<?php
+/**
+ * XML file reader for the page data importer
+ *
+ * @file
+ */
+
/**
* implements Special:Import
* @ingroup SpecialPage
* Creates an ImportXMLReader drawing from the source provided
*/
function __construct( $source ) {
- $this->reader = new XMLReader();
-
+ $this->reader = new XMLReader2();
+
stream_wrapper_register( 'uploadsource', 'UploadSourceAdapter' );
$id = UploadSourceAdapter::registerSource( $source );
$this->reader->open( "uploadsource://$id" );
-
+
// Default callbacks
$this->setRevisionCallback( array( $this, "importRevision" ) );
$this->setUploadCallback( array( $this, 'importUpload' ) );
$this->setLogItemCallback( array( $this, 'importLogItem' ) );
+ $this->setPageOutCallback( array( $this, 'finishImportPage' ) );
}
private function throwXmlError( $err ) {
}
private function debug( $data ) {
- if( $this->mDebug ) {
+ if( $this->mDebug ) {
wfDebug( "IMPORT: $data\n" );
- }
+ }
}
-
+
private function warn( $data ) {
wfDebug( "IMPORT: $data\n" );
}
$wgOut->addHTML( "<li>" . htmlspecialchars( $data ) . "</li>\n" );
}
}
-
+
/**
* Set debug mode...
*/
$this->mUploadCallback = $callback;
return $previous;
}
-
+
/**
* Sets the action to perform as each log item reached.
* @param $callback callback
$this->mLogItemCallback = $callback;
return $previous;
}
-
+
/**
* Sets the action to perform when site info is encountered
* @param $callback callback
return false;
}
}
-
+
/**
* Default per-revision callback, performs the import.
* @param $revision WikiRevision
$dbw = wfGetDB( DB_MASTER );
return $dbw->deadlockLoop( array( $revision, 'importOldRevision' ) );
}
-
+
/**
* Default per-revision callback, performs the import.
* @param $rev WikiRevision
return false;
}
+ /**
+ * Mostly for hook use
+ */
+ public function finishImportPage( $title, $origTitle, $revCount, $sRevCount, $pageInfo ) {
+ $args = func_get_args();
+ return wfRunHooks( 'AfterImportPage', $args );
+ }
+
/**
* Alternate per-revision callback, for debugging.
* @param $revision WikiRevision
$this->debug( "-- Comment: " . $revision->comment );
$this->debug( "-- Text: " . $revision->text );
}
-
+
/**
* Notify the callback function when a new <page> is reached.
* @param $title Title
* Notify the callback function when a </page> is closed.
* @param $title Title
* @param $origTitle Title
- * @param $revisionCount int
- * @param $successCount Int: number of revisions for which callback returned true
+ * @param $revCount Integer
+ * @param $sucCount Int: number of revisions for which callback returned true
+ * @param $pageInfo Array: associative array of page information
*/
- private function pageOutCallback( $title, $origTitle, $revisionCount, $successCount ) {
+ private function pageOutCallback( $title, $origTitle, $revCount, $sucCount, $pageInfo ) {
if( isset( $this->mPageOutCallback ) ) {
- call_user_func_array( $this->mPageOutCallback,
- array( $title, $origTitle, $revisionCount, $successCount ) );
+ $args = func_get_args();
+ call_user_func_array( $this->mPageOutCallback, $args );
}
}
-
+
/**
* Notify the callback function of a revision
* @param $revision A WikiRevision object
return false;
}
}
-
+
/**
* Notify the callback function of a new log item
* @param $revision A WikiRevision object
return false;
}
}
-
+
/**
* Shouldn't something like this be built-in to XMLReader?
* Fetches text contents of the current element, assuming
* @access private
*/
private function nodeContents() {
- if( $this->reader->isEmptyElement ) {
- return "";
- }
- $buffer = "";
- while( $this->reader->read() ) {
- switch( $this->reader->nodeType ) {
- case XmlReader::TEXT:
- case XmlReader::SIGNIFICANT_WHITESPACE:
- $buffer .= $this->reader->value;
- break;
- case XmlReader::END_ELEMENT:
- return $buffer;
- }
- }
- return $this->close();
+ return $this->reader->nodeContents();
}
# --------------
-
+
/** Left in for debugging */
private function dumpElement() {
static $lookup = null;
$xmlReaderConstants = array(
"NONE",
"ELEMENT",
- "ATTRIBUTE",
+ "ATTRIBUTE",
"TEXT",
"CDATA",
"ENTITY_REF",
"XML_DECLARATION",
);
$lookup = array();
-
+
foreach( $xmlReaderConstants as $name ) {
$lookup[constant("XmlReader::$name")] = $name;
}
}
-
- print( var_dump(
+
+ print( var_dump(
$lookup[$this->reader->nodeType],
$this->reader->name,
$this->reader->value
/**
* Primary entry point
*/
- public function doImport() {
+ public function doImport() {
$this->reader->read();
-
+
if ( $this->reader->name != 'mediawiki' ) {
throw new MWException( "Expected <mediawiki> tag, got ".
$this->reader->name );
}
$this->debug( "<mediawiki> tag is correct." );
-
+
$this->debug( "Starting primary dump processing loop." );
-
+
$keepReading = $this->reader->read();
$skip = false;
while ( $keepReading ) {
$tag = $this->reader->name;
$type = $this->reader->nodeType;
-
+
if ( !wfRunHooks( 'ImportHandleToplevelXMLTag', $this->reader ) ) {
// Do nothing
} elseif ( $tag == 'mediawiki' && $type == XmlReader::END_ELEMENT ) {
$this->handleSiteInfo();
} elseif ( $tag == 'page' ) {
$this->handlePage();
- } elseif ( $tag == 'logitem' ) {
- $this->handleLogItem();
+ } elseif ( $tag == 'logitem' ) {
+ $this->handleLogItem();
} elseif ( $tag != '#text' ) {
$this->warn( "Unhandled top-level XML tag $tag" );
-
+
$skip = true;
}
-
+
if ($skip) {
$keepReading = $this->reader->next();
$skip = false;
return true;
}
-
+
private function handleSiteInfo() {
// Site info is useful, but not actually used for dump imports.
// Includes a quick short-circuit to save performance.
}
throw new MWException( "SiteInfo tag is not yet handled, do not set mSiteInfoCallback" );
}
-
+
private function handleLogItem() {
$this->debug( "Enter log item handler." );
$logInfo = array();
-
+
// Fields that can just be stuffed in the pageInfo object
$normalFields = array( 'id', 'comment', 'type', 'action', 'timestamp',
'logtitle', 'params' );
-
+
while ( $this->reader->read() ) {
if ( $this->reader->nodeType == XmlReader::END_ELEMENT &&
$this->reader->name == 'logitem') {
break;
}
-
+
$tag = $this->reader->name;
-
+
if ( !wfRunHooks( 'ImportHandleLogItemXMLTag',
$this->reader, $logInfo ) ) {
// Do nothing
- } if ( in_array( $tag, $normalFields ) ) {
+ } elseif ( in_array( $tag, $normalFields ) ) {
$logInfo[$tag] = $this->nodeContents();
} elseif ( $tag == 'contributor' ) {
$logInfo['contributor'] = $this->handleContributor();
$this->warn( "Unhandled log-item XML tag $tag" );
}
}
-
+
$this->processLogItem( $logInfo );
}
-
+
private function processLogItem( $logInfo ) {
$revision = new WikiRevision;
-
+
$revision->setID( $logInfo['id'] );
$revision->setType( $logInfo['type'] );
$revision->setAction( $logInfo['action'] );
$revision->setTimestamp( $logInfo['timestamp'] );
$revision->setParams( $logInfo['params'] );
$revision->setTitle( Title::newFromText( $logInfo['logtitle'] ) );
-
+
if ( isset( $logInfo['comment'] ) ) {
$revision->setComment( $logInfo['comment'] );
}
-
+
if ( isset( $logInfo['contributor']['ip'] ) ) {
$revision->setUserIP( $logInfo['contributor']['ip'] );
}
if ( isset( $logInfo['contributor']['username'] ) ) {
$revision->setUserName( $logInfo['contributor']['username'] );
}
-
+
return $this->logItemCallback( $revision );
}
-
+
private function handlePage() {
// Handle page data.
$this->debug( "Enter page handler." );
$pageInfo = array( 'revisionCount' => 0, 'successfulRevisionCount' => 0 );
-
+
// Fields that can just be stuffed in the pageInfo object
$normalFields = array( 'title', 'id', 'redirect', 'restrictions' );
-
+
$skip = false;
$badTitle = false;
-
+
while ( $skip ? $this->reader->next() : $this->reader->read() ) {
if ( $this->reader->nodeType == XmlReader::END_ELEMENT &&
$this->reader->name == 'page') {
break;
}
-
+
$tag = $this->reader->name;
-
+
if ( $badTitle ) {
// The title is invalid, bail out of this page
$skip = true;
- } elseif ( !wfRunHooks( 'ImportHandlePageXMLTag', $this->reader,
- $pageInfo ) ) {
+ } elseif ( !wfRunHooks( 'ImportHandlePageXMLTag', array( $this->reader,
+ &$pageInfo ) ) ) {
// Do nothing
- } if ( in_array( $tag, $normalFields ) ) {
+ } elseif ( in_array( $tag, $normalFields ) ) {
$pageInfo[$tag] = $this->nodeContents();
if ( $tag == 'title' ) {
$title = $this->processTitle( $pageInfo['title'] );
-
+
if ( !$title ) {
$badTitle = true;
$skip = true;
}
-
+
$this->pageCallback( $title );
list( $pageInfo['_title'], $origTitle ) = $title;
}
$skip = true;
}
}
-
+
$this->pageOutCallback( $pageInfo['_title'], $origTitle,
$pageInfo['revisionCount'],
- $pageInfo['successfulRevisionCount'] );
+ $pageInfo['successfulRevisionCount'],
+ $pageInfo );
}
-
+
private function handleRevision( &$pageInfo ) {
$this->debug( "Enter revision handler" );
$revisionInfo = array();
-
+
$normalFields = array( 'id', 'timestamp', 'comment', 'minor', 'text' );
-
+
$skip = false;
-
+
while ( $skip ? $this->reader->next() : $this->reader->read() ) {
if ( $this->reader->nodeType == XmlReader::END_ELEMENT &&
$this->reader->name == 'revision') {
break;
}
-
+
$tag = $this->reader->name;
-
+
if ( !wfRunHooks( 'ImportHandleRevisionXMLTag', $this->reader,
$pageInfo, $revisionInfo ) ) {
// Do nothing
- } if ( in_array( $tag, $normalFields ) ) {
+ } elseif ( in_array( $tag, $normalFields ) ) {
$revisionInfo[$tag] = $this->nodeContents();
} elseif ( $tag == 'contributor' ) {
$revisionInfo['contributor'] = $this->handleContributor();
$skip = true;
}
}
-
+
$pageInfo['revisionCount']++;
if ( $this->processRevision( $pageInfo, $revisionInfo ) ) {
$pageInfo['successfulRevisionCount']++;
}
}
-
+
private function processRevision( $pageInfo, $revisionInfo ) {
$revision = new WikiRevision;
-
+
$revision->setID( $revisionInfo['id'] );
$revision->setText( $revisionInfo['text'] );
$revision->setTitle( $pageInfo['_title'] );
$revision->setTimestamp( $revisionInfo['timestamp'] );
-
+
if ( isset( $revisionInfo['comment'] ) ) {
$revision->setComment( $revisionInfo['comment'] );
}
-
+
if ( isset( $revisionInfo['minor'] ) )
$revision->setMinor( true );
-
+
if ( isset( $revisionInfo['contributor']['ip'] ) ) {
$revision->setUserIP( $revisionInfo['contributor']['ip'] );
}
if ( isset( $revisionInfo['contributor']['username'] ) ) {
$revision->setUserName( $revisionInfo['contributor']['username'] );
}
-
+
return $this->revisionCallback( $revision );
}
-
+
private function handleUpload( &$pageInfo ) {
$this->debug( "Enter upload handler" );
$uploadInfo = array();
-
+
$normalFields = array( 'timestamp', 'comment', 'filename', 'text',
'src', 'size' );
-
+
$skip = false;
-
+
while ( $skip ? $this->reader->next() : $this->reader->read() ) {
if ( $this->reader->nodeType == XmlReader::END_ELEMENT &&
$this->reader->name == 'upload') {
break;
}
-
+
$tag = $this->reader->name;
-
+
if ( !wfRunHooks( 'ImportHandleUploadXMLTag', $this->reader,
$pageInfo, $revisionInfo ) ) {
// Do nothing
- } if ( in_array( $tag, $normalFields ) ) {
+ } elseif ( in_array( $tag, $normalFields ) ) {
$uploadInfo[$tag] = $this->nodeContents();
} elseif ( $tag == 'contributor' ) {
$uploadInfo['contributor'] = $this->handleContributor();
$skip = true;
}
}
-
+
return $this->processUpload( $pageInfo, $uploadInfo );
}
-
+
private function processUpload( $pageInfo, $uploadInfo ) {
$revision = new WikiRevision;
-
+
$revision->setTitle( $pageInfo['_title'] );
$revision->setID( $uploadInfo['id'] );
$revision->setTimestamp( $uploadInfo['timestamp'] );
$revision->setSrc( $uploadInfo['src'] );
$revision->setSize( intval( $uploadInfo['size'] ) );
$revision->setComment( $uploadInfo['comment'] );
-
+
if ( isset( $uploadInfo['contributor']['ip'] ) ) {
$revision->setUserIP( $revisionInfo['contributor']['ip'] );
}
if ( isset( $uploadInfo['contributor']['username'] ) ) {
$revision->setUserName( $revisionInfo['contributor']['username'] );
}
-
+
return $this->uploadCallback( $revision );
}
-
+
private function handleContributor() {
$fields = array( 'id', 'ip', 'username' );
$info = array();
-
+
while ( $this->reader->read() ) {
if ( $this->reader->nodeType == XmlReader::END_ELEMENT &&
$this->reader->name == 'contributor') {
break;
}
-
+
$tag = $this->reader->name;
-
+
if ( in_array( $tag, $fields ) ) {
$info[$tag] = $this->nodeContents();
}
}
-
+
return $info;
}
-
+
private function processTitle( $text ) {
$workTitle = $text;
$origTitle = Title::newFromText( $workTitle );
- $title = null;
-
+
if( !is_null( $this->mTargetNamespace ) && !is_null( $origTitle ) ) {
$title = Title::makeTitle( $this->mTargetNamespace,
$origTitle->getDBkey() );
} else {
$title = Title::newFromText( $workTitle );
}
-
+
if( is_null( $title ) ) {
// Invalid page title? Ignore the page
$this->notice( "Skipping invalid page title '$workTitle'" );
+ return false;
} elseif( $title->getInterwiki() != '' ) {
$this->notice( "Skipping interwiki page title '$workTitle'" );
- $title = null;
+ return false;
}
-
+
return array( $origTitle, $title );
}
}
/** This is a horrible hack used to keep source compatibility */
class UploadSourceAdapter {
static $sourceRegistrations = array();
-
+
private $mSource;
private $mBuffer;
private $mPosition;
-
+
static function registerSource( $source ) {
$id = wfGenerateToken();
-
+
self::$sourceRegistrations[$id] = $source;
-
+
return $id;
}
-
+
function stream_open( $path, $mode, $options, &$opened_path ) {
$url = parse_url($path);
$id = $url['host'];
-
+
if ( !isset( self::$sourceRegistrations[$id] ) ) {
return false;
}
-
+
$this->mSource = self::$sourceRegistrations[$id];
-
+
return true;
}
-
+
function stream_read( $count ) {
$return = '';
$leave = false;
-
+
while ( !$leave && !$this->mSource->atEnd() &&
strlen($this->mBuffer) < $count ) {
$read = $this->mSource->readChunk();
-
+
if ( !strlen($read) ) {
- $leave = true;
+ $leave = true;
}
-
+
$this->mBuffer .= $read;
}
-
+
if ( strlen($this->mBuffer) ) {
$return = substr( $this->mBuffer, 0, $count );
$this->mBuffer = substr( $this->mBuffer, $count );
}
-
+
$this->mPosition += strlen($return);
-
+
return $return;
}
-
+
function stream_write( $data ) {
return false;
}
-
+
function stream_tell() {
return $this->mPosition;
}
function url_stat() {
$result = array();
-
+
$result['dev'] = $result[0] = 0;
$result['ino'] = $result[1] = 0;
$result['mode'] = $result[2] = 0;
$result['ctime'] = $result[10] = 0;
$result['blksize'] = $result[11] = 0;
$result['blocks'] = $result[12] = 0;
-
+
return $result;
}
}
+
+class XMLReader2 extends XMLReader {
+ function nodeContents() {
+ if( $this->isEmptyElement ) {
+ return "";
+ }
+ $buffer = "";
+ while( $this->read() ) {
+ switch( $this->nodeType ) {
+ case XmlReader::TEXT:
+ case XmlReader::SIGNIFICANT_WHITESPACE:
+ $buffer .= $this->value;
+ break;
+ case XmlReader::END_ELEMENT:
+ return $buffer;
+ }
+ }
+ return $this->close();
+ }
+}