/**
* The document catalog can also have a /Version parameter which overrides the version specified
* in the header if, and only if it is greater.
*
* @param parsedTrailer the parsed catalog in the trailer
*/
private void readVersionInTrailer(COSDictionary parsedTrailer) {
COSObject root = (COSObject) parsedTrailer.getItem(COSName.ROOT);
if (root != null) {
COSName version = (COSName) root.getItem(COSName.VERSION);
if (version != null) {
float trailerVersion = Float.valueOf(version.getName());
if (trailerVersion > document.getVersion()) {
document.setVersion(trailerVersion);
}
}
}
}
/**
* This will parse the next object from the stream and add it to the local state.
*
* @return Returns true if the processed object had an endOfFile marker
* @throws IOException If an IO error occurs.
*/
private boolean parseObject() throws IOException {
long currentObjByteOffset = pdfSource.getOffset();
boolean isEndOfFile = false;
skipSpaces();
// peek at the next character to determine the type of object we are parsing
char peekedChar = (char) pdfSource.peek();
// ignore endobj and endstream sections.
while (peekedChar == 'e') {
// there are times when there are multiple endobj, so lets
// just read them and move on.
readString();
skipSpaces();
currentObjByteOffset = pdfSource.getOffset();
peekedChar = (char) pdfSource.peek();
}
if (pdfSource.isEOF()) {
// "Skipping because of EOF" );
// end of file we will return a false and call it a day.
}
// xref table. Note: The contents of the Xref table are currently ignored
else if (peekedChar == 'x') {
parseXrefTable(currentObjByteOffset);
}
// Note: startxref can occur in either a trailer section or by itself
else if (peekedChar == 't' || peekedChar == 's') {
if (peekedChar == 't') {
parseTrailer();
peekedChar = (char) pdfSource.peek();
}
if (peekedChar == 's') {
parseStartXref();
// readString() calls skipSpaces() will skip comments... that's
// bad for us b/c the %%EOF flag is a comment
while (isWhitespace(pdfSource.peek()) && !pdfSource.isEOF()) {
pdfSource.read(); // read (get rid of) all the whitespace
}
String eof = "";
if (!pdfSource.isEOF()) {
eof = readLine(); // if there's more data to read, get the EOF flag
}
// verify that EOF exists (see PDFBOX-979 for documentation on special cases)
if (!"%%EOF".equals(eof)) {
if (eof.startsWith("%%EOF")) {
// content after marker -> unread with first space byte for read newline
pdfSource.unread(SPACE_BYTE); // we read a whole line; add space as newline replacement
pdfSource.unread(eof.substring(5).getBytes("ISO-8859-1"));
} else {
// PDF does not conform to spec, we should warn someone
Log.w("expected='%%EOF' actual='" + eof + "'", PDFBox.LOG_TAG);
// if we're not at the end of a file, just put it back and move on
if (!pdfSource.isEOF()) {
pdfSource.unread(
SPACE_BYTE); // we read a whole line; add space as newline replacement
pdfSource.unread(eof.getBytes("ISO-8859-1"));
}
}
}
isEndOfFile = true;
}
}
// we are going to parse an normal object
else {
long number = -1;
int genNum = -1;
String objectKey = null;
boolean missingObjectNumber = false;
try {
char peeked = (char) pdfSource.peek();
if (peeked == '<') {
missingObjectNumber = true;
} else {
number = readObjectNumber();
}
} catch (IOException e) {
// ok for some reason "GNU Ghostscript 5.10" puts two endobj
// statements after an object, of course this is nonsense
// but because we want to support as many PDFs as possible
// we will simply try again
number = readObjectNumber();
}
if (!missingObjectNumber) {
skipSpaces();
genNum = readGenerationNumber();
objectKey = readString(3);
// System.out.println( "parseObject() num=" + number +
// " genNumber=" + genNum + " key='" + objectKey + "'" );
if (!objectKey.equals("obj")) {
if (!isContinueOnError(null) || !objectKey.equals("o")) {
throw new IOException("expected='obj' actual='" + objectKey + "' " + pdfSource);
}
// assume that "o" was meant to be "obj" (this is a workaround for
// PDFBOX-773 attached PDF Andersens_Fairy_Tales.pdf).
}
} else {
number = -1;
genNum = -1;
}
skipSpaces();
COSBase pb = parseDirObject();
String endObjectKey = readString();
if (endObjectKey.equals("stream")) {
pdfSource.unread(endObjectKey.getBytes("ISO-8859-1"));
pdfSource.unread(' ');
if (pb instanceof COSDictionary) {
pb = parseCOSStream((COSDictionary) pb, getDocument().getScratchFile());
// test for XRef type
final COSStream strmObj = (COSStream) pb;
final COSName objectType = (COSName) strmObj.getItem(COSName.TYPE);
if (objectType != null && objectType.equals(COSName.XREF)) {
// XRef stream
parseXrefStream(strmObj, currentObjByteOffset);
}
} else {
// this is not legal
// the combination of a dict and the stream/endstream forms a complete stream object
throw new IOException("stream not preceded by dictionary");
}
skipSpaces();
endObjectKey = readLine();
}
COSObjectKey key = new COSObjectKey(number, genNum);
COSObject pdfObject = document.getObjectFromPool(key);
if (pdfObject.getObject() == null) {
pdfObject.setObject(pb);
}
/*
* If the object we returned already has a baseobject, then we have a conflict
* which we will resolve using information after we parse the xref table.
*/
else {
addObjectToConflicts(currentObjByteOffset, key, pb);
}
if (!endObjectKey.equals("endobj")) {
if (endObjectKey.startsWith("endobj")) {
/*
* Some PDF files don't contain a new line after endobj so we
* need to make sure that the next object number is getting read separately
* and not part of the endobj keyword. Ex. Some files would have "endobj28"
* instead of "endobj"
*/
pdfSource.unread(
SPACE_BYTE); // add a space first in place of the newline consumed by readline()
pdfSource.unread(endObjectKey.substring(6).getBytes("ISO-8859-1"));
} else if (endObjectKey.trim().endsWith("endobj")) {
/*
* Some PDF files contain junk (like ">> ", in the case of a PDF
* I found which was created by Exstream Dialogue Version 5.0.039)
* in which case we ignore the data before endobj and just move on
*/
Log.w("expected='endobj' actual='" + endObjectKey + "' ", PDFBox.LOG_TAG);
} else if (!pdfSource.isEOF()) {
// It is possible that the endobj is missing, there
// are several PDFs out there that do that so. Unread
// and assume that endobj was missing
pdfSource.unread(
SPACE_BYTE); // add a space first in place of the newline consumed by readline()
pdfSource.unread(endObjectKey.getBytes("ISO-8859-1"));
}
}
skipSpaces();
}
return isEndOfFile;
}
