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