1<?php 2namespace dokuwiki\Search; 3 4use dokuwiki\Extension\Event; 5use dokuwiki\Search\Indexer; 6use dokuwiki\Utf8; 7 8/** 9 * Class DokuWiki Fulltext Search 10 * 11 * @license GPL 2 (http://www.gnu.org/licenses/gpl.html) 12 * @author Andreas Gohr <andi@splitbrain.org> 13 */ 14class FulltextSearch 15{ 16 /** 17 * Fulltext Search constructor. prevent direct object creation 18 */ 19 protected function __construct() {} 20 21 /** 22 * The fulltext search 23 * 24 * Returns a list of matching documents for the given query 25 * 26 * refactored into ft_pageSearch(), _ft_pageSearch() and trigger_event() 27 * 28 * @param string $query 29 * @param array $highlight 30 * @param string $sort 31 * @param int|string $after only show results with mtime after this date, 32 * accepts timestap or strtotime arguments 33 * @param int|string $before only show results with mtime before this date, 34 * accepts timestap or strtotime arguments 35 * 36 * @return array 37 */ 38 public static function pageSearch($query, &$highlight, $sort = null, $after = null, $before = null) 39 { 40 if ($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 $action = static::class.'::callback_pageSearch'; 51 return Event::createAndTrigger('SEARCH_QUERY_FULLPAGE', $data, $action); 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 $data event data 61 * @return array matching documents 62 */ 63 public static function callback_pageSearch(&$data) 64 { 65 $Indexer = Indexer::getInstance(); 66 67 // parse the given query 68 $q = static::queryParser($Indexer, $data['query']); 69 $data['highlight'] = $q['highlight']; 70 71 if (empty($q['parsed_ary'])) return array(); 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 $pages_all = array(); 78 foreach ($Indexer->getPages() as $id) { 79 $pages_all[$id] = 0; // base: 0 hit 80 } 81 82 // process the query 83 $stack = array(); 84 foreach ($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 $stack[] = (array) $lookup[$word]; 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 $pages_matched = array(); 100 foreach (array_keys($pages) as $id) { 101 $evdata = array( 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 $text = Utf8\PhpString::strtolower($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 $pages_matched = array(); 124 foreach (array_keys($pages_all) as $id) { 125 if (strpos($id, $ns) === 0) { 126 $pages_matched[$id] = 0; // namespace: always 0 hit 127 } 128 } 129 $stack[] = $pages_matched; 130 break; 131 case 'AND': // and operation 132 list($pages1, $pages2) = array_splice($stack, -2); 133 $stack[] = static::resultCombine(array($pages1, $pages2)); 134 break; 135 case 'OR': // or operation 136 list($pages1, $pages2) = array_splice($stack, -2); 137 $stack[] = static::resultUnite(array($pages1, $pages2)); 138 break; 139 case 'NOT': // not operation (unary) 140 $pages = array_pop($stack); 141 $stack[] = static::resultComplement(array($pages_all, $pages)); 142 break; 143 } 144 } 145 $docs = array_pop($stack); 146 147 if (empty($docs)) return array(); 148 149 // check: settings, acls, existence 150 foreach (array_keys($docs) as $id) { 151 if (isHiddenPage($id) 152 || auth_quickaclcheck($id) < AUTH_READ 153 || !page_exists($id, '', false) 154 ) { 155 unset($docs[$id]); 156 } 157 } 158 159 $docs = static::filterResultsByTime($docs, $data['after'], $data['before']); 160 161 if ($data['sort'] === 'mtime') { 162 uksort($docs, static::class.'::pagemtimesorter'); 163 } else { 164 // sort docs by count 165 arsort($docs); 166 } 167 168 return $docs; 169 } 170 171 /** 172 * Quicksearch for pagenames 173 * 174 * By default it only matches the pagename and ignores the 175 * namespace. This can be changed with the second parameter. 176 * The third parameter allows to search in titles as well. 177 * 178 * The function always returns titles as well 179 * 180 * @triggers SEARCH_QUERY_PAGELOOKUP 181 * @author Andreas Gohr <andi@splitbrain.org> 182 * @author Adrian Lang <lang@cosmocode.de> 183 * 184 * @param string $id page id 185 * @param bool $in_ns match against namespace as well? 186 * @param bool $in_title search in title? 187 * @param int|string $after only show results with mtime after this date, 188 * accepts timestap or strtotime arguments 189 * @param int|string $before only show results with mtime before this date, 190 * accepts timestap or strtotime arguments 191 * 192 * @return string[] 193 */ 194 public static function pageLookup($id, $in_ns=false, $in_title=false, $after = null, $before = null) 195 { 196 $data = [ 197 'id' => $id, 198 'in_ns' => $in_ns, 199 'in_title' => $in_title, 200 'after' => $after, 201 'before' => $before 202 ]; 203 $data['has_titles'] = true; // for plugin backward compatibility check 204 $action = static::class.'::callback_pageLookup'; 205 return Event::createAndTrigger('SEARCH_QUERY_PAGELOOKUP', $data, $action); 206 } 207 208 /** 209 * Returns list of pages as array(pageid => First Heading) 210 * 211 * @param array &$data event data 212 * @return string[] 213 */ 214 public static function callback_pageLookup(&$data) 215 { 216 $Indexer = Indexer::getInstance(); 217 218 // split out original parameters 219 $id = $data['id']; 220 $parsedQuery = static::queryParser($Indexer, $id); 221 if (count($parsedQuery['ns']) > 0) { 222 $ns = cleanID($parsedQuery['ns'][0]) . ':'; 223 $id = implode(' ', $parsedQuery['highlight']); 224 } 225 226 $in_ns = $data['in_ns']; 227 $in_title = $data['in_title']; 228 $cleaned = cleanID($id); 229 230 $pages = array(); 231 if ($id !== '' && $cleaned !== '') { 232 $page_idx = $Indexer->getPages(); 233 foreach ($page_idx as $p_id) { 234 if ((strpos($in_ns ? $p_id : noNSorNS($p_id), $cleaned) !== false)) { 235 if (!isset($pages[$p_id])) { 236 $pages[$p_id] = p_get_first_heading($p_id, METADATA_DONT_RENDER); 237 } 238 } 239 } 240 if ($in_title) { 241 $func = static::class.'::pageLookupTitleCompare'; 242 foreach ($Indexer->lookupKey('title', $id, $func) as $p_id) { 243 if (!isset($pages[$p_id])) { 244 $pages[$p_id] = p_get_first_heading($p_id, METADATA_DONT_RENDER); 245 } 246 } 247 } 248 } 249 250 if (isset($ns)) { 251 foreach (array_keys($pages) as $p_id) { 252 if (strpos($p_id, $ns) !== 0) { 253 unset($pages[$p_id]); 254 } 255 } 256 } 257 258 // discard hidden pages 259 // discard nonexistent pages 260 // check ACL permissions 261 foreach (array_keys($pages) as $idx) { 262 if (!isVisiblePage($idx) || !page_exists($idx) || auth_quickaclcheck($idx) < AUTH_READ) { 263 unset($pages[$idx]); 264 } 265 } 266 267 $pages = static::filterResultsByTime($pages, $data['after'], $data['before']); 268 269 uksort($pages, static::class.'::pagesorter'); 270 return $pages; 271 } 272 273 /** 274 * @param array $results search results in the form pageid => value 275 * @param int|string $after only returns results with mtime after this date, 276 * accepts timestap or strtotime arguments 277 * @param int|string $before only returns results with mtime after this date, 278 * accepts timestap or strtotime arguments 279 * 280 * @return array 281 */ 282 protected static function filterResultsByTime(array $results, $after, $before) 283 { 284 if ($after || $before) { 285 $after = is_int($after) ? $after : strtotime($after); 286 $before = is_int($before) ? $before : strtotime($before); 287 288 foreach ($results as $id => $value) { 289 $mTime = filemtime(wikiFN($id)); 290 if ($after && $after > $mTime) { 291 unset($results[$id]); 292 continue; 293 } 294 if ($before && $before < $mTime) { 295 unset($results[$id]); 296 } 297 } 298 } 299 300 return $results; 301 } 302 303 /** 304 * Tiny helper function for comparing the searched title with the title 305 * from the search index. This function is a wrapper around stripos with 306 * adapted argument order and return value. 307 * 308 * @param string $search searched title 309 * @param string $title title from index 310 * @return bool 311 */ 312 protected static function pageLookupTitleCompare($search, $title) 313 { 314 return stripos($title, $search) !== false; 315 } 316 317 /** 318 * Sort pages based on their namespace level first, then on their string 319 * values. This makes higher hierarchy pages rank higher than lower hierarchy 320 * pages. 321 * 322 * @param string $a 323 * @param string $b 324 * @return int Returns < 0 if $a is less than $b; > 0 if $a is greater than $b, 325 * and 0 if they are equal. 326 */ 327 protected static function pagesorter($a, $b) 328 { 329 $ac = count(explode(':',$a)); 330 $bc = count(explode(':',$b)); 331 if ($ac < $bc) { 332 return -1; 333 } elseif ($ac > $bc) { 334 return 1; 335 } 336 return strcmp ($a,$b); 337 } 338 339 /** 340 * Sort pages by their mtime, from newest to oldest 341 * 342 * @param string $a 343 * @param string $b 344 * 345 * @return int Returns < 0 if $a is newer than $b, > 0 if $b is newer than $a 346 * and 0 if they are of the same age 347 */ 348 protected static function pagemtimesorter($a, $b) 349 { 350 $mtimeA = filemtime(wikiFN($a)); 351 $mtimeB = filemtime(wikiFN($b)); 352 return $mtimeB - $mtimeA; 353 } 354 355 /** 356 * Creates a snippet extract 357 * 358 * @author Andreas Gohr <andi@splitbrain.org> 359 * @triggers FULLTEXT_SNIPPET_CREATE 360 * 361 * @param string $id page id 362 * @param array $highlight 363 * @return mixed 364 */ 365 public static function snippet($id, $highlight) 366 { 367 $text = rawWiki($id); 368 $text = str_replace("\xC2\xAD",'',$text); // remove soft-hyphens 369 $evdata = array( 370 'id' => $id, 371 'text' => &$text, 372 'highlight' => &$highlight, 373 'snippet' => '', 374 ); 375 376 $evt = new Event('FULLTEXT_SNIPPET_CREATE', $evdata); 377 if ($evt->advise_before()) { 378 $match = array(); 379 $snippets = array(); 380 $utf8_offset = $offset = $end = 0; 381 $len = Utf8\PhpString::strlen($text); 382 383 // build a regexp from the phrases to highlight 384 $re1 = '(' . 385 join( 386 '|', 387 array_map( 388 static::class.'::snippet_re_preprocess', 389 array_map( 390 'preg_quote_cb', 391 array_filter((array) $highlight) 392 ) 393 ) 394 ) . 395 ')'; 396 $re2 = "$re1.{0,75}(?!\\1)$re1"; 397 $re3 = "$re1.{0,45}(?!\\1)$re1.{0,45}(?!\\1)(?!\\2)$re1"; 398 399 for ($cnt=4; $cnt--;) { 400 if (0) { 401 } elseif (preg_match('/'.$re3.'/iu', $text, $match, PREG_OFFSET_CAPTURE, $offset)) { 402 } elseif (preg_match('/'.$re2.'/iu', $text, $match, PREG_OFFSET_CAPTURE, $offset)) { 403 } elseif (preg_match('/'.$re1.'/iu', $text, $match, PREG_OFFSET_CAPTURE, $offset)) { 404 } else { 405 break; 406 } 407 408 list($str, $idx) = $match[0]; 409 410 // convert $idx (a byte offset) into a utf8 character offset 411 $utf8_idx = Utf8\PhpString::strlen(substr($text, 0, $idx)); 412 $utf8_len = Utf8\PhpString::strlen($str); 413 414 // establish context, 100 bytes surrounding the match string 415 // first look to see if we can go 100 either side, 416 // then drop to 50 adding any excess if the other side can't go to 50, 417 $pre = min($utf8_idx - $utf8_offset, 100); 418 $post = min($len - $utf8_idx - $utf8_len, 100); 419 420 if ($pre > 50 && $post > 50) { 421 $pre = $post = 50; 422 } elseif ($pre > 50) { 423 $pre = min($pre, 100 - $post); 424 } elseif ($post > 50) { 425 $post = min($post, 100 - $pre); 426 } elseif ($offset == 0) { 427 // both are less than 50, means the context is the whole string 428 // make it so and break out of this loop - there is no need for the 429 // complex snippet calculations 430 $snippets = array($text); 431 break; 432 } 433 434 // establish context start and end points, try to append to previous 435 // context if possible 436 $start = $utf8_idx - $pre; 437 $append = ($start < $end) ? $end : false; // still the end of the previous context snippet 438 $end = $utf8_idx + $utf8_len + $post; // now set it to the end of this context 439 440 if ($append) { 441 $snippets[count($snippets)-1] .= Utf8\PhpString::substr($text, $append, $end-$append); 442 } else { 443 $snippets[] = Utf8\PhpString::substr($text, $start, $end-$start); 444 } 445 446 // set $offset for next match attempt 447 // continue matching after the current match 448 // if the current match is not the longest possible match starting at the current offset 449 // this prevents further matching of this snippet but for possible matches of length 450 // smaller than match length + context (at least 50 characters) this match is part of the context 451 $utf8_offset = $utf8_idx + $utf8_len; 452 $offset = $idx + strlen(Utf8\PhpString::substr($text, $utf8_idx, $utf8_len)); 453 $offset = Utf8\Clean::correctIdx($text, $offset); 454 } 455 456 $m = "\1"; 457 $snippets = preg_replace('/'.$re1.'/iu', $m.'$1'.$m, $snippets); 458 $snippet = preg_replace( 459 '/' . $m . '([^' . $m . ']*?)' . $m . '/iu', 460 '<strong class="search_hit">$1</strong>', 461 hsc(join('... ', $snippets)) 462 ); 463 464 $evdata['snippet'] = $snippet; 465 } 466 $evt->advise_after(); 467 unset($evt); 468 469 return $evdata['snippet']; 470 } 471 472 /** 473 * Wraps a search term in regex boundary checks. 474 * 475 * @param string $term 476 * @return string 477 */ 478 public static function snippet_re_preprocess($term) 479 { 480 // do not process asian terms where word boundaries are not explicit 481 if (Utf8\Asian::isAsianWords($term)) return $term; 482 483 if (UTF8_PROPERTYSUPPORT) { 484 // unicode word boundaries 485 // see http://stackoverflow.com/a/2449017/172068 486 $BL = '(?<!\pL)'; 487 $BR = '(?!\pL)'; 488 } else { 489 // not as correct as above, but at least won't break 490 $BL = '\b'; 491 $BR = '\b'; 492 } 493 494 if (substr($term, 0, 2) == '\\*') { 495 $term = substr($term, 2); 496 } else { 497 $term = $BL.$term; 498 } 499 500 if (substr($term, -2, 2) == '\\*') { 501 $term = substr($term, 0, -2); 502 } else { 503 $term = $term.$BR; 504 } 505 506 if ($term == $BL || $term == $BR || $term == $BL.$BR) { 507 $term = ''; 508 } 509 return $term; 510 } 511 512 /** 513 * Combine found documents and sum up their scores 514 * 515 * This function is used to combine searched words with a logical 516 * AND. Only documents available in all arrays are returned. 517 * 518 * based upon PEAR's PHP_Compat function for array_intersect_key() 519 * 520 * @param array $args An array of page arrays 521 * @return array 522 */ 523 protected static function resultCombine($args) 524 { 525 $array_count = count($args); 526 if ($array_count == 1) { 527 return $args[0]; 528 } 529 530 $result = array(); 531 if ($array_count > 1) { 532 foreach ($args[0] as $key => $value) { 533 $result[$key] = $value; 534 for ($i = 1; $i !== $array_count; $i++) { 535 if (!isset($args[$i][$key])) { 536 unset($result[$key]); 537 break; 538 } 539 $result[$key] += $args[$i][$key]; 540 } 541 } 542 } 543 return $result; 544 } 545 546 /** 547 * Unites found documents and sum up their scores 548 * based upon resultCombine() method 549 * 550 * @param array $args An array of page arrays 551 * @return array 552 * 553 * @author Kazutaka Miyasaka <kazmiya@gmail.com> 554 */ 555 protected static function resultUnite($args) 556 { 557 $array_count = count($args); 558 if ($array_count === 1) { 559 return $args[0]; 560 } 561 562 $result = $args[0]; 563 for ($i = 1; $i !== $array_count; $i++) { 564 foreach (array_keys($args[$i]) as $id) { 565 $result[$id] += $args[$i][$id]; 566 } 567 } 568 return $result; 569 } 570 571 /** 572 * Computes the difference of documents using page id for comparison 573 * nearly identical to PHP5's array_diff_key() 574 * 575 * @param array $args An array of page arrays 576 * @return array 577 * 578 * @author Kazutaka Miyasaka <kazmiya@gmail.com> 579 */ 580 protected static function resultComplement($args) 581 { 582 $array_count = count($args); 583 if ($array_count === 1) { 584 return $args[0]; 585 } 586 587 $result = $args[0]; 588 foreach (array_keys($result) as $id) { 589 for ($i = 1; $i !== $array_count; $i++) { 590 if (isset($args[$i][$id])) unset($result[$id]); 591 } 592 } 593 return $result; 594 } 595 596 /** 597 * Parses a search query and builds an array of search formulas 598 * 599 * @author Andreas Gohr <andi@splitbrain.org> 600 * @author Kazutaka Miyasaka <kazmiya@gmail.com> 601 * 602 * @param Indexer $Indexer 603 * @param string $query search query 604 * @return array of search formulas 605 */ 606 public static function queryParser($Indexer, $query) 607 { 608 /** 609 * parse a search query and transform it into intermediate representation 610 * 611 * in a search query, you can use the following expressions: 612 * 613 * words: 614 * include 615 * -exclude 616 * phrases: 617 * "phrase to be included" 618 * -"phrase you want to exclude" 619 * namespaces: 620 * @include:namespace (or ns:include:namespace) 621 * ^exclude:namespace (or -ns:exclude:namespace) 622 * groups: 623 * () 624 * -() 625 * operators: 626 * and ('and' is the default operator: you can always omit this) 627 * or (or pipe symbol '|', lower precedence than 'and') 628 * 629 * e.g. a query [ aa "bb cc" @dd:ee ] means "search pages which contain 630 * a word 'aa', a phrase 'bb cc' and are within a namespace 'dd:ee'". 631 * this query is equivalent to [ -(-aa or -"bb cc" or -ns:dd:ee) ] 632 * as long as you don't mind hit counts. 633 * 634 * intermediate representation consists of the following parts: 635 * 636 * ( ) - group 637 * AND - logical and 638 * OR - logical or 639 * NOT - logical not 640 * W+:, W-:, W_: - word (underscore: no need to highlight) 641 * P+:, P-: - phrase (minus sign: logically in NOT group) 642 * N+:, N-: - namespace 643 */ 644 $parsed_query = ''; 645 $parens_level = 0; 646 $terms = preg_split('/(-?".*?")/u', Utf8\PhpString::strtolower($query), 647 -1, PREG_SPLIT_DELIM_CAPTURE | PREG_SPLIT_NO_EMPTY 648 ); 649 650 foreach ($terms as $term) { 651 $parsed = ''; 652 if (preg_match('/^(-?)"(.+)"$/u', $term, $matches)) { 653 // phrase-include and phrase-exclude 654 $not = $matches[1] ? 'NOT' : ''; 655 $parsed = $not . static::termParser($Indexer, $matches[2], false, true); 656 } else { 657 // fix incomplete phrase 658 $term = str_replace('"', ' ', $term); 659 660 // fix parentheses 661 $term = str_replace(')' , ' ) ', $term); 662 $term = str_replace('(' , ' ( ', $term); 663 $term = str_replace('- (', ' -(', $term); 664 665 // treat pipe symbols as 'OR' operators 666 $term = str_replace('|', ' or ', $term); 667 668 // treat ideographic spaces (U+3000) as search term separators 669 // FIXME: some more separators? 670 $term = preg_replace('/[ \x{3000}]+/u', ' ', $term); 671 $term = trim($term); 672 if ($term === '') continue; 673 674 $tokens = explode(' ', $term); 675 foreach ($tokens as $token) { 676 if ($token === '(') { 677 // parenthesis-include-open 678 $parsed .= '('; 679 ++$parens_level; 680 } elseif ($token === '-(') { 681 // parenthesis-exclude-open 682 $parsed .= 'NOT('; 683 ++$parens_level; 684 } elseif ($token === ')') { 685 // parenthesis-any-close 686 if ($parens_level === 0) continue; 687 $parsed .= ')'; 688 $parens_level--; 689 } elseif ($token === 'and') { 690 // logical-and (do nothing) 691 } elseif ($token === 'or') { 692 // logical-or 693 $parsed .= 'OR'; 694 } elseif (preg_match('/^(?:\^|-ns:)(.+)$/u', $token, $matches)) { 695 // namespace-exclude 696 $parsed .= 'NOT(N+:'.$matches[1].')'; 697 } elseif (preg_match('/^(?:@|ns:)(.+)$/u', $token, $matches)) { 698 // namespace-include 699 $parsed .= '(N+:'.$matches[1].')'; 700 } elseif (preg_match('/^-(.+)$/', $token, $matches)) { 701 // word-exclude 702 $parsed .= 'NOT('.static::termParser($Indexer, $matches[1]).')'; 703 } else { 704 // word-include 705 $parsed .= static::termParser($Indexer, $token); 706 } 707 } 708 } 709 $parsed_query .= $parsed; 710 } 711 712 // cleanup (very sensitive) 713 $parsed_query .= str_repeat(')', $parens_level); 714 do { 715 $parsed_query_old = $parsed_query; 716 $parsed_query = preg_replace('/(NOT)?\(\)/u', '', $parsed_query); 717 } while ($parsed_query !== $parsed_query_old); 718 $parsed_query = preg_replace('/(NOT|OR)+\)/u', ')' , $parsed_query); 719 $parsed_query = preg_replace('/(OR)+/u' , 'OR' , $parsed_query); 720 $parsed_query = preg_replace('/\(OR/u' , '(' , $parsed_query); 721 $parsed_query = preg_replace('/^OR|OR$/u' , '' , $parsed_query); 722 $parsed_query = preg_replace('/\)(NOT)?\(/u' , ')AND$1(', $parsed_query); 723 724 // adjustment: make highlightings right 725 $parens_level = 0; 726 $notgrp_levels = array(); 727 $parsed_query_new = ''; 728 $tokens = preg_split('/(NOT\(|[()])/u', $parsed_query, 729 -1, PREG_SPLIT_DELIM_CAPTURE | PREG_SPLIT_NO_EMPTY 730 ); 731 foreach ($tokens as $token) { 732 if ($token === 'NOT(') { 733 $notgrp_levels[] = ++$parens_level; 734 } elseif ($token === '(') { 735 ++$parens_level; 736 } elseif ($token === ')') { 737 if ($parens_level-- === end($notgrp_levels)) array_pop($notgrp_levels); 738 } elseif (count($notgrp_levels) % 2 === 1) { 739 // turn highlight-flag off if terms are logically in "NOT" group 740 $token = preg_replace('/([WPN])\+\:/u', '$1-:', $token); 741 } 742 $parsed_query_new .= $token; 743 } 744 $parsed_query = $parsed_query_new; 745 746 /** 747 * convert infix notation string into postfix (Reverse Polish notation) array 748 * by Shunting-yard algorithm 749 * 750 * see: http://en.wikipedia.org/wiki/Reverse_Polish_notation 751 * see: http://en.wikipedia.org/wiki/Shunting-yard_algorithm 752 */ 753 $parsed_ary = array(); 754 $ope_stack = array(); 755 $ope_precedence = array(')' => 1, 'OR' => 2, 'AND' => 3, 'NOT' => 4, '(' => 5); 756 $ope_regex = '/([()]|OR|AND|NOT)/u'; 757 758 $tokens = preg_split($ope_regex, $parsed_query, 759 -1, PREG_SPLIT_DELIM_CAPTURE | PREG_SPLIT_NO_EMPTY 760 ); 761 foreach ($tokens as $token) { 762 if (preg_match($ope_regex, $token)) { 763 // operator 764 $last_ope = end($ope_stack); 765 while ($last_ope !== false 766 && $ope_precedence[$token] <= $ope_precedence[$last_ope] 767 && $last_ope != '(' 768 ) { 769 $parsed_ary[] = array_pop($ope_stack); 770 $last_ope = end($ope_stack); 771 } 772 if ($token == ')') { 773 array_pop($ope_stack); // this array_pop always deletes '(' 774 } else { 775 $ope_stack[] = $token; 776 } 777 } else { 778 // operand 779 $token_decoded = str_replace(['OP','CP'], ['(',')'], $token); 780 $parsed_ary[] = $token_decoded; 781 } 782 } 783 $parsed_ary = array_values(array_merge($parsed_ary, array_reverse($ope_stack))); 784 785 // cleanup: each double "NOT" in RPN array actually does nothing 786 $parsed_ary_count = count($parsed_ary); 787 for ($i = 1; $i < $parsed_ary_count; ++$i) { 788 if ($parsed_ary[$i] === 'NOT' && $parsed_ary[$i - 1] === 'NOT') { 789 unset($parsed_ary[$i], $parsed_ary[$i - 1]); 790 } 791 } 792 $parsed_ary = array_values($parsed_ary); 793 794 // build return value 795 $q = array(); 796 $q['query'] = $query; 797 $q['parsed_str'] = $parsed_query; 798 $q['parsed_ary'] = $parsed_ary; 799 800 foreach ($q['parsed_ary'] as $token) { 801 if ($token[2] !== ':') continue; 802 $body = substr($token, 3); 803 804 switch (substr($token, 0, 3)) { 805 case 'N+:': 806 $q['ns'][] = $body; // for backward compatibility 807 break; 808 case 'N-:': 809 $q['notns'][] = $body; // for backward compatibility 810 break; 811 case 'W_:': 812 $q['words'][] = $body; 813 break; 814 case 'W-:': 815 $q['words'][] = $body; 816 $q['not'][] = $body; // for backward compatibility 817 break; 818 case 'W+:': 819 $q['words'][] = $body; 820 $q['highlight'][] = $body; 821 $q['and'][] = $body; // for backward compatibility 822 break; 823 case 'P-:': 824 $q['phrases'][] = $body; 825 break; 826 case 'P+:': 827 $q['phrases'][] = $body; 828 $q['highlight'][] = $body; 829 break; 830 } 831 } 832 foreach (['words', 'phrases', 'highlight', 'ns', 'notns', 'and', 'not'] as $key) { 833 $q[$key] = empty($q[$key]) ? array() : array_values(array_unique($q[$key])); 834 } 835 836 return $q; 837 } 838 839 /** 840 * Transforms given search term into intermediate representation 841 * 842 * This function is used in ft_queryParser() and not for general purpose use. 843 * 844 * @author Kazutaka Miyasaka <kazmiya@gmail.com> 845 * 846 * @param Indexer $Indexer 847 * @param string $term 848 * @param bool $consider_asian 849 * @param bool $phrase_mode 850 * @return string 851 */ 852 protected static function termParser($Indexer, $term, $consider_asian = true, $phrase_mode = false) 853 { 854 $parsed = ''; 855 if ($consider_asian) { 856 // successive asian characters need to be searched as a phrase 857 $words = Utf8\Asian::splitAsianWords($term); 858 foreach ($words as $word) { 859 $phrase_mode = $phrase_mode ? true : Utf8\Asian::isAsianWords($word); 860 $parsed .= static::termParser($Indexer, $word, false, $phrase_mode); 861 } 862 } else { 863 $term_noparen = str_replace(['(',')'], ' ', $term); 864 $words = $Indexer->tokenizer($term_noparen, true); 865 866 // W_: no need to highlight 867 if (empty($words)) { 868 $parsed = '()'; // important: do not remove 869 } elseif ($words[0] === $term) { 870 $parsed = '(W+:'.$words[0].')'; 871 } elseif ($phrase_mode) { 872 $term_encoded = str_replace(['(',')'], ['OP','CP'], $term); 873 $parsed = '((W_:'.implode(')(W_:', $words).')(P+:'.$term_encoded.'))'; 874 } else { 875 $parsed = '((W+:'.implode(')(W+:', $words).'))'; 876 } 877 } 878 return $parsed; 879 } 880 881 /** 882 * Recreate a search query string based on parsed parts, 883 * doesn't support negated phrases and `OR` searches 884 * 885 * @param array $and 886 * @param array $not 887 * @param array $phrases 888 * @param array $ns 889 * @param array $notns 890 * 891 * @return string 892 */ 893 public static function queryUnparser_simple( 894 array $and, array $not, array $phrases, array $ns, array $notns 895 ) { 896 $query = implode(' ', $and); 897 898 if (!empty($not)) { 899 $query .= ' -' . implode(' -', $not); 900 } 901 if (!empty($phrases)) { 902 $query .= ' "' . implode('" "', $phrases) . '"'; 903 } 904 if (!empty($ns)) { 905 $query .= ' @' . implode(' @', $ns); 906 } 907 if (!empty($notns)) { 908 $query .= ' ^' . implode(' ^', $notns); 909 } 910 return $query; 911 } 912} 913