WikiExporter.php

Go to the documentation of this file.

<?php
use Wikimedia\Rdbms\ResultWrapper;
use Wikimedia\Rdbms\IDatabase;
 
class WikiExporter {
    public $list_authors = false;
 
    public $dumpUploads = false;
 
    public $dumpUploadFileContents = false;
 
    public $author_list = "";
 
    const FULL = 1;
    const CURRENT = 2;
    const STABLE = 4; // extension defined
    const LOGS = 8;
    const RANGE = 16;
 
    const BUFFER = 0;
    const STREAM = 1;
 
    const TEXT = 0;
    const STUB = 1;
 
    public $buffer;
 
    public $text;
 
    public $sink;
 
    public static function schemaVersion() {
        return "0.10";
    }
 
    function __construct( $db, $history = self::CURRENT,
            $buffer = self::BUFFER, $text = self::TEXT ) {
        $this->db = $db;
        $this->history = $history;
        $this->buffer = $buffer;
        $this->writer = new XmlDumpWriter();
        $this->sink = new DumpOutput();
        $this->text = $text;
    }
 
    public function setOutputSink( &$sink ) {
        $this->sink =& $sink;
    }
 
    public function openStream() {
        $output = $this->writer->openStream();
        $this->sink->writeOpenStream( $output );
    }
 
    public function closeStream() {
        $output = $this->writer->closeStream();
        $this->sink->writeCloseStream( $output );
    }
 
    public function allPages() {
        $this->dumpFrom( '' );
    }
 
    public function pagesByRange( $start, $end, $orderRevs ) {
        if ( $orderRevs ) {
            $condition = 'rev_page >= ' . intval( $start );
            if ( $end ) {
                $condition .= ' AND rev_page < ' . intval( $end );
            }
        } else {
            $condition = 'page_id >= ' . intval( $start );
            if ( $end ) {
                $condition .= ' AND page_id < ' . intval( $end );
            }
        }
        $this->dumpFrom( $condition, $orderRevs );
    }
 
    public function revsByRange( $start, $end ) {
        $condition = 'rev_id >= ' . intval( $start );
        if ( $end ) {
            $condition .= ' AND rev_id < ' . intval( $end );
        }
        $this->dumpFrom( $condition );
    }
 
    public function pageByTitle( $title ) {
        $this->dumpFrom(
            'page_namespace=' . $title->getNamespace() .
            ' AND page_title=' . $this->db->addQuotes( $title->getDBkey() ) );
    }
 
    public function pageByName( $name ) {
        $title = Title::newFromText( $name );
        if ( is_null( $title ) ) {
            throw new MWException( "Can't export invalid title" );
        } else {
            $this->pageByTitle( $title );
        }
    }
 
    public function pagesByName( $names ) {
        foreach ( $names as $name ) {
            $this->pageByName( $name );
        }
    }
 
    public function allLogs() {
        $this->dumpFrom( '' );
    }
 
    public function logsByRange( $start, $end ) {
        $condition = 'log_id >= ' . intval( $start );
        if ( $end ) {
            $condition .= ' AND log_id < ' . intval( $end );
        }
        $this->dumpFrom( $condition );
    }
 
    protected function do_list_authors( $cond ) {
        $this->author_list = "<contributors>";
        // rev_deleted
 
        $res = $this->db->select(
            [ 'page', 'revision' ],
            [ 'DISTINCT rev_user_text', 'rev_user' ],
            [
                $this->db->bitAnd( 'rev_deleted', Revision::DELETED_USER ) . ' = 0',
                $cond,
                'page_id = rev_id',
            ],
            __METHOD__
        );
 
        foreach ( $res as $row ) {
            $this->author_list .= "<contributor>" .
                "<username>" .
                htmlentities( $row->rev_user_text ) .
                "</username>" .
                "<id>" .
                $row->rev_user .
                "</id>" .
                "</contributor>";
        }
        $this->author_list .= "</contributors>";
    }
 
    protected function dumpFrom( $cond = '', $orderRevs = false ) {
        # For logging dumps...
        if ( $this->history & self::LOGS ) {
            $where = [];
            # Hide private logs
            $hideLogs = LogEventsList::getExcludeClause( $this->db );
            if ( $hideLogs ) {
                $where[] = $hideLogs;
            }
            # Add on any caller specified conditions
            if ( $cond ) {
                $where[] = $cond;
            }
            # Get logging table name for logging.* clause
            $logging = $this->db->tableName( 'logging' );
 
            if ( $this->buffer == self::STREAM ) {
                $prev = $this->db->bufferResults( false );
            }
            $result = null; // Assuring $result is not undefined, if exception occurs early
 
            $commentQuery = CommentStore::newKey( 'log_comment' )->getJoin();
 
            try {
                $result = $this->db->select( [ 'logging', 'user' ] + $commentQuery['tables'],
                    [ "{$logging}.*", 'user_name' ] + $commentQuery['fields'], // grab the user name
                    $where,
                    __METHOD__,
                    [ 'ORDER BY' => 'log_id', 'USE INDEX' => [ 'logging' => 'PRIMARY' ] ],
                    [ 'user' => [ 'JOIN', 'user_id = log_user' ] ] + $commentQuery['joins']
                );
                $this->outputLogStream( $result );
                if ( $this->buffer == self::STREAM ) {
                    $this->db->bufferResults( $prev );
                }
            } catch ( Exception $e ) {
                // Throwing the exception does not reliably free the resultset, and
                // would also leave the connection in unbuffered mode.
 
                // Freeing result
                try {
                    if ( $result ) {
                        $result->free();
                    }
                } catch ( Exception $e2 ) {
                    // Already in panic mode -> ignoring $e2 as $e has
                    // higher priority
                }
 
                // Putting database back in previous buffer mode
                try {
                    if ( $this->buffer == self::STREAM ) {
                        $this->db->bufferResults( $prev );
                    }
                } catch ( Exception $e2 ) {
                    // Already in panic mode -> ignoring $e2 as $e has
                    // higher priority
                }
 
                // Inform caller about problem
                throw $e;
            }
        # For page dumps...
        } else {
            $tables = [ 'page', 'revision' ];
            $opts = [ 'ORDER BY' => 'page_id ASC' ];
            $opts['USE INDEX'] = [];
            $join = [];
            if ( is_array( $this->history ) ) {
                # Time offset/limit for all pages/history...
                $revJoin = 'page_id=rev_page';
                # Set time order
                if ( $this->history['dir'] == 'asc' ) {
                    $op = '>';
                    $opts['ORDER BY'] = 'rev_timestamp ASC';
                } else {
                    $op = '<';
                    $opts['ORDER BY'] = 'rev_timestamp DESC';
                }
                # Set offset
                if ( !empty( $this->history['offset'] ) ) {
                    $revJoin .= " AND rev_timestamp $op " .
                        $this->db->addQuotes( $this->db->timestamp( $this->history['offset'] ) );
                }
                $join['revision'] = [ 'INNER JOIN', $revJoin ];
                # Set query limit
                if ( !empty( $this->history['limit'] ) ) {
                    $opts['LIMIT'] = intval( $this->history['limit'] );
                }
            } elseif ( $this->history & self::FULL ) {
                # Full history dumps...
                # query optimization for history stub dumps
                if ( $this->text == self::STUB && $orderRevs ) {
                    $tables = [ 'revision', 'page' ];
                    $opts[] = 'STRAIGHT_JOIN';
                    $opts['ORDER BY'] = [ 'rev_page ASC', 'rev_id ASC' ];
                    $opts['USE INDEX']['revision'] = 'rev_page_id';
                    $join['page'] = [ 'INNER JOIN', 'rev_page=page_id' ];
                } else {
                    $join['revision'] = [ 'INNER JOIN', 'page_id=rev_page' ];
                }
            } elseif ( $this->history & self::CURRENT ) {
                # Latest revision dumps...
                if ( $this->list_authors && $cond != '' ) { // List authors, if so desired
                    $this->do_list_authors( $cond );
                }
                $join['revision'] = [ 'INNER JOIN', 'page_id=rev_page AND page_latest=rev_id' ];
            } elseif ( $this->history & self::STABLE ) {
                # "Stable" revision dumps...
                # Default JOIN, to be overridden...
                $join['revision'] = [ 'INNER JOIN', 'page_id=rev_page AND page_latest=rev_id' ];
                # One, and only one hook should set this, and return false
                if ( Hooks::run( 'WikiExporter::dumpStableQuery', [ &$tables, &$opts, &$join ] ) ) {
                    throw new MWException( __METHOD__ . " given invalid history dump type." );
                }
            } elseif ( $this->history & self::RANGE ) {
                # Dump of revisions within a specified range
                $join['revision'] = [ 'INNER JOIN', 'page_id=rev_page' ];
                $opts['ORDER BY'] = [ 'rev_page ASC', 'rev_id ASC' ];
            } else {
                # Unknown history specification parameter?
                throw new MWException( __METHOD__ . " given invalid history dump type." );
            }
            # Query optimization hacks
            if ( $cond == '' ) {
                $opts[] = 'STRAIGHT_JOIN';
                $opts['USE INDEX']['page'] = 'PRIMARY';
            }
            # Build text join options
            if ( $this->text != self::STUB ) { // 1-pass
                $tables[] = 'text';
                $join['text'] = [ 'INNER JOIN', 'rev_text_id=old_id' ];
            }
 
            if ( $this->buffer == self::STREAM ) {
                $prev = $this->db->bufferResults( false );
            }
            $result = null; // Assuring $result is not undefined, if exception occurs early
            try {
                Hooks::run( 'ModifyExportQuery',
                        [ $this->db, &$tables, &$cond, &$opts, &$join ] );
 
                $commentQuery = CommentStore::newKey( 'rev_comment' )->getJoin();
 
                # Do the query!
                $result = $this->db->select(
                    $tables + $commentQuery['tables'],
                    [ '*' ] + $commentQuery['fields'],
                    $cond,
                    __METHOD__,
                    $opts,
                    $join + $commentQuery['joins']
                );
                # Output dump results
                $this->outputPageStream( $result );
 
                if ( $this->buffer == self::STREAM ) {
                    $this->db->bufferResults( $prev );
                }
            } catch ( Exception $e ) {
                // Throwing the exception does not reliably free the resultset, and
                // would also leave the connection in unbuffered mode.
 
                // Freeing result
                try {
                    if ( $result ) {
                        $result->free();
                    }
                } catch ( Exception $e2 ) {
                    // Already in panic mode -> ignoring $e2 as $e has
                    // higher priority
                }
 
                // Putting database back in previous buffer mode
                try {
                    if ( $this->buffer == self::STREAM ) {
                        $this->db->bufferResults( $prev );
                    }
                } catch ( Exception $e2 ) {
                    // Already in panic mode -> ignoring $e2 as $e has
                    // higher priority
                }
 
                // Inform caller about problem
                throw $e;
            }
        }
    }
 
    protected function outputPageStream( $resultset ) {
        $last = null;
        foreach ( $resultset as $row ) {
            if ( $last === null ||
                $last->page_namespace != $row->page_namespace ||
                $last->page_title != $row->page_title ) {
                if ( $last !== null ) {
                    $output = '';
                    if ( $this->dumpUploads ) {
                        $output .= $this->writer->writeUploads( $last, $this->dumpUploadFileContents );
                    }
                    $output .= $this->writer->closePage();
                    $this->sink->writeClosePage( $output );
                }
                $output = $this->writer->openPage( $row );
                $this->sink->writeOpenPage( $row, $output );
                $last = $row;
            }
            $output = $this->writer->writeRevision( $row );
            $this->sink->writeRevision( $row, $output );
        }
        if ( $last !== null ) {
            $output = '';
            if ( $this->dumpUploads ) {
                $output .= $this->writer->writeUploads( $last, $this->dumpUploadFileContents );
            }
            $output .= $this->author_list;
            $output .= $this->writer->closePage();
            $this->sink->writeClosePage( $output );
        }
    }
 
    protected function outputLogStream( $resultset ) {
        foreach ( $resultset as $row ) {
            $output = $this->writer->writeLogItem( $row );
            $this->sink->writeLogItem( $row, $output );
        }
    }
}