maintenance/rebuildlinks.inc

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