def readObjectHeader(self, stream):
# Should never be necessary to read out whitespace, since the
# cross-reference table should put us in the right spot to read the
# object header. In reality... some files have stupid cross reference
# tables that are off by whitespace bytes.
extra = False
utils.skipOverComment(stream)
extra |= utils.skipOverWhitespace(stream)
stream.seek(-1, 1)
idnum = readUntilWhitespace(stream)
extra |= utils.skipOverWhitespace(stream)
stream.seek(-1, 1)
generation = readUntilWhitespace(stream)
stream.read(3)
readNonWhitespace(stream)
stream.seek(-1, 1)
if (extra and self.strict):
#not a fatal error
warnings.warn(
"Superfluous whitespace found in "
"object header %s %s" % (idnum, generation),
utils.PdfReadWarning)
return int(idnum), int(generation)
def readFromStream(stream, pdf):
debug = False
tmp = stream.read(2)
if tmp != b_("<<"):
raise utils.PdfReadError(
"Dictionary read error at byte %s: stream must begin with '<<'"
% utils.hexStr(stream.tell()))
data = {}
while True:
tok = readNonWhitespace(stream)
if tok == b_('\x00'):
continue
elif tok == b_('%'):
stream.seek(-1, 1)
skipOverComment(stream)
continue
if not tok:
# stream has truncated prematurely
raise PdfStreamError("Stream has ended unexpectedly")
if debug: print(("Tok:", tok))
if tok == b_(">"):
stream.read(1)
break
stream.seek(-1, 1)
key = readObject(stream, pdf)
tok = readNonWhitespace(stream)
stream.seek(-1, 1)
value = readObject(stream, pdf)
if not data.get(key):
data[key] = value
elif pdf.strict:
# multiple definitions of key not permitted
raise utils.PdfReadError("Multiple definitions in dictionary at byte %s for key %s" \
% (utils.hexStr(stream.tell()), key))
else:
warnings.warn("Multiple definitions in dictionary at byte %s for key %s" \
% (utils.hexStr(stream.tell()), key), utils.PdfReadWarning)
pos = stream.tell()
s = readNonWhitespace(stream)
if s == b_('s') and stream.read(5) == b_('tream'):
eol = stream.read(1)
# odd PDF file output has spaces after 'stream' keyword but before EOL.
# patch provided by Danial Sandler
while eol == b_(' '):
eol = stream.read(1)
assert eol in (b_("\n"), b_("\r"))
if eol == b_("\r"):
# read \n after
if stream.read(1) != b_('\n'):
stream.seek(-1, 1)
# this is a stream object, not a dictionary
assert "/Length" in data
length = data["/Length"]
if debug: print(data)
if isinstance(length, IndirectObject):
t = stream.tell()
length = pdf.getObject(length)
stream.seek(t, 0)
data["__streamdata__"] = stream.read(length)
if debug: print("here")
#if debug: print(binascii.hexlify(data["__streamdata__"]))
e = readNonWhitespace(stream)
ndstream = stream.read(8)
if (e + ndstream) != b_("endstream"):
# (sigh) - the odd PDF file has a length that is too long, so
# we need to read backwards to find the "endstream" ending.
# ReportLab (unknown version) generates files with this bug,
# and Python users into PDF files tend to be our audience.
# we need to do this to correct the streamdata and chop off
# an extra character.
pos = stream.tell()
stream.seek(-10, 1)
end = stream.read(9)
if end == b_("endstream"):
# we found it by looking back one character further.
data["__streamdata__"] = data["__streamdata__"][:-1]
else:
if debug: print(("E", e, ndstream, debugging.toHex(end)))
stream.seek(pos, 0)
raise utils.PdfReadError(
"Unable to find 'endstream' marker after stream at byte %s."
% utils.hexStr(stream.tell()))
else:
stream.seek(pos, 0)
if "__streamdata__" in data:
return StreamObject.initializeFromDictionary(data)
else:
retval = DictionaryObject()
retval.update(data)
return retval
def readFromStream(stream, pdf):
debug = False
tmp = stream.read(2)
if tmp != b_("<<"):
raise utils.PdfReadError("Dictionary read error at byte %s: stream must begin with '<<'" % utils.hexStr(stream.tell()))
data = {}
while True:
tok = readNonWhitespace(stream)
if tok == b_('\x00'):
continue
elif tok == b_('%'):
stream.seek(-1, 1)
skipOverComment(stream)
continue
if not tok:
# stream has truncated prematurely
raise PdfStreamError("Stream has ended unexpectedly")
if debug: print(("Tok:", tok))
if tok == b_(">"):
stream.read(1)
break
stream.seek(-1, 1)
key = readObject(stream, pdf)
tok = readNonWhitespace(stream)
stream.seek(-1, 1)
value = readObject(stream, pdf)
if not data.get(key):
data[key] = value
elif pdf.strict:
# multiple definitions of key not permitted
raise utils.PdfReadError("Multiple definitions in dictionary at byte %s for key %s" \
% (utils.hexStr(stream.tell()), key))
else:
warnings.warn("Multiple definitions in dictionary at byte %s for key %s" \
% (utils.hexStr(stream.tell()), key), utils.PdfReadWarning)
pos = stream.tell()
s = readNonWhitespace(stream)
if s == b_('s') and stream.read(5) == b_('tream'):
eol = stream.read(1)
# odd PDF file output has spaces after 'stream' keyword but before EOL.
# patch provided by Danial Sandler
while eol == b_(' '):
eol = stream.read(1)
assert eol in (b_("\n"), b_("\r"))
if eol == b_("\r"):
# read \n after
if stream.read(1) != b_('\n'):
stream.seek(-1, 1)
# this is a stream object, not a dictionary
assert "/Length" in data
length = data["/Length"]
if debug: print(data)
if isinstance(length, IndirectObject):
t = stream.tell()
length = pdf.getObject(length)
stream.seek(t, 0)
data["__streamdata__"] = stream.read(length)
if debug: print("here")
#if debug: print(binascii.hexlify(data["__streamdata__"]))
e = readNonWhitespace(stream)
ndstream = stream.read(8)
if (e + ndstream) != b_("endstream"):
# (sigh) - the odd PDF file has a length that is too long, so
# we need to read backwards to find the "endstream" ending.
# ReportLab (unknown version) generates files with this bug,
# and Python users into PDF files tend to be our audience.
# we need to do this to correct the streamdata and chop off
# an extra character.
pos = stream.tell()
stream.seek(-10, 1)
end = stream.read(9)
if end == b_("endstream"):
# we found it by looking back one character further.
data["__streamdata__"] = data["__streamdata__"][:-1]
else:
if debug: print(("E", e, ndstream, debugging.toHex(end)))
stream.seek(pos, 0)
raise utils.PdfReadError("Unable to find 'endstream' marker after stream at byte %s." % utils.hexStr(stream.tell()))
else:
stream.seek(pos, 0)
if "__streamdata__" in data:
return StreamObject.initializeFromDictionary(data)
else:
retval = DictionaryObject()
retval.update(data)
return retval
def readObjectHeader(self, stream):
# Should never be necessary to read out whitespace, since the
# cross-reference table should put us in the right spot to read the
# object header. In reality... some files have stupid cross reference
# tables that are off by whitespace bytes.
extra = False
utils.skipOverComment(stream)
extra |= utils.skipOverWhitespace(stream)
stream.seek(-1, 1)
idnum = readUntilWhitespace(stream)
extra |= utils.skipOverWhitespace(stream)
stream.seek(-1, 1)
generation = readUntilWhitespace(stream)
stream.read(3)
readNonWhitespace(stream)
stream.seek(-1, 1)
if (extra and self.strict):
#not a fatal error
warnings.warn("Superfluous whitespace found in "
"object header %s %s" % (idnum, generation),
utils.PdfReadWarning)
return int(idnum), int(generation)