# Turn this on if you've got memory to burn
$wgUseMemoryTables = false;
-function rebuildLinkTablesPass1()
+# Buffer this many rows before inserting them all in one sweep. More
+# than about 1000 will probably not increase speed significantly on
+# most setups.
+/* private */ $rowbuf_size = 2000; // 2000 rows ~ 80 kB
+
+function rebuildLinkTables()
{
- global $wgLang, $wgUseMemoryTables;
+ global $wgLang, $wgUseMemoryTables, $rowbuf_size;
+
+ print "This script may take many hours to complete. If you abort during that time,\n";
+ print "your wiki will be in an inconsistent state and you may have problems running\n";
+ print "this script again. If you are going to abort, this is the time to do it.\n\n";
+ print "Press control-c to abort (will proceed automatically in 15 seconds)\n";
+ sleep(15);
+
$count = 0;
- print "Rebuilding link tables (pass 1).\n";
+ print "Rebuilding link tables.\n";
- $sql = "DROP TABLE IF EXISTS rebuildlinks";
+ print "Setting AUTOCOMMIT=1\n";
+ wfQuery("SET SESSION AUTOCOMMIT=1", DB_WRITE);
+
+ print "Deleting old data in links table.\n";
+ $sql = "DELETE FROM links";
wfQuery( $sql, DB_WRITE );
- $sql = "CREATE TABLE rebuildlinks (
- rl_f_id int(8) unsigned NOT NULL default 0,
- rl_f_title varchar(255) binary NOT NULL default '',
- rl_to varchar(255) binary NOT NULL default '',
- INDEX rl_to (rl_to) )";
- if( $wgUseMemoryTables ) $sql .= " TYPE=heap";
+ print "Deleting old data in brokenlinks table.\n";
+ $sql = "DELETE FROM brokenlinks";
+ wfQuery( $sql, DB_WRITE );
+
+ print "Deleting old data in imagelinks table.\n";
+ $sql = "DELETE FROM imagelinks";
wfQuery( $sql, DB_WRITE );
- $sql = "LOCK TABLES cur READ, rebuildlinks WRITE, interwiki READ, user_newtalk READ";
+ print "\nAdding temporary unique index on links, brokenlinks and imagelinks.\n";
+ print "->If build aborts now, you probably aborted a previous build. If that is\n";
+ print " the case, you can clean up the remains with the following SQL commands,\n";
+ print " and then try again.\n";
+ print " ALTER TABLE links DROP INDEX tmp_unique;\n";
+ print " ALTER TABLE brokenlinks DROP INDEX tmp_unique;\n";
+ print " ALTER TABLE imagelinks DROP INDEX tmp_unique;\n\n";
+
+ $sql = "ALTER TABLE links ADD UNIQUE tmp_unique (l_from, l_to)";
+ wfQuery( $sql, DB_WRITE );
+ $sql = "ALTER TABLE brokenlinks ADD UNIQUE tmp_unique (bl_from, bl_to)";
+ wfQuery( $sql, DB_WRITE );
+ $sql = "ALTER TABLE imagelinks ADD UNIQUE tmp_unique (il_from, il_to(244))";
wfQuery( $sql, DB_WRITE );
+ print "Temporary unique index added ok. Forget what I said.\n\n";
- $sql = "DELETE FROM rebuildlinks";
+ print "Locking tables\n";
+ $sql = "LOCK TABLES cur READ, interwiki READ, user_newtalk READ, " .
+ "links WRITE, brokenlinks WRITE, imagelinks WRITE";
wfQuery( $sql, DB_WRITE );
- $sql = "SELECT cur_id,cur_namespace,cur_title,cur_text FROM cur";
- $res = wfQuery( $sql, DB_WRITE );
- $total = wfNumRows( $res );
+ print "Finding number of articles to process\n";
+ $sql = "SELECT COUNT(*) as count FROM cur";
+ $res = wfQuery( $sql, DB_READ );
+ $obj = wfFetchObject( $res );
+ $total = $obj->count;
+
+ print "Finding highest article id\n";
+ $sql = "SELECT MIN(cur_id) AS min, MAX(cur_id) AS max FROM cur";
+ $res = wfQuery( $sql, DB_READ );
+ $obj = wfFetchObject( $res );
+
+ $cur_pulser = new SelectPulser("SELECT cur_id,cur_namespace,cur_title,cur_text " .
+ "FROM cur WHERE cur_id ",
+ $obj->min, $obj->max, $rowbuf_size);
+
+ $brokenlinks_inserter = new InsertBuffer(
+ "INSERT IGNORE INTO brokenlinks (bl_from,bl_to) VALUES " , $rowbuf_size);
+
+ $links_inserter = new InsertBuffer(
+ "INSERT IGNORE INTO links (l_from,l_to) VALUES ", $rowbuf_size);
+
+ $imagelinks_inserter = new InsertBuffer("INSERT IGNORE INTO imagelinks ".
+ "(il_from,il_to) VALUES ", $rowbuf_size);
+
+ print "Starting processing\n";
+
+ $ins = $wgLang->getNsText( Namespace::getImage() );
+ $inslen = strlen($ins)+1;
$tc = Title::legalChars();
- while ( $row = wfFetchObject( $res ) ) {
- $id = $row->cur_id;
+
+ $start_time = time();
+ while ( $row = $cur_pulser->next() ) {
+ $from_id = $row->cur_id;
$ns = $wgLang->getNsText( $row->cur_namespace );
- if ( "" == $ns ) {
- $title = addslashes( $row->cur_title );
- } else {
- $title = addslashes( "$ns:{$row->cur_title}" );
+
+ $raw_title = $row->cur_title;
+ if ( "" != $ns ) {
+ $raw_title = "$ns:{$raw_title}";
}
+ $title = addslashes( $raw_title );
$text = $row->cur_text;
+
$numlinks = preg_match_all( "/\\[\\[([{$tc}]+)(]|\\|)/", $text,
$m, PREG_PATTERN_ORDER );
- if ( 0 != $numlinks ) {
- $first = true;
- $sql = "INSERT INTO rebuildlinks (rl_f_id,rl_f_title,rl_to) VALUES ";
- for ( $i = 0; $i < $numlinks; ++$i ) {
- if( preg_match( '/^(http|https|ftp|mailto|news):/', $m[1][$i] ) ) {
- # an URL link; not for us!
- continue;
- }
- # FIXME: Handle subpage links
- $nt = Title::newFromText( $m[1][$i] );
- if (! $nt)
- {
- $txt = $m[1][$i];
- print "error in '$ns:{$row->cur_title}' :\t'$txt'\n";
- continue;
- }
- if( $nt->getInterwiki() != "" ) {
- # Interwiki links are not stored in the link tables
- continue;
- }
- if( $nt->getNamespace() == Namespace::getSpecial() ) {
- # Special links not stored in link tables
- continue;
- }
- if( $nt->getNamespace() == Namespace::getMedia() ) {
- # treat media: links as image: links
- $nt = Title::makeTitle( Namespace::getImage(), $nt->getDBkey() );
- }
-
- if (!$first)
- $sql .= ",";
- else
- $first = false;
-
- $dest = addslashes( $nt->getPrefixedDBkey() );
- $sql .= "({$id},'{$title}','{$dest}')";
+ for ( $i = 0; $i < $numlinks; ++$i ) {
+ if( preg_match( '/^(http|https|ftp|mailto|news):/', $m[1][$i] ) ) {
+ # an URL link; not for us!
+ continue;
+ }
+
+ # FIXME: Handle subpage links
+ $nt = Title::newFromText( $m[1][$i] );
+
+ if (! $nt)
+ {
+ $txt = $m[1][$i];
+ print "error in '$ns:{$row->cur_title}' :\t'$txt'\n";
+ continue;
+ }
+ if( $nt->getInterwiki() != "" ) {
+ # Interwiki links are not stored in the link tables
+ continue;
}
+ if( $nt->getNamespace() == Namespace::getSpecial() ) {
+ # Special links not stored in link tables
+ continue;
+ }
+ if( $nt->getNamespace() == Namespace::getMedia() ) {
+ # treat media: links as image: links
+ $nt = Title::makeTitle( Namespace::getImage(), $nt->getDBkey() );
+ }
+
+ $dest = addslashes( $nt->getPrefixedDBkey() );
+ $dest_id = $nt->getArticleID();
- if (! $first) { wfQuery( $sql, DB_WRITE ); }
+ if ( 0 == strncmp( "$ins:", $raw_title, $inslen ) ) {
+ $iname = addslashes( substr( $raw_title, $inslen ) );
+ $imagelinks_inserter->insert( "('{$title}','{$iname}')" );
+ } else if ( 0 == $dest_id ) {
+ $brokenlinks_inserter->insert( "({$from_id},'{$dest}')" );
+ } else {
+ $links_inserter->insert( "('{$title}',{$dest_id})" );
+ }
}
+
+ if ( ( $count % 10 ) == 0 )
+ print ".";
+
if ( ( ++$count % 1000 ) == 0 ) {
- print "$count of $total articles scanned.\n";
+ $dt = time() - $start_time;
+ $start_time = time();
+ $rps = ($dt == 0 ? "lots of" : intval(1000/$dt));
+ print "\n$count of $total articles scanned ({$rps} articles per second)\n";
}
+
}
- print "$total articles scanned.\n";
- mysql_free_result( $res );
- $sql = "UNLOCK TABLES";
- wfQuery( $sql, DB_WRITE );
-}
+ $imagelinks_inserter->flush();
+ $links_inserter->flush();
+ $brokenlinks_inserter->flush();
-function rebuildLinkTablesPass2()
-{
- global $wgLang;
- $count = 0;
- print "Rebuilding link tables (pass 2).\n";
+ print "$total articles scanned.\n";
- $sql = "LOCK TABLES cur READ, rebuildlinks READ, interwiki READ, " .
- "links WRITE, brokenlinks WRITE, imagelinks WRITE";
+ print "Removing temporary unique indexes from tables links, brokenlinks and imagelinks.\n";
+ $sql = "ALTER TABLE links DROP INDEX tmp_unique";
wfQuery( $sql, DB_WRITE );
-
- $sql = "DELETE FROM links";
+ $sql = "ALTER TABLE brokenlinks DROP INDEX tmp_unique";
wfQuery( $sql, DB_WRITE );
-
- $sql = "DELETE FROM brokenlinks";
+ $sql = "ALTER TABLE imagelinks DROP INDEX tmp_unique";
wfQuery( $sql, DB_WRITE );
- $sql = "DELETE FROM imagelinks";
+ $sql = "UNLOCK TABLES";
wfQuery( $sql, DB_WRITE );
+ print "Done\n";
+}
- $ins = $wgLang->getNsText( Namespace::getImage() );
- $inslen = strlen($ins)+1;
- $sql = "SELECT rl_f_title,rl_to FROM rebuildlinks " .
- "WHERE rl_to LIKE '$ins:%'";
- $res = wfQuery( $sql, DB_WRITE );
-
- $sql = "INSERT INTO imagelinks (il_from,il_to) VALUES ";
- $first = true;
- while ( $row = wfFetchObject( $res ) )
- {
- $iname = addslashes( substr( $row->rl_to, $inslen ) );
- $pname = addslashes( $row->rl_f_title );
-
- if ( ! $first )
- $sql .= ",";
- else
- $first = false;
-
- $sql .= "('{$pname}','{$iname}')";
+# InsertBuffer increases performance slightly by inserting many rows
+# at once. The gain is small (<5%) when running against a local, idle
+# database, but may be significant in other circumstances. It also
+# limits the number of inserted rows uppwards, which should avoid
+# problems with huge articles and certain mysql settings that limits
+# the size of queries. It's also convenient.
+
+class InsertBuffer {
+ /* private */ var $mBuf, $mSql, $mBufcount, $mMaxsize;
+
+ function InsertBuffer( $sql, $bufsize ){
+ $this->mSql = $sql;
+ $this->mBuf = array();
+ $this->mBufcount = 0;
+ $this->mMaxsize = $bufsize;
}
- wfFreeResult( $res );
- if ( ! $first ) { wfQuery( $sql, DB_WRITE ); }
- $sql = "SELECT DISTINCT rl_to FROM rebuildlinks ORDER BY rl_to";
- $res = wfQuery( $sql, DB_WRITE );
- $count = 0;
- $total = wfNumRows( $res );
-
- while ( $row = wfFetchObject( $res ) ) {
- if ( 0 == strncmp( "$ins:", $row->rl_to, $inslen ) ) { continue; }
-
- $nt = Title::newFromDBkey( $row->rl_to );
- if (! $nt)
- {
- print "error pass2: '{$row->rl_to}'\n";
- continue;
+ function insert( $value ){
+ // print $this->mSql . " -> " . $value . "\n";
+ $this->mBuf[] = $value;
+ $this->mBufcount++;
+ if($this->mBufcount > $this->mMaxsize){
+ $this->flush();
}
- $id = $nt->getArticleID();
- $to = addslashes( $row->rl_to );
-
- if ( 0 == $id ) {
- $sql = "SELECT DISTINCT rl_f_id FROM rebuildlinks WHERE rl_to='{$to}'";
- $res2 = wfQuery( $sql, DB_WRITE );
+ }
- $sql = "INSERT INTO brokenlinks (bl_from,bl_to) VALUES ";
- $first = true;
- while ( $row2 = wfFetchObject( $res2 ) )
- {
- if (! $first)
- $sql .= ",";
- else
- $first = false;
+ function flush(){
+ if( $this->mBufcount > 0 ){
+ $sql = $this->mSql . implode(",", $this->mBuf);
+ wfQuery( $sql, DB_WRITE );
+ $this->mBuf = array();
+ $this->mBufcount = 0;
+ // print "Wrote query of size " . strlen( $sql ) . "\n";
+ }
+ }
+
+}
- $from = $row2->rl_f_id;
- $sql .= "({$from},'{$to}')";
+# Select parts from a large table by using the "BETWEEN X AND Y"
+# operator on the id column. Avoids buffering the whole thing in
+# RAM. It's also convenient.
+class SelectPulser {
+ /* private */ var $mSql, $mSetsize, $mPos, $mMax, $mSet;
+
+ function SelectPulser( $sql, $min, $max, $setsize) {
+ $this->mSql = $sql;
+ $this->mSet = array();
+ $this->mPos = $min;
+ $this->mMax = $max;
+ $this->mSetsize = $setsize;
+ }
+
+ function next(){
+ $result = current( $this->mSet );
+ next( $this->mSet );
+ if( false !== $result ){
+ return $result;
+ }
+ while( $this->mPos <= $this->mMax ){
+ $this->mSet = array();
+ $sql = $this->mSql . " BETWEEN " . $this->mPos .
+ " AND " . ($this->mPos + $this->mSetsize - 1);
+ $this->mPos += $this->mSetsize;
+
+ $res = wfQuery( $sql, DB_READ );
+ while ( $row = wfFetchObject( $res ) ) {
+ $this->mSet[] = $row;
}
- wfFreeResult( $res2 );
- if ( ! $first ) { wfQuery( $sql, DB_WRITE ); }
- } else {
- $sql = "SELECT DISTINCT rl_f_title FROM rebuildlinks WHERE rl_to='{$to}'";
- $res2 = wfQuery( $sql, DB_WRITE );
-
- $sql = "INSERT INTO links (l_from,l_to) VALUES ";
- $first = true;
- while ( $row2 = wfFetchObject( $res2 ) )
- {
- if (! $first)
- $sql .= ",";
- else
- $first = false;
- $from = addslashes( $row2->rl_f_title );
- $sql .= "('{$from}',{$id})";
- }
- wfFreeResult( $res2 );
- if ( ! $first ) { wfQuery( $sql, DB_WRITE ); }
- }
- if ( ( ++$count % 1000 ) == 0 ) {
- print "$count of $total titles processed.\n";
+ if( count( $this->mSet ) > 0 ){
+ return $this->next();
+ }
}
+ return false;
}
- wfFreeResult( $res );
-
- $sql = "UNLOCK TABLES";
- wfQuery( $sql, DB_WRITE );
-
- $sql = "DROP TABLE rebuildlinks";
- wfQuery( $sql, DB_WRITE );
}
+
?>