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.'); 10require_once(DOKU_INC.'inc/indexer.php'); 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 */ 25function ft_pageSearch($query,&$highlight){ 26 27 $data['query'] = $query; 28 $data['highlight'] =& $highlight; 29 30 return trigger_event('SEARCH_QUERY_FULLPAGE', $data, '_ft_pageSearch'); 31} 32 33/** 34 * Returns a list of matching documents for the given query 35 * 36 * @author Andreas Gohr <andi@splitbrain.org> 37 * @author Kazutaka Miyasaka <kazmiya@gmail.com> 38 */ 39function _ft_pageSearch(&$data) { 40 // parse the given query 41 $q = ft_queryParser($data['query']); 42 $data['highlight'] = $q['highlight']; 43 44 if (empty($q['parsed_ary'])) return array(); 45 46 // lookup all words found in the query 47 $lookup = idx_lookup($q['words']); 48 49 // get all pages in this dokuwiki site (!: includes nonexistent pages) 50 $pages_all = array(); 51 foreach (idx_getIndex('page', '') as $id) { 52 $pages_all[trim($id)] = 0; // base: 0 hit 53 } 54 55 // process the query 56 $stack = array(); 57 foreach ($q['parsed_ary'] as $token) { 58 switch (substr($token, 0, 3)) { 59 case 'W+:': 60 case 'W-:': 61 case 'W_:': // word 62 $word = substr($token, 3); 63 $stack[] = (array) $lookup[$word]; 64 break; 65 case 'P+:': 66 case 'P-:': // phrase 67 $phrase = substr($token, 3); 68 // since phrases are always parsed as ((W1)(W2)...(P)), 69 // the end($stack) always points the pages that contain 70 // all words in this phrase 71 $pages = end($stack); 72 $pages_matched = array(); 73 foreach(array_keys($pages) as $id){ 74 $text = utf8_strtolower(rawWiki($id)); 75 if (strpos($text, $phrase) !== false) { 76 $pages_matched[$id] = 0; // phrase: always 0 hit 77 } 78 } 79 $stack[] = $pages_matched; 80 break; 81 case 'N+:': 82 case 'N-:': // namespace 83 $ns = substr($token, 3); 84 $pages_matched = array(); 85 foreach (array_keys($pages_all) as $id) { 86 if (strpos($id, $ns) === 0) { 87 $pages_matched[$id] = 0; // namespace: always 0 hit 88 } 89 } 90 $stack[] = $pages_matched; 91 break; 92 case 'AND': // and operation 93 list($pages1, $pages2) = array_splice($stack, -2); 94 $stack[] = ft_resultCombine(array($pages1, $pages2)); 95 break; 96 case 'OR': // or operation 97 list($pages1, $pages2) = array_splice($stack, -2); 98 $stack[] = ft_resultUnite(array($pages1, $pages2)); 99 break; 100 case 'NOT': // not operation (unary) 101 $pages = array_pop($stack); 102 $stack[] = ft_resultComplement(array($pages_all, $pages)); 103 break; 104 } 105 } 106 $docs = array_pop($stack); 107 108 if (empty($docs)) return array(); 109 110 // check: settings, acls, existence 111 foreach (array_keys($docs) as $id) { 112 if (isHiddenPage($id) || auth_quickaclcheck($id) < AUTH_READ || !page_exists($id, '', false)) { 113 unset($docs[$id]); 114 } 115 } 116 117 // sort docs by count 118 arsort($docs); 119 120 return $docs; 121} 122 123/** 124 * Returns the backlinks for a given page 125 * 126 * Does a quick lookup with the fulltext index, then 127 * evaluates the instructions of the found pages 128 */ 129function ft_backlinks($id){ 130 global $conf; 131 $swfile = DOKU_INC.'inc/lang/'.$conf['lang'].'/stopwords.txt'; 132 $stopwords = @file_exists($swfile) ? file($swfile) : array(); 133 134 $result = array(); 135 136 // quick lookup of the pagename 137 $page = noNS($id); 138 $matches = idx_lookup(idx_tokenizer($page,$stopwords)); // pagename may contain specials (_ or .) 139 $docs = array_keys(ft_resultCombine(array_values($matches))); 140 $docs = array_filter($docs,'isVisiblePage'); // discard hidden pages 141 if(!count($docs)) return $result; 142 require_once(DOKU_INC.'inc/parserutils.php'); 143 144 // check metadata for matching links 145 foreach($docs as $match){ 146 // metadata relation reference links are already resolved 147 $links = p_get_metadata($match,'relation references'); 148 if (isset($links[$id])) $result[] = $match; 149 } 150 151 if(!count($result)) return $result; 152 153 // check ACL permissions 154 foreach(array_keys($result) as $idx){ 155 if(auth_quickaclcheck($result[$idx]) < AUTH_READ){ 156 unset($result[$idx]); 157 } 158 } 159 160 sort($result); 161 return $result; 162} 163 164/** 165 * Returns the pages that use a given media file 166 * 167 * Does a quick lookup with the fulltext index, then 168 * evaluates the instructions of the found pages 169 * 170 * Aborts after $max found results 171 */ 172function ft_mediause($id,$max){ 173 global $conf; 174 $swfile = DOKU_INC.'inc/lang/'.$conf['lang'].'/stopwords.txt'; 175 $stopwords = @file_exists($swfile) ? file($swfile) : array(); 176 177 if(!$max) $max = 1; // need to find at least one 178 179 $result = array(); 180 181 // quick lookup of the mediafile 182 $media = noNS($id); 183 $matches = idx_lookup(idx_tokenizer($media,$stopwords)); 184 $docs = array_keys(ft_resultCombine(array_values($matches))); 185 if(!count($docs)) return $result; 186 187 // go through all found pages 188 $found = 0; 189 $pcre = preg_quote($media,'/'); 190 foreach($docs as $doc){ 191 $ns = getNS($doc); 192 preg_match_all('/\{\{([^|}]*'.$pcre.'[^|}]*)(|[^}]+)?\}\}/i',rawWiki($doc),$matches); 193 foreach($matches[1] as $img){ 194 $img = trim($img); 195 if(preg_match('/^https?:\/\//i',$img)) continue; // skip external images 196 list($img) = explode('?',$img); // remove any parameters 197 resolve_mediaid($ns,$img,$exists); // resolve the possibly relative img 198 199 if($img == $id){ // we have a match 200 $result[] = $doc; 201 $found++; 202 break; 203 } 204 } 205 if($found >= $max) break; 206 } 207 208 sort($result); 209 return $result; 210} 211 212 213 214/** 215 * Quicksearch for pagenames 216 * 217 * By default it only matches the pagename and ignores the 218 * namespace. This can be changed with the second parameter 219 * 220 * refactored into ft_pageLookup(), _ft_pageLookup() and trigger_event() 221 * 222 * @author Andreas Gohr <andi@splitbrain.org> 223 */ 224function ft_pageLookup($id,$pageonly=true){ 225 $data = array('id' => $id, 'pageonly' => $pageonly); 226 return trigger_event('SEARCH_QUERY_PAGELOOKUP',$data,'_ft_pageLookup'); 227} 228 229function _ft_pageLookup(&$data){ 230 // split out original parameterrs 231 $id = $data['id']; 232 $pageonly = $data['pageonly']; 233 234 global $conf; 235 $id = preg_quote($id,'/'); 236 $pages = file($conf['indexdir'].'/page.idx'); 237 if($id) $pages = array_values(preg_grep('/'.$id.'/',$pages)); 238 239 $cnt = count($pages); 240 for($i=0; $i<$cnt; $i++){ 241 if($pageonly){ 242 if(!preg_match('/'.$id.'/',noNS($pages[$i]))){ 243 unset($pages[$i]); 244 continue; 245 } 246 } 247 if(!page_exists($pages[$i])){ 248 unset($pages[$i]); 249 continue; 250 } 251 } 252 253 $pages = array_filter($pages,'isVisiblePage'); // discard hidden pages 254 if(!count($pages)) return array(); 255 256 // check ACL permissions 257 foreach(array_keys($pages) as $idx){ 258 if(auth_quickaclcheck(trim($pages[$idx])) < AUTH_READ){ 259 unset($pages[$idx]); 260 } 261 } 262 263 $pages = array_map('trim',$pages); 264 usort($pages,'ft_pagesorter'); 265 return $pages; 266} 267 268/** 269 * Sort pages based on their namespace level first, then on their string 270 * values. This makes higher hierarchy pages rank higher than lower hierarchy 271 * pages. 272 */ 273function ft_pagesorter($a, $b){ 274 $ac = count(explode(':',$a)); 275 $bc = count(explode(':',$b)); 276 if($ac < $bc){ 277 return -1; 278 }elseif($ac > $bc){ 279 return 1; 280 } 281 return strcmp ($a,$b); 282} 283 284/** 285 * Creates a snippet extract 286 * 287 * @author Andreas Gohr <andi@splitbrain.org> 288 * @triggers FULLTEXT_SNIPPET_CREATE 289 */ 290function ft_snippet($id,$highlight){ 291 $text = rawWiki($id); 292 $evdata = array( 293 'id' => $id, 294 'text' => &$text, 295 'highlight' => &$highlight, 296 'snippet' => '', 297 ); 298 299 $evt = new Doku_Event('FULLTEXT_SNIPPET_CREATE',$evdata); 300 if ($evt->advise_before()) { 301 $match = array(); 302 $snippets = array(); 303 $utf8_offset = $offset = $end = 0; 304 $len = utf8_strlen($text); 305 306 // build a regexp from the phrases to highlight 307 $re1 = '('.join('|',array_map('preg_quote_cb',array_filter((array) $highlight))).')'; 308 $re2 = "$re1.{0,75}(?!\\1)$re1"; 309 $re3 = "$re1.{0,45}(?!\\1)$re1.{0,45}(?!\\1)(?!\\2)$re1"; 310 311 for ($cnt=4; $cnt--;) { 312 if (0) { 313 } else if (preg_match('/'.$re3.'/iu',$text,$match,PREG_OFFSET_CAPTURE,$offset)) { 314 } else if (preg_match('/'.$re2.'/iu',$text,$match,PREG_OFFSET_CAPTURE,$offset)) { 315 } else if (preg_match('/'.$re1.'/iu',$text,$match,PREG_OFFSET_CAPTURE,$offset)) { 316 } else { 317 break; 318 } 319 320 list($str,$idx) = $match[0]; 321 322 // convert $idx (a byte offset) into a utf8 character offset 323 $utf8_idx = utf8_strlen(substr($text,0,$idx)); 324 $utf8_len = utf8_strlen($str); 325 326 // establish context, 100 bytes surrounding the match string 327 // first look to see if we can go 100 either side, 328 // then drop to 50 adding any excess if the other side can't go to 50, 329 $pre = min($utf8_idx-$utf8_offset,100); 330 $post = min($len-$utf8_idx-$utf8_len,100); 331 332 if ($pre>50 && $post>50) { 333 $pre = $post = 50; 334 } else if ($pre>50) { 335 $pre = min($pre,100-$post); 336 } else if ($post>50) { 337 $post = min($post, 100-$pre); 338 } else { 339 // both are less than 50, means the context is the whole string 340 // make it so and break out of this loop - there is no need for the 341 // complex snippet calculations 342 $snippets = array($text); 343 break; 344 } 345 346 // establish context start and end points, try to append to previous 347 // context if possible 348 $start = $utf8_idx - $pre; 349 $append = ($start < $end) ? $end : false; // still the end of the previous context snippet 350 $end = $utf8_idx + $utf8_len + $post; // now set it to the end of this context 351 352 if ($append) { 353 $snippets[count($snippets)-1] .= utf8_substr($text,$append,$end-$append); 354 } else { 355 $snippets[] = utf8_substr($text,$start,$end-$start); 356 } 357 358 // set $offset for next match attempt 359 // substract strlen to avoid splitting a potential search success, 360 // this is an approximation as the search pattern may match strings 361 // of varying length and it will fail if the context snippet 362 // boundary breaks a matching string longer than the current match 363 $utf8_offset = $utf8_idx + $post; 364 $offset = $idx + strlen(utf8_substr($text,$utf8_idx,$post)); 365 $offset = utf8_correctIdx($text,$offset); 366 } 367 368 $m = "\1"; 369 $snippets = preg_replace('/'.$re1.'/iu',$m.'$1'.$m,$snippets); 370 $snippet = preg_replace('/'.$m.'([^'.$m.']*?)'.$m.'/iu','<strong class="search_hit">$1</strong>',hsc(join('... ',$snippets))); 371 372 $evdata['snippet'] = $snippet; 373 } 374 $evt->advise_after(); 375 unset($evt); 376 377 return $evdata['snippet']; 378} 379 380/** 381 * Combine found documents and sum up their scores 382 * 383 * This function is used to combine searched words with a logical 384 * AND. Only documents available in all arrays are returned. 385 * 386 * based upon PEAR's PHP_Compat function for array_intersect_key() 387 * 388 * @param array $args An array of page arrays 389 */ 390function ft_resultCombine($args){ 391 $array_count = count($args); 392 if($array_count == 1){ 393 return $args[0]; 394 } 395 396 $result = array(); 397 if ($array_count > 1) { 398 foreach ($args[0] as $key => $value) { 399 $result[$key] = $value; 400 for ($i = 1; $i !== $array_count; $i++) { 401 if (!isset($args[$i][$key])) { 402 unset($result[$key]); 403 break; 404 } 405 $result[$key] += $args[$i][$key]; 406 } 407 } 408 } 409 return $result; 410} 411 412/** 413 * Unites found documents and sum up their scores 414 * 415 * based upon ft_resultCombine() function 416 * 417 * @param array $args An array of page arrays 418 * @author Kazutaka Miyasaka <kazmiya@gmail.com> 419 */ 420function ft_resultUnite($args) { 421 $array_count = count($args); 422 if ($array_count === 1) { 423 return $args[0]; 424 } 425 426 $result = $args[0]; 427 for ($i = 1; $i !== $array_count; $i++) { 428 foreach (array_keys($args[$i]) as $id) { 429 $result[$id] += $args[$i][$id]; 430 } 431 } 432 return $result; 433} 434 435/** 436 * Computes the difference of documents using page id for comparison 437 * 438 * nearly identical to PHP5's array_diff_key() 439 * 440 * @param array $args An array of page arrays 441 * @author Kazutaka Miyasaka <kazmiya@gmail.com> 442 */ 443function ft_resultComplement($args) { 444 $array_count = count($args); 445 if ($array_count === 1) { 446 return $args[0]; 447 } 448 449 $result = $args[0]; 450 foreach (array_keys($result) as $id) { 451 for ($i = 1; $i !== $array_count; $i++) { 452 if (isset($args[$i][$id])) unset($result[$id]); 453 } 454 } 455 return $result; 456} 457 458/** 459 * Parses a search query and builds an array of search formulas 460 * 461 * @author Andreas Gohr <andi@splitbrain.org> 462 * @author Kazutaka Miyasaka <kazmiya@gmail.com> 463 */ 464function ft_queryParser($query){ 465 global $conf; 466 $swfile = DOKU_INC.'inc/lang/'.$conf['lang'].'/stopwords.txt'; 467 $stopwords = @file_exists($swfile) ? file($swfile) : array(); 468 469 /** 470 * parse a search query and transform it into intermediate representation 471 * 472 * in a search query, you can use the following expressions: 473 * 474 * words: 475 * include 476 * -exclude 477 * phrases: 478 * "phrase to be included" 479 * -"phrase you want to exclude" 480 * namespaces: 481 * @include:namespace (or ns:include:namespace) 482 * ^exclude:namespace (or -ns:exclude:namespace) 483 * groups: 484 * () 485 * -() 486 * operators: 487 * and ('and' is the default operator: you can always omit this) 488 * or (or pipe symbol '|', lower precedence than 'and') 489 * 490 * e.g. a query [ aa "bb cc" @dd:ee ] means "search pages which contain 491 * a word 'aa', a phrase 'bb cc' and are within a namespace 'dd:ee'". 492 * this query is equivalent to [ -(-aa or -"bb cc" or -ns:dd:ee) ] 493 * as long as you don't mind hit counts. 494 * 495 * intermediate representation consists of the following parts: 496 * 497 * ( ) - group 498 * AND - logical and 499 * OR - logical or 500 * NOT - logical not 501 * W+:, W-:, W_: - word (underscore: no need to highlight) 502 * P+:, P-: - phrase (minus sign: logically in NOT group) 503 * N+:, N-: - namespace 504 */ 505 $parsed_query = ''; 506 $parens_level = 0; 507 $terms = preg_split('/(-?".*?")/u', utf8_strtolower($query), -1, PREG_SPLIT_DELIM_CAPTURE | PREG_SPLIT_NO_EMPTY); 508 509 foreach ($terms as $term) { 510 $parsed = ''; 511 if (preg_match('/^(-?)"(.+)"$/u', $term, $matches)) { 512 // phrase-include and phrase-exclude 513 $not = $matches[1] ? 'NOT' : ''; 514 $parsed = $not.ft_termParser($matches[2], $stopwords, false, true); 515 } else { 516 // fix incomplete phrase 517 $term = str_replace('"', ' ', $term); 518 519 // fix parentheses 520 $term = str_replace(')' , ' ) ', $term); 521 $term = str_replace('(' , ' ( ', $term); 522 $term = str_replace('- (', ' -(', $term); 523 524 // treat pipe symbols as 'OR' operators 525 $term = str_replace('|', ' or ', $term); 526 527 // treat ideographic spaces (U+3000) as search term separators 528 // FIXME: some more separators? 529 $term = preg_replace('/[ \x{3000}]+/u', ' ', $term); 530 $term = trim($term); 531 if ($term === '') continue; 532 533 $tokens = explode(' ', $term); 534 foreach ($tokens as $token) { 535 if ($token === '(') { 536 // parenthesis-include-open 537 $parsed .= '('; 538 ++$parens_level; 539 } elseif ($token === '-(') { 540 // parenthesis-exclude-open 541 $parsed .= 'NOT('; 542 ++$parens_level; 543 } elseif ($token === ')') { 544 // parenthesis-any-close 545 if ($parens_level === 0) continue; 546 $parsed .= ')'; 547 $parens_level--; 548 } elseif ($token === 'and') { 549 // logical-and (do nothing) 550 } elseif ($token === 'or') { 551 // logical-or 552 $parsed .= 'OR'; 553 } elseif (preg_match('/^(?:\^|-ns:)(.+)$/u', $token, $matches)) { 554 // namespace-exclude 555 $parsed .= 'NOT(N+:'.$matches[1].')'; 556 } elseif (preg_match('/^(?:@|ns:)(.+)$/u', $token, $matches)) { 557 // namespace-include 558 $parsed .= '(N+:'.$matches[1].')'; 559 } elseif (preg_match('/^-(.+)$/', $token, $matches)) { 560 // word-exclude 561 $parsed .= 'NOT('.ft_termParser($matches[1], $stopwords).')'; 562 } else { 563 // word-include 564 $parsed .= ft_termParser($token, $stopwords); 565 } 566 } 567 } 568 $parsed_query .= $parsed; 569 } 570 571 // cleanup (very sensitive) 572 $parsed_query .= str_repeat(')', $parens_level); 573 do { 574 $parsed_query_old = $parsed_query; 575 $parsed_query = preg_replace('/(NOT)?\(\)/u', '', $parsed_query); 576 } while ($parsed_query !== $parsed_query_old); 577 $parsed_query = preg_replace('/(NOT|OR)+\)/u', ')' , $parsed_query); 578 $parsed_query = preg_replace('/(OR)+/u' , 'OR' , $parsed_query); 579 $parsed_query = preg_replace('/\(OR/u' , '(' , $parsed_query); 580 $parsed_query = preg_replace('/^OR|OR$/u' , '' , $parsed_query); 581 $parsed_query = preg_replace('/\)(NOT)?\(/u' , ')AND$1(', $parsed_query); 582 583 // adjustment: make highlightings right 584 $parens_level = 0; 585 $notgrp_levels = array(); 586 $parsed_query_new = ''; 587 $tokens = preg_split('/(NOT\(|[()])/u', $parsed_query, -1, PREG_SPLIT_DELIM_CAPTURE | PREG_SPLIT_NO_EMPTY); 588 foreach ($tokens as $token) { 589 if ($token === 'NOT(') { 590 $notgrp_levels[] = ++$parens_level; 591 } elseif ($token === '(') { 592 ++$parens_level; 593 } elseif ($token === ')') { 594 if ($parens_level-- === end($notgrp_levels)) array_pop($notgrp_levels); 595 } elseif (count($notgrp_levels) % 2 === 1) { 596 // turn highlight-flag off if terms are logically in "NOT" group 597 $token = preg_replace('/([WPN])\+\:/u', '$1-:', $token); 598 } 599 $parsed_query_new .= $token; 600 } 601 $parsed_query = $parsed_query_new; 602 603 /** 604 * convert infix notation string into postfix (Reverse Polish notation) array 605 * by Shunting-yard algorithm 606 * 607 * see: http://en.wikipedia.org/wiki/Reverse_Polish_notation 608 * see: http://en.wikipedia.org/wiki/Shunting-yard_algorithm 609 */ 610 $parsed_ary = array(); 611 $ope_stack = array(); 612 $ope_precedence = array(')' => 1, 'OR' => 2, 'AND' => 3, 'NOT' => 4, '(' => 5); 613 $ope_regex = '/([()]|OR|AND|NOT)/u'; 614 615 $tokens = preg_split($ope_regex, $parsed_query, -1, PREG_SPLIT_DELIM_CAPTURE | PREG_SPLIT_NO_EMPTY); 616 foreach ($tokens as $token) { 617 if (preg_match($ope_regex, $token)) { 618 // operator 619 $last_ope = end($ope_stack); 620 while ($ope_precedence[$token] <= $ope_precedence[$last_ope] && $last_ope != '(') { 621 $parsed_ary[] = array_pop($ope_stack); 622 $last_ope = end($ope_stack); 623 } 624 if ($token == ')') { 625 array_pop($ope_stack); // this array_pop always deletes '(' 626 } else { 627 $ope_stack[] = $token; 628 } 629 } else { 630 // operand 631 $token_decoded = str_replace(array('OP', 'CP'), array('(', ')'), $token); 632 $parsed_ary[] = $token_decoded; 633 } 634 } 635 $parsed_ary = array_values(array_merge($parsed_ary, array_reverse($ope_stack))); 636 637 // cleanup: each double "NOT" in RPN array actually does nothing 638 $parsed_ary_count = count($parsed_ary); 639 for ($i = 1; $i < $parsed_ary_count; ++$i) { 640 if ($parsed_ary[$i] === 'NOT' && $parsed_ary[$i - 1] === 'NOT') { 641 unset($parsed_ary[$i], $parsed_ary[$i - 1]); 642 } 643 } 644 $parsed_ary = array_values($parsed_ary); 645 646 // build return value 647 $q = array(); 648 $q['query'] = $query; 649 $q['parsed_str'] = $parsed_query; 650 $q['parsed_ary'] = $parsed_ary; 651 652 foreach ($q['parsed_ary'] as $token) { 653 if ($token[2] !== ':') continue; 654 $body = substr($token, 3); 655 656 switch (substr($token, 0, 3)) { 657 case 'N+:': 658 $q['ns'][] = $body; // for backward compatibility 659 break; 660 case 'N-:': 661 $q['notns'][] = $body; // for backward compatibility 662 break; 663 case 'W_:': 664 $q['words'][] = $body; 665 break; 666 case 'W-:': 667 $q['words'][] = $body; 668 $q['not'][] = $body; // for backward compatibility 669 break; 670 case 'W+:': 671 $q['words'][] = $body; 672 $q['highlight'][] = str_replace('*', '', $body); 673 $q['and'][] = $body; // for backward compatibility 674 break; 675 case 'P-:': 676 $q['phrases'][] = $body; 677 break; 678 case 'P+:': 679 $q['phrases'][] = $body; 680 $q['highlight'][] = str_replace('*', '', $body); 681 break; 682 } 683 } 684 foreach (array('words', 'phrases', 'highlight', 'ns', 'notns', 'and', 'not') as $key) { 685 $q[$key] = empty($q[$key]) ? array() : array_values(array_unique($q[$key])); 686 } 687 688 return $q; 689} 690 691/** 692 * Transforms given search term into intermediate representation 693 * 694 * This function is used in ft_queryParser() and not for general purpose use. 695 * 696 * @author Kazutaka Miyasaka <kazmiya@gmail.com> 697 */ 698function ft_termParser($term, &$stopwords, $consider_asian = true, $phrase_mode = false) { 699 $parsed = ''; 700 if ($consider_asian) { 701 // successive asian characters need to be searched as a phrase 702 $words = preg_split('/('.IDX_ASIAN.'+)/u', $term, -1, PREG_SPLIT_DELIM_CAPTURE | PREG_SPLIT_NO_EMPTY); 703 foreach ($words as $word) { 704 if (preg_match('/'.IDX_ASIAN.'/u', $word)) $phrase_mode = true; 705 $parsed .= ft_termParser($word, $stopwords, false, $phrase_mode); 706 } 707 } else { 708 $term_noparen = str_replace(array('(', ')'), ' ', $term); 709 $words = idx_tokenizer($term_noparen, $stopwords, true); 710 711 // W_: no need to highlight 712 if (empty($words)) { 713 $parsed = '()'; // important: do not remove 714 } elseif ($words[0] === $term) { 715 $parsed = '(W+:'.$words[0].')'; 716 } elseif ($phrase_mode) { 717 $term_encoded = str_replace(array('(', ')'), array('OP', 'CP'), $term); 718 $parsed = '((W_:'.implode(')(W_:', $words).')(P+:'.$term_encoded.'))'; 719 } else { 720 $parsed = '((W+:'.implode(')(W+:', $words).'))'; 721 } 722 } 723 return $parsed; 724} 725 726//Setup VIM: ex: et ts=4 enc=utf-8 : 727