From c9ebbcd09bf45be8773e616a5384eb62155afd5a Mon Sep 17 00:00:00 2001 From: Andrew Garrett Date: Wed, 12 May 2010 13:28:13 +0000 Subject: [PATCH] Rewrite of XML Dump Processing: * Full rewrite of the WikiImporter class to a new class, XMLDumpImporter, which uses XMLReader instead of the deprecated PHP SAX interface. * Fixes for the following bugs: ** XML Dump Processor chokes on unrecognised tags. ** Special:Import does not allow you to import a dump including only log entries. * Added hooks to allow extensions to handle tags in any part of the XML document. * Fully source-compatible with the previous interface. --- includes/AutoLoader.php | 2 +- includes/Import.php | 652 -------------------------- includes/ImportXMLReader.php | 703 ++++++++++++++++++++++++++++ includes/api/ApiImport.php | 4 +- includes/specials/SpecialImport.php | 20 +- languages/messages/MessagesEn.php | 1 + maintenance/importDump.php | 2 +- maintenance/renderDump.php | 2 +- 8 files changed, 727 insertions(+), 659 deletions(-) create mode 100644 includes/ImportXMLReader.php diff --git a/includes/AutoLoader.php b/includes/AutoLoader.php index 55f9c48415..8f69c0d5fb 100644 --- a/includes/AutoLoader.php +++ b/includes/AutoLoader.php @@ -605,7 +605,7 @@ $wgAutoloadLocalClasses = array( 'WantedPagesPage' => 'includes/specials/SpecialWantedpages.php', 'WantedTemplatesPage' => 'includes/specials/SpecialWantedtemplates.php', 'WhatLinksHerePage' => 'includes/specials/SpecialWhatlinkshere.php', - 'WikiImporter' => 'includes/Import.php', + 'ImportXMLReader' => 'includes/ImportXMLReader.php', 'WikiRevision' => 'includes/Import.php', 'WithoutInterwikiPage' => 'includes/specials/SpecialWithoutinterwiki.php', diff --git a/includes/Import.php b/includes/Import.php index 45908a6699..605fd81d90 100644 --- a/includes/Import.php +++ b/includes/Import.php @@ -370,658 +370,6 @@ class WikiRevision { } -/** - * implements Special:Import - * @ingroup SpecialPage - */ -class WikiImporter { - var $mDebug = false; - var $mSource = null; - var $mPageCallback = null; - var $mPageOutCallback = null; - var $mRevisionCallback = null; - var $mLogItemCallback = null; - var $mUploadCallback = null; - var $mTargetNamespace = null; - var $mXmlNamespace = false; - var $lastfield; - var $tagStack = array(); - - function __construct( $source ) { - $this->setRevisionCallback( array( $this, "importRevision" ) ); - $this->setUploadCallback( array( $this, "importUpload" ) ); - $this->setLogItemCallback( array( $this, "importLogItem" ) ); - $this->mSource = $source; - } - - function throwXmlError( $err ) { - $this->debug( "FAILURE: $err" ); - wfDebug( "WikiImporter XML error: $err\n" ); - } - - function handleXmlNamespace ( $parser, $data, $prefix=false, $uri=false ) { - if( preg_match( '/www.mediawiki.org/',$prefix ) ) { - $prefix = str_replace( '/','\/',$prefix ); - $this->mXmlNamespace='/^'.$prefix.':/'; - } - } - - function stripXmlNamespace($name) { - if( $this->mXmlNamespace ) { - return(preg_replace($this->mXmlNamespace,'',$name,1)); - } - else { - return($name); - } - } - - # -------------- - - function doImport() { - if( empty( $this->mSource ) ) { - return new WikiErrorMsg( "importnotext" ); - } - - $parser = xml_parser_create_ns( "UTF-8" ); - - # case folding violates XML standard, turn it off - xml_parser_set_option( $parser, XML_OPTION_CASE_FOLDING, false ); - - xml_set_object( $parser, $this ); - xml_set_element_handler( $parser, "in_start", "" ); - xml_set_start_namespace_decl_handler( $parser, "handleXmlNamespace" ); - - $offset = 0; // for context extraction on error reporting - do { - $chunk = $this->mSource->readChunk(); - if( !xml_parse( $parser, $chunk, $this->mSource->atEnd() ) ) { - wfDebug( "WikiImporter::doImport encountered XML parsing error\n" ); - return new WikiXmlError( $parser, wfMsgHtml( 'import-parse-failure' ), $chunk, $offset ); - } - $offset += strlen( $chunk ); - } while( $chunk !== false && !$this->mSource->atEnd() ); - xml_parser_free( $parser ); - - return true; - } - - function debug( $data ) { - if( $this->mDebug ) { - wfDebug( "IMPORT: $data\n" ); - } - } - - function notice( $data ) { - global $wgCommandLineMode; - if( $wgCommandLineMode ) { - print "$data\n"; - } else { - global $wgOut; - $wgOut->addHTML( "
  • " . htmlspecialchars( $data ) . "
    • \n" ); - } - } - - /** - * Set debug mode... - */ - function setDebug( $debug ) { - $this->mDebug = $debug; - } - - /** - * Sets the action to perform as each new page in the stream is reached. - * @param $callback callback - * @return callback - */ - function setPageCallback( $callback ) { - $previous = $this->mPageCallback; - $this->mPageCallback = $callback; - return $previous; - } - - /** - * Sets the action to perform as each page in the stream is completed. - * Callback accepts the page title (as a Title object), a second object - * with the original title form (in case it's been overridden into a - * local namespace), and a count of revisions. - * - * @param $callback callback - * @return callback - */ - function setPageOutCallback( $callback ) { - $previous = $this->mPageOutCallback; - $this->mPageOutCallback = $callback; - return $previous; - } - - /** - * Sets the action to perform as each page revision is reached. - * @param $callback callback - * @return callback - */ - function setRevisionCallback( $callback ) { - $previous = $this->mRevisionCallback; - $this->mRevisionCallback = $callback; - return $previous; - } - - /** - * Sets the action to perform as each file upload version is reached. - * @param $callback callback - * @return callback - */ - function setUploadCallback( $callback ) { - $previous = $this->mUploadCallback; - $this->mUploadCallback = $callback; - return $previous; - } - - /** - * Sets the action to perform as each log item reached. - * @param $callback callback - * @return callback - */ - function setLogItemCallback( $callback ) { - $previous = $this->mLogItemCallback; - $this->mLogItemCallback = $callback; - return $previous; - } - - /** - * Set a target namespace to override the defaults - */ - function setTargetNamespace( $namespace ) { - if( is_null( $namespace ) ) { - // Don't override namespaces - $this->mTargetNamespace = null; - } elseif( $namespace >= 0 ) { - // FIXME: Check for validity - $this->mTargetNamespace = intval( $namespace ); - } else { - return false; - } - } - - /** - * Default per-revision callback, performs the import. - * @param $revision WikiRevision - * @private - */ - function importRevision( $revision ) { - $dbw = wfGetDB( DB_MASTER ); - return $dbw->deadlockLoop( array( $revision, 'importOldRevision' ) ); - } - - /** - * Default per-revision callback, performs the import. - * @param $rev WikiRevision - * @private - */ - function importLogItem( $rev ) { - $dbw = wfGetDB( DB_MASTER ); - return $dbw->deadlockLoop( array( $rev, 'importLogItem' ) ); - } - - /** - * Dummy for now... - */ - function importUpload( $revision ) { - //$dbw = wfGetDB( DB_MASTER ); - //return $dbw->deadlockLoop( array( $revision, 'importUpload' ) ); - return false; - } - - /** - * Alternate per-revision callback, for debugging. - * @param $revision WikiRevision - * @private - */ - function debugRevisionHandler( &$revision ) { - $this->debug( "Got revision:" ); - if( is_object( $revision->title ) ) { - $this->debug( "-- Title: " . $revision->title->getPrefixedText() ); - } else { - $this->debug( "-- Title: " ); - } - $this->debug( "-- User: " . $revision->user_text ); - $this->debug( "-- Timestamp: " . $revision->timestamp ); - $this->debug( "-- Comment: " . $revision->comment ); - $this->debug( "-- Text: " . $revision->text ); - } - - /** - * Notify the callback function when a new is reached. - * @param $title Title - * @private - */ - function pageCallback( $title ) { - if( is_callable( $this->mPageCallback ) ) { - call_user_func( $this->mPageCallback, $title ); - } - } - - /** - * Notify the callback function when a is closed. - * @param $title Title - * @param $origTitle Title - * @param $revisionCount int - * @param $successCount Int: number of revisions for which callback returned true - * @private - */ - function pageOutCallback( $title, $origTitle, $revisionCount, $successCount ) { - if( is_callable( $this->mPageOutCallback ) ) { - call_user_func( $this->mPageOutCallback, $title, $origTitle, - $revisionCount, $successCount ); - } - } - - # XML parser callbacks from here out -- beware! - function donothing( $parser, $x, $y="" ) { - #$this->debug( "donothing" ); - } - - function in_start( $parser, $name, $attribs ) { - $name = $this->stripXmlNamespace($name); - $this->debug( "in_start $name" ); - if( $name != "mediawiki" ) { - return $this->throwXMLerror( "Expected , got <$name>" ); - } - xml_set_element_handler( $parser, "in_mediawiki", "out_mediawiki" ); - } - - function in_mediawiki( $parser, $name, $attribs ) { - $name = $this->stripXmlNamespace($name); - $this->debug( "in_mediawiki $name" ); - if( $name == 'siteinfo' ) { - xml_set_element_handler( $parser, "in_siteinfo", "out_siteinfo" ); - } elseif( $name == 'page' ) { - $this->push( $name ); - $this->workRevisionCount = 0; - $this->workSuccessCount = 0; - $this->uploadCount = 0; - $this->uploadSuccessCount = 0; - xml_set_element_handler( $parser, "in_page", "out_page" ); - } elseif( $name == 'logitem' ) { - $this->push( $name ); - $this->workRevision = new WikiRevision; - xml_set_element_handler( $parser, "in_logitem", "out_logitem" ); - } else { - return $this->throwXMLerror( "Expected , got <$name>" ); - } - } - function out_mediawiki( $parser, $name ) { - $name = $this->stripXmlNamespace($name); - $this->debug( "out_mediawiki $name" ); - if( $name != "mediawiki" ) { - return $this->throwXMLerror( "Expected , got " ); - } - xml_set_element_handler( $parser, "donothing", "donothing" ); - } - - - function in_siteinfo( $parser, $name, $attribs ) { - // no-ops for now - $name = $this->stripXmlNamespace($name); - $this->debug( "in_siteinfo $name" ); - switch( $name ) { - case "sitename": - case "base": - case "generator": - case "case": - case "namespaces": - case "namespace": - break; - default: - return $this->throwXMLerror( "Element <$name> not allowed in ." ); - } - } - - function out_siteinfo( $parser, $name ) { - $name = $this->stripXmlNamespace($name); - if( $name == "siteinfo" ) { - xml_set_element_handler( $parser, "in_mediawiki", "out_mediawiki" ); - } - } - - - function in_page( $parser, $name, $attribs ) { - $name = $this->stripXmlNamespace($name); - $this->debug( "in_page $name" ); - switch( $name ) { - case "id": - case "title": - case "redirect": - case "restrictions": - $this->appendfield = $name; - $this->appenddata = ""; - xml_set_element_handler( $parser, "in_nothing", "out_append" ); - xml_set_character_data_handler( $parser, "char_append" ); - break; - case "revision": - $this->push( "revision" ); - if( is_object( $this->pageTitle ) ) { - $this->workRevision = new WikiRevision; - $this->workRevision->setTitle( $this->pageTitle ); - $this->workRevisionCount++; - } else { - // Skipping items due to invalid page title - $this->workRevision = null; - } - xml_set_element_handler( $parser, "in_revision", "out_revision" ); - break; - case "upload": - $this->push( "upload" ); - if( is_object( $this->pageTitle ) ) { - $this->workRevision = new WikiRevision; - $this->workRevision->setTitle( $this->pageTitle ); - $this->uploadCount++; - } else { - // Skipping items due to invalid page title - $this->workRevision = null; - } - xml_set_element_handler( $parser, "in_upload", "out_upload" ); - break; - default: - return $this->throwXMLerror( "Element <$name> not allowed in a ." ); - } - } - - function out_page( $parser, $name ) { - $name = $this->stripXmlNamespace($name); - $this->debug( "out_page $name" ); - $this->pop(); - if( $name != "page" ) { - return $this->throwXMLerror( "Expected , got " ); - } - xml_set_element_handler( $parser, "in_mediawiki", "out_mediawiki" ); - - $this->pageOutCallback( $this->pageTitle, $this->origTitle, - $this->workRevisionCount, $this->workSuccessCount ); - - $this->workTitle = null; - $this->workRevision = null; - $this->workRevisionCount = 0; - $this->workSuccessCount = 0; - $this->pageTitle = null; - $this->origTitle = null; - } - - function in_nothing( $parser, $name, $attribs ) { - $name = $this->stripXmlNamespace($name); - $this->debug( "in_nothing $name" ); - return $this->throwXMLerror( "No child elements allowed here; got <$name>" ); - } - - function char_append( $parser, $data ) { - $this->debug( "char_append '$data'" ); - $this->appenddata .= $data; - } - - function out_append( $parser, $name ) { - $name = $this->stripXmlNamespace($name); - $this->debug( "out_append $name" ); - if( $name != $this->appendfield ) { - return $this->throwXMLerror( "Expected appendfield}>, got " ); - } - - switch( $this->appendfield ) { - case "title": - $this->workTitle = $this->appenddata; - $this->origTitle = Title::newFromText( $this->workTitle ); - if( !is_null( $this->mTargetNamespace ) && !is_null( $this->origTitle ) ) { - $this->pageTitle = Title::makeTitle( $this->mTargetNamespace, - $this->origTitle->getDBkey() ); - } else { - $this->pageTitle = Title::newFromText( $this->workTitle ); - } - if( is_null( $this->pageTitle ) ) { - // Invalid page title? Ignore the page - $this->notice( "Skipping invalid page title '$this->workTitle'" ); - } elseif( $this->pageTitle->getInterwiki() != '' ) { - $this->notice( "Skipping interwiki page title '$this->workTitle'" ); - $this->pageTitle = null; - } else { - $this->pageCallback( $this->workTitle ); - } - break; - case "id": - if ( $this->parentTag() == 'revision' || $this->parentTag() == 'logitem' ) { - if( $this->workRevision ) - $this->workRevision->setID( $this->appenddata ); - } - break; - case "text": - if( $this->workRevision ) - $this->workRevision->setText( $this->appenddata ); - break; - case "username": - if( $this->workRevision ) - $this->workRevision->setUsername( $this->appenddata ); - break; - case "ip": - if( $this->workRevision ) - $this->workRevision->setUserIP( $this->appenddata ); - break; - case "timestamp": - if( $this->workRevision ) - $this->workRevision->setTimestamp( $this->appenddata ); - break; - case "comment": - if( $this->workRevision ) - $this->workRevision->setComment( $this->appenddata ); - break; - case "type": - if( $this->workRevision ) - $this->workRevision->setType( $this->appenddata ); - break; - case "action": - if( $this->workRevision ) - $this->workRevision->setAction( $this->appenddata ); - break; - case "logtitle": - if( $this->workRevision ) - $this->workRevision->setTitle( Title::newFromText( $this->appenddata ) ); - break; - case "params": - if( $this->workRevision ) - $this->workRevision->setParams( $this->appenddata ); - break; - case "minor": - if( $this->workRevision ) - $this->workRevision->setMinor( true ); - break; - case "filename": - if( $this->workRevision ) - $this->workRevision->setFilename( $this->appenddata ); - break; - case "src": - if( $this->workRevision ) - $this->workRevision->setSrc( $this->appenddata ); - break; - case "size": - if( $this->workRevision ) - $this->workRevision->setSize( intval( $this->appenddata ) ); - break; - default: - $this->debug( "Bad append: {$this->appendfield}" ); - } - $this->appendfield = ""; - $this->appenddata = ""; - - $parent = $this->parentTag(); - xml_set_element_handler( $parser, "in_$parent", "out_$parent" ); - xml_set_character_data_handler( $parser, "donothing" ); - } - - function in_revision( $parser, $name, $attribs ) { - $name = $this->stripXmlNamespace($name); - $this->debug( "in_revision $name" ); - switch( $name ) { - case "id": - case "timestamp": - case "comment": - case "minor": - case "text": - $this->appendfield = $name; - xml_set_element_handler( $parser, "in_nothing", "out_append" ); - xml_set_character_data_handler( $parser, "char_append" ); - break; - case "contributor": - $this->push( "contributor" ); - xml_set_element_handler( $parser, "in_contributor", "out_contributor" ); - break; - default: - return $this->throwXMLerror( "Element <$name> not allowed in a ." ); - } - } - - function out_revision( $parser, $name ) { - $name = $this->stripXmlNamespace($name); - $this->debug( "out_revision $name" ); - $this->pop(); - if( $name != "revision" ) { - return $this->throwXMLerror( "Expected , got " ); - } - xml_set_element_handler( $parser, "in_page", "out_page" ); - - if( $this->workRevision ) { - $ok = call_user_func_array( $this->mRevisionCallback, - array( $this->workRevision, $this ) ); - if( $ok ) { - $this->workSuccessCount++; - } - } - } - - function in_logitem( $parser, $name, $attribs ) { - $name = $this->stripXmlNamespace($name); - $this->debug( "in_logitem $name" ); - switch( $name ) { - case "id": - case "timestamp": - case "comment": - case "type": - case "action": - case "logtitle": - case "params": - $this->appendfield = $name; - xml_set_element_handler( $parser, "in_nothing", "out_append" ); - xml_set_character_data_handler( $parser, "char_append" ); - break; - case "contributor": - $this->push( "contributor" ); - xml_set_element_handler( $parser, "in_contributor", "out_contributor" ); - break; - default: - return $this->throwXMLerror( "Element <$name> not allowed in a ." ); - } - } - - function out_logitem( $parser, $name ) { - $name = $this->stripXmlNamespace($name); - $this->debug( "out_logitem $name" ); - $this->pop(); - if( $name != "logitem" ) { - return $this->throwXMLerror( "Expected , got " ); - } - xml_set_element_handler( $parser, "in_mediawiki", "out_mediawiki" ); - - if( $this->workRevision ) { - $ok = call_user_func_array( $this->mLogItemCallback, - array( $this->workRevision, $this ) ); - if( $ok ) { - $this->workSuccessCount++; - } - } - } - - function in_upload( $parser, $name, $attribs ) { - $name = $this->stripXmlNamespace($name); - $this->debug( "in_upload $name" ); - switch( $name ) { - case "timestamp": - case "comment": - case "text": - case "filename": - case "src": - case "size": - $this->appendfield = $name; - xml_set_element_handler( $parser, "in_nothing", "out_append" ); - xml_set_character_data_handler( $parser, "char_append" ); - break; - case "contributor": - $this->push( "contributor" ); - xml_set_element_handler( $parser, "in_contributor", "out_contributor" ); - break; - default: - return $this->throwXMLerror( "Element <$name> not allowed in an ." ); - } - } - - function out_upload( $parser, $name ) { - $name = $this->stripXmlNamespace($name); - $this->debug( "out_revision $name" ); - $this->pop(); - if( $name != "upload" ) { - return $this->throwXMLerror( "Expected , got " ); - } - xml_set_element_handler( $parser, "in_page", "out_page" ); - - if( $this->workRevision ) { - $ok = call_user_func_array( $this->mUploadCallback, - array( $this->workRevision, $this ) ); - if( $ok ) { - $this->workUploadSuccessCount++; - } - } - } - - function in_contributor( $parser, $name, $attribs ) { - $name = $this->stripXmlNamespace($name); - $this->debug( "in_contributor $name" ); - switch( $name ) { - case "username": - case "ip": - case "id": - $this->appendfield = $name; - xml_set_element_handler( $parser, "in_nothing", "out_append" ); - xml_set_character_data_handler( $parser, "char_append" ); - break; - default: - $this->throwXMLerror( "Invalid tag <$name> in " ); - } - } - - function out_contributor( $parser, $name ) { - $name = $this->stripXmlNamespace($name); - $this->debug( "out_contributor $name" ); - $this->pop(); - if( $name != "contributor" ) { - return $this->throwXMLerror( "Expected , got " ); - } - $parent = $this->parentTag(); - xml_set_element_handler( $parser, "in_$parent", "out_$parent" ); - } - - private function push( $name ) { - array_push( $this->tagStack, $name ); - $this->debug( "PUSH $name" ); - } - - private function pop() { - $name = array_pop( $this->tagStack ); - $this->debug( "POP $name" ); - return $name; - } - - private function parentTag() { - $name = $this->tagStack[count( $this->tagStack ) - 1]; - $this->debug( "PARENT $name" ); - return $name; - } - -} - /** * @todo document (e.g. one-sentence class description). * @ingroup SpecialPage diff --git a/includes/ImportXMLReader.php b/includes/ImportXMLReader.php new file mode 100644 index 0000000000..605d262b5c --- /dev/null +++ b/includes/ImportXMLReader.php @@ -0,0 +1,703 @@ +reader = new XMLReader(); + + 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' ) ); + } + + function throwXmlError( $err ) { + $this->debug( "FAILURE: $err" ); + wfDebug( "WikiImporter XML error: $err\n" ); + } + + function debug( $data ) { + if( $this->mDebug ) { + wfDebug( "IMPORT: $data\n" ); + } + } + + function warn( $data ) { + wfDebug( "IMPORT: $data\n" ); + } + + function notice( $data ) { + global $wgCommandLineMode; + if( $wgCommandLineMode ) { + print "$data\n"; + } else { + global $wgOut; + $wgOut->addHTML( "
  • " . htmlspecialchars( $data ) . "
    • \n" ); + } + } + + /** + * Set debug mode... + */ + function setDebug( $debug ) { + $this->mDebug = $debug; + } + + /** + * Sets the action to perform as each new page in the stream is reached. + * @param $callback callback + * @return callback + */ + function setPageCallback( $callback ) { + $previous = $this->mPageCallback; + $this->mPageCallback = $callback; + return $previous; + } + + /** + * Sets the action to perform as each page in the stream is completed. + * Callback accepts the page title (as a Title object), a second object + * with the original title form (in case it's been overridden into a + * local namespace), and a count of revisions. + * + * @param $callback callback + * @return callback + */ + function setPageOutCallback( $callback ) { + $previous = $this->mPageOutCallback; + $this->mPageOutCallback = $callback; + return $previous; + } + + /** + * Sets the action to perform as each page revision is reached. + * @param $callback callback + * @return callback + */ + function setRevisionCallback( $callback ) { + $previous = $this->mRevisionCallback; + $this->mRevisionCallback = $callback; + return $previous; + } + + /** + * Sets the action to perform as each file upload version is reached. + * @param $callback callback + * @return callback + */ + function setUploadCallback( $callback ) { + $previous = $this->mUploadCallback; + $this->mUploadCallback = $callback; + return $previous; + } + + /** + * Sets the action to perform as each log item reached. + * @param $callback callback + * @return callback + */ + function setLogItemCallback( $callback ) { + $previous = $this->mLogItemCallback; + $this->mLogItemCallback = $callback; + return $previous; + } + + /** + * Sets the action to perform when site info is encountered + * @param $callback callback + * @return callback + */ + function setSiteInfoCallback( $callback ) { + $previous = $this->mSiteInfoCallback; + $this->mSiteInfoCallback = $callback; + return $previous; + } + + /** + * Set a target namespace to override the defaults + */ + function setTargetNamespace( $namespace ) { + if( is_null( $namespace ) ) { + // Don't override namespaces + $this->mTargetNamespace = null; + } elseif( $namespace >= 0 ) { + // FIXME: Check for validity + $this->mTargetNamespace = intval( $namespace ); + } else { + return false; + } + } + + /** + * Default per-revision callback, performs the import. + * @param $revision WikiRevision + * @private + */ + function importRevision( $revision ) { + $dbw = wfGetDB( DB_MASTER ); + return $dbw->deadlockLoop( array( $revision, 'importOldRevision' ) ); + } + + /** + * Default per-revision callback, performs the import. + * @param $rev WikiRevision + * @private + */ + function importLogItem( $rev ) { + $dbw = wfGetDB( DB_MASTER ); + return $dbw->deadlockLoop( array( $rev, 'importLogItem' ) ); + } + + /** + * Dummy for now... + */ + function importUpload( $revision ) { + //$dbw = wfGetDB( DB_MASTER ); + //return $dbw->deadlockLoop( array( $revision, 'importUpload' ) ); + return false; + } + + /** + * Alternate per-revision callback, for debugging. + * @param $revision WikiRevision + * @private + */ + function debugRevisionHandler( &$revision ) { + $this->debug( "Got revision:" ); + if( is_object( $revision->title ) ) { + $this->debug( "-- Title: " . $revision->title->getPrefixedText() ); + } else { + $this->debug( "-- Title: " ); + } + $this->debug( "-- User: " . $revision->user_text ); + $this->debug( "-- Timestamp: " . $revision->timestamp ); + $this->debug( "-- Comment: " . $revision->comment ); + $this->debug( "-- Text: " . $revision->text ); + } + + /** + * Notify the callback function when a new is reached. + * @param $title Title + * @private + */ + function pageCallback( $title ) { + if( is_callable( $this->mPageCallback ) ) { + call_user_func( $this->mPageCallback, $title ); + } + } + + /** + * Notify the callback function when a is closed. + * @param $title Title + * @param $origTitle Title + * @param $revisionCount int + * @param $successCount Int: number of revisions for which callback returned true + * @private + */ + function pageOutCallback( $title, $origTitle, $revisionCount, $successCount ) { + if( is_callable( $this->mPageOutCallback ) ) { + call_user_func_array( $this->mPageOutCallback, + array( $title, $origTitle, $revisionCount, $successCount ) ); + } + } + + function revisionCallback( $revision ) { + if ( is_callable( $this->mRevisionCallback ) ) { + return call_user_func_array( $this->mRevisionCallback, + array( $revision, $this ) ); + } else { + return false; + } + } + + function logItemCallback( $revision ) { + if ( is_callable( $this->mLogItemCallback ) ) { + return call_user_func_array( $this->mLogItemCallback, + array( $revision, $this ) ); + } else { + return false; + } + } + + /** + * Shouldn't something like this be built-in to XMLReader? + * Fetches text contents of the current element, assuming + * no sub-elements or such scary things. + * @return string + * @access 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(); + } + + # -------------- + + function dumpElement() { + static $lookup = null; + if (!$lookup) { + $xmlReaderConstants = array( + "NONE", + "ELEMENT", + "ATTRIBUTE", + "TEXT", + "CDATA", + "ENTITY_REF", + "ENTITY", + "PI", + "COMMENT", + "DOC", + "DOC_TYPE", + "DOC_FRAGMENT", + "NOTATION", + "WHITESPACE", + "SIGNIFICANT_WHITESPACE", + "END_ELEMENT", + "END_ENTITY", + "XML_DECLARATION", + ); + $lookup = array(); + + foreach( $xmlReaderConstants as $name ) { + $lookup[constant("XmlReader::$name")] = $name; + } + } + + print( var_dump( + $lookup[$this->reader->nodeType], + $this->reader->name, + $this->reader->value + )."\n\n" ); + } + + function doImport() { + $this->reader->read(); + + if ( $this->reader->name != 'mediawiki' ) { + throw new MWException( "Expected tag, got ". + $this->reader->name ); + } + $this->debug( " 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 ) { + break; + } elseif ( $tag == 'siteinfo' ) { + $this->handleSiteInfo(); + } elseif ( $tag == 'page' ) { + $this->handlePage(); + } 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; + $this->debug( "Skip" ); + } else { + $keepReading = $this->reader->read(); + } + } + + return true; + } + + function handleSiteInfo() { + // Site info is useful, but not actually used for dump imports. + // Includes a quick short-circuit to save performance. + if ( ! $this->mSiteInfoCallback ) { + $this->reader->next(); + return true; + } + throw new MWException( "SiteInfo tag is not yet handled, do not set mSiteInfoCallback" ); + } + + 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 ) ) { + $logInfo[$tag] = $this->nodeContents(); + } elseif ( $tag == 'contributor' ) { + $logInfo['contributor'] = $this->handleContributor(); + } elseif ( $tag != '#text' ) { + $this->warn( "Unhandled log-item XML tag $tag" ); + } + } + + $this->processLogItem( $logInfo ); + } + + 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 ); + } + + 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 ) ) { + // Do nothing + } if ( 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; + } + } elseif ( $tag == 'revision' ) { + $this->handleRevision( $pageInfo ); + } elseif ( $tag == 'upload' ) { + $this->handleUpload( $pageInfo ); + } elseif ( $tag != '#text' ) { + $this->warn( "Unhandled page XML tag $tag" ); + $skip = true; + } + } + + $this->pageOutCallback( $pageInfo['_title'], $origTitle, + $pageInfo['revisionCount'], + $pageInfo['successfulRevisionCount'] ); + } + + function handleRevision( &$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 ) ) { + $revisionInfo[$tag] = $this->nodeContents(); + } elseif ( $tag == 'contributor' ) { + $revisionInfo['contributor'] = $this->handleContributor(); + } elseif ( $tag != '#text' ) { + $this->warn( "Unhandled revision XML tag $tag" ); + $skip = true; + } + } + + $pageInfo['revisionCount']++; + if ( $this->processRevision( $pageInfo, $revisionInfo ) ) { + $pageInfo['successfulRevisionCount']++; + } + } + + 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( $revision ); + } + + function handleUpload( &$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 ) ) { + $uploadInfo[$tag] = $this->nodeContents(); + } elseif ( $tag == 'contributor' ) { + $uploadInfo['contributor'] = $this->handleContributor(); + } elseif ( $tag != '#text' ) { + $this->warn( "Unhandled upload XML tag $tag" ); + $skip = true; + } + } + + return $this->processUpload( $pageInfo, $uploadInfo ); + } + + function processUpload( $pageInfo, $uploadInfo ) { + $revision = new WikiRevision; + + $revision->setTitle( $pageInfo['_title'] ); + $revision->setID( $uploadInfo['id'] ); + $revision->setTimestamp( $uploadInfo['timestamp'] ); + $revision->setText( $uploadInfo['text'] ); + $revision->setFilename( $uploadInfo['filename'] ); + $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 ); + } + + 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; + } + + 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'" ); + } elseif( $title->getInterwiki() != '' ) { + $this->notice( "Skipping interwiki page title '$workTitle'" ); + $title = null; + } + + return array( $origTitle, $title ); + } +} + +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 && !$this->mSource->atEnd() && + count($this->mBuffer) < $count ) { + $read = $this->mSource->readChunk(); + + if ( !count($read) ) { + $leave = true; + } + + $this->mBuffer .= $read; + } + + if ( count($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 stream_eof() { + return $this->mSource->atEnd(); + } + + function url_stat() { + $result = array(); + + $result['dev'] = $result[0] = 0; + $result['ino'] = $result[1] = 0; + $result['mode'] = $result[2] = 0; + $result['nlink'] = $result[3] = 0; + $result['uid'] = $result[4] = 0; + $result['gid'] = $result[5] = 0; + $result['rdev'] = $result[6] = 0; + $result['size'] = $result[7] = 0; + $result['atime'] = $result[8] = 0; + $result['mtime'] = $result[9] = 0; + $result['ctime'] = $result[10] = 0; + $result['blksize'] = $result[11] = 0; + $result['blocks'] = $result[12] = 0; + + return $result; + } +} diff --git a/includes/api/ApiImport.php b/includes/api/ApiImport.php index 8b7d1c13bb..5135ecf2d2 100644 --- a/includes/api/ApiImport.php +++ b/includes/api/ApiImport.php @@ -74,7 +74,7 @@ class ApiImport extends ApiBase { $this->dieUsageMsg( array( 'import-unknownerror', $source->getMessage() ) ); } - $importer = new WikiImporter( $source ); + $importer = new ImportXMLReader( $source ); if ( isset( $params['namespace'] ) ) { $importer->setTargetNamespace( $params['namespace'] ); } @@ -193,4 +193,4 @@ class ApiImportReporter extends ImportReporter { function getData() { return $this->mResultArr; } -} \ No newline at end of file +} diff --git a/includes/specials/SpecialImport.php b/includes/specials/SpecialImport.php index 6beeab7fd1..5e5cea4074 100644 --- a/includes/specials/SpecialImport.php +++ b/includes/specials/SpecialImport.php @@ -107,7 +107,7 @@ class SpecialImport extends SpecialPage { } else { $wgOut->addWikiMsg( "importstart" ); - $importer = new WikiImporter( $source ); + $importer = new ImportXMLReader( $source ); if( !is_null( $this->namespace ) ) { $importer->setTargetNamespace( $this->namespace ); } @@ -274,9 +274,13 @@ class SpecialImport extends SpecialPage { */ class ImportReporter { private $reason=false; + private $mOriginalLogCallback = null; + private $mLogItemCount = 0; function __construct( $importer, $upload, $interwiki , $reason=false ) { $importer->setPageOutCallback( array( $this, 'reportPage' ) ); + $this->mOriginalLogCallback = + $importer->setLogItemCallback( array( $this, 'reportLogItem' ) ); $this->mPageCount = 0; $this->mIsUpload = $upload; $this->mInterwiki = $interwiki; @@ -287,6 +291,13 @@ class ImportReporter { global $wgOut; $wgOut->addHTML( "
      \n" ); } + + function reportLogItem( /* ... */ ) { + $this->mLogItemCount++; + if ( is_callable( $this->mOriginalLogCallback ) ) { + call_user_func_array( $this->mOriginalLogCallback, func_get_args() ); + } + } function reportPage( $title, $origTitle, $revisionCount, $successCount ) { global $wgOut, $wgUser, $wgLang, $wgContLang; @@ -340,7 +351,12 @@ class ImportReporter { function close() { global $wgOut; - if( $this->mPageCount == 0 ) { + + if ( $this->mLogItemCount > 0 ) { + $msg = wfMsgExt( 'imported-log-entries', 'parseinline', + $this->mLogItemCount ); + $wgOut->addHTML( Xml::tags( 'li', null, $msg ) ); + } elseif( $this->mPageCount == 0 && $this->mLogItemCount == 0 ) { $wgOut->addHTML( "
    \n" ); return new WikiErrorMsg( "importnopages" ); } diff --git a/languages/messages/MessagesEn.php b/languages/messages/MessagesEn.php index 63e3c4e844..055d8269cd 100644 --- a/languages/messages/MessagesEn.php +++ b/languages/messages/MessagesEn.php @@ -3203,6 +3203,7 @@ Save it to your computer and upload it here.', 'importstart' => 'Importing pages...', 'import-revision-count' => '$1 {{PLURAL:$1|revision|revisions}}', 'importnopages' => 'No pages to import.', +'imported-log-entries' => 'Imported $1 {{PLURAL:$1|log entry|log entries}}.', 'importfailed' => 'Import failed: $1', 'importunknownsource' => 'Unknown import source type', 'importcantopen' => 'Could not open import file', diff --git a/maintenance/importDump.php b/maintenance/importDump.php index 714d76d875..c4f7c73424 100644 --- a/maintenance/importDump.php +++ b/maintenance/importDump.php @@ -140,7 +140,7 @@ class BackupReader { $this->startTime = wfTime(); $source = new ImportStreamSource( $handle ); - $importer = new WikiImporter( $source ); + $importer = new ImportXMLReader( $source ); $importer->setDebug( $this->debug ); $importer->setPageCallback( array( &$this, 'reportPage' ) ); diff --git a/maintenance/renderDump.php b/maintenance/renderDump.php index d36953f89f..c2e7d68db5 100644 --- a/maintenance/renderDump.php +++ b/maintenance/renderDump.php @@ -46,7 +46,7 @@ class DumpRenderer extends Maintenance { $this->startTime = wfTime(); $source = new ImportStreamSource( $this->getStdin() ); - $importer = new WikiImporter( $source ); + $importer = new ImportXMLReader( $source ); $importer->setRevisionCallback( array( &$this, 'handleRevision' ) ); -- 2.20.1