PDF to Markdown

pdf.script.php

<?php

use com\adobe\pdf\{PdfReader, Content, Ref, Tokens, CharacterMap};
use io\streams\FileInputStream;
use lang\FormatException;
use util\Objects;
use util\cmd\Console;

try {
  $reader= new PdfReader(new FileInputStream($argv[1]));
} catch (Throwable $e) {
  Console::writeLine($argv[1], ': ', $e);
  return 2;
}

$objects= [];
$trailer= [];
Console::$err->write('Parsing ', $argv[1], ' [');
foreach ($reader->objects() as $kind => $value) {
  Console::$err->write('.');
  if ('object' === $kind) {
    $objects[$value['id']->hashCode()]= $value['dict'];
  } else if ('trailer' === $kind) {
    $trailer+= $value;
  }
}
Console::$err->writeLine(']');

// Handle files without `trailer` entry, e.g. linearized PDFs
if (empty($trailer)) {
  foreach ($objects as $object) {
    if ('XRef' === ($object['Type'] ?? null)) {
      $trailer= $object;
      break;
    }
  }
}

// Handle encryption
Console::writeLine('Trailer: ', $trailer);
if (isset($trailer['Encrypt'])) {
  $encrypt= $objects[$trailer['Encrypt']->hashCode()];
  Console::writeLine('Cannot handle encrypted PDFs: ', $encrypt);
  return 1;
}

// Now, unpack object streams and add them to the lookup table
foreach ($objects as $id => $object) {
  if ('ObjStm' === ($object['Type'] ?? null)) {
    $n= $object['N'];
    $tokens= new Tokens($object['$stream']);

    $numbers= [];
    for ($i= 0; $i < $n; $i++) {
      $number= $tokens->token()[1];
      $offset= $tokens->token()[1];
      $numbers[$i]= $number.'_0';
    }

    for ($i= 0; $i < $n; $i++) {
      $objects[$numbers[$i]]= $tokens->value();
    }
  }
}

if ($info= ($trailer['Info'] ?? null)) {
  Console::writeLine('Info: ', $objects[$info->hashCode()]);
}

$root= $objects[$trailer['Root']->hashCode()];
Console::writeLine('Root: ', $root);

$pages= $objects[$root['Pages']->hashCode()];
Console::writeLine('Pages: ', $pages);

$objectOf= function($arg) use(&$objects) {
  return $arg instanceof Ref ? $objects[$arg->hashCode()] : $arg;
};

$contentsOf= function($kids, $suffix= '') use(&$contentsOf, &$objectOf, &$objects) {
  foreach ($kids as $i => $ref) {
    $page= $objects[$ref->hashCode()];
    $number= $suffix.($i + 1);

    if ('Pages' === $page['Type']) {
      yield from $contentsOf($page['Kids'], $number.'.');
      continue;
    }

    // A single content object or an array of references
    $contents= $objectOf($page['Contents']);
    if ($stream= $contents['$stream'] ?? null) {
      $streams= [$stream];
    } else {
      $streams= [];
      foreach ($contents as $ref) {
        $streams[]= $objects[$ref->hashCode()]['$stream'];
      }
    }
    yield $number => $page + ['$contents' => $streams];
  }
};

$stringOf= function($bytes, $encoding, $unicode= null) use(&$stringOf) {
  switch ($encoding) {
    case null: return $bytes;
    case 'MacRomanEncoding': return iconv('macintosh', \xp::ENCODING, $bytes);
    case 'WinAnsiEncoding': return iconv('cp1252', \xp::ENCODING, $bytes);
    case 'Identity-H': return $unicode->translate($bytes);
    default:

      // FIXME: Respect differences
      return $stringOf($bytes, $encoding['BaseEncoding'] ?? null);
  }
};

$textOf= function($token, array $format) use(&$objects, &$objectOf, &$stringOf) {
  if ($descendants= $format['font']['DescendantFonts'] ?? null) {
    // TODO https://github.com/adobe-type-tools/cmap-resources
  }

  $bytes= 'hex' === $token[0] ? hex2bin($token[1]) : $token[1];
  switch ($format['font']['Subtype']) {
    case 'Type0': case 'Type1': case 'TrueType':
      return $stringOf(
        $bytes,
        $objectOf($format['font']['Encoding'] ?? null),
        $format['font']['$unicode'] ?? null
      );

    case 'Type3':
      if ($map= $format['font']['$unicode'] ?? null) {
        return $format['font']['$unicode']->translate($bytes);
      }
      throw new FormatException('Cannot handle type 3 fonts without unicode');

    default:
      throw new FormatException('Unknown font type '.$format['font']['Subtype']);
  }
};

foreach ($contentsOf($pages['Kids']) as $number => $page) {
  Console::writeLine('- Page #', $number, ' contents: ', Objects::stringOf($page['Contents'], '  '));

  // Resolve fonts
  $fonts= [];
  if ($resources= $objectOf($page['Resources'])['Font'] ?? null) {
    foreach (is_array($resources) ? $resources : $objectOf($resources) as $id => $ref) {
      $font= $objects[$ref->hashCode()];
      // Console::writeLine('  Font <', $id, '> := ', Objects::stringOf($font, '  '));

      if ($ref= $font['ToUnicode'] ?? null) {
        $font['$unicode']= new CharacterMap($objects[$ref->hashCode()]['$stream']);
      }
      $fonts[$id]= $font;
    }
  }

  // Select no font for the beginning
  $font= null;

  // Resolve objects
  $xobjects= [];
  if ($resources= $objectOf($page['Resources'])['XObject'] ?? null) {
    foreach (is_array($resources) ? $resources : $objectOf($resources) as $id => $ref) {
      $xobjects[$id]= $objects[$ref->hashCode()];
    }
  }

  // DEBUG
  // foreach ($page['$contents'] as $i => $stream) {
  //   file_put_contents('text-block.'.$number.'_'.$i, $stream->bytes());
  // }

  $paragraph= '';
  $content= new Content(...$page['$contents']);
  foreach ($content->operations() as $op => $arguments) {

    // DEBUG
    // Console::$err->writeLine("\e[2m", '[', $number, ' | ', $op, ']: ', $arguments, "\e[0m");

    if ('BT' === $op) {
      // NOOP
    } else if ('ET' === $op) {
      Console::writeLine('  ', str_replace("\n", "\n  ", $paragraph));
      Console::writeLine();
      $paragraph= '';
    } else if ('Tf' === $op) {
      $id= $arguments[0][1];
      $size= $arguments[1][1];
      $font= ['font' => $fonts[$id] ?? '@'.$id, 'size' => $size];
    } else if ('Tj' === $op) {
      $paragraph.= $textOf($arguments[0], $font);
    } else if ('TJ' === $op) {
      foreach ($arguments as $argument) {
        if ('hex' === $argument[0] || 'string' === $argument[0]) {
          $paragraph.= $textOf($argument, $font);
        }
      }
    } else if ("'" === $op) {
      $paragraph.= "\n".$textOf($arguments[0], $font);;
    } else if ('"' === $op) {
      $paragraph.= "\n".$textOf($arguments[2], $font);;
    } else if ('TD' === $op) {
      $paragraph.= "\n";
    } else if ('Do' === $op) {
      $id= $arguments[0][1];
      $xobject= $xobjects[$id];
      if ('Image' === $xobject['Subtype']) {
        Console::writeLine('  ', "![{$id}]({$xobject['Subtype']}:{$xobject['Width']}x{$xobject['Height']})\n");
      }
    }
  }
}

Uses https://github.com/xp-forge/pdf-parser