inc/fulltext.php

   1 <?php
   2
   3 /**
   4  * DokuWiki fulltextsearch functions using the index
   5  *
   6  * @license    GPL 2 (http://www.gnu.org/licenses/gpl.html)
   7  * @author     Andreas Gohr <andi@splitbrain.org>
   8  */
   9
  10 use dokuwiki\Utf8\Asian;
  11 use dokuwiki\Search\Indexer;
  12 use dokuwiki\Extension\Event;
  13 use dokuwiki\Utf8\Clean;
  14 use dokuwiki\Utf8\PhpString;
  15 use dokuwiki\Utf8\Sort;
  16
  17 /**
  18  * create snippets for the first few results only
  19  */
  20 if (!defined('FT_SNIPPET_NUMBER')) define('FT_SNIPPET_NUMBER', 15);
  21
  22 /**
  23  * The fulltext search
  24  *
  25  * Returns a list of matching documents for the given query
  26  *
  27  * refactored into ft_pageSearch(), _ft_pageSearch() and trigger_event()
  28  *
  29  * @param string     $query
  30  * @param array      $highlight
  31  * @param string     $sort
  32  * @param int|string $after  only show results with mtime after this date, accepts timestap or strtotime arguments
  33  * @param int|string $before only show results with mtime before this date, accepts timestap or strtotime arguments
  34  *
  35  * @return array
  36  */
  37 function ft_pageSearch($query, &$highlight, $sort = null, $after = null, $before = null)
  38 {
  39
  40     if ($sort === null) {
  41         $sort = 'hits';
  42     }
  43     $data = [
  44         'query' => $query,
  45         'sort' => $sort,
  46         'after' => $after,
  47         'before' => $before
  48     ];
  49     $data['highlight'] =& $highlight;
  50
  51     return Event::createAndTrigger('SEARCH_QUERY_FULLPAGE', $data, '_ft_pageSearch');
  52 }
  53
  54 /**
  55  * Returns a list of matching documents for the given query
  56  *
  57  * @author Andreas Gohr <andi@splitbrain.org>
  58  * @author Kazutaka Miyasaka <kazmiya@gmail.com>
  59  *
  60  * @param array $data event data
  61  * @return array matching documents
  62  */
  63 function _ft_pageSearch(&$data)
  64 {
  65     $Indexer = idx_get_indexer();
  66
  67     // parse the given query
  68     $q = ft_queryParser($Indexer, $data['query']);
  69     $data['highlight'] = $q['highlight'];
  70
  71     if (empty($q['parsed_ary'])) return [];
  72
  73     // lookup all words found in the query
  74     $lookup = $Indexer->lookup($q['words']);
  75
  76     // get all pages in this dokuwiki site (!: includes nonexistent pages)
  77     $pages_all = [];
  78     foreach ($Indexer->getPages() as $id) {
  79         $pages_all[$id] = 0; // base: 0 hit
  80     }
  81
  82     // process the query
  83     $stack = [];
  84     foreach ($q['parsed_ary'] as $token) {
  85         switch (substr($token, 0, 3)) {
  86             case 'W+:':
  87             case 'W-:':
  88             case 'W_:': // word
  89                 $word    = substr($token, 3);
  90                 if (isset($lookup[$word])) {
  91                     $stack[] = (array)$lookup[$word];
  92                 }
  93                 break;
  94             case 'P+:':
  95             case 'P-:': // phrase
  96                 $phrase = substr($token, 3);
  97                 // since phrases are always parsed as ((W1)(W2)...(P)),
  98                 // the end($stack) always points the pages that contain
  99                 // all words in this phrase
 100                 $pages  = end($stack);
 101                 $pages_matched = [];
 102                 foreach (array_keys($pages) as $id) {
 103                     $evdata = [
 104                         'id' => $id,
 105                         'phrase' => $phrase,
 106                         'text' => rawWiki($id)
 107                     ];
 108                     $evt = new Event('FULLTEXT_PHRASE_MATCH', $evdata);
 109                     if ($evt->advise_before() && $evt->result !== true) {
 110                         $text = PhpString::strtolower($evdata['text']);
 111                         if (strpos($text, $phrase) !== false) {
 112                             $evt->result = true;
 113                         }
 114                     }
 115                     $evt->advise_after();
 116                     if ($evt->result === true) {
 117                         $pages_matched[$id] = 0; // phrase: always 0 hit
 118                     }
 119                 }
 120                 $stack[] = $pages_matched;
 121                 break;
 122             case 'N+:':
 123             case 'N-:': // namespace
 124                 $ns = cleanID(substr($token, 3)) . ':';
 125                 $pages_matched = [];
 126                 foreach (array_keys($pages_all) as $id) {
 127                     if (strpos($id, $ns) === 0) {
 128                         $pages_matched[$id] = 0; // namespace: always 0 hit
 129                     }
 130                 }
 131                 $stack[] = $pages_matched;
 132                 break;
 133             case 'AND': // and operation
 134                 $pages = array_splice($stack, -2);
 135                 if ($pages === []) {
 136                     break;
 137                 }
 138                 $stack[] = ft_resultCombine($pages);
 139                 break;
 140             case 'OR':  // or operation
 141                 $pages = array_splice($stack, -2);
 142                 if ($pages === []) {
 143                     break;
 144                 }
 145                 $stack[] = ft_resultUnite($pages);
 146                 break;
 147             case 'NOT': // not operation (unary)
 148                 $pages   = array_pop($stack);
 149                 $stack[] = ft_resultComplement([$pages_all, $pages]);
 150                 break;
 151         }
 152     }
 153     $docs = array_pop($stack);
 154
 155     if (empty($docs)) return [];
 156
 157     // check: settings, acls, existence
 158     foreach (array_keys($docs) as $id) {
 159         if (isHiddenPage($id) || auth_quickaclcheck($id) < AUTH_READ || !page_exists($id, '', false)) {
 160             unset($docs[$id]);
 161         }
 162     }
 163
 164     $docs = _ft_filterResultsByTime($docs, $data['after'], $data['before']);
 165
 166     if ($data['sort'] === 'mtime') {
 167         uksort($docs, 'ft_pagemtimesorter');
 168     } else {
 169         // sort docs by count
 170         uksort($docs, 'ft_pagesorter');
 171         arsort($docs);
 172     }
 173
 174     return $docs;
 175 }
 176
 177 /**
 178  * Returns the backlinks for a given page
 179  *
 180  * Uses the metadata index.
 181  *
 182  * @param string $id           The id for which links shall be returned
 183  * @param bool   $ignore_perms Ignore the fact that pages are hidden or read-protected
 184  * @return array The pages that contain links to the given page
 185  */
 186 function ft_backlinks($id, $ignore_perms = false)
 187 {
 188     $result = idx_get_indexer()->lookupKey('relation_references', $id);
 189
 190     if ($result === []) return $result;
 191
 192     // check ACL permissions
 193     foreach (array_keys($result) as $idx) {
 194         if (
 195             (!$ignore_perms && (
 196                 isHiddenPage($result[$idx]) || auth_quickaclcheck($result[$idx]) < AUTH_READ
 197             )) || !page_exists($result[$idx], '', false)
 198         ) {
 199             unset($result[$idx]);
 200         }
 201     }
 202
 203     Sort::sort($result);
 204     return $result;
 205 }
 206
 207 /**
 208  * Returns the pages that use a given media file
 209  *
 210  * Uses the relation media metadata property and the metadata index.
 211  *
 212  * Note that before 2013-07-31 the second parameter was the maximum number of results and
 213  * permissions were ignored. That's why the parameter is now checked to be explicitely set
 214  * to true (with type bool) in order to be compatible with older uses of the function.
 215  *
 216  * @param string $id           The media id to look for
 217  * @param bool   $ignore_perms Ignore hidden pages and acls (optional, default: false)
 218  * @return array A list of pages that use the given media file
 219  */
 220 function ft_mediause($id, $ignore_perms = false)
 221 {
 222     $result = idx_get_indexer()->lookupKey('relation_media', $id);
 223
 224     if ($result === []) return $result;
 225
 226     // check ACL permissions
 227     foreach (array_keys($result) as $idx) {
 228         if (
 229             (!$ignore_perms && (
 230                     isHiddenPage($result[$idx]) || auth_quickaclcheck($result[$idx]) < AUTH_READ
 231                 )) || !page_exists($result[$idx], '', false)
 232         ) {
 233             unset($result[$idx]);
 234         }
 235     }
 236
 237     Sort::sort($result);
 238     return $result;
 239 }
 240
 241
 242 /**
 243  * Quicksearch for pagenames
 244  *
 245  * By default it only matches the pagename and ignores the
 246  * namespace. This can be changed with the second parameter.
 247  * The third parameter allows to search in titles as well.
 248  *
 249  * The function always returns titles as well
 250  *
 251  * @triggers SEARCH_QUERY_PAGELOOKUP
 252  * @author   Andreas Gohr <andi@splitbrain.org>
 253  * @author   Adrian Lang <lang@cosmocode.de>
 254  *
 255  * @param string     $id       page id
 256  * @param bool       $in_ns    match against namespace as well?
 257  * @param bool       $in_title search in title?
 258  * @param int|string $after    only show results with mtime after this date, accepts timestap or strtotime arguments
 259  * @param int|string $before   only show results with mtime before this date, accepts timestap or strtotime arguments
 260  *
 261  * @return string[]
 262  */
 263 function ft_pageLookup($id, $in_ns = false, $in_title = false, $after = null, $before = null)
 264 {
 265     $data = [
 266         'id' => $id,
 267         'in_ns' => $in_ns,
 268         'in_title' => $in_title,
 269         'after' => $after,
 270         'before' => $before
 271     ];
 272     $data['has_titles'] = true; // for plugin backward compatibility check
 273     return Event::createAndTrigger('SEARCH_QUERY_PAGELOOKUP', $data, '_ft_pageLookup');
 274 }
 275
 276 /**
 277  * Returns list of pages as array(pageid => First Heading)
 278  *
 279  * @param array &$data event data
 280  * @return string[]
 281  */
 282 function _ft_pageLookup(&$data)
 283 {
 284     // split out original parameters
 285     $id = $data['id'];
 286     $Indexer = idx_get_indexer();
 287     $parsedQuery = ft_queryParser($Indexer, $id);
 288     if (count($parsedQuery['ns']) > 0) {
 289         $ns = cleanID($parsedQuery['ns'][0]) . ':';
 290         $id = implode(' ', $parsedQuery['highlight']);
 291     }
 292     if (count($parsedQuery['notns']) > 0) {
 293         $notns = cleanID($parsedQuery['notns'][0]) . ':';
 294         $id = implode(' ', $parsedQuery['highlight']);
 295     }
 296
 297     $in_ns    = $data['in_ns'];
 298     $in_title = $data['in_title'];
 299     $cleaned = cleanID($id);
 300
 301     $Indexer = idx_get_indexer();
 302     $page_idx = $Indexer->getPages();
 303
 304     $pages = [];
 305     if ($id !== '' && $cleaned !== '') {
 306         foreach ($page_idx as $p_id) {
 307             if ((strpos($in_ns ? $p_id : noNSorNS($p_id), $cleaned) !== false)) {
 308                 if (!isset($pages[$p_id]))
 309                     $pages[$p_id] = p_get_first_heading($p_id, METADATA_DONT_RENDER);
 310             }
 311         }
 312         if ($in_title) {
 313             foreach ($Indexer->lookupKey('title', $id, '_ft_pageLookupTitleCompare') as $p_id) {
 314                 if (!isset($pages[$p_id]))
 315                     $pages[$p_id] = p_get_first_heading($p_id, METADATA_DONT_RENDER);
 316             }
 317         }
 318     }
 319
 320     if (isset($ns)) {
 321         foreach (array_keys($pages) as $p_id) {
 322             if (strpos($p_id, $ns) !== 0) {
 323                 unset($pages[$p_id]);
 324             }
 325         }
 326     }
 327     if (isset($notns)) {
 328         foreach (array_keys($pages) as $p_id) {
 329             if (strpos($p_id, $notns) === 0) {
 330                 unset($pages[$p_id]);
 331             }
 332         }
 333     }
 334
 335     // discard hidden pages
 336     // discard nonexistent pages
 337     // check ACL permissions
 338     foreach (array_keys($pages) as $idx) {
 339         if (
 340             !isVisiblePage($idx) || !page_exists($idx) ||
 341             auth_quickaclcheck($idx) < AUTH_READ
 342         ) {
 343             unset($pages[$idx]);
 344         }
 345     }
 346
 347     $pages = _ft_filterResultsByTime($pages, $data['after'], $data['before']);
 348
 349     uksort($pages, 'ft_pagesorter');
 350     return $pages;
 351 }
 352
 353
 354 /**
 355  * @param array      $results search results in the form pageid => value
 356  * @param int|string $after   only returns results with mtime after this date, accepts timestap or strtotime arguments
 357  * @param int|string $before  only returns results with mtime after this date, accepts timestap or strtotime arguments
 358  *
 359  * @return array
 360  */
 361 function _ft_filterResultsByTime(array $results, $after, $before)
 362 {
 363     if ($after || $before) {
 364         $after = is_int($after) ? $after : strtotime($after);
 365         $before = is_int($before) ? $before : strtotime($before);
 366
 367         foreach (array_keys($results) as $id) {
 368             $mTime = filemtime(wikiFN($id));
 369             if ($after && $after > $mTime) {
 370                 unset($results[$id]);
 371                 continue;
 372             }
 373             if ($before && $before < $mTime) {
 374                 unset($results[$id]);
 375             }
 376         }
 377     }
 378
 379     return $results;
 380 }
 381
 382 /**
 383  * Tiny helper function for comparing the searched title with the title
 384  * from the search index. This function is a wrapper around stripos with
 385  * adapted argument order and return value.
 386  *
 387  * @param string $search searched title
 388  * @param string $title  title from index
 389  * @return bool
 390  */
 391 function _ft_pageLookupTitleCompare($search, $title)
 392 {
 393     if (Clean::isASCII($search)) {
 394         $pos = stripos($title, $search);
 395     } else {
 396         $pos = PhpString::strpos(
 397             PhpString::strtolower($title),
 398             PhpString::strtolower($search)
 399         );
 400     }
 401
 402     return $pos !== false;
 403 }
 404
 405 /**
 406  * Sort pages based on their namespace level first, then on their string
 407  * values. This makes higher hierarchy pages rank higher than lower hierarchy
 408  * pages.
 409  *
 410  * @param string $a
 411  * @param string $b
 412  * @return int Returns < 0 if $a is less than $b; > 0 if $a is greater than $b, and 0 if they are equal.
 413  */
 414 function ft_pagesorter($a, $b)
 415 {
 416     $ac = count(explode(':', $a));
 417     $bc = count(explode(':', $b));
 418     if ($ac < $bc) {
 419         return -1;
 420     } elseif ($ac > $bc) {
 421         return 1;
 422     }
 423     return Sort::strcmp($a, $b);
 424 }
 425
 426 /**
 427  * Sort pages by their mtime, from newest to oldest
 428  *
 429  * @param string $a
 430  * @param string $b
 431  *
 432  * @return int Returns < 0 if $a is newer than $b, > 0 if $b is newer than $a and 0 if they are of the same age
 433  */
 434 function ft_pagemtimesorter($a, $b)
 435 {
 436     $mtimeA = filemtime(wikiFN($a));
 437     $mtimeB = filemtime(wikiFN($b));
 438     return $mtimeB - $mtimeA;
 439 }
 440
 441 /**
 442  * Creates a snippet extract
 443  *
 444  * @author Andreas Gohr <andi@splitbrain.org>
 445  * @triggers FULLTEXT_SNIPPET_CREATE
 446  *
 447  * @param string $id page id
 448  * @param array $highlight
 449  * @return mixed
 450  */
 451 function ft_snippet($id, $highlight)
 452 {
 453     $text = rawWiki($id);
 454     $text = str_replace("\xC2\xAD", '', $text);
 455      // remove soft-hyphens
 456     $evdata = [
 457         'id'        => $id,
 458         'text'      => &$text,
 459         'highlight' => &$highlight,
 460         'snippet'   => ''
 461     ];
 462
 463     $evt = new Event('FULLTEXT_SNIPPET_CREATE', $evdata);
 464     if ($evt->advise_before()) {
 465         $match = [];
 466         $snippets = [];
 467         $utf8_offset = 0;
 468         $offset = 0;
 469         $end = 0;
 470         $len = PhpString::strlen($text);
 471
 472         // build a regexp from the phrases to highlight
 473         $re1 = '(' .
 474             implode(
 475                 '|',
 476                 array_map(
 477                     'ft_snippet_re_preprocess',
 478                     array_map(
 479                         'preg_quote_cb',
 480                         array_filter((array) $highlight)
 481                     )
 482                 )
 483             ) .
 484             ')';
 485         $re2 = "$re1.{0,75}(?!\\1)$re1";
 486         $re3 = "$re1.{0,45}(?!\\1)$re1.{0,45}(?!\\1)(?!\\2)$re1";
 487
 488         for ($cnt = 4; $cnt--;) {
 489             if (0) {
 490             } elseif (preg_match('/' . $re3 . '/iu', $text, $match, PREG_OFFSET_CAPTURE, $offset)) {
 491             } elseif (preg_match('/' . $re2 . '/iu', $text, $match, PREG_OFFSET_CAPTURE, $offset)) {
 492             } elseif (preg_match('/' . $re1 . '/iu', $text, $match, PREG_OFFSET_CAPTURE, $offset)) {
 493             } else {
 494                 break;
 495             }
 496
 497             [$str, $idx] = $match[0];
 498
 499             // convert $idx (a byte offset) into a utf8 character offset
 500             $utf8_idx = PhpString::strlen(substr($text, 0, $idx));
 501             $utf8_len = PhpString::strlen($str);
 502
 503             // establish context, 100 bytes surrounding the match string
 504             // first look to see if we can go 100 either side,
 505             // then drop to 50 adding any excess if the other side can't go to 50,
 506             $pre = min($utf8_idx - $utf8_offset, 100);
 507             $post = min($len - $utf8_idx - $utf8_len, 100);
 508
 509             if ($pre > 50 && $post > 50) {
 510                 $pre = 50;
 511                 $post = 50;
 512             } elseif ($pre > 50) {
 513                 $pre = min($pre, 100 - $post);
 514             } elseif ($post > 50) {
 515                 $post = min($post, 100 - $pre);
 516             } elseif ($offset == 0) {
 517                 // both are less than 50, means the context is the whole string
 518                 // make it so and break out of this loop - there is no need for the
 519                 // complex snippet calculations
 520                 $snippets = [$text];
 521                 break;
 522             }
 523
 524             // establish context start and end points, try to append to previous
 525             // context if possible
 526             $start = $utf8_idx - $pre;
 527             $append = ($start < $end) ? $end : false;  // still the end of the previous context snippet
 528             $end = $utf8_idx + $utf8_len + $post;      // now set it to the end of this context
 529
 530             if ($append) {
 531                 $snippets[count($snippets) - 1] .= PhpString::substr($text, $append, $end - $append);
 532             } else {
 533                 $snippets[] = PhpString::substr($text, $start, $end - $start);
 534             }
 535
 536             // set $offset for next match attempt
 537             // continue matching after the current match
 538             // if the current match is not the longest possible match starting at the current offset
 539             // this prevents further matching of this snippet but for possible matches of length
 540             // smaller than match length + context (at least 50 characters) this match is part of the context
 541             $utf8_offset = $utf8_idx + $utf8_len;
 542             $offset = $idx + strlen(PhpString::substr($text, $utf8_idx, $utf8_len));
 543             $offset = Clean::correctIdx($text, $offset);
 544         }
 545
 546         $m = "\1";
 547         $snippets = preg_replace('/' . $re1 . '/iu', $m . '$1' . $m, $snippets);
 548         $snippet = preg_replace(
 549             '/' . $m . '([^' . $m . ']*?)' . $m . '/iu',
 550             '<strong class="search_hit">$1</strong>',
 551             hsc(implode('... ', $snippets))
 552         );
 553
 554         $evdata['snippet'] = $snippet;
 555     }
 556     $evt->advise_after();
 557     unset($evt);
 558
 559     return $evdata['snippet'];
 560 }
 561
 562 /**
 563  * Wraps a search term in regex boundary checks.
 564  *
 565  * @param string $term
 566  * @return string
 567  */
 568 function ft_snippet_re_preprocess($term)
 569 {
 570     // do not process asian terms where word boundaries are not explicit
 571     if (Asian::isAsianWords($term)) return $term;
 572
 573     if (UTF8_PROPERTYSUPPORT) {
 574         // unicode word boundaries
 575         // see http://stackoverflow.com/a/2449017/172068
 576         $BL = '(?<!\pL)';
 577         $BR = '(?!\pL)';
 578     } else {
 579         // not as correct as above, but at least won't break
 580         $BL = '\b';
 581         $BR = '\b';
 582     }
 583
 584     if (str_starts_with($term, '\\*')) {
 585         $term = substr($term, 2);
 586     } else {
 587         $term = $BL . $term;
 588     }
 589
 590     if (str_ends_with($term, '\\*')) {
 591         $term = substr($term, 0, -2);
 592     } else {
 593         $term .= $BR;
 594     }
 595
 596     if ($term == $BL || $term == $BR || $term == $BL . $BR) $term = '';
 597     return $term;
 598 }
 599
 600 /**
 601  * Combine found documents and sum up their scores
 602  *
 603  * This function is used to combine searched words with a logical
 604  * AND. Only documents available in all arrays are returned.
 605  *
 606  * based upon PEAR's PHP_Compat function for array_intersect_key()
 607  *
 608  * @param array $args An array of page arrays
 609  * @return array
 610  */
 611 function ft_resultCombine($args)
 612 {
 613     $array_count = count($args);
 614     if ($array_count == 1) {
 615         return $args[0];
 616     }
 617
 618     $result = [];
 619     if ($array_count > 1) {
 620         foreach ($args[0] as $key => $value) {
 621             $result[$key] = $value;
 622             for ($i = 1; $i !== $array_count; $i++) {
 623                 if (!isset($args[$i][$key])) {
 624                     unset($result[$key]);
 625                     break;
 626                 }
 627                 $result[$key] += $args[$i][$key];
 628             }
 629         }
 630     }
 631     return $result;
 632 }
 633
 634 /**
 635  * Unites found documents and sum up their scores
 636  *
 637  * based upon ft_resultCombine() function
 638  *
 639  * @param array $args An array of page arrays
 640  * @return array
 641  *
 642  * @author Kazutaka Miyasaka <kazmiya@gmail.com>
 643  */
 644 function ft_resultUnite($args)
 645 {
 646     $array_count = count($args);
 647     if ($array_count === 1) {
 648         return $args[0];
 649     }
 650
 651     $result = $args[0];
 652     for ($i = 1; $i !== $array_count; $i++) {
 653         foreach (array_keys($args[$i]) as $id) {
 654             $result[$id] += $args[$i][$id];
 655         }
 656     }
 657     return $result;
 658 }
 659
 660 /**
 661  * Computes the difference of documents using page id for comparison
 662  *
 663  * nearly identical to PHP5's array_diff_key()
 664  *
 665  * @param array $args An array of page arrays
 666  * @return array
 667  *
 668  * @author Kazutaka Miyasaka <kazmiya@gmail.com>
 669  */
 670 function ft_resultComplement($args)
 671 {
 672     $array_count = count($args);
 673     if ($array_count === 1) {
 674         return $args[0];
 675     }
 676
 677     $result = $args[0];
 678     foreach (array_keys($result) as $id) {
 679         for ($i = 1; $i !== $array_count; $i++) {
 680             if (isset($args[$i][$id])) unset($result[$id]);
 681         }
 682     }
 683     return $result;
 684 }
 685
 686 /**
 687  * Parses a search query and builds an array of search formulas
 688  *
 689  * @author Andreas Gohr <andi@splitbrain.org>
 690  * @author Kazutaka Miyasaka <kazmiya@gmail.com>
 691  *
 692  * @param Indexer $Indexer
 693  * @param string                  $query search query
 694  * @return array of search formulas
 695  */
 696 function ft_queryParser($Indexer, $query)
 697 {
 698     /**
 699      * parse a search query and transform it into intermediate representation
 700      *
 701      * in a search query, you can use the following expressions:
 702      *
 703      *   words:
 704      *     include
 705      *     -exclude
 706      *   phrases:
 707      *     "phrase to be included"
 708      *     -"phrase you want to exclude"
 709      *   namespaces:
 710      *     @include:namespace (or ns:include:namespace)
 711      *     ^exclude:namespace (or -ns:exclude:namespace)
 712      *   groups:
 713      *     ()
 714      *     -()
 715      *   operators:
 716      *     and ('and' is the default operator: you can always omit this)
 717      *     or  (or pipe symbol '|', lower precedence than 'and')
 718      *
 719      * e.g. a query [ aa "bb cc" @dd:ee ] means "search pages which contain
 720      *      a word 'aa', a phrase 'bb cc' and are within a namespace 'dd:ee'".
 721      *      this query is equivalent to [ -(-aa or -"bb cc" or -ns:dd:ee) ]
 722      *      as long as you don't mind hit counts.
 723      *
 724      * intermediate representation consists of the following parts:
 725      *
 726      *   ( )           - group
 727      *   AND           - logical and
 728      *   OR            - logical or
 729      *   NOT           - logical not
 730      *   W+:, W-:, W_: - word      (underscore: no need to highlight)
 731      *   P+:, P-:      - phrase    (minus sign: logically in NOT group)
 732      *   N+:, N-:      - namespace
 733      */
 734     $parsed_query = '';
 735     $parens_level = 0;
 736     $terms = preg_split(
 737         '/(-?".*?")/u',
 738         PhpString::strtolower($query),
 739         -1,
 740         PREG_SPLIT_DELIM_CAPTURE | PREG_SPLIT_NO_EMPTY
 741     );
 742
 743     foreach ($terms as $term) {
 744         $parsed = '';
 745         if (preg_match('/^(-?)"(.+)"$/u', $term, $matches)) {
 746             // phrase-include and phrase-exclude
 747             $not = $matches[1] ? 'NOT' : '';
 748             $parsed = $not . ft_termParser($Indexer, $matches[2], false, true);
 749         } else {
 750             // fix incomplete phrase
 751             $term = str_replace('"', ' ', $term);
 752
 753             // fix parentheses
 754             $term = str_replace(')', ' ) ', $term);
 755             $term = str_replace('(', ' ( ', $term);
 756             $term = str_replace('- (', ' -(', $term);
 757
 758             // treat pipe symbols as 'OR' operators
 759             $term = str_replace('|', ' or ', $term);
 760
 761             // treat ideographic spaces (U+3000) as search term separators
 762             // FIXME: some more separators?
 763             $term = preg_replace('/[ \x{3000}]+/u', ' ', $term);
 764             $term = trim($term);
 765             if ($term === '') continue;
 766
 767             $tokens = explode(' ', $term);
 768             foreach ($tokens as $token) {
 769                 if ($token === '(') {
 770                     // parenthesis-include-open
 771                     $parsed .= '(';
 772                     ++$parens_level;
 773                 } elseif ($token === '-(') {
 774                     // parenthesis-exclude-open
 775                     $parsed .= 'NOT(';
 776                     ++$parens_level;
 777                 } elseif ($token === ')') {
 778                     // parenthesis-any-close
 779                     if ($parens_level === 0) continue;
 780                     $parsed .= ')';
 781                     $parens_level--;
 782                 } elseif ($token === 'and') {
 783                     // logical-and (do nothing)
 784                 } elseif ($token === 'or') {
 785                     // logical-or
 786                     $parsed .= 'OR';
 787                 } elseif (preg_match('/^(?:\^|-ns:)(.+)$/u', $token, $matches)) {
 788                     // namespace-exclude
 789                     $parsed .= 'NOT(N+:' . $matches[1] . ')';
 790                 } elseif (preg_match('/^(?:@|ns:)(.+)$/u', $token, $matches)) {
 791                     // namespace-include
 792                     $parsed .= '(N+:' . $matches[1] . ')';
 793                 } elseif (preg_match('/^-(.+)$/', $token, $matches)) {
 794                     // word-exclude
 795                     $parsed .= 'NOT(' . ft_termParser($Indexer, $matches[1]) . ')';
 796                 } else {
 797                     // word-include
 798                     $parsed .= ft_termParser($Indexer, $token);
 799                 }
 800             }
 801         }
 802         $parsed_query .= $parsed;
 803     }
 804
 805     // cleanup (very sensitive)
 806     $parsed_query .= str_repeat(')', $parens_level);
 807     do {
 808         $parsed_query_old = $parsed_query;
 809         $parsed_query = preg_replace('/(NOT)?\(\)/u', '', $parsed_query);
 810     } while ($parsed_query !== $parsed_query_old);
 811     $parsed_query = preg_replace('/(NOT|OR)+\)/u', ')', $parsed_query);
 812     $parsed_query = preg_replace('/(OR)+/u', 'OR', $parsed_query);
 813     $parsed_query = preg_replace('/\(OR/u', '(', $parsed_query);
 814     $parsed_query = preg_replace('/^OR|OR$/u', '', $parsed_query);
 815     $parsed_query = preg_replace('/\)(NOT)?\(/u', ')AND$1(', $parsed_query);
 816
 817     // adjustment: make highlightings right
 818     $parens_level     = 0;
 819     $notgrp_levels    = [];
 820     $parsed_query_new = '';
 821     $tokens = preg_split('/(NOT\(|[()])/u', $parsed_query, -1, PREG_SPLIT_DELIM_CAPTURE | PREG_SPLIT_NO_EMPTY);
 822     foreach ($tokens as $token) {
 823         if ($token === 'NOT(') {
 824             $notgrp_levels[] = ++$parens_level;
 825         } elseif ($token === '(') {
 826             ++$parens_level;
 827         } elseif ($token === ')') {
 828             if ($parens_level-- === end($notgrp_levels)) array_pop($notgrp_levels);
 829         } elseif (count($notgrp_levels) % 2 === 1) {
 830             // turn highlight-flag off if terms are logically in "NOT" group
 831             $token = preg_replace('/([WPN])\+\:/u', '$1-:', $token);
 832         }
 833         $parsed_query_new .= $token;
 834     }
 835     $parsed_query = $parsed_query_new;
 836
 837     /**
 838      * convert infix notation string into postfix (Reverse Polish notation) array
 839      * by Shunting-yard algorithm
 840      *
 841      * see: http://en.wikipedia.org/wiki/Reverse_Polish_notation
 842      * see: http://en.wikipedia.org/wiki/Shunting-yard_algorithm
 843      */
 844     $parsed_ary     = [];
 845     $ope_stack      = [];
 846     $ope_precedence = [')' => 1, 'OR' => 2, 'AND' => 3, 'NOT' => 4, '(' => 5];
 847     $ope_regex      = '/([()]|OR|AND|NOT)/u';
 848
 849     $tokens = preg_split($ope_regex, $parsed_query, -1, PREG_SPLIT_DELIM_CAPTURE | PREG_SPLIT_NO_EMPTY);
 850     foreach ($tokens as $token) {
 851         if (preg_match($ope_regex, $token)) {
 852             // operator
 853             $last_ope = end($ope_stack);
 854             while ($last_ope !== false && $ope_precedence[$token] <= $ope_precedence[$last_ope] && $last_ope != '(') {
 855                 $parsed_ary[] = array_pop($ope_stack);
 856                 $last_ope = end($ope_stack);
 857             }
 858             if ($token == ')') {
 859                 array_pop($ope_stack); // this array_pop always deletes '('
 860             } else {
 861                 $ope_stack[] = $token;
 862             }
 863         } else {
 864             // operand
 865             $token_decoded = str_replace(['OP', 'CP'], ['(', ')'], $token);
 866             $parsed_ary[] = $token_decoded;
 867         }
 868     }
 869     $parsed_ary = array_values([...$parsed_ary, ...array_reverse($ope_stack)]);
 870
 871     // cleanup: each double "NOT" in RPN array actually does nothing
 872     $parsed_ary_count = count($parsed_ary);
 873     for ($i = 1; $i < $parsed_ary_count; ++$i) {
 874         if ($parsed_ary[$i] === 'NOT' && $parsed_ary[$i - 1] === 'NOT') {
 875             unset($parsed_ary[$i], $parsed_ary[$i - 1]);
 876         }
 877     }
 878     $parsed_ary = array_values($parsed_ary);
 879
 880     // build return value
 881     $q = [];
 882     $q['query']      = $query;
 883     $q['parsed_str'] = $parsed_query;
 884     $q['parsed_ary'] = $parsed_ary;
 885
 886     foreach ($q['parsed_ary'] as $token) {
 887         if (strlen($token) < 3 || $token[2] !== ':') continue;
 888         $body = substr($token, 3);
 889
 890         switch (substr($token, 0, 3)) {
 891             case 'N+:':
 892                      $q['ns'][]        = $body; // for backward compatibility
 893                 break;
 894             case 'N-:':
 895                      $q['notns'][]     = $body; // for backward compatibility
 896                 break;
 897             case 'W_:':
 898                      $q['words'][]     = $body;
 899                 break;
 900             case 'W-:':
 901                      $q['words'][]     = $body;
 902                      $q['not'][]       = $body; // for backward compatibility
 903                 break;
 904             case 'W+:':
 905                      $q['words'][]     = $body;
 906                      $q['highlight'][] = $body;
 907                      $q['and'][]       = $body; // for backward compatibility
 908                 break;
 909             case 'P-:':
 910                      $q['phrases'][]   = $body;
 911                 break;
 912             case 'P+:':
 913                      $q['phrases'][]   = $body;
 914                      $q['highlight'][] = $body;
 915                 break;
 916         }
 917     }
 918     foreach (['words', 'phrases', 'highlight', 'ns', 'notns', 'and', 'not'] as $key) {
 919         $q[$key] = empty($q[$key]) ? [] : array_values(array_unique($q[$key]));
 920     }
 921
 922     return $q;
 923 }
 924
 925 /**
 926  * Transforms given search term into intermediate representation
 927  *
 928  * This function is used in ft_queryParser() and not for general purpose use.
 929  *
 930  * @author Kazutaka Miyasaka <kazmiya@gmail.com>
 931  *
 932  * @param Indexer $Indexer
 933  * @param string                  $term
 934  * @param bool                    $consider_asian
 935  * @param bool                    $phrase_mode
 936  * @return string
 937  */
 938 function ft_termParser($Indexer, $term, $consider_asian = true, $phrase_mode = false)
 939 {
 940     $parsed = '';
 941     if ($consider_asian) {
 942         // successive asian characters need to be searched as a phrase
 943         $words = Asian::splitAsianWords($term);
 944         foreach ($words as $word) {
 945             $phrase_mode = $phrase_mode ? true : Asian::isAsianWords($word);
 946             $parsed .= ft_termParser($Indexer, $word, false, $phrase_mode);
 947         }
 948     } else {
 949         $term_noparen = str_replace(['(', ')'], ' ', $term);
 950         $words = $Indexer->tokenizer($term_noparen, true);
 951
 952         // W_: no need to highlight
 953         if (empty($words)) {
 954             $parsed = '()'; // important: do not remove
 955         } elseif ($words[0] === $term) {
 956             $parsed = '(W+:' . $words[0] . ')';
 957         } elseif ($phrase_mode) {
 958             $term_encoded = str_replace(['(', ')'], ['OP', 'CP'], $term);
 959             $parsed = '((W_:' . implode(')(W_:', $words) . ')(P+:' . $term_encoded . '))';
 960         } else {
 961             $parsed = '((W+:' . implode(')(W+:', $words) . '))';
 962         }
 963     }
 964     return $parsed;
 965 }
 966
 967 /**
 968  * Recreate a search query string based on parsed parts, doesn't support negated phrases and `OR` searches
 969  *
 970  * @param array $and
 971  * @param array $not
 972  * @param array $phrases
 973  * @param array $ns
 974  * @param array $notns
 975  *
 976  * @return string
 977  */
 978 function ft_queryUnparser_simple(array $and, array $not, array $phrases, array $ns, array $notns)
 979 {
 980     $query = implode(' ', $and);
 981     if ($not !== []) {
 982         $query .= ' -' . implode(' -', $not);
 983     }
 984
 985     if ($phrases !== []) {
 986         $query .= ' "' . implode('" "', $phrases) . '"';
 987     }
 988
 989     if ($ns !== []) {
 990         $query .= ' @' . implode(' @', $ns);
 991     }
 992
 993     if ($notns !== []) {
 994         $query .= ' ^' . implode(' ^', $notns);
 995     }
 996
 997     return $query;
 998 }
 999
1000 //Setup VIM: ex: et ts=4 :