def test04(self):
"""Test token seek and tell."""
filepath = os.path.join(TokenStreamTest.path, 'obj_stream3.dat')
with open(filepath, 'rb') as f:
tk = TokenStream(filepath, f)
tok = tk.next_token()
self.assertEqual(EToken.INTEGER, tok.type)
self.assertEqual(98, tok.data)
# Memorize position after doing a first next_token(), this works
xpos = tk.tell()
tok = tk.next_token()
self.assertEqual(EToken.INTEGER, tok.type)
self.assertEqual(73, tok.data)
tok = tk.next_token()
self.assertEqual(EToken.INTEGER, tok.type)
self.assertEqual(5, tok.data)
tk.seek(xpos)
tok = tk.next_token()
self.assertEqual(EToken.INTEGER, tok.type)
self.assertEqual(73, tok.data)
def test05(self):
"""Test token seek and tell."""
filepath = os.path.join(TokenStreamTest.path, 'obj_stream3.dat')
with open(filepath, 'rb') as f:
tk = TokenStream(filepath, f)
# Memorize position at the beginning, this bugs
xpos = tk.tell()
# tk.seek(0)
# print(f'test05: seek(0): cc="{chr(tk.cc)}", bf.s_pos={tk.bf.s_pos}')
# tk.seek(1)
# print(f'test05: seek(1), cc="{chr(tk.cc)}", bf.s_pos={tk.bf.s_pos}')
tok = tk.next_token()
self.assertEqual(EToken.INTEGER, tok.type)
self.assertEqual(98, tok.data)
tok = tk.next_token()
self.assertEqual(EToken.INTEGER, tok.type)
self.assertEqual(73, tok.data)
tok = tk.next_token()
self.assertEqual(EToken.INTEGER, tok.type)
self.assertEqual(5, tok.data)
tok = tk.next_token()
self.assertEqual(EToken.INTEGER, tok.type)
self.assertEqual(19, tok.data)
pos2 = tk.tell()
# Go back
tk.seek(xpos)
tok = tk.next_token()
self.assertEqual(EToken.INTEGER, tok.type)
self.assertEqual(98, tok.data)
tok = tk.next_token()
self.assertEqual(EToken.INTEGER, tok.type)
self.assertEqual(73, tok.data)
# Move forward
tk.seek(pos2)
tok = tk.next_token()
self.assertEqual(EToken.INTEGER, tok.type)
self.assertEqual(18, tok.data)
tok = tk.next_token()
self.assertEqual(EToken.INTEGER, tok.type)
self.assertEqual(33, tok.data)
tok = tk.next_token()
self.assertEqual(EToken.INTEGER, tok.type)
self.assertEqual(45, tok.data)
tok = tk.next_token()
self.assertEqual(EToken.INTEGER, tok.type)
self.assertEqual(66, tok.data)
tok = tk.next_token()
self.assertEqual(EToken.INTEGER, tok.type)
self.assertEqual(13, tok.data)
tok = tk.next_token()
self.assertEqual(EToken.INTEGER, tok.type)
self.assertEqual(2, tok.data)
tok = tk.next_token()
self.assertEqual(EToken.OBJ_REF, tok.type)
class ObjectStream:
# Initializer
def __init__(self, filepath, f):
self.tk = TokenStream(filepath, f)
self.f = f
self.tok = self.tk.next_token()
# The xref table will be a property of the object stream ?
def seek(self, offset):
self.tk.seek(offset)
# Normal init
self.tok = self.tk.next_token()
#---------------------------------------------------------------------------
# get_indirect_obj_def
#---------------------------------------------------------------------------
def get_indirect_obj_def(self):
"""Found the opening OBJECT_BEGIN token, now get the entire object."""
# self.tok has an EToken.OBJECT_BEGIN, parse the following tokens.
# Return is done with the closing token (already analyzed) in self.tok.
tok = self.tok
# Get the defined (internal) object
self.tok = self.tk.next_token()
if tok.type == EToken.EOF:
return PdfObject(EObject.EOF)
elif tok.type == EToken.ERROR:
return PdfObject(EObject.ERROR)
# Get the defined (internal) object
obj = self.next_object()
if obj.type in [EObject.ERROR, EObject.EOF]:
return obj
# self.tok holds the next token, read but not yet analyzed
tok = self.tok
# Ignore any end-if-line marker
if tok.type in [EToken.CR, EToken.LF, EToken.CRLF]:
tok = self.tk.next_token()
if tok.type == EToken.EOF:
return PdfObject(EObject.EOF)
elif tok.type == EToken.ERROR:
return PdfObject(EObject.ERROR)
if tok.type == EToken.OBJECT_END:
return obj
#---------------------------------------------------------------------------
# get_array
#---------------------------------------------------------------------------
def get_array(self):
"""Found the opening ARRAY_BEGIN token, now get the entire array."""
# self.tok has an EToken.ARRAY_BEGIN, parse the following tokens.
# Return is done with the closing token (already analyzed) in self.tok.
# Prepare an array object
arr = []
# FIXME shouldn't I ignore end-of-line characters ?
tok = self.tk.next_token()
while True:
if tok.type == EToken.ARRAY_END:
# It's a python array, but the elements are PdfObjects
return PdfObject(EObject.ARRAY, arr)
if tok.type == EToken.ERROR:
return PdfObject(EObject.ERROR)
if tok.type == EToken.EOF:
return PdfObject(EObject.EOF)
# Ignore end-if-line markers
if tok.type in [EToken.CR, EToken.LF, EToken.CRLF]:
tok = self.tk.next_token()
continue
self.tok = tok
obj = self.next_object()
# self.tok holds the next token, read but not yet analyzed
if obj.type in [EObject.ERROR, EObject.EOF]:
return obj
# self.tok holds the next token, read but not yet analyzed
tok = self.tok
arr.append(obj)
#---------------------------------------------------------------------------
# get_dictionary
#---------------------------------------------------------------------------
def get_dictionary(self):
"""Found the opening DICT_BEGIN token, now get the entire dictionary."""
# self.tok has an EToken.DICT_BEGIN, parse the following tokens.
# Return is done with the closing token (already analyzed) in self.tok.
# Prepare a dictionary object
d = {}
tok = self.tk.next_token()
while True:
if tok.type == EToken.DICT_END:
self.tok = tok
# It's a python dictionary, but the values are PdfObjects
return PdfObject(EObject.DICTIONARY, d)
if tok.type == EToken.ERROR:
return PdfObject(EObject.ERROR)
if tok.type == EToken.EOF:
return PdfObject(EObject.EOF)
# Ignore end-if-line markers
if tok.type in [EToken.CR, EToken.LF, EToken.CRLF]:
tok = self.tk.next_token()
elif tok.type == EToken.NAME:
tok2 = self.tk.next_token()
self.tok = tok2
obj = self.next_object()
# FIXME: can any bytes object be decoded like this ?
# FIXME: I've lost the keys' original bytes object
d[tok.data.decode('unicode_escape')] = obj
# The next token is already stored in self.tok, but it hasn't
# been analyzed yet.
tok = self.tok
else:
return PdfObject(EObject.ERROR)
#---------------------------------------------------------------------------
# get_stream
#---------------------------------------------------------------------------
# FIXME define a proper stream class, with the dictionary in it
def get_stream(self, length):
"""Found the opening STREAM_BEGIN token, now get all the data."""
# self.tok has an EToken.STREAM_BEGIN, parse the following tokens.
# Return is done with the closing token (already analyzed) in self.tok.
# FIXME I need to stop testing EOF and ERROR after every single
# next_XXX() function call, use exceptions instead.
# Get the token that follows 'stream' (CRLF or LF)
tok = self.tk.next_token()
if tok.type == EToken.EOF:
return PdfObject(EObject.EOF)
# "The keyword stream that follows the stream dictionary shall be
# followed by an end-of-line marker consisting of either a CARRIAGE
# RETURN and a LINE FEED or just a LINE FEED, and not by a CARRIAGE
# RETURN alone". PDF spec, § 7.3.8.1, page 19
if tok.type not in [EToken.LF, EToken.CRLF]:
return PdfObject(EObject.ERROR)
# Get the token with the stream data
tok = self.tk.next_stream(length)
if tok.type == EToken.EOF:
return PdfObject(EObject.EOF)
s = tok.data
# "There should be an end-of-line marker after the data and before
# endstream; this marker shall not be included in the stream length".
# PDF spec, § 7.3.8.1, page 19
tok = self.tk.next_token()
if tok.type == EToken.EOF:
return PdfObject(EObject.EOF)
if tok.type not in [EToken.CR, EToken.LF, EToken.CRLF]:
return PdfObject(EObject.ERROR)
# Get the closing STREAM_END
tok = self.tk.next_token()
if tok.type == EToken.EOF:
return PdfObject(EObject.EOF)
if tok.type != EToken.STREAM_END:
return PdfObject(EObject.ERROR)
# Return the stream data object, with the closing _END token
return PdfObject(EObject.STREAM, data=s)
#---------------------------------------------------------------------------
# deflate_stream
#---------------------------------------------------------------------------
def deflate_stream(self, s, columns=None, predictor=None, W=None):
"""Decode stream s, encoded with flate, with predictor and W params."""
# s: original compressed data stream (stripped)
# collumns: integer
# predictor: integer with values in { 1, 2, 10-15 }
# W: python array of integers
# First, deflate the string
zd = zlib.decompress(s)
if not predictor:
# No DecodeParms, so we assume no predictor
# False means we have not done the un-predicting, just return zd
return False, zd
if predictor != 12:
print(f'Predictor value {predictor} not supported (currently only 12)')
return False, zd
# From https://forums.adobe.com/thread/664902: "Strip off the last 10
# characters of the string. This is the CRC and is unnecessary to
# extract the raw data". Not doing this, at this point.
# Sum up the column widths. For the example above [1 2 1] would be
# 4. This is one less than the number of bytes in each row.
n = sum(W) # n == 4
width = n+1
# Split the string into rows by the column width: sum+1, or in our
# example, 5.
# Is the uncompressed stream length a multiple of this width ?
if len(zd)%(width) == 0:
print(f'*** Uncompressed len(zd)={len(zd)}, width={width}'
+ f', {len(zd)}={len(zd)//(width)}*{width}')
else:
print(f'*** Uncompressed len(zd)={len(zd)}, width={width}'
+ ', not a multiple')
# zd is a bytes object
prev = [0]*width
nrows = len(zd)//(width) # 86
arr = []
for r in range(nrows): # 0..85
bs = ''
rowdata = [x for x in zd[r*width:(r+1)*width]] # array of ints
for i in range(1, width):
rowdata[i] = (rowdata[i] + prev[i]) % 256
bs += format(rowdata[i], '08b') # Convert to binary string
prev = rowdata # Update prev for next pass
# Split the string according to W
# print(f'{bs} len={len(bs)}')
begin = 0
end = 8*W[0]
type = int(bs[begin:end], 2)
begin = 8*W[0]
end = 8*(W[0] + W[1])
fld1 = int(bs[begin:end], 2)
begin = 8*(W[0] + W[1])
end = begin + 8*W[2]
fld2 = int(bs[begin:end], 2)
arr.append((type, fld1, fld2))
# True means we have done the un-predicting, so what we return is an
# array of 3-uples"
return True, arr
#---------------------------------------------------------------------------
# get_xref_section
#---------------------------------------------------------------------------
def get_xref_section(self):
"""Parse a cross reference section into an object"""
# self.tok has a EToken.XREF_SECTION, parse the following tokens.
# "Each cross-reference section shall begin with a line containing the
# keyword xref": this implies an end-of-line marker after 'xref'
tok = self.tk.next_token()
if tok.type not in [EToken.CR, EToken.LF, EToken.CRLF]:
self.tok = tok # FIXME this way, self.tok will be analyzed again
return PdfObject(EObject.ERROR)
# Loop over cross-reference subsections
self.xref_sec = XrefSection()
while True:
# Get a special token representing the sub-section header
tok = self.tk.get_subsection_header()
if tok.type == EToken.EOF:
return PdfObject(EObject.EOF)
if tok.type == EToken.ERROR:
return PdfObject(EObject.ERROR)
if tok.type == EToken.UNEXPECTED:
# Couldn't parse the line as a sub-section header, this means
# that the sub-section is over. The xref is stored as a
# property of this ObjectSTream, and it is also returned.
# State has been rolled back, so prepare to continue
self.tok = self.tk.next_token()
return PdfObject(EObject.XREF_SECTION, self.xref_sec)
# Sub-section header was successfully parsed
first_objn, entry_cnt = tok.data
# I'm assuming entry_cnt is not 0.
subs = XrefSubSection(first_objn, entry_cnt)
for i in range(entry_cnt):
# Get a special token representing a sub-section entry
tok = self.tk.get_subsection_entry()
if tok.type == EToken.EOF:
return PdfObject(EObject.EOF)
if tok.type == EToken.ERROR:
return PdfObject(EObject.ERROR)
subs.entries.append(tok.data)
# Finish off the this sub-section
self.xref_sec.sub_sections.append(subs)
#---------------------------------------------------------------------------
# get_cross_reference
#---------------------------------------------------------------------------
def get_cross_reference(self):
"""Parse a cross reference section into an object"""
# The current token from the stream should be either a XREF_SECTION
# (for a traditional cross_reference table) or an INTEGER, introducing
# an indirect object definition, for a cross-reference stream
# (available in PDF 1.5 and later)
tok = self.tok
if tok.type == EToken.EOF:
return PdfObject(EObject.EOF)
# Traditional
if tok.type == EToken.XREF_SECTION:
return self.get_xref_section()
# Available in PDF 1.5 and later
if tok.type == EToken.INTEGER:
obj = self.next_object()
if obj.type == EObject.IND_OBJ_DEF:
return obj
# Any other case is an error, because we were expecting to find a
# cross-reference table, modern or traditional.
return PdfObject(EObject.ERROR)
#---------------------------------------------------------------------------
# next_object
#---------------------------------------------------------------------------
def next_object(self):
"""Get the next object as a PdfObject."""
# Invariant: tok has been read from the stream, but not yet analyzed. It
# is stored (persisted in between calls) in self.tok. This means that
# every time control leaves this function (through return), it must
# read, but not analyze, the next token, and store it in self.tok.
tok = self.tok
# Ignore CRLF (why do I parse the tokens then ?)
while tok.type in [EToken.CR, EToken.LF, EToken.CRLF]:
tok = self.tok = self.tk.next_token()
# Have we reached EOF ?
if tok.type == EToken.EOF:
return PdfObject(EObject.EOF)
elif tok.type == EToken.ERROR:
return PdfObject(EObject.ERROR)
elif tok.type == EToken.VERSION_MARKER:
self.tok = self.tk.next_token()
return PdfObject(EObject.VERSION_MARKER, data=tok.data)
# Now analyze tok: is it a boolean ?
elif tok.type == EToken.TRUE:
self.tok = self.tk.next_token()
return PdfObject(EObject.BOOLEAN, True)
elif tok.type == EToken.FALSE:
self.tok = self.tk.next_token()
return PdfObject(EObject.BOOLEAN, False)
# Is it an integer number ?
elif tok.type == EToken.INTEGER:
# Attempt to find the longest match first. Object definitions and
# references are two integers plus another token, they must be
# parsed first, and if not found, then we'll settle for the simple
# integer.
# Lookahead 1 token. If we find another integer, keep looking.
# If we find an OBJECT_BEGIN, then we have an indirect object
# definition.
# If we find an OBJ_REF, then we have an indirect reference.
pos = self.tk.tell()
tok2 = self.tk.next_token()
if tok2.type == EToken.INTEGER:
# Keep looking
tok3 = self.tk.next_token()
if tok3.type == EToken.OBJECT_BEGIN:
# Start creating the object with the object number (from
# tok) and generation number (from tok2)
# Get the defined (internal) object
self.tok = tok3
obj = self.get_indirect_obj_def()
if obj.type in [EObject.ERROR, EObject.EOF]:
return obj
self.tok = self.tk.next_token()
return PdfObject(EObject.IND_OBJ_DEF,
data=dict(obj=obj, objn=tok.data, gen=tok2.data))
elif tok3.type == EToken.OBJ_REF:
# self.tk.next_token() # peeked tok2
# self.tk.next_token() # peeked tok3
self.tok = self.tk.next_token()
return PdfObject(EObject.IND_OBJ_REF,
data=dict(objn=tok.data, gen=tok2.data))
# Ignore tok2, we re-read it anyway
self.tk.seek(pos)
x = tok.data
self.tok = self.tk.next_token()
return PdfObject(EObject.INTEGER, x)
# Is it a real number ?
elif tok.type == EToken.REAL:
self.tok = self.tk.next_token()
return PdfObject(EObject.REAL, tok.data)
# Is it a string ?
elif tok.type in [EToken.LITERAL_STRING, EToken.HEX_STRING]:
self.tok = self.tk.next_token()
return PdfObject(EObject.STRING, tok.data) # bytearray
# Is it a name ?
elif tok.type == EToken.NAME:
self.tok = self.tk.next_token()
return PdfObject(EObject.NAME, tok.data) # bytearray
# Is it an array ?
elif tok.type == EToken.ARRAY_BEGIN:
# self.tok already has the right value, tok was taken from there
obj = self.get_array()
# self.tok == ARRAY_END
if obj.type in [EObject.ERROR, EObject.EOF]:
return obj
self.tok = self.tk.next_token()
return obj
# Is it a dictionary ? or a (dictionary, stream) couple ?
elif tok.type == EToken.DICT_BEGIN:
# self.tok already has the right value, tok was taken from there
obj = self.get_dictionary()
# self.tok == DICT_END
if obj.type in [EObject.ERROR, EObject.EOF]:
return obj
while True:
self.tok = self.tk.next_token()
if self.tok.type not in [EToken.CR, EToken.LF, EToken.CRLF]:
break
if self.tok.type != EToken.STREAM_BEGIN:
return obj # return the dict
# We have found a STREAM_BEGIN token, so 'obj' is the stream
# dictionary
# FIXME this may not be right. Length is given as an indirect
# object ref, we must have parsed all the xref tables at this point
# if we want to parse this stream.
o = obj.data['Length']
if o.type == EObject.INTEGER:
ln = o.data
elif o.type == EObject.IND_OBJ_REF:
ln = self.deref_object(o)
else:
return PdfObject(EObject.ERROR)
obj2 = self.get_stream(ln)
# FIXME use exceptions instead
if obj2.type in [EObject.ERROR, EObject.EOF]:
return obj2
self.tok = self.tk.next_token()
return PdfObject(EObject.COUPLE, data=(obj, obj2))
# Is it a xref section ?
elif tok.type == EToken.XREF_SECTION:
obj = self.get_xref_section()
# self.tok already holds the next token
return obj
# Is it a trailer ?
elif tok.type == EToken.TRAILER:
tok = self.tk.next_token()
# Ignore CRLF (why do I parse the tokens then ?)
while tok.type in [EToken.CR, EToken.LF, EToken.CRLF]:
tok = self.tk.next_token()
if tok.type != EToken.DICT_BEGIN:
# FIXME specify once and for all which token I want to see when
# an error has been detected. The question is "how do I recover
# from this error ?"
self.tok = self.tk.next_token()
return PdfObject(EObject.ERROR)
obj = self.get_dictionary()
self.tok = self.tk.next_token()
return PdfObject(EObject.TRAILER, data=obj)
elif tok.type == EToken.STARTXREF:
self.tok = self.tk.next_token()
return PdfObject(EObject.STARTXREF)
elif tok.type == EToken.EOF_MARKER:
self.tok = self.tk.next_token()
return PdfObject(EObject.EOF_MARKER)
# Is it a stream ? Wrong. Streams are preceded by a dictionary.
elif tok.type == EToken.STREAM_BEGIN:
return PdfObject(EObject.ERROR)
# Is it null ?
elif tok.type == EToken.NULL:
self.tok = self.tk.next_token()
return PdfObject(EObject.NULL)
# Nothing that was expected here
else:
self.tok = self.tk.next_token()
return PdfObject(EObject.ERROR)
#---------------------------------------------------------------------------
# deref_object - read an indirect object from the file
#---------------------------------------------------------------------------
def deref_object(self, o):
"""Find an object's definition from a reference."""
if o.type != EObject.IND_OBJ_REF:
print(f'Expecting an indirect object reference, got "{o.type}"'
+ ' instead')
return None
if not self.xref_sec:
return None
# Now use objn to search the xref table for the file offset where
# this catalog dictionary object can be found; seek the file to
# that offset, and do another ob.next_object()
# Catalog dictionary object is found at this offset, go there
entry = self.xref_sec.get_object(o.data['objn'], o.data['gen'])
if not entry:
return None
offset, _, _ = entry
self.seek(offset)
# Now read the next char, this will be the beginning of
# "6082 0 obj^M<</Metadata 6125 0 R ..." where 6082 is the objn
o = self.next_object()
if o.type != EObject.IND_OBJ_DEF:
print(f'Expecting an indirect object definition, got "{o.type}"'
+ ' instead')
return None
# The indirect object definition surrounds the object we want
return o.data['obj']
