maintenance/backupPrefetch.inc

   1 <?php
   2 /**
   3  * Helper class for the --prefetch option of dumpTextPass.php
   4  *
   5  * Copyright © 2005 Brion Vibber <brion@pobox.com>
   6  * https://www.mediawiki.org/
   7  *
   8  * This program is free software; you can redistribute it and/or modify
   9  * it under the terms of the GNU General Public License as published by
  10  * the Free Software Foundation; either version 2 of the License, or
  11  * (at your option) any later version.
  12  *
  13  * This program is distributed in the hope that it will be useful,
  14  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  15  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
  16  * GNU General Public License for more details.
  17  *
  18  * You should have received a copy of the GNU General Public License along
  19  * with this program; if not, write to the Free Software Foundation, Inc.,
  20  * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
  21  * http://www.gnu.org/copyleft/gpl.html
  22  *
  23  * @file
  24  * @ingroup Maintenance
  25  */
  26
  27 /**
  28  * Readahead helper for making large MediaWiki data dumps;
  29  * reads in a previous XML dump to sequentially prefetch text
  30  * records already normalized and decompressed.
  31  *
  32  * This can save load on the external database servers, hopefully.
  33  *
  34  * Assumes that dumps will be recorded in the canonical order:
  35  * - ascending by page_id
  36  * - ascending by rev_id within each page
  37  * - text contents are immutable and should not change once
  38  *   recorded, so the previous dump is a reliable source
  39  *
  40  * @ingroup Maintenance
  41  */
  42 class BaseDump {
  43         /** @var XMLReader */
  44         protected $reader = null;
  45         protected $atEnd = false;
  46         protected $atPageEnd = false;
  47         protected $lastPage = 0;
  48         protected $lastRev = 0;
  49         protected $infiles = null;
  50
  51         public function __construct( $infile ) {
  52                 $this->infiles = explode( ';', $infile );
  53                 $this->reader = new XMLReader();
  54                 $infile = array_shift( $this->infiles );
  55                 if ( defined( 'LIBXML_PARSEHUGE' ) ) {
  56                         $this->reader->open( $infile, null, LIBXML_PARSEHUGE );
  57                 } else {
  58                         $this->reader->open( $infile );
  59                 }
  60         }
  61
  62         /**
  63          * Attempts to fetch the text of a particular page revision
  64          * from the dump stream. May return null if the page is
  65          * unavailable.
  66          *
  67          * @param int $page ID number of page to read
  68          * @param int $rev ID number of revision to read
  69          * @return string|null
  70          */
  71         function prefetch( $page, $rev ) {
  72                 $page = intval( $page );
  73                 $rev = intval( $rev );
  74                 while ( $this->lastPage < $page && !$this->atEnd ) {
  75                         $this->debug( "BaseDump::prefetch at page $this->lastPage, looking for $page" );
  76                         $this->nextPage();
  77                 }
  78                 if ( $this->lastPage > $page || $this->atEnd ) {
  79                         $this->debug( "BaseDump::prefetch already past page $page "
  80                                 . "looking for rev $rev  [$this->lastPage, $this->lastRev]" );
  81
  82                         return null;
  83                 }
  84                 while ( $this->lastRev < $rev && !$this->atEnd && !$this->atPageEnd ) {
  85                         $this->debug( "BaseDump::prefetch at page $this->lastPage, rev $this->lastRev, "
  86                                 . "looking for $page, $rev" );
  87                         $this->nextRev();
  88                 }
  89                 if ( $this->lastRev == $rev && !$this->atEnd ) {
  90                         $this->debug( "BaseDump::prefetch hit on $page, $rev [$this->lastPage, $this->lastRev]" );
  91
  92                         return $this->nextText();
  93                 } else {
  94                         $this->debug( "BaseDump::prefetch already past rev $rev on page $page "
  95                                 . "[$this->lastPage, $this->lastRev]" );
  96
  97                         return null;
  98                 }
  99         }
 100
 101         function debug( $str ) {
 102                 wfDebug( $str . "\n" );
 103                 // global $dumper;
 104                 // $dumper->progress( $str );
 105         }
 106
 107         /**
 108          * @access private
 109          */
 110         function nextPage() {
 111                 if ( $this->skipTo( 'page', 'mediawiki' ) ) {
 112                         if ( $this->skipTo( 'id' ) ) {
 113                                 $this->lastPage = intval( $this->nodeContents() );
 114                                 $this->lastRev = 0;
 115                                 $this->atPageEnd = false;
 116                         }
 117                 } else {
 118                         $this->close();
 119                         if ( count( $this->infiles ) ) {
 120                                 $infile = array_shift( $this->infiles );
 121                                 $this->reader->open( $infile );
 122                                 $this->atEnd = false;
 123                         }
 124                 }
 125         }
 126
 127         /**
 128          * @access private
 129          */
 130         function nextRev() {
 131                 if ( $this->skipTo( 'revision' ) ) {
 132                         if ( $this->skipTo( 'id' ) ) {
 133                                 $this->lastRev = intval( $this->nodeContents() );
 134                         }
 135                 } else {
 136                         $this->atPageEnd = true;
 137                 }
 138         }
 139
 140         /**
 141          * @access private
 142          * @return string
 143          */
 144         function nextText() {
 145                 $this->skipTo( 'text' );
 146
 147                 return strval( $this->nodeContents() );
 148         }
 149
 150         /**
 151          * @access private
 152          * @param string $name
 153          * @param string $parent
 154          * @return bool|null
 155          */
 156         function skipTo( $name, $parent = 'page' ) {
 157                 if ( $this->atEnd ) {
 158                         return false;
 159                 }
 160                 while ( $this->reader->read() ) {
 161                         if ( $this->reader->nodeType == XMLReader::ELEMENT
 162                                 && $this->reader->name == $name
 163                         ) {
 164                                 return true;
 165                         }
 166                         if ( $this->reader->nodeType == XMLReader::END_ELEMENT
 167                                 && $this->reader->name == $parent
 168                         ) {
 169                                 $this->debug( "BaseDump::skipTo found </$parent> searching for <$name>" );
 170
 171                                 return false;
 172                         }
 173                 }
 174
 175                 return $this->close();
 176         }
 177
 178         /**
 179          * Shouldn't something like this be built-in to XMLReader?
 180          * Fetches text contents of the current element, assuming
 181          * no sub-elements or such scary things.
 182          *
 183          * @return string
 184          * @access private
 185          */
 186         function nodeContents() {
 187                 if ( $this->atEnd ) {
 188                         return null;
 189                 }
 190                 if ( $this->reader->isEmptyElement ) {
 191                         return "";
 192                 }
 193                 $buffer = "";
 194                 while ( $this->reader->read() ) {
 195                         switch ( $this->reader->nodeType ) {
 196                                 case XMLReader::TEXT:
 197                                 // case XMLReader::WHITESPACE:
 198                                 case XMLReader::SIGNIFICANT_WHITESPACE:
 199                                         $buffer .= $this->reader->value;
 200                                         break;
 201                                 case XMLReader::END_ELEMENT:
 202                                         return $buffer;
 203                         }
 204                 }
 205
 206                 return $this->close();
 207         }
 208
 209         /**
 210          * @access private
 211          * @return null
 212          */
 213         function close() {
 214                 $this->reader->close();
 215                 $this->atEnd = true;
 216
 217                 return null;
 218         }
 219 }