1<?php 2 3namespace dokuwiki\Search; 4 5use dokuwiki\Utf8\Asian; 6use dokuwiki\Utf8\Clean; 7use dokuwiki\Utf8\PhpString; 8use dokuwiki\Extension\Event; 9 10/** 11 * Class that encapsulates operations on the indexer database. 12 * 13 * @author Tom N Harris <tnharris@whoopdedo.org> 14 */ 15class Indexer 16{ 17 /** 18 * @var array $pidCache Cache for getPID() 19 */ 20 protected $pidCache = []; 21 22 /** 23 * Adds the contents of a page to the fulltext index 24 * 25 * The added text replaces previous words for the same page. 26 * An empty value erases the page. 27 * 28 * @param string $page a page name 29 * @param string $text the body of the page 30 * @return string|boolean the function completed successfully 31 * 32 * @author Tom N Harris <tnharris@whoopdedo.org> 33 * @author Andreas Gohr <andi@splitbrain.org> 34 */ 35 public function addPageWords($page, $text) 36 { 37 if (!$this->lock()) 38 return "locked"; 39 40 // load known documents 41 $pid = $this->getPIDNoLock($page); 42 if ($pid === false) { 43 $this->unlock(); 44 return false; 45 } 46 47 $pagewords = []; 48 // get word usage in page 49 $words = $this->getPageWords($text); 50 if ($words === false) { 51 $this->unlock(); 52 return false; 53 } 54 55 if (!empty($words)) { 56 foreach (array_keys($words) as $wlen) { 57 $index = $this->getIndex('i', $wlen); 58 foreach ($words[$wlen] as $wid => $freq) { 59 $idx = ($wid < count($index)) ? $index[$wid] : ''; 60 $index[$wid] = $this->updateTuple($idx, $pid, $freq); 61 $pagewords[] = "$wlen*$wid"; 62 } 63 if (!$this->saveIndex('i', $wlen, $index)) { 64 $this->unlock(); 65 return false; 66 } 67 } 68 } 69 70 // Remove obsolete index entries 71 $pageword_idx = $this->getIndexKey('pageword', '', $pid); 72 if ($pageword_idx !== '') { 73 $oldwords = explode(':', $pageword_idx); 74 $delwords = array_diff($oldwords, $pagewords); 75 $upwords = []; 76 foreach ($delwords as $word) { 77 if ($word != '') { 78 [$wlen, $wid] = explode('*', $word); 79 $wid = (int)$wid; 80 $upwords[$wlen][] = $wid; 81 } 82 } 83 foreach ($upwords as $wlen => $widx) { 84 $index = $this->getIndex('i', $wlen); 85 foreach ($widx as $wid) { 86 $index[$wid] = $this->updateTuple($index[$wid], $pid, 0); 87 } 88 $this->saveIndex('i', $wlen, $index); 89 } 90 } 91 // Save the reverse index 92 $pageword_idx = implode(':', $pagewords); 93 if (!$this->saveIndexKey('pageword', '', $pid, $pageword_idx)) { 94 $this->unlock(); 95 return false; 96 } 97 98 $this->unlock(); 99 return true; 100 } 101 102 /** 103 * Split the words in a page and add them to the index. 104 * 105 * @param string $text content of the page 106 * @return array list of word IDs and number of times used 107 * 108 * @author Andreas Gohr <andi@splitbrain.org> 109 * @author Christopher Smith <chris@jalakai.co.uk> 110 * @author Tom N Harris <tnharris@whoopdedo.org> 111 */ 112 protected function getPageWords($text) 113 { 114 115 $tokens = $this->tokenizer($text); 116 $tokens = array_count_values($tokens); // count the frequency of each token 117 118 $words = []; 119 foreach ($tokens as $w => $c) { 120 $l = wordlen($w); 121 if (isset($words[$l])) { 122 $words[$l][$w] = $c + ($words[$l][$w] ?? 0); 123 } else { 124 $words[$l] = [$w => $c]; 125 } 126 } 127 128 // arrive here with $words = array(wordlen => array(word => frequency)) 129 $index = []; //resulting index 130 foreach (array_keys($words) as $wlen) { 131 $word_idx = $this->getIndex('w', $wlen); 132 $word_idx_modified = false; 133 foreach ($words[$wlen] as $word => $freq) { 134 $word = (string)$word; 135 $wid = array_search($word, $word_idx, true); 136 if ($wid === false) { 137 $wid = count($word_idx); 138 $word_idx[] = $word; 139 $word_idx_modified = true; 140 } 141 if (!isset($index[$wlen])) 142 $index[$wlen] = []; 143 $index[$wlen][$wid] = $freq; 144 } 145 // save back the word index 146 if ($word_idx_modified && !$this->saveIndex('w', $wlen, $word_idx)) 147 return false; 148 } 149 150 return $index; 151 } 152 153 /** 154 * Add/update keys to/of the metadata index. 155 * 156 * Adding new keys does not remove other keys for the page. 157 * An empty value will erase the key. 158 * The $key parameter can be an array to add multiple keys. $value will 159 * not be used if $key is an array. 160 * 161 * @param string $page a page name 162 * @param mixed $key a key string or array of key=>value pairs 163 * @param mixed $value the value or list of values 164 * @return boolean|string the function completed successfully 165 * 166 * @author Tom N Harris <tnharris@whoopdedo.org> 167 * @author Michael Hamann <michael@content-space.de> 168 */ 169 public function addMetaKeys($page, $key, $value = null) 170 { 171 if (!is_array($key)) { 172 $key = [$key => $value]; 173 } elseif (!is_null($value)) { 174 // $key is array, but $value is not null 175 trigger_error("array passed to addMetaKeys but value is not null", E_USER_WARNING); 176 } 177 178 if (!$this->lock()) 179 return "locked"; 180 181 // load known documents 182 $pid = $this->getPIDNoLock($page); 183 if ($pid === false) { 184 $this->unlock(); 185 return false; 186 } 187 188 // Special handling for titles so the index file is simpler 189 if (isset($key['title'])) { 190 $value = $key['title']; 191 if (is_array($value)) { 192 $value = $value[0]; 193 } 194 $this->saveIndexKey('title', '', $pid, $value); 195 unset($key['title']); 196 } 197 198 foreach ($key as $name => $values) { 199 $metaname = idx_cleanName($name); 200 $this->addIndexKey('metadata', '', $metaname); 201 $metaidx = $this->getIndex($metaname . '_i', ''); 202 $metawords = $this->getIndex($metaname . '_w', ''); 203 $addwords = false; 204 205 if (!is_array($values)) $values = [$values]; 206 207 $val_idx = $this->getIndexKey($metaname . '_p', '', $pid); 208 if ($val_idx !== '') { 209 $val_idx = explode(':', $val_idx); 210 // -1 means remove, 0 keep, 1 add 211 $val_idx = array_combine($val_idx, array_fill(0, count($val_idx), -1)); 212 } else { 213 $val_idx = []; 214 } 215 216 foreach ($values as $val) { 217 $val = (string)$val; 218 if ($val !== "") { 219 $id = array_search($val, $metawords, true); 220 if ($id === false) { 221 // didn't find $val, so we'll add it to the end of metawords and create a placeholder in metaidx 222 $id = count($metawords); 223 $metawords[$id] = $val; 224 $metaidx[$id] = ''; 225 $addwords = true; 226 } 227 // test if value is already in the index 228 if (isset($val_idx[$id]) && $val_idx[$id] <= 0) { 229 $val_idx[$id] = 0; 230 } else { // else add it 231 $val_idx[$id] = 1; 232 } 233 } 234 } 235 236 if ($addwords) { 237 $this->saveIndex($metaname . '_w', '', $metawords); 238 } 239 $vals_changed = false; 240 foreach ($val_idx as $id => $action) { 241 if ($action == -1) { 242 $metaidx[$id] = $this->updateTuple($metaidx[$id], $pid, 0); 243 $vals_changed = true; 244 unset($val_idx[$id]); 245 } elseif ($action == 1) { 246 $metaidx[$id] = $this->updateTuple($metaidx[$id], $pid, 1); 247 $vals_changed = true; 248 } 249 } 250 251 if ($vals_changed) { 252 $this->saveIndex($metaname . '_i', '', $metaidx); 253 $val_idx = implode(':', array_keys($val_idx)); 254 $this->saveIndexKey($metaname . '_p', '', $pid, $val_idx); 255 } 256 257 unset($metaidx); 258 unset($metawords); 259 } 260 261 $this->unlock(); 262 return true; 263 } 264 265 /** 266 * Rename a page in the search index without changing the indexed content. This function doesn't check if the 267 * old or new name exists in the filesystem. It returns an error if the old page isn't in the page list of the 268 * indexer and it deletes all previously indexed content of the new page. 269 * 270 * @param string $oldpage The old page name 271 * @param string $newpage The new page name 272 * @return string|bool If the page was successfully renamed, can be a message in the case of an error 273 */ 274 public function renamePage($oldpage, $newpage) 275 { 276 if (!$this->lock()) return 'locked'; 277 278 $pages = $this->getPages(); 279 280 $id = array_search($oldpage, $pages, true); 281 if ($id === false) { 282 $this->unlock(); 283 return 'page is not in index'; 284 } 285 286 $new_id = array_search($newpage, $pages, true); 287 if ($new_id !== false) { 288 // make sure the page is not in the index anymore 289 if (!$this->deletePageNoLock($newpage)) { 290 return false; 291 } 292 293 $pages[$new_id] = 'deleted:' . time() . random_int(0, 9999); 294 } 295 296 $pages[$id] = $newpage; 297 298 // update index 299 if (!$this->saveIndex('page', '', $pages)) { 300 $this->unlock(); 301 return false; 302 } 303 304 // reset the pid cache 305 $this->pidCache = []; 306 307 $this->unlock(); 308 return true; 309 } 310 311 /** 312 * Renames a meta value in the index. This doesn't change the meta value in the pages, it assumes that all pages 313 * will be updated. 314 * 315 * @param string $key The metadata key of which a value shall be changed 316 * @param string $oldvalue The old value that shall be renamed 317 * @param string $newvalue The new value to which the old value shall be renamed, if exists values will be merged 318 * @return bool|string If renaming the value has been successful, false or error message on error. 319 */ 320 public function renameMetaValue($key, $oldvalue, $newvalue) 321 { 322 if (!$this->lock()) return 'locked'; 323 324 // change the relation references index 325 $metavalues = $this->getIndex($key, '_w'); 326 $oldid = array_search($oldvalue, $metavalues, true); 327 if ($oldid !== false) { 328 $newid = array_search($newvalue, $metavalues, true); 329 if ($newid !== false) { 330 // free memory 331 unset($metavalues); 332 333 // okay, now we have two entries for the same value. we need to merge them. 334 $indexline = $this->getIndexKey($key . '_i', '', $oldid); 335 if ($indexline != '') { 336 $newindexline = $this->getIndexKey($key . '_i', '', $newid); 337 $pagekeys = $this->getIndex($key . '_p', ''); 338 $parts = explode(':', $indexline); 339 foreach ($parts as $part) { 340 [$id, $count] = explode('*', $part); 341 $newindexline = $this->updateTuple($newindexline, $id, $count); 342 343 $keyline = explode(':', $pagekeys[$id]); 344 // remove old meta value 345 $keyline = array_diff($keyline, [$oldid]); 346 // add new meta value when not already present 347 if (!in_array($newid, $keyline)) { 348 $keyline[] = $newid; 349 } 350 $pagekeys[$id] = implode(':', $keyline); 351 } 352 $this->saveIndex($key . '_p', '', $pagekeys); 353 unset($pagekeys); 354 $this->saveIndexKey($key . '_i', '', $oldid, ''); 355 $this->saveIndexKey($key . '_i', '', $newid, $newindexline); 356 } 357 } else { 358 $metavalues[$oldid] = $newvalue; 359 if (!$this->saveIndex($key . '_w', '', $metavalues)) { 360 $this->unlock(); 361 return false; 362 } 363 } 364 } 365 366 $this->unlock(); 367 return true; 368 } 369 370 /** 371 * Remove a page from the index 372 * 373 * Erases entries in all known indexes. 374 * 375 * @param string $page a page name 376 * @return string|boolean the function completed successfully 377 * 378 * @author Tom N Harris <tnharris@whoopdedo.org> 379 */ 380 public function deletePage($page) 381 { 382 if (!$this->lock()) 383 return "locked"; 384 385 $result = $this->deletePageNoLock($page); 386 387 $this->unlock(); 388 389 return $result; 390 } 391 392 /** 393 * Remove a page from the index without locking the index, only use this function if the index is already locked 394 * 395 * Erases entries in all known indexes. 396 * 397 * @param string $page a page name 398 * @return boolean the function completed successfully 399 * 400 * @author Tom N Harris <tnharris@whoopdedo.org> 401 */ 402 protected function deletePageNoLock($page) 403 { 404 // load known documents 405 $pid = $this->getPIDNoLock($page); 406 if ($pid === false) { 407 return false; 408 } 409 410 // Remove obsolete index entries 411 $pageword_idx = $this->getIndexKey('pageword', '', $pid); 412 if ($pageword_idx !== '') { 413 $delwords = explode(':', $pageword_idx); 414 $upwords = []; 415 foreach ($delwords as $word) { 416 if ($word != '') { 417 [$wlen, $wid] = explode('*', $word); 418 $wid = (int)$wid; 419 $upwords[$wlen][] = $wid; 420 } 421 } 422 foreach ($upwords as $wlen => $widx) { 423 $index = $this->getIndex('i', $wlen); 424 foreach ($widx as $wid) { 425 $index[$wid] = $this->updateTuple($index[$wid], $pid, 0); 426 } 427 $this->saveIndex('i', $wlen, $index); 428 } 429 } 430 // Save the reverse index 431 if (!$this->saveIndexKey('pageword', '', $pid, "")) { 432 return false; 433 } 434 435 $this->saveIndexKey('title', '', $pid, ""); 436 $keyidx = $this->getIndex('metadata', ''); 437 foreach ($keyidx as $metaname) { 438 $val_idx = explode(':', $this->getIndexKey($metaname . '_p', '', $pid)); 439 $meta_idx = $this->getIndex($metaname . '_i', ''); 440 foreach ($val_idx as $id) { 441 if ($id === '') continue; 442 $meta_idx[$id] = $this->updateTuple($meta_idx[$id], $pid, 0); 443 } 444 $this->saveIndex($metaname . '_i', '', $meta_idx); 445 $this->saveIndexKey($metaname . '_p', '', $pid, ''); 446 } 447 448 return true; 449 } 450 451 /** 452 * Clear the whole index 453 * 454 * @return bool If the index has been cleared successfully 455 */ 456 public function clear() 457 { 458 global $conf; 459 460 if (!$this->lock()) return false; 461 462 @unlink($conf['indexdir'] . '/page.idx'); 463 @unlink($conf['indexdir'] . '/title.idx'); 464 @unlink($conf['indexdir'] . '/pageword.idx'); 465 @unlink($conf['indexdir'] . '/metadata.idx'); 466 $dir = @opendir($conf['indexdir']); 467 if ($dir !== false) { 468 while (($f = readdir($dir)) !== false) { 469 if ( 470 str_ends_with($f, '.idx') && 471 (str_starts_with($f, 'i') || 472 str_starts_with($f, 'w') || 473 str_ends_with($f, '_w.idx') || 474 str_ends_with($f, '_i.idx') || 475 str_ends_with($f, '_p.idx')) 476 ) 477 @unlink($conf['indexdir'] . "/$f"); 478 } 479 } 480 @unlink($conf['indexdir'] . '/lengths.idx'); 481 482 // clear the pid cache 483 $this->pidCache = []; 484 485 $this->unlock(); 486 return true; 487 } 488 489 /** 490 * Split the text into words for fulltext search 491 * 492 * TODO: does this also need &$stopwords ? 493 * 494 * @triggers INDEXER_TEXT_PREPARE 495 * This event allows plugins to modify the text before it gets tokenized. 496 * Plugins intercepting this event should also intercept INDEX_VERSION_GET 497 * 498 * @param string $text plain text 499 * @param boolean $wc are wildcards allowed? 500 * @return array list of words in the text 501 * 502 * @author Tom N Harris <tnharris@whoopdedo.org> 503 * @author Andreas Gohr <andi@splitbrain.org> 504 */ 505 public function tokenizer($text, $wc = false) 506 { 507 $wc = ($wc) ? '' : '\*'; 508 $stopwords =& idx_get_stopwords(); 509 510 // prepare the text to be tokenized 511 $evt = new Event('INDEXER_TEXT_PREPARE', $text); 512 if ($evt->advise_before(true)) { 513 if (preg_match('/[^0-9A-Za-z ]/u', $text)) { 514 $text = Asian::separateAsianWords($text); 515 } 516 } 517 $evt->advise_after(); 518 unset($evt); 519 520 $text = strtr( 521 $text, 522 ["\r" => ' ', "\n" => ' ', "\t" => ' ', "\xC2\xAD" => ''] 523 ); 524 if (preg_match('/[^0-9A-Za-z ]/u', $text)) 525 $text = Clean::stripspecials($text, ' ', '\._\-:' . $wc); 526 527 $wordlist = explode(' ', $text); 528 foreach ($wordlist as $i => $word) { 529 $wordlist[$i] = (preg_match('/[^0-9A-Za-z]/u', $word)) ? 530 PhpString::strtolower($word) : strtolower($word); 531 } 532 533 foreach ($wordlist as $i => $word) { 534 if ( 535 (!is_numeric($word) && strlen($word) < IDX_MINWORDLENGTH) 536 || in_array($word, $stopwords, true) 537 ) 538 unset($wordlist[$i]); 539 } 540 return array_values($wordlist); 541 } 542 543 /** 544 * Get the numeric PID of a page 545 * 546 * @param string $page The page to get the PID for 547 * @return bool|int The page id on success, false on error 548 */ 549 public function getPID($page) 550 { 551 // return PID without locking when it is in the cache 552 if (isset($this->pidCache[$page])) return $this->pidCache[$page]; 553 554 if (!$this->lock()) 555 return false; 556 557 // load known documents 558 $pid = $this->getPIDNoLock($page); 559 if ($pid === false) { 560 $this->unlock(); 561 return false; 562 } 563 564 $this->unlock(); 565 return $pid; 566 } 567 568 /** 569 * Get the numeric PID of a page without locking the index. 570 * Only use this function when the index is already locked. 571 * 572 * @param string $page The page to get the PID for 573 * @return bool|int The page id on success, false on error 574 */ 575 protected function getPIDNoLock($page) 576 { 577 // avoid expensive addIndexKey operation for the most recently requested pages by using a cache 578 if (isset($this->pidCache[$page])) return $this->pidCache[$page]; 579 $pid = $this->addIndexKey('page', '', $page); 580 // limit cache to 10 entries by discarding the oldest element as in DokuWiki usually only the most recently 581 // added item will be requested again 582 if (count($this->pidCache) > 10) array_shift($this->pidCache); 583 $this->pidCache[$page] = $pid; 584 return $pid; 585 } 586 587 /** 588 * Get the page id of a numeric PID 589 * 590 * @param int $pid The PID to get the page id for 591 * @return string The page id 592 */ 593 public function getPageFromPID($pid) 594 { 595 return $this->getIndexKey('page', '', $pid); 596 } 597 598 /** 599 * Find pages in the fulltext index containing the words, 600 * 601 * The search words must be pre-tokenized, meaning only letters and 602 * numbers with an optional wildcard 603 * 604 * The returned array will have the original tokens as key. The values 605 * in the returned list is an array with the page names as keys and the 606 * number of times that token appears on the page as value. 607 * 608 * @param array $tokens list of words to search for 609 * @return array list of page names with usage counts 610 * 611 * @author Tom N Harris <tnharris@whoopdedo.org> 612 * @author Andreas Gohr <andi@splitbrain.org> 613 */ 614 public function lookup(&$tokens) 615 { 616 $result = []; 617 $wids = $this->getIndexWords($tokens, $result); 618 if (empty($wids)) return []; 619 // load known words and documents 620 $page_idx = $this->getIndex('page', ''); 621 $docs = []; 622 foreach (array_keys($wids) as $wlen) { 623 $wids[$wlen] = array_unique($wids[$wlen]); 624 $index = $this->getIndex('i', $wlen); 625 foreach ($wids[$wlen] as $ixid) { 626 if ($ixid < count($index)) 627 $docs["$wlen*$ixid"] = $this->parseTuples($page_idx, $index[$ixid]); 628 } 629 } 630 // merge found pages into final result array 631 $final = []; 632 foreach ($result as $word => $res) { 633 $final[$word] = []; 634 foreach ($res as $wid) { 635 // handle the case when ($ixid < count($index)) has been false 636 // and thus $docs[$wid] hasn't been set. 637 if (!isset($docs[$wid])) continue; 638 $hits = &$docs[$wid]; 639 foreach ($hits as $hitkey => $hitcnt) { 640 // make sure the document still exists 641 if (!page_exists($hitkey, '', false)) continue; 642 if (!isset($final[$word][$hitkey])) 643 $final[$word][$hitkey] = $hitcnt; 644 else $final[$word][$hitkey] += $hitcnt; 645 } 646 } 647 } 648 return $final; 649 } 650 651 /** 652 * Find pages containing a metadata key. 653 * 654 * The metadata values are compared as case-sensitive strings. Pass a 655 * callback function that returns true or false to use a different 656 * comparison function. The function will be called with the $value being 657 * searched for as the first argument, and the word in the index as the 658 * second argument. The function preg_match can be used directly if the 659 * values are regexes. 660 * 661 * @param string $key name of the metadata key to look for 662 * @param string $value search term to look for, must be a string or array of strings 663 * @param callback $func comparison function 664 * @return array lists with page names, keys are query values if $value is array 665 * 666 * @author Tom N Harris <tnharris@whoopdedo.org> 667 * @author Michael Hamann <michael@content-space.de> 668 */ 669 public function lookupKey($key, &$value, $func = null) 670 { 671 if (!is_array($value)) 672 $value_array = [$value]; 673 else $value_array =& $value; 674 675 // the matching ids for the provided value(s) 676 $value_ids = []; 677 678 $metaname = idx_cleanName($key); 679 680 // get all words in order to search the matching ids 681 if ($key == 'title') { 682 $words = $this->getIndex('title', ''); 683 } else { 684 $words = $this->getIndex($metaname . '_w', ''); 685 } 686 687 if (!is_null($func)) { 688 foreach ($value_array as $val) { 689 foreach ($words as $i => $word) { 690 if (call_user_func_array($func, [$val, $word])) 691 $value_ids[$i][] = $val; 692 } 693 } 694 } else { 695 foreach ($value_array as $val) { 696 $xval = $val; 697 $caret = '^'; 698 $dollar = '$'; 699 // check for wildcards 700 if (str_starts_with($xval, '*')) { 701 $xval = substr($xval, 1); 702 $caret = ''; 703 } 704 if (str_ends_with($xval, '*')) { 705 $xval = substr($xval, 0, -1); 706 $dollar = ''; 707 } 708 if (!$caret || !$dollar) { 709 $re = $caret . preg_quote($xval, '/') . $dollar; 710 foreach (array_keys(preg_grep('/' . $re . '/', $words)) as $i) 711 $value_ids[$i][] = $val; 712 } elseif (($i = array_search($val, $words, true)) !== false) { 713 $value_ids[$i][] = $val; 714 } 715 } 716 } 717 718 unset($words); // free the used memory 719 720 // initialize the result so it won't be null 721 $result = []; 722 foreach ($value_array as $val) { 723 $result[$val] = []; 724 } 725 726 $page_idx = $this->getIndex('page', ''); 727 728 // Special handling for titles 729 if ($key == 'title') { 730 foreach ($value_ids as $pid => $val_list) { 731 $page = $page_idx[$pid]; 732 foreach ($val_list as $val) { 733 $result[$val][] = $page; 734 } 735 } 736 } else { 737 // load all lines and pages so the used lines can be taken and matched with the pages 738 $lines = $this->getIndex($metaname . '_i', ''); 739 740 foreach ($value_ids as $value_id => $val_list) { 741 // parse the tuples of the form page_id*1:page2_id*1 and so on, return value 742 // is an array with page_id => 1, page2_id => 1 etc. so take the keys only 743 $pages = array_keys($this->parseTuples($page_idx, $lines[$value_id])); 744 foreach ($val_list as $val) { 745 $result[$val] = [...$result[$val], ...$pages]; 746 } 747 } 748 } 749 if (!is_array($value)) $result = $result[$value]; 750 return $result; 751 } 752 753 /** 754 * Find the index ID of each search term. 755 * 756 * The query terms should only contain valid characters, with a '*' at 757 * either the beginning or end of the word (or both). 758 * The $result parameter can be used to merge the index locations with 759 * the appropriate query term. 760 * 761 * @param array $words The query terms. 762 * @param array $result Set to word => array("length*id" ...) 763 * @return array Set to length => array(id ...) 764 * 765 * @author Tom N Harris <tnharris@whoopdedo.org> 766 */ 767 protected function getIndexWords(&$words, &$result) 768 { 769 $tokens = []; 770 $tokenlength = []; 771 $tokenwild = []; 772 foreach ($words as $word) { 773 $result[$word] = []; 774 $caret = '^'; 775 $dollar = '$'; 776 $xword = $word; 777 $wlen = wordlen($word); 778 779 // check for wildcards 780 if (str_starts_with($xword, '*')) { 781 $xword = substr($xword, 1); 782 $caret = ''; 783 --$wlen; 784 } 785 if (str_ends_with($xword, '*')) { 786 $xword = substr($xword, 0, -1); 787 $dollar = ''; 788 --$wlen; 789 } 790 if ($wlen < IDX_MINWORDLENGTH && $caret && $dollar && !is_numeric($xword)) 791 continue; 792 if (!isset($tokens[$xword])) 793 $tokenlength[$wlen][] = $xword; 794 if (!$caret || !$dollar) { 795 $re = $caret . preg_quote($xword, '/') . $dollar; 796 $tokens[$xword][] = [$word, '/' . $re . '/']; 797 if (!isset($tokenwild[$xword])) 798 $tokenwild[$xword] = $wlen; 799 } else { 800 $tokens[$xword][] = [$word, null]; 801 } 802 } 803 asort($tokenwild); 804 // $tokens = array( base word => array( [ query term , regexp ] ... ) ... ) 805 // $tokenlength = array( base word length => base word ... ) 806 // $tokenwild = array( base word => base word length ... ) 807 $length_filter = $tokenwild === [] ? $tokenlength : min(array_keys($tokenlength)); 808 $indexes_known = $this->indexLengths($length_filter); 809 if ($tokenwild !== []) sort($indexes_known); 810 // get word IDs 811 $wids = []; 812 foreach ($indexes_known as $ixlen) { 813 $word_idx = $this->getIndex('w', $ixlen); 814 // handle exact search 815 if (isset($tokenlength[$ixlen])) { 816 foreach ($tokenlength[$ixlen] as $xword) { 817 $wid = array_search($xword, $word_idx, true); 818 if ($wid !== false) { 819 $wids[$ixlen][] = $wid; 820 foreach ($tokens[$xword] as $w) 821 $result[$w[0]][] = "$ixlen*$wid"; 822 } 823 } 824 } 825 // handle wildcard search 826 foreach ($tokenwild as $xword => $wlen) { 827 if ($wlen >= $ixlen) break; 828 foreach ($tokens[$xword] as $w) { 829 if (is_null($w[1])) continue; 830 foreach (array_keys(preg_grep($w[1], $word_idx)) as $wid) { 831 $wids[$ixlen][] = $wid; 832 $result[$w[0]][] = "$ixlen*$wid"; 833 } 834 } 835 } 836 } 837 return $wids; 838 } 839 840 /** 841 * Return a list of all pages 842 * Warning: pages may not exist! 843 * 844 * @param string $key list only pages containing the metadata key (optional) 845 * @return array list of page names 846 * 847 * @author Tom N Harris <tnharris@whoopdedo.org> 848 */ 849 public function getPages($key = null) 850 { 851 $page_idx = $this->getIndex('page', ''); 852 if (is_null($key)) return $page_idx; 853 854 $metaname = idx_cleanName($key); 855 856 // Special handling for titles 857 if ($key == 'title') { 858 $title_idx = $this->getIndex('title', ''); 859 array_splice($page_idx, count($title_idx)); 860 foreach ($title_idx as $i => $title) 861 if ($title === "") unset($page_idx[$i]); 862 return array_values($page_idx); 863 } 864 865 $pages = []; 866 $lines = $this->getIndex($metaname . '_i', ''); 867 foreach ($lines as $line) { 868 $pages = array_merge($pages, $this->parseTuples($page_idx, $line)); 869 } 870 return array_keys($pages); 871 } 872 873 /** 874 * Return a list of words sorted by number of times used 875 * 876 * @param int $min bottom frequency threshold 877 * @param int $max upper frequency limit. No limit if $max<$min 878 * @param int $minlen minimum length of words to count 879 * @param string $key metadata key to list. Uses the fulltext index if not given 880 * @return array list of words as the keys and frequency as values 881 * 882 * @author Tom N Harris <tnharris@whoopdedo.org> 883 */ 884 public function histogram($min = 1, $max = 0, $minlen = 3, $key = null) 885 { 886 if ($min < 1) 887 $min = 1; 888 if ($max < $min) 889 $max = 0; 890 891 $result = []; 892 893 if ($key == 'title') { 894 $index = $this->getIndex('title', ''); 895 $index = array_count_values($index); 896 foreach ($index as $val => $cnt) { 897 if ($cnt >= $min && (!$max || $cnt <= $max) && strlen($val) >= $minlen) 898 $result[$val] = $cnt; 899 } 900 } elseif (!is_null($key)) { 901 $metaname = idx_cleanName($key); 902 $index = $this->getIndex($metaname . '_i', ''); 903 $val_idx = []; 904 foreach ($index as $wid => $line) { 905 $freq = $this->countTuples($line); 906 if ($freq >= $min && (!$max || $freq <= $max)) 907 $val_idx[$wid] = $freq; 908 } 909 if (!empty($val_idx)) { 910 $words = $this->getIndex($metaname . '_w', ''); 911 foreach ($val_idx as $wid => $freq) { 912 if (strlen($words[$wid]) >= $minlen) 913 $result[$words[$wid]] = $freq; 914 } 915 } 916 } else { 917 $lengths = idx_listIndexLengths(); 918 foreach ($lengths as $length) { 919 if ($length < $minlen) continue; 920 $index = $this->getIndex('i', $length); 921 $words = null; 922 foreach ($index as $wid => $line) { 923 $freq = $this->countTuples($line); 924 if ($freq >= $min && (!$max || $freq <= $max)) { 925 if ($words === null) 926 $words = $this->getIndex('w', $length); 927 $result[$words[$wid]] = $freq; 928 } 929 } 930 } 931 } 932 933 arsort($result); 934 return $result; 935 } 936 937 /** 938 * Lock the indexer. 939 * 940 * @author Tom N Harris <tnharris@whoopdedo.org> 941 * 942 * @return bool|string 943 */ 944 protected function lock() 945 { 946 global $conf; 947 $status = true; 948 $run = 0; 949 $lock = $conf['lockdir'] . '/_indexer.lock'; 950 while (!@mkdir($lock)) { 951 usleep(50); 952 if (is_dir($lock) && time() - @filemtime($lock) > 60 * 5) { 953 // looks like a stale lock - remove it 954 if (!@rmdir($lock)) { 955 $status = "removing the stale lock failed"; 956 return false; 957 } else { 958 $status = "stale lock removed"; 959 } 960 } elseif ($run++ == 1000) { 961 // we waited 5 seconds for that lock 962 return false; 963 } 964 } 965 if ($conf['dperm']) { 966 chmod($lock, $conf['dperm']); 967 } 968 return $status; 969 } 970 971 /** 972 * Release the indexer lock. 973 * 974 * @author Tom N Harris <tnharris@whoopdedo.org> 975 * 976 * @return bool 977 */ 978 protected function unlock() 979 { 980 global $conf; 981 @rmdir($conf['lockdir'] . '/_indexer.lock'); 982 return true; 983 } 984 985 /** 986 * Retrieve the entire index. 987 * 988 * The $suffix argument is for an index that is split into 989 * multiple parts. Different index files should use different 990 * base names. 991 * 992 * @param string $idx name of the index 993 * @param string $suffix subpart identifier 994 * @return array list of lines without CR or LF 995 * 996 * @author Tom N Harris <tnharris@whoopdedo.org> 997 */ 998 protected function getIndex($idx, $suffix) 999 { 1000 global $conf; 1001 $fn = $conf['indexdir'] . '/' . $idx . $suffix . '.idx'; 1002 if (!file_exists($fn)) return []; 1003 return file($fn, FILE_IGNORE_NEW_LINES); 1004 } 1005 1006 /** 1007 * Replace the contents of the index with an array. 1008 * 1009 * @param string $idx name of the index 1010 * @param string $suffix subpart identifier 1011 * @param array $lines list of lines without LF 1012 * @return bool If saving succeeded 1013 * 1014 * @author Tom N Harris <tnharris@whoopdedo.org> 1015 */ 1016 protected function saveIndex($idx, $suffix, &$lines) 1017 { 1018 global $conf; 1019 $fn = $conf['indexdir'] . '/' . $idx . $suffix; 1020 $fh = @fopen($fn . '.tmp', 'w'); 1021 if (!$fh) return false; 1022 fwrite($fh, implode("\n", $lines)); 1023 if (!empty($lines)) 1024 fwrite($fh, "\n"); 1025 fclose($fh); 1026 if ($conf['fperm']) 1027 chmod($fn . '.tmp', $conf['fperm']); 1028 io_rename($fn . '.tmp', $fn . '.idx'); 1029 return true; 1030 } 1031 1032 /** 1033 * Retrieve a line from the index. 1034 * 1035 * @param string $idx name of the index 1036 * @param string $suffix subpart identifier 1037 * @param int $id the line number 1038 * @return string a line with trailing whitespace removed 1039 * 1040 * @author Tom N Harris <tnharris@whoopdedo.org> 1041 */ 1042 protected function getIndexKey($idx, $suffix, $id) 1043 { 1044 global $conf; 1045 $fn = $conf['indexdir'] . '/' . $idx . $suffix . '.idx'; 1046 if (!file_exists($fn)) return ''; 1047 $fh = @fopen($fn, 'r'); 1048 if (!$fh) return ''; 1049 $ln = -1; 1050 while (($line = fgets($fh)) !== false) { 1051 if (++$ln == $id) break; 1052 } 1053 fclose($fh); 1054 return rtrim((string)$line); 1055 } 1056 1057 /** 1058 * Write a line into the index. 1059 * 1060 * @param string $idx name of the index 1061 * @param string $suffix subpart identifier 1062 * @param int $id the line number 1063 * @param string $line line to write 1064 * @return bool If saving succeeded 1065 * 1066 * @author Tom N Harris <tnharris@whoopdedo.org> 1067 */ 1068 protected function saveIndexKey($idx, $suffix, $id, $line) 1069 { 1070 global $conf; 1071 if (!str_ends_with($line, "\n")) 1072 $line .= "\n"; 1073 $fn = $conf['indexdir'] . '/' . $idx . $suffix; 1074 $fh = @fopen($fn . '.tmp', 'w'); 1075 if (!$fh) return false; 1076 $ih = @fopen($fn . '.idx', 'r'); 1077 if ($ih) { 1078 $ln = -1; 1079 while (($curline = fgets($ih)) !== false) { 1080 fwrite($fh, (++$ln == $id) ? $line : $curline); 1081 } 1082 if ($id > $ln) { 1083 while ($id > ++$ln) 1084 fwrite($fh, "\n"); 1085 fwrite($fh, $line); 1086 } 1087 fclose($ih); 1088 } else { 1089 $ln = -1; 1090 while ($id > ++$ln) 1091 fwrite($fh, "\n"); 1092 fwrite($fh, $line); 1093 } 1094 fclose($fh); 1095 if ($conf['fperm']) 1096 chmod($fn . '.tmp', $conf['fperm']); 1097 io_rename($fn . '.tmp', $fn . '.idx'); 1098 return true; 1099 } 1100 1101 /** 1102 * Retrieve or insert a value in the index. 1103 * 1104 * @param string $idx name of the index 1105 * @param string $suffix subpart identifier 1106 * @param string $value line to find in the index 1107 * @return int|bool line number of the value in the index or false if writing the index failed 1108 * 1109 * @author Tom N Harris <tnharris@whoopdedo.org> 1110 */ 1111 protected function addIndexKey($idx, $suffix, $value) 1112 { 1113 $index = $this->getIndex($idx, $suffix); 1114 $id = array_search($value, $index, true); 1115 if ($id === false) { 1116 $id = count($index); 1117 $index[$id] = $value; 1118 if (!$this->saveIndex($idx, $suffix, $index)) { 1119 trigger_error("Failed to write $idx index", E_USER_ERROR); 1120 return false; 1121 } 1122 } 1123 return $id; 1124 } 1125 1126 /** 1127 * Get the list of lengths indexed in the wiki. 1128 * 1129 * Read the index directory or a cache file and returns 1130 * a sorted array of lengths of the words used in the wiki. 1131 * 1132 * @author YoBoY <yoboy.leguesh@gmail.com> 1133 * 1134 * @return array 1135 */ 1136 protected function listIndexLengths() 1137 { 1138 return idx_listIndexLengths(); 1139 } 1140 1141 /** 1142 * Get the word lengths that have been indexed. 1143 * 1144 * Reads the index directory and returns an array of lengths 1145 * that there are indices for. 1146 * 1147 * @author YoBoY <yoboy.leguesh@gmail.com> 1148 * 1149 * @param array|int $filter 1150 * @return array 1151 */ 1152 protected function indexLengths($filter) 1153 { 1154 global $conf; 1155 $idx = []; 1156 if (is_array($filter)) { 1157 // testing if index files exist only 1158 $path = $conf['indexdir'] . "/i"; 1159 foreach (array_keys($filter) as $key) { 1160 if (file_exists($path . $key . '.idx')) 1161 $idx[] = $key; 1162 } 1163 } else { 1164 $lengths = idx_listIndexLengths(); 1165 foreach ($lengths as $length) { 1166 // keep all the values equal or superior 1167 if ((int)$length >= (int)$filter) 1168 $idx[] = $length; 1169 } 1170 } 1171 return $idx; 1172 } 1173 1174 /** 1175 * Insert or replace a tuple in a line. 1176 * 1177 * @author Tom N Harris <tnharris@whoopdedo.org> 1178 * 1179 * @param string $line 1180 * @param string|int $id 1181 * @param int $count 1182 * @return string 1183 */ 1184 protected function updateTuple($line, $id, $count) 1185 { 1186 if ($line != '') { 1187 $line = preg_replace('/(^|:)' . preg_quote($id, '/') . '\*\d*/', '', $line); 1188 } 1189 $line = trim($line, ':'); 1190 if ($count) { 1191 if ($line) { 1192 return "$id*$count:" . $line; 1193 } else { 1194 return "$id*$count"; 1195 } 1196 } 1197 return $line; 1198 } 1199 1200 /** 1201 * Split a line into an array of tuples. 1202 * 1203 * @author Tom N Harris <tnharris@whoopdedo.org> 1204 * @author Andreas Gohr <andi@splitbrain.org> 1205 * 1206 * @param array $keys 1207 * @param string $line 1208 * @return array 1209 */ 1210 protected function parseTuples(&$keys, $line) 1211 { 1212 $result = []; 1213 if ($line == '') return $result; 1214 $parts = explode(':', $line); 1215 foreach ($parts as $tuple) { 1216 if ($tuple === '') continue; 1217 [$key, $cnt] = explode('*', $tuple); 1218 if (!$cnt) continue; 1219 if (isset($keys[$key])) { 1220 $key = $keys[$key]; 1221 if ($key === false || is_null($key)) continue; 1222 } 1223 $result[$key] = $cnt; 1224 } 1225 return $result; 1226 } 1227 1228 /** 1229 * Sum the counts in a list of tuples. 1230 * 1231 * @author Tom N Harris <tnharris@whoopdedo.org> 1232 * 1233 * @param string $line 1234 * @return int 1235 */ 1236 protected function countTuples($line) 1237 { 1238 $freq = 0; 1239 $parts = explode(':', $line); 1240 foreach ($parts as $tuple) { 1241 if ($tuple === '') continue; 1242 [/* pid */, $cnt] = explode('*', $tuple); 1243 $freq += (int)$cnt; 1244 } 1245 return $freq; 1246 } 1247} 1248