) = array_splice($stack, -2);
 128                  $stack[] = ft_resultCombine(array($pages1, $pages2));
 129                  break;
 130              case 'OR':  // or operation
 131                  list($pages1, $pages2) = array_splice($stack, -2);
 132                  $stack[] = ft_resultUnite(array($pages1, $pages2));
 133                  break;
 134              case 'NOT': // not operation (unary)
 135                  $pages   = array_pop($stack);
 136                  $stack[] = ft_resultComplement(array($pages_all, $pages));
 137                  break;
 138          }
 139      }
 140      $docs = array_pop($stack);
 141  
 142      if (empty($docs)) return array();
 143  
 144      // check: settings, acls, existence
 145      foreach (array_keys($docs) as $id) {
 146          if (isHiddenPage($id) || auth_quickaclcheck($id) < AUTH_READ || !page_exists($id, '', false)) {
 147              unset($docs[$id]);
 148          }
 149      }
 150  
 151      $docs = _ft_filterResultsByTime($docs, $data['after'], $data['before']);
 152  
 153      if ($data['sort'] === 'mtime') {
 154          uksort($docs, 'ft_pagemtimesorter');
 155      } else {
 156          // sort docs by count
 157          uksort($docs, 'ft_pagesorter');
 158          arsort($docs);
 159      }
 160  
 161      return $docs;
 162  }
 163  
 164  /**
 165   * Returns the backlinks for a given page
 166   *
 167   * Uses the metadata index.
 168   *
 169   * @param string $id           The id for which links shall be returned
 170   * @param bool   $ignore_perms Ignore the fact that pages are hidden or read-protected
 171   * @return array The pages that contain links to the given page
 172   */
 173  function ft_backlinks($id, $ignore_perms = false){
 174      $result = idx_get_indexer()->lookupKey('relation_references', $id);
 175  
 176      if(!count($result)) return $result;
 177  
 178      // check ACL permissions
 179      foreach(array_keys($result) as $idx){
 180          if(($ignore_perms !== true && (
 181                  isHiddenPage($result[$idx]) || auth_quickaclcheck($result[$idx]) < AUTH_READ
 182              )) || !page_exists($result[$idx], '', false)){
 183              unset($result[$idx]);
 184          }
 185      }
 186  
 187      Sort::sort($result);
 188      return $result;
 189  }
 190  
 191  /**
 192   * Returns the pages that use a given media file
 193   *
 194   * Uses the relation media metadata property and the metadata index.
 195   *
 196   * Note that before 2013-07-31 the second parameter was the maximum number of results and
 197   * permissions were ignored. That's why the parameter is now checked to be explicitely set
 198   * to true (with type bool) in order to be compatible with older uses of the function.
 199   *
 200   * @param string $id           The media id to look for
 201   * @param bool   $ignore_perms Ignore hidden pages and acls (optional, default: false)
 202   * @return array A list of pages that use the given media file
 203   */
 204  function ft_mediause($id, $ignore_perms = false){
 205      $result = idx_get_indexer()->lookupKey('relation_media', $id);
 206  
 207      if(!count($result)) return $result;
 208  
 209      // check ACL permissions
 210      foreach(array_keys($result) as $idx){
 211          if(($ignore_perms !== true && (
 212                      isHiddenPage($result[$idx]) || auth_quickaclcheck($result[$idx]) < AUTH_READ
 213                  )) || !page_exists($result[$idx], '', false)){
 214              unset($result[$idx]);
 215          }
 216      }
 217  
 218      Sort::sort($result);
 219      return $result;
 220  }
 221  
 222  
 223  /**
 224   * Quicksearch for pagenames
 225   *
 226   * By default it only matches the pagename and ignores the
 227   * namespace. This can be changed with the second parameter.
 228   * The third parameter allows to search in titles as well.
 229   *
 230   * The function always returns titles as well
 231   *
 232   * @triggers SEARCH_QUERY_PAGELOOKUP
 233   * @author   Andreas Gohr <andi@splitbrain.org>
 234   * @author   Adrian Lang <lang@cosmocode.de>
 235   *
 236   * @param string     $id       page id
 237   * @param bool       $in_ns    match against namespace as well?
 238   * @param bool       $in_title search in title?
 239   * @param int|string $after    only show results with mtime after this date, accepts timestap or strtotime arguments
 240   * @param int|string $before   only show results with mtime before this date, accepts timestap or strtotime arguments
 241   *
 242   * @return string[]
 243   */
 244  function ft_pageLookup($id, $in_ns=false, $in_title=false, $after = null, $before = null){
 245      $data = [
 246          'id' => $id,
 247          'in_ns' => $in_ns,
 248          'in_title' => $in_title,
 249          'after' => $after,
 250          'before' => $before
 251      ];
 252      $data['has_titles'] = true; // for plugin backward compatibility check
 253      return Event::createAndTrigger('SEARCH_QUERY_PAGELOOKUP', $data, '_ft_pageLookup');
 254  }
 255  
 256  /**
 257   * Returns list of pages as array(pageid => First Heading)
 258   *
 259   * @param array &$data event data
 260   * @return string[]
 261   */
 262  function _ft_pageLookup(&$data){
 263      // split out original parameters
 264      $id = $data['id'];
 265      $Indexer = idx_get_indexer();
 266      $parsedQuery = ft_queryParser($Indexer, $id);
 267      if (count($parsedQuery['ns']) > 0) {
 268          $ns = cleanID($parsedQuery['ns'][0]) . ':';
 269          $id = implode(' ', $parsedQuery['highlight']);
 270      }
 271      if (count($parsedQuery['notns']) > 0) {
 272          $notns = cleanID($parsedQuery['notns'][0]) . ':';
 273          $id = implode(' ', $parsedQuery['highlight']);
 274      }
 275  
 276      $in_ns    = $data['in_ns'];
 277      $in_title = $data['in_title'];
 278      $cleaned = cleanID($id);
 279  
 280      $Indexer = idx_get_indexer();
 281      $page_idx = $Indexer->getPages();
 282  
 283      $pages = array();
 284      if ($id !== '' && $cleaned !== '') {
 285          foreach ($page_idx as $p_id) {
 286              if ((strpos($in_ns ? $p_id : noNSorNS($p_id), $cleaned) !== false)) {
 287                  if (!isset($pages[$p_id]))
 288                      $pages[$p_id] = p_get_first_heading($p_id, METADATA_DONT_RENDER);
 289              }
 290          }
 291          if ($in_title) {
 292              foreach ($Indexer->lookupKey('title', $id, '_ft_pageLookupTitleCompare') as $p_id) {
 293                  if (!isset($pages[$p_id]))
 294                      $pages[$p_id] = p_get_first_heading($p_id, METADATA_DONT_RENDER);
 295              }
 296          }
 297      }
 298  
 299      if (isset($ns)) {
 300          foreach (array_keys($pages) as $p_id) {
 301              if (strpos($p_id, $ns) !== 0) {
 302                  unset($pages[$p_id]);
 303              }
 304          }
 305      }
 306      if (isset($notns)) {
 307          foreach (array_keys($pages) as $p_id) {
 308              if (strpos($p_id, $notns) === 0) {
 309                  unset($pages[$p_id]);
 310              }
 311          }
 312      }
 313  
 314      // discard hidden pages
 315      // discard nonexistent pages
 316      // check ACL permissions
 317      foreach(array_keys($pages) as $idx){
 318          if(!isVisiblePage($idx) || !page_exists($idx) ||
 319             auth_quickaclcheck($idx) < AUTH_READ) {
 320              unset($pages[$idx]);
 321          }
 322      }
 323  
 324      $pages = _ft_filterResultsByTime($pages, $data['after'], $data['before']);
 325  
 326      uksort($pages,'ft_pagesorter');
 327      return $pages;
 328  }
 329  
 330  
 331  /**
 332   * @param array      $results search results in the form pageid => value
 333   * @param int|string $after   only returns results with mtime after this date, accepts timestap or strtotime arguments
 334   * @param int|string $before  only returns results with mtime after this date, accepts timestap or strtotime arguments
 335   *
 336   * @return array
 337   */
 338  function _ft_filterResultsByTime(array $results, $after, $before) {
 339      if ($after || $before) {
 340          $after = is_int($after) ? $after : strtotime($after);
 341          $before = is_int($before) ? $before : strtotime($before);
 342  
 343          foreach ($results as $id => $value) {
 344              $mTime = filemtime(wikiFN($id));
 345              if ($after && $after > $mTime) {
 346                  unset($results[$id]);
 347                  continue;
 348              }
 349              if ($before && $before < $mTime) {
 350                  unset($results[$id]);
 351              }
 352          }
 353      }
 354  
 355      return $results;
 356  }
 357  
 358  /**
 359   * Tiny helper function for comparing the searched title with the title
 360   * from the search index. This function is a wrapper around stripos with
 361   * adapted argument order and return value.
 362   *
 363   * @param string $search searched title
 364   * @param string $title  title from index
 365   * @return bool
 366   */
 367  function _ft_pageLookupTitleCompare($search, $title) {
 368      return stripos($title, $search) !== false;
 369  }
 370  
 371  /**
 372   * Sort pages based on their namespace level first, then on their string
 373   * values. This makes higher hierarchy pages rank higher than lower hierarchy
 374   * pages.
 375   *
 376   * @param string $a
 377   * @param string $b
 378   * @return int Returns < 0 if $a is less than $b; > 0 if $a is greater than $b, and 0 if they are equal.
 379   */
 380  function ft_pagesorter($a, $b){
 381      $ac = count(explode(':',$a));
 382      $bc = count(explode(':',$b));
 383      if($ac < $bc){
 384          return -1;
 385      }elseif($ac > $bc){
 386          return 1;
 387      }
 388      return Sort::strcmp($a,$b);
 389  }
 390  
 391  /**
 392   * Sort pages by their mtime, from newest to oldest
 393   *
 394   * @param string $a
 395   * @param string $b
 396   *
 397   * @return int Returns < 0 if $a is newer than $b, > 0 if $b is newer than $a and 0 if they are of the same age
 398   */
 399  function ft_pagemtimesorter($a, $b) {
 400      $mtimeA = filemtime(wikiFN($a));
 401      $mtimeB = filemtime(wikiFN($b));
 402      return $mtimeB - $mtimeA;
 403  }
 404  
 405  /**
 406   * Creates a snippet extract
 407   *
 408   * @author Andreas Gohr <andi@splitbrain.org>
 409   * @triggers FULLTEXT_SNIPPET_CREATE
 410   *
 411   * @param string $id page id
 412   * @param array $highlight
 413   * @return mixed
 414   */
 415  function ft_snippet($id,$highlight){
 416      $text = rawWiki($id);
 417      $text = str_replace("\xC2\xAD",'',$text); // remove soft-hyphens
 418      $evdata = array(
 419              'id'        => $id,
 420              'text'      => &$text,
 421              'highlight' => &$highlight,
 422              'snippet'   => '',
 423              );
 424  
 425      $evt = new Event('FULLTEXT_SNIPPET_CREATE',$evdata);
 426      if ($evt->advise_before()) {
 427          $match = array();
 428          $snippets = array();
 429          $utf8_offset = $offset = $end = 0;
 430          $len = \dokuwiki\Utf8\PhpString::strlen($text);
 431  
 432          // build a regexp from the phrases to highlight
 433          $re1 = '(' .
 434              join(
 435                  '|',
 436                  array_map(
 437                      'ft_snippet_re_preprocess',
 438                      array_map(
 439                          'preg_quote_cb',
 440                          array_filter((array) $highlight)
 441                      )
 442                  )
 443              ) .
 444              ')';
 445          $re2 = "$re1.{0,75}(?!\\1)$re1";
 446          $re3 = "$re1.{0,45}(?!\\1)$re1.{0,45}(?!\\1)(?!\\2)$re1";
 447  
 448          for ($cnt=4; $cnt--;) {
 449              if (0) {
 450              } else if (preg_match('/'.$re3.'/iu',$text,$match,PREG_OFFSET_CAPTURE,$offset)) {
 451              } else if (preg_match('/'.$re2.'/iu',$text,$match,PREG_OFFSET_CAPTURE,$offset)) {
 452              } else if (preg_match('/'.$re1.'/iu',$text,$match,PREG_OFFSET_CAPTURE,$offset)) {
 453              } else {
 454                  break;
 455              }
 456  
 457              list($str,$idx) = $match[0];
 458  
 459              // convert $idx (a byte offset) into a utf8 character offset
 460              $utf8_idx = \dokuwiki\Utf8\PhpString::strlen(substr($text,0,$idx));
 461              $utf8_len = \dokuwiki\Utf8\PhpString::strlen($str);
 462  
 463              // establish context, 100 bytes surrounding the match string
 464              // first look to see if we can go 100 either side,
 465              // then drop to 50 adding any excess if the other side can't go to 50,
 466              $pre = min($utf8_idx-$utf8_offset,100);
 467              $post = min($len-$utf8_idx-$utf8_len,100);
 468  
 469              if ($pre>50 && $post>50) {
 470                  $pre = $post = 50;
 471              } else if ($pre>50) {
 472                  $pre = min($pre,100-$post);
 473              } else if ($post>50) {
 474                  $post = min($post, 100-$pre);
 475              } else if ($offset == 0) {
 476                  // both are less than 50, means the context is the whole string
 477                  // make it so and break out of this loop - there is no need for the
 478                  // complex snippet calculations
 479                  $snippets = array($text);
 480                  break;
 481              }
 482  
 483              // establish context start and end points, try to append to previous
 484              // context if possible
 485              $start = $utf8_idx - $pre;
 486              $append = ($start < $end) ? $end : false;  // still the end of the previous context snippet
 487              $end = $utf8_idx + $utf8_len + $post;      // now set it to the end of this context
 488  
 489              if ($append) {
 490                  $snippets[count($snippets)-1] .= \dokuwiki\Utf8\PhpString::substr($text,$append,$end-$append);
 491              } else {
 492                  $snippets[] = \dokuwiki\Utf8\PhpString::substr($text,$start,$end-$start);
 493              }
 494  
 495              // set $offset for next match attempt
 496              // continue matching after the current match
 497              // if the current match is not the longest possible match starting at the current offset
 498              // this prevents further matching of this snippet but for possible matches of length
 499              // smaller than match length + context (at least 50 characters) this match is part of the context
 500              $utf8_offset = $utf8_idx + $utf8_len;
 501              $offset = $idx + strlen(\dokuwiki\Utf8\PhpString::substr($text,$utf8_idx,$utf8_len));
 502              $offset = \dokuwiki\Utf8\Clean::correctIdx($text,$offset);
 503          }
 504  
 505          $m = "\1";
 506          $snippets = preg_replace('/'.$re1.'/iu',$m.'$1'.$m,$snippets);
 507          $snippet = preg_replace(
 508              '/' . $m . '([^' . $m . ']*?)' . $m . '/iu',
 509              '<strong class="search_hit">$1</strong>',
 510              hsc(join('... ', $snippets))
 511          );
 512  
 513          $evdata['snippet'] = $snippet;
 514      }
 515      $evt->advise_after();
 516      unset($evt);
 517  
 518      return $evdata['snippet'];
 519  }
 520  
 521  /**
 522   * Wraps a search term in regex boundary checks.
 523   *
 524   * @param string $term
 525   * @return string
 526   */
 527  function ft_snippet_re_preprocess($term) {
 528      // do not process asian terms where word boundaries are not explicit
 529      if(\dokuwiki\Utf8\Asian::isAsianWords($term)) return $term;
 530  
 531      if (UTF8_PROPERTYSUPPORT) {
 532          // unicode word boundaries
 533          // see http://stackoverflow.com/a/2449017/172068
 534          $BL = '(?<!\pL)';
 535          $BR = '(?!\pL)';
 536      } else {
 537          // not as correct as above, but at least won't break
 538          $BL = '\b';
 539          $BR = '\b';
 540      }
 541  
 542      if(substr($term,0,2) == '\\*'){
 543          $term = substr($term,2);
 544      }else{
 545          $term = $BL.$term;
 546      }
 547  
 548      if(substr($term,-2,2) == '\\*'){
 549          $term = substr($term,0,-2);
 550      }else{
 551          $term = $term.$BR;
 552      }
 553  
 554      if($term == $BL || $term == $BR || $term == $BL.$BR) $term = '';
 555      return $term;
 556  }
 557  
 558  /**
 559   * Combine found documents and sum up their scores
 560   *
 561   * This function is used to combine searched words with a logical
 562   * AND. Only documents available in all arrays are returned.
 563   *
 564   * based upon PEAR's PHP_Compat function for array_intersect_key()
 565   *
 566   * @param array $args An array of page arrays
 567   * @return array
 568   */
 569  function ft_resultCombine($args){
 570      $array_count = count($args);
 571      if($array_count == 1){
 572          return $args[0];
 573      }
 574  
 575      $result = array();
 576      if ($array_count > 1) {
 577          foreach ($args[0] as $key => $value) {
 578              $result[$key] = $value;
 579              for ($i = 1; $i !== $array_count; $i++) {
 580                  if (!isset($args[$i][$key])) {
 581                      unset($result[$key]);
 582                      break;
 583                  }
 584                  $result[$key] += $args[$i][$key];
 585              }
 586          }
 587      }
 588      return $result;
 589  }
 590  
 591  /**
 592   * Unites found documents and sum up their scores
 593   *
 594   * based upon ft_resultCombine() function
 595   *
 596   * @param array $args An array of page arrays
 597   * @return array
 598   *
 599   * @author Kazutaka Miyasaka <kazmiya@gmail.com>
 600   */
 601  function ft_resultUnite($args) {
 602      $array_count = count($args);
 603      if ($array_count === 1) {
 604          return $args[0];
 605      }
 606  
 607      $result = $args[0];
 608      for ($i = 1; $i !== $array_count; $i++) {
 609          foreach (array_keys($args[$i]) as $id) {
 610              $result[$id] += $args[$i][$id];
 611          }
 612      }
 613      return $result;
 614  }
 615  
 616  /**
 617   * Computes the difference of documents using page id for comparison
 618   *
 619   * nearly identical to PHP5's array_diff_key()
 620   *
 621   * @param array $args An array of page arrays
 622   * @return array
 623   *
 624   * @author Kazutaka Miyasaka <kazmiya@gmail.com>
 625   */
 626  function ft_resultComplement($args) {
 627      $array_count = count($args);
 628      if ($array_count === 1) {
 629          return $args[0];
 630      }
 631  
 632      $result = $args[0];
 633      foreach (array_keys($result) as $id) {
 634          for ($i = 1; $i !== $array_count; $i++) {
 635              if (isset($args[$i][$id])) unset($result[$id]);
 636          }
 637      }
 638      return $result;
 639  }
 640  
 641  /**
 642   * Parses a search query and builds an array of search formulas
 643   *
 644   * @author Andreas Gohr <andi@splitbrain.org>
 645   * @author Kazutaka Miyasaka <kazmiya@gmail.com>
 646   *
 647   * @param dokuwiki\Search\Indexer $Indexer
 648   * @param string                  $query search query
 649   * @return array of search formulas
 650   */
 651  function ft_queryParser($Indexer, $query){
 652      /**
 653       * parse a search query and transform it into intermediate representation
 654       *
 655       * in a search query, you can use the following expressions:
 656       *
 657       *   words:
 658       *     include
 659       *     -exclude
 660       *   phrases:
 661       *     "phrase to be included"
 662       *     -"phrase you want to exclude"
 663       *   namespaces:
 664       *     @include:namespace (or ns:include:namespace)
 665       *     ^exclude:namespace (or -ns:exclude:namespace)
 666       *   groups:
 667       *     ()
 668       *     -()
 669       *   operators:
 670       *     and ('and' is the default operator: you can always omit this)
 671       *     or  (or pipe symbol '|', lower precedence than 'and')
 672       *
 673       * e.g. a query [ aa "bb cc" @dd:ee ] means "search pages which contain
 674       *      a word 'aa', a phrase 'bb cc' and are within a namespace 'dd:ee'".
 675       *      this query is equivalent to [ -(-aa or -"bb cc" or -ns:dd:ee) ]
 676       *      as long as you don't mind hit counts.
 677       *
 678       * intermediate representation consists of the following parts:
 679       *
 680       *   ( )           - group
 681       *   AND           - logical and
 682       *   OR            - logical or
 683       *   NOT           - logical not
 684       *   W+:, W-:, W_: - word      (underscore: no need to highlight)
 685       *   P+:, P-:      - phrase    (minus sign: logically in NOT group)
 686       *   N+:, N-:      - namespace
 687       */
 688      $parsed_query = '';
 689      $parens_level = 0;
 690      $terms = preg_split('/(-?".*?")/u', \dokuwiki\Utf8\PhpString::strtolower($query),
 691          -1, PREG_SPLIT_DELIM_CAPTURE | PREG_SPLIT_NO_EMPTY);
 692  
 693      foreach ($terms as $term) {
 694          $parsed = '';
 695          if (preg_match('/^(-?)"(.+)"$/u', $term, $matches)) {
 696              // phrase-include and phrase-exclude
 697              $not = $matches[1] ? 'NOT' : '';
 698              $parsed = $not.ft_termParser($Indexer, $matches[2], false, true);
 699          } else {
 700              // fix incomplete phrase
 701              $term = str_replace('"', ' ', $term);
 702  
 703              // fix parentheses
 704              $term = str_replace(')'  , ' ) ', $term);
 705              $term = str_replace('('  , ' ( ', $term);
 706              $term = str_replace('- (', ' -(', $term);
 707  
 708              // treat pipe symbols as 'OR' operators
 709              $term = str_replace('|', ' or ', $term);
 710  
 711              // treat ideographic spaces (U+3000) as search term separators
 712              // FIXME: some more separators?
 713              $term = preg_replace('/[ \x{3000}]+/u', ' ',  $term);
 714              $term = trim($term);
 715              if ($term === '') continue;
 716  
 717              $tokens = explode(' ', $term);
 718              foreach ($tokens as $token) {
 719                  if ($token === '(') {
 720                      // parenthesis-include-open
 721                      $parsed .= '(';
 722                      ++$parens_level;
 723                  } elseif ($token === '-(') {
 724                      // parenthesis-exclude-open
 725                      $parsed .= 'NOT(';
 726                      ++$parens_level;
 727                  } elseif ($token === ')') {
 728                      // parenthesis-any-close
 729                      if ($parens_level === 0) continue;
 730                      $parsed .= ')';
 731                      $parens_level--;
 732                  } elseif ($token === 'and') {
 733                      // logical-and (do nothing)
 734                  } elseif ($token === 'or') {
 735                      // logical-or
 736                      $parsed .= 'OR';
 737                  } elseif (preg_match('/^(?:\^|-ns:)(.+)$/u', $token, $matches)) {
 738                      // namespace-exclude
 739                      $parsed .= 'NOT(N+:'.$matches[1].')';
 740                  } elseif (preg_match('/^(?:@|ns:)(.+)$/u', $token, $matches)) {
 741                      // namespace-include
 742                      $parsed .= '(N+:'.$matches[1].')';
 743                  } elseif (preg_match('/^-(.+)$/', $token, $matches)) {
 744                      // word-exclude
 745                      $parsed .= 'NOT('.ft_termParser($Indexer, $matches[1]).')';
 746                  } else {
 747                      // word-include
 748                      $parsed .= ft_termParser($Indexer, $token);
 749                  }
 750              }
 751          }
 752          $parsed_query .= $parsed;
 753      }
 754  
 755      // cleanup (very sensitive)
 756      $parsed_query .= str_repeat(')', $parens_level);
 757      do {
 758          $parsed_query_old = $parsed_query;
 759          $parsed_query = preg_replace('/(NOT)?\(\)/u', '', $parsed_query);
 760      } while ($parsed_query !== $parsed_query_old);
 761      $parsed_query = preg_replace('/(NOT|OR)+\)/u', ')'      , $parsed_query);
 762      $parsed_query = preg_replace('/(OR)+/u'      , 'OR'     , $parsed_query);
 763      $parsed_query = preg_replace('/\(OR/u'       , '('      , $parsed_query);
 764      $parsed_query = preg_replace('/^OR|OR$/u'    , ''       , $parsed_query);
 765      $parsed_query = preg_replace('/\)(NOT)?\(/u' , ')AND$1(', $parsed_query);
 766  
 767      // adjustment: make highlightings right
 768      $parens_level     = 0;
 769      $notgrp_levels    = array();
 770      $parsed_query_new = '';
 771      $tokens = preg_split('/(NOT\(|[()])/u', $parsed_query, -1, PREG_SPLIT_DELIM_CAPTURE | PREG_SPLIT_NO_EMPTY);
 772      foreach ($tokens as $token) {
 773          if ($token === 'NOT(') {
 774              $notgrp_levels[] = ++$parens_level;
 775          } elseif ($token === '(') {
 776              ++$parens_level;
 777          } elseif ($token === ')') {
 778              if ($parens_level-- === end($notgrp_levels)) array_pop($notgrp_levels);
 779          } elseif (count($notgrp_levels) % 2 === 1) {
 780              // turn highlight-flag off if terms are logically in "NOT" group
 781              $token = preg_replace('/([WPN])\+\:/u', '$1-:', $token);
 782          }
 783          $parsed_query_new .= $token;
 784      }
 785      $parsed_query = $parsed_query_new;
 786  
 787      /**
 788       * convert infix notation string into postfix (Reverse Polish notation) array
 789       * by Shunting-yard algorithm
 790       *
 791       * see: http://en.wikipedia.org/wiki/Reverse_Polish_notation
 792       * see: http://en.wikipedia.org/wiki/Shunting-yard_algorithm
 793       */
 794      $parsed_ary     = array();
 795      $ope_stack      = array();
 796      $ope_precedence = array(')' => 1, 'OR' => 2, 'AND' => 3, 'NOT' => 4, '(' => 5);
 797      $ope_regex      = '/([()]|OR|AND|NOT)/u';
 798  
 799      $tokens = preg_split($ope_regex, $parsed_query, -1, PREG_SPLIT_DELIM_CAPTURE | PREG_SPLIT_NO_EMPTY);
 800      foreach ($tokens as $token) {
 801          if (preg_match($ope_regex, $token)) {
 802              // operator
 803              $last_ope = end($ope_stack);
 804              while ($last_ope !== false && $ope_precedence[$token] <= $ope_precedence[$last_ope] && $last_ope != '(') {
 805                  $parsed_ary[] = array_pop($ope_stack);
 806                  $last_ope = end($ope_stack);
 807              }
 808              if ($token == ')') {
 809                  array_pop($ope_stack); // this array_pop always deletes '('
 810              } else {
 811                  $ope_stack[] = $token;
 812              }
 813          } else {
 814              // operand
 815              $token_decoded = str_replace(array('OP', 'CP'), array('(', ')'), $token);
 816              $parsed_ary[] = $token_decoded;
 817          }
 818      }
 819      $parsed_ary = array_values(array_merge($parsed_ary, array_reverse($ope_stack)));
 820  
 821      // cleanup: each double "NOT" in RPN array actually does nothing
 822      $parsed_ary_count = count($parsed_ary);
 823      for ($i = 1; $i < $parsed_ary_count; ++$i) {
 824          if ($parsed_ary[$i] === 'NOT' && $parsed_ary[$i - 1] === 'NOT') {
 825              unset($parsed_ary[$i], $parsed_ary[$i - 1]);
 826          }
 827      }
 828      $parsed_ary = array_values($parsed_ary);
 829  
 830      // build return value
 831      $q = array();
 832      $q['query']      = $query;
 833      $q['parsed_str'] = $parsed_query;
 834      $q['parsed_ary'] = $parsed_ary;
 835  
 836      foreach ($q['parsed_ary'] as $token) {
 837          if ($token[2] !== ':') continue;
 838          $body = substr($token, 3);
 839  
 840          switch (substr($token, 0, 3)) {
 841              case 'N+:':
 842                       $q['ns'][]        = $body; // for backward compatibility
 843                       break;
 844              case 'N-:':
 845                       $q['notns'][]     = $body; // for backward compatibility
 846                       break;
 847              case 'W_:':
 848                       $q['words'][]     = $body;
 849                       break;
 850              case 'W-:':
 851                       $q['words'][]     = $body;
 852                       $q['not'][]       = $body; // for backward compatibility
 853                       break;
 854              case 'W+:':
 855                       $q['words'][]     = $body;
 856                       $q['highlight'][] = $body;
 857                       $q['and'][]       = $body; // for backward compatibility
 858                       break;
 859              case 'P-:':
 860                       $q['phrases'][]   = $body;
 861                       break;
 862              case 'P+:':
 863                       $q['phrases'][]   = $body;
 864                       $q['highlight'][] = $body;
 865                       break;
 866          }
 867      }
 868      foreach (array('words', 'phrases', 'highlight', 'ns', 'notns', 'and', 'not') as $key) {
 869          $q[$key] = empty($q[$key]) ? array() : array_values(array_unique($q[$key]));
 870      }
 871  
 872      return $q;
 873  }
 874  
 875  /**
 876   * Transforms given search term into intermediate representation
 877   *
 878   * This function is used in ft_queryParser() and not for general purpose use.
 879   *
 880   * @author Kazutaka Miyasaka <kazmiya@gmail.com>
 881   *
 882   * @param dokuwiki\Search\Indexer $Indexer
 883   * @param string                  $term
 884   * @param bool                    $consider_asian
 885   * @param bool                    $phrase_mode
 886   * @return string
 887   */
 888  function ft_termParser($Indexer, $term, $consider_asian = true, $phrase_mode = false) {
 889      $parsed = '';
 890      if ($consider_asian) {
 891          // successive asian characters need to be searched as a phrase
 892          $words = \dokuwiki\Utf8\Asian::splitAsianWords($term);
 893          foreach ($words as $word) {
 894              $phrase_mode = $phrase_mode ? true : \dokuwiki\Utf8\Asian::isAsianWords($word);
 895              $parsed .= ft_termParser($Indexer, $word, false, $phrase_mode);
 896          }
 897      } else {
 898          $term_noparen = str_replace(array('(', ')'), ' ', $term);
 899          $words = $Indexer->tokenizer($term_noparen, true);
 900  
 901          // W_: no need to highlight
 902          if (empty($words)) {
 903              $parsed = '()'; // important: do not remove
 904          } elseif ($words[0] === $term) {
 905              $parsed = '(W+:'.$words[0].')';
 906          } elseif ($phrase_mode) {
 907              $term_encoded = str_replace(array('(', ')'), array('OP', 'CP'), $term);
 908              $parsed = '((W_:'.implode(')(W_:', $words).')(P+:'.$term_encoded.'))';
 909          } else {
 910              $parsed = '((W+:'.implode(')(W+:', $words).'))';
 911          }
 912      }
 913      return $parsed;
 914  }
 915  
 916  /**
 917   * Recreate a search query string based on parsed parts, doesn't support negated phrases and `OR` searches
 918   *
 919   * @param array $and
 920   * @param array $not
 921   * @param array $phrases
 922   * @param array $ns
 923   * @param array $notns
 924   *
 925   * @return string
 926   */
 927  function ft_queryUnparser_simple(array $and, array $not, array $phrases, array $ns, array $notns) {
 928      $query = implode(' ', $and);
 929      if (!empty($not)) {
 930          $query .= ' -' . implode(' -', $not);
 931      }
 932  
 933      if (!empty($phrases)) {
 934          $query .= ' "' . implode('" "', $phrases) . '"';
 935      }
 936  
 937      if (!empty($ns)) {
 938          $query .= ' @' . implode(' @', $ns);
 939      }
 940  
 941      if (!empty($notns)) {
 942          $query .= ' ^' . implode(' ^', $notns);
 943      }
 944  
 945      return $query;
 946  }
 947  
 948  //Setup VIM: ex: et ts=4 :