public long CheckXRefOffset(long startXRefOffset, bool isLenientParsing)
{
// repair mode isn't available in non-lenient mode
if (!isLenientParsing)
{
return(startXRefOffset);
}
source.Seek(startXRefOffset);
ReadHelper.SkipSpaces(source);
if (source.Peek() == 'x' && ReadHelper.IsString(source, "xref"))
{
return(startXRefOffset);
}
if (startXRefOffset > 0)
{
if (CheckXRefStreamOffset(source, startXRefOffset, true, pool))
{
return(startXRefOffset);
}
return(CalculateXRefFixedOffset(startXRefOffset));
}
// can't find a valid offset
return(-1);
}
public static string ReadLine(IRandomAccessRead reader)
{
if (reader == null)
{
throw new ArgumentNullException(nameof(reader));
}
if (reader.IsEof())
{
throw new InvalidOperationException("Error: End-of-File, expected line");
}
var buffer = new StringBuilder(11);
int c;
while ((c = reader.Read()) != -1)
{
// CR and LF are valid EOLs
if (IsEndOfLine(c))
{
break;
}
buffer.Append((char)c);
}
// CR+LF is also a valid EOL
if (IsCarriageReturn(c) && IsLineFeed(reader.Peek()))
{
reader.Read();
}
return(buffer.ToString());
}
public PdfDictionary Parse(IRandomAccessRead reader, IBaseParser baseParser, CosObjectPool pool)
{
if (reader == null)
{
throw new ArgumentNullException(nameof(reader));
}
if (baseParser == null)
{
throw new ArgumentNullException(nameof(baseParser));
}
if (pool == null)
{
throw new ArgumentNullException(nameof(pool));
}
ReadHelper.ReadExpectedChar(reader, '<');
ReadHelper.ReadExpectedChar(reader, '<');
ReadHelper.SkipSpaces(reader);
var dictionary = new PdfDictionary();
var done = false;
while (!done)
{
ReadHelper.SkipSpaces(reader);
var c = (char)reader.Peek();
switch (c)
{
case '>':
done = true;
break;
case '/':
var nameValue = ParseCosDictionaryNameValuePair(reader, baseParser, pool);
if (nameValue.key != null && nameValue.value != null)
{
dictionary.Set(nameValue.key, nameValue.value);
}
break;
default:
if (ReadUntilEnd(reader))
{
return(new PdfDictionary());
}
break;
}
}
ReadHelper.ReadExpectedString(reader, ">>");
return(dictionary);
}
private (CosName key, CosBase value) ParseCosDictionaryNameValuePair(IRandomAccessRead reader, IBaseParser baseParser, CosObjectPool pool)
{
var key = nameParser.Parse(reader);
var value = ParseValue(reader, baseParser, pool);
ReadHelper.SkipSpaces(reader);
if ((char)reader.Peek() == 'd')
{
// if the next string is 'def' then we are parsing a cmap stream
// and want to ignore it, otherwise throw an exception.
var potentialDef = ReadHelper.ReadString(reader);
if (!potentialDef.Equals("def"))
{
reader.Unread(OtherEncodings.StringAsLatin1Bytes(potentialDef));
}
else
{
ReadHelper.SkipSpaces(reader);
}
}
if (value == null)
{
log?.Warn("Bad Dictionary Declaration " + ReadHelper.ReadString(reader));
return(null, null);
}
// label this item as direct, to avoid signature problems.
value.Direct = true;
return(key, value);
}
public int Peek()
{
if (Throw)
{
throw new InvalidOperationException();
}
return(reader.Peek());
}
public int peek()
{
return(reader.Peek());
}
public bool TryParse(IRandomAccessRead source, long offset, bool isLenientParsing, CosObjectPool pool, out CrossReferenceTablePartBuilder builder)
{
builder = null;
var tableStartOffset = source.GetPosition();
if (source.Peek() != 'x')
{
return(false);
}
var xref = ReadHelper.ReadString(source);
if (!xref.Trim().Equals("xref"))
{
return(false);
}
// check for trailer after xref
var str = ReadHelper.ReadString(source);
byte[] b = OtherEncodings.StringAsLatin1Bytes(str);
source.Rewind(b.Length);
if (str.StartsWith("trailer"))
{
log.Warn("skipping empty xref table");
return(false);
}
builder = new CrossReferenceTablePartBuilder
{
Offset = offset,
XRefType = CrossReferenceType.Table
};
// Tables can have multiple sections. Each starts with a starting object id and a count.
while (true)
{
if (!TableSubsectionDefinition.TryRead(log, source, out var subsectionDefinition))
{
log.Warn($"Unexpected subsection definition in the cross-reference table at offset {offset}");
if (isLenientParsing)
{
break;
}
return(false);
}
var currentObjectId = subsectionDefinition.FirstNumber;
ReadHelper.SkipSpaces(source);
for (var i = 0; i < subsectionDefinition.Count; i++)
{
if (source.IsEof() || ReadHelper.IsEndOfName((char)source.Peek()))
{
break;
}
if (source.Peek() == 't')
{
break;
}
//Ignore table contents
var currentLine = ReadHelper.ReadLine(source);
var splitString = currentLine.Split(new[] { ' ' }, StringSplitOptions.RemoveEmptyEntries);
if (splitString.Length < 3)
{
log.Warn("invalid xref line: " + currentLine);
break;
}
// This supports the corrupt table as reported in PDFBOX-474 (XXXX XXX XX n)
if (splitString[splitString.Length - 1].Equals(InUseEntry))
{
try
{
var objectOffset = long.Parse(splitString[0]);
if (objectOffset >= tableStartOffset && objectOffset <= source.GetPosition())
{
// PDFBOX-3923: offset points inside this table - that can't be good
throw new InvalidOperationException(
$"Object offset {objectOffset} is within its own cross-reference table for object {currentObjectId}");
}
var generation = int.Parse(splitString[1]);
builder.Add(currentObjectId, generation, objectOffset);
}
catch (FormatException e)
{
throw new InvalidOperationException("Bad", e);
}
}
else if (!splitString[2].Equals(FreeEntry))
{
throw new InvalidOperationException(
$"Corrupt cross-reference table entry for object {currentObjectId}. The indicator was not 'n' or 'f' but {splitString[2]}.");
}
currentObjectId++;
ReadHelper.SkipSpaces(source);
}
ReadHelper.SkipSpaces(source);
if (!ReadHelper.IsDigit(source))
{
break;
}
}
if (!TryParseTrailer(source, isLenientParsing, pool, out var trailer))
{
throw new InvalidOperationException($"Something went wrong trying to read the XREF table at {offset}.");
}
builder.Dictionary = trailer;
builder.Previous = trailer.GetLongOrDefault(CosName.PREV);
return(true);
}
private bool TryParseTrailer(IRandomAccessRead source, bool isLenientParsing, CosObjectPool pool, out PdfDictionary trailer)
{
trailer = null;
// parse the last trailer.
var trailerOffset = source.GetPosition();
// PDFBOX-1739 skip extra xref entries in RegisSTAR documents
if (isLenientParsing)
{
int nextCharacter = source.Peek();
while (nextCharacter != 't' && ReadHelper.IsDigit(nextCharacter))
{
if (source.GetPosition() == trailerOffset)
{
// warn only the first time
//LOG.warn("Expected trailer object at position " + trailerOffset
// + ", keep trying");
}
ReadHelper.ReadLine(source);
nextCharacter = source.Peek();
}
}
if (source.Peek() != 't')
{
return(false);
}
//read "trailer"
long currentOffset = source.GetPosition();
string nextLine = ReadHelper.ReadLine(source);
if (!nextLine.Trim().Equals("trailer"))
{
// in some cases the EOL is missing and the trailer immediately
// continues with "<<" or with a blank character
// even if this does not comply with PDF reference we want to support as many PDFs as possible
// Acrobat reader can also deal with this.
if (nextLine.StartsWith("trailer"))
{
// we can't just unread a portion of the read data as we don't know if the EOL consist of 1 or 2 bytes
int len = "trailer".Length;
// jump back right after "trailer"
source.Seek(currentOffset + len);
}
else
{
return(false);
}
}
// in some cases the EOL is missing and the trailer continues with " <<"
// even if this does not comply with PDF reference we want to support as many PDFs as possible
// Acrobat reader can also deal with this.
ReadHelper.SkipSpaces(source);
PdfDictionary parsedTrailer = dictionaryParser.Parse(source, baseParser, pool);
trailer = parsedTrailer;
ReadHelper.SkipSpaces(source);
return(true);
}
public IReadOnlyDictionary <CosObjectKey, long> GetObjectLocations()
{
if (objectLocations != null)
{
return(objectLocations);
}
var lastEndOfFile = GetLastEndOfFileMarker();
var results = new Dictionary <CosObjectKey, long>();
var originPosition = reader.GetPosition();
long currentOffset = MinimumSearchOffset;
long lastObjectId = long.MinValue;
int lastGenerationId = int.MinValue;
long lastObjOffset = long.MinValue;
byte[] objString = OtherEncodings.StringAsLatin1Bytes(" obj");
byte[] endobjString = OtherEncodings.StringAsLatin1Bytes("endobj");
bool endobjFound = false;
do
{
reader.Seek(currentOffset);
if (ReadHelper.IsString(reader, objString))
{
long tempOffset = currentOffset - 1;
reader.Seek(tempOffset);
int generationId = reader.Peek();
// is the next char a digit?
if (ReadHelper.IsDigit(generationId))
{
generationId -= 48;
tempOffset--;
reader.Seek(tempOffset);
if (ReadHelper.IsSpace(reader))
{
while (tempOffset > MinimumSearchOffset && ReadHelper.IsSpace(reader))
{
reader.Seek(--tempOffset);
}
bool objectIdFound = false;
while (tempOffset > MinimumSearchOffset && ReadHelper.IsDigit(reader))
{
reader.Seek(--tempOffset);
objectIdFound = true;
}
if (objectIdFound)
{
reader.Read();
long objectId = ObjectHelper.ReadObjectNumber(reader);
if (lastObjOffset > 0)
{
// add the former object ID only if there was a subsequent object ID
results[new CosObjectKey(lastObjectId, lastGenerationId)] = lastObjOffset;
}
lastObjectId = objectId;
lastGenerationId = generationId;
lastObjOffset = tempOffset + 1;
currentOffset += objString.Length - 1;
endobjFound = false;
}
}
}
}
else if (ReadHelper.IsString(reader, "endobj"))
{
endobjFound = true;
currentOffset += endobjString.Length - 1;
}
currentOffset++;
} while (currentOffset < lastEndOfFile && !reader.IsEof());
if ((lastEndOfFile < long.MaxValue || endobjFound) && lastObjOffset > 0)
{
// if the pdf wasn't cut off in the middle or if the last object ends with a "endobj" marker
// the last object id has to be added here so that it can't get lost as there isn't any subsequent object id
results[new CosObjectKey(lastObjectId, lastGenerationId)] = lastObjOffset;
}
// reestablish origin position
reader.Seek(originPosition);
objectLocations = results;
return(objectLocations);
}
public static bool IsSpace(IRandomAccessRead reader)
{
return(IsSpace(reader.Peek()));
}
public static bool IsDigit(IRandomAccessRead reader)
{
return(IsDigit(reader.Peek()));
}
public COSArray Parse(IRandomAccessRead reader, CosBaseParser baseParser, CosObjectPool pool)
{
ReadHelper.ReadExpectedChar(reader, '[');
var po = new COSArray();
CosBase pbo;
ReadHelper.SkipSpaces(reader);
int i;
while (((i = reader.Peek()) > 0) && ((char)i != ']'))
{
pbo = baseParser.Parse(reader, pool);
if (pbo is CosObject)
{
// We have to check if the expected values are there or not PDFBOX-385
if (po.get(po.size() - 1) is CosInt)
{
var genNumber = (CosInt)po.remove(po.size() - 1);
if (po.get(po.size() - 1) is CosInt)
{
var number = (CosInt)po.remove(po.size() - 1);
CosObjectKey key = new CosObjectKey(number.AsLong(), genNumber.AsInt());
pbo = pool.Get(key);
}
else
{
// the object reference is somehow wrong
pbo = null;
}
}
else
{
pbo = null;
}
}
if (pbo != null)
{
po.add(pbo);
}
else
{
//it could be a bad object in the array which is just skipped
// LOG.warn("Corrupt object reference at offset " + seqSource.getPosition());
// This could also be an "endobj" or "endstream" which means we can assume that
// the array has ended.
string isThisTheEnd = ReadHelper.ReadString(reader);
reader.Unread(OtherEncodings.StringAsLatin1Bytes(isThisTheEnd));
if (string.Equals(isThisTheEnd, "endobj") || string.Equals(isThisTheEnd, "endstream"))
{
return(po);
}
}
ReadHelper.SkipSpaces(reader);
}
// read ']'
reader.Read();
ReadHelper.SkipSpaces(reader);
return(po);
}
private void bfSearchForObjects(IRandomAccessRead source)
{
bfSearchForLastEOFMarker(source);
bfSearchCOSObjectKeyOffsets = new Dictionary <CosObjectKey, long>();
long originOffset = source.GetPosition();
long currentOffset = MINIMUM_SEARCH_OFFSET;
long lastObjectId = long.MinValue;
int lastGenID = int.MinValue;
long lastObjOffset = long.MinValue;
char[] objString = " obj".ToCharArray();
char[] endobjString = "endobj".ToCharArray();
bool endobjFound = false;
do
{
source.Seek(currentOffset);
if (ReadHelper.IsString(source, "obj"))
{
long tempOffset = currentOffset - 1;
source.Seek(tempOffset);
int genID = source.Peek();
// is the next char a digit?
if (ReadHelper.IsDigit(genID))
{
genID -= 48;
tempOffset--;
source.Seek(tempOffset);
if (ReadHelper.IsSpace(source))
{
while (tempOffset > MINIMUM_SEARCH_OFFSET && ReadHelper.IsSpace(source))
{
source.Seek(--tempOffset);
}
bool objectIDFound = false;
while (tempOffset > MINIMUM_SEARCH_OFFSET && ReadHelper.IsDigit(source))
{
source.Seek(--tempOffset);
objectIDFound = true;
}
if (objectIDFound)
{
source.Read();
long objectId = ObjectHelper.ReadObjectNumber(source);
if (lastObjOffset > 0)
{
// add the former object ID only if there was a subsequent object ID
bfSearchCOSObjectKeyOffsets[new CosObjectKey(lastObjectId, lastGenID)] = lastObjOffset;
}
lastObjectId = objectId;
lastGenID = genID;
lastObjOffset = tempOffset + 1;
currentOffset += objString.Length - 1;
endobjFound = false;
}
}
}
}
else if (ReadHelper.IsString(source, "endobj"))
{
endobjFound = true;
currentOffset += endobjString.Length - 1;
}
currentOffset++;
} while (currentOffset < lastEOFMarker && !source.IsEof());
if ((lastEOFMarker < long.MaxValue || endobjFound) && lastObjOffset > 0)
{
// if the pdf wasn't cut off in the middle or if the last object ends with a "endobj" marker
// the last object id has to be added here so that it can't get lost as there isn't any subsequent object id
bfSearchCOSObjectKeyOffsets[new CosObjectKey(lastObjectId, lastGenID)] = lastObjOffset;
}
// reestablish origin position
source.Seek(originOffset);
}
public CosBase Parse(IRandomAccessRead reader, CosObjectPool pool)
{
CosBase retval = null;
ReadHelper.SkipSpaces(reader);
int nextByte = reader.Peek();
if (nextByte == -1)
{
return(null);
}
char c = (char)nextByte;
switch (c)
{
case '<':
{
// pull off first left bracket
int leftBracket = reader.Read();
// check for second left bracket
c = (char)reader.Peek();
reader.Unread(leftBracket);
if (c == '<')
{
retval = dictionaryParser.Parse(reader, this, pool);
ReadHelper.SkipSpaces(reader);
}
else
{
retval = stringParser.Parse(reader);
}
break;
}
case '[':
{
// array
retval = arrayParser.Parse(reader, this, pool);
break;
}
case '(':
retval = stringParser.Parse(reader);
break;
case '/':
// name
retval = nameParser.Parse(reader);
break;
case 'n':
{
// null
ReadHelper.ReadExpectedString(reader, "null");
retval = CosNull.Null;
break;
}
case 't':
{
string truestring = OtherEncodings.BytesAsLatin1String(reader.ReadFully(4));
if (truestring.Equals("true"))
{
retval = PdfBoolean.True;
}
else
{
throw new IOException("expected true actual='" + truestring + "' " + reader +
"' at offset " + reader.GetPosition());
}
break;
}
case 'f':
{
string falsestring = OtherEncodings.BytesAsLatin1String(reader.ReadFully(5));
if (falsestring.Equals("false"))
{
retval = PdfBoolean.False;
}
else
{
throw new IOException("expected false actual='" + falsestring + "' " + reader +
"' at offset " + reader.GetPosition());
}
break;
}
case 'R':
reader.Read();
retval = new CosObject(null);
break;
default:
if (char.IsDigit(c) || c == '-' || c == '+' || c == '.')
{
StringBuilder buf = new StringBuilder();
int ic = reader.Read();
c = (char)ic;
while (char.IsDigit(c) ||
c == '-' ||
c == '+' ||
c == '.' ||
c == 'E' ||
c == 'e')
{
buf.Append(c);
ic = reader.Read();
c = (char)ic;
}
if (ic != -1)
{
reader.Unread(ic);
}
retval = CosNumberFactory.get(buf.ToString()) as CosBase;
}
else
{
//This is not suppose to happen, but we will allow for it
//so we are more compatible with POS writers that don't
//follow the spec
string badstring = ReadHelper.ReadString(reader);
if (badstring == string.Empty)
{
int peek = reader.Peek();
// we can end up in an infinite loop otherwise
throw new IOException("Unknown dir object c='" + c +
"' cInt=" + (int)c + " peek='" + (char)peek
+ "' peekInt=" + peek + " at offset " + reader.GetPosition());
}
// if it's an endstream/endobj, we want to put it back so the caller will see it
if (string.Equals("endobj", badstring) || string.Equals("endstream", badstring))
{
reader.Unread(OtherEncodings.StringAsLatin1Bytes(badstring));
}
}
break;
}
return(retval);
}