DjVuImage.php

Go to the documentation of this file.

<?php
use MediaWiki\MainConfigNames;
use MediaWiki\MediaWikiServices;
use MediaWiki\Shell\Shell;
use Wikimedia\AtEase\AtEase;
 
class DjVuImage {
 
    private const DJVUTXT_MEMORY_LIMIT = 300000000;
 
    private $mFilename;
 
    public function __construct( $filename ) {
        $this->mFilename = $filename;
    }
 
    public function isValid() {
        $info = $this->getInfo();
 
        return $info !== false;
    }
 
    public function getImageSize() {
        $data = $this->getInfo();
 
        if ( $data !== false ) {
            return [
                'width' => $data['width'],
                'height' => $data['height']
            ];
        } else {
            return [];
        }
    }
 
    // ---------
 
    public function dump() {
        $file = fopen( $this->mFilename, 'rb' );
        $header = fread( $file, 12 );
        $arr = unpack( 'a4magic/a4chunk/NchunkLength', $header );
        $chunk = $arr['chunk'];
        $chunkLength = $arr['chunkLength'];
        echo "$chunk $chunkLength\n";
        $this->dumpForm( $file, $chunkLength, 1 );
        fclose( $file );
    }
 
    private function dumpForm( $file, $length, $indent ) {
        $start = ftell( $file );
        $secondary = fread( $file, 4 );
        echo str_repeat( ' ', $indent * 4 ) . "($secondary)\n";
        while ( ftell( $file ) - $start < $length ) {
            $chunkHeader = fread( $file, 8 );
            if ( $chunkHeader == '' ) {
                break;
            }
            $arr = unpack( 'a4chunk/NchunkLength', $chunkHeader );
            $chunk = $arr['chunk'];
            $chunkLength = $arr['chunkLength'];
            echo str_repeat( ' ', $indent * 4 ) . "$chunk $chunkLength\n";
 
            if ( $chunk == 'FORM' ) {
                $this->dumpForm( $file, $chunkLength, $indent + 1 );
            } else {
                fseek( $file, $chunkLength, SEEK_CUR );
                if ( $chunkLength & 1 ) {
                    // Padding byte between chunks
                    fseek( $file, 1, SEEK_CUR );
                }
            }
        }
    }
 
    private function getInfo() {
        AtEase::suppressWarnings();
        $file = fopen( $this->mFilename, 'rb' );
        AtEase::restoreWarnings();
        if ( $file === false ) {
            wfDebug( __METHOD__ . ": missing or failed file read" );
 
            return false;
        }
 
        $header = fread( $file, 16 );
        $info = false;
 
        if ( strlen( $header ) < 16 ) {
            wfDebug( __METHOD__ . ": too short file header" );
        } else {
            $arr = unpack( 'a4magic/a4form/NformLength/a4subtype', $header );
 
            $subtype = $arr['subtype'];
            if ( $arr['magic'] != 'AT&T' ) {
                wfDebug( __METHOD__ . ": not a DjVu file" );
            } elseif ( $subtype == 'DJVU' ) {
                // Single-page document
                $info = $this->getPageInfo( $file );
            } elseif ( $subtype == 'DJVM' ) {
                // Multi-page document
                $info = $this->getMultiPageInfo( $file, $arr['formLength'] );
            } else {
                wfDebug( __METHOD__ . ": unrecognized DJVU file type '{$arr['subtype']}'" );
            }
        }
        fclose( $file );
 
        return $info;
    }
 
    private function readChunk( $file ) {
        $header = fread( $file, 8 );
        if ( strlen( $header ) < 8 ) {
            return [ false, 0 ];
        } else {
            $arr = unpack( 'a4chunk/Nlength', $header );
 
            return [ $arr['chunk'], $arr['length'] ];
        }
    }
 
    private function skipChunk( $file, $chunkLength ) {
        fseek( $file, $chunkLength, SEEK_CUR );
 
        if ( ( $chunkLength & 1 ) && !feof( $file ) ) {
            // padding byte
            fseek( $file, 1, SEEK_CUR );
        }
    }
 
    private function getMultiPageInfo( $file, $formLength ) {
        // For now, we'll just look for the first page in the file
        // and report its information, hoping others are the same size.
        $start = ftell( $file );
        do {
            list( $chunk, $length ) = $this->readChunk( $file );
            if ( !$chunk ) {
                break;
            }
 
            if ( $chunk == 'FORM' ) {
                $subtype = fread( $file, 4 );
                if ( $subtype == 'DJVU' ) {
                    wfDebug( __METHOD__ . ": found first subpage" );
 
                    return $this->getPageInfo( $file );
                }
                $this->skipChunk( $file, $length - 4 );
            } else {
                wfDebug( __METHOD__ . ": skipping '$chunk' chunk" );
                $this->skipChunk( $file, $length );
            }
        } while ( $length != 0 && !feof( $file ) && ftell( $file ) - $start < $formLength );
 
        wfDebug( __METHOD__ . ": multi-page DJVU file contained no pages" );
 
        return false;
    }
 
    private function getPageInfo( $file ) {
        list( $chunk, $length ) = $this->readChunk( $file );
        if ( $chunk != 'INFO' ) {
            wfDebug( __METHOD__ . ": expected INFO chunk, got '$chunk'" );
 
            return false;
        }
 
        if ( $length < 9 ) {
            wfDebug( __METHOD__ . ": INFO should be 9 or 10 bytes, found $length" );
 
            return false;
        }
        $data = fread( $file, $length );
        if ( strlen( $data ) < $length ) {
            wfDebug( __METHOD__ . ": INFO chunk cut off" );
 
            return false;
        }
 
        $arr = unpack(
            'nwidth/' .
            'nheight/' .
            'Cminor/' .
            'Cmajor/' .
            'vresolution/' .
            'Cgamma', $data );
 
        # Newer files have rotation info in byte 10, but we don't use it yet.
 
        return [
            'width' => $arr['width'],
            'height' => $arr['height'],
            'version' => "{$arr['major']}.{$arr['minor']}",
            'resolution' => $arr['resolution'],
            'gamma' => $arr['gamma'] / 10.0 ];
    }
 
    public function retrieveMetaData() {
        $djvuDump = MediaWikiServices::getInstance()->getMainConfig()->get( MainConfigNames::DjvuDump );
        $djvuTxt = MediaWikiServices::getInstance()->getMainConfig()->get( MainConfigNames::DjvuTxt );
        if ( !$this->isValid() ) {
            return false;
        }
 
        if ( isset( $djvuDump ) ) {
            # djvudump is faster than djvutoxml (now abandoned) as of version 3.5
            # https://sourceforge.net/p/djvu/bugs/71/
            $cmd = Shell::escape( $djvuDump ) . ' ' . Shell::escape( $this->mFilename );
            $dump = wfShellExec( $cmd );
            $json = [ 'data' => $this->convertDumpToJSON( $dump ) ];
        } else {
            $json = null;
        }
        # Text layer
        if ( isset( $djvuTxt ) ) {
            $cmd = Shell::escape( $djvuTxt ) . ' --detail=page ' . Shell::escape( $this->mFilename );
            wfDebug( __METHOD__ . ": $cmd" );
            $retval = '';
            $txt = wfShellExec( $cmd, $retval, [], [ 'memory' => self::DJVUTXT_MEMORY_LIMIT ] );
            $json['text'] = [];
            if ( $retval == 0 ) {
                # Strip some control characters
                # Ignore carriage returns
                $txt = preg_replace( "/\\\\013/", "", $txt );
                # Replace runs of OCR region separators with a single extra line break
                $txt = preg_replace( "/(?:\\\\(035|037))+/", "\n", $txt );
 
                $reg = <<<EOR
                    /\‍(page\s[\d-]*\s[\d-]*\s[\d-]*\s[\d-]*\s*"
                    ((?>    # Text to match is composed of atoms of either:
                        \\\\. # - any escaped character
                        |     # - any character different from " and \
                        [^"\\\\]+
                    )*?)
                    "\s*\‍)
                    | # Or page can be empty ; in this case, djvutxt dumps ()
                    \‍(\s*()\)/sx
EOR;
                $matches = [];
                preg_match_all( $reg, $txt, $matches );
                $json['text'] = array_map( [ $this, 'pageTextCallback' ], $matches[1] );
            }
        }
 
        return $json;
    }
 
    private function pageTextCallback( string $match ) {
        # Get rid of invalid UTF-8
        $val = UtfNormal\Validator::cleanUp( stripcslashes( $match ) );
        $val = str_replace( '�', '', $val );
        return $val;
    }
 
    private function convertDumpToJSON( $dump ) {
        if ( strval( $dump ) == '' ) {
            return false;
        }
 
        $dump = str_replace( "\r", '', $dump );
        $line = strtok( $dump, "\n" );
        $m = false;
        $good = false;
        $result = [];
        if ( preg_match( '/^( *)FORM:DJVU/', $line, $m ) ) {
            # Single-page
            $parsed = $this->parseFormDjvu( $line );
            if ( $parsed ) {
                $good = true;
            } else {
                return false;
            }
            $result['pages'] = [ $parsed ];
        } elseif ( preg_match( '/^( *)FORM:DJVM/', $line, $m ) ) {
            # Multi-page
            $parentLevel = strlen( $m[1] );
            # Find DIRM
            $line = strtok( "\n" );
            $result['pages'] = [];
            while ( $line !== false ) {
                $childLevel = strspn( $line, ' ' );
                if ( $childLevel <= $parentLevel ) {
                    # End of chunk
                    break;
                }
 
                if ( preg_match( '/^ *DIRM.*indirect/', $line ) ) {
                    wfDebug( "Indirect multi-page DjVu document, bad for server!" );
 
                    return false;
                }
 
                if ( preg_match( '/^ *FORM:DJVU/', $line ) ) {
                    # Found page
                    $parsed = $this->parseFormDjvu( $line );
                    if ( $parsed ) {
                        $good = true;
                    } else {
                        return false;
                    }
                    $result['pages'][] = $parsed;
                }
                $line = strtok( "\n" );
            }
        }
        if ( !$good ) {
            return false;
        }
 
        return $result;
    }
 
    private function parseFormDjvu( $line ) {
        $parentLevel = strspn( $line, ' ' );
        $line = strtok( "\n" );
        # Find INFO
        while ( $line !== false ) {
            $childLevel = strspn( $line, ' ' );
            if ( $childLevel <= $parentLevel ) {
                # End of chunk
                break;
            }
 
            if ( preg_match(
                '/^ *INFO *\[\d*\] *DjVu *(\d+)x(\d+), *\w*, *(\d+) *dpi, *gamma=([0-9.-]+)/',
                $line,
                $m
            ) ) {
                return [
                    'height' => (int)$m[2],
                    'width' => (int)$m[1],
                    'dpi' => (float)$m[3],
                    'gamma' => (float)$m[4],
                ];
            }
            $line = strtok( "\n" );
        }
 
        # Not found
        return false;
    }
}