1<?php 2/** 3 * Common DokuWiki functions 4 * 5 * @license GPL 2 (http://www.gnu.org/licenses/gpl.html) 6 * @author Andreas Gohr <andi@splitbrain.org> 7 */ 8 9 if(!defined('DOKU_INC')) define('DOKU_INC',realpath(dirname(__FILE__).'/../').'/'); 10 require_once(DOKU_CONF.'dokuwiki.php'); 11 require_once(DOKU_INC.'inc/io.php'); 12 require_once(DOKU_INC.'inc/utf8.php'); 13 require_once(DOKU_INC.'inc/parserutils.php'); 14 15// Asian characters are handled as words. The following regexp defines the 16// Unicode-Ranges for Asian characters 17// Ranges taken from http://en.wikipedia.org/wiki/Unicode_block 18// I'm no language expert. If you think some ranges are wrongly chosen or 19// a range is missing, please contact me 20define('IDX_ASIAN','['. 21 '\x{0E00}-\x{0E7F}'. // Thai 22 '\x{2E80}-\x{D7AF}'. // CJK -> Hangul 23 '\x{F900}-\x{FAFF}'. // CJK Compatibility Ideographs 24 '\x{FE30}-\x{FE4F}'. // CJK Compatibility Forms 25 ']'); 26 27 28/** 29 * Write a list of strings to an index file. 30 * 31 * @author Tom N Harris <tnharris@whoopdedo.org> 32 */ 33function idx_saveIndex($pre, $wlen, $idx){ 34 global $conf; 35 $fn = $conf['indexdir'].'/'.$pre.$wlen; 36 $fh = @fopen($fn.'.tmp','w'); 37 if(!$fh) return false; 38 fwrite($fh,join('',$idx)); 39 fclose($fh); 40 if($conf['fperm']) chmod($fn.'.tmp', $conf['fperm']); 41 io_rename($fn.'.tmp', $fn.'.idx'); 42 return true; 43} 44 45/** 46 * Read the list of words in an index (if it exists). 47 * 48 * @author Tom N Harris <tnharris@whoopdedo.org> 49 */ 50function idx_getIndex($pre, $wlen){ 51 global $conf; 52 $fn = $conf['indexdir'].'/'.$pre.$wlen.'.idx'; 53 if(!@file_exists($fn)) return array(); 54 return file($fn); 55} 56 57/** 58 * Create an empty index file if it doesn't exist yet. 59 * 60 * @author Tom N Harris <tnharris@whoopdedo.org> 61 */ 62function idx_touchIndex($pre, $wlen){ 63 global $conf; 64 $fn = $conf['indexdir'].'/'.$pre.$wlen.'.idx'; 65 if(!@file_exists($fn)){ 66 touch($fn); 67 if($conf['fperm']) chmod($fn, $conf['fperm']); 68 } 69} 70 71/** 72 * Split a page into words 73 * 74 * Returns an array of word counts, false if an error occured. 75 * Array is keyed on the word length, then the word index. 76 * 77 * @author Andreas Gohr <andi@splitbrain.org> 78 * @author Christopher Smith <chris@jalakai.co.uk> 79 */ 80function idx_getPageWords($page){ 81 global $conf; 82 $swfile = DOKU_INC.'inc/lang/'.$conf['lang'].'/stopwords.txt'; 83 if(@file_exists($swfile)){ 84 $stopwords = file($swfile); 85 }else{ 86 $stopwords = array(); 87 } 88 89 $body = rawWiki($page); 90 $body = strtr($body, "\r\n\t", ' '); 91 $tokens = explode(' ', $body); 92 $tokens = array_count_values($tokens); // count the frequency of each token 93 94// ensure the deaccented or romanised page names of internal links are added to the token array 95// (this is necessary for the backlink function -- there maybe a better way!) 96 if ($conf['deaccent']) { 97 $links = p_get_metadata($page,'relation references'); 98 99 if (!empty($links)) { 100 $tmp = join(' ',array_keys($links)); // make a single string 101 $tmp = strtr($tmp, ':', ' '); // replace namespace separator with a space 102 $link_tokens = array_unique(explode(' ', $tmp)); // break into tokens 103 104 foreach ($link_tokens as $link_token) { 105 if (isset($tokens[$link_token])) continue; 106 $tokens[$link_token] = 1; 107 } 108 } 109 } 110 111 $words = array(); 112 foreach ($tokens as $word => $count) { 113 $arr = idx_tokenizer($word,$stopwords); 114 $arr = array_count_values($arr); 115 foreach ($arr as $w => $c) { 116 $l = strlen($w); 117 if(isset($words[$l])){ 118 $words[$l][$w] = $c * $count + (isset($words[$l][$w]) ? $words[$l][$w] : 0); 119 }else{ 120 $words[$l] = array($w => $c * $count); 121 } 122 } 123 } 124 125 // arrive here with $words = array(wordlen => array(word => frequency)) 126 127 $index = array(); //resulting index 128 foreach (array_keys($words) as $wlen){ 129 $word_idx = idx_getIndex('w',$wlen); 130 foreach ($words[$wlen] as $word => $freq) { 131 $wid = array_search("$word\n",$word_idx); 132 if(!is_int($wid)){ 133 $word_idx[] = "$word\n"; 134 $wid = count($word_idx)-1; 135 } 136 if(!isset($index[$wlen])) 137 $index[$wlen] = array(); 138 $index[$wlen][$wid] = $freq; 139 } 140 141 // save back word index 142 if(!idx_saveIndex('w',$wlen,$word_idx)){ 143 trigger_error("Failed to write word index", E_USER_ERROR); 144 return false; 145 } 146 } 147 148 return $index; 149} 150 151/** 152 * Adds/updates the search for the given page 153 * 154 * This is the core function of the indexer which does most 155 * of the work. This function needs to be called with proper 156 * locking! 157 * 158 * @author Andreas Gohr <andi@splitbrain.org> 159 */ 160function idx_addPage($page){ 161 global $conf; 162 163 // load known documents 164 $page_idx = idx_getIndex('page',''); 165 166 // get page id (this is the linenumber in page.idx) 167 $pid = array_search("$page\n",$page_idx); 168 if(!is_int($pid)){ 169 $page_idx[] = "$page\n"; 170 $pid = count($page_idx)-1; 171 // page was new - write back 172 if (!idx_saveIndex('page','',$page_idx)) 173 return false; 174 } 175 176 // get word usage in page 177 $words = idx_getPageWords($page); 178 if($words === false) return false; 179 if(!count($words)) return true; 180 181 foreach(array_keys($words) as $wlen){ 182 // Open index and temp file 183 $fn = $conf['indexdir']."/i$wlen"; 184 idx_touchIndex('i',$wlen); 185 $idx = fopen($fn.'.idx','r'); 186 $tmp = fopen($fn.'.tmp','w'); 187 if(!$idx || !$tmp){ 188 trigger_error("Failed to open index files", E_USER_ERROR); 189 return false; 190 } 191 192 // copy from index to temp file, modifying where needed 193 $lno = 0; 194 $line = ''; 195 while (!feof($idx)) { 196 // read full line 197 $line .= fgets($idx, 4096); 198 if(substr($line,-1) != "\n") continue; 199 200 // write a new Line to temp file 201 idx_writeIndexLine($tmp,$line,$pid,$words[$wlen][$lno]); 202 203 $line = ''; // reset line buffer 204 $lno++; // increase linecounter 205 } 206 fclose($idx); 207 208 // add missing lines (usually index and word should contain 209 // the same number of lines, however if the page contained 210 // new words the word file has some more lines which need to 211 // be added here 212 $word_idx = idx_getIndex('w',$wlen); 213 $wcnt = count($word_idx); 214 for($lno; $lno<$wcnt; $lno++){ 215 idx_writeIndexLine($tmp,'',$pid,$words[$wlen][$lno]); 216 } 217 218 // close the temp file and move it over to be the new one 219 fclose($tmp); 220 if($conf['fperm']) chmod($fn.'.tmp', $conf['fperm']); 221 // try rename first (fast) fallback to copy (slow) 222 io_rename($fn.'.tmp', $fn.'.idx'); 223 } 224 225 return true; 226} 227 228/** 229 * Write a new index line to the filehandle 230 * 231 * This function writes an line for the index file to the 232 * given filehandle. It removes the given document from 233 * the given line and readds it when $count is >0. 234 * 235 * @author Andreas Gohr <andi@splitbrain.org> 236 */ 237function idx_writeIndexLine($fh,$line,$pid,$count){ 238 $line = trim($line); 239 240 if($line != ''){ 241 $parts = explode(':',$line); 242 // remove doc from given line 243 foreach($parts as $part){ 244 if($part == '') continue; 245 list($doc,$cnt) = explode('*',$part); 246 if($doc != $pid){ 247 fwrite($fh,"$doc*$cnt:"); 248 } 249 } 250 } 251 252 // add doc 253 if ($count){ 254 fwrite($fh,"$pid*$count"); 255 } 256 257 // add newline 258 fwrite($fh,"\n"); 259} 260 261/** 262 * Get the word lengths that have been indexed. 263 * 264 * Reads the index directory and returns an array of lengths 265 * that there are indices for. 266 * 267 * @author Tom N Harris <tnharris@whoopdedo.org> 268 */ 269function idx_indexLengths($minlen){ 270 global $conf; 271 $dir = @opendir($conf['indexdir']); 272 if($dir===false) 273 return array(); 274 $idx = array(); 275 // Exact match first. 276 if(@file_exists($conf['indexdir']."/i$minlen.idx")) 277 $idx[] = $minlen; 278 while (($f = readdir($dir)) !== false) { 279 if (substr($f,0,1) == 'i' && substr($f,-4) == '.idx'){ 280 $i = substr($f,1,-4); 281 if (is_numeric($i) && $i > $minlen) 282 $idx[] = $i; 283 } 284 } 285 closedir($dir); 286 return $idx; 287} 288 289/** 290 * Lookup words in index 291 * 292 * Takes an array of word and will return a list of matching 293 * documents for each one. 294 * 295 * Important: No ACL checking is done here! All results are 296 * returned, regardless of permissions 297 * 298 * @author Andreas Gohr <andi@splitbrain.org> 299 */ 300function idx_lookup($words){ 301 global $conf; 302 303 $result = array(); 304 305 // load known words and documents 306 $page_idx = idx_getIndex('page',''); 307 308 // get word IDs 309 $wids = array(); 310 foreach($words as $word){ 311 $result[$word] = array(); 312 $wild = 0; 313 $xword = $word; 314 $wlen = strlen($word); 315 316 // check for wildcards 317 if(substr($xword,0,1) == '*'){ 318 $xword = substr($xword,1); 319 $wild = 1; 320 $ptn = '/'.preg_quote($xword,'/').'$/'; 321 $wlen -= 1; 322# $l = -1*strlen($xword)-1; 323 } 324 if(substr($xword,-1,1) == '*'){ 325 $xword = substr($xword,0,-1); 326 $wild += 2; 327 $wlen -= 1; 328 } 329 if ($wlen < 3 && $wild == 0 && !is_numeric($xword)) continue; 330 331 // look for the ID(s) for the given word 332 if($wild){ // handle wildcard search 333 foreach (idx_indexLengths($wlen) as $ixlen){ 334 $word_idx = idx_getIndex('w',$ixlen); 335 $cnt = count($word_idx); 336 for($wid=0; $wid<$cnt; $wid++){ 337 $iword = $word_idx[$wid]; 338 if( (($wild==3) && is_int(strpos($iword,$xword))) || 339# (($wild==1) && ("$xword\n" == substr($iword,$l))) || 340 (($wild==1) && preg_match($ptn,$iword)) || 341# (($wild==2) && ($xword == substr($iword,0,strlen($xword)))) 342 (($wild==2) && (0 === strpos($iword,$xword))) 343 344 ){ 345 if(!isset($wids[$ixlen])) $wids[$ixlen] = array(); 346 $wids[$ixlen][] = $wid; 347 $result[$word][] = "$ixlen*$wid"; 348 } 349 } 350 } 351 }else{ // handle exact search 352 $word_idx = idx_getIndex('w',$wlen); 353 $wid = array_search("$word\n",$word_idx); 354 if(is_int($wid)){ 355 $wids[$wlen] = array($wid); 356 $result[$word][] = "$wlen*$wid"; 357 }else{ 358 $result[$word] = array(); 359 } 360 } 361 } 362 363 $docs = array(); // hold docs found 364 foreach(array_keys($wids) as $wlen){ 365 sort($wids[$wlen]); 366 $wids[$wlen] = array_unique($wids[$wlen]); 367 368 // Open index 369 idx_touchIndex('i',$wlen); 370 $idx = fopen($conf['indexdir']."/i$wlen.idx",'r'); 371 if(!$idx){ 372 msg("Failed to open index file",-1); 373 return false; 374 } 375 376 // Walk the index til the lines are found 377 $lno = 0; 378 $line = ''; 379 $ixids =& $wids[$wlen]; 380 $srch = array_shift($ixids); // which word do we look for? 381 while (!feof($idx)) { 382 // read full line 383 $line .= fgets($idx, 4096); 384 if(substr($line,-1) != "\n") continue; 385 if($lno > $srch) break; // shouldn't happen 386 387 // do we want this line? 388 if($lno == $srch){ 389 // add docs to list 390 $docs["$wlen*$srch"] = idx_parseIndexLine($page_idx,$line); 391 392 $srch = array_shift($ixids); // next word to look up 393 if($srch == null) break; // no more words 394 } 395 396 $line = ''; // reset line buffer 397 $lno++; // increase linecounter 398 } 399 fclose($idx); 400 } 401 402 403 // merge found pages into final result array 404 $final = array(); 405 foreach(array_keys($result) as $word){ 406 $final[$word] = array(); 407 foreach($result[$word] as $wid){ 408 $hits = &$docs[$wid]; 409 foreach ($hits as $hitkey => $hitcnt) { 410 $final[$word][$hitkey] = $hitcnt + $final[$word][$hitkey]; 411 } 412 } 413 } 414 return $final; 415} 416 417/** 418 * Returns a list of documents and counts from a index line 419 * 420 * It omits docs with a count of 0 and pages that no longer 421 * exist. 422 * 423 * @param array $page_idx The list of known pages 424 * @param string $line A line from the main index 425 * @author Andreas Gohr <andi@splitbrain.org> 426 */ 427function idx_parseIndexLine(&$page_idx,$line){ 428 $result = array(); 429 430 $line = trim($line); 431 if($line == '') return $result; 432 433 $parts = explode(':',$line); 434 foreach($parts as $part){ 435 if($part == '') continue; 436 list($doc,$cnt) = explode('*',$part); 437 if(!$cnt) continue; 438 $doc = trim($page_idx[$doc]); 439 if(!$doc) continue; 440 // make sure the document still exists 441 if(!@file_exists(wikiFN($doc,'',false))) continue; 442 443 $result[$doc] = $cnt; 444 } 445 return $result; 446} 447 448/** 449 * Tokenizes a string into an array of search words 450 * 451 * Uses the same algorithm as idx_getPageWords() 452 * 453 * @param string $string the query as given by the user 454 * @param arrayref $stopwords array of stopwords 455 * @param boolean $wc are wildcards allowed? 456 * 457 * @todo make combined function to use alone or in getPageWords 458 */ 459function idx_tokenizer($string,&$stopwords,$wc=false){ 460 $words = array(); 461 $wc = ($wc) ? '' : $wc = '\*'; 462 463 if(preg_match('/[^0-9A-Za-z]/u', $string)){ 464 // handle asian chars as single words (may fail on older PHP version) 465 $asia = @preg_replace('/('.IDX_ASIAN.')/u','\1 ',$string); 466 if(!is_null($asia)) $string = $asia; //recover from regexp failure 467 468 $arr = explode(' ', utf8_stripspecials($string,' ','\._\-:'.$wc)); 469 foreach ($arr as $w) { 470 if (!is_numeric($w) && strlen($w) < 3) continue; 471 $w = utf8_strtolower($w); 472 if($stopwords && is_int(array_search("$w\n",$stopwords))) continue; 473 $words[] = $w; 474 } 475 }else{ 476 $w = $string; 477 if (!is_numeric($w) && strlen($w) < 3) return $words; 478 $w = strtolower($w); 479 if(is_int(array_search("$w\n",$stopwords))) return $words; 480 $words[] = $w; 481 } 482 483 return $words; 484} 485 486//Setup VIM: ex: et ts=4 enc=utf-8 : 487