XmlDumpWriter.php

Go to the documentation of this file.

<?php
use MediaWiki\HookContainer\HookRunner;
use MediaWiki\MediaWikiServices;
use MediaWiki\Revision\RevisionRecord;
use MediaWiki\Revision\RevisionStore;
use MediaWiki\Revision\SlotRecord;
use MediaWiki\Revision\SuppressedDataException;
use MediaWiki\Storage\SqlBlobStore;
use Wikimedia\Assert\Assert;
use Wikimedia\IPUtils;
 
class XmlDumpWriter {
 
    public const WRITE_CONTENT = 0;
 
    public const WRITE_STUB = 1;
 
    private const WRITE_STUB_DELETED = 2;
 
    public static $supportedSchemas = [
        XML_DUMP_SCHEMA_VERSION_10,
        XML_DUMP_SCHEMA_VERSION_11
    ];
 
    private $schemaVersion;
 
    private $currentTitle = null;
 
    private $contentMode;
 
    private $hookRunner;
 
    public function __construct(
        $contentMode = self::WRITE_CONTENT,
        $schemaVersion = XML_DUMP_SCHEMA_VERSION_11
    ) {
        Assert::parameter(
            in_array( $contentMode, [ self::WRITE_CONTENT, self::WRITE_STUB ] ),
            '$contentMode',
            'must be one of the following constants: WRITE_CONTENT or WRITE_STUB.'
        );
 
        Assert::parameter(
            in_array( $schemaVersion, self::$supportedSchemas ),
            '$schemaVersion',
            'must be one of the following schema versions: '
                . implode( ',', self::$supportedSchemas )
        );
 
        $this->contentMode = $contentMode;
        $this->schemaVersion = $schemaVersion;
        $this->hookRunner = new HookRunner( MediaWikiServices::getInstance()->getHookContainer() );
    }
 
    public function openStream() {
        $ver = $this->schemaVersion;
        return Xml::element( 'mediawiki', [
            'xmlns'              => "http://www.mediawiki.org/xml/export-$ver/",
            'xmlns:xsi'          => "http://www.w3.org/2001/XMLSchema-instance",
            /*
             * When a new version of the schema is created, it needs staging on mediawiki.org.
             * This requires a change in the operations/mediawiki-config git repo.
             *
             * Create a changeset like https://gerrit.wikimedia.org/r/#/c/149643/ in which
             * you copy in the new xsd file.
             *
             * After it is reviewed, merged and deployed (sync-docroot), the index.html needs purging.
             * echo "https://www.mediawiki.org/xml/index.html" | mwscript purgeList.php --wiki=aawiki
             */
            'xsi:schemaLocation' => "http://www.mediawiki.org/xml/export-$ver/ " .
                "http://www.mediawiki.org/xml/export-$ver.xsd",
            'version' => $ver,
            'xml:lang' => MediaWikiServices::getInstance()->getContentLanguage()->getHtmlCode() ],
            null ) .
            "\n" .
            $this->siteInfo();
    }
 
    private function siteInfo() {
        $info = [
            $this->sitename(),
            $this->dbname(),
            $this->homelink(),
            $this->generator(),
            $this->caseSetting(),
            $this->namespaces() ];
        return "  <siteinfo>\n    " .
            implode( "\n    ", $info ) .
            "\n  </siteinfo>\n";
    }
 
    private function sitename() {
        $sitename = MediaWikiServices::getInstance()->getMainConfig()->get( 'Sitename' );
        return Xml::element( 'sitename', [], $sitename );
    }
 
    private function dbname() {
        $dbname = MediaWikiServices::getInstance()->getMainConfig()->get( 'DBname' );
        return Xml::element( 'dbname', [], $dbname );
    }
 
    private function generator() {
        return Xml::element( 'generator', [], 'MediaWiki ' . MW_VERSION );
    }
 
    private function homelink() {
        return Xml::element( 'base', [], Title::newMainPage()->getCanonicalURL() );
    }
 
    private function caseSetting() {
        $capitalLinks = MediaWikiServices::getInstance()->getMainConfig()->get( 'CapitalLinks' );
        // "case-insensitive" option is reserved for future
        $sensitivity = $capitalLinks ? 'first-letter' : 'case-sensitive';
        return Xml::element( 'case', [], $sensitivity );
    }
 
    private function namespaces() {
        $spaces = "<namespaces>\n";
        $nsInfo = MediaWikiServices::getInstance()->getNamespaceInfo();
        foreach (
            MediaWikiServices::getInstance()->getContentLanguage()->getFormattedNamespaces()
            as $ns => $title
        ) {
            $spaces .= '      ' .
                Xml::element( 'namespace',
                    [
                        'key' => $ns,
                        'case' => $nsInfo->isCapitalized( $ns )
                            ? 'first-letter' : 'case-sensitive',
                    ], $title ) . "\n";
        }
        $spaces .= "    </namespaces>";
        return $spaces;
    }
 
    public function closeStream() {
        return "</mediawiki>\n";
    }
 
    public function openPage( $row ) {
        $out = "  <page>\n";
        $this->currentTitle = Title::newFromRow( $row );
        $canonicalTitle = self::canonicalTitle( $this->currentTitle );
        $out .= '    ' . Xml::elementClean( 'title', [], $canonicalTitle ) . "\n";
        $out .= '    ' . Xml::element( 'ns', [], strval( $row->page_namespace ) ) . "\n";
        $out .= '    ' . Xml::element( 'id', [], strval( $row->page_id ) ) . "\n";
        if ( $row->page_is_redirect ) {
            $services = MediaWikiServices::getInstance();
            $page = $services->getWikiPageFactory()->newFromTitle( $this->currentTitle );
            $redirectStore = $services->getRedirectStore();
            $redirect = $this->invokeLenient(
                static function () use ( $page, $redirectStore ) {
                    return $redirectStore->getRedirectTarget( $page );
                },
                'Failed to get redirect target of page ' . $page->getId()
            );
            if ( $redirect instanceof Title && $redirect->isValidRedirectTarget() ) {
                $out .= '    ';
                $out .= Xml::element( 'redirect', [ 'title' => self::canonicalTitle( $redirect ) ] );
                $out .= "\n";
            }
        }
 
        if ( $row->page_restrictions != '' ) {
            $out .= '    ' . Xml::element( 'restrictions', [],
                strval( $row->page_restrictions ) ) . "\n";
        }
 
        $this->hookRunner->onXmlDumpWriterOpenPage( $this, $out, $row, $this->currentTitle );
 
        return $out;
    }
 
    public function closePage() {
        if ( $this->currentTitle !== null ) {
            $linkCache = MediaWikiServices::getInstance()->getLinkCache();
            // In rare cases, link cache has the same key for some pages which
            // might be read as part of the same batch. T220424 and T220316
            $linkCache->clearLink( $this->currentTitle );
        }
        return "  </page>\n";
    }
 
    private function getRevisionStore() {
        return MediaWikiServices::getInstance()->getRevisionStore();
    }
 
    private function getBlobStore() {
        // @phan-suppress-next-line PhanTypeMismatchReturnSuperType
        return MediaWikiServices::getInstance()->getBlobStore();
    }
 
    private function invokeLenient( $callback, $warning ) {
        try {
            return $callback();
        } catch ( SuppressedDataException $ex ) {
            return null;
        } catch ( Exception $ex ) {
            if ( $ex instanceof MWException || $ex instanceof RuntimeException ||
                $ex instanceof InvalidArgumentException ) {
                MWDebug::warning( $warning . ': ' . $ex->getMessage() );
                return null;
            } else {
                throw $ex;
            }
        }
    }
 
    public function writeRevision( $row, $slotRows = null ) {
        $rev = $this->getRevisionStore()->newRevisionFromRowAndSlots(
            $row,
            $slotRows,
            0,
            $this->currentTitle
        );
 
        $out = "    <revision>\n";
        $out .= "      " . Xml::element( 'id', null, strval( $rev->getId() ) ) . "\n";
 
        if ( $rev->getParentId() ) {
            $out .= "      " . Xml::element( 'parentid', null, strval( $rev->getParentId() ) ) . "\n";
        }
 
        $out .= $this->writeTimestamp( $rev->getTimestamp() );
 
        if ( $rev->isDeleted( RevisionRecord::DELETED_USER ) ) {
            $out .= "      " . Xml::element( 'contributor', [ 'deleted' => 'deleted' ] ) . "\n";
        } else {
            // empty values get written out as uid 0, see T224221
            $user = $rev->getUser();
            $out .= $this->writeContributor(
                $user ? $user->getId() : 0,
                $user ? $user->getName() : ''
            );
        }
 
        if ( $rev->isMinor() ) {
            $out .= "      <minor/>\n";
        }
        if ( $rev->isDeleted( RevisionRecord::DELETED_COMMENT ) ) {
            $out .= "      " . Xml::element( 'comment', [ 'deleted' => 'deleted' ] ) . "\n";
        } else {
            if ( $rev->getComment()->text != '' ) {
                $out .= "      "
                    // @phan-suppress-next-line SecurityCheck-DoubleEscaped getComment is polluted by truncate
                    . Xml::elementClean( 'comment', [], strval( $rev->getComment()->text ) )
                    . "\n";
            }
        }
 
        $contentMode = $rev->isDeleted( RevisionRecord::DELETED_TEXT ) ? self::WRITE_STUB_DELETED
            : $this->contentMode;
 
        $slots = $rev->getSlots()->getSlots();
 
        // use predictable order, put main slot first
        ksort( $slots );
        $out .= $this->writeSlot( $slots[SlotRecord::MAIN], $contentMode );
 
        foreach ( $slots as $role => $slot ) {
            if ( $role === SlotRecord::MAIN ) {
                continue;
            }
            $out .= $this->writeSlot( $slot, $contentMode );
        }
 
        if ( $rev->isDeleted( RevisionRecord::DELETED_TEXT ) ) {
            $out .= "      <sha1/>\n";
        } else {
            $sha1 = $this->invokeLenient(
                static function () use ( $rev ) {
                    return $rev->getSha1();
                },
                'failed to determine sha1 for revision ' . $rev->getId()
            );
            $out .= "      " . Xml::element( 'sha1', null, strval( $sha1 ) ) . "\n";
        }
 
        // Avoid PHP 7.1 warning from passing $this by reference
        $writer = $this;
        $text = '';
        if ( $contentMode === self::WRITE_CONTENT ) {
            $content = $this->invokeLenient(
                static function () use ( $rev ) {
                    return $rev->getContent( SlotRecord::MAIN, RevisionRecord::RAW );
                },
                'Failed to load main slot content of revision ' . $rev->getId()
            );
 
            $text = $content ? $content->serialize() : '';
        }
        $this->hookRunner->onXmlDumpWriterWriteRevision( $writer, $out, $row, $text, $rev );
 
        $out .= "    </revision>\n";
 
        return $out;
    }
 
    private function writeSlot( SlotRecord $slot, $contentMode ) {
        $isMain = $slot->getRole() === SlotRecord::MAIN;
        $isV11 = $this->schemaVersion >= XML_DUMP_SCHEMA_VERSION_11;
 
        if ( !$isV11 && !$isMain ) {
            // ignore extra slots
            return '';
        }
 
        $out = '';
        $indent = '      ';
 
        if ( !$isMain ) {
            // non-main slots are wrapped into an additional element.
            $out .= '      ' . Xml::openElement( 'content' ) . "\n";
            $indent .= '  ';
            $out .= $indent . Xml::element( 'role', null, strval( $slot->getRole() ) ) . "\n";
        }
 
        if ( $isV11 ) {
            $out .= $indent . Xml::element( 'origin', null, strval( $slot->getOrigin() ) ) . "\n";
        }
 
        $contentModel = $slot->getModel();
        $contentHandler = MediaWikiServices::getInstance()
            ->getContentHandlerFactory()
            ->getContentHandler( $contentModel );
        $contentFormat = $contentHandler->getDefaultFormat();
 
        // XXX: The content format is only relevant when actually outputting serialized content.
        // It should probably be an attribute on the text tag.
        $out .= $indent . Xml::element( 'model', null, strval( $contentModel ) ) . "\n";
        $out .= $indent . Xml::element( 'format', null, strval( $contentFormat ) ) . "\n";
 
        $textAttributes = [
            'bytes' => $this->invokeLenient(
                static function () use ( $slot ) {
                    return $slot->getSize();
                },
                'failed to determine size for slot ' . $slot->getRole() . ' of revision '
                . $slot->getRevision()
            ) ?: '0'
        ];
 
        if ( $isV11 ) {
            $textAttributes['sha1'] = $this->invokeLenient(
                static function () use ( $slot ) {
                    return $slot->getSha1();
                },
                'failed to determine sha1 for slot ' . $slot->getRole() . ' of revision '
                . $slot->getRevision()
            ) ?: '';
        }
 
        if ( $contentMode === self::WRITE_CONTENT ) {
            $content = $this->invokeLenient(
                static function () use ( $slot ) {
                    return $slot->getContent();
                },
                'failed to load content for slot ' . $slot->getRole() . ' of revision '
                . $slot->getRevision()
            );
 
            if ( $content === null ) {
                $out .= $indent . Xml::element( 'text', $textAttributes ) . "\n";
            } else {
                $out .= $this->writeText( $content, $textAttributes, $indent );
            }
        } elseif ( $contentMode === self::WRITE_STUB_DELETED ) {
            // write <text> placeholder tag
            $textAttributes['deleted'] = 'deleted';
            $out .= $indent . Xml::element( 'text', $textAttributes ) . "\n";
        } else {
            // write <text> stub tag
            if ( $isV11 ) {
                $textAttributes['location'] = $slot->getAddress();
            }
 
            if ( $isMain ) {
                // Output the numerical text ID if possible, for backwards compatibility.
                // Note that this is currently the ONLY reason we have a BlobStore here at all.
                // When removing this line, check whether the BlobStore has become unused.
                try {
                    // NOTE: this will only work for addresses of the form "tt:12345".
                    // If we want to support other kinds of addresses in the future,
                    // we will have to silently ignore failures here.
                    // For now, this fails for "tt:0", which is present in the WMF production
                    // database as of July 2019, due to data corruption.
                    $textId = $this->getBlobStore()->getTextIdFromAddress( $slot->getAddress() );
                } catch ( InvalidArgumentException $ex ) {
                    MWDebug::warning( 'Bad content address for slot ' . $slot->getRole()
                        . ' of revision ' . $slot->getRevision() . ': ' . $ex->getMessage() );
                    $textId = 0;
                }
 
                if ( is_int( $textId ) ) {
                    $textAttributes['id'] = $textId;
                }
            }
 
            $out .= $indent . Xml::element( 'text', $textAttributes ) . "\n";
        }
 
        if ( !$isMain ) {
            $out .= '      ' . Xml::closeElement( 'content' ) . "\n";
        }
 
        return $out;
    }
 
    private function writeText( Content $content, $textAttributes, $indent ) {
        $out = '';
 
        $contentHandler = $content->getContentHandler();
        $contentFormat = $contentHandler->getDefaultFormat();
 
        if ( $content instanceof TextContent ) {
            // HACK: For text based models, bypass the serialization step. This allows extensions (like Flow)
            // that use incompatible combinations of serialization format and content model.
            $data = $content->getNativeData();
        } else {
            $data = $content->serialize( $contentFormat );
        }
 
        $data = $contentHandler->exportTransform( $data, $contentFormat );
        $textAttributes['bytes'] = $size = strlen( $data ); // make sure to use the actual size
        $textAttributes['xml:space'] = 'preserve';
        $out .= $indent . Xml::elementClean( 'text', $textAttributes, strval( $data ) ) . "\n";
 
        return $out;
    }
 
    public function writeLogItem( $row ) {
        $out = "  <logitem>\n";
        $out .= "    " . Xml::element( 'id', null, strval( $row->log_id ) ) . "\n";
 
        $out .= $this->writeTimestamp( $row->log_timestamp, "    " );
 
        if ( $row->log_deleted & LogPage::DELETED_USER ) {
            $out .= "    " . Xml::element( 'contributor', [ 'deleted' => 'deleted' ] ) . "\n";
        } else {
            $out .= $this->writeContributor( $row->actor_user, $row->actor_name, "    " );
        }
 
        if ( $row->log_deleted & LogPage::DELETED_COMMENT ) {
            $out .= "    " . Xml::element( 'comment', [ 'deleted' => 'deleted' ] ) . "\n";
        } else {
            $comment = CommentStore::getStore()->getComment( 'log_comment', $row )->text;
            if ( $comment != '' ) {
                // @phan-suppress-next-line SecurityCheck-DoubleEscaped CommentStore is polluted by truncate
                $out .= "    " . Xml::elementClean( 'comment', null, strval( $comment ) ) . "\n";
            }
        }
 
        $out .= "    " . Xml::element( 'type', null, strval( $row->log_type ) ) . "\n";
        $out .= "    " . Xml::element( 'action', null, strval( $row->log_action ) ) . "\n";
 
        if ( $row->log_deleted & LogPage::DELETED_ACTION ) {
            $out .= "    " . Xml::element( 'text', [ 'deleted' => 'deleted' ] ) . "\n";
        } else {
            $title = Title::makeTitle( $row->log_namespace, $row->log_title );
            $out .= "    " . Xml::elementClean( 'logtitle', null, self::canonicalTitle( $title ) ) . "\n";
            $out .= "    " . Xml::elementClean( 'params',
                [ 'xml:space' => 'preserve' ],
                strval( $row->log_params ) ) . "\n";
        }
 
        $out .= "  </logitem>\n";
 
        return $out;
    }
 
    public function writeTimestamp( $timestamp, $indent = "      " ) {
        $ts = wfTimestamp( TS_ISO_8601, $timestamp );
        return $indent . Xml::element( 'timestamp', null, $ts ) . "\n";
    }
 
    public function writeContributor( $id, $text, $indent = "      " ) {
        $out = $indent . "<contributor>\n";
        if ( $id || !IPUtils::isValid( $text ) ) {
            $out .= $indent . "  " . Xml::elementClean( 'username', null, strval( $text ) ) . "\n";
            $out .= $indent . "  " . Xml::element( 'id', null, strval( $id ) ) . "\n";
        } else {
            $out .= $indent . "  " . Xml::elementClean( 'ip', null, strval( $text ) ) . "\n";
        }
        $out .= $indent . "</contributor>\n";
        return $out;
    }
 
    public function writeUploads( $row, $dumpContents = false ) {
        if ( $row->page_namespace == NS_FILE ) {
            $img = MediaWikiServices::getInstance()->getRepoGroup()->getLocalRepo()
                ->newFile( $row->page_title );
            if ( $img && $img->exists() ) {
                $out = '';
                foreach ( array_reverse( $img->getHistory() ) as $ver ) {
                    $out .= $this->writeUpload( $ver, $dumpContents );
                }
                $out .= $this->writeUpload( $img, $dumpContents );
                return $out;
            }
        }
        return '';
    }
 
    private function writeUpload( $file, $dumpContents = false ) {
        if ( $file->isOld() ) {
            '@phan-var OldLocalFile $file';
            $archiveName = "      " .
                Xml::element( 'archivename', null, $file->getArchiveName() ) . "\n";
        } else {
            $archiveName = '';
        }
        if ( $dumpContents ) {
            $be = $file->getRepo()->getBackend();
            # Dump file as base64
            # Uses only XML-safe characters, so does not need escaping
            # @todo Too bad this loads the contents into memory (script might swap)
            $contents = '      <contents encoding="base64">' .
                chunk_split( base64_encode(
                    $be->getFileContents( [ 'src' => $file->getPath() ] ) ) ) .
                "      </contents>\n";
        } else {
            $contents = '';
        }
        $uploader = $file->getUploader( File::FOR_PUBLIC );
        if ( $uploader ) {
            $uploader = $this->writeContributor( $uploader->getId(), $uploader->getName() );
        } else {
            $uploader = Xml::element( 'contributor', [ 'deleted' => 'deleted' ] ) . "\n";
        }
        $comment = $file->getDescription( File::FOR_PUBLIC );
        if ( ( $comment ?? '' ) !== '' ) {
            $comment = Xml::elementClean( 'comment', null, $comment );
        } else {
            $comment = Xml::element( 'comment', [ 'deleted' => 'deleted' ] );
        }
        return "    <upload>\n" .
            $this->writeTimestamp( $file->getTimestamp() ) .
            $uploader .
            "      " . $comment . "\n" .
            "      " . Xml::element( 'filename', null, $file->getName() ) . "\n" .
            $archiveName .
            "      " . Xml::element( 'src', null, $file->getCanonicalUrl() ) . "\n" .
            "      " . Xml::element( 'size', null, $file->getSize() ) . "\n" .
            "      " . Xml::element( 'sha1base36', null, $file->getSha1() ) . "\n" .
            "      " . Xml::element( 'rel', null, $file->getRel() ) . "\n" .
            $contents .
            "    </upload>\n";
    }
 
    public static function canonicalTitle( Title $title ) {
        if ( $title->isExternal() ) {
            return $title->getPrefixedText();
        }
 
        $prefix = MediaWikiServices::getInstance()->getContentLanguage()->
            getFormattedNsText( $title->getNamespace() );
 
        // @todo Emit some kind of warning to the user if $title->getNamespace() !==
        // NS_MAIN and $prefix === '' (viz. pages in an unregistered namespace)
 
        if ( $prefix !== '' ) {
            $prefix .= ':';
        }
 
        return $prefix . $title->getText();
    }
}