<?php
/**
- * @package MediaWiki
- * @subpackage Maintenance
+ * @addtogroup Maintenance
*/
define( 'REPORTING_INTERVAL', 10 );
# Destination directory
var $dest;
+ # Skip existing files
+ var $noOverwrite = false;
+
# Show interlanguage links?
var $interwiki = true;
# Make a copy of all images encountered
var $makeSnapshot = false;
+ # Don't image description pages in doEverything()
+ var $noSharedDesc = false;
+
# Make links assuming the script path is in the same directory as
# the destination
var $alternateScriptPath = false;
# Has setupGlobals been called?
var $setupDone = false;
+ # Has to compress html pages
+ var $compress = false;
+
# List of raw pages used in the current article
var $rawPages;
var $sliceNumerator = 1, $sliceDenominator = 1;
+ # Max page ID, lazy initialised
+ var $maxPageID = false;
+
+ # UDP profiling
+ var $udpProfile, $udpProfileCounter = 0, $udpProfileInit = false;
+
function DumpHTML( $settings = array() ) {
foreach ( $settings as $var => $value ) {
$this->$var = $value;
return;
}
$this->doArticles();
- $this->doLocalImageDescriptions();
- $this->doSharedImageDescriptions();
$this->doCategories();
$this->doRedirects();
if ( $this->sliceNumerator == 1 ) {
$this->doSpecials();
}
+ $this->doLocalImageDescriptions();
+
+ if ( !$this->noSharedDesc ) {
+ $this->doSharedImageDescriptions();
+ }
$this->setCheckpoint( 'everything', 'done' );
}
* Skip categories and images, they will be done separately
*/
function doArticles() {
- $fname = 'DumpHTML::doArticles';
-
if ( $this->endID === false ) {
- $dbr =& wfGetDB( DB_SLAVE );
- $this->endID = $dbr->selectField( 'page', 'max(page_id)', false, $fname );
+ $end = $this->getMaxPageID();
+ } else {
+ $end = $this->endID;
}
+ $start = $this->startID;
- # Slice the range
- list( $start, $end ) = $this->sliceRange( $this->startID, $this->endID );
-
# Start from the checkpoint
$cp = $this->getCheckpoint( 'article' );
if ( $cp == 'done' ) {
print "Starting from page_id $start of $end\n";
}
+ # Move the start point to the correct slice if it isn't there already
+ $start = $this->modSliceStart( $start );
+
$this->setupGlobals();
$mainPageObj = Title::newMainPage();
$mainPage = $mainPageObj->getPrefixedDBkey();
- for ($id = $start; $id <= $end; $id++) {
+ for ( $id = $start, $i = 0; $id <= $end; $id += $this->sliceDenominator, $i++ ) {
wfWaitForSlaves( 20 );
- if ( !($id % REPORTING_INTERVAL) ) {
+ if ( !( $i % REPORTING_INTERVAL) ) {
print "Processing ID: $id\r";
$this->setCheckpoint( 'article', $id );
}
- if ( !($id % (REPORTING_INTERVAL*10) ) ) {
+ if ( !($i % (REPORTING_INTERVAL*10) ) ) {
print "\n";
}
$title = Title::newFromID( $id );
if ( $title ) {
$ns = $title->getNamespace() ;
- if ( $ns != NS_CATEGORY && $title->getPrefixedDBkey() != $mainPage ) {
+ if ( $ns != NS_CATEGORY && $ns != NS_MEDIAWIKI &&
+ $title->getPrefixedDBkey() != $mainPage ) {
$this->doArticle( $title );
}
}
$this->setupGlobals();
print "Special:Categories...";
- $this->doArticle( Title::makeTitle( NS_SPECIAL, 'Categories' ) );
+ $this->doArticle( SpecialPage::getTitleFor( 'Categories' ) );
print "\n";
}
function doImageDescriptions() {
$this->doLocalImageDescriptions();
- $this->doSharedImageDescriptions();
+ if ( !$this->noSharedDesc ) {
+ $this->doSharedImageDescriptions();
+ }
}
/**
* have a local image
*/
function doLocalImageDescriptions() {
- global $wgSharedUploadDirectory;
+ $chunkSize = 1000;
- $dbr =& wfGetDB( DB_SLAVE );
+ $dbr = wfGetDB( DB_SLAVE );
$cp = $this->getCheckpoint( 'local image' );
if ( $cp == 'done' ) {
}
$this->setupGlobals();
-
- $res = $dbr->select( 'image', array( 'img_name' ), $conds, __METHOD__,
- array( 'ORDER BY' => 'img_name' ) );
-
$i = 0;
- $num = $dbr->numRows( $res );
- while ( $row = $dbr->fetchObject( $res ) ) {
- // Slice the result set with a filter
- if ( !$this->sliceFilter( $row->img_name ) ) {
- continue;
- }
- wfWaitForSlaves( 10 );
- if ( !( ++$i % REPORTING_INTERVAL ) ) {
- print "Done $i of $num\r";
- if ( $row->img_name !== 'done' ) {
- $this->setCheckpoint( 'local image', $row->img_name );
+ do {
+ $res = $dbr->select( 'image', array( 'img_name' ), $conds, __METHOD__,
+ array( 'ORDER BY' => 'img_name', 'LIMIT' => $chunkSize ) );
+ $numRows = $dbr->numRows( $res );
+
+ while ( $row = $dbr->fetchObject( $res ) ) {
+ # Update conds for the next chunk query
+ $conds = array( 'img_name > ' . $dbr->addQuotes( $row->img_name ) );
+
+ // Slice the result set with a filter
+ if ( !$this->sliceFilter( $row->img_name ) ) {
+ continue;
}
+
+ wfWaitForSlaves( 10 );
+ if ( !( ++$i % REPORTING_INTERVAL ) ) {
+ print "{$row->img_name}\n";
+ if ( $row->img_name !== 'done' ) {
+ $this->setCheckpoint( 'local image', $row->img_name );
+ }
+ }
+ $title = Title::makeTitle( NS_IMAGE, $row->img_name );
+ if ( $title->getArticleID() ) {
+ // Already done by dumpHTML
+ continue;
+ }
+ $this->doArticle( $title );
}
- $title = Title::makeTitle( NS_IMAGE, $row->img_name );
- if ( $title->getArticleID() ) {
- // Already done by dumpHTML
- continue;
- }
- $this->doArticle( $title );
- }
+ $dbr->freeResult( $res );
+ } while ( $numRows );
+
$this->setCheckpoint( 'local image', 'done' );
print "\n";
}
print "Writing description pages for commons images\n";
}
-
$this->setupGlobals();
$i = 0;
for ( $hash = $start; $hash <= $end; $hash++ ) {
$this->setCheckpoint( 'shared image', $hash );
- $dir = sprintf( "%01x/%02x", intval( $hash / 16 ), $hash );
- $paths = array_merge( glob( "{$this->sharedStaticDirectory}/$dir/*" ),
- glob( "{$this->sharedStaticDirectory}/thumb/$dir/*" ) );
-
- foreach ( $paths as $path ) {
- $file = wfBaseName( $path );
+ $dir = sprintf( "%s/%01x/%02x", $this->sharedStaticDirectory,
+ intval( $hash / 16 ), $hash );
+ $handle = @opendir( $dir );
+ while ( $handle && $file = readdir( $handle ) ) {
+ if ( $file[0] == '.' ) {
+ continue;
+ }
if ( !(++$i % REPORTING_INTERVAL ) ) {
print "$i\r";
}
- $title = Title::makeTitle( NS_IMAGE, $file );
+ $title = Title::makeTitleSafe( NS_IMAGE, $file );
$this->doArticle( $title );
}
+ if ( $handle ) {
+ closedir( $handle );
+ }
}
$this->setCheckpoint( 'shared image', 'done' );
print "\n";
}
function doCategories() {
- $fname = 'DumpHTML::doCategories';
+ $chunkSize = 1000;
+
$this->setupGlobals();
- $dbr =& wfGetDB( DB_SLAVE );
- $sql = 'SELECT DISTINCT cl_to FROM ' . $dbr->tableName( 'categorylinks' );
-
+ $dbr = wfGetDB( DB_SLAVE );
+
$cp = $this->getCheckpoint( 'category' );
if ( $cp == 'done' ) {
print "Category pages already done\n";
return;
} elseif ( $cp !== false ) {
print "Resuming category page dump from $cp\n";
- $sql .= ' WHERE cl_to >= ' . $dbr->addQuotes( $cp );
+ $conds = array( 'cl_to >= ' . $dbr->addQuotes( $cp ) );
+ } else {
+ print "Starting category pages\n";
+ $conds = false;
}
- $sql .= ' ORDER BY cl_to';
- print "Selecting categories...";
- $res = $dbr->query( $sql, $fname );
-
- print "\nWriting " . $dbr->numRows( $res ). " category pages\n";
$i = 0;
- while ( $row = $dbr->fetchObject( $res ) ) {
- // Filter pages from other slices
- if ( !$this->sliceFilter( $row->cl_to ) ) {
- continue;
- }
+ do {
+ $res = $dbr->select( 'categorylinks', 'DISTINCT cl_to', $conds, __METHOD__,
+ array( 'ORDER BY' => 'cl_to', 'LIMIT' => $chunkSize ) );
+ $numRows = $dbr->numRows( $res );
+
+ while ( $row = $dbr->fetchObject( $res ) ) {
+ // Set conditions for next chunk
+ $conds = array( 'cl_to > ' . $dbr->addQuotes( $row->cl_to ) );
+
+ // Filter pages from other slices
+ if ( !$this->sliceFilter( $row->cl_to ) ) {
+ continue;
+ }
- wfWaitForSlaves( 10 );
- if ( !(++$i % REPORTING_INTERVAL ) ) {
- print "{$row->cl_to}\n";
- if ( $row->cl_to != 'done' ) {
- $this->setCheckpoint( 'category', $row->cl_to );
+ wfWaitForSlaves( 10 );
+ if ( !(++$i % REPORTING_INTERVAL ) ) {
+ print "{$row->cl_to}\n";
+ if ( $row->cl_to != 'done' ) {
+ $this->setCheckpoint( 'category', $row->cl_to );
+ }
}
+ $title = Title::makeTitle( NS_CATEGORY, $row->cl_to );
+ $this->doArticle( $title );
}
- $title = Title::makeTitle( NS_CATEGORY, $row->cl_to );
- $this->doArticle( $title );
- }
+ $dbr->freeResult( $res );
+ } while ( $numRows );
+
$this->setCheckpoint( 'category', 'done' );
print "\n";
}
function doRedirects() {
print "Doing redirects...\n";
- $fname = 'DumpHTML::doRedirects';
-
-
- $dbr =& wfGetDB( DB_SLAVE );
- $end = $dbr->selectField( 'page', 'max(page_id)', false, $fname );
- list( $start, $end ) = $this->sliceRange( 1, $end );
+ $chunkSize = 10000;
+ $end = $this->getMaxPageID();
$cp = $this->getCheckpoint( 'redirect' );
if ( $cp == 'done' ) {
print "Redirects already done\n";
} elseif ( $cp !== false ) {
print "Resuming redirect generation from page_id $cp\n";
$start = intval( $cp );
+ } else {
+ $start = 1;
}
-
- $conds = array(
- 'page_is_redirect' => 1,
- "page_id BETWEEN $start AND $end"
- );
$this->setupGlobals();
- $res = $dbr->select( 'page', array( 'page_id', 'page_namespace', 'page_title' ),
- $conds, $fname );
- $num = $dbr->numRows( $res );
- print "$num redirects to do...\n";
+ $dbr = wfGetDB( DB_SLAVE );
$i = 0;
- while ( $row = $dbr->fetchObject( $res ) ) {
- $title = Title::makeTitle( $row->page_namespace, $row->page_title );
- if ( !(++$i % (REPORTING_INTERVAL*10) ) ) {
- print "Done $i of $num (ID {$row->page_id})\n";
- $this->setCheckpoint( 'redirect', $row->page_id );
+
+ for ( $chunkStart = $start; $chunkStart <= $end; $chunkStart += $chunkSize ) {
+ $chunkEnd = min( $end, $chunkStart + $chunkSize - 1 );
+ $conds = array(
+ 'page_is_redirect' => 1,
+ "page_id BETWEEN $chunkStart AND $chunkEnd"
+ );
+ # Modulo slicing in SQL
+ if ( $this->sliceDenominator != 1 ) {
+ $n = intval( $this->sliceNumerator );
+ $m = intval( $this->sliceDenominator );
+ $conds[] = "page_id % $m = $n";
+ }
+ $res = $dbr->select( 'page', array( 'page_id', 'page_namespace', 'page_title' ),
+ $conds, __METHOD__ );
+
+ while ( $row = $dbr->fetchObject( $res ) ) {
+ $title = Title::makeTitle( $row->page_namespace, $row->page_title );
+ if ( !(++$i % (REPORTING_INTERVAL*10) ) ) {
+ printf( "Done %d redirects (%2.3f%%)\n", $i, $row->page_id / $end * 100 );
+ $this->setCheckpoint( 'redirect', $row->page_id );
+ }
+ $this->doArticle( $title );
}
- $this->doArticle( $title );
+ $dbr->freeResult( $res );
}
$this->setCheckpoint( 'redirect', 'done' );
}
/** Write an article specified by title */
function doArticle( $title ) {
- global $wgTitle, $wgSharedUploadPath, $wgSharedUploadDirectory;
- global $wgUploadDirectory;
+ if ( $this->noOverwrite ) {
+ $fileName = $this->dest.'/'.$this->getHashedFilename( $title );
+ if ( file_exists( $fileName ) ) {
+ return;
+ }
+ }
+
+ $this->profile();
$this->rawPages = array();
$text = $this->getArticleHTML( $title );
print "Writing $file\n";
$file = fopen( $path, 'w' );
if ( !$file ) {
- print("Can't open file $fullName for writing\n");
+ print("Can't open file $path for writing\n");
continue;
}
fwrite( $file, $text );
fclose( $file );
}
}
+
+ wfIncrStats( 'dumphtml_article' );
}
/** Write the given text to the file identified by the given title object */
- function writeArticle( &$title, $text ) {
+ function writeArticle( $title, $text ) {
$filename = $this->getHashedFilename( $title );
+
+ # Temporary hack for current dump, this should be moved to
+ # getFriendlyName() at the earliest opportunity.
+ #
+ # Limit filename length to 255 characters, so it works on ext3.
+ # Titles are in fact limited to 255 characters, but dumpHTML
+ # adds a suffix which may put them over the limit.
+ $length = strlen( $filename );
+ if ( $length > 255 ) {
+ print "Warning: Filename too long ($length bytes). Skipping.\n";
+ return;
+ }
+
$fullName = "{$this->dest}/$filename";
$fullDir = dirname( $fullName );
+ if ( $this->compress ) {
+ $fullName .= ".gz";
+ $text = gzencode( $text, 9 );
+ }
+
wfMkdirParents( $fullDir, 0755 );
wfSuppressWarnings();
/** Set up globals required for parsing */
function setupGlobals( $currentDepth = NULL ) {
- global $wgUser, $wgTitle, $wgStylePath, $wgArticlePath, $wgMathPath;
+ global $wgUser, $wgStylePath, $wgArticlePath, $wgMathPath;
global $wgUploadPath, $wgLogo, $wgMaxCredits, $wgSharedUploadPath;
global $wgHideInterlanguageLinks, $wgUploadDirectory, $wgThumbnailScriptPath;
global $wgSharedThumbnailScriptPath, $wgEnableParserCache, $wgHooks, $wgServer;
$wgUser->setOption( 'skin', $this->skin );
$wgUser->setOption( 'editsection', 0 );
- if ( $this->makeSnapshot ) {
- $this->destUploadDirectory = "{$this->dest}/{$this->imageRel}";
- if ( realpath( $this->destUploadDirectory == $wgUploadDirectory ) ) {
- $this->makeSnapshot = false;
- }
+ $this->destUploadDirectory = "{$this->dest}/{$this->imageRel}";
+ if ( realpath( $this->destUploadDirectory ) == realpath( $wgUploadDirectory ) ) {
+ print "Disabling image snapshot because the destination is the same as the source\n";
+ $this->makeSnapshot = false;
}
-
$this->sharedStaticDirectory = "{$this->destUploadDirectory}/shared";
$this->setupDone = true;
}
/** Reads the content of a title object, executes the skin and captures the result */
- function getArticleHTML( &$title ) {
+ function getArticleHTML( $title ) {
global $wgOut, $wgTitle, $wgArticle, $wgUser;
$linkCache =& LinkCache::singleton();
}
}
+
$sk =& $wgUser->getSkin();
ob_start();
$sk->outputPage( $wgOut );
if ( !file_exists( $destLoc ) ) {
wfMkdirParents( dirname( $destLoc ), 0755 );
if ( function_exists( 'symlink' ) && !$this->forceCopy ) {
- symlink( $sourceLoc, $destLoc );
+ if ( !symlink( $sourceLoc, $destLoc ) ) {
+ print "Warning: unable to create symlink at $destLoc\n";
+ }
} else {
- copy( $sourceLoc, $destLoc );
+ if ( !copy( $sourceLoc, $destLoc ) ) {
+ print "Warning: unable to copy $sourceLoc to $destLoc\n";
+ }
}
}
}
$url = str_replace( '$1', "../$iw/" . wfUrlencode( $this->getHashedFilename( $title ) ),
$wgArticlePath );
}
+ $url .= $this->compress ? ".gz" : "";
return false;
} else {
return true;
$url = false;
if ( $query != '' ) {
+ $params = array();
parse_str( $query, $params );
if ( isset($params['action']) && $params['action'] == 'raw' ) {
if ( $params['gen'] == 'css' || $params['gen'] == 'js' ) {
} else {
$file = $this->getFriendlyName( $title->getPrefixedDBkey() );
// Clean up Monobook.css etc.
+ $matches = array();
if ( preg_match( '/^(.*)\.(css|js)_[0-9a-f]{4}$/', $file, $matches ) ) {
$file = $matches[1] . '.' . $matches[2];
}
if ( $url === false ) {
$url = str_replace( '$1', wfUrlencode( $this->getHashedFilename( $title ) ), $wgArticlePath );
}
-
+ $url .= $this->compress ? ".gz" : "";
return false;
}
}
# Split into characters
+ $m = array();
preg_match_all( '/./us', $dbk, $m );
$chars = $m[0];
return array( $sliceStart, $sliceEnd );
}
+ /**
+ * Adjust a start point so that it belongs to the current slice, where slices are defined by integer modulo
+ * @param integer $start
+ * @param integer $base The true start of the range; the minimum start
+ */
+ function modSliceStart( $start, $base = 1 ) {
+ return $start - ( $start % $this->sliceDenominator ) + $this->sliceNumerator - 1 + $base;
+ }
+
/**
* Determine whether a string belongs to the current slice, based on hash
*/
$text = '';
return false;
}
+
+ function getMaxPageID() {
+ if ( $this->maxPageID === false ) {
+ $dbr = wfGetDB( DB_SLAVE );
+ $this->maxPageID = $dbr->selectField( 'page', 'max(page_id)', false, __METHOD__ );
+ }
+ return $this->maxPageID;
+ }
+
+ function profile() {
+ global $wgProfiler;
+
+ if ( !$this->udpProfile ) {
+ return;
+ }
+ if ( !$this->udpProfileInit ) {
+ $this->udpProfileInit = true;
+ } elseif ( $this->udpProfileCounter == 1 % $this->udpProfile ) {
+ $wgProfiler->getFunctionReport();
+ $wgProfiler = new DumpHTML_ProfilerStub;
+ }
+ if ( $this->udpProfileCounter == 0 ) {
+ $wgProfiler = new ProfilerSimpleUDP;
+ $wgProfiler->setProfileID( 'dumpHTML' );
+ }
+ $this->udpProfileCounter = ( $this->udpProfileCounter + 1 ) % $this->udpProfile;
+ }
+}
+
+class DumpHTML_ProfilerStub {
+ function profileIn() {}
+ function profileOut() {}
+ function getOutput() {}
+ function close() {}
+ function getFunctionReport() {}
}
/** XML parser callback */