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[] = $pages_matched;
 132                 break;
 133             case 'AND': // and operation
 134                 [$pages1, $pages2] = array_splice($stack, -2);
 135                 $stack[] = ft_resultCombine([$pages1, $pages2]);
 136                 break;
 137             case 'OR':  // or operation
 138                 [$pages1, $pages2] = array_splice($stack, -2);
 139                 $stack[] = ft_resultUnite([$pages1, $pages2]);
 140                 break;
 141             case 'NOT': // not operation (unary)
 142                 $pages   = array_pop($stack);
 143                 $stack[] = ft_resultComplement([$pages_all, $pages]);
 144                 break;
 145         }
 146     }
 147     $docs = array_pop($stack);
 148
 149     if (empty($docs)) return [];
 150
 151     // check: settings, acls, existence
 152     foreach (array_keys($docs) as $id) {
 153         if (isHiddenPage($id) || auth_quickaclcheck($id) < AUTH_READ || !page_exists($id, '', false)) {
 154             unset($docs[$id]);
 155         }
 156     }
 157
 158     $docs = _ft_filterResultsByTime($docs, $data['after'], $data['before']);
 159
 160     if ($data['sort'] === 'mtime') {
 161         uksort($docs, 'ft_pagemtimesorter');
 162     } else {
 163         // sort docs by count
 164         uksort($docs, 'ft_pagesorter');
 165         arsort($docs);
 166     }
 167
 168     return $docs;
 169 }
 170
 171 /**
 172  * Returns the backlinks for a given page
 173  *
 174  * Uses the metadata index.
 175  *
 176  * @param string $id           The id for which links shall be returned
 177  * @param bool   $ignore_perms Ignore the fact that pages are hidden or read-protected
 178  * @return array The pages that contain links to the given page
 179  */
 180 function ft_backlinks($id, $ignore_perms = false)
 181 {
 182     $result = idx_get_indexer()->lookupKey('relation_references', $id);
 183
 184     if ($result === []) return $result;
 185
 186     // check ACL permissions
 187     foreach (array_keys($result) as $idx) {
 188         if (
 189             (!$ignore_perms && (
 190                 isHiddenPage($result[$idx]) || auth_quickaclcheck($result[$idx]) < AUTH_READ
 191             )) || !page_exists($result[$idx], '', false)
 192         ) {
 193             unset($result[$idx]);
 194         }
 195     }
 196
 197     Sort::sort($result);
 198     return $result;
 199 }
 200
 201 /**
 202  * Returns the pages that use a given media file
 203  *
 204  * Uses the relation media metadata property and the metadata index.
 205  *
 206  * Note that before 2013-07-31 the second parameter was the maximum number of results and
 207  * permissions were ignored. That's why the parameter is now checked to be explicitely set
 208  * to true (with type bool) in order to be compatible with older uses of the function.
 209  *
 210  * @param string $id           The media id to look for
 211  * @param bool   $ignore_perms Ignore hidden pages and acls (optional, default: false)
 212  * @return array A list of pages that use the given media file
 213  */
 214 function ft_mediause($id, $ignore_perms = false)
 215 {
 216     $result = idx_get_indexer()->lookupKey('relation_media', $id);
 217
 218     if ($result === []) return $result;
 219
 220     // check ACL permissions
 221     foreach (array_keys($result) as $idx) {
 222         if (
 223             (!$ignore_perms && (
 224                     isHiddenPage($result[$idx]) || auth_quickaclcheck($result[$idx]) < AUTH_READ
 225                 )) || !page_exists($result[$idx], '', false)
 226         ) {
 227             unset($result[$idx]);
 228         }
 229     }
 230
 231     Sort::sort($result);
 232     return $result;
 233 }
 234
 235
 236 /**
 237  * Quicksearch for pagenames
 238  *
 239  * By default it only matches the pagename and ignores the
 240  * namespace. This can be changed with the second parameter.
 241  * The third parameter allows to search in titles as well.
 242  *
 243  * The function always returns titles as well
 244  *
 245  * @triggers SEARCH_QUERY_PAGELOOKUP
 246  * @author   Andreas Gohr <andi@splitbrain.org>
 247  * @author   Adrian Lang <lang@cosmocode.de>
 248  *
 249  * @param string     $id       page id
 250  * @param bool       $in_ns    match against namespace as well?
 251  * @param bool       $in_title search in title?
 252  * @param int|string $after    only show results with mtime after this date, accepts timestap or strtotime arguments
 253  * @param int|string $before   only show results with mtime before this date, accepts timestap or strtotime arguments
 254  *
 255  * @return string[]
 256  */
 257 function ft_pageLookup($id, $in_ns = false, $in_title = false, $after = null, $before = null)
 258 {
 259     $data = [
 260         'id' => $id,
 261         'in_ns' => $in_ns,
 262         'in_title' => $in_title,
 263         'after' => $after,
 264         'before' => $before
 265     ];
 266     $data['has_titles'] = true; // for plugin backward compatibility check
 267     return Event::createAndTrigger('SEARCH_QUERY_PAGELOOKUP', $data, '_ft_pageLookup');
 268 }
 269
 270 /**
 271  * Returns list of pages as array(pageid => First Heading)
 272  *
 273  * @param array &$data event data
 274  * @return string[]
 275  */
 276 function _ft_pageLookup(&$data)
 277 {
 278     // split out original parameters
 279     $id = $data['id'];
 280     $Indexer = idx_get_indexer();
 281     $parsedQuery = ft_queryParser($Indexer, $id);
 282     if (count($parsedQuery['ns']) > 0) {
 283         $ns = cleanID($parsedQuery['ns'][0]) . ':';
 284         $id = implode(' ', $parsedQuery['highlight']);
 285     }
 286     if (count($parsedQuery['notns']) > 0) {
 287         $notns = cleanID($parsedQuery['notns'][0]) . ':';
 288         $id = implode(' ', $parsedQuery['highlight']);
 289     }
 290
 291     $in_ns    = $data['in_ns'];
 292     $in_title = $data['in_title'];
 293     $cleaned = cleanID($id);
 294
 295     $Indexer = idx_get_indexer();
 296     $page_idx = $Indexer->getPages();
 297
 298     $pages = [];
 299     if ($id !== '' && $cleaned !== '') {
 300         foreach ($page_idx as $p_id) {
 301             if ((strpos($in_ns ? $p_id : noNSorNS($p_id), $cleaned) !== false)) {
 302                 if (!isset($pages[$p_id]))
 303                     $pages[$p_id] = p_get_first_heading($p_id, METADATA_DONT_RENDER);
 304             }
 305         }
 306         if ($in_title) {
 307             foreach ($Indexer->lookupKey('title', $id, '_ft_pageLookupTitleCompare') as $p_id) {
 308                 if (!isset($pages[$p_id]))
 309                     $pages[$p_id] = p_get_first_heading($p_id, METADATA_DONT_RENDER);
 310             }
 311         }
 312     }
 313
 314     if (isset($ns)) {
 315         foreach (array_keys($pages) as $p_id) {
 316             if (strpos($p_id, $ns) !== 0) {
 317                 unset($pages[$p_id]);
 318             }
 319         }
 320     }
 321     if (isset($notns)) {
 322         foreach (array_keys($pages) as $p_id) {
 323             if (strpos($p_id, $notns) === 0) {
 324                 unset($pages[$p_id]);
 325             }
 326         }
 327     }
 328
 329     // discard hidden pages
 330     // discard nonexistent pages
 331     // check ACL permissions
 332     foreach (array_keys($pages) as $idx) {
 333         if (
 334             !isVisiblePage($idx) || !page_exists($idx) ||
 335             auth_quickaclcheck($idx) < AUTH_READ
 336         ) {
 337             unset($pages[$idx]);
 338         }
 339     }
 340
 341     $pages = _ft_filterResultsByTime($pages, $data['after'], $data['before']);
 342
 343     uksort($pages, 'ft_pagesorter');
 344     return $pages;
 345 }
 346
 347
 348 /**
 349  * @param array      $results search results in the form pageid => value
 350  * @param int|string $after   only returns results with mtime after this date, accepts timestap or strtotime arguments
 351  * @param int|string $before  only returns results with mtime after this date, accepts timestap or strtotime arguments
 352  *
 353  * @return array
 354  */
 355 function _ft_filterResultsByTime(array $results, $after, $before)
 356 {
 357     if ($after || $before) {
 358         $after = is_int($after) ? $after : strtotime($after);
 359         $before = is_int($before) ? $before : strtotime($before);
 360
 361         foreach (array_keys($results) as $id) {
 362             $mTime = filemtime(wikiFN($id));
 363             if ($after && $after > $mTime) {
 364                 unset($results[$id]);
 365                 continue;
 366             }
 367             if ($before && $before < $mTime) {
 368                 unset($results[$id]);
 369             }
 370         }
 371     }
 372
 373     return $results;
 374 }
 375
 376 /**
 377  * Tiny helper function for comparing the searched title with the title
 378  * from the search index. This function is a wrapper around stripos with
 379  * adapted argument order and return value.
 380  *
 381  * @param string $search searched title
 382  * @param string $title  title from index
 383  * @return bool
 384  */
 385 function _ft_pageLookupTitleCompare($search, $title)
 386 {
 387     if (Clean::isASCII($search)) {
 388         $pos = stripos($title, $search);
 389     } else {
 390         $pos = PhpString::strpos(
 391             PhpString::strtolower($title),
 392             PhpString::strtolower($search)
 393         );
 394     }
 395
 396     return $pos !== false;
 397 }
 398
 399 /**
 400  * Sort pages based on their namespace level first, then on their string
 401  * values. This makes higher hierarchy pages rank higher than lower hierarchy
 402  * pages.
 403  *
 404  * @param string $a
 405  * @param string $b
 406  * @return int Returns < 0 if $a is less than $b; > 0 if $a is greater than $b, and 0 if they are equal.
 407  */
 408 function ft_pagesorter($a, $b)
 409 {
 410     $ac = count(explode(':', $a));
 411     $bc = count(explode(':', $b));
 412     if ($ac < $bc) {
 413         return -1;
 414     } elseif ($ac > $bc) {
 415         return 1;
 416     }
 417     return Sort::strcmp($a, $b);
 418 }
 419
 420 /**
 421  * Sort pages by their mtime, from newest to oldest
 422  *
 423  * @param string $a
 424  * @param string $b
 425  *
 426  * @return int Returns < 0 if $a is newer than $b, > 0 if $b is newer than $a and 0 if they are of the same age
 427  */
 428 function ft_pagemtimesorter($a, $b)
 429 {
 430     $mtimeA = filemtime(wikiFN($a));
 431     $mtimeB = filemtime(wikiFN($b));
 432     return $mtimeB - $mtimeA;
 433 }
 434
 435 /**
 436  * Creates a snippet extract
 437  *
 438  * @author Andreas Gohr <andi@splitbrain.org>
 439  * @triggers FULLTEXT_SNIPPET_CREATE
 440  *
 441  * @param string $id page id
 442  * @param array $highlight
 443  * @return mixed
 444  */
 445 function ft_snippet($id, $highlight)
 446 {
 447     $text = rawWiki($id);
 448     $text = str_replace("\xC2\xAD", '', $text);
 449      // remove soft-hyphens
 450     $evdata = [
 451         'id'        => $id,
 452         'text'      => &$text,
 453         'highlight' => &$highlight,
 454         'snippet'   => ''
 455     ];
 456
 457     $evt = new Event('FULLTEXT_SNIPPET_CREATE', $evdata);
 458     if ($evt->advise_before()) {
 459         $match = [];
 460         $snippets = [];
 461         $utf8_offset = 0;
 462         $offset = 0;
 463         $end = 0;
 464         $len = PhpString::strlen($text);
 465
 466         // build a regexp from the phrases to highlight
 467         $re1 = '(' .
 468             implode(
 469                 '|',
 470                 array_map(
 471                     'ft_snippet_re_preprocess',
 472                     array_map(
 473                         'preg_quote_cb',
 474                         array_filter((array) $highlight)
 475                     )
 476                 )
 477             ) .
 478             ')';
 479         $re2 = "$re1.{0,75}(?!\\1)$re1";
 480         $re3 = "$re1.{0,45}(?!\\1)$re1.{0,45}(?!\\1)(?!\\2)$re1";
 481
 482         for ($cnt = 4; $cnt--;) {
 483             if (0) {
 484             } elseif (preg_match('/' . $re3 . '/iu', $text, $match, PREG_OFFSET_CAPTURE, $offset)) {
 485             } elseif (preg_match('/' . $re2 . '/iu', $text, $match, PREG_OFFSET_CAPTURE, $offset)) {
 486             } elseif (preg_match('/' . $re1 . '/iu', $text, $match, PREG_OFFSET_CAPTURE, $offset)) {
 487             } else {
 488                 break;
 489             }
 490
 491             [$str, $idx] = $match[0];
 492
 493             // convert $idx (a byte offset) into a utf8 character offset
 494             $utf8_idx = PhpString::strlen(substr($text, 0, $idx));
 495             $utf8_len = PhpString::strlen($str);
 496
 497             // establish context, 100 bytes surrounding the match string
 498             // first look to see if we can go 100 either side,
 499             // then drop to 50 adding any excess if the other side can't go to 50,
 500             $pre = min($utf8_idx - $utf8_offset, 100);
 501             $post = min($len - $utf8_idx - $utf8_len, 100);
 502
 503             if ($pre > 50 && $post > 50) {
 504                 $pre = 50;
 505                 $post = 50;
 506             } elseif ($pre > 50) {
 507                 $pre = min($pre, 100 - $post);
 508             } elseif ($post > 50) {
 509                 $post = min($post, 100 - $pre);
 510             } elseif ($offset == 0) {
 511                 // both are less than 50, means the context is the whole string
 512                 // make it so and break out of this loop - there is no need for the
 513                 // complex snippet calculations
 514                 $snippets = [$text];
 515                 break;
 516             }
 517
 518             // establish context start and end points, try to append to previous
 519             // context if possible
 520             $start = $utf8_idx - $pre;
 521             $append = ($start < $end) ? $end : false;  // still the end of the previous context snippet
 522             $end = $utf8_idx + $utf8_len + $post;      // now set it to the end of this context
 523
 524             if ($append) {
 525                 $snippets[count($snippets) - 1] .= PhpString::substr($text, $append, $end - $append);
 526             } else {
 527                 $snippets[] = PhpString::substr($text, $start, $end - $start);
 528             }
 529
 530             // set $offset for next match attempt
 531             // continue matching after the current match
 532             // if the current match is not the longest possible match starting at the current offset
 533             // this prevents further matching of this snippet but for possible matches of length
 534             // smaller than match length + context (at least 50 characters) this match is part of the context
 535             $utf8_offset = $utf8_idx + $utf8_len;
 536             $offset = $idx + strlen(PhpString::substr($text, $utf8_idx, $utf8_len));
 537             $offset = Clean::correctIdx($text, $offset);
 538         }
 539
 540         $m = "\1";
 541         $snippets = preg_replace('/' . $re1 . '/iu', $m . '$1' . $m, $snippets);
 542         $snippet = preg_replace(
 543             '/' . $m . '([^' . $m . ']*?)' . $m . '/iu',
 544             '<strong class="search_hit">$1</strong>',
 545             hsc(implode('... ', $snippets))
 546         );
 547
 548         $evdata['snippet'] = $snippet;
 549     }
 550     $evt->advise_after();
 551     unset($evt);
 552
 553     return $evdata['snippet'];
 554 }
 555
 556 /**
 557  * Wraps a search term in regex boundary checks.
 558  *
 559  * @param string $term
 560  * @return string
 561  */
 562 function ft_snippet_re_preprocess($term)
 563 {
 564     // do not process asian terms where word boundaries are not explicit
 565     if (Asian::isAsianWords($term)) return $term;
 566
 567     if (UTF8_PROPERTYSUPPORT) {
 568         // unicode word boundaries
 569         // see http://stackoverflow.com/a/2449017/172068
 570         $BL = '(?<!\pL)';
 571         $BR = '(?!\pL)';
 572     } else {
 573         // not as correct as above, but at least won't break
 574         $BL = '\b';
 575         $BR = '\b';
 576     }
 577
 578     if (str_starts_with($term, '\\*')) {
 579         $term = substr($term, 2);
 580     } else {
 581         $term = $BL . $term;
 582     }
 583
 584     if (str_ends_with($term, '\\*')) {
 585         $term = substr($term, 0, -2);
 586     } else {
 587         $term .= $BR;
 588     }
 589
 590     if ($term == $BL || $term == $BR || $term == $BL . $BR) $term = '';
 591     return $term;
 592 }
 593
 594 /**
 595  * Combine found documents and sum up their scores
 596  *
 597  * This function is used to combine searched words with a logical
 598  * AND. Only documents available in all arrays are returned.
 599  *
 600  * based upon PEAR's PHP_Compat function for array_intersect_key()
 601  *
 602  * @param array $args An array of page arrays
 603  * @return array
 604  */
 605 function ft_resultCombine($args)
 606 {
 607     $array_count = count($args);
 608     if ($array_count == 1) {
 609         return $args[0];
 610     }
 611
 612     $result = [];
 613     if ($array_count > 1) {
 614         foreach ($args[0] as $key => $value) {
 615             $result[$key] = $value;
 616             for ($i = 1; $i !== $array_count; $i++) {
 617                 if (!isset($args[$i][$key])) {
 618                     unset($result[$key]);
 619                     break;
 620                 }
 621                 $result[$key] += $args[$i][$key];
 622             }
 623         }
 624     }
 625     return $result;
 626 }
 627
 628 /**
 629  * Unites found documents and sum up their scores
 630  *
 631  * based upon ft_resultCombine() function
 632  *
 633  * @param array $args An array of page arrays
 634  * @return array
 635  *
 636  * @author Kazutaka Miyasaka <kazmiya@gmail.com>
 637  */
 638 function ft_resultUnite($args)
 639 {
 640     $array_count = count($args);
 641     if ($array_count === 1) {
 642         return $args[0];
 643     }
 644
 645     $result = $args[0];
 646     for ($i = 1; $i !== $array_count; $i++) {
 647         foreach (array_keys($args[$i]) as $id) {
 648             $result[$id] += $args[$i][$id];
 649         }
 650     }
 651     return $result;
 652 }
 653
 654 /**
 655  * Computes the difference of documents using page id for comparison
 656  *
 657  * nearly identical to PHP5's array_diff_key()
 658  *
 659  * @param array $args An array of page arrays
 660  * @return array
 661  *
 662  * @author Kazutaka Miyasaka <kazmiya@gmail.com>
 663  */
 664 function ft_resultComplement($args)
 665 {
 666     $array_count = count($args);
 667     if ($array_count === 1) {
 668         return $args[0];
 669     }
 670
 671     $result = $args[0];
 672     foreach (array_keys($result) as $id) {
 673         for ($i = 1; $i !== $array_count; $i++) {
 674             if (isset($args[$i][$id])) unset($result[$id]);
 675         }
 676     }
 677     return $result;
 678 }
 679
 680 /**
 681  * Parses a search query and builds an array of search formulas
 682  *
 683  * @author Andreas Gohr <andi@splitbrain.org>
 684  * @author Kazutaka Miyasaka <kazmiya@gmail.com>
 685  *
 686  * @param Indexer $Indexer
 687  * @param string                  $query search query
 688  * @return array of search formulas
 689  */
 690 function ft_queryParser($Indexer, $query)
 691 {
 692     /**
 693      * parse a search query and transform it into intermediate representation
 694      *
 695      * in a search query, you can use the following expressions:
 696      *
 697      *   words:
 698      *     include
 699      *     -exclude
 700      *   phrases:
 701      *     "phrase to be included"
 702      *     -"phrase you want to exclude"
 703      *   namespaces:
 704      *     @include:namespace (or ns:include:namespace)
 705      *     ^exclude:namespace (or -ns:exclude:namespace)
 706      *   groups:
 707      *     ()
 708      *     -()
 709      *   operators:
 710      *     and ('and' is the default operator: you can always omit this)
 711      *     or  (or pipe symbol '|', lower precedence than 'and')
 712      *
 713      * e.g. a query [ aa "bb cc" @dd:ee ] means "search pages which contain
 714      *      a word 'aa', a phrase 'bb cc' and are within a namespace 'dd:ee'".
 715      *      this query is equivalent to [ -(-aa or -"bb cc" or -ns:dd:ee) ]
 716      *      as long as you don't mind hit counts.
 717      *
 718      * intermediate representation consists of the following parts:
 719      *
 720      *   ( )           - group
 721      *   AND           - logical and
 722      *   OR            - logical or
 723      *   NOT           - logical not
 724      *   W+:, W-:, W_: - word      (underscore: no need to highlight)
 725      *   P+:, P-:      - phrase    (minus sign: logically in NOT group)
 726      *   N+:, N-:      - namespace
 727      */
 728     $parsed_query = '';
 729     $parens_level = 0;
 730     $terms = preg_split(
 731         '/(-?".*?")/u',
 732         PhpString::strtolower($query),
 733         -1,
 734         PREG_SPLIT_DELIM_CAPTURE | PREG_SPLIT_NO_EMPTY
 735     );
 736
 737     foreach ($terms as $term) {
 738         $parsed = '';
 739         if (preg_match('/^(-?)"(.+)"$/u', $term, $matches)) {
 740             // phrase-include and phrase-exclude
 741             $not = $matches[1] ? 'NOT' : '';
 742             $parsed = $not . ft_termParser($Indexer, $matches[2], false, true);
 743         } else {
 744             // fix incomplete phrase
 745             $term = str_replace('"', ' ', $term);
 746
 747             // fix parentheses
 748             $term = str_replace(')', ' ) ', $term);
 749             $term = str_replace('(', ' ( ', $term);
 750             $term = str_replace('- (', ' -(', $term);
 751
 752             // treat pipe symbols as 'OR' operators
 753             $term = str_replace('|', ' or ', $term);
 754
 755             // treat ideographic spaces (U+3000) as search term separators
 756             // FIXME: some more separators?
 757             $term = preg_replace('/[ \x{3000}]+/u', ' ', $term);
 758             $term = trim($term);
 759             if ($term === '') continue;
 760
 761             $tokens = explode(' ', $term);
 762             foreach ($tokens as $token) {
 763                 if ($token === '(') {
 764                     // parenthesis-include-open
 765                     $parsed .= '(';
 766                     ++$parens_level;
 767                 } elseif ($token === '-(') {
 768                     // parenthesis-exclude-open
 769                     $parsed .= 'NOT(';
 770                     ++$parens_level;
 771                 } elseif ($token === ')') {
 772                     // parenthesis-any-close
 773                     if ($parens_level === 0) continue;
 774                     $parsed .= ')';
 775                     $parens_level--;
 776                 } elseif ($token === 'and') {
 777                     // logical-and (do nothing)
 778                 } elseif ($token === 'or') {
 779                     // logical-or
 780                     $parsed .= 'OR';
 781                 } elseif (preg_match('/^(?:\^|-ns:)(.+)$/u', $token, $matches)) {
 782                     // namespace-exclude
 783                     $parsed .= 'NOT(N+:' . $matches[1] . ')';
 784                 } elseif (preg_match('/^(?:@|ns:)(.+)$/u', $token, $matches)) {
 785                     // namespace-include
 786                     $parsed .= '(N+:' . $matches[1] . ')';
 787                 } elseif (preg_match('/^-(.+)$/', $token, $matches)) {
 788                     // word-exclude
 789                     $parsed .= 'NOT(' . ft_termParser($Indexer, $matches[1]) . ')';
 790                 } else {
 791                     // word-include
 792                     $parsed .= ft_termParser($Indexer, $token);
 793                 }
 794             }
 795         }
 796         $parsed_query .= $parsed;
 797     }
 798
 799     // cleanup (very sensitive)
 800     $parsed_query .= str_repeat(')', $parens_level);
 801     do {
 802         $parsed_query_old = $parsed_query;
 803         $parsed_query = preg_replace('/(NOT)?\(\)/u', '', $parsed_query);
 804     } while ($parsed_query !== $parsed_query_old);
 805     $parsed_query = preg_replace('/(NOT|OR)+\)/u', ')', $parsed_query);
 806     $parsed_query = preg_replace('/(OR)+/u', 'OR', $parsed_query);
 807     $parsed_query = preg_replace('/\(OR/u', '(', $parsed_query);
 808     $parsed_query = preg_replace('/^OR|OR$/u', '', $parsed_query);
 809     $parsed_query = preg_replace('/\)(NOT)?\(/u', ')AND$1(', $parsed_query);
 810
 811     // adjustment: make highlightings right
 812     $parens_level     = 0;
 813     $notgrp_levels    = [];
 814     $parsed_query_new = '';
 815     $tokens = preg_split('/(NOT\(|[()])/u', $parsed_query, -1, PREG_SPLIT_DELIM_CAPTURE | PREG_SPLIT_NO_EMPTY);
 816     foreach ($tokens as $token) {
 817         if ($token === 'NOT(') {
 818             $notgrp_levels[] = ++$parens_level;
 819         } elseif ($token === '(') {
 820             ++$parens_level;
 821         } elseif ($token === ')') {
 822             if ($parens_level-- === end($notgrp_levels)) array_pop($notgrp_levels);
 823         } elseif (count($notgrp_levels) % 2 === 1) {
 824             // turn highlight-flag off if terms are logically in "NOT" group
 825             $token = preg_replace('/([WPN])\+\:/u', '$1-:', $token);
 826         }
 827         $parsed_query_new .= $token;
 828     }
 829     $parsed_query = $parsed_query_new;
 830
 831     /**
 832      * convert infix notation string into postfix (Reverse Polish notation) array
 833      * by Shunting-yard algorithm
 834      *
 835      * see: http://en.wikipedia.org/wiki/Reverse_Polish_notation
 836      * see: http://en.wikipedia.org/wiki/Shunting-yard_algorithm
 837      */
 838     $parsed_ary     = [];
 839     $ope_stack      = [];
 840     $ope_precedence = [')' => 1, 'OR' => 2, 'AND' => 3, 'NOT' => 4, '(' => 5];
 841     $ope_regex      = '/([()]|OR|AND|NOT)/u';
 842
 843     $tokens = preg_split($ope_regex, $parsed_query, -1, PREG_SPLIT_DELIM_CAPTURE | PREG_SPLIT_NO_EMPTY);
 844     foreach ($tokens as $token) {
 845         if (preg_match($ope_regex, $token)) {
 846             // operator
 847             $last_ope = end($ope_stack);
 848             while ($last_ope !== false && $ope_precedence[$token] <= $ope_precedence[$last_ope] && $last_ope != '(') {
 849                 $parsed_ary[] = array_pop($ope_stack);
 850                 $last_ope = end($ope_stack);
 851             }
 852             if ($token == ')') {
 853                 array_pop($ope_stack); // this array_pop always deletes '('
 854             } else {
 855                 $ope_stack[] = $token;
 856             }
 857         } else {
 858             // operand
 859             $token_decoded = str_replace(['OP', 'CP'], ['(', ')'], $token);
 860             $parsed_ary[] = $token_decoded;
 861         }
 862     }
 863     $parsed_ary = array_values([...$parsed_ary, ...array_reverse($ope_stack)]);
 864
 865     // cleanup: each double "NOT" in RPN array actually does nothing
 866     $parsed_ary_count = count($parsed_ary);
 867     for ($i = 1; $i < $parsed_ary_count; ++$i) {
 868         if ($parsed_ary[$i] === 'NOT' && $parsed_ary[$i - 1] === 'NOT') {
 869             unset($parsed_ary[$i], $parsed_ary[$i - 1]);
 870         }
 871     }
 872     $parsed_ary = array_values($parsed_ary);
 873
 874     // build return value
 875     $q = [];
 876     $q['query']      = $query;
 877     $q['parsed_str'] = $parsed_query;
 878     $q['parsed_ary'] = $parsed_ary;
 879
 880     foreach ($q['parsed_ary'] as $token) {
 881         if (strlen($token) < 3 || $token[2] !== ':') continue;
 882         $body = substr($token, 3);
 883
 884         switch (substr($token, 0, 3)) {
 885             case 'N+:':
 886                      $q['ns'][]        = $body; // for backward compatibility
 887                 break;
 888             case 'N-:':
 889                      $q['notns'][]     = $body; // for backward compatibility
 890                 break;
 891             case 'W_:':
 892                      $q['words'][]     = $body;
 893                 break;
 894             case 'W-:':
 895                      $q['words'][]     = $body;
 896                      $q['not'][]       = $body; // for backward compatibility
 897                 break;
 898             case 'W+:':
 899                      $q['words'][]     = $body;
 900                      $q['highlight'][] = $body;
 901                      $q['and'][]       = $body; // for backward compatibility
 902                 break;
 903             case 'P-:':
 904                      $q['phrases'][]   = $body;
 905                 break;
 906             case 'P+:':
 907                      $q['phrases'][]   = $body;
 908                      $q['highlight'][] = $body;
 909                 break;
 910         }
 911     }
 912     foreach (['words', 'phrases', 'highlight', 'ns', 'notns', 'and', 'not'] as $key) {
 913         $q[$key] = empty($q[$key]) ? [] : array_values(array_unique($q[$key]));
 914     }
 915
 916     return $q;
 917 }
 918
 919 /**
 920  * Transforms given search term into intermediate representation
 921  *
 922  * This function is used in ft_queryParser() and not for general purpose use.
 923  *
 924  * @author Kazutaka Miyasaka <kazmiya@gmail.com>
 925  *
 926  * @param Indexer $Indexer
 927  * @param string                  $term
 928  * @param bool                    $consider_asian
 929  * @param bool                    $phrase_mode
 930  * @return string
 931  */
 932 function ft_termParser($Indexer, $term, $consider_asian = true, $phrase_mode = false)
 933 {
 934     $parsed = '';
 935     if ($consider_asian) {
 936         // successive asian characters need to be searched as a phrase
 937         $words = Asian::splitAsianWords($term);
 938         foreach ($words as $word) {
 939             $phrase_mode = $phrase_mode ? true : Asian::isAsianWords($word);
 940             $parsed .= ft_termParser($Indexer, $word, false, $phrase_mode);
 941         }
 942     } else {
 943         $term_noparen = str_replace(['(', ')'], ' ', $term);
 944         $words = $Indexer->tokenizer($term_noparen, true);
 945
 946         // W_: no need to highlight
 947         if (empty($words)) {
 948             $parsed = '()'; // important: do not remove
 949         } elseif ($words[0] === $term) {
 950             $parsed = '(W+:' . $words[0] . ')';
 951         } elseif ($phrase_mode) {
 952             $term_encoded = str_replace(['(', ')'], ['OP', 'CP'], $term);
 953             $parsed = '((W_:' . implode(')(W_:', $words) . ')(P+:' . $term_encoded . '))';
 954         } else {
 955             $parsed = '((W+:' . implode(')(W+:', $words) . '))';
 956         }
 957     }
 958     return $parsed;
 959 }
 960
 961 /**
 962  * Recreate a search query string based on parsed parts, doesn't support negated phrases and `OR` searches
 963  *
 964  * @param array $and
 965  * @param array $not
 966  * @param array $phrases
 967  * @param array $ns
 968  * @param array $notns
 969  *
 970  * @return string
 971  */
 972 function ft_queryUnparser_simple(array $and, array $not, array $phrases, array $ns, array $notns)
 973 {
 974     $query = implode(' ', $and);
 975     if ($not !== []) {
 976         $query .= ' -' . implode(' -', $not);
 977     }
 978
 979     if ($phrases !== []) {
 980         $query .= ' "' . implode('" "', $phrases) . '"';
 981     }
 982
 983     if ($ns !== []) {
 984         $query .= ' @' . implode(' @', $ns);
 985     }
 986
 987     if ($notns !== []) {
 988         $query .= ' ^' . implode(' ^', $notns);
 989     }
 990
 991     return $query;
 992 }
 993
 994 //Setup VIM: ex: et ts=4 :