def decode(self):
"""
TIFF 6.0 specification explains in sufficient details the steps to
implement the LZW encode() and decode() algorithms.
:rtype: bytes
"""
# TO-DO Make return value type bytes, as instructed by ISO 32000
cW = self.CLEARDICT
output = b""
while True:
pW = cW
cW = self._readCode()
if cW == -1:
raise PdfReadError(
"Missed the stop code in during LZW decoding")
if cW == self.STOP:
break
elif cW == self.CLEARDICT:
self._resetDict()
elif pW == self.CLEARDICT:
output += self.dict[cW]
else:
if cW < self.dictindex:
output += self.dict[cW]
if version_info > (3, 0):
p = self.dict[pW] + bytes([self.dict[cW][0]])
else:
p = self.dict[pW] + self.dict[cW][0]
self._addCodeToTable(p)
else:
if version_info > (3, 0):
p = self.dict[pW] + bytes([self.dict[pW][0]])
else:
p = self.dict[pW] + self.dict[pW][0]
output += p
self._addCodeToTable(p)
return output
def decode(data, decodeParms=None):
"""
:param data: flate-encoded data.
:param decodeParms: a dictionary of parameter values.
:return: the flate-decoded data.
:rtype: bytes
"""
data = decompress(data)
predictor = 1
if decodeParms:
try:
predictor = decodeParms.get("/Predictor", 1)
except AttributeError:
pass # Usually an array with a null object was read
# predictor 1 == no predictor
if predictor != 1:
# The /Columns param. has 1 as the default value; see ISO 32000,
# §7.4.4.3 LZWDecode and FlateDecode Parameters, Table 8
columns = decodeParms.get("/Columns", 1)
# PNG prediction:
if 10 <= predictor <= 15:
output = BytesIO()
# PNG prediction can vary from row to row
row_length = columns + 1
assert len(data) % row_length == 0
prev_rowdata = (0, ) * row_length
for row in range(len(data) // row_length):
rowdata = [
pypdfOrd(x)
for x in data[(row * row_length):((row + 1) *
row_length)]
]
filterByte = rowdata[0]
if filterByte == 0:
pass
elif filterByte == 1:
for i in range(2, row_length):
rowdata[i] = (rowdata[i] + rowdata[i - 1]) % 256
elif filterByte == 2:
for i in range(1, row_length):
rowdata[i] = (rowdata[i] + prev_rowdata[i]) % 256
elif filterByte == 3:
for i in range(1, row_length):
left = rowdata[i - 1] if i > 1 else 0
floor = math.floor(left + prev_rowdata[i]) / 2
rowdata[i] = (rowdata[i] + int(floor)) % 256
elif filterByte == 4:
for i in range(1, row_length):
left = rowdata[i - 1] if i > 1 else 0
up = prev_rowdata[i]
up_left = prev_rowdata[i - 1] if i > 1 else 0
paeth = paethPredictor(left, up, up_left)
rowdata[i] = (rowdata[i] + paeth) % 256
else:
# Unsupported PNG filter
raise PdfReadError("Unsupported PNG filter %r" %
filterByte)
prev_rowdata = rowdata
for d in rowdata:
if version_info < (3, 0):
output.write(chr(d))
else:
output.write(bytes([d]))
data = output.getvalue()
else:
# unsupported predictor
raise PdfReadError("Unsupported flatedecode predictor %r" %
predictor)
return data
def decode(data, decodeParms=None):
"""
:param data: flate-encoded data.
:param decodeParms: a dictionary of parameter values.
:return: the flate-decoded data.
:rtype: bytes
"""
data = decompress(data)
predictor = 1
if decodeParms:
try:
predictor = decodeParms.get("/Predictor", 1)
except AttributeError:
pass # Usually an array with a null object was read
# predictor 1 == no predictor
if predictor != 1:
# The /Columns param. has 1 as the default value; see ISO 32000,
# §7.4.4.3 LZWDecode and FlateDecode Parameters, Table 8
columns = decodeParms.get("/Columns", 1)
colors = decodeParms.get("/Colors", 8)
# PNG prediction:
if 10 <= predictor <= 15:
output = BytesIO(
) ### BytesIO au lieu de StringIO, car PIL attend des bytes (Pas de mon fait)
# PNG prediction can vary from row to row
row_length = columns * colors + 1
assert len(data) % row_length == 0
prev_rowdata = (0, ) * row_length
for row in range(len(data) // row_length):
rowdata = [
pypdfOrd(x)
for x in data[(row * row_length):((row + 1) *
row_length)]
]
filterByte = rowdata[0]
if filterByte == 0:
pass
elif filterByte == 1:
for i in range(1, row_length):
if i <= 3: ### On vise hors de la limite gauche, donc 0 (as per les specs)
rowdata[i] = (rowdata[i] + 0) % 256
else:
rowdata[i] = (
rowdata[i] + rowdata[i - colors]
) % 256 ### Pas le byte d'avant ! Son homologue dans le pixel d'avant
### (TODO : vérifier si le bitdepth < 8, auquel cas il faut
### bel et bien prendre simplement le byte d'avant)
elif filterByte == 2:
for i in range(1, row_length):
rowdata[i] = (rowdata[i] + prev_rowdata[i]) % 256
elif filterByte == 3:
for i in range(1, row_length):
left = rowdata[i - 1] if i > 1 else 0
floor = math.floor(left + prev_rowdata[i]) / 2
rowdata[i] = (rowdata[i] + int(floor)) % 256
elif filterByte == 4:
for i in range(1, row_length):
left = rowdata[i - 1] if i > 1 else 0
up = prev_rowdata[i]
up_left = prev_rowdata[i - 1] if i > 1 else 0
paeth = paethPredictor(left, up, up_left)
rowdata[i] = (rowdata[i] + paeth) % 256
else:
# Unsupported PNG filter
raise PdfReadError("Unsupported PNG filter %r" %
filterByte)
prev_rowdata = rowdata
for d in rowdata[
1:]: ### Ignorer le premier bit de chaque scanline
if version_info < (3, 0):
output.write(chr(d))
else:
output.write(bytes([d]))
data = output.getvalue()
else:
# unsupported predictor
raise PdfReadError("Unsupported flatedecode predictor %r" %
predictor)
return data