1<?php 2/** 3 * DokuWiki fulltextsearch functions using the index 4 * 5 * @license GPL 2 (http://www.gnu.org/licenses/gpl.html) 6 * @author Andreas Gohr <andi@splitbrain.org> 7 */ 8 9use dokuwiki\Extension\Event; 10use dokuwiki\Utf8\Sort; 11 12/** 13 * create snippets for the first few results only 14 */ 15if(!defined('FT_SNIPPET_NUMBER')) define('FT_SNIPPET_NUMBER',15); 16 17/** 18 * The fulltext search 19 * 20 * Returns a list of matching documents for the given query 21 * 22 * refactored into ft_pageSearch(), _ft_pageSearch() and trigger_event() 23 * 24 * @param string $query 25 * @param array $highlight 26 * @param string $sort 27 * @param int|string $after only show results with mtime after this date, accepts timestap or strtotime arguments 28 * @param int|string $before only show results with mtime before this date, accepts timestap or strtotime arguments 29 * 30 * @return array 31 */ 32function ft_pageSearch($query,&$highlight, $sort = null, $after = null, $before = null){ 33 34 if ($sort === null) { 35 $sort = 'hits'; 36 } 37 $data = [ 38 'query' => $query, 39 'sort' => $sort, 40 'after' => $after, 41 'before' => $before 42 ]; 43 $data['highlight'] =& $highlight; 44 45 return Event::createAndTrigger('SEARCH_QUERY_FULLPAGE', $data, '_ft_pageSearch'); 46} 47 48/** 49 * Returns a list of matching documents for the given query 50 * 51 * @author Andreas Gohr <andi@splitbrain.org> 52 * @author Kazutaka Miyasaka <kazmiya@gmail.com> 53 * 54 * @param array $data event data 55 * @return array matching documents 56 */ 57function _ft_pageSearch(&$data) { 58 $Indexer = idx_get_indexer(); 59 60 // parse the given query 61 $q = ft_queryParser($Indexer, $data['query']); 62 $data['highlight'] = $q['highlight']; 63 64 if (empty($q['parsed_ary'])) return array(); 65 66 // lookup all words found in the query 67 $lookup = $Indexer->lookup($q['words']); 68 69 // get all pages in this dokuwiki site (!: includes nonexistent pages) 70 $pages_all = array(); 71 foreach ($Indexer->getPages() as $id) { 72 $pages_all[$id] = 0; // base: 0 hit 73 } 74 75 // process the query 76 $stack = array(); 77 foreach ($q['parsed_ary'] as $token) { 78 switch (substr($token, 0, 3)) { 79 case 'W+:': 80 case 'W-:': 81 case 'W_:': // word 82 $word = substr($token, 3); 83 $stack[] = (array) $lookup[$word]; 84 break; 85 case 'P+:': 86 case 'P-:': // phrase 87 $phrase = substr($token, 3); 88 // since phrases are always parsed as ((W1)(W2)...(P)), 89 // the end($stack) always points the pages that contain 90 // all words in this phrase 91 $pages = end($stack); 92 $pages_matched = array(); 93 foreach(array_keys($pages) as $id){ 94 $evdata = array( 95 'id' => $id, 96 'phrase' => $phrase, 97 'text' => rawWiki($id) 98 ); 99 $evt = new Event('FULLTEXT_PHRASE_MATCH',$evdata); 100 if ($evt->advise_before() && $evt->result !== true) { 101 $text = \dokuwiki\Utf8\PhpString::strtolower($evdata['text']); 102 if (strpos($text, $phrase) !== false) { 103 $evt->result = true; 104 } 105 } 106 $evt->advise_after(); 107 if ($evt->result === true) { 108 $pages_matched[$id] = 0; // phrase: always 0 hit 109 } 110 } 111 $stack[] = $pages_matched; 112 break; 113 case 'N+:': 114 case 'N-:': // namespace 115 $ns = cleanID(substr($token, 3)) . ':'; 116 $pages_matched = array(); 117 foreach (array_keys($pages_all) as $id) { 118 if (strpos($id, $ns) === 0) { 119 $pages_matched[$id] = 0; // namespace: always 0 hit 120 } 121 } 122 $stack[] = $pages_matched; 123 break; 124 case 'AND': // and operation 125 list($pages1, $pages2) = array_splice($stack, -2); 126 $stack[] = ft_resultCombine(array($pages1, $pages2)); 127 break; 128 case 'OR': // or operation 129 list($pages1, $pages2) = array_splice($stack, -2); 130 $stack[] = ft_resultUnite(array($pages1, $pages2)); 131 break; 132 case 'NOT': // not operation (unary) 133 $pages = array_pop($stack); 134 $stack[] = ft_resultComplement(array($pages_all, $pages)); 135 break; 136 } 137 } 138 $docs = array_pop($stack); 139 140 if (empty($docs)) return array(); 141 142 // check: settings, acls, existence 143 foreach (array_keys($docs) as $id) { 144 if (isHiddenPage($id) || auth_quickaclcheck($id) < AUTH_READ || !page_exists($id, '', false)) { 145 unset($docs[$id]); 146 } 147 } 148 149 $docs = _ft_filterResultsByTime($docs, $data['after'], $data['before']); 150 151 if ($data['sort'] === 'mtime') { 152 uksort($docs, 'ft_pagemtimesorter'); 153 } else { 154 // sort docs by count 155 uksort($docs, 'ft_pagesorter'); 156 arsort($docs); 157 } 158 159 return $docs; 160} 161 162/** 163 * Returns the backlinks for a given page 164 * 165 * Uses the metadata index. 166 * 167 * @param string $id The id for which links shall be returned 168 * @param bool $ignore_perms Ignore the fact that pages are hidden or read-protected 169 * @return array The pages that contain links to the given page 170 */ 171function ft_backlinks($id, $ignore_perms = false){ 172 $result = idx_get_indexer()->lookupKey('relation_references', $id); 173 174 if(!count($result)) return $result; 175 176 // check ACL permissions 177 foreach(array_keys($result) as $idx){ 178 if(($ignore_perms !== true && ( 179 isHiddenPage($result[$idx]) || auth_quickaclcheck($result[$idx]) < AUTH_READ 180 )) || !page_exists($result[$idx], '', false)){ 181 unset($result[$idx]); 182 } 183 } 184 185 Sort::sort($result); 186 return $result; 187} 188 189/** 190 * Returns the pages that use a given media file 191 * 192 * Uses the relation media metadata property and the metadata index. 193 * 194 * Note that before 2013-07-31 the second parameter was the maximum number of results and 195 * permissions were ignored. That's why the parameter is now checked to be explicitely set 196 * to true (with type bool) in order to be compatible with older uses of the function. 197 * 198 * @param string $id The media id to look for 199 * @param bool $ignore_perms Ignore hidden pages and acls (optional, default: false) 200 * @return array A list of pages that use the given media file 201 */ 202function ft_mediause($id, $ignore_perms = false){ 203 $result = idx_get_indexer()->lookupKey('relation_media', $id); 204 205 if(!count($result)) return $result; 206 207 // check ACL permissions 208 foreach(array_keys($result) as $idx){ 209 if(($ignore_perms !== true && ( 210 isHiddenPage($result[$idx]) || auth_quickaclcheck($result[$idx]) < AUTH_READ 211 )) || !page_exists($result[$idx], '', false)){ 212 unset($result[$idx]); 213 } 214 } 215 216 Sort::sort($result); 217 return $result; 218} 219 220 221/** 222 * Quicksearch for pagenames 223 * 224 * By default it only matches the pagename and ignores the 225 * namespace. This can be changed with the second parameter. 226 * The third parameter allows to search in titles as well. 227 * 228 * The function always returns titles as well 229 * 230 * @triggers SEARCH_QUERY_PAGELOOKUP 231 * @author Andreas Gohr <andi@splitbrain.org> 232 * @author Adrian Lang <lang@cosmocode.de> 233 * 234 * @param string $id page id 235 * @param bool $in_ns match against namespace as well? 236 * @param bool $in_title search in title? 237 * @param int|string $after only show results with mtime after this date, accepts timestap or strtotime arguments 238 * @param int|string $before only show results with mtime before this date, accepts timestap or strtotime arguments 239 * 240 * @return string[] 241 */ 242function ft_pageLookup($id, $in_ns=false, $in_title=false, $after = null, $before = null){ 243 $data = [ 244 'id' => $id, 245 'in_ns' => $in_ns, 246 'in_title' => $in_title, 247 'after' => $after, 248 'before' => $before 249 ]; 250 $data['has_titles'] = true; // for plugin backward compatibility check 251 return Event::createAndTrigger('SEARCH_QUERY_PAGELOOKUP', $data, '_ft_pageLookup'); 252} 253 254/** 255 * Returns list of pages as array(pageid => First Heading) 256 * 257 * @param array &$data event data 258 * @return string[] 259 */ 260function _ft_pageLookup(&$data){ 261 // split out original parameters 262 $id = $data['id']; 263 $Indexer = idx_get_indexer(); 264 $parsedQuery = ft_queryParser($Indexer, $id); 265 if (count($parsedQuery['ns']) > 0) { 266 $ns = cleanID($parsedQuery['ns'][0]) . ':'; 267 $id = implode(' ', $parsedQuery['highlight']); 268 } 269 270 $in_ns = $data['in_ns']; 271 $in_title = $data['in_title']; 272 $cleaned = cleanID($id); 273 274 $Indexer = idx_get_indexer(); 275 $page_idx = $Indexer->getPages(); 276 277 $pages = array(); 278 if ($id !== '' && $cleaned !== '') { 279 foreach ($page_idx as $p_id) { 280 if ((strpos($in_ns ? $p_id : noNSorNS($p_id), $cleaned) !== false)) { 281 if (!isset($pages[$p_id])) 282 $pages[$p_id] = p_get_first_heading($p_id, METADATA_DONT_RENDER); 283 } 284 } 285 if ($in_title) { 286 foreach ($Indexer->lookupKey('title', $id, '_ft_pageLookupTitleCompare') as $p_id) { 287 if (!isset($pages[$p_id])) 288 $pages[$p_id] = p_get_first_heading($p_id, METADATA_DONT_RENDER); 289 } 290 } 291 } 292 293 if (isset($ns)) { 294 foreach (array_keys($pages) as $p_id) { 295 if (strpos($p_id, $ns) !== 0) { 296 unset($pages[$p_id]); 297 } 298 } 299 } 300 301 // discard hidden pages 302 // discard nonexistent pages 303 // check ACL permissions 304 foreach(array_keys($pages) as $idx){ 305 if(!isVisiblePage($idx) || !page_exists($idx) || 306 auth_quickaclcheck($idx) < AUTH_READ) { 307 unset($pages[$idx]); 308 } 309 } 310 311 $pages = _ft_filterResultsByTime($pages, $data['after'], $data['before']); 312 313 uksort($pages,'ft_pagesorter'); 314 return $pages; 315} 316 317 318/** 319 * @param array $results search results in the form pageid => value 320 * @param int|string $after only returns results with mtime after this date, accepts timestap or strtotime arguments 321 * @param int|string $before only returns results with mtime after this date, accepts timestap or strtotime arguments 322 * 323 * @return array 324 */ 325function _ft_filterResultsByTime(array $results, $after, $before) { 326 if ($after || $before) { 327 $after = is_int($after) ? $after : strtotime($after); 328 $before = is_int($before) ? $before : strtotime($before); 329 330 foreach ($results as $id => $value) { 331 $mTime = filemtime(wikiFN($id)); 332 if ($after && $after > $mTime) { 333 unset($results[$id]); 334 continue; 335 } 336 if ($before && $before < $mTime) { 337 unset($results[$id]); 338 } 339 } 340 } 341 342 return $results; 343} 344 345/** 346 * Tiny helper function for comparing the searched title with the title 347 * from the search index. This function is a wrapper around stripos with 348 * adapted argument order and return value. 349 * 350 * @param string $search searched title 351 * @param string $title title from index 352 * @return bool 353 */ 354function _ft_pageLookupTitleCompare($search, $title) { 355 return stripos($title, $search) !== false; 356} 357 358/** 359 * Sort pages based on their namespace level first, then on their string 360 * values. This makes higher hierarchy pages rank higher than lower hierarchy 361 * pages. 362 * 363 * @param string $a 364 * @param string $b 365 * @return int Returns < 0 if $a is less than $b; > 0 if $a is greater than $b, and 0 if they are equal. 366 */ 367function ft_pagesorter($a, $b){ 368 $ac = count(explode(':',$a)); 369 $bc = count(explode(':',$b)); 370 if($ac < $bc){ 371 return -1; 372 }elseif($ac > $bc){ 373 return 1; 374 } 375 return Sort::strcmp($a,$b); 376} 377 378/** 379 * Sort pages by their mtime, from newest to oldest 380 * 381 * @param string $a 382 * @param string $b 383 * 384 * @return int Returns < 0 if $a is newer than $b, > 0 if $b is newer than $a and 0 if they are of the same age 385 */ 386function ft_pagemtimesorter($a, $b) { 387 $mtimeA = filemtime(wikiFN($a)); 388 $mtimeB = filemtime(wikiFN($b)); 389 return $mtimeB - $mtimeA; 390} 391 392/** 393 * Creates a snippet extract 394 * 395 * @author Andreas Gohr <andi@splitbrain.org> 396 * @triggers FULLTEXT_SNIPPET_CREATE 397 * 398 * @param string $id page id 399 * @param array $highlight 400 * @return mixed 401 */ 402function ft_snippet($id,$highlight){ 403 $text = rawWiki($id); 404 $text = str_replace("\xC2\xAD",'',$text); // remove soft-hyphens 405 $evdata = array( 406 'id' => $id, 407 'text' => &$text, 408 'highlight' => &$highlight, 409 'snippet' => '', 410 ); 411 412 $evt = new Event('FULLTEXT_SNIPPET_CREATE',$evdata); 413 if ($evt->advise_before()) { 414 $match = array(); 415 $snippets = array(); 416 $utf8_offset = $offset = $end = 0; 417 $len = \dokuwiki\Utf8\PhpString::strlen($text); 418 419 // build a regexp from the phrases to highlight 420 $re1 = '(' . 421 join( 422 '|', 423 array_map( 424 'ft_snippet_re_preprocess', 425 array_map( 426 'preg_quote_cb', 427 array_filter((array) $highlight) 428 ) 429 ) 430 ) . 431 ')'; 432 $re2 = "$re1.{0,75}(?!\\1)$re1"; 433 $re3 = "$re1.{0,45}(?!\\1)$re1.{0,45}(?!\\1)(?!\\2)$re1"; 434 435 for ($cnt=4; $cnt--;) { 436 if (0) { 437 } else if (preg_match('/'.$re3.'/iu',$text,$match,PREG_OFFSET_CAPTURE,$offset)) { 438 } else if (preg_match('/'.$re2.'/iu',$text,$match,PREG_OFFSET_CAPTURE,$offset)) { 439 } else if (preg_match('/'.$re1.'/iu',$text,$match,PREG_OFFSET_CAPTURE,$offset)) { 440 } else { 441 break; 442 } 443 444 list($str,$idx) = $match[0]; 445 446 // convert $idx (a byte offset) into a utf8 character offset 447 $utf8_idx = \dokuwiki\Utf8\PhpString::strlen(substr($text,0,$idx)); 448 $utf8_len = \dokuwiki\Utf8\PhpString::strlen($str); 449 450 // establish context, 100 bytes surrounding the match string 451 // first look to see if we can go 100 either side, 452 // then drop to 50 adding any excess if the other side can't go to 50, 453 $pre = min($utf8_idx-$utf8_offset,100); 454 $post = min($len-$utf8_idx-$utf8_len,100); 455 456 if ($pre>50 && $post>50) { 457 $pre = $post = 50; 458 } else if ($pre>50) { 459 $pre = min($pre,100-$post); 460 } else if ($post>50) { 461 $post = min($post, 100-$pre); 462 } else if ($offset == 0) { 463 // both are less than 50, means the context is the whole string 464 // make it so and break out of this loop - there is no need for the 465 // complex snippet calculations 466 $snippets = array($text); 467 break; 468 } 469 470 // establish context start and end points, try to append to previous 471 // context if possible 472 $start = $utf8_idx - $pre; 473 $append = ($start < $end) ? $end : false; // still the end of the previous context snippet 474 $end = $utf8_idx + $utf8_len + $post; // now set it to the end of this context 475 476 if ($append) { 477 $snippets[count($snippets)-1] .= \dokuwiki\Utf8\PhpString::substr($text,$append,$end-$append); 478 } else { 479 $snippets[] = \dokuwiki\Utf8\PhpString::substr($text,$start,$end-$start); 480 } 481 482 // set $offset for next match attempt 483 // continue matching after the current match 484 // if the current match is not the longest possible match starting at the current offset 485 // this prevents further matching of this snippet but for possible matches of length 486 // smaller than match length + context (at least 50 characters) this match is part of the context 487 $utf8_offset = $utf8_idx + $utf8_len; 488 $offset = $idx + strlen(\dokuwiki\Utf8\PhpString::substr($text,$utf8_idx,$utf8_len)); 489 $offset = \dokuwiki\Utf8\Clean::correctIdx($text,$offset); 490 } 491 492 $m = "\1"; 493 $snippets = preg_replace('/'.$re1.'/iu',$m.'$1'.$m,$snippets); 494 $snippet = preg_replace( 495 '/' . $m . '([^' . $m . ']*?)' . $m . '/iu', 496 '<strong class="search_hit">$1</strong>', 497 hsc(join('... ', $snippets)) 498 ); 499 500 $evdata['snippet'] = $snippet; 501 } 502 $evt->advise_after(); 503 unset($evt); 504 505 return $evdata['snippet']; 506} 507 508/** 509 * Wraps a search term in regex boundary checks. 510 * 511 * @param string $term 512 * @return string 513 */ 514function ft_snippet_re_preprocess($term) { 515 // do not process asian terms where word boundaries are not explicit 516 if(\dokuwiki\Utf8\Asian::isAsianWords($term)) return $term; 517 518 if (UTF8_PROPERTYSUPPORT) { 519 // unicode word boundaries 520 // see http://stackoverflow.com/a/2449017/172068 521 $BL = '(?<!\pL)'; 522 $BR = '(?!\pL)'; 523 } else { 524 // not as correct as above, but at least won't break 525 $BL = '\b'; 526 $BR = '\b'; 527 } 528 529 if(substr($term,0,2) == '\\*'){ 530 $term = substr($term,2); 531 }else{ 532 $term = $BL.$term; 533 } 534 535 if(substr($term,-2,2) == '\\*'){ 536 $term = substr($term,0,-2); 537 }else{ 538 $term = $term.$BR; 539 } 540 541 if($term == $BL || $term == $BR || $term == $BL.$BR) $term = ''; 542 return $term; 543} 544 545/** 546 * Combine found documents and sum up their scores 547 * 548 * This function is used to combine searched words with a logical 549 * AND. Only documents available in all arrays are returned. 550 * 551 * based upon PEAR's PHP_Compat function for array_intersect_key() 552 * 553 * @param array $args An array of page arrays 554 * @return array 555 */ 556function ft_resultCombine($args){ 557 $array_count = count($args); 558 if($array_count == 1){ 559 return $args[0]; 560 } 561 562 $result = array(); 563 if ($array_count > 1) { 564 foreach ($args[0] as $key => $value) { 565 $result[$key] = $value; 566 for ($i = 1; $i !== $array_count; $i++) { 567 if (!isset($args[$i][$key])) { 568 unset($result[$key]); 569 break; 570 } 571 $result[$key] += $args[$i][$key]; 572 } 573 } 574 } 575 return $result; 576} 577 578/** 579 * Unites found documents and sum up their scores 580 * 581 * based upon ft_resultCombine() function 582 * 583 * @param array $args An array of page arrays 584 * @return array 585 * 586 * @author Kazutaka Miyasaka <kazmiya@gmail.com> 587 */ 588function ft_resultUnite($args) { 589 $array_count = count($args); 590 if ($array_count === 1) { 591 return $args[0]; 592 } 593 594 $result = $args[0]; 595 for ($i = 1; $i !== $array_count; $i++) { 596 foreach (array_keys($args[$i]) as $id) { 597 $result[$id] += $args[$i][$id]; 598 } 599 } 600 return $result; 601} 602 603/** 604 * Computes the difference of documents using page id for comparison 605 * 606 * nearly identical to PHP5's array_diff_key() 607 * 608 * @param array $args An array of page arrays 609 * @return array 610 * 611 * @author Kazutaka Miyasaka <kazmiya@gmail.com> 612 */ 613function ft_resultComplement($args) { 614 $array_count = count($args); 615 if ($array_count === 1) { 616 return $args[0]; 617 } 618 619 $result = $args[0]; 620 foreach (array_keys($result) as $id) { 621 for ($i = 1; $i !== $array_count; $i++) { 622 if (isset($args[$i][$id])) unset($result[$id]); 623 } 624 } 625 return $result; 626} 627 628/** 629 * Parses a search query and builds an array of search formulas 630 * 631 * @author Andreas Gohr <andi@splitbrain.org> 632 * @author Kazutaka Miyasaka <kazmiya@gmail.com> 633 * 634 * @param dokuwiki\Search\Indexer $Indexer 635 * @param string $query search query 636 * @return array of search formulas 637 */ 638function ft_queryParser($Indexer, $query){ 639 /** 640 * parse a search query and transform it into intermediate representation 641 * 642 * in a search query, you can use the following expressions: 643 * 644 * words: 645 * include 646 * -exclude 647 * phrases: 648 * "phrase to be included" 649 * -"phrase you want to exclude" 650 * namespaces: 651 * @include:namespace (or ns:include:namespace) 652 * ^exclude:namespace (or -ns:exclude:namespace) 653 * groups: 654 * () 655 * -() 656 * operators: 657 * and ('and' is the default operator: you can always omit this) 658 * or (or pipe symbol '|', lower precedence than 'and') 659 * 660 * e.g. a query [ aa "bb cc" @dd:ee ] means "search pages which contain 661 * a word 'aa', a phrase 'bb cc' and are within a namespace 'dd:ee'". 662 * this query is equivalent to [ -(-aa or -"bb cc" or -ns:dd:ee) ] 663 * as long as you don't mind hit counts. 664 * 665 * intermediate representation consists of the following parts: 666 * 667 * ( ) - group 668 * AND - logical and 669 * OR - logical or 670 * NOT - logical not 671 * W+:, W-:, W_: - word (underscore: no need to highlight) 672 * P+:, P-: - phrase (minus sign: logically in NOT group) 673 * N+:, N-: - namespace 674 */ 675 $parsed_query = ''; 676 $parens_level = 0; 677 $terms = preg_split('/(-?".*?")/u', \dokuwiki\Utf8\PhpString::strtolower($query), 678 -1, PREG_SPLIT_DELIM_CAPTURE | PREG_SPLIT_NO_EMPTY); 679 680 foreach ($terms as $term) { 681 $parsed = ''; 682 if (preg_match('/^(-?)"(.+)"$/u', $term, $matches)) { 683 // phrase-include and phrase-exclude 684 $not = $matches[1] ? 'NOT' : ''; 685 $parsed = $not.ft_termParser($Indexer, $matches[2], false, true); 686 } else { 687 // fix incomplete phrase 688 $term = str_replace('"', ' ', $term); 689 690 // fix parentheses 691 $term = str_replace(')' , ' ) ', $term); 692 $term = str_replace('(' , ' ( ', $term); 693 $term = str_replace('- (', ' -(', $term); 694 695 // treat pipe symbols as 'OR' operators 696 $term = str_replace('|', ' or ', $term); 697 698 // treat ideographic spaces (U+3000) as search term separators 699 // FIXME: some more separators? 700 $term = preg_replace('/[ \x{3000}]+/u', ' ', $term); 701 $term = trim($term); 702 if ($term === '') continue; 703 704 $tokens = explode(' ', $term); 705 foreach ($tokens as $token) { 706 if ($token === '(') { 707 // parenthesis-include-open 708 $parsed .= '('; 709 ++$parens_level; 710 } elseif ($token === '-(') { 711 // parenthesis-exclude-open 712 $parsed .= 'NOT('; 713 ++$parens_level; 714 } elseif ($token === ')') { 715 // parenthesis-any-close 716 if ($parens_level === 0) continue; 717 $parsed .= ')'; 718 $parens_level--; 719 } elseif ($token === 'and') { 720 // logical-and (do nothing) 721 } elseif ($token === 'or') { 722 // logical-or 723 $parsed .= 'OR'; 724 } elseif (preg_match('/^(?:\^|-ns:)(.+)$/u', $token, $matches)) { 725 // namespace-exclude 726 $parsed .= 'NOT(N+:'.$matches[1].')'; 727 } elseif (preg_match('/^(?:@|ns:)(.+)$/u', $token, $matches)) { 728 // namespace-include 729 $parsed .= '(N+:'.$matches[1].')'; 730 } elseif (preg_match('/^-(.+)$/', $token, $matches)) { 731 // word-exclude 732 $parsed .= 'NOT('.ft_termParser($Indexer, $matches[1]).')'; 733 } else { 734 // word-include 735 $parsed .= ft_termParser($Indexer, $token); 736 } 737 } 738 } 739 $parsed_query .= $parsed; 740 } 741 742 // cleanup (very sensitive) 743 $parsed_query .= str_repeat(')', $parens_level); 744 do { 745 $parsed_query_old = $parsed_query; 746 $parsed_query = preg_replace('/(NOT)?\(\)/u', '', $parsed_query); 747 } while ($parsed_query !== $parsed_query_old); 748 $parsed_query = preg_replace('/(NOT|OR)+\)/u', ')' , $parsed_query); 749 $parsed_query = preg_replace('/(OR)+/u' , 'OR' , $parsed_query); 750 $parsed_query = preg_replace('/\(OR/u' , '(' , $parsed_query); 751 $parsed_query = preg_replace('/^OR|OR$/u' , '' , $parsed_query); 752 $parsed_query = preg_replace('/\)(NOT)?\(/u' , ')AND$1(', $parsed_query); 753 754 // adjustment: make highlightings right 755 $parens_level = 0; 756 $notgrp_levels = array(); 757 $parsed_query_new = ''; 758 $tokens = preg_split('/(NOT\(|[()])/u', $parsed_query, -1, PREG_SPLIT_DELIM_CAPTURE | PREG_SPLIT_NO_EMPTY); 759 foreach ($tokens as $token) { 760 if ($token === 'NOT(') { 761 $notgrp_levels[] = ++$parens_level; 762 } elseif ($token === '(') { 763 ++$parens_level; 764 } elseif ($token === ')') { 765 if ($parens_level-- === end($notgrp_levels)) array_pop($notgrp_levels); 766 } elseif (count($notgrp_levels) % 2 === 1) { 767 // turn highlight-flag off if terms are logically in "NOT" group 768 $token = preg_replace('/([WPN])\+\:/u', '$1-:', $token); 769 } 770 $parsed_query_new .= $token; 771 } 772 $parsed_query = $parsed_query_new; 773 774 /** 775 * convert infix notation string into postfix (Reverse Polish notation) array 776 * by Shunting-yard algorithm 777 * 778 * see: http://en.wikipedia.org/wiki/Reverse_Polish_notation 779 * see: http://en.wikipedia.org/wiki/Shunting-yard_algorithm 780 */ 781 $parsed_ary = array(); 782 $ope_stack = array(); 783 $ope_precedence = array(')' => 1, 'OR' => 2, 'AND' => 3, 'NOT' => 4, '(' => 5); 784 $ope_regex = '/([()]|OR|AND|NOT)/u'; 785 786 $tokens = preg_split($ope_regex, $parsed_query, -1, PREG_SPLIT_DELIM_CAPTURE | PREG_SPLIT_NO_EMPTY); 787 foreach ($tokens as $token) { 788 if (preg_match($ope_regex, $token)) { 789 // operator 790 $last_ope = end($ope_stack); 791 while ($last_ope !== false && $ope_precedence[$token] <= $ope_precedence[$last_ope] && $last_ope != '(') { 792 $parsed_ary[] = array_pop($ope_stack); 793 $last_ope = end($ope_stack); 794 } 795 if ($token == ')') { 796 array_pop($ope_stack); // this array_pop always deletes '(' 797 } else { 798 $ope_stack[] = $token; 799 } 800 } else { 801 // operand 802 $token_decoded = str_replace(array('OP', 'CP'), array('(', ')'), $token); 803 $parsed_ary[] = $token_decoded; 804 } 805 } 806 $parsed_ary = array_values(array_merge($parsed_ary, array_reverse($ope_stack))); 807 808 // cleanup: each double "NOT" in RPN array actually does nothing 809 $parsed_ary_count = count($parsed_ary); 810 for ($i = 1; $i < $parsed_ary_count; ++$i) { 811 if ($parsed_ary[$i] === 'NOT' && $parsed_ary[$i - 1] === 'NOT') { 812 unset($parsed_ary[$i], $parsed_ary[$i - 1]); 813 } 814 } 815 $parsed_ary = array_values($parsed_ary); 816 817 // build return value 818 $q = array(); 819 $q['query'] = $query; 820 $q['parsed_str'] = $parsed_query; 821 $q['parsed_ary'] = $parsed_ary; 822 823 foreach ($q['parsed_ary'] as $token) { 824 if ($token[2] !== ':') continue; 825 $body = substr($token, 3); 826 827 switch (substr($token, 0, 3)) { 828 case 'N+:': 829 $q['ns'][] = $body; // for backward compatibility 830 break; 831 case 'N-:': 832 $q['notns'][] = $body; // for backward compatibility 833 break; 834 case 'W_:': 835 $q['words'][] = $body; 836 break; 837 case 'W-:': 838 $q['words'][] = $body; 839 $q['not'][] = $body; // for backward compatibility 840 break; 841 case 'W+:': 842 $q['words'][] = $body; 843 $q['highlight'][] = $body; 844 $q['and'][] = $body; // for backward compatibility 845 break; 846 case 'P-:': 847 $q['phrases'][] = $body; 848 break; 849 case 'P+:': 850 $q['phrases'][] = $body; 851 $q['highlight'][] = $body; 852 break; 853 } 854 } 855 foreach (array('words', 'phrases', 'highlight', 'ns', 'notns', 'and', 'not') as $key) { 856 $q[$key] = empty($q[$key]) ? array() : array_values(array_unique($q[$key])); 857 } 858 859 return $q; 860} 861 862/** 863 * Transforms given search term into intermediate representation 864 * 865 * This function is used in ft_queryParser() and not for general purpose use. 866 * 867 * @author Kazutaka Miyasaka <kazmiya@gmail.com> 868 * 869 * @param dokuwiki\Search\Indexer $Indexer 870 * @param string $term 871 * @param bool $consider_asian 872 * @param bool $phrase_mode 873 * @return string 874 */ 875function ft_termParser($Indexer, $term, $consider_asian = true, $phrase_mode = false) { 876 $parsed = ''; 877 if ($consider_asian) { 878 // successive asian characters need to be searched as a phrase 879 $words = \dokuwiki\Utf8\Asian::splitAsianWords($term); 880 foreach ($words as $word) { 881 $phrase_mode = $phrase_mode ? true : \dokuwiki\Utf8\Asian::isAsianWords($word); 882 $parsed .= ft_termParser($Indexer, $word, false, $phrase_mode); 883 } 884 } else { 885 $term_noparen = str_replace(array('(', ')'), ' ', $term); 886 $words = $Indexer->tokenizer($term_noparen, true); 887 888 // W_: no need to highlight 889 if (empty($words)) { 890 $parsed = '()'; // important: do not remove 891 } elseif ($words[0] === $term) { 892 $parsed = '(W+:'.$words[0].')'; 893 } elseif ($phrase_mode) { 894 $term_encoded = str_replace(array('(', ')'), array('OP', 'CP'), $term); 895 $parsed = '((W_:'.implode(')(W_:', $words).')(P+:'.$term_encoded.'))'; 896 } else { 897 $parsed = '((W+:'.implode(')(W+:', $words).'))'; 898 } 899 } 900 return $parsed; 901} 902 903/** 904 * Recreate a search query string based on parsed parts, doesn't support negated phrases and `OR` searches 905 * 906 * @param array $and 907 * @param array $not 908 * @param array $phrases 909 * @param array $ns 910 * @param array $notns 911 * 912 * @return string 913 */ 914function ft_queryUnparser_simple(array $and, array $not, array $phrases, array $ns, array $notns) { 915 $query = implode(' ', $and); 916 if (!empty($not)) { 917 $query .= ' -' . implode(' -', $not); 918 } 919 920 if (!empty($phrases)) { 921 $query .= ' "' . implode('" "', $phrases) . '"'; 922 } 923 924 if (!empty($ns)) { 925 $query .= ' @' . implode(' @', $ns); 926 } 927 928 if (!empty($notns)) { 929 $query .= ' ^' . implode(' ^', $notns); 930 } 931 932 return $query; 933} 934 935//Setup VIM: ex: et ts=4 : 936