def _readInlineImage(self, stream):
# begin reading just after the "BI" - begin image
# first read the dictionary of settings.
settings = DictionaryObject()
while True:
tok = readNonWhitespace(stream)
stream.seek(-1, 1)
if tok == "I":
# "ID" - begin of image data
break
key = readObject(stream, self.pdf)
tok = readNonWhitespace(stream)
stream.seek(-1, 1)
value = readObject(stream, self.pdf)
settings[key] = value
# left at beginning of ID
tmp = stream.read(3)
assert tmp[:2] == "ID"
data = ""
while True:
tok = stream.read(1)
if tok == "E":
next = stream.read(1)
if next == "I":
break
else:
stream.seek(-1, 1)
data += tok
else:
data += tok
readNonWhitespace(stream)
stream.seek(-1, 1)
return {"settings": settings, "data": data}
def _readInlineImage(self, stream):
# begin reading just after the "BI" - begin image
# first read the dictionary of settings.
settings = DictionaryObject()
while True:
tok = readNonWhitespace(stream)
stream.seek(-1, 1)
if tok == "I":
# "ID" - begin of image data
break
key = readObject(stream, self.pdf)
tok = readNonWhitespace(stream)
stream.seek(-1, 1)
value = readObject(stream, self.pdf)
settings[key] = value
# left at beginning of ID
tmp = stream.read(3)
assert tmp[:2] == "ID"
data = ""
while True:
tok = stream.read(1)
if tok == "E":
next = stream.read(1)
if next == "I":
break
else:
stream.seek(-1, 1)
data += tok
else:
data += tok
readNonWhitespace(stream)
stream.seek(-1, 1)
return {"settings": settings, "data": data}
def __parseContentStream(self, stream):
stream.seek(0, 0)
operands = []
while True:
peek = readNonWhitespace(stream)
if peek == '':
break
stream.seek(-1, 1)
if peek.isalpha() or peek == "'" or peek == '"':
operator = ""
while True:
tok = stream.read(1)
if tok.isspace() or tok in NameObject.delimiterCharacters:
stream.seek(-1, 1)
break
elif tok == '':
break
operator += tok
if operator == "BI":
# begin inline image - a completely different parsing
# mechanism is required, of course... thanks buddy...
assert operands == []
ii = self._readInlineImage(stream)
self.operations.append((ii, "INLINE IMAGE"))
else:
self.operations.append((operands, operator))
operands = []
elif peek == '%':
# If we encounter a comment in the content stream, we have to
# handle it here. Typically, readObject will handle
# encountering a comment -- but readObject assumes that
# following the comment must be the object we're trying to
# read. In this case, it could be an operator instead.
while peek not in ('\r', '\n'):
peek = stream.read(1)
else:
operands.append(readObject(stream, None))
def __parseContentStream(self, stream):
stream.seek(0, 0)
operands = []
while True:
peek = readNonWhitespace(stream)
if peek == '':
break
stream.seek(-1, 1)
if peek.isalpha() or peek == "'" or peek == '"':
operator = ""
while True:
tok = stream.read(1)
if tok.isspace() or tok in NameObject.delimiterCharacters:
stream.seek(-1, 1)
break
elif tok == '':
break
operator += tok
if operator == "BI":
# begin inline image - a completely different parsing
# mechanism is required, of course... thanks buddy...
assert operands == []
ii = self._readInlineImage(stream)
self.operations.append((ii, "INLINE IMAGE"))
else:
self.operations.append((operands, operator))
operands = []
elif peek == '%':
# If we encounter a comment in the content stream, we have to
# handle it here. Typically, readObject will handle
# encountering a comment -- but readObject assumes that
# following the comment must be the object we're trying to
# read. In this case, it could be an operator instead.
while peek not in ('\r', '\n'):
peek = stream.read(1)
else:
operands.append(readObject(stream, None))
def read(self, stream):
# start at the end:
stream.seek(-1, 2)
line = b_('')
while not line:
line = self.readNextEndLine(stream)
if line[:5] != b_("%%EOF"):
raise utils.PdfReadError, "EOF marker not found"
# find startxref entry - the location of the xref table
line = self.readNextEndLine(stream)
startxref = int(line)
line = self.readNextEndLine(stream)
if line[:9] != b_("startxref"):
raise utils.PdfReadError, "startxref not found"
# read all cross reference tables and their trailers
self.xref = {}
self.xref_objStm = {}
self.trailer = DictionaryObject()
while 1:
# load the xref table
stream.seek(startxref, 0)
x = stream.read(1)
if x == b_("x"):
# standard cross-reference table
ref = stream.read(4)
if ref[:3] != b_("ref"):
raise utils.PdfReadError, "xref table read error"
readNonWhitespace(stream)
stream.seek(-1, 1)
# check if the first time looking at the xref table
firsttime = True
while True:
num = readObject(stream, self)
if firsttime and num != 0:
self.xrefIndex = num
warnings.warn("Xref table not zero-indexed. ID "
"numbers for objects will %sbe "
"corrected." %
("" if not self.strict else "not "),
utils.PdfReadWarning)
# if table not zero indexed, could be due to
# error from when PDF was created
# which will lead to mismatched indices later on
firsttime = False
readNonWhitespace(stream)
stream.seek(-1, 1)
size = readObject(stream, self)
readNonWhitespace(stream)
stream.seek(-1, 1)
cnt = 0
while cnt < size:
line = stream.read(20)
# It's very clear in section 3.4.3 of the PDF spec
# that all cross-reference table lines are a fixed
# 20 bytes (as of PDF 1.7). However, some files have
# 21-byte entries (or more) due to the use of \r\n
# (CRLF) EOL's. Detect that case, and adjust the line
# until it does not begin with a \r (CR) or \n (LF).
while line[0] in b_("\x0D\x0A"):
stream.seek(-20 + 1, 1)
line = stream.read(20)
# On the other hand, some malformed PDF files
# use a single character EOL without a preceeding
# space. Detect that case, and seek the stream
# back one character. (0-9 means we've bled into
# the next xref entry, t means we've bled into the
# text "trailer"):
if line[-1] in b_("0123456789t"):
stream.seek(-1, 1)
offset, generation = line[:16].split(b_(" "))
offset, generation = int(offset), int(generation)
self.xref.setdefault(generation, {})
if num in self.xref[generation]:
# It really seems like we should allow the last
# xref table in the file to override previous
# ones. Since we read the file backwards, assume
# any existing key is already set correctly.
pass
else:
self.xref[generation][num] = offset
cnt += 1
num += 1
readNonWhitespace(stream)
stream.seek(-1, 1)
trailertag = stream.read(7)
if trailertag != b_("trailer"):
# more xrefs!
stream.seek(-7, 1)
else:
break
readNonWhitespace(stream)
stream.seek(-1, 1)
newTrailer = readObject(stream, self)
for key, value in newTrailer.items():
self.trailer.setdefault(key, value)
if "/Prev" in newTrailer:
startxref = newTrailer["/Prev"]
else:
break
elif x.isdigit():
# PDF 1.5+ Cross-Reference Stream
stream.seek(-1, 1)
idnum, generation = self.readObjectHeader(stream)
xrefstream = readObject(stream, self)
assert xrefstream["/Type"] == "/XRef"
self.cacheIndirectObject(generation, idnum, xrefstream)
streamData = StringIO(xrefstream.getData())
idx_pairs = xrefstream.get("/Index",
[0, xrefstream.get("/Size")])
entrySizes = xrefstream.get("/W")
for num, size in self._pairs(idx_pairs):
cnt = 0
while cnt < size:
for i in range(len(entrySizes)):
d = streamData.read(entrySizes[i])
di = convertToInt(d, entrySizes[i])
if i == 0:
xref_type = di
elif i == 1:
if xref_type == 0:
# next_free_object = di
pass
elif xref_type == 1:
byte_offset = di
elif xref_type == 2:
objstr_num = di
elif i == 2:
if xref_type == 0:
# next_generation = di
pass
elif xref_type == 1:
generation = di
elif xref_type == 2:
obstr_idx = di
if xref_type == 0:
pass
elif xref_type == 1:
if generation not in self.xref:
self.xref[generation] = {}
if not num in self.xref[generation]:
self.xref[generation][num] = byte_offset
elif xref_type == 2:
if not num in self.xref_objStm:
self.xref_objStm[num] = [objstr_num, obstr_idx]
cnt += 1
num += 1
trailerKeys = "/Root", "/Encrypt", "/Info", "/ID"
for key in trailerKeys:
if key in xrefstream and key not in self.trailer:
self.trailer[NameObject(key)] = xrefstream.raw_get(key)
if "/Prev" in xrefstream:
startxref = xrefstream["/Prev"]
else:
break
else:
# bad xref character at startxref. Let's see if we can find
# the xref table nearby, as we've observed this error with an
# off-by-one before.
stream.seek(-11, 1)
tmp = stream.read(20)
xref_loc = tmp.find(b_("xref"))
if xref_loc != -1:
startxref -= (10 - xref_loc)
continue
else:
# no xref table found at specified location
assert False
break
# if not zero-indexed, verify that the table is correct
# change it if necessary
if self.xrefIndex and not self.strict:
loc = stream.tell()
for gen in self.xref:
if gen == 65535:
continue
for id in self.xref[gen]:
stream.seek(self.xref[gen][id], 0)
pid, pgen = self.readObjectHeader(stream)
if pid == id - self.xrefIndex:
self._zeroXref(gen)
break
# if not, then either it's just plain wrong,
# or the non-zero-index is actually correct
stream.seek(loc, 0) # return to where it was
def getObject(self, indirectReference):
retval = self.resolvedObjects.get(indirectReference.generation,
{}).get(indirectReference.idnum,
None)
if retval is not None:
return retval
if indirectReference.generation == 0 \
and indirectReference.idnum in self.xref_objStm:
# indirect reference to object in object stream
# read the entire object stream into memory
stmnum, idx = self.xref_objStm[indirectReference.idnum]
objStm = IndirectObject(stmnum, 0, self).getObject()
assert objStm['/Type'] == '/ObjStm'
assert idx < objStm['/N']
streamData = StringIO(objStm.getData())
for i in range(objStm['/N']):
objnum = NumberObject.readFromStream(streamData)
readNonWhitespace(streamData)
streamData.seek(-1, 1)
offset = NumberObject.readFromStream(streamData)
readNonWhitespace(streamData)
streamData.seek(-1, 1)
t = streamData.tell()
streamData.seek(objStm['/First']+offset, 0)
obj = readObject(streamData, self)
self.resolvedObjects[0][objnum] = obj
streamData.seek(t, 0)
return self.resolvedObjects[0][indirectReference.idnum]
if indirectReference.idnum \
not in self.xref[indirectReference.generation]:
warnings.warn("Object %d %d not defined." % (
indirectReference.idnum, indirectReference.generation),
utils.PdfReadWarning)
return None
start = self.xref[indirectReference.generation][
indirectReference.idnum]
self.stream.seek(start, 0)
idnum, generation = self.readObjectHeader(self.stream)
try:
assert idnum == indirectReference.idnum
except AssertionError:
if self.xrefIndex:
# Xref table probably had bad indexes due to not
# being zero-indexed
if self.strict:
raise utils.PdfReadError(
"Expected object ID (%d %d) does "
"not match actual (%d %d); xref "
"table not zero-indexed." % (
indirectReference.idnum,
indirectReference.generation,
idnum,
generation))
else:
# should not happen since the xref table is corrected in
# non-strict mode
pass
else: # some other problem
raise utils.PdfReadError("Expected object ID (%d %d) does not "
" match actual (%d %d)." % (
indirectReference.idnum,
indirectReference.generation,
idnum, generation))
assert generation == indirectReference.generation
retval = readObject(self.stream, self)
# override encryption is used for the /Encrypt dictionary
if not self._override_encryption and self.isEncrypted:
# if we don't have the encryption key:
if not hasattr(self, '_decryption_key'):
raise Exception("file has not been decrypted")
# otherwise, decrypt here...
pack1 = struct.pack("<i", indirectReference.idnum)[:3]
pack2 = struct.pack("<i", indirectReference.generation)[:2]
key = self._decryption_key + pack1 + pack2
assert len(key) == (len(self._decryption_key) + 5)
md5_hash = md5(key).digest()
key = md5_hash[:min(16, len(self._decryption_key) + 5)]
retval = self._decryptObject(retval, key)
self.cacheIndirectObject(generation, idnum, retval)
return retval
def read(self, stream):
# start at the end:
stream.seek(-1, 2)
line = b_('')
while not line:
line = self.readNextEndLine(stream)
if line[:5] != b_("%%EOF"):
raise utils.PdfReadError, "EOF marker not found"
# find startxref entry - the location of the xref table
line = self.readNextEndLine(stream)
startxref = int(line)
line = self.readNextEndLine(stream)
if line[:9] != b_("startxref"):
raise utils.PdfReadError, "startxref not found"
# read all cross reference tables and their trailers
self.xref = {}
self.xref_objStm = {}
self.trailer = DictionaryObject()
while 1:
# load the xref table
stream.seek(startxref, 0)
x = stream.read(1)
if x == b_("x"):
# standard cross-reference table
ref = stream.read(4)
if ref[:3] != b_("ref"):
raise utils.PdfReadError, "xref table read error"
readNonWhitespace(stream)
stream.seek(-1, 1)
# check if the first time looking at the xref table
firsttime = True
while True:
num = readObject(stream, self)
if firsttime and num != 0:
self.xrefIndex = num
warnings.warn(
"Xref table not zero-indexed. ID "
"numbers for objects will %sbe "
"corrected." % ("" if not self.strict else "not "),
utils.PdfReadWarning)
# if table not zero indexed, could be due to
# error from when PDF was created
# which will lead to mismatched indices later on
firsttime = False
readNonWhitespace(stream)
stream.seek(-1, 1)
size = readObject(stream, self)
readNonWhitespace(stream)
stream.seek(-1, 1)
cnt = 0
while cnt < size:
line = stream.read(20)
# It's very clear in section 3.4.3 of the PDF spec
# that all cross-reference table lines are a fixed
# 20 bytes (as of PDF 1.7). However, some files have
# 21-byte entries (or more) due to the use of \r\n
# (CRLF) EOL's. Detect that case, and adjust the line
# until it does not begin with a \r (CR) or \n (LF).
while line[0] in b_("\x0D\x0A"):
stream.seek(-20 + 1, 1)
line = stream.read(20)
# On the other hand, some malformed PDF files
# use a single character EOL without a preceeding
# space. Detect that case, and seek the stream
# back one character. (0-9 means we've bled into
# the next xref entry, t means we've bled into the
# text "trailer"):
if line[-1] in b_("0123456789t"):
stream.seek(-1, 1)
offset, generation = line[:16].split(b_(" "))
offset, generation = int(offset), int(generation)
self.xref.setdefault(generation, {})
if num in self.xref[generation]:
# It really seems like we should allow the last
# xref table in the file to override previous
# ones. Since we read the file backwards, assume
# any existing key is already set correctly.
pass
else:
self.xref[generation][num] = offset
cnt += 1
num += 1
readNonWhitespace(stream)
stream.seek(-1, 1)
trailertag = stream.read(7)
if trailertag != b_("trailer"):
# more xrefs!
stream.seek(-7, 1)
else:
break
readNonWhitespace(stream)
stream.seek(-1, 1)
newTrailer = readObject(stream, self)
for key, value in newTrailer.items():
self.trailer.setdefault(key, value)
if "/Prev" in newTrailer:
startxref = newTrailer["/Prev"]
else:
break
elif x.isdigit():
# PDF 1.5+ Cross-Reference Stream
stream.seek(-1, 1)
idnum, generation = self.readObjectHeader(stream)
xrefstream = readObject(stream, self)
assert xrefstream["/Type"] == "/XRef"
self.cacheIndirectObject(generation, idnum, xrefstream)
streamData = StringIO(xrefstream.getData())
idx_pairs = xrefstream.get("/Index",
[0, xrefstream.get("/Size")])
entrySizes = xrefstream.get("/W")
for num, size in self._pairs(idx_pairs):
cnt = 0
while cnt < size:
for i in range(len(entrySizes)):
d = streamData.read(entrySizes[i])
di = convertToInt(d, entrySizes[i])
if i == 0:
xref_type = di
elif i == 1:
if xref_type == 0:
# next_free_object = di
pass
elif xref_type == 1:
byte_offset = di
elif xref_type == 2:
objstr_num = di
elif i == 2:
if xref_type == 0:
# next_generation = di
pass
elif xref_type == 1:
generation = di
elif xref_type == 2:
obstr_idx = di
if xref_type == 0:
pass
elif xref_type == 1:
if generation not in self.xref:
self.xref[generation] = {}
if not num in self.xref[generation]:
self.xref[generation][num] = byte_offset
elif xref_type == 2:
if not num in self.xref_objStm:
self.xref_objStm[num] = [objstr_num, obstr_idx]
cnt += 1
num += 1
trailerKeys = "/Root", "/Encrypt", "/Info", "/ID"
for key in trailerKeys:
if key in xrefstream and key not in self.trailer:
self.trailer[NameObject(key)] = xrefstream.raw_get(key)
if "/Prev" in xrefstream:
startxref = xrefstream["/Prev"]
else:
break
else:
# bad xref character at startxref. Let's see if we can find
# the xref table nearby, as we've observed this error with an
# off-by-one before.
stream.seek(-11, 1)
tmp = stream.read(20)
xref_loc = tmp.find(b_("xref"))
if xref_loc != -1:
startxref -= (10 - xref_loc)
continue
else:
# no xref table found at specified location
assert False
break
# if not zero-indexed, verify that the table is correct
# change it if necessary
if self.xrefIndex and not self.strict:
loc = stream.tell()
for gen in self.xref:
if gen == 65535:
continue
for id in self.xref[gen]:
stream.seek(self.xref[gen][id], 0)
pid, pgen = self.readObjectHeader(stream)
if pid == id - self.xrefIndex:
self._zeroXref(gen)
break
# if not, then either it's just plain wrong,
# or the non-zero-index is actually correct
stream.seek(loc, 0) # return to where it was
def getObject(self, indirectReference):
retval = self.resolvedObjects.get(indirectReference.generation,
{}).get(indirectReference.idnum,
None)
if retval is not None:
return retval
if indirectReference.generation == 0 \
and indirectReference.idnum in self.xref_objStm:
# indirect reference to object in object stream
# read the entire object stream into memory
stmnum, idx = self.xref_objStm[indirectReference.idnum]
objStm = IndirectObject(stmnum, 0, self).getObject()
assert objStm['/Type'] == '/ObjStm'
assert idx < objStm['/N']
streamData = StringIO(objStm.getData())
for i in range(objStm['/N']):
objnum = NumberObject.readFromStream(streamData)
readNonWhitespace(streamData)
streamData.seek(-1, 1)
offset = NumberObject.readFromStream(streamData)
readNonWhitespace(streamData)
streamData.seek(-1, 1)
t = streamData.tell()
streamData.seek(objStm['/First'] + offset, 0)
obj = readObject(streamData, self)
self.resolvedObjects[0][objnum] = obj
streamData.seek(t, 0)
return self.resolvedObjects[0][indirectReference.idnum]
if indirectReference.idnum \
not in self.xref[indirectReference.generation]:
warnings.warn(
"Object %d %d not defined." %
(indirectReference.idnum, indirectReference.generation),
utils.PdfReadWarning)
return None
start = self.xref[indirectReference.generation][
indirectReference.idnum]
self.stream.seek(start, 0)
idnum, generation = self.readObjectHeader(self.stream)
try:
assert idnum == indirectReference.idnum
except AssertionError:
if self.xrefIndex:
# Xref table probably had bad indexes due to not
# being zero-indexed
if self.strict:
raise utils.PdfReadError(
"Expected object ID (%d %d) does "
"not match actual (%d %d); xref "
"table not zero-indexed." %
(indirectReference.idnum, indirectReference.generation,
idnum, generation))
else:
# should not happen since the xref table is corrected in
# non-strict mode
pass
else: # some other problem
raise utils.PdfReadError(
"Expected object ID (%d %d) does not "
" match actual (%d %d)." %
(indirectReference.idnum, indirectReference.generation,
idnum, generation))
assert generation == indirectReference.generation
retval = readObject(self.stream, self)
# override encryption is used for the /Encrypt dictionary
if not self._override_encryption and self.isEncrypted:
# if we don't have the encryption key:
if not hasattr(self, '_decryption_key'):
raise Exception("file has not been decrypted")
# otherwise, decrypt here...
pack1 = struct.pack("<i", indirectReference.idnum)[:3]
pack2 = struct.pack("<i", indirectReference.generation)[:2]
key = self._decryption_key + pack1 + pack2
assert len(key) == (len(self._decryption_key) + 5)
md5_hash = md5(key).digest()
key = md5_hash[:min(16, len(self._decryption_key) + 5)]
retval = self._decryptObject(retval, key)
self.cacheIndirectObject(generation, idnum, retval)
return retval