private long GetLastEndOfFileMarker()
{
var originalOffset = reader.GetPosition();
var searchTerm = OtherEncodings.StringAsLatin1Bytes("%%EOF");
var minimumEndOffset = reader.Length() - searchTerm.Length;
reader.Seek(minimumEndOffset);
while (reader.GetPosition() > 0)
{
if (ReadHelper.IsString(reader, searchTerm))
{
var position = reader.GetPosition();
reader.Seek(originalOffset);
return(position);
}
reader.Seek(minimumEndOffset--);
}
reader.Seek(originalOffset);
return(long.MaxValue);
}
private bool validateStreamLength(IRandomAccessRead source, long streamLength, long fileLength)
{
bool streamLengthIsValid = true;
long originOffset = source.GetPosition();
long expectedEndOfStream = originOffset + streamLength;
if (expectedEndOfStream > fileLength)
{
streamLengthIsValid = false;
//LOG.warn("The end of the stream is out of range, using workaround to read the stream, "
// + "stream start position: " + originOffset + ", length: " + streamLength
// + ", expected end position: " + expectedEndOfStream);
}
else
{
source.Seek(expectedEndOfStream);
ReadHelper.SkipSpaces(source);
if (!ReadHelper.IsString(source, "endstream"))
{
streamLengthIsValid = false;
//LOG.warn("The end of the stream doesn't point to the correct offset, using workaround to read the stream, "
// + "stream start position: " + originOffset + ", length: " + streamLength
// + ", expected end position: " + expectedEndOfStream);
}
source.Seek(originOffset);
}
return(streamLengthIsValid);
}
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 static CosBase ParseValue(IRandomAccessRead reader, IBaseParser baseParser, CosObjectPool pool)
{
var numOffset = reader.GetPosition();
var value = baseParser.Parse(reader, pool);
ReadHelper.SkipSpaces(reader);
// proceed if the given object is a number and the following is a number as well
if (!(value is ICosNumber) || !ReadHelper.IsDigit(reader))
{
return(value);
}
// read the remaining information of the object number
var genOffset = reader.GetPosition();
var generationNumber = baseParser.Parse(reader, pool);
ReadHelper.SkipSpaces(reader);
ReadHelper.ReadExpectedChar(reader, 'R');
if (!(value is CosInt))
{
throw new InvalidOperationException("expected number, actual=" + value + " at offset " + numOffset);
}
if (!(generationNumber is CosInt))
{
throw new InvalidOperationException("expected number, actual=" + value + " at offset " + genOffset);
}
var key = new CosObjectKey(((CosInt)value).AsLong(), ((CosInt)generationNumber).AsInt());
// dereference the object
return(pool.Get(key));
}
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 static int ReadGenerationNumber(IRandomAccessRead reader)
{
int retval = ReadHelper.ReadInt(reader);
if (retval < 0 || retval > GenerationNumberThreshold)
{
throw new FormatException("Generation Number '" + retval + "' has more than 5 digits");
}
return(retval);
}
public static long ReadObjectNumber(IRandomAccessRead reader)
{
long retval = ReadHelper.ReadLong(reader);
if (retval < 0 || retval >= ObjectNumberThreshold)
{
throw new FormatException($"Object Number \'{retval}\' has more than 10 digits or is negative");
}
return(retval);
}
public PdfRawStream Parse(IRandomAccessRead reader, PdfDictionary streamDictionary, bool isLenientParsing, IPdfObjectParser parser)
{
PdfRawStream result;
// read 'stream'; this was already tested in parseObjectsDynamically()
ReadHelper.ReadExpectedString(reader, "stream");
skipWhiteSpaces(reader);
// This needs to be streamDictionary.getItem because when we are parsing, the underlying object might still be null.
ICosNumber streamLength = GetLength(reader, streamDictionary.GetItemOrDefault(CosName.LENGTH), streamDictionary.GetName(CosName.TYPE), isLenientParsing, parser);
ValidateStreamLength(reader, isLenientParsing, streamLength);
// get output stream to copy data to
using (var stream = new MemoryStream())
using (var writer = new BinaryWriter(stream))
{
if (streamLength != null && validateStreamLength(reader, streamLength.AsLong(), reader.Length()))
{
ReadValidStream(reader, writer, streamLength);
}
else
{
ReadUntilEndStream(reader, writer);
}
result = new PdfRawStream(stream.ToArray(), streamDictionary);
}
String endStream = ReadHelper.ReadString(reader);
if (endStream.Equals("endobj") && isLenientParsing)
{
log.Warn($"stream ends with \'endobj\' instead of \'endstream\' at offset {reader.GetPosition()}");
// avoid follow-up warning about missing endobj
reader.Rewind("endobj".Length);
}
else if (endStream.Length > 9 && isLenientParsing && endStream.Substring(0, 9).Equals("endstream"))
{
log.Warn("stream ends with '" + endStream + "' instead of 'endstream' at offset " + reader.GetPosition());
// unread the "extra" bytes
reader.Rewind(OtherEncodings.StringAsLatin1Bytes(endStream.Substring(9)).Length);
}
else if (!endStream.Equals("endstream"))
{
throw new InvalidOperationException("Error reading stream, expected='endstream' actual='"
+ endStream + "' at offset " + reader.GetPosition());
}
return(result);
}
/// <summary>
/// This will parse a PDF HEX string with fail fast semantic meaning that we stop if a not allowed character is found.
/// This is necessary in order to detect malformed input and be able to skip to next object start.
/// We assume starting '<' was already read.
/// </summary>
private static CosString ParseHexString(IRandomAccessRead reader)
{
var sBuf = new StringBuilder();
while (true)
{
int c = reader.Read();
if (ReadHelper.IsHexDigit((char)c))
{
sBuf.Append((char)c);
}
else if (c == '>')
{
break;
}
else if (c < 0)
{
throw new IOException("Missing closing bracket for hex string. Reached EOS.");
}
else if (c == ' ' || c == '\n' || c == '\t' || c == '\r' || c == '\b' || c == '\f')
{
}
else
{
// if invalid chars was found: discard last
// hex character if it is not part of a pair
if (sBuf.Length % 2 != 0)
{
sBuf.Remove(sBuf.Length - 1, 1);
}
// read till the closing bracket was found
do
{
c = reader.Read();
}while (c != '>' && c >= 0);
// might have reached EOF while looking for the closing bracket
// this can happen for malformed PDFs only. Make sure that there is
// no endless loop.
if (c < 0)
{
throw new IOException("Missing closing bracket for hex string. Reached EOS.");
}
// exit loop
break;
}
}
return(CosString.ParseHex(sBuf.ToString()));
}
public CosName Parse([NotNull] IRandomAccessRead reader)
{
if (reader == null)
{
throw new ArgumentNullException(nameof(reader));
}
ReadHelper.ReadExpectedChar(reader, '/');
using (var memoryStream = new MemoryStream())
using (var writer = new BinaryWriter(memoryStream))
{
int c = reader.Read();
while (c != -1)
{
byte ch = (byte)c;
if (ch == '#')
{
int ch1 = reader.Read();
int ch2 = reader.Read();
// Prior to PDF v1.2, the # was not a special character. Also,
// it has been observed that various PDF tools do not follow the
// spec with respect to the # escape, even though they report
// PDF versions of 1.2 or later. The solution here is that we
// interpret the # as an escape only when it is followed by two
// valid hex digits.
if (ReadHelper.IsHexDigit((char)ch1) && ReadHelper.IsHexDigit((char)ch2))
{
string hex = "" + (char)ch1 + (char)ch2;
try
{
var byteToWrite = (byte)Convert.ToInt32(hex, 16);
writer.Write(byteToWrite);
}
catch (FormatException e)
{
throw new IOException("Error: expected hex digit, actual='" + hex + "'", e);
}
c = reader.Read();
}
else
{
// check for premature EOF
if (ch2 == -1 || ch1 == -1)
{
//LOG.error("Premature EOF in BaseParser#parseCosName");
c = -1;
break;
}
reader.Unread(ch2);
c = ch1;
writer.Write(ch);
}
}
else if (ReadHelper.IsEndOfName(ch))
{
break;
}
else
{
writer.Write(ch);
c = reader.Read();
}
}
if (c != -1)
{
reader.Unread(c);
}
byte[] bytes = memoryStream.ToArray();
var str = ReadHelper.IsValidUtf8(bytes) ? Encoding.UTF8.GetString(memoryStream.ToArray()) : Encoding.GetEncoding("windows-1252").GetString(memoryStream.ToArray());
return(CosName.Create(str));
}
}
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 CosString Parse(IRandomAccessRead seqSource)
{
char nextChar = (char)seqSource.Read();
if (nextChar == '<')
{
return(ParseHexString(seqSource));
}
if (nextChar != '(')
{
throw new IOException("parseCOSstring string should start with '(' or '<' and not '" +
nextChar + "' " + seqSource);
}
using (var memoryStream = new MemoryStream())
using (var writer = new StreamWriter(memoryStream))
{
// This is the number of braces read
int braces = 1;
int c = seqSource.Read();
while (braces > 0 && c != -1)
{
char ch = (char)c;
int nextc = -2; // not yet read
if (ch == ')')
{
braces--;
braces = CheckForEndOfString(seqSource, braces);
if (braces != 0)
{
writer.Write(ch);
}
}
else if (ch == '(')
{
braces++;
writer.Write(ch);
}
else if (ch == '\\')
{
//patched by ram
char next = (char)seqSource.Read();
switch (next)
{
case 'n':
writer.Write('\n');
break;
case 'r':
writer.Write('\r');
break;
case 't':
writer.Write('\t');
break;
case 'b':
writer.Write('\b');
break;
case 'f':
writer.Write('\f');
break;
case ')':
// PDFBox 276 /Title (c:\)
braces = CheckForEndOfString(seqSource, braces);
if (braces != 0)
{
writer.Write(next);
}
else
{
writer.Write('\\');
}
break;
case '(':
case '\\':
writer.Write(next);
break;
case '0':
case '1':
case '2':
case '3':
case '4':
case '5':
case '6':
case '7':
{
var octal = new StringBuilder();
octal.Append(next);
c = seqSource.Read();
char digit = (char)c;
if (digit >= '0' && digit <= '7')
{
octal.Append(digit);
c = seqSource.Read();
digit = (char)c;
if (digit >= '0' && digit <= '7')
{
octal.Append(digit);
}
else
{
nextc = c;
}
}
else
{
nextc = c;
}
int character;
try
{
character = Convert.ToInt32(octal.ToString(), 8);
}
catch (FormatException e)
{
throw new IOException("Error: Expected octal character, actual='" + octal + "'", e);
}
writer.Write(character);
break;
}
default:
if (c == ReadHelper.AsciiCarriageReturn || c == ReadHelper.AsciiLineFeed)
{
// this is a break in the line so ignore it and the newline and continue
c = seqSource.Read();
while (ReadHelper.IsEndOfLine(c) && c != -1)
{
c = seqSource.Read();
}
nextc = c;
break;
}
// dropping the backslash
// see 7.3.4.2 Literal strings for further information
writer.Write(next);
break;
}
}
else
{
writer.Write(ch);
}
if (nextc != -2)
{
c = nextc;
}
else
{
c = seqSource.Read();
}
}
if (c != -1)
{
seqSource.Unread(c);
}
writer.Flush();
return(new CosString(memoryStream.ToArray()));
}
}
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);
}
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);
}
