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($"stream ends with \'endobj\' instead of \'endstream\' at offset {reader.GetPosition()}");
// avoid follow-up warning about missing endobj
reader.Rewind("endobj".Length);
}
else if (endStream.Length > 9 && isLenientParsing && endStream.Substring(0, 9).Equals("endstream"))
{
log.Warn("stream ends with '" + endStream + "' instead of 'endstream' at offset " + reader.GetPosition());
// unread the "extra" bytes
reader.Rewind(OtherEncodings.StringAsLatin1Bytes(endStream.Substring(9)).Length);
}
else if (!endStream.Equals("endstream"))
{
throw new InvalidOperationException("Error reading stream, expected='endstream' actual='"
+ endStream + "' at offset " + reader.GetPosition());
}
return(result);
}
public void Rewind(int bytes)
{
if (Throw)
{
throw new InvalidOperationException();
}
reader.Rewind(bytes);
}
private CosBase ParseObjectFromFile(long offset, IRandomAccessRead reader,
CosObjectKey key,
CosObjectPool pool,
bool isLenientParsing)
{
reader.Seek(offset);
var objectNumber = ObjectHelper.ReadObjectNumber(reader);
var objectGeneration = ObjectHelper.ReadGenerationNumber(reader);
ReadHelper.ReadExpectedString(reader, "obj", true);
if (objectNumber != key.Number || objectGeneration != key.Generation)
{
throw new InvalidOperationException($"Xref for {key} points to object {objectNumber} {objectGeneration} at {offset}");
}
ReadHelper.SkipSpaces(reader);
var baseObject = baseParser.Parse(reader, pool);
var endObjectKey = ReadHelper.ReadString(reader);
var atStreamStart = string.Equals(endObjectKey, "stream");
if (atStreamStart)
{
var streamStartBytes = OtherEncodings.StringAsLatin1Bytes(endObjectKey);
reader.Rewind(streamStartBytes.Length);
baseObject = ReadNormalObjectStream(reader, baseObject, offset, isLenientParsing, out endObjectKey);
}
if (!string.Equals(endObjectKey, "endobj"))
{
var message =
$"Object ({objectNumber}:{objectGeneration}) at offset {offset} does not end with \'endobj\' but with \'{endObjectKey}\'";
if (isLenientParsing)
{
log.Warn(message);
}
else
{
throw new InvalidOperationException(message);
}
}
return(baseObject);
}
public void unread(int b)
{
reader.Rewind(1);
}
public bool TryParse(IRandomAccessRead source, long offset, bool isLenientParsing, CosObjectPool pool, out CrossReferenceTablePartBuilder builder)
{
builder = null;
var tableStartOffset = source.GetPosition();
if (source.Peek() != 'x')
{
return(false);
}
var xref = ReadHelper.ReadString(source);
if (!xref.Trim().Equals("xref"))
{
return(false);
}
// check for trailer after xref
var str = ReadHelper.ReadString(source);
byte[] b = OtherEncodings.StringAsLatin1Bytes(str);
source.Rewind(b.Length);
if (str.StartsWith("trailer"))
{
log.Warn("skipping empty xref table");
return(false);
}
builder = new CrossReferenceTablePartBuilder
{
Offset = offset,
XRefType = CrossReferenceType.Table
};
// Tables can have multiple sections. Each starts with a starting object id and a count.
while (true)
{
if (!TableSubsectionDefinition.TryRead(log, source, out var subsectionDefinition))
{
log.Warn($"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 void ReadUntilEndStream(IRandomAccessRead source, BinaryWriter output)
{
int bufSize;
int charMatchCount = 0;
byte[] keyw = ENDSTREAM;
// last character position of shortest keyword ('endobj')
int quickTestOffset = 5;
// read next chunk into buffer; already matched chars are added to beginning of buffer
while ((bufSize = source.Read(strmBuf, charMatchCount, STRMBUFLEN - charMatchCount)) > 0)
{
bufSize += charMatchCount;
int bIdx = charMatchCount;
int quickTestIdx;
// iterate over buffer, trying to find keyword match
for (int maxQuicktestIdx = bufSize - quickTestOffset; bIdx < bufSize; bIdx++)
{
// reduce compare operations by first test last character we would have to
// match if current one matches; if it is not a character from keywords
// we can move behind the test character; this shortcut is inspired by the
// Boyer-Moore string search algorithm and can reduce parsing time by approx. 20%
quickTestIdx = bIdx + quickTestOffset;
if (charMatchCount == 0 && quickTestIdx < maxQuicktestIdx)
{
byte ch = strmBuf[quickTestIdx];
if ((ch > 't') || (ch < 'a'))
{
// last character we would have to match if current character would match
// is not a character from keywords -> jump behind and start over
bIdx = quickTestIdx;
continue;
}
}
// could be negative - but we only compare to ASCII
byte ch1 = strmBuf[bIdx];
if (ch1 == keyw[charMatchCount])
{
if (++charMatchCount == keyw.Length)
{
// match found
bIdx++;
break;
}
}
else
{
if ((charMatchCount == 3) && (ch1 == ENDOBJ[charMatchCount]))
{
// maybe ENDSTREAM is missing but we could have ENDOBJ
keyw = ENDOBJ;
charMatchCount++;
}
else
{
// no match; incrementing match start by 1 would be dumb since we already know
// matched chars depending on current char read we may already have beginning
// of a new match: 'e': first char matched; 'n': if we are at match position
// idx 7 we already read 'e' thus 2 chars matched for each other char we have
// to start matching first keyword char beginning with next read position
charMatchCount = (ch1 == 'e') ? 1 : ((ch1 == 'n') && (charMatchCount == 7)) ? 2 : 0;
// search again for 'endstream'
keyw = ENDSTREAM;
}
}
}
int contentBytes = Math.Max(0, bIdx - charMatchCount);
// write buffer content until first matched char to output stream
if (contentBytes > 0)
{
output.Write(strmBuf, 0, contentBytes);
}
if (charMatchCount == keyw.Length)
{
// keyword matched; unread matched keyword (endstream/endobj) and following buffered content
source.Rewind(bufSize - contentBytes);
break;
}
// copy matched chars at start of buffer
Array.Copy(keyw, 0, strmBuf, 0, charMatchCount);
}
// this writes a lonely CR or drops trailing CR LF and LF
// output.flush();
}
