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 use dokuwiki\Utf8\Asian;
   9 use dokuwiki\Search\Indexer;
  10 use dokuwiki\Extension\Event;
  11 use dokuwiki\Utf8\Clean;
  12 use dokuwiki\Utf8\PhpString;
  13 use dokuwiki\Utf8\Sort;
  14
  15 /**
  16  * create snippets for the first few results only
  17  */
  18 if (!defined('FT_SNIPPET_NUMBER')) define('FT_SNIPPET_NUMBER', 15);
  19
  20 /**
  21  * The fulltext search
  22  *
  23  * Returns a list of matching documents for the given query
  24  *
  25  * refactored into ft_pageSearch(), _ft_pageSearch() and trigger_event()
  26  *
  27  * @param string     $query
  28  * @param array      $highlight
  29  * @param string     $sort
  30  * @param int|string $after  only show results with mtime after this date, accepts timestap or strtotime arguments
  31  * @param int|string $before only show results with mtime before this date, accepts timestap or strtotime arguments
  32  *
  33  * @return array
  34  */
  35 function ft_pageSearch($query, &$highlight, $sort = null, $after = null, $before = null)
  36 {
  37
  38     if ($sort === null) {
  39         $sort = 'hits';
  40     }
  41     $data = [
  42         'query' => $query,
  43         'sort' => $sort,
  44         'after' => $after,
  45         'before' => $before
  46     ];
  47     $data['highlight'] =& $highlight;
  48
  49     return Event::createAndTrigger('SEARCH_QUERY_FULLPAGE', $data, '_ft_pageSearch');
  50 }
  51
  52 /**
  53  * Returns a list of matching documents for the given query
  54  *
  55  * @author Andreas Gohr <andi@splitbrain.org>
  56  * @author Kazutaka Miyasaka <kazmiya@gmail.com>
  57  *
  58  * @param array $data event data
  59  * @return array matching documents
  60  */
  61 function _ft_pageSearch(&$data)
  62 {
  63     $Indexer = idx_get_indexer();
  64
  65     // parse the given query
  66     $q = ft_queryParser($Indexer, $data['query']);
  67     $data['highlight'] = $q['highlight'];
  68
  69     if (empty($q['parsed_ary'])) return [];
  70
  71     // lookup all words found in the query
  72     $lookup = $Indexer->lookup($q['words']);
  73
  74     // get all pages in this dokuwiki site (!: includes nonexistent pages)
  75     $pages_all = [];
  76     foreach ($Indexer->getPages() as $id) {
  77         $pages_all[$id] = 0; // base: 0 hit
  78     }
  79
  80     // process the query
  81     $stack = [];
  82     foreach ($q['parsed_ary'] as $token) {
  83         switch (substr($token, 0, 3)) {
  84             case 'W+:':
  85             case 'W-:':
  86             case 'W_:': // word
  87                 $word    = substr($token, 3);
  88                 if (isset($lookup[$word])) {
  89                     $stack[] = (array)$lookup[$word];
  90                 }
  91                 break;
  92             case 'P+:':
  93             case 'P-:': // phrase
  94                 $phrase = substr($token, 3);
  95                 // since phrases are always parsed as ((W1)(W2)...(P)),
  96                 // the end($stack) always points the pages that contain
  97                 // all words in this phrase
  98                 $pages  = end($stack);
  99                 $pages_matched = [];
 100                 foreach (array_keys($pages) as $id) {
 101                     $evdata = [
 102                         'id' => $id,
 103                         'phrase' => $phrase,
 104                         'text' => rawWiki($id)
 105                     ];
 106                     $evt = new Event('FULLTEXT_PHRASE_MATCH', $evdata);
 107                     if ($evt->advise_before() && $evt->result !== true) {
 108                         $text = PhpString::strtolower($evdata['text']);
 109                         if (strpos($text, $phrase) !== false) {
 110                             $evt->result = true;
 111                         }
 112                     }
 113                     $evt->advise_after();
 114                     if ($evt->result === true) {
 115                         $pages_matched[$id] = 0; // phrase: always 0 hit
 116                     }
 117                 }
 118                 $stack[] = $pages_matched;
 119                 break;
 120             case 'N+:':
 121             case 'N-:': // namespace
 122                 $ns = cleanID(substr($token, 3)) . ':';
 123                 $pages_matched = [];
 124                 foreach (array_keys($pages_all) as $id) {
 125                     if (strpos($id, $ns) === 0) {
 126                         $pages_matched[$id] = 0; // namespace: always 0 hit
 127                     }
 128                 }
 129                 $stack[] = $pages_matched;
 130                 break;
 131             case 'AND': // and operation
 132                 [$pages1, $pages2] = array_splice($stack, -2);
 133                 $stack[] = ft_resultCombine([$pages1, $pages2]);
 134                 break;
 135             case 'OR':  // or operation
 136                 [$pages1, $pages2] = array_splice($stack, -2);
 137                 $stack[] = ft_resultUnite([$pages1, $pages2]);
 138                 break;
 139             case 'NOT': // not operation (unary)
 140                 $pages   = array_pop($stack);
 141                 $stack[] = ft_resultComplement([$pages_all, $pages]);
 142                 break;
 143         }
 144     }
 145     $docs = array_pop($stack);
 146
 147     if (empty($docs)) return [];
 148
 149     // check: settings, acls, existence
 150     foreach (array_keys($docs) as $id) {
 151         if (isHiddenPage($id) || auth_quickaclcheck($id) < AUTH_READ || !page_exists($id, '', false)) {
 152             unset($docs[$id]);
 153         }
 154     }
 155
 156     $docs = _ft_filterResultsByTime($docs, $data['after'], $data['before']);
 157
 158     if ($data['sort'] === 'mtime') {
 159         uksort($docs, 'ft_pagemtimesorter');
 160     } else {
 161         // sort docs by count
 162         uksort($docs, 'ft_pagesorter');
 163         arsort($docs);
 164     }
 165
 166     return $docs;
 167 }
 168
 169 /**
 170  * Returns the backlinks for a given page
 171  *
 172  * Uses the metadata index.
 173  *
 174  * @param string $id           The id for which links shall be returned
 175  * @param bool   $ignore_perms Ignore the fact that pages are hidden or read-protected
 176  * @return array The pages that contain links to the given page
 177  */
 178 function ft_backlinks($id, $ignore_perms = false)
 179 {
 180     $result = idx_get_indexer()->lookupKey('relation_references', $id);
 181
 182     if ($result === []) return $result;
 183
 184     // check ACL permissions
 185     foreach (array_keys($result) as $idx) {
 186         if (
 187             (!$ignore_perms && (
 188                 isHiddenPage($result[$idx]) || auth_quickaclcheck($result[$idx]) < AUTH_READ
 189             )) || !page_exists($result[$idx], '', false)
 190         ) {
 191             unset($result[$idx]);
 192         }
 193     }
 194
 195     Sort::sort($result);
 196     return $result;
 197 }
 198
 199 /**
 200  * Returns the pages that use a given media file
 201  *
 202  * Uses the relation media metadata property and the metadata index.
 203  *
 204  * Note that before 2013-07-31 the second parameter was the maximum number of results and
 205  * permissions were ignored. That's why the parameter is now checked to be explicitely set
 206  * to true (with type bool) in order to be compatible with older uses of the function.
 207  *
 208  * @param string $id           The media id to look for
 209  * @param bool   $ignore_perms Ignore hidden pages and acls (optional, default: false)
 210  * @return array A list of pages that use the given media file
 211  */
 212 function ft_mediause($id, $ignore_perms = false)
 213 {
 214     $result = idx_get_indexer()->lookupKey('relation_media', $id);
 215
 216     if ($result === []) return $result;
 217
 218     // check ACL permissions
 219     foreach (array_keys($result) as $idx) {
 220         if (
 221             (!$ignore_perms && (
 222                     isHiddenPage($result[$idx]) || auth_quickaclcheck($result[$idx]) < AUTH_READ
 223                 )) || !page_exists($result[$idx], '', false)
 224         ) {
 225             unset($result[$idx]);
 226         }
 227     }
 228
 229     Sort::sort($result);
 230     return $result;
 231 }
 232
 233
 234 /**
 235  * Quicksearch for pagenames
 236  *
 237  * By default it only matches the pagename and ignores the
 238  * namespace. This can be changed with the second parameter.
 239  * The third parameter allows to search in titles as well.
 240  *
 241  * The function always returns titles as well
 242  *
 243  * @triggers SEARCH_QUERY_PAGELOOKUP
 244  * @author   Andreas Gohr <andi@splitbrain.org>
 245  * @author   Adrian Lang <lang@cosmocode.de>
 246  *
 247  * @param string     $id       page id
 248  * @param bool       $in_ns    match against namespace as well?
 249  * @param bool       $in_title search in title?
 250  * @param int|string $after    only show results with mtime after this date, accepts timestap or strtotime arguments
 251  * @param int|string $before   only show results with mtime before this date, accepts timestap or strtotime arguments
 252  *
 253  * @return string[]
 254  */
 255 function ft_pageLookup($id, $in_ns = false, $in_title = false, $after = null, $before = null)
 256 {
 257     $data = [
 258         'id' => $id,
 259         'in_ns' => $in_ns,
 260         'in_title' => $in_title,
 261         'after' => $after,
 262         'before' => $before
 263     ];
 264     $data['has_titles'] = true; // for plugin backward compatibility check
 265     return Event::createAndTrigger('SEARCH_QUERY_PAGELOOKUP', $data, '_ft_pageLookup');
 266 }
 267
 268 /**
 269  * Returns list of pages as array(pageid => First Heading)
 270  *
 271  * @param array &$data event data
 272  * @return string[]
 273  */
 274 function _ft_pageLookup(&$data)
 275 {
 276     // split out original parameters
 277     $id = $data['id'];
 278     $Indexer = idx_get_indexer();
 279     $parsedQuery = ft_queryParser($Indexer, $id);
 280     if (count($parsedQuery['ns']) > 0) {
 281         $ns = cleanID($parsedQuery['ns'][0]) . ':';
 282         $id = implode(' ', $parsedQuery['highlight']);
 283     }
 284     if (count($parsedQuery['notns']) > 0) {
 285         $notns = cleanID($parsedQuery['notns'][0]) . ':';
 286         $id = implode(' ', $parsedQuery['highlight']);
 287     }
 288
 289     $in_ns    = $data['in_ns'];
 290     $in_title = $data['in_title'];
 291     $cleaned = cleanID($id);
 292
 293     $Indexer = idx_get_indexer();
 294     $page_idx = $Indexer->getPages();
 295
 296     $pages = [];
 297     if ($id !== '' && $cleaned !== '') {
 298         foreach ($page_idx as $p_id) {
 299             if ((strpos($in_ns ? $p_id : noNSorNS($p_id), $cleaned) !== false)) {
 300                 if (!isset($pages[$p_id]))
 301                     $pages[$p_id] = p_get_first_heading($p_id, METADATA_DONT_RENDER);
 302             }
 303         }
 304         if ($in_title) {
 305             foreach ($Indexer->lookupKey('title', $id, '_ft_pageLookupTitleCompare') as $p_id) {
 306                 if (!isset($pages[$p_id]))
 307                     $pages[$p_id] = p_get_first_heading($p_id, METADATA_DONT_RENDER);
 308             }
 309         }
 310     }
 311
 312     if (isset($ns)) {
 313         foreach (array_keys($pages) as $p_id) {
 314             if (strpos($p_id, $ns) !== 0) {
 315                 unset($pages[$p_id]);
 316             }
 317         }
 318     }
 319     if (isset($notns)) {
 320         foreach (array_keys($pages) as $p_id) {
 321             if (strpos($p_id, $notns) === 0) {
 322                 unset($pages[$p_id]);
 323             }
 324         }
 325     }
 326
 327     // discard hidden pages
 328     // discard nonexistent pages
 329     // check ACL permissions
 330     foreach (array_keys($pages) as $idx) {
 331         if (
 332             !isVisiblePage($idx) || !page_exists($idx) ||
 333             auth_quickaclcheck($idx) < AUTH_READ
 334         ) {
 335             unset($pages[$idx]);
 336         }
 337     }
 338
 339     $pages = _ft_filterResultsByTime($pages, $data['after'], $data['before']);
 340
 341     uksort($pages, 'ft_pagesorter');
 342     return $pages;
 343 }
 344
 345
 346 /**
 347  * @param array      $results search results in the form pageid => value
 348  * @param int|string $after   only returns results with mtime after this date, accepts timestap or strtotime arguments
 349  * @param int|string $before  only returns results with mtime after this date, accepts timestap or strtotime arguments
 350  *
 351  * @return array
 352  */
 353 function _ft_filterResultsByTime(array $results, $after, $before)
 354 {
 355     if ($after || $before) {
 356         $after = is_int($after) ? $after : strtotime($after);
 357         $before = is_int($before) ? $before : strtotime($before);
 358
 359         foreach (array_keys($results) as $id) {
 360             $mTime = filemtime(wikiFN($id));
 361             if ($after && $after > $mTime) {
 362                 unset($results[$id]);
 363                 continue;
 364             }
 365             if ($before && $before < $mTime) {
 366                 unset($results[$id]);
 367             }
 368         }
 369     }
 370
 371     return $results;
 372 }
 373
 374 /**
 375  * Tiny helper function for comparing the searched title with the title
 376  * from the search index. This function is a wrapper around stripos with
 377  * adapted argument order and return value.
 378  *
 379  * @param string $search searched title
 380  * @param string $title  title from index
 381  * @return bool
 382  */
 383 function _ft_pageLookupTitleCompare($search, $title)
 384 {
 385     if (Clean::isASCII($search)) {
 386         $pos = stripos($title, $search);
 387     } else {
 388         $pos = PhpString::strpos(
 389             PhpString::strtolower($title),
 390             PhpString::strtolower($search)
 391         );
 392     }
 393
 394     return $pos !== false;
 395 }
 396
 397 /**
 398  * Sort pages based on their namespace level first, then on their string
 399  * values. This makes higher hierarchy pages rank higher than lower hierarchy
 400  * pages.
 401  *
 402  * @param string $a
 403  * @param string $b
 404  * @return int Returns < 0 if $a is less than $b; > 0 if $a is greater than $b, and 0 if they are equal.
 405  */
 406 function ft_pagesorter($a, $b)
 407 {
 408     $ac = count(explode(':', $a));
 409     $bc = count(explode(':', $b));
 410     if ($ac < $bc) {
 411         return -1;
 412     } elseif ($ac > $bc) {
 413         return 1;
 414     }
 415     return Sort::strcmp($a, $b);
 416 }
 417
 418 /**
 419  * Sort pages by their mtime, from newest to oldest
 420  *
 421  * @param string $a
 422  * @param string $b
 423  *
 424  * @return int Returns < 0 if $a is newer than $b, > 0 if $b is newer than $a and 0 if they are of the same age
 425  */
 426 function ft_pagemtimesorter($a, $b)
 427 {
 428     $mtimeA = filemtime(wikiFN($a));
 429     $mtimeB = filemtime(wikiFN($b));
 430     return $mtimeB - $mtimeA;
 431 }
 432
 433 /**
 434  * Creates a snippet extract
 435  *
 436  * @author Andreas Gohr <andi@splitbrain.org>
 437  * @triggers FULLTEXT_SNIPPET_CREATE
 438  *
 439  * @param string $id page id
 440  * @param array $highlight
 441  * @return mixed
 442  */
 443 function ft_snippet($id, $highlight)
 444 {
 445     $text = rawWiki($id);
 446     $text = str_replace("\xC2\xAD", '', $text);
 447      // remove soft-hyphens
 448     $evdata = [
 449         'id'        => $id,
 450         'text'      => &$text,
 451         'highlight' => &$highlight,
 452         'snippet'   => ''
 453     ];
 454
 455     $evt = new Event('FULLTEXT_SNIPPET_CREATE', $evdata);
 456     if ($evt->advise_before()) {
 457         $match = [];
 458         $snippets = [];
 459         $utf8_offset = 0;
 460         $offset = 0;
 461         $end = 0;
 462         $len = PhpString::strlen($text);
 463
 464         // build a regexp from the phrases to highlight
 465         $re1 = '(' .
 466             implode(
 467                 '|',
 468                 array_map(
 469                     'ft_snippet_re_preprocess',
 470                     array_map(
 471                         'preg_quote_cb',
 472                         array_filter((array) $highlight)
 473                     )
 474                 )
 475             ) .
 476             ')';
 477         $re2 = "$re1.{0,75}(?!\\1)$re1";
 478         $re3 = "$re1.{0,45}(?!\\1)$re1.{0,45}(?!\\1)(?!\\2)$re1";
 479
 480         for ($cnt=4; $cnt--;) {
 481             if (0) {
 482             } elseif (preg_match('/'.$re3.'/iu', $text, $match, PREG_OFFSET_CAPTURE, $offset)) {
 483             } elseif (preg_match('/'.$re2.'/iu', $text, $match, PREG_OFFSET_CAPTURE, $offset)) {
 484             } elseif (preg_match('/'.$re1.'/iu', $text, $match, PREG_OFFSET_CAPTURE, $offset)) {
 485             } else {
 486                 break;
 487             }
 488
 489             [$str, $idx] = $match[0];
 490
 491             // convert $idx (a byte offset) into a utf8 character offset
 492             $utf8_idx = PhpString::strlen(substr($text, 0, $idx));
 493             $utf8_len = PhpString::strlen($str);
 494
 495             // establish context, 100 bytes surrounding the match string
 496             // first look to see if we can go 100 either side,
 497             // then drop to 50 adding any excess if the other side can't go to 50,
 498             $pre = min($utf8_idx-$utf8_offset, 100);
 499             $post = min($len-$utf8_idx-$utf8_len, 100);
 500
 501             if ($pre>50 && $post>50) {
 502                 $pre = 50;
 503                 $post = 50;
 504             } elseif ($pre>50) {
 505                 $pre = min($pre, 100-$post);
 506             } elseif ($post>50) {
 507                 $post = min($post, 100-$pre);
 508             } elseif ($offset == 0) {
 509                 // both are less than 50, means the context is the whole string
 510                 // make it so and break out of this loop - there is no need for the
 511                 // complex snippet calculations
 512                 $snippets = [$text];
 513                 break;
 514             }
 515
 516             // establish context start and end points, try to append to previous
 517             // context if possible
 518             $start = $utf8_idx - $pre;
 519             $append = ($start < $end) ? $end : false;  // still the end of the previous context snippet
 520             $end = $utf8_idx + $utf8_len + $post;      // now set it to the end of this context
 521
 522             if ($append) {
 523                 $snippets[count($snippets)-1] .= PhpString::substr($text, $append, $end-$append);
 524             } else {
 525                 $snippets[] = PhpString::substr($text, $start, $end-$start);
 526             }
 527
 528             // set $offset for next match attempt
 529             // continue matching after the current match
 530             // if the current match is not the longest possible match starting at the current offset
 531             // this prevents further matching of this snippet but for possible matches of length
 532             // smaller than match length + context (at least 50 characters) this match is part of the context
 533             $utf8_offset = $utf8_idx + $utf8_len;
 534             $offset = $idx + strlen(PhpString::substr($text, $utf8_idx, $utf8_len));
 535             $offset = Clean::correctIdx($text, $offset);
 536         }
 537
 538         $m = "\1";
 539         $snippets = preg_replace('/'.$re1.'/iu', $m.'$1'.$m, $snippets);
 540         $snippet = preg_replace(
 541             '/' . $m . '([^' . $m . ']*?)' . $m . '/iu',
 542             '<strong class="search_hit">$1</strong>',
 543             hsc(implode('... ', $snippets))
 544         );
 545
 546         $evdata['snippet'] = $snippet;
 547     }
 548     $evt->advise_after();
 549     unset($evt);
 550
 551     return $evdata['snippet'];
 552 }
 553
 554 /**
 555  * Wraps a search term in regex boundary checks.
 556  *
 557  * @param string $term
 558  * @return string
 559  */
 560 function ft_snippet_re_preprocess($term)
 561 {
 562     // do not process asian terms where word boundaries are not explicit
 563     if (Asian::isAsianWords($term)) return $term;
 564
 565     if (UTF8_PROPERTYSUPPORT) {
 566         // unicode word boundaries
 567         // see http://stackoverflow.com/a/2449017/172068
 568         $BL = '(?<!\pL)';
 569         $BR = '(?!\pL)';
 570     } else {
 571         // not as correct as above, but at least won't break
 572         $BL = '\b';
 573         $BR = '\b';
 574     }
 575
 576     if (substr($term, 0, 2) == '\\*') {
 577         $term = substr($term, 2);
 578     } else {
 579         $term = $BL.$term;
 580     }
 581
 582     if (substr($term, -2, 2) == '\\*') {
 583         $term = substr($term, 0, -2);
 584     } else {
 585         $term .= $BR;
 586     }
 587
 588     if ($term == $BL || $term == $BR || $term == $BL.$BR) $term = '';
 589     return $term;
 590 }
 591
 592 /**
 593  * Combine found documents and sum up their scores
 594  *
 595  * This function is used to combine searched words with a logical
 596  * AND. Only documents available in all arrays are returned.
 597  *
 598  * based upon PEAR's PHP_Compat function for array_intersect_key()
 599  *
 600  * @param array $args An array of page arrays
 601  * @return array
 602  */
 603 function ft_resultCombine($args)
 604 {
 605     $array_count = count($args);
 606     if ($array_count == 1) {
 607         return $args[0];
 608     }
 609
 610     $result = [];
 611     if ($array_count > 1) {
 612         foreach ($args[0] as $key => $value) {
 613             $result[$key] = $value;
 614             for ($i = 1; $i !== $array_count; $i++) {
 615                 if (!isset($args[$i][$key])) {
 616                     unset($result[$key]);
 617                     break;
 618                 }
 619                 $result[$key] += $args[$i][$key];
 620             }
 621         }
 622     }
 623     return $result;
 624 }
 625
 626 /**
 627  * Unites found documents and sum up their scores
 628  *
 629  * based upon ft_resultCombine() function
 630  *
 631  * @param array $args An array of page arrays
 632  * @return array
 633  *
 634  * @author Kazutaka Miyasaka <kazmiya@gmail.com>
 635  */
 636 function ft_resultUnite($args)
 637 {
 638     $array_count = count($args);
 639     if ($array_count === 1) {
 640         return $args[0];
 641     }
 642
 643     $result = $args[0];
 644     for ($i = 1; $i !== $array_count; $i++) {
 645         foreach (array_keys($args[$i]) as $id) {
 646             $result[$id] += $args[$i][$id];
 647         }
 648     }
 649     return $result;
 650 }
 651
 652 /**
 653  * Computes the difference of documents using page id for comparison
 654  *
 655  * nearly identical to PHP5's array_diff_key()
 656  *
 657  * @param array $args An array of page arrays
 658  * @return array
 659  *
 660  * @author Kazutaka Miyasaka <kazmiya@gmail.com>
 661  */
 662 function ft_resultComplement($args)
 663 {
 664     $array_count = count($args);
 665     if ($array_count === 1) {
 666         return $args[0];
 667     }
 668
 669     $result = $args[0];
 670     foreach (array_keys($result) as $id) {
 671         for ($i = 1; $i !== $array_count; $i++) {
 672             if (isset($args[$i][$id])) unset($result[$id]);
 673         }
 674     }
 675     return $result;
 676 }
 677
 678 /**
 679  * Parses a search query and builds an array of search formulas
 680  *
 681  * @author Andreas Gohr <andi@splitbrain.org>
 682  * @author Kazutaka Miyasaka <kazmiya@gmail.com>
 683  *
 684  * @param Indexer $Indexer
 685  * @param string                  $query search query
 686  * @return array of search formulas
 687  */
 688 function ft_queryParser($Indexer, $query)
 689 {
 690     /**
 691      * parse a search query and transform it into intermediate representation
 692      *
 693      * in a search query, you can use the following expressions:
 694      *
 695      *   words:
 696      *     include
 697      *     -exclude
 698      *   phrases:
 699      *     "phrase to be included"
 700      *     -"phrase you want to exclude"
 701      *   namespaces:
 702      *     @include:namespace (or ns:include:namespace)
 703      *     ^exclude:namespace (or -ns:exclude:namespace)
 704      *   groups:
 705      *     ()
 706      *     -()
 707      *   operators:
 708      *     and ('and' is the default operator: you can always omit this)
 709      *     or  (or pipe symbol '|', lower precedence than 'and')
 710      *
 711      * e.g. a query [ aa "bb cc" @dd:ee ] means "search pages which contain
 712      *      a word 'aa', a phrase 'bb cc' and are within a namespace 'dd:ee'".
 713      *      this query is equivalent to [ -(-aa or -"bb cc" or -ns:dd:ee) ]
 714      *      as long as you don't mind hit counts.
 715      *
 716      * intermediate representation consists of the following parts:
 717      *
 718      *   ( )           - group
 719      *   AND           - logical and
 720      *   OR            - logical or
 721      *   NOT           - logical not
 722      *   W+:, W-:, W_: - word      (underscore: no need to highlight)
 723      *   P+:, P-:      - phrase    (minus sign: logically in NOT group)
 724      *   N+:, N-:      - namespace
 725      */
 726     $parsed_query = '';
 727     $parens_level = 0;
 728     $terms = preg_split(
 729         '/(-?".*?")/u',
 730         PhpString::strtolower($query),
 731         -1,
 732         PREG_SPLIT_DELIM_CAPTURE | PREG_SPLIT_NO_EMPTY
 733     );
 734
 735     foreach ($terms as $term) {
 736         $parsed = '';
 737         if (preg_match('/^(-?)"(.+)"$/u', $term, $matches)) {
 738             // phrase-include and phrase-exclude
 739             $not = $matches[1] ? 'NOT' : '';
 740             $parsed = $not.ft_termParser($Indexer, $matches[2], false, true);
 741         } else {
 742             // fix incomplete phrase
 743             $term = str_replace('"', ' ', $term);
 744
 745             // fix parentheses
 746             $term = str_replace(')', ' ) ', $term);
 747             $term = str_replace('(', ' ( ', $term);
 748             $term = str_replace('- (', ' -(', $term);
 749
 750             // treat pipe symbols as 'OR' operators
 751             $term = str_replace('|', ' or ', $term);
 752
 753             // treat ideographic spaces (U+3000) as search term separators
 754             // FIXME: some more separators?
 755             $term = preg_replace('/[ \x{3000}]+/u', ' ', $term);
 756             $term = trim($term);
 757             if ($term === '') continue;
 758
 759             $tokens = explode(' ', $term);
 760             foreach ($tokens as $token) {
 761                 if ($token === '(') {
 762                     // parenthesis-include-open
 763                     $parsed .= '(';
 764                     ++$parens_level;
 765                 } elseif ($token === '-(') {
 766                     // parenthesis-exclude-open
 767                     $parsed .= 'NOT(';
 768                     ++$parens_level;
 769                 } elseif ($token === ')') {
 770                     // parenthesis-any-close
 771                     if ($parens_level === 0) continue;
 772                     $parsed .= ')';
 773                     $parens_level--;
 774                 } elseif ($token === 'and') {
 775                     // logical-and (do nothing)
 776                 } elseif ($token === 'or') {
 777                     // logical-or
 778                     $parsed .= 'OR';
 779                 } elseif (preg_match('/^(?:\^|-ns:)(.+)$/u', $token, $matches)) {
 780                     // namespace-exclude
 781                     $parsed .= 'NOT(N+:'.$matches[1].')';
 782                 } elseif (preg_match('/^(?:@|ns:)(.+)$/u', $token, $matches)) {
 783                     // namespace-include
 784                     $parsed .= '(N+:'.$matches[1].')';
 785                 } elseif (preg_match('/^-(.+)$/', $token, $matches)) {
 786                     // word-exclude
 787                     $parsed .= 'NOT('.ft_termParser($Indexer, $matches[1]).')';
 788                 } else {
 789                     // word-include
 790                     $parsed .= ft_termParser($Indexer, $token);
 791                 }
 792             }
 793         }
 794         $parsed_query .= $parsed;
 795     }
 796
 797     // cleanup (very sensitive)
 798     $parsed_query .= str_repeat(')', $parens_level);
 799     do {
 800         $parsed_query_old = $parsed_query;
 801         $parsed_query = preg_replace('/(NOT)?\(\)/u', '', $parsed_query);
 802     } while ($parsed_query !== $parsed_query_old);
 803     $parsed_query = preg_replace('/(NOT|OR)+\)/u', ')', $parsed_query);
 804     $parsed_query = preg_replace('/(OR)+/u', 'OR', $parsed_query);
 805     $parsed_query = preg_replace('/\(OR/u', '(', $parsed_query);
 806     $parsed_query = preg_replace('/^OR|OR$/u', '', $parsed_query);
 807     $parsed_query = preg_replace('/\)(NOT)?\(/u', ')AND$1(', $parsed_query);
 808
 809     // adjustment: make highlightings right
 810     $parens_level     = 0;
 811     $notgrp_levels    = [];
 812     $parsed_query_new = '';
 813     $tokens = preg_split('/(NOT\(|[()])/u', $parsed_query, -1, PREG_SPLIT_DELIM_CAPTURE | PREG_SPLIT_NO_EMPTY);
 814     foreach ($tokens as $token) {
 815         if ($token === 'NOT(') {
 816             $notgrp_levels[] = ++$parens_level;
 817         } elseif ($token === '(') {
 818             ++$parens_level;
 819         } elseif ($token === ')') {
 820             if ($parens_level-- === end($notgrp_levels)) array_pop($notgrp_levels);
 821         } elseif (count($notgrp_levels) % 2 === 1) {
 822             // turn highlight-flag off if terms are logically in "NOT" group
 823             $token = preg_replace('/([WPN])\+\:/u', '$1-:', $token);
 824         }
 825         $parsed_query_new .= $token;
 826     }
 827     $parsed_query = $parsed_query_new;
 828
 829     /**
 830      * convert infix notation string into postfix (Reverse Polish notation) array
 831      * by Shunting-yard algorithm
 832      *
 833      * see: http://en.wikipedia.org/wiki/Reverse_Polish_notation
 834      * see: http://en.wikipedia.org/wiki/Shunting-yard_algorithm
 835      */
 836     $parsed_ary     = [];
 837     $ope_stack      = [];
 838     $ope_precedence = [')' => 1, 'OR' => 2, 'AND' => 3, 'NOT' => 4, '(' => 5];
 839     $ope_regex      = '/([()]|OR|AND|NOT)/u';
 840
 841     $tokens = preg_split($ope_regex, $parsed_query, -1, PREG_SPLIT_DELIM_CAPTURE | PREG_SPLIT_NO_EMPTY);
 842     foreach ($tokens as $token) {
 843         if (preg_match($ope_regex, $token)) {
 844             // operator
 845             $last_ope = end($ope_stack);
 846             while ($last_ope !== false && $ope_precedence[$token] <= $ope_precedence[$last_ope] && $last_ope != '(') {
 847                 $parsed_ary[] = array_pop($ope_stack);
 848                 $last_ope = end($ope_stack);
 849             }
 850             if ($token == ')') {
 851                 array_pop($ope_stack); // this array_pop always deletes '('
 852             } else {
 853                 $ope_stack[] = $token;
 854             }
 855         } else {
 856             // operand
 857             $token_decoded = str_replace(['OP', 'CP'], ['(', ')'], $token);
 858             $parsed_ary[] = $token_decoded;
 859         }
 860     }
 861     $parsed_ary = array_values([...$parsed_ary, ...array_reverse($ope_stack)]);
 862
 863     // cleanup: each double "NOT" in RPN array actually does nothing
 864     $parsed_ary_count = count($parsed_ary);
 865     for ($i = 1; $i < $parsed_ary_count; ++$i) {
 866         if ($parsed_ary[$i] === 'NOT' && $parsed_ary[$i - 1] === 'NOT') {
 867             unset($parsed_ary[$i], $parsed_ary[$i - 1]);
 868         }
 869     }
 870     $parsed_ary = array_values($parsed_ary);
 871
 872     // build return value
 873     $q = [];
 874     $q['query']      = $query;
 875     $q['parsed_str'] = $parsed_query;
 876     $q['parsed_ary'] = $parsed_ary;
 877
 878     foreach ($q['parsed_ary'] as $token) {
 879         if (strlen($token) < 3 || $token[2] !== ':') continue;
 880         $body = substr($token, 3);
 881
 882         switch (substr($token, 0, 3)) {
 883             case 'N+:':
 884                      $q['ns'][]        = $body; // for backward compatibility
 885                 break;
 886             case 'N-:':
 887                      $q['notns'][]     = $body; // for backward compatibility
 888                 break;
 889             case 'W_:':
 890                      $q['words'][]     = $body;
 891                 break;
 892             case 'W-:':
 893                      $q['words'][]     = $body;
 894                      $q['not'][]       = $body; // for backward compatibility
 895                 break;
 896             case 'W+:':
 897                      $q['words'][]     = $body;
 898                      $q['highlight'][] = $body;
 899                      $q['and'][]       = $body; // for backward compatibility
 900                 break;
 901             case 'P-:':
 902                      $q['phrases'][]   = $body;
 903                 break;
 904             case 'P+:':
 905                      $q['phrases'][]   = $body;
 906                      $q['highlight'][] = $body;
 907                 break;
 908         }
 909     }
 910     foreach (['words', 'phrases', 'highlight', 'ns', 'notns', 'and', 'not'] as $key) {
 911         $q[$key] = empty($q[$key]) ? [] : array_values(array_unique($q[$key]));
 912     }
 913
 914     return $q;
 915 }
 916
 917 /**
 918  * Transforms given search term into intermediate representation
 919  *
 920  * This function is used in ft_queryParser() and not for general purpose use.
 921  *
 922  * @author Kazutaka Miyasaka <kazmiya@gmail.com>
 923  *
 924  * @param Indexer $Indexer
 925  * @param string                  $term
 926  * @param bool                    $consider_asian
 927  * @param bool                    $phrase_mode
 928  * @return string
 929  */
 930 function ft_termParser($Indexer, $term, $consider_asian = true, $phrase_mode = false)
 931 {
 932     $parsed = '';
 933     if ($consider_asian) {
 934         // successive asian characters need to be searched as a phrase
 935         $words = Asian::splitAsianWords($term);
 936         foreach ($words as $word) {
 937             $phrase_mode = $phrase_mode ? true : Asian::isAsianWords($word);
 938             $parsed .= ft_termParser($Indexer, $word, false, $phrase_mode);
 939         }
 940     } else {
 941         $term_noparen = str_replace(['(', ')'], ' ', $term);
 942         $words = $Indexer->tokenizer($term_noparen, true);
 943
 944         // W_: no need to highlight
 945         if (empty($words)) {
 946             $parsed = '()'; // important: do not remove
 947         } elseif ($words[0] === $term) {
 948             $parsed = '(W+:'.$words[0].')';
 949         } elseif ($phrase_mode) {
 950             $term_encoded = str_replace(['(', ')'], ['OP', 'CP'], $term);
 951             $parsed = '((W_:'.implode(')(W_:', $words).')(P+:'.$term_encoded.'))';
 952         } else {
 953             $parsed = '((W+:'.implode(')(W+:', $words).'))';
 954         }
 955     }
 956     return $parsed;
 957 }
 958
 959 /**
 960  * Recreate a search query string based on parsed parts, doesn't support negated phrases and `OR` searches
 961  *
 962  * @param array $and
 963  * @param array $not
 964  * @param array $phrases
 965  * @param array $ns
 966  * @param array $notns
 967  *
 968  * @return string
 969  */
 970 function ft_queryUnparser_simple(array $and, array $not, array $phrases, array $ns, array $notns)
 971 {
 972     $query = implode(' ', $and);
 973     if ($not !== []) {
 974         $query .= ' -' . implode(' -', $not);
 975     }
 976
 977     if ($phrases !== []) {
 978         $query .= ' "' . implode('" "', $phrases) . '"';
 979     }
 980
 981     if ($ns !== []) {
 982         $query .= ' @' . implode(' @', $ns);
 983     }
 984
 985     if ($notns !== []) {
 986         $query .= ' ^' . implode(' ^', $notns);
 987     }
 988
 989     return $query;
 990 }
 991
 992 //Setup VIM: ex: et ts=4 :