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 */ 24function ft_pageSearch($query,&$highlight){ 25 26 $data['query'] = $query; 27 $data['highlight'] =& $highlight; 28 29 return trigger_event('SEARCH_QUERY_FULLPAGE', $data, '_ft_pageSearch'); 30} 31 32/** 33 * Returns a list of matching documents for the given query 34 * 35 * @author Andreas Gohr <andi@splitbrain.org> 36 * @author Kazutaka Miyasaka <kazmiya@gmail.com> 37 */ 38function _ft_pageSearch(&$data) { 39 $Indexer = idx_get_indexer(); 40 41 // parse the given query 42 $q = ft_queryParser($Indexer, $data['query']); 43 $data['highlight'] = $q['highlight']; 44 45 if (empty($q['parsed_ary'])) return array(); 46 47 // lookup all words found in the query 48 $lookup = $Indexer->lookup($q['words']); 49 50 // get all pages in this dokuwiki site (!: includes nonexistent pages) 51 $pages_all = array(); 52 foreach ($Indexer->getPages() as $id) { 53 $pages_all[$id] = 0; // base: 0 hit 54 } 55 56 // process the query 57 $stack = array(); 58 foreach ($q['parsed_ary'] as $token) { 59 switch (substr($token, 0, 3)) { 60 case 'W+:': 61 case 'W-:': 62 case 'W_:': // word 63 $word = substr($token, 3); 64 $stack[] = (array) $lookup[$word]; 65 break; 66 case 'P+:': 67 case 'P-:': // phrase 68 $phrase = substr($token, 3); 69 // since phrases are always parsed as ((W1)(W2)...(P)), 70 // the end($stack) always points the pages that contain 71 // all words in this phrase 72 $pages = end($stack); 73 $pages_matched = array(); 74 foreach(array_keys($pages) as $id){ 75 $text = utf8_strtolower(rawWiki($id)); 76 if (strpos($text, $phrase) !== false) { 77 $pages_matched[$id] = 0; // phrase: always 0 hit 78 } 79 } 80 $stack[] = $pages_matched; 81 break; 82 case 'N+:': 83 case 'N-:': // namespace 84 $ns = substr($token, 3); 85 $pages_matched = array(); 86 foreach (array_keys($pages_all) as $id) { 87 if (strpos($id, $ns) === 0) { 88 $pages_matched[$id] = 0; // namespace: always 0 hit 89 } 90 } 91 $stack[] = $pages_matched; 92 break; 93 case 'AND': // and operation 94 list($pages1, $pages2) = array_splice($stack, -2); 95 $stack[] = ft_resultCombine(array($pages1, $pages2)); 96 break; 97 case 'OR': // or operation 98 list($pages1, $pages2) = array_splice($stack, -2); 99 $stack[] = ft_resultUnite(array($pages1, $pages2)); 100 break; 101 case 'NOT': // not operation (unary) 102 $pages = array_pop($stack); 103 $stack[] = ft_resultComplement(array($pages_all, $pages)); 104 break; 105 } 106 } 107 $docs = array_pop($stack); 108 109 if (empty($docs)) return array(); 110 111 // check: settings, acls, existence 112 foreach (array_keys($docs) as $id) { 113 if (isHiddenPage($id) || auth_quickaclcheck($id) < AUTH_READ || !page_exists($id, '', false)) { 114 unset($docs[$id]); 115 } 116 } 117 118 // sort docs by count 119 arsort($docs); 120 121 return $docs; 122} 123 124/** 125 * Returns the backlinks for a given page 126 * 127 * Uses the metadata index. 128 */ 129function ft_backlinks($id){ 130 $result = array(); 131 132 $result = idx_get_indexer()->lookupKey('relation_references', $id); 133 134 if(!count($result)) return $result; 135 136 // check ACL permissions 137 foreach(array_keys($result) as $idx){ 138 if(auth_quickaclcheck($result[$idx]) < AUTH_READ){ 139 unset($result[$idx]); 140 } 141 } 142 143 sort($result); 144 return $result; 145} 146 147/** 148 * Returns the pages that use a given media file 149 * 150 * Does a quick lookup with the fulltext index, then 151 * evaluates the instructions of the found pages 152 * 153 * Aborts after $max found results 154 */ 155function ft_mediause($id,$max){ 156 if(!$max) $max = 1; // need to find at least one 157 158 $result = array(); 159 160 // quick lookup of the mediafile 161 // FIXME use metadata key lookup 162 $media = noNS($id); 163 $matches = idx_lookup(idx_tokenizer($media)); 164 $docs = array_keys(ft_resultCombine(array_values($matches))); 165 if(!count($docs)) return $result; 166 167 // go through all found pages 168 $found = 0; 169 $pcre = preg_quote($media,'/'); 170 foreach($docs as $doc){ 171 $ns = getNS($doc); 172 preg_match_all('/\{\{([^|}]*'.$pcre.'[^|}]*)(|[^}]+)?\}\}/i',rawWiki($doc),$matches); 173 foreach($matches[1] as $img){ 174 $img = trim($img); 175 if(preg_match('/^https?:\/\//i',$img)) continue; // skip external images 176 list($img) = explode('?',$img); // remove any parameters 177 resolve_mediaid($ns,$img,$exists); // resolve the possibly relative img 178 179 if($img == $id){ // we have a match 180 $result[] = $doc; 181 $found++; 182 break; 183 } 184 } 185 if($found >= $max) break; 186 } 187 188 sort($result); 189 return $result; 190} 191 192 193 194/** 195 * Quicksearch for pagenames 196 * 197 * By default it only matches the pagename and ignores the 198 * namespace. This can be changed with the second parameter. 199 * The third parameter allows to search in titles as well. 200 * 201 * The function always returns titles as well 202 * 203 * @triggers SEARCH_QUERY_PAGELOOKUP 204 * @author Andreas Gohr <andi@splitbrain.org> 205 * @author Adrian Lang <lang@cosmocode.de> 206 */ 207function ft_pageLookup($id, $in_ns=false, $in_title=false){ 208 $data = compact('id', 'in_ns', 'in_title'); 209 $data['has_titles'] = true; // for plugin backward compatibility check 210 return trigger_event('SEARCH_QUERY_PAGELOOKUP', $data, '_ft_pageLookup'); 211} 212 213function _ft_pageLookup(&$data){ 214 // split out original parameters 215 $id = $data['id']; 216 if (preg_match('/(?:^| )@(\w+)/', $id, $matches)) { 217 $ns = cleanID($matches[1]) . ':'; 218 $id = str_replace($matches[0], '', $id); 219 } 220 221 $in_ns = $data['in_ns']; 222 $in_title = $data['in_title']; 223 $cleaned = cleanID($id); 224 225 $Indexer = idx_get_indexer(); 226 $page_idx = $Indexer->getPages(); 227 228 $pages = array(); 229 if ($id !== '' && $cleaned !== '') { 230 foreach ($page_idx as $p_id) { 231 if ((strpos($in_ns ? $p_id : noNSorNS($p_id), $cleaned) !== false)) { 232 if (!isset($pages[$p_id])) 233 $pages[$p_id] = p_get_first_heading($p_id, false); 234 } 235 } 236 if ($in_title) { 237 $wildcard_id = "*$id*"; 238 foreach ($Indexer->lookupKey('title', $wildcard_id) as $p_id) { 239 if (!isset($pages[$p_id])) 240 $pages[$p_id] = p_get_first_heading($p_id, false); 241 } 242 } 243 } 244 if (isset($ns)) { 245 foreach ($page_idx as $p_id) { 246 if (strpos($p_id, $ns) === 0) { 247 if (!isset($pages[$p_id])) 248 $pages[$p_id] = p_get_first_heading($p_id, false); 249 } 250 } 251 } 252 253 // discard hidden pages 254 // discard nonexistent pages 255 // check ACL permissions 256 foreach(array_keys($pages) as $idx){ 257 if(!isVisiblePage($idx) || !page_exists($idx) || 258 auth_quickaclcheck($idx) < AUTH_READ) { 259 unset($pages[$idx]); 260 } 261 } 262 263 uksort($pages,'ft_pagesorter'); 264 return $pages; 265} 266 267/** 268 * Sort pages based on their namespace level first, then on their string 269 * values. This makes higher hierarchy pages rank higher than lower hierarchy 270 * pages. 271 */ 272function ft_pagesorter($a, $b){ 273 $ac = count(explode(':',$a)); 274 $bc = count(explode(':',$b)); 275 if($ac < $bc){ 276 return -1; 277 }elseif($ac > $bc){ 278 return 1; 279 } 280 return strcmp ($a,$b); 281} 282 283/** 284 * Creates a snippet extract 285 * 286 * @author Andreas Gohr <andi@splitbrain.org> 287 * @triggers FULLTEXT_SNIPPET_CREATE 288 */ 289function ft_snippet($id,$highlight){ 290 $text = rawWiki($id); 291 $text = str_replace("\xC2\xAD",'',$text); // remove soft-hyphens 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('ft_snippet_re_preprocess', 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 $evdata['snippet']; 378} 379 380/** 381 * Wraps a search term in regex boundary checks. 382 */ 383function ft_snippet_re_preprocess($term) { 384 // do not process asian terms where word boundaries are not explicit 385 if(preg_match('/'.IDX_ASIAN.'/u',$term)){ 386 return $term; 387 } 388 389 if(substr($term,0,2) == '\\*'){ 390 $term = substr($term,2); 391 }else{ 392 $term = '\b'.$term; 393 } 394 395 if(substr($term,-2,2) == '\\*'){ 396 $term = substr($term,0,-2); 397 }else{ 398 $term = $term.'\b'; 399 } 400 return $term; 401} 402 403/** 404 * Combine found documents and sum up their scores 405 * 406 * This function is used to combine searched words with a logical 407 * AND. Only documents available in all arrays are returned. 408 * 409 * based upon PEAR's PHP_Compat function for array_intersect_key() 410 * 411 * @param array $args An array of page arrays 412 */ 413function ft_resultCombine($args){ 414 $array_count = count($args); 415 if($array_count == 1){ 416 return $args[0]; 417 } 418 419 $result = array(); 420 if ($array_count > 1) { 421 foreach ($args[0] as $key => $value) { 422 $result[$key] = $value; 423 for ($i = 1; $i !== $array_count; $i++) { 424 if (!isset($args[$i][$key])) { 425 unset($result[$key]); 426 break; 427 } 428 $result[$key] += $args[$i][$key]; 429 } 430 } 431 } 432 return $result; 433} 434 435/** 436 * Unites found documents and sum up their scores 437 * 438 * based upon ft_resultCombine() function 439 * 440 * @param array $args An array of page arrays 441 * @author Kazutaka Miyasaka <kazmiya@gmail.com> 442 */ 443function ft_resultUnite($args) { 444 $array_count = count($args); 445 if ($array_count === 1) { 446 return $args[0]; 447 } 448 449 $result = $args[0]; 450 for ($i = 1; $i !== $array_count; $i++) { 451 foreach (array_keys($args[$i]) as $id) { 452 $result[$id] += $args[$i][$id]; 453 } 454 } 455 return $result; 456} 457 458/** 459 * Computes the difference of documents using page id for comparison 460 * 461 * nearly identical to PHP5's array_diff_key() 462 * 463 * @param array $args An array of page arrays 464 * @author Kazutaka Miyasaka <kazmiya@gmail.com> 465 */ 466function ft_resultComplement($args) { 467 $array_count = count($args); 468 if ($array_count === 1) { 469 return $args[0]; 470 } 471 472 $result = $args[0]; 473 foreach (array_keys($result) as $id) { 474 for ($i = 1; $i !== $array_count; $i++) { 475 if (isset($args[$i][$id])) unset($result[$id]); 476 } 477 } 478 return $result; 479} 480 481/** 482 * Parses a search query and builds an array of search formulas 483 * 484 * @author Andreas Gohr <andi@splitbrain.org> 485 * @author Kazutaka Miyasaka <kazmiya@gmail.com> 486 */ 487function ft_queryParser($Indexer, $query){ 488 /** 489 * parse a search query and transform it into intermediate representation 490 * 491 * in a search query, you can use the following expressions: 492 * 493 * words: 494 * include 495 * -exclude 496 * phrases: 497 * "phrase to be included" 498 * -"phrase you want to exclude" 499 * namespaces: 500 * @include:namespace (or ns:include:namespace) 501 * ^exclude:namespace (or -ns:exclude:namespace) 502 * groups: 503 * () 504 * -() 505 * operators: 506 * and ('and' is the default operator: you can always omit this) 507 * or (or pipe symbol '|', lower precedence than 'and') 508 * 509 * e.g. a query [ aa "bb cc" @dd:ee ] means "search pages which contain 510 * a word 'aa', a phrase 'bb cc' and are within a namespace 'dd:ee'". 511 * this query is equivalent to [ -(-aa or -"bb cc" or -ns:dd:ee) ] 512 * as long as you don't mind hit counts. 513 * 514 * intermediate representation consists of the following parts: 515 * 516 * ( ) - group 517 * AND - logical and 518 * OR - logical or 519 * NOT - logical not 520 * W+:, W-:, W_: - word (underscore: no need to highlight) 521 * P+:, P-: - phrase (minus sign: logically in NOT group) 522 * N+:, N-: - namespace 523 */ 524 $parsed_query = ''; 525 $parens_level = 0; 526 $terms = preg_split('/(-?".*?")/u', utf8_strtolower($query), -1, PREG_SPLIT_DELIM_CAPTURE | PREG_SPLIT_NO_EMPTY); 527 528 foreach ($terms as $term) { 529 $parsed = ''; 530 if (preg_match('/^(-?)"(.+)"$/u', $term, $matches)) { 531 // phrase-include and phrase-exclude 532 $not = $matches[1] ? 'NOT' : ''; 533 $parsed = $not.ft_termParser($Indexer, $matches[2], false, true); 534 } else { 535 // fix incomplete phrase 536 $term = str_replace('"', ' ', $term); 537 538 // fix parentheses 539 $term = str_replace(')' , ' ) ', $term); 540 $term = str_replace('(' , ' ( ', $term); 541 $term = str_replace('- (', ' -(', $term); 542 543 // treat pipe symbols as 'OR' operators 544 $term = str_replace('|', ' or ', $term); 545 546 // treat ideographic spaces (U+3000) as search term separators 547 // FIXME: some more separators? 548 $term = preg_replace('/[ \x{3000}]+/u', ' ', $term); 549 $term = trim($term); 550 if ($term === '') continue; 551 552 $tokens = explode(' ', $term); 553 foreach ($tokens as $token) { 554 if ($token === '(') { 555 // parenthesis-include-open 556 $parsed .= '('; 557 ++$parens_level; 558 } elseif ($token === '-(') { 559 // parenthesis-exclude-open 560 $parsed .= 'NOT('; 561 ++$parens_level; 562 } elseif ($token === ')') { 563 // parenthesis-any-close 564 if ($parens_level === 0) continue; 565 $parsed .= ')'; 566 $parens_level--; 567 } elseif ($token === 'and') { 568 // logical-and (do nothing) 569 } elseif ($token === 'or') { 570 // logical-or 571 $parsed .= 'OR'; 572 } elseif (preg_match('/^(?:\^|-ns:)(.+)$/u', $token, $matches)) { 573 // namespace-exclude 574 $parsed .= 'NOT(N+:'.$matches[1].')'; 575 } elseif (preg_match('/^(?:@|ns:)(.+)$/u', $token, $matches)) { 576 // namespace-include 577 $parsed .= '(N+:'.$matches[1].')'; 578 } elseif (preg_match('/^-(.+)$/', $token, $matches)) { 579 // word-exclude 580 $parsed .= 'NOT('.ft_termParser($Indexer, $matches[1]).')'; 581 } else { 582 // word-include 583 $parsed .= ft_termParser($Indexer, $token); 584 } 585 } 586 } 587 $parsed_query .= $parsed; 588 } 589 590 // cleanup (very sensitive) 591 $parsed_query .= str_repeat(')', $parens_level); 592 do { 593 $parsed_query_old = $parsed_query; 594 $parsed_query = preg_replace('/(NOT)?\(\)/u', '', $parsed_query); 595 } while ($parsed_query !== $parsed_query_old); 596 $parsed_query = preg_replace('/(NOT|OR)+\)/u', ')' , $parsed_query); 597 $parsed_query = preg_replace('/(OR)+/u' , 'OR' , $parsed_query); 598 $parsed_query = preg_replace('/\(OR/u' , '(' , $parsed_query); 599 $parsed_query = preg_replace('/^OR|OR$/u' , '' , $parsed_query); 600 $parsed_query = preg_replace('/\)(NOT)?\(/u' , ')AND$1(', $parsed_query); 601 602 // adjustment: make highlightings right 603 $parens_level = 0; 604 $notgrp_levels = array(); 605 $parsed_query_new = ''; 606 $tokens = preg_split('/(NOT\(|[()])/u', $parsed_query, -1, PREG_SPLIT_DELIM_CAPTURE | PREG_SPLIT_NO_EMPTY); 607 foreach ($tokens as $token) { 608 if ($token === 'NOT(') { 609 $notgrp_levels[] = ++$parens_level; 610 } elseif ($token === '(') { 611 ++$parens_level; 612 } elseif ($token === ')') { 613 if ($parens_level-- === end($notgrp_levels)) array_pop($notgrp_levels); 614 } elseif (count($notgrp_levels) % 2 === 1) { 615 // turn highlight-flag off if terms are logically in "NOT" group 616 $token = preg_replace('/([WPN])\+\:/u', '$1-:', $token); 617 } 618 $parsed_query_new .= $token; 619 } 620 $parsed_query = $parsed_query_new; 621 622 /** 623 * convert infix notation string into postfix (Reverse Polish notation) array 624 * by Shunting-yard algorithm 625 * 626 * see: http://en.wikipedia.org/wiki/Reverse_Polish_notation 627 * see: http://en.wikipedia.org/wiki/Shunting-yard_algorithm 628 */ 629 $parsed_ary = array(); 630 $ope_stack = array(); 631 $ope_precedence = array(')' => 1, 'OR' => 2, 'AND' => 3, 'NOT' => 4, '(' => 5); 632 $ope_regex = '/([()]|OR|AND|NOT)/u'; 633 634 $tokens = preg_split($ope_regex, $parsed_query, -1, PREG_SPLIT_DELIM_CAPTURE | PREG_SPLIT_NO_EMPTY); 635 foreach ($tokens as $token) { 636 if (preg_match($ope_regex, $token)) { 637 // operator 638 $last_ope = end($ope_stack); 639 while ($ope_precedence[$token] <= $ope_precedence[$last_ope] && $last_ope != '(') { 640 $parsed_ary[] = array_pop($ope_stack); 641 $last_ope = end($ope_stack); 642 } 643 if ($token == ')') { 644 array_pop($ope_stack); // this array_pop always deletes '(' 645 } else { 646 $ope_stack[] = $token; 647 } 648 } else { 649 // operand 650 $token_decoded = str_replace(array('OP', 'CP'), array('(', ')'), $token); 651 $parsed_ary[] = $token_decoded; 652 } 653 } 654 $parsed_ary = array_values(array_merge($parsed_ary, array_reverse($ope_stack))); 655 656 // cleanup: each double "NOT" in RPN array actually does nothing 657 $parsed_ary_count = count($parsed_ary); 658 for ($i = 1; $i < $parsed_ary_count; ++$i) { 659 if ($parsed_ary[$i] === 'NOT' && $parsed_ary[$i - 1] === 'NOT') { 660 unset($parsed_ary[$i], $parsed_ary[$i - 1]); 661 } 662 } 663 $parsed_ary = array_values($parsed_ary); 664 665 // build return value 666 $q = array(); 667 $q['query'] = $query; 668 $q['parsed_str'] = $parsed_query; 669 $q['parsed_ary'] = $parsed_ary; 670 671 foreach ($q['parsed_ary'] as $token) { 672 if ($token[2] !== ':') continue; 673 $body = substr($token, 3); 674 675 switch (substr($token, 0, 3)) { 676 case 'N+:': 677 $q['ns'][] = $body; // for backward compatibility 678 break; 679 case 'N-:': 680 $q['notns'][] = $body; // for backward compatibility 681 break; 682 case 'W_:': 683 $q['words'][] = $body; 684 break; 685 case 'W-:': 686 $q['words'][] = $body; 687 $q['not'][] = $body; // for backward compatibility 688 break; 689 case 'W+:': 690 $q['words'][] = $body; 691 $q['highlight'][] = $body; 692 $q['and'][] = $body; // for backward compatibility 693 break; 694 case 'P-:': 695 $q['phrases'][] = $body; 696 break; 697 case 'P+:': 698 $q['phrases'][] = $body; 699 $q['highlight'][] = $body; 700 break; 701 } 702 } 703 foreach (array('words', 'phrases', 'highlight', 'ns', 'notns', 'and', 'not') as $key) { 704 $q[$key] = empty($q[$key]) ? array() : array_values(array_unique($q[$key])); 705 } 706 707 return $q; 708} 709 710/** 711 * Transforms given search term into intermediate representation 712 * 713 * This function is used in ft_queryParser() and not for general purpose use. 714 * 715 * @author Kazutaka Miyasaka <kazmiya@gmail.com> 716 */ 717function ft_termParser($Indexer, $term, $consider_asian = true, $phrase_mode = false) { 718 $parsed = ''; 719 if ($consider_asian) { 720 // successive asian characters need to be searched as a phrase 721 $words = preg_split('/('.IDX_ASIAN.'+)/u', $term, -1, PREG_SPLIT_DELIM_CAPTURE | PREG_SPLIT_NO_EMPTY); 722 foreach ($words as $word) { 723 $phrase_mode = $phrase_mode ? true : preg_match('/'.IDX_ASIAN.'/u', $word); 724 $parsed .= ft_termParser($Indexer, $word, false, $phrase_mode); 725 } 726 } else { 727 $term_noparen = str_replace(array('(', ')'), ' ', $term); 728 $words = $Indexer->tokenizer($term_noparen, true); 729 730 // W_: no need to highlight 731 if (empty($words)) { 732 $parsed = '()'; // important: do not remove 733 } elseif ($words[0] === $term) { 734 $parsed = '(W+:'.$words[0].')'; 735 } elseif ($phrase_mode) { 736 $term_encoded = str_replace(array('(', ')'), array('OP', 'CP'), $term); 737 $parsed = '((W_:'.implode(')(W_:', $words).')(P+:'.$term_encoded.'))'; 738 } else { 739 $parsed = '((W+:'.implode(')(W+:', $words).'))'; 740 } 741 } 742 return $parsed; 743} 744 745//Setup VIM: ex: et ts=4 : 746