maintenance/rebuildlinks.inc

   1 <?php
   2
   3 # Functions for rebuilding the link tracking tables; must
   4 # be included within a script that also includes the Setup.
   5 # See rebuildlinks.php, for example.
   6 #
   7
   8 # Buffer this many rows before inserting them all in one sweep. More
   9 # than about 1000 will probably not increase speed significantly on
  10 # most setups.
  11 /* private */ $rowbuf_size = 1000; // 1000 rows ~40 kB
  12
  13 function rebuildLinkTables()
  14 {
  15         error_reporting (E_ALL);
  16         global $wgLang, $wgLinkCache, $rowbuf_size;
  17
  18         print "This script may take several hours to complete. If you abort during that time,\n";
  19         print "your wiki will be in an inconsistent state. If you are going to abort, this is\n";
  20         print "the time to do it.\n\n";
  21         print "Press control-c to abort (will proceed automatically in 15 seconds)\n";
  22         sleep(15);
  23
  24         $count = 0;
  25         print "Rebuilding link tables.\n";
  26
  27         print "Setting AUTOCOMMIT=1\n";
  28         wfQuery("SET SESSION AUTOCOMMIT=1", DB_WRITE);
  29
  30         print "Extracting often used data from cur (may take a few minutes)\n";
  31         $sql = "CREATE TEMPORARY TABLE cur_fast SELECT cur_namespace, cur_title, cur_id FROM cur";
  32         wfQuery( $sql, DB_WRITE );
  33         $sql = "ALTER TABLE cur_fast ADD INDEX(cur_namespace, cur_title)";
  34         wfQuery( $sql, DB_WRITE );
  35
  36         print "Locking tables\n";
  37         $sql = "LOCK TABLES cur READ, cur_fast READ, interwiki READ, user_newtalk READ, " .
  38                 "links WRITE, brokenlinks WRITE, imagelinks WRITE";
  39         wfQuery( $sql, DB_WRITE );
  40
  41
  42         print "Deleting old data in links table.\n";
  43         $sql = "DELETE FROM links";
  44         wfQuery( $sql, DB_WRITE );
  45
  46         print "Deleting old data in brokenlinks table.\n";
  47         $sql = "DELETE FROM brokenlinks";
  48         wfQuery( $sql, DB_WRITE );
  49
  50         print "Deleting old data in imagelinks table.\n";
  51         $sql = "DELETE FROM imagelinks";
  52         wfQuery( $sql, DB_WRITE );
  53
  54         print "Finding number of articles to process... ";
  55         $sql = "SELECT COUNT(*) as count FROM cur";
  56         $res = wfQuery( $sql, DB_READ );
  57         $obj = wfFetchObject( $res );
  58         $total = $obj->count;
  59         print "$total\n";
  60
  61         print "Finding highest article id\n";
  62         $sql = "SELECT MIN(cur_id) AS min, MAX(cur_id) AS max FROM cur";
  63         $res = wfQuery( $sql, DB_READ );
  64         $obj = wfFetchObject( $res );
  65
  66         $cur_pulser = new SelectPulser("SELECT cur_id,cur_namespace,cur_title,cur_text " .
  67                 "FROM cur WHERE cur_id ",
  68                 $obj->min, $obj->max, 100);
  69
  70         $brokenlinks_inserter = new InsertBuffer(
  71                 "INSERT IGNORE INTO brokenlinks (bl_from,bl_to) VALUES " , $rowbuf_size);
  72
  73         $links_inserter = new InsertBuffer(
  74                 "INSERT IGNORE INTO links (l_from,l_to) VALUES ", $rowbuf_size);
  75
  76         $imagelinks_inserter = new InsertBuffer("INSERT IGNORE INTO imagelinks ".
  77                 "(il_from,il_to) VALUES ", $rowbuf_size);
  78
  79         print "Starting processing\n";
  80
  81         $ins = $wgLang->getNsText( Namespace::getImage() );
  82         $inslen = strlen($ins)+1;
  83
  84         $tc = Title::legalChars();
  85
  86         $titleCache = new MRUCache( 10000 );
  87         $titlecount = 0;
  88         $start_time = time();
  89
  90         while ( $row = $cur_pulser->next() ) {
  91
  92                 $from_id = intval($row->cur_id);
  93                 $ns = $wgLang->getNsText( $row->cur_namespace );
  94                 $from_full_title = $row->cur_title;
  95                 if ( "" != $ns ) {
  96                         $from_full_title = "$ns:{$from_full_title}";
  97                 }
  98                 $from_full_title_with_slashes = addslashes( $from_full_title );
  99                 $text = $row->cur_text;
 100
 101                 $numlinks = preg_match_all( "/\\[\\[([{$tc}]+)(]|\\|)/", $text,
 102                   $m, PREG_PATTERN_ORDER );
 103
 104                 $seen_dbtitles = array(); // seen links (normalized and with ns, see below)
 105                 $titles_ready_for_insertion = array();
 106                 $titles_needing_curdata = array();
 107                 $titles_needing_curdata_pos = array();
 108                 $links_corresponding_to_titles = array();
 109
 110                 for ( $i = 0 ; $i < $numlinks; ++$i ) {
 111                         $link = $m[1][$i];
 112                         if( preg_match( '/^(http|https|ftp|mailto|news):/', $m[1][$i] ) ) {
 113                                 # an URL link; not for us!
 114                                 continue;
 115                         }
 116
 117                         # FIXME: Handle subpage links
 118                         $nt = $titleCache->get( $link );
 119                         if( $nt != false ){
 120                                 // Only process each unique link once per page
 121                                 $nt_key = $nt->getDBkey() . $nt->getNamespace();
 122                                 if( isset( $seen_dbtitles[$nt_key] ) )
 123                                         continue;
 124                                 $seen_dbtitles[$nt_key] = 1;
 125
 126                                 $titles_ready_for_insertion[] = $nt;
 127                         } else {
 128                                 $nt = Title::newFromText( $link );
 129                                 if (! $nt) {
 130                                         // Invalid link, probably something like "[[  ]]"
 131                                         continue;
 132                                 }
 133
 134                                 // Only process each unique link once per page
 135                                 $nt_key = $nt->getDBkey() . $nt->getNamespace();
 136                                 if( isset( $seen_dbtitles[$nt_key] ) )
 137                                         continue;
 138                                 $seen_dbtitles[$nt_key] = 1;
 139
 140                                 if( $nt->getInterwiki() != "" ) {
 141                                         # Interwiki links are not stored in the link tables
 142                                         continue;
 143                                 }
 144                                 if( $nt->getNamespace() == Namespace::getSpecial() ) {
 145                                         # Special links not stored in link tables
 146                                         continue;
 147                                 }
 148                                 if( $nt->getNamespace() == Namespace::getMedia() ) {
 149                                         # treat media: links as image: links
 150                                         $nt = Title::makeTitle( Namespace::getImage(), $nt->getDBkey() );
 151                                 }
 152                                 $nt->mArticleID = 0; // assume broken link until proven otherwise
 153
 154                                 $pos = array_push($titles_needing_curdata, $nt) - 1;
 155                                 $titles_needing_curdata_pos[$nt->getDBkey() . $nt->getNamespace()] = $pos;
 156                                 $links_corresponding_to_titles[] = $link;
 157                                 unset( $link ); // useless outside this loop, but tempting
 158                         }
 159                 }
 160
 161
 162                 if ( count( $titles_needing_curdata ) > 0 ){
 163                         $parts = array();
 164                         foreach ($titles_needing_curdata as $nt ) {
 165                                 $parts[] = " (cur_namespace = " . $nt->getNamespace() . " AND " .
 166                                         "cur_title='" . wfStrencode( $nt->getDBkey() ) . "')";
 167                         }
 168                         $sql = "SELECT cur_namespace, cur_title, cur_id FROM cur_fast WHERE " .
 169                                 implode(" OR ", $parts);
 170                         $res = wfQuery( $sql, DB_WRITE );
 171                         while($row = wfFetchObject( $res ) ){
 172                                 $pos = $titles_needing_curdata_pos[$row->cur_title . $row->cur_namespace];
 173                                 $titles_needing_curdata[$pos]->mArticleID = intval($row->cur_id);
 174                         }
 175                         for( $k = 0; $k < count( $titles_needing_curdata ) ; $k++) {
 176                                 $tmplink = $links_corresponding_to_titles[$k];
 177                                 $titleCache->set( $tmplink, $titles_needing_curdata[$k] );
 178                                 $titles_ready_for_insertion[] = $titles_needing_curdata[$k];
 179                         }
 180                 }
 181
 182                 foreach ( $titles_ready_for_insertion as $nt ) {
 183                         $dest_noslashes = $nt->getPrefixedDBkey();
 184                         $dest = addslashes( $dest_noslashes );
 185                         $dest_id = $nt->getArticleID();
 186                         $from = $from_full_title_with_slashes;
 187
 188                         # print "\nLINK '$from_full_title' ($from_id) -> '$dest' ($dest_id)\n";
 189
 190                         if ( 0 == strncmp( "$ins:", $dest_noslashes, $inslen ) ) {
 191                                 $iname = addslashes( substr( $dest_noslashes, $inslen ) );
 192                                 $imagelinks_inserter->insert( "('{$from}','{$iname}')" );
 193                         } else if ( 0 == $dest_id ) {
 194                                 $brokenlinks_inserter->insert( "({$from_id},'{$dest}')" );
 195                         } else {
 196                                 $links_inserter->insert( "('{$from}',{$dest_id})" );
 197                         }
 198                         $titlecount++;
 199                 }
 200
 201                 if ( ( $count % 20 ) == 0 )
 202                         print ".";
 203
 204                 if ( ( ++$count % 1000 ) == 0 ) {
 205                         $dt = time() - $start_time;
 206                         $start_time = time();
 207                         $rps = persec(1000, $dt);
 208                         $tps = persec($titlecount, $dt);
 209                         $titlecount = 0;
 210                         print "\n$count of $total articles scanned ({$rps} articles ".
 211                                 "and {$tps} titles per second)\n";
 212                         print "Title cache hits: " . $titleCache->getPerformance() . "%\n";
 213
 214                 }
 215
 216         }
 217
 218         print "\nFlushing insertion buffers...";
 219         $imagelinks_inserter->flush();
 220         $links_inserter->flush();
 221         $brokenlinks_inserter->flush();
 222         print "ok\n";
 223
 224         print "$count articles scanned.\n";
 225
 226         $sql = "UNLOCK TABLES";
 227         wfQuery( $sql, DB_WRITE );
 228         print "Done\n";
 229 }
 230
 231 /* private */ function persec($n, $t){
 232         if($n == 0)
 233                 return "zero";
 234         if($t == 0)
 235                 return "lots of";
 236         return intval($n/$t);
 237 }
 238
 239 # InsertBuffer increases performance slightly by inserting many rows
 240 # at once. The gain is small (<5%) when running against a local, idle
 241 # database, but may be significant in other circumstances. It also
 242 # limits the number of inserted rows uppwards, which should avoid
 243 # problems with huge articles and certain mysql settings that limits
 244 # the size of queries. It's also convenient.
 245
 246 class InsertBuffer {
 247         /* private */ var $mBuf, $mSql, $mBufcount, $mMaxsize;
 248
 249         function InsertBuffer( $sql, $bufsize ){
 250                 $this->mSql = $sql;
 251                 $this->mBuf = array();
 252                 $this->mBufcount = 0;
 253                 $this->mMaxsize = $bufsize;
 254         }
 255
 256         function insert( $value ){
 257                 // print $this->mSql . " -> " . $value . "\n";
 258                 $this->mBuf[] = $value;
 259                 $this->mBufcount++;
 260                 if($this->mBufcount > $this->mMaxsize){
 261                         $this->flush();
 262                 }
 263         }
 264
 265         function flush(){
 266                 if( $this->mBufcount > 0 ){
 267                         $sql = $this->mSql . implode(",", $this->mBuf);
 268                         wfQuery( $sql, DB_WRITE );
 269                         $this->mBuf = array();
 270                         $this->mBufcount = 0;
 271                         // print "Wrote query of size " . strlen( $sql ) . "\n";
 272                 }
 273         }
 274
 275 }
 276
 277 # Select parts from a large table by using the "BETWEEN X AND Y"
 278 # operator on the id column. Avoids buffering the whole thing in
 279 # RAM. It's also convenient.
 280
 281 class SelectPulser {
 282         /* private */ var $mSql, $mSetsize, $mPos, $mMax, $mSet;
 283
 284         function SelectPulser( $sql, $min, $max, $setsize) {
 285                 $this->mSql = $sql;
 286                 $this->mSet = array();
 287                 $this->mPos = $min;
 288                 $this->mMax = $max;
 289                 $this->mSetsize = $setsize;
 290         }
 291
 292         function next(){
 293                 $result = current( $this->mSet );
 294                 next( $this->mSet );
 295                 if( false !== $result ){
 296                         return $result;
 297                 }
 298                 while( $this->mPos <= $this->mMax ){
 299                         $this->mSet = array();
 300                         $sql = $this->mSql . " BETWEEN " . $this->mPos .
 301                                 " AND " . ($this->mPos + $this->mSetsize - 1);
 302                         $this->mPos += $this->mSetsize;
 303
 304                         $res = wfQuery( $sql, DB_READ );
 305                         while ( $row = wfFetchObject( $res ) ) {
 306                                 $this->mSet[] = $row;
 307                         }
 308                         wfFreeResult( $res );
 309                         if( count( $this->mSet ) > 0 ){
 310                                 return $this->next();
 311                         }
 312                 }
 313                 return false;
 314         }
 315 }
 316
 317 # A simple MRU for general cacheing.
 318
 319 class MRUCache {
 320         /* private */ var $mMru, $mCache, $mSize, $mPurgefreq, $nexti;
 321         /* private */ var $hits, $misses;
 322
 323         function MRUCache( $size, $purgefreq = -1 ) {
 324                 // purgefreq is 1/10 of $size if not stated
 325                 $purgefreq = ($purgefreq == -1 ? intval($size/10) : $purgefreq);
 326                 $purgefreq = ($purgefreq <= 0 ? 1 : $purgefreq);
 327
 328                 $this->mSize = $size;
 329                 $this->mMru = array();
 330                 $this->mCache = array();
 331                 $this->mPurgefreq = $purgefreq;
 332                 $this->nexti = 1;
 333                 print "purgefreq = " . $this->mPurgefreq . "\n";
 334         }
 335
 336         function get( $key ){
 337                 if ( ! array_key_exists( $key, $this->mCache) ){
 338                         $this->misses++;
 339                         return false;
 340                 }
 341                 $this->hits++;
 342                 $this->mMru[$key] = $this->nexti++;
 343                 return $this->mCache[$key];
 344         }
 345
 346         function set( $key, $value ){
 347                 $this->mMru[$key] = $this->nexti++;
 348                 $this->mCache[$key] = $value;
 349
 350                 if($this->nexti % $this->mPurgefreq == 0)
 351                         $this->purge();
 352         }
 353
 354         function purge(){
 355                 $to_remove = count( $this->mMru ) - $this->mSize;
 356                 if( $to_remove <= 0 ){
 357                         return;
 358                 }
 359                 asort( $this->mMru );
 360                 $removed = array_splice( $this->mMru, 0, $to_remove );
 361                 foreach( array_keys( $removed ) as $key ){
 362                         unset( $this->mCache[$key] );
 363                 }
 364         }
 365
 366         function getPerformance(){
 367                 $tot = $this->hits + $this->misses;
 368                 if($tot > 0)
 369                         return intval(100.0 * $this->hits / $tot);
 370                 else
 371                         return 0;
 372         }
 373 }
 374
 375 ?>