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 9if(!defined('DOKU_INC')) die('meh.'); 10 11/** 12 * create snippets for the first few results only 13 */ 14if(!defined('FT_SNIPPET_NUMBER')) define('FT_SNIPPET_NUMBER',15); 15 16/** 17 * The fulltext search 18 * 19 * Returns a list of matching documents for the given query 20 * 21 * refactored into ft_pageSearch(), _ft_pageSearch() and trigger_event() 22 * 23 * @param string $query 24 * @param array $highlight 25 * @param string $sort 26 * @param int|string $after only show results with an modified time after this date, accepts timestap or strtotime arguments 27 * @param int|string $before only show results with an modified time before this date, accepts timestap or strtotime arguments 28 * 29 * @return array 30 */ 31function ft_pageSearch($query,&$highlight, $sort = null, $after = null, $before = null){ 32 33 if ($sort === null) { 34 $sort = 'hits'; 35 } 36 $data = [ 37 'query' => $query, 38 'sort' => $sort, 39 'after' => $after, 40 'before' => $before 41 ]; 42 $data['highlight'] =& $highlight; 43 44 return trigger_event('SEARCH_QUERY_FULLPAGE', $data, '_ft_pageSearch'); 45} 46 47/** 48 * Returns a list of matching documents for the given query 49 * 50 * @author Andreas Gohr <andi@splitbrain.org> 51 * @author Kazutaka Miyasaka <kazmiya@gmail.com> 52 * 53 * @param array $data event data 54 * @return array matching documents 55 */ 56function _ft_pageSearch(&$data) { 57 $Indexer = idx_get_indexer(); 58 59 // parse the given query 60 $q = ft_queryParser($Indexer, $data['query']); 61 $data['highlight'] = $q['highlight']; 62 63 if (empty($q['parsed_ary'])) return array(); 64 65 // lookup all words found in the query 66 $lookup = $Indexer->lookup($q['words']); 67 68 // get all pages in this dokuwiki site (!: includes nonexistent pages) 69 $pages_all = array(); 70 foreach ($Indexer->getPages() as $id) { 71 $pages_all[$id] = 0; // base: 0 hit 72 } 73 74 // process the query 75 $stack = array(); 76 foreach ($q['parsed_ary'] as $token) { 77 switch (substr($token, 0, 3)) { 78 case 'W+:': 79 case 'W-:': 80 case 'W_:': // word 81 $word = substr($token, 3); 82 $stack[] = (array) $lookup[$word]; 83 break; 84 case 'P+:': 85 case 'P-:': // phrase 86 $phrase = substr($token, 3); 87 // since phrases are always parsed as ((W1)(W2)...(P)), 88 // the end($stack) always points the pages that contain 89 // all words in this phrase 90 $pages = end($stack); 91 $pages_matched = array(); 92 foreach(array_keys($pages) as $id){ 93 $evdata = array( 94 'id' => $id, 95 'phrase' => $phrase, 96 'text' => rawWiki($id) 97 ); 98 $evt = new Doku_Event('FULLTEXT_PHRASE_MATCH',$evdata); 99 if ($evt->advise_before() && $evt->result !== true) { 100 $text = utf8_strtolower($evdata['text']); 101 if (strpos($text, $phrase) !== false) { 102 $evt->result = true; 103 } 104 } 105 $evt->advise_after(); 106 if ($evt->result === true) { 107 $pages_matched[$id] = 0; // phrase: always 0 hit 108 } 109 } 110 $stack[] = $pages_matched; 111 break; 112 case 'N+:': 113 case 'N-:': // namespace 114 $ns = cleanID(substr($token, 3)) . ':'; 115 $pages_matched = array(); 116 foreach (array_keys($pages_all) as $id) { 117 if (strpos($id, $ns) === 0) { 118 $pages_matched[$id] = 0; // namespace: always 0 hit 119 } 120 } 121 $stack[] = $pages_matched; 122 break; 123 case 'AND': // and operation 124 list($pages1, $pages2) = array_splice($stack, -2); 125 $stack[] = ft_resultCombine(array($pages1, $pages2)); 126 break; 127 case 'OR': // or operation 128 list($pages1, $pages2) = array_splice($stack, -2); 129 $stack[] = ft_resultUnite(array($pages1, $pages2)); 130 break; 131 case 'NOT': // not operation (unary) 132 $pages = array_pop($stack); 133 $stack[] = ft_resultComplement(array($pages_all, $pages)); 134 break; 135 } 136 } 137 $docs = array_pop($stack); 138 139 if (empty($docs)) return array(); 140 141 // check: settings, acls, existence 142 foreach (array_keys($docs) as $id) { 143 if (isHiddenPage($id) || auth_quickaclcheck($id) < AUTH_READ || !page_exists($id, '', false)) { 144 unset($docs[$id]); 145 } 146 } 147 148 $docs = _ft_filterResultsByTime($docs, $data['after'], $data['before']); 149 150 if ($data['sort'] === 'mtime') { 151 uksort($docs, 'ft_pagemtimesorter'); 152 } else { 153 // sort docs by count 154 arsort($docs); 155 } 156 157 return $docs; 158} 159 160/** 161 * Returns the backlinks for a given page 162 * 163 * Uses the metadata index. 164 * 165 * @param string $id The id for which links shall be returned 166 * @param bool $ignore_perms Ignore the fact that pages are hidden or read-protected 167 * @return array The pages that contain links to the given page 168 */ 169function ft_backlinks($id, $ignore_perms = false){ 170 $result = idx_get_indexer()->lookupKey('relation_references', $id); 171 172 if(!count($result)) return $result; 173 174 // check ACL permissions 175 foreach(array_keys($result) as $idx){ 176 if(($ignore_perms !== true && ( 177 isHiddenPage($result[$idx]) || auth_quickaclcheck($result[$idx]) < AUTH_READ 178 )) || !page_exists($result[$idx], '', false)){ 179 unset($result[$idx]); 180 } 181 } 182 183 sort($result); 184 return $result; 185} 186 187/** 188 * Returns the pages that use a given media file 189 * 190 * Uses the relation media metadata property and the metadata index. 191 * 192 * Note that before 2013-07-31 the second parameter was the maximum number of results and 193 * permissions were ignored. That's why the parameter is now checked to be explicitely set 194 * to true (with type bool) in order to be compatible with older uses of the function. 195 * 196 * @param string $id The media id to look for 197 * @param bool $ignore_perms Ignore hidden pages and acls (optional, default: false) 198 * @return array A list of pages that use the given media file 199 */ 200function ft_mediause($id, $ignore_perms = false){ 201 $result = idx_get_indexer()->lookupKey('relation_media', $id); 202 203 if(!count($result)) return $result; 204 205 // check ACL permissions 206 foreach(array_keys($result) as $idx){ 207 if(($ignore_perms !== true && ( 208 isHiddenPage($result[$idx]) || auth_quickaclcheck($result[$idx]) < AUTH_READ 209 )) || !page_exists($result[$idx], '', false)){ 210 unset($result[$idx]); 211 } 212 } 213 214 sort($result); 215 return $result; 216} 217 218 219/** 220 * Quicksearch for pagenames 221 * 222 * By default it only matches the pagename and ignores the 223 * namespace. This can be changed with the second parameter. 224 * The third parameter allows to search in titles as well. 225 * 226 * The function always returns titles as well 227 * 228 * @triggers SEARCH_QUERY_PAGELOOKUP 229 * @author Andreas Gohr <andi@splitbrain.org> 230 * @author Adrian Lang <lang@cosmocode.de> 231 * 232 * @param string $id page id 233 * @param bool $in_ns match against namespace as well? 234 * @param bool $in_title search in title? 235 * @param int|string $after only show results with an modified time after this date, accepts timestap or strtotime arguments 236 * @param int|string $before only show results with an modified time before this date, accepts timestap or strtotime arguments 237 * 238 * @return string[] 239 */ 240function ft_pageLookup($id, $in_ns=false, $in_title=false, $after = null, $before = null){ 241 $data = [ 242 'id' => $id, 243 'in_ns' => $in_ns, 244 'in_title' => $in_title, 245 'after' => $after, 246 'before' => $before 247 ]; 248 $data['has_titles'] = true; // for plugin backward compatibility check 249 return trigger_event('SEARCH_QUERY_PAGELOOKUP', $data, '_ft_pageLookup'); 250} 251 252/** 253 * Returns list of pages as array(pageid => First Heading) 254 * 255 * @param array &$data event data 256 * @return string[] 257 */ 258function _ft_pageLookup(&$data){ 259 // split out original parameters 260 $id = $data['id']; 261 $Indexer = idx_get_indexer(); 262 $parsedQuery = ft_queryParser($Indexer, $id); 263 if (count($parsedQuery['ns']) > 0) { 264 $ns = cleanID($parsedQuery['ns'][0]) . ':'; 265 $id = implode(' ', $parsedQuery['highlight']); 266 } 267 268 $in_ns = $data['in_ns']; 269 $in_title = $data['in_title']; 270 $cleaned = cleanID($id); 271 272 $Indexer = idx_get_indexer(); 273 $page_idx = $Indexer->getPages(); 274 275 $pages = array(); 276 if ($id !== '' && $cleaned !== '') { 277 foreach ($page_idx as $p_id) { 278 if ((strpos($in_ns ? $p_id : noNSorNS($p_id), $cleaned) !== false)) { 279 if (!isset($pages[$p_id])) 280 $pages[$p_id] = p_get_first_heading($p_id, METADATA_DONT_RENDER); 281 } 282 } 283 if ($in_title) { 284 foreach ($Indexer->lookupKey('title', $id, '_ft_pageLookupTitleCompare') as $p_id) { 285 if (!isset($pages[$p_id])) 286 $pages[$p_id] = p_get_first_heading($p_id, METADATA_DONT_RENDER); 287 } 288 } 289 } 290 291 if (isset($ns)) { 292 foreach (array_keys($pages) as $p_id) { 293 if (strpos($p_id, $ns) !== 0) { 294 unset($pages[$p_id]); 295 } 296 } 297 } 298 299 // discard hidden pages 300 // discard nonexistent pages 301 // check ACL permissions 302 foreach(array_keys($pages) as $idx){ 303 if(!isVisiblePage($idx) || !page_exists($idx) || 304 auth_quickaclcheck($idx) < AUTH_READ) { 305 unset($pages[$idx]); 306 } 307 } 308 309 $pages = _ft_filterResultsByTime($pages, $data['after'], $data['before']); 310 311 uksort($pages,'ft_pagesorter'); 312 return $pages; 313} 314 315 316/** 317 * @param array $results search results in the form pageid => value 318 * @param int|string $after only returns results with an modified time after this date, accepts timestap or strtotime arguments 319 * @param int|string $before only returns results with an modified time after this date, accepts timestap or strtotime arguments 320 * 321 * @return array 322 */ 323function _ft_filterResultsByTime(array $results, $after, $before) { 324 if ($after || $before) { 325 $after = is_int($after) ? $after : strtotime($after); 326 $before = is_int($before) ? $before : strtotime($before); 327 328 foreach ($results as $id => $value) { 329 $mTime = filemtime(wikiFN($id)); 330 if ($after && $after > $mTime) { 331 unset($results[$id]); 332 continue; 333 } 334 if ($before && $before < $mTime) { 335 unset($results[$id]); 336 } 337 } 338 } 339 340 return $results; 341} 342 343/** 344 * Tiny helper function for comparing the searched title with the title 345 * from the search index. This function is a wrapper around stripos with 346 * adapted argument order and return value. 347 * 348 * @param string $search searched title 349 * @param string $title title from index 350 * @return bool 351 */ 352function _ft_pageLookupTitleCompare($search, $title) { 353 return stripos($title, $search) !== false; 354} 355 356/** 357 * Sort pages based on their namespace level first, then on their string 358 * values. This makes higher hierarchy pages rank higher than lower hierarchy 359 * pages. 360 * 361 * @param string $a 362 * @param string $b 363 * @return int Returns < 0 if $a is less than $b; > 0 if $a is greater than $b, and 0 if they are equal. 364 */ 365function ft_pagesorter($a, $b){ 366 $ac = count(explode(':',$a)); 367 $bc = count(explode(':',$b)); 368 if($ac < $bc){ 369 return -1; 370 }elseif($ac > $bc){ 371 return 1; 372 } 373 return strcmp ($a,$b); 374} 375 376/** 377 * Sort pages by their mtime, from newest to oldest 378 * 379 * @param string $a 380 * @param string $b 381 * 382 * @return int Returns < 0 if $a is newer than $b, > 0 if $b is newer than $a and 0 if they are of the same age 383 */ 384function ft_pagemtimesorter($a, $b) { 385 $mtimeA = filemtime(wikiFN($a)); 386 $mtimeB = filemtime(wikiFN($b)); 387 return $mtimeB - $mtimeA; 388} 389 390/** 391 * Creates a snippet extract 392 * 393 * @author Andreas Gohr <andi@splitbrain.org> 394 * @triggers FULLTEXT_SNIPPET_CREATE 395 * 396 * @param string $id page id 397 * @param array $highlight 398 * @return mixed 399 */ 400function ft_snippet($id,$highlight){ 401 $text = rawWiki($id); 402 $text = str_replace("\xC2\xAD",'',$text); // remove soft-hyphens 403 $evdata = array( 404 'id' => $id, 405 'text' => &$text, 406 'highlight' => &$highlight, 407 'snippet' => '', 408 ); 409 410 $evt = new Doku_Event('FULLTEXT_SNIPPET_CREATE',$evdata); 411 if ($evt->advise_before()) { 412 $match = array(); 413 $snippets = array(); 414 $utf8_offset = $offset = $end = 0; 415 $len = utf8_strlen($text); 416 417 // build a regexp from the phrases to highlight 418 $re1 = '('.join('|',array_map('ft_snippet_re_preprocess', array_map('preg_quote_cb',array_filter((array) $highlight)))).')'; 419 $re2 = "$re1.{0,75}(?!\\1)$re1"; 420 $re3 = "$re1.{0,45}(?!\\1)$re1.{0,45}(?!\\1)(?!\\2)$re1"; 421 422 for ($cnt=4; $cnt--;) { 423 if (0) { 424 } else if (preg_match('/'.$re3.'/iu',$text,$match,PREG_OFFSET_CAPTURE,$offset)) { 425 } else if (preg_match('/'.$re2.'/iu',$text,$match,PREG_OFFSET_CAPTURE,$offset)) { 426 } else if (preg_match('/'.$re1.'/iu',$text,$match,PREG_OFFSET_CAPTURE,$offset)) { 427 } else { 428 break; 429 } 430 431 list($str,$idx) = $match[0]; 432 433 // convert $idx (a byte offset) into a utf8 character offset 434 $utf8_idx = utf8_strlen(substr($text,0,$idx)); 435 $utf8_len = utf8_strlen($str); 436 437 // establish context, 100 bytes surrounding the match string 438 // first look to see if we can go 100 either side, 439 // then drop to 50 adding any excess if the other side can't go to 50, 440 $pre = min($utf8_idx-$utf8_offset,100); 441 $post = min($len-$utf8_idx-$utf8_len,100); 442 443 if ($pre>50 && $post>50) { 444 $pre = $post = 50; 445 } else if ($pre>50) { 446 $pre = min($pre,100-$post); 447 } else if ($post>50) { 448 $post = min($post, 100-$pre); 449 } else if ($offset == 0) { 450 // both are less than 50, means the context is the whole string 451 // make it so and break out of this loop - there is no need for the 452 // complex snippet calculations 453 $snippets = array($text); 454 break; 455 } 456 457 // establish context start and end points, try to append to previous 458 // context if possible 459 $start = $utf8_idx - $pre; 460 $append = ($start < $end) ? $end : false; // still the end of the previous context snippet 461 $end = $utf8_idx + $utf8_len + $post; // now set it to the end of this context 462 463 if ($append) { 464 $snippets[count($snippets)-1] .= utf8_substr($text,$append,$end-$append); 465 } else { 466 $snippets[] = utf8_substr($text,$start,$end-$start); 467 } 468 469 // set $offset for next match attempt 470 // continue matching after the current match 471 // if the current match is not the longest possible match starting at the current offset 472 // this prevents further matching of this snippet but for possible matches of length 473 // smaller than match length + context (at least 50 characters) this match is part of the context 474 $utf8_offset = $utf8_idx + $utf8_len; 475 $offset = $idx + strlen(utf8_substr($text,$utf8_idx,$utf8_len)); 476 $offset = utf8_correctIdx($text,$offset); 477 } 478 479 $m = "\1"; 480 $snippets = preg_replace('/'.$re1.'/iu',$m.'$1'.$m,$snippets); 481 $snippet = preg_replace('/'.$m.'([^'.$m.']*?)'.$m.'/iu','<strong class="search_hit">$1</strong>',hsc(join('... ',$snippets))); 482 483 $evdata['snippet'] = $snippet; 484 } 485 $evt->advise_after(); 486 unset($evt); 487 488 return $evdata['snippet']; 489} 490 491/** 492 * Wraps a search term in regex boundary checks. 493 * 494 * @param string $term 495 * @return string 496 */ 497function ft_snippet_re_preprocess($term) { 498 // do not process asian terms where word boundaries are not explicit 499 if(preg_match('/'.IDX_ASIAN.'/u',$term)){ 500 return $term; 501 } 502 503 if (UTF8_PROPERTYSUPPORT) { 504 // unicode word boundaries 505 // see http://stackoverflow.com/a/2449017/172068 506 $BL = '(?<!\pL)'; 507 $BR = '(?!\pL)'; 508 } else { 509 // not as correct as above, but at least won't break 510 $BL = '\b'; 511 $BR = '\b'; 512 } 513 514 if(substr($term,0,2) == '\\*'){ 515 $term = substr($term,2); 516 }else{ 517 $term = $BL.$term; 518 } 519 520 if(substr($term,-2,2) == '\\*'){ 521 $term = substr($term,0,-2); 522 }else{ 523 $term = $term.$BR; 524 } 525 526 if($term == $BL || $term == $BR || $term == $BL.$BR) $term = ''; 527 return $term; 528} 529 530/** 531 * Combine found documents and sum up their scores 532 * 533 * This function is used to combine searched words with a logical 534 * AND. Only documents available in all arrays are returned. 535 * 536 * based upon PEAR's PHP_Compat function for array_intersect_key() 537 * 538 * @param array $args An array of page arrays 539 * @return array 540 */ 541function ft_resultCombine($args){ 542 $array_count = count($args); 543 if($array_count == 1){ 544 return $args[0]; 545 } 546 547 $result = array(); 548 if ($array_count > 1) { 549 foreach ($args[0] as $key => $value) { 550 $result[$key] = $value; 551 for ($i = 1; $i !== $array_count; $i++) { 552 if (!isset($args[$i][$key])) { 553 unset($result[$key]); 554 break; 555 } 556 $result[$key] += $args[$i][$key]; 557 } 558 } 559 } 560 return $result; 561} 562 563/** 564 * Unites found documents and sum up their scores 565 * 566 * based upon ft_resultCombine() function 567 * 568 * @param array $args An array of page arrays 569 * @return array 570 * 571 * @author Kazutaka Miyasaka <kazmiya@gmail.com> 572 */ 573function ft_resultUnite($args) { 574 $array_count = count($args); 575 if ($array_count === 1) { 576 return $args[0]; 577 } 578 579 $result = $args[0]; 580 for ($i = 1; $i !== $array_count; $i++) { 581 foreach (array_keys($args[$i]) as $id) { 582 $result[$id] += $args[$i][$id]; 583 } 584 } 585 return $result; 586} 587 588/** 589 * Computes the difference of documents using page id for comparison 590 * 591 * nearly identical to PHP5's array_diff_key() 592 * 593 * @param array $args An array of page arrays 594 * @return array 595 * 596 * @author Kazutaka Miyasaka <kazmiya@gmail.com> 597 */ 598function ft_resultComplement($args) { 599 $array_count = count($args); 600 if ($array_count === 1) { 601 return $args[0]; 602 } 603 604 $result = $args[0]; 605 foreach (array_keys($result) as $id) { 606 for ($i = 1; $i !== $array_count; $i++) { 607 if (isset($args[$i][$id])) unset($result[$id]); 608 } 609 } 610 return $result; 611} 612 613/** 614 * Parses a search query and builds an array of search formulas 615 * 616 * @author Andreas Gohr <andi@splitbrain.org> 617 * @author Kazutaka Miyasaka <kazmiya@gmail.com> 618 * 619 * @param Doku_Indexer $Indexer 620 * @param string $query search query 621 * @return array of search formulas 622 */ 623function ft_queryParser($Indexer, $query){ 624 /** 625 * parse a search query and transform it into intermediate representation 626 * 627 * in a search query, you can use the following expressions: 628 * 629 * words: 630 * include 631 * -exclude 632 * phrases: 633 * "phrase to be included" 634 * -"phrase you want to exclude" 635 * namespaces: 636 * @include:namespace (or ns:include:namespace) 637 * ^exclude:namespace (or -ns:exclude:namespace) 638 * groups: 639 * () 640 * -() 641 * operators: 642 * and ('and' is the default operator: you can always omit this) 643 * or (or pipe symbol '|', lower precedence than 'and') 644 * 645 * e.g. a query [ aa "bb cc" @dd:ee ] means "search pages which contain 646 * a word 'aa', a phrase 'bb cc' and are within a namespace 'dd:ee'". 647 * this query is equivalent to [ -(-aa or -"bb cc" or -ns:dd:ee) ] 648 * as long as you don't mind hit counts. 649 * 650 * intermediate representation consists of the following parts: 651 * 652 * ( ) - group 653 * AND - logical and 654 * OR - logical or 655 * NOT - logical not 656 * W+:, W-:, W_: - word (underscore: no need to highlight) 657 * P+:, P-: - phrase (minus sign: logically in NOT group) 658 * N+:, N-: - namespace 659 */ 660 $parsed_query = ''; 661 $parens_level = 0; 662 $terms = preg_split('/(-?".*?")/u', utf8_strtolower($query), -1, PREG_SPLIT_DELIM_CAPTURE | PREG_SPLIT_NO_EMPTY); 663 664 foreach ($terms as $term) { 665 $parsed = ''; 666 if (preg_match('/^(-?)"(.+)"$/u', $term, $matches)) { 667 // phrase-include and phrase-exclude 668 $not = $matches[1] ? 'NOT' : ''; 669 $parsed = $not.ft_termParser($Indexer, $matches[2], false, true); 670 } else { 671 // fix incomplete phrase 672 $term = str_replace('"', ' ', $term); 673 674 // fix parentheses 675 $term = str_replace(')' , ' ) ', $term); 676 $term = str_replace('(' , ' ( ', $term); 677 $term = str_replace('- (', ' -(', $term); 678 679 // treat pipe symbols as 'OR' operators 680 $term = str_replace('|', ' or ', $term); 681 682 // treat ideographic spaces (U+3000) as search term separators 683 // FIXME: some more separators? 684 $term = preg_replace('/[ \x{3000}]+/u', ' ', $term); 685 $term = trim($term); 686 if ($term === '') continue; 687 688 $tokens = explode(' ', $term); 689 foreach ($tokens as $token) { 690 if ($token === '(') { 691 // parenthesis-include-open 692 $parsed .= '('; 693 ++$parens_level; 694 } elseif ($token === '-(') { 695 // parenthesis-exclude-open 696 $parsed .= 'NOT('; 697 ++$parens_level; 698 } elseif ($token === ')') { 699 // parenthesis-any-close 700 if ($parens_level === 0) continue; 701 $parsed .= ')'; 702 $parens_level--; 703 } elseif ($token === 'and') { 704 // logical-and (do nothing) 705 } elseif ($token === 'or') { 706 // logical-or 707 $parsed .= 'OR'; 708 } elseif (preg_match('/^(?:\^|-ns:)(.+)$/u', $token, $matches)) { 709 // namespace-exclude 710 $parsed .= 'NOT(N+:'.$matches[1].')'; 711 } elseif (preg_match('/^(?:@|ns:)(.+)$/u', $token, $matches)) { 712 // namespace-include 713 $parsed .= '(N+:'.$matches[1].')'; 714 } elseif (preg_match('/^-(.+)$/', $token, $matches)) { 715 // word-exclude 716 $parsed .= 'NOT('.ft_termParser($Indexer, $matches[1]).')'; 717 } else { 718 // word-include 719 $parsed .= ft_termParser($Indexer, $token); 720 } 721 } 722 } 723 $parsed_query .= $parsed; 724 } 725 726 // cleanup (very sensitive) 727 $parsed_query .= str_repeat(')', $parens_level); 728 do { 729 $parsed_query_old = $parsed_query; 730 $parsed_query = preg_replace('/(NOT)?\(\)/u', '', $parsed_query); 731 } while ($parsed_query !== $parsed_query_old); 732 $parsed_query = preg_replace('/(NOT|OR)+\)/u', ')' , $parsed_query); 733 $parsed_query = preg_replace('/(OR)+/u' , 'OR' , $parsed_query); 734 $parsed_query = preg_replace('/\(OR/u' , '(' , $parsed_query); 735 $parsed_query = preg_replace('/^OR|OR$/u' , '' , $parsed_query); 736 $parsed_query = preg_replace('/\)(NOT)?\(/u' , ')AND$1(', $parsed_query); 737 738 // adjustment: make highlightings right 739 $parens_level = 0; 740 $notgrp_levels = array(); 741 $parsed_query_new = ''; 742 $tokens = preg_split('/(NOT\(|[()])/u', $parsed_query, -1, PREG_SPLIT_DELIM_CAPTURE | PREG_SPLIT_NO_EMPTY); 743 foreach ($tokens as $token) { 744 if ($token === 'NOT(') { 745 $notgrp_levels[] = ++$parens_level; 746 } elseif ($token === '(') { 747 ++$parens_level; 748 } elseif ($token === ')') { 749 if ($parens_level-- === end($notgrp_levels)) array_pop($notgrp_levels); 750 } elseif (count($notgrp_levels) % 2 === 1) { 751 // turn highlight-flag off if terms are logically in "NOT" group 752 $token = preg_replace('/([WPN])\+\:/u', '$1-:', $token); 753 } 754 $parsed_query_new .= $token; 755 } 756 $parsed_query = $parsed_query_new; 757 758 /** 759 * convert infix notation string into postfix (Reverse Polish notation) array 760 * by Shunting-yard algorithm 761 * 762 * see: http://en.wikipedia.org/wiki/Reverse_Polish_notation 763 * see: http://en.wikipedia.org/wiki/Shunting-yard_algorithm 764 */ 765 $parsed_ary = array(); 766 $ope_stack = array(); 767 $ope_precedence = array(')' => 1, 'OR' => 2, 'AND' => 3, 'NOT' => 4, '(' => 5); 768 $ope_regex = '/([()]|OR|AND|NOT)/u'; 769 770 $tokens = preg_split($ope_regex, $parsed_query, -1, PREG_SPLIT_DELIM_CAPTURE | PREG_SPLIT_NO_EMPTY); 771 foreach ($tokens as $token) { 772 if (preg_match($ope_regex, $token)) { 773 // operator 774 $last_ope = end($ope_stack); 775 while ($last_ope !== false && $ope_precedence[$token] <= $ope_precedence[$last_ope] && $last_ope != '(') { 776 $parsed_ary[] = array_pop($ope_stack); 777 $last_ope = end($ope_stack); 778 } 779 if ($token == ')') { 780 array_pop($ope_stack); // this array_pop always deletes '(' 781 } else { 782 $ope_stack[] = $token; 783 } 784 } else { 785 // operand 786 $token_decoded = str_replace(array('OP', 'CP'), array('(', ')'), $token); 787 $parsed_ary[] = $token_decoded; 788 } 789 } 790 $parsed_ary = array_values(array_merge($parsed_ary, array_reverse($ope_stack))); 791 792 // cleanup: each double "NOT" in RPN array actually does nothing 793 $parsed_ary_count = count($parsed_ary); 794 for ($i = 1; $i < $parsed_ary_count; ++$i) { 795 if ($parsed_ary[$i] === 'NOT' && $parsed_ary[$i - 1] === 'NOT') { 796 unset($parsed_ary[$i], $parsed_ary[$i - 1]); 797 } 798 } 799 $parsed_ary = array_values($parsed_ary); 800 801 // build return value 802 $q = array(); 803 $q['query'] = $query; 804 $q['parsed_str'] = $parsed_query; 805 $q['parsed_ary'] = $parsed_ary; 806 807 foreach ($q['parsed_ary'] as $token) { 808 if ($token[2] !== ':') continue; 809 $body = substr($token, 3); 810 811 switch (substr($token, 0, 3)) { 812 case 'N+:': 813 $q['ns'][] = $body; // for backward compatibility 814 break; 815 case 'N-:': 816 $q['notns'][] = $body; // for backward compatibility 817 break; 818 case 'W_:': 819 $q['words'][] = $body; 820 break; 821 case 'W-:': 822 $q['words'][] = $body; 823 $q['not'][] = $body; // for backward compatibility 824 break; 825 case 'W+:': 826 $q['words'][] = $body; 827 $q['highlight'][] = $body; 828 $q['and'][] = $body; // for backward compatibility 829 break; 830 case 'P-:': 831 $q['phrases'][] = $body; 832 break; 833 case 'P+:': 834 $q['phrases'][] = $body; 835 $q['highlight'][] = $body; 836 break; 837 } 838 } 839 foreach (array('words', 'phrases', 'highlight', 'ns', 'notns', 'and', 'not') as $key) { 840 $q[$key] = empty($q[$key]) ? array() : array_values(array_unique($q[$key])); 841 } 842 843 return $q; 844} 845 846/** 847 * Transforms given search term into intermediate representation 848 * 849 * This function is used in ft_queryParser() and not for general purpose use. 850 * 851 * @author Kazutaka Miyasaka <kazmiya@gmail.com> 852 * 853 * @param Doku_Indexer $Indexer 854 * @param string $term 855 * @param bool $consider_asian 856 * @param bool $phrase_mode 857 * @return string 858 */ 859function ft_termParser($Indexer, $term, $consider_asian = true, $phrase_mode = false) { 860 $parsed = ''; 861 if ($consider_asian) { 862 // successive asian characters need to be searched as a phrase 863 $words = preg_split('/('.IDX_ASIAN.'+)/u', $term, -1, PREG_SPLIT_DELIM_CAPTURE | PREG_SPLIT_NO_EMPTY); 864 foreach ($words as $word) { 865 $phrase_mode = $phrase_mode ? true : preg_match('/'.IDX_ASIAN.'/u', $word); 866 $parsed .= ft_termParser($Indexer, $word, false, $phrase_mode); 867 } 868 } else { 869 $term_noparen = str_replace(array('(', ')'), ' ', $term); 870 $words = $Indexer->tokenizer($term_noparen, true); 871 872 // W_: no need to highlight 873 if (empty($words)) { 874 $parsed = '()'; // important: do not remove 875 } elseif ($words[0] === $term) { 876 $parsed = '(W+:'.$words[0].')'; 877 } elseif ($phrase_mode) { 878 $term_encoded = str_replace(array('(', ')'), array('OP', 'CP'), $term); 879 $parsed = '((W_:'.implode(')(W_:', $words).')(P+:'.$term_encoded.'))'; 880 } else { 881 $parsed = '((W+:'.implode(')(W+:', $words).'))'; 882 } 883 } 884 return $parsed; 885} 886 887/** 888 * Recreate a search query string based on parsed parts, doesn't support negated phrases and `OR` searches 889 * 890 * @param array $and 891 * @param array $not 892 * @param array $phrases 893 * @param array $ns 894 * @param array $notns 895 * 896 * @return string 897 */ 898function ft_queryUnparser_simple(array $and, array $not, array $phrases, array $ns, array $notns) { 899 $query = implode(' ', $and); 900 if (!empty($not)) { 901 $query .= ' -' . implode(' -', $not); 902 } 903 904 if (!empty($phrases)) { 905 $query .= ' "' . implode('" "', $phrases) . '"'; 906 } 907 908 if (!empty($ns)) { 909 $query .= ' @' . implode(' @', $ns); 910 } 911 912 if (!empty($notns)) { 913 $query .= ' ^' . implode(' ^', $notns); 914 } 915 916 return $query; 917} 918 919//Setup VIM: ex: et ts=4 : 920