backupTextPass.inc

Go to the documentation of this file.

<?php
require_once __DIR__ . '/backup.inc';

class TextPassDumper extends BackupDumper {
    public $prefetch = null;

    // when we spend more than maxTimeAllowed seconds on this run, we continue
    // processing until we write out the next complete page, then save output file(s),
    // rename it/them and open new one(s)
    public $maxTimeAllowed = 0; // 0 = no limit

    protected $input = "php://stdin";
    protected $history = WikiExporter::FULL;
    protected $fetchCount = 0;
    protected $prefetchCount = 0;
    protected $prefetchCountLast = 0;
    protected $fetchCountLast = 0;

    protected $maxFailures = 5;
    protected $maxConsecutiveFailedTextRetrievals = 200;
    protected $failureTimeout = 5; // Seconds to sleep after db failure

    protected $php = "php";
    protected $spawn = false;

    protected $spawnProc = false;

    protected $spawnWrite = false;

    protected $spawnRead = false;

    protected $spawnErr = false;

    protected $xmlwriterobj = false;

    protected $timeExceeded = false;
    protected $firstPageWritten = false;
    protected $lastPageWritten = false;
    protected $checkpointJustWritten = false;
    protected $checkpointFiles = array();

    protected $db;

    function rotateDb() {
        // Cleaning up old connections
        if ( isset( $this->lb ) ) {
            $this->lb->closeAll();
            unset( $this->lb );
        }

        if ( $this->forcedDb !== null ) {
            $this->db = $this->forcedDb;

            return;
        }

        if ( isset( $this->db ) && $this->db->isOpen() ) {
            throw new MWException( 'DB is set and has not been closed by the Load Balancer' );
        }

        unset( $this->db );

        // Trying to set up new connection.
        // We do /not/ retry upon failure, but delegate to encapsulating logic, to avoid
        // individually retrying at different layers of code.

        // 1. The LoadBalancer.
        try {
            $this->lb = wfGetLBFactory()->newMainLB();
        } catch ( Exception $e ) {
            throw new MWException( __METHOD__
                . " rotating DB failed to obtain new load balancer (" . $e->getMessage() . ")" );
        }

        // 2. The Connection, through the load balancer.
        try {
            $this->db = $this->lb->getConnection( DB_SLAVE, 'dump' );
        } catch ( Exception $e ) {
            throw new MWException( __METHOD__
                . " rotating DB failed to obtain new database (" . $e->getMessage() . ")" );
        }
    }

    function initProgress( $history = WikiExporter::FULL ) {
        parent::initProgress();
        $this->timeOfCheckpoint = $this->startTime;
    }

    function dump( $history, $text = WikiExporter::TEXT ) {
        // Notice messages will foul up your XML output even if they're
        // relatively harmless.
        if ( ini_get( 'display_errors' ) ) {
            ini_set( 'display_errors', 'stderr' );
        }

        $this->initProgress( $this->history );

        // We are trying to get an initial database connection to avoid that the
        // first try of this request's first call to getText fails. However, if
        // obtaining a good DB connection fails it's not a serious issue, as
        // getText does retry upon failure and can start without having a working
        // DB connection.
        try {
            $this->rotateDb();
        } catch ( Exception $e ) {
            // We do not even count this as failure. Just let eventual
            // watchdogs know.
            $this->progress( "Getting initial DB connection failed (" .
                $e->getMessage() . ")" );
        }

        $this->egress = new ExportProgressFilter( $this->sink, $this );

        // it would be nice to do it in the constructor, oh well. need egress set
        $this->finalOptionCheck();

        // we only want this so we know how to close a stream :-P
        $this->xmlwriterobj = new XmlDumpWriter();

        $input = fopen( $this->input, "rt" );
        $this->readDump( $input );

        if ( $this->spawnProc ) {
            $this->closeSpawn();
        }

        $this->report( true );
    }

    function processOption( $opt, $val, $param ) {
        global $IP;
        $url = $this->processFileOpt( $val, $param );

        switch ( $opt ) {
            case 'prefetch':
                require_once "$IP/maintenance/backupPrefetch.inc";
                $this->prefetch = new BaseDump( $url );
                break;
            case 'stub':
                $this->input = $url;
                break;
            case 'maxtime':
                $this->maxTimeAllowed = intval( $val ) * 60;
                break;
            case 'checkpointfile':
                $this->checkpointFiles[] = $val;
                break;
            case 'current':
                $this->history = WikiExporter::CURRENT;
                break;
            case 'full':
                $this->history = WikiExporter::FULL;
                break;
            case 'spawn':
                $this->spawn = true;
                if ( $val ) {
                    $this->php = $val;
                }
                break;
        }
    }

    function processFileOpt( $val, $param ) {
        $fileURIs = explode( ';', $param );
        foreach ( $fileURIs as $URI ) {
            switch ( $val ) {
                case "file":
                    $newURI = $URI;
                    break;
                case "gzip":
                    $newURI = "compress.zlib://$URI";
                    break;
                case "bzip2":
                    $newURI = "compress.bzip2://$URI";
                    break;
                case "7zip":
                    $newURI = "mediawiki.compress.7z://$URI";
                    break;
                default:
                    $newURI = $URI;
            }
            $newFileURIs[] = $newURI;
        }
        $val = implode( ';', $newFileURIs );

        return $val;
    }

    function showReport() {
        if ( !$this->prefetch ) {
            parent::showReport();

            return;
        }

        if ( $this->reporting ) {
            $now = wfTimestamp( TS_DB );
            $nowts = microtime( true );
            $deltaAll = $nowts - $this->startTime;
            $deltaPart = $nowts - $this->lastTime;
            $this->pageCountPart = $this->pageCount - $this->pageCountLast;
            $this->revCountPart = $this->revCount - $this->revCountLast;

            if ( $deltaAll ) {
                $portion = $this->revCount / $this->maxCount;
                $eta = $this->startTime + $deltaAll / $portion;
                $etats = wfTimestamp( TS_DB, intval( $eta ) );
                if ( $this->fetchCount ) {
                    $fetchRate = 100.0 * $this->prefetchCount / $this->fetchCount;
                } else {
                    $fetchRate = '-';
                }
                $pageRate = $this->pageCount / $deltaAll;
                $revRate = $this->revCount / $deltaAll;
            } else {
                $pageRate = '-';
                $revRate = '-';
                $etats = '-';
                $fetchRate = '-';
            }
            if ( $deltaPart ) {
                if ( $this->fetchCountLast ) {
                    $fetchRatePart = 100.0 * $this->prefetchCountLast / $this->fetchCountLast;
                } else {
                    $fetchRatePart = '-';
                }
                $pageRatePart = $this->pageCountPart / $deltaPart;
                $revRatePart = $this->revCountPart / $deltaPart;
            } else {
                $fetchRatePart = '-';
                $pageRatePart = '-';
                $revRatePart = '-';
            }
            $this->progress( sprintf(
                "%s: %s (ID %d) %d pages (%0.1f|%0.1f/sec all|curr), "
                    . "%d revs (%0.1f|%0.1f/sec all|curr), %0.1f%%|%0.1f%% "
                    . "prefetched (all|curr), ETA %s [max %d]",
                $now, wfWikiID(), $this->ID, $this->pageCount, $pageRate,
                $pageRatePart, $this->revCount, $revRate, $revRatePart,
                $fetchRate, $fetchRatePart, $etats, $this->maxCount
            ) );
            $this->lastTime = $nowts;
            $this->revCountLast = $this->revCount;
            $this->prefetchCountLast = $this->prefetchCount;
            $this->fetchCountLast = $this->fetchCount;
        }
    }

    function setTimeExceeded() {
        $this->timeExceeded = true;
    }

    function checkIfTimeExceeded() {
        if ( $this->maxTimeAllowed
            && ( $this->lastTime - $this->timeOfCheckpoint > $this->maxTimeAllowed )
        ) {
            return true;
        }

        return false;
    }

    function finalOptionCheck() {
        if ( ( $this->checkpointFiles && !$this->maxTimeAllowed )
            || ( $this->maxTimeAllowed && !$this->checkpointFiles )
        ) {
            throw new MWException( "Options checkpointfile and maxtime must be specified together.\n" );
        }
        foreach ( $this->checkpointFiles as $checkpointFile ) {
            $count = substr_count( $checkpointFile, "%s" );
            if ( $count != 2 ) {
                throw new MWException( "Option checkpointfile must contain two '%s' "
                    . "for substitution of first and last pageids, count is $count instead, "
                    . "file is $checkpointFile.\n" );
            }
        }

        if ( $this->checkpointFiles ) {
            $filenameList = (array)$this->egress->getFilenames();
            if ( count( $filenameList ) != count( $this->checkpointFiles ) ) {
                throw new MWException( "One checkpointfile must be specified "
                    . "for each output option, if maxtime is used.\n" );
            }
        }
    }

    function readDump( $input ) {
        $this->buffer = "";
        $this->openElement = false;
        $this->atStart = true;
        $this->state = "";
        $this->lastName = "";
        $this->thisPage = 0;
        $this->thisRev = 0;

        $parser = xml_parser_create( "UTF-8" );
        xml_parser_set_option( $parser, XML_OPTION_CASE_FOLDING, false );

        xml_set_element_handler(
            $parser,
            array( &$this, 'startElement' ),
            array( &$this, 'endElement' )
        );
        xml_set_character_data_handler( $parser, array( &$this, 'characterData' ) );

        $offset = 0; // for context extraction on error reporting
        $bufferSize = 512 * 1024;
        do {
            if ( $this->checkIfTimeExceeded() ) {
                $this->setTimeExceeded();
            }
            $chunk = fread( $input, $bufferSize );
            if ( !xml_parse( $parser, $chunk, feof( $input ) ) ) {
                wfDebug( "TextDumpPass::readDump encountered XML parsing error\n" );

                $byte = xml_get_current_byte_index( $parser );
                $msg = wfMessage( 'xml-error-string',
                    'XML import parse failure',
                    xml_get_current_line_number( $parser ),
                    xml_get_current_column_number( $parser ),
                    $byte . ( is_null( $chunk ) ? null : ( '; "' . substr( $chunk, $byte - $offset, 16 ) . '"' ) ),
                    xml_error_string( xml_get_error_code( $parser ) ) )->escaped();

                xml_parser_free( $parser );

                throw new MWException( $msg );
            }
            $offset += strlen( $chunk );
        } while ( $chunk !== false && !feof( $input ) );
        if ( $this->maxTimeAllowed ) {
            $filenameList = (array)$this->egress->getFilenames();
            // we wrote some stuff after last checkpoint that needs renamed
            if ( file_exists( $filenameList[0] ) ) {
                $newFilenames = array();
                # we might have just written the header and footer and had no
                # pages or revisions written... perhaps they were all deleted
                # there's no pageID 0 so we use that. the caller is responsible
                # for deciding what to do with a file containing only the
                # siteinfo information and the mw tags.
                if ( !$this->firstPageWritten ) {
                    $firstPageID = str_pad( 0, 9, "0", STR_PAD_LEFT );
                    $lastPageID = str_pad( 0, 9, "0", STR_PAD_LEFT );
                } else {
                    $firstPageID = str_pad( $this->firstPageWritten, 9, "0", STR_PAD_LEFT );
                    $lastPageID = str_pad( $this->lastPageWritten, 9, "0", STR_PAD_LEFT );
                }

                $filenameCount = count( $filenameList );
                for ( $i = 0; $i < $filenameCount; $i++ ) {
                    $checkpointNameFilledIn = sprintf( $this->checkpointFiles[$i], $firstPageID, $lastPageID );
                    $fileinfo = pathinfo( $filenameList[$i] );
                    $newFilenames[] = $fileinfo['dirname'] . '/' . $checkpointNameFilledIn;
                }
                $this->egress->closeAndRename( $newFilenames );
            }
        }
        xml_parser_free( $parser );

        return true;
    }

    function getText( $id ) {
        global $wgContentHandlerUseDB;

        $prefetchNotTried = true; // Whether or not we already tried to get the text via prefetch.
        $text = false; // The candidate for a good text. false if no proper value.
        $failures = 0; // The number of times, this invocation of getText already failed.

        // The number of times getText failed without yielding a good text in between.
        static $consecutiveFailedTextRetrievals = 0;

        $this->fetchCount++;

        // To allow to simply return on success and do not have to worry about book keeping,
        // we assume, this fetch works (possible after some retries). Nevertheless, we koop
        // the old value, so we can restore it, if problems occur (See after the while loop).
        $oldConsecutiveFailedTextRetrievals = $consecutiveFailedTextRetrievals;
        $consecutiveFailedTextRetrievals = 0;

        while ( $failures < $this->maxFailures ) {

            // As soon as we found a good text for the $id, we will return immediately.
            // Hence, if we make it past the try catch block, we know that we did not
            // find a good text.

            try {
                // Step 1: Get some text (or reuse from previous iteratuon if checking
                //         for plausibility failed)

                // Trying to get prefetch, if it has not been tried before
                if ( $text === false && isset( $this->prefetch ) && $prefetchNotTried ) {
                    $prefetchNotTried = false;
                    $tryIsPrefetch = true;
                    $text = $this->prefetch->prefetch( intval( $this->thisPage ),
                        intval( $this->thisRev ) );
                    if ( $text === null ) {
                        $text = false;
                    }
                }

                if ( $text === false ) {
                    // Fallback to asking the database
                    $tryIsPrefetch = false;
                    if ( $this->spawn ) {
                        $text = $this->getTextSpawned( $id );
                    } else {
                        $text = $this->getTextDb( $id );
                    }

                    // No more checks for texts from DB for now.
                    // If we received something that is not false,
                    // We treat it as good text, regardless of whether it actually is or is not
                    if ( $text !== false ) {
                        return $text;
                    }
                }

                if ( $text === false ) {
                    throw new MWException( "Generic error while obtaining text for id " . $id );
                }

                // We received a good candidate for the text of $id via some method

                // Step 2: Checking for plausibility and return the text if it is
                //         plausible
                $revID = intval( $this->thisRev );
                if ( !isset( $this->db ) ) {
                    throw new MWException( "No database available" );
                }

                $revLength = strlen( $text );
                if ( $wgContentHandlerUseDB ) {
                    $row = $this->db->selectRow(
                        'revision',
                        array( 'rev_len', 'rev_content_model' ),
                        array( 'rev_id' => $revID ),
                        __METHOD__
                    );
                    if ( $row ) {
                        // only check the length for the wikitext content handler,
                        // it's a wasted (and failed) check otherwise
                        if ( $row->rev_content_model == CONTENT_MODEL_WIKITEXT ) {
                            $revLength = $row->rev_len;
                        }
                    }
                } else {
                    $revLength = $this->db->selectField( 'revision', 'rev_len', array( 'rev_id' => $revID ) );
                }

                if ( strlen( $text ) == $revLength ) {
                    if ( $tryIsPrefetch ) {
                        $this->prefetchCount++;
                    }

                    return $text;
                }

                $text = false;
                throw new MWException( "Received text is unplausible for id " . $id );
            } catch ( Exception $e ) {
                $msg = "getting/checking text " . $id . " failed (" . $e->getMessage() . ")";
                if ( $failures + 1 < $this->maxFailures ) {
                    $msg .= " (Will retry " . ( $this->maxFailures - $failures - 1 ) . " more times)";
                }
                $this->progress( $msg );
            }

            // Something went wrong; we did not a text that was plausible :(
            $failures++;

            // A failure in a prefetch hit does not warrant resetting db connection etc.
            if ( !$tryIsPrefetch ) {
                // After backing off for some time, we try to reboot the whole process as
                // much as possible to not carry over failures from one part to the other
                // parts
                sleep( $this->failureTimeout );
                try {
                    $this->rotateDb();
                    if ( $this->spawn ) {
                        $this->closeSpawn();
                        $this->openSpawn();
                    }
                } catch ( Exception $e ) {
                    $this->progress( "Rebooting getText infrastructure failed (" . $e->getMessage() . ")" .
                        " Trying to continue anyways" );
                }
            }
        }

        // Retirieving a good text for $id failed (at least) maxFailures times.
        // We abort for this $id.

        // Restoring the consecutive failures, and maybe aborting, if the dump
        // is too broken.
        $consecutiveFailedTextRetrievals = $oldConsecutiveFailedTextRetrievals + 1;
        if ( $consecutiveFailedTextRetrievals > $this->maxConsecutiveFailedTextRetrievals ) {
            throw new MWException( "Graceful storage failure" );
        }

        return "";
    }

    private function getTextDb( $id ) {
        global $wgContLang;
        if ( !isset( $this->db ) ) {
            throw new MWException( __METHOD__ . "No database available" );
        }
        $row = $this->db->selectRow( 'text',
            array( 'old_text', 'old_flags' ),
            array( 'old_id' => $id ),
            __METHOD__ );
        $text = Revision::getRevisionText( $row );
        if ( $text === false ) {
            return false;
        }
        $stripped = str_replace( "\r", "", $text );
        $normalized = $wgContLang->normalize( $stripped );

        return $normalized;
    }

    private function getTextSpawned( $id ) {
        wfSuppressWarnings();
        if ( !$this->spawnProc ) {
            // First time?
            $this->openSpawn();
        }
        $text = $this->getTextSpawnedOnce( $id );
        wfRestoreWarnings();

        return $text;
    }

    function openSpawn() {
        global $IP;

        if ( file_exists( "$IP/../multiversion/MWScript.php" ) ) {
            $cmd = implode( " ",
                array_map( 'wfEscapeShellArg',
                    array(
                        $this->php,
                        "$IP/../multiversion/MWScript.php",
                        "fetchText.php",
                        '--wiki', wfWikiID() ) ) );
        } else {
            $cmd = implode( " ",
                array_map( 'wfEscapeShellArg',
                    array(
                        $this->php,
                        "$IP/maintenance/fetchText.php",
                        '--wiki', wfWikiID() ) ) );
        }
        $spec = array(
            0 => array( "pipe", "r" ),
            1 => array( "pipe", "w" ),
            2 => array( "file", "/dev/null", "a" ) );
        $pipes = array();

        $this->progress( "Spawning database subprocess: $cmd" );
        $this->spawnProc = proc_open( $cmd, $spec, $pipes );
        if ( !$this->spawnProc ) {
            // shit
            $this->progress( "Subprocess spawn failed." );

            return false;
        }
        list(
            $this->spawnWrite, // -> stdin
            $this->spawnRead, // <- stdout
        ) = $pipes;

        return true;
    }

    private function closeSpawn() {
        wfSuppressWarnings();
        if ( $this->spawnRead ) {
            fclose( $this->spawnRead );
        }
        $this->spawnRead = false;
        if ( $this->spawnWrite ) {
            fclose( $this->spawnWrite );
        }
        $this->spawnWrite = false;
        if ( $this->spawnErr ) {
            fclose( $this->spawnErr );
        }
        $this->spawnErr = false;
        if ( $this->spawnProc ) {
            pclose( $this->spawnProc );
        }
        $this->spawnProc = false;
        wfRestoreWarnings();
    }

    private function getTextSpawnedOnce( $id ) {
        global $wgContLang;

        $ok = fwrite( $this->spawnWrite, "$id\n" );
        // $this->progress( ">> $id" );
        if ( !$ok ) {
            return false;
        }

        $ok = fflush( $this->spawnWrite );
        // $this->progress( ">> [flush]" );
        if ( !$ok ) {
            return false;
        }

        // check that the text id they are sending is the one we asked for
        // this avoids out of sync revision text errors we have encountered in the past
        $newId = fgets( $this->spawnRead );
        if ( $newId === false ) {
            return false;
        }
        if ( $id != intval( $newId ) ) {
            return false;
        }

        $len = fgets( $this->spawnRead );
        // $this->progress( "<< " . trim( $len ) );
        if ( $len === false ) {
            return false;
        }

        $nbytes = intval( $len );
        // actual error, not zero-length text
        if ( $nbytes < 0 ) {
            return false;
        }

        $text = "";

        // Subprocess may not send everything at once, we have to loop.
        while ( $nbytes > strlen( $text ) ) {
            $buffer = fread( $this->spawnRead, $nbytes - strlen( $text ) );
            if ( $buffer === false ) {
                break;
            }
            $text .= $buffer;
        }

        $gotbytes = strlen( $text );
        if ( $gotbytes != $nbytes ) {
            $this->progress( "Expected $nbytes bytes from database subprocess, got $gotbytes " );

            return false;
        }

        // Do normalization in the dump thread...
        $stripped = str_replace( "\r", "", $text );
        $normalized = $wgContLang->normalize( $stripped );

        return $normalized;
    }

    function startElement( $parser, $name, $attribs ) {
        $this->checkpointJustWritten = false;

        $this->clearOpenElement( null );
        $this->lastName = $name;

        if ( $name == 'revision' ) {
            $this->state = $name;
            $this->egress->writeOpenPage( null, $this->buffer );
            $this->buffer = "";
        } elseif ( $name == 'page' ) {
            $this->state = $name;
            if ( $this->atStart ) {
                $this->egress->writeOpenStream( $this->buffer );
                $this->buffer = "";
                $this->atStart = false;
            }
        }

        if ( $name == "text" && isset( $attribs['id'] ) ) {
            $text = $this->getText( $attribs['id'] );
            $this->openElement = array( $name, array( 'xml:space' => 'preserve' ) );
            if ( strlen( $text ) > 0 ) {
                $this->characterData( $parser, $text );
            }
        } else {
            $this->openElement = array( $name, $attribs );
        }
    }

    function endElement( $parser, $name ) {
        $this->checkpointJustWritten = false;

        if ( $this->openElement ) {
            $this->clearOpenElement( "" );
        } else {
            $this->buffer .= "</$name>";
        }

        if ( $name == 'revision' ) {
            $this->egress->writeRevision( null, $this->buffer );
            $this->buffer = "";
            $this->thisRev = "";
        } elseif ( $name == 'page' ) {
            if ( !$this->firstPageWritten ) {
                $this->firstPageWritten = trim( $this->thisPage );
            }
            $this->lastPageWritten = trim( $this->thisPage );
            if ( $this->timeExceeded ) {
                $this->egress->writeClosePage( $this->buffer );
                // nasty hack, we can't just write the chardata after the
                // page tag, it will include leading blanks from the next line
                $this->egress->sink->write( "\n" );

                $this->buffer = $this->xmlwriterobj->closeStream();
                $this->egress->writeCloseStream( $this->buffer );

                $this->buffer = "";
                $this->thisPage = "";
                // this could be more than one file if we had more than one output arg

                $filenameList = (array)$this->egress->getFilenames();
                $newFilenames = array();
                $firstPageID = str_pad( $this->firstPageWritten, 9, "0", STR_PAD_LEFT );
                $lastPageID = str_pad( $this->lastPageWritten, 9, "0", STR_PAD_LEFT );
                $filenamesCount = count( $filenameList );
                for ( $i = 0; $i < $filenamesCount; $i++ ) {
                    $checkpointNameFilledIn = sprintf( $this->checkpointFiles[$i], $firstPageID, $lastPageID );
                    $fileinfo = pathinfo( $filenameList[$i] );
                    $newFilenames[] = $fileinfo['dirname'] . '/' . $checkpointNameFilledIn;
                }
                $this->egress->closeRenameAndReopen( $newFilenames );
                $this->buffer = $this->xmlwriterobj->openStream();
                $this->timeExceeded = false;
                $this->timeOfCheckpoint = $this->lastTime;
                $this->firstPageWritten = false;
                $this->checkpointJustWritten = true;
            } else {
                $this->egress->writeClosePage( $this->buffer );
                $this->buffer = "";
                $this->thisPage = "";
            }
        } elseif ( $name == 'mediawiki' ) {
            $this->egress->writeCloseStream( $this->buffer );
            $this->buffer = "";
        }
    }

    function characterData( $parser, $data ) {
        $this->clearOpenElement( null );
        if ( $this->lastName == "id" ) {
            if ( $this->state == "revision" ) {
                $this->thisRev .= $data;
            } elseif ( $this->state == "page" ) {
                $this->thisPage .= $data;
            }
        }
        // have to skip the newline left over from closepagetag line of
        // end of checkpoint files. nasty hack!!
        if ( $this->checkpointJustWritten ) {
            if ( $data[0] == "\n" ) {
                $data = substr( $data, 1 );
            }
            $this->checkpointJustWritten = false;
        }
        $this->buffer .= htmlspecialchars( $data );
    }

    function clearOpenElement( $style ) {
        if ( $this->openElement ) {
            $this->buffer .= Xml::element( $this->openElement[0], $this->openElement[1], $style );
            $this->openElement = false;
        }
    }
}