#!/usr/bin/env python

# olefile (formerly OleFileIO_PL) version 0.42 2015-01-25
#
# Module to read/write Microsoft OLE2 files (also called Structured Storage or
# Microsoft Compound Document File Format), such as Microsoft Office 97-2003
# documents, Image Composer and FlashPix files, Outlook messages, ...
# This version is compatible with Python 2.6+ and 3.x
#
# Project website: http://www.decalage.info/olefile
#
# olefile is copyright (c) 2005-2015 Philippe Lagadec (http://www.decalage.info)
#
# olefile is based on the OleFileIO module from the PIL library v1.1.6
# See: http://www.pythonware.com/products/pil/index.htm
#
# The Python Imaging Library (PIL) is
# Copyright (c) 1997-2005 by Secret Labs AB
# Copyright (c) 1995-2005 by Fredrik Lundh
#
# See source code and LICENSE.txt for information on usage and redistribution.


# Since OleFileIO_PL v0.30, only Python 2.6+ and 3.x is supported
# This import enables print() as a function rather than a keyword
# (main requirement to be compatible with Python 3.x)
# The comment on the line below should be printed on Python 2.5 or older:
from __future__ import print_function   # This version of olefile requires Python 2.6+ or 3.x.


__author__  = "Philippe Lagadec"
__date__    = "2015-01-25"
__version__ = '0.42b'

#--- LICENSE ------------------------------------------------------------------

# olefile (formerly OleFileIO_PL) is copyright (c) 2005-2015 Philippe Lagadec
# (http://www.decalage.info)
#
# All rights reserved.
#
# Redistribution and use in source and binary forms, with or without modification,
# are permitted provided that the following conditions are met:
#
#  * Redistributions of source code must retain the above copyright notice, this
#    list of conditions and the following disclaimer.
#  * Redistributions in binary form must reproduce the above copyright notice,
#    this list of conditions and the following disclaimer in the documentation
#    and/or other materials provided with the distribution.
#
# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
# ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
# WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
# DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
# FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
# DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
# SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
# CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
# OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.

# ----------
# PIL License:
#
# olefile is based on source code from the OleFileIO module of the Python
# Imaging Library (PIL) published by Fredrik Lundh under the following license:

# The Python Imaging Library (PIL) is
#    Copyright (c) 1997-2005 by Secret Labs AB
#    Copyright (c) 1995-2005 by Fredrik Lundh
#
# By obtaining, using, and/or copying this software and/or its associated
# documentation, you agree that you have read, understood, and will comply with
# the following terms and conditions:
#
# Permission to use, copy, modify, and distribute this software and its
# associated documentation for any purpose and without fee is hereby granted,
# provided that the above copyright notice appears in all copies, and that both
# that copyright notice and this permission notice appear in supporting
# documentation, and that the name of Secret Labs AB or the author(s) not be used
# in advertising or publicity pertaining to distribution of the software
# without specific, written prior permission.
#
# SECRET LABS AB AND THE AUTHORS DISCLAIMS ALL WARRANTIES WITH REGARD TO THIS
# SOFTWARE, INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS.
# IN NO EVENT SHALL SECRET LABS AB OR THE AUTHORS BE LIABLE FOR ANY SPECIAL,
# INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM
# LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR
# OTHER TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR
# PERFORMANCE OF THIS SOFTWARE.

#-----------------------------------------------------------------------------
# CHANGELOG: (only olefile/OleFileIO_PL changes compared to PIL 1.1.6)
# 2005-05-11 v0.10 PL: - a few fixes for Python 2.4 compatibility
#                        (all changes flagged with [PL])
# 2006-02-22 v0.11 PL: - a few fixes for some Office 2003 documents which raise
#                        exceptions in _OleStream.__init__()
# 2006-06-09 v0.12 PL: - fixes for files above 6.8MB (DIFAT in loadfat)
#                      - added some constants
#                      - added header values checks
#                      - added some docstrings
#                      - getsect: bugfix in case sectors >512 bytes
#                      - getsect: added conformity checks
#                      - DEBUG_MODE constant to activate debug display
# 2007-09-04 v0.13 PL: - improved/translated (lots of) comments
#                      - updated license
#                      - converted tabs to 4 spaces
# 2007-11-19 v0.14 PL: - added OleFileIO._raise_defect() to adapt sensitivity
#                      - improved _unicode() to use Python 2.x unicode support
#                      - fixed bug in _OleDirectoryEntry
# 2007-11-25 v0.15 PL: - added safety checks to detect FAT loops
#                      - fixed _OleStream which didn't check stream size
#                      - added/improved many docstrings and comments
#                      - moved helper functions _unicode and _clsid out of
#                        OleFileIO class
#                      - improved OleFileIO._find() to add Unix path syntax
#                      - OleFileIO._find() is now case-insensitive
#                      - added get_type() and get_rootentry_name()
#                      - rewritten loaddirectory and _OleDirectoryEntry
# 2007-11-27 v0.16 PL: - added _OleDirectoryEntry.kids_dict
#                      - added detection of duplicate filenames in storages
#                      - added detection of duplicate references to streams
#                      - added get_size() and exists() to _OleDirectoryEntry
#                      - added isOleFile to check header before parsing
#                      - added __all__ list to control public keywords in pydoc
# 2007-12-04 v0.17 PL: - added _load_direntry to fix a bug in loaddirectory
#                      - improved _unicode(), added workarounds for Python <2.3
#                      - added set_debug_mode and -d option to set debug mode
#                      - fixed bugs in OleFileIO.open and _OleDirectoryEntry
#                      - added safety check in main for large or binary
#                        properties
#                      - allow size>0 for storages for some implementations
# 2007-12-05 v0.18 PL: - fixed several bugs in handling of FAT, MiniFAT and
#                        streams
#                      - added option '-c' in main to check all streams
# 2009-12-10 v0.19 PL: - bugfix for 32 bit arrays on 64 bits platforms
#                        (thanks to Ben G. and Martijn for reporting the bug)
# 2009-12-11 v0.20 PL: - bugfix in OleFileIO.open when filename is not plain str
# 2010-01-22 v0.21 PL: - added support for big-endian CPUs such as PowerPC Macs
# 2012-02-16 v0.22 PL: - fixed bug in getproperties, patch by chuckleberryfinn
#                        (https://bitbucket.org/decalage/olefileio_pl/issue/7)
#                      - added close method to OleFileIO (fixed issue #2)
# 2012-07-25 v0.23 PL: - added support for file-like objects (patch by mete0r_kr)
# 2013-05-05 v0.24 PL: - getproperties: added conversion from filetime to python
#                        datetime
#                      - main: displays properties with date format
#                      - new class OleMetadata to parse standard properties
#                      - added get_metadata method
# 2013-05-07 v0.24 PL: - a few improvements in OleMetadata
# 2013-05-24 v0.25 PL: - getproperties: option to not convert some timestamps
#                      - OleMetaData: total_edit_time is now a number of seconds,
#                        not a timestamp
#                      - getproperties: added support for VT_BOOL, VT_INT, V_UINT
#                      - getproperties: filter out null chars from strings
#                      - getproperties: raise non-fatal defects instead of
#                        exceptions when properties cannot be parsed properly
# 2013-05-27       PL: - getproperties: improved exception handling
#                      - _raise_defect: added option to set exception type
#                      - all non-fatal issues are now recorded, and displayed
#                        when run as a script
# 2013-07-11 v0.26 PL: - added methods to get modification and creation times
#                        of a directory entry or a storage/stream
#                      - fixed parsing of direntry timestamps
# 2013-07-24       PL: - new options in listdir to list storages and/or streams
# 2014-02-04 v0.30 PL: - upgraded code to support Python 3.x by Martin Panter
#                      - several fixes for Python 2.6 (xrange, MAGIC)
#                      - reused i32 from Pillow's _binary
# 2014-07-18 v0.31     - preliminary support for 4K sectors
# 2014-07-27 v0.31 PL: - a few improvements in OleFileIO.open (header parsing)
#                      - Fixed loadfat for large files with 4K sectors (issue #3)
# 2014-07-30 v0.32 PL: - added write_sect to write sectors to disk
#                      - added write_mode option to OleFileIO.__init__ and open
# 2014-07-31       PL: - fixed padding in write_sect for Python 3, added checks
#                      - added write_stream to write a stream to disk
# 2014-09-26 v0.40 PL: - renamed OleFileIO_PL to olefile
# 2014-11-09       NE: - added support for Jython (Niko Ehrenfeuchter)
# 2014-11-13 v0.41 PL: - improved isOleFile and OleFileIO.open to support OLE
#                        data in a string buffer and file-like objects.
# 2014-11-21       PL: - updated comments according to Pillow's commits
# 2015-01-24 v0.42 PL: - changed the default path name encoding from Latin-1
#                        to UTF-8 on Python 2.x (Unicode on Python 3.x)
#                      - added path_encoding option to override the default
#                      - fixed a bug in _list when a storage is empty

#-----------------------------------------------------------------------------
# TODO (for version 1.0):
# + get rid of print statements, to simplify Python 2.x and 3.x support
# + add is_stream and is_storage
# + remove leading and trailing slashes where a path is used
# + add functions path_list2str and path_str2list
# + fix how all the methods handle unicode str and/or bytes as arguments
# + add path attrib to _OleDirEntry, set it once and for all in init or
#   append_kids (then listdir/_list can be simplified)
# - TESTS with Linux, MacOSX, Python 1.5.2, various files, PIL, ...
# - add underscore to each private method, to avoid their display in
#   pydoc/epydoc documentation - Remove it for classes to be documented
# - replace all raised exceptions with _raise_defect (at least in OleFileIO)
# - merge code from _OleStream and OleFileIO.getsect to read sectors
#   (maybe add a class for FAT and MiniFAT ?)
# - add method to check all streams (follow sectors chains without storing all
#   stream in memory, and report anomalies)
# - use _OleDirectoryEntry.kids_dict to improve _find and _list ?
# - fix Unicode names handling (find some way to stay compatible with Py1.5.2)
#   => if possible avoid converting names to Latin-1
# - review DIFAT code: fix handling of DIFSECT blocks in FAT (not stop)
# - rewrite OleFileIO.getproperties
# - improve docstrings to show more sample uses
# - see also original notes and FIXME below
# - remove all obsolete FIXMEs
# - OleMetadata: fix version attrib according to
#   http://msdn.microsoft.com/en-us/library/dd945671%28v=office.12%29.aspx

# IDEAS:
# - in OleFileIO._open and _OleStream, use size=None instead of 0x7FFFFFFF for
#   streams with unknown size
# - use arrays of int instead of long integers for FAT/MiniFAT, to improve
#   performance and reduce memory usage ? (possible issue with values >2^31)
# - provide tests with unittest (may need write support to create samples)
# - move all debug code (and maybe dump methods) to a separate module, with
#   a class which inherits OleFileIO ?
# - fix docstrings to follow epydoc format
# - add support for big endian byte order ?
# - create a simple OLE explorer with wxPython

# FUTURE EVOLUTIONS to add write support:
# see issue #6 on Bitbucket:
# https://bitbucket.org/decalage/olefileio_pl/issue/6/improve-olefileio_pl-to-write-ole-files

#-----------------------------------------------------------------------------
# NOTES from PIL 1.1.6:

# History:
# 1997-01-20 fl   Created
# 1997-01-22 fl   Fixed 64-bit portability quirk
# 2003-09-09 fl   Fixed typo in OleFileIO.loadfat (noted by Daniel Haertle)
# 2004-02-29 fl   Changed long hex constants to signed integers
#
# Notes:
# FIXME: sort out sign problem (eliminate long hex constants)
# FIXME: change filename to use "a/b/c" instead of ["a", "b", "c"]
# FIXME: provide a glob mechanism function (using fnmatchcase)
#
# Literature:
#
# "FlashPix Format Specification, Appendix A", Kodak and Microsoft,
#  September 1996.
#
# Quotes:
#
# "If this document and functionality of the Software conflict,
#  the actual functionality of the Software represents the correct
#  functionality" -- Microsoft, in the OLE format specification

#------------------------------------------------------------------------------


import io
import sys
import struct
import array
import os.path
import datetime

#=== COMPATIBILITY WORKAROUNDS ================================================

# [PL] Define explicitly the public API to avoid private objects in pydoc:
#TODO: add more
# __all__ = ['OleFileIO', 'isOleFile', 'MAGIC']

# For Python 3.x, need to redefine long as int:
if str is not bytes:
    long = int

# Need to make sure we use xrange both on Python 2 and 3.x:
try:
    # on Python 2 we need xrange:
    iterrange = xrange
except:
    # no xrange, for Python 3 it was renamed as range:
    iterrange = range

# [PL] workaround to fix an issue with array item size on 64 bits systems:
if array.array('L').itemsize == 4:
    # on 32 bits platforms, long integers in an array are 32 bits:
    UINT32 = 'L'
elif array.array('I').itemsize == 4:
    # on 64 bits platforms, integers in an array are 32 bits:
    UINT32 = 'I'
elif array.array('i').itemsize == 4:
    # On 64 bit Jython, signed integers ('i') are the only way to store our 32
    # bit values in an array in a *somewhat* reasonable way, as the otherwise
    # perfectly suited 'H' (unsigned int, 32 bits) results in a completely
    # unusable behaviour. This is most likely caused by the fact that Java
    # doesn't have unsigned values, and thus Jython's "array" implementation,
    # which is based on "jarray", doesn't have them either.
    # NOTE: to trick Jython into converting the values it would normally
    # interpret as "signed" into "unsigned", a binary-and operation with
    # 0xFFFFFFFF can be used. This way it is possible to use the same comparing
    # operations on all platforms / implementations. The corresponding code
    # lines are flagged with a 'JYTHON-WORKAROUND' tag below.
    UINT32 = 'i'
else:
    raise ValueError('Need to fix a bug with 32 bit arrays, please contact author...')


# [PL] These workarounds were inspired from the Path module
# (see http://www.jorendorff.com/articles/python/path/)
try:
    basestring
except NameError:
    basestring = str

# [PL] Experimental setting: if True, OLE filenames will be kept in Unicode
# if False (default PIL behaviour), all filenames are converted to Latin-1.
KEEP_UNICODE_NAMES = True

if sys.version_info[0] < 3:
    # On Python 2.x, the default encoding for path names is UTF-8:
    DEFAULT_PATH_ENCODING = 'utf-8'
else:
    # On Python 3.x, the default encoding for path names is Unicode (None):
    DEFAULT_PATH_ENCODING = None


#=== DEBUGGING ===============================================================

#TODO: replace this by proper logging

# [PL] DEBUG display mode: False by default, use set_debug_mode() or "-d" on
# command line to change it.
DEBUG_MODE = False


def debug_print(msg):
    print(msg)


def debug_pass(msg):
    pass


debug = debug_pass


def set_debug_mode(debug_mode):
    """
    Set debug mode on or off, to control display of debugging messages.
    :param mode: True or False
    """
    global DEBUG_MODE, debug
    DEBUG_MODE = debug_mode
    if debug_mode:
        debug = debug_print
    else:
        debug = debug_pass


#=== CONSTANTS ===============================================================

# magic bytes that should be at the beginning of every OLE file:
MAGIC = b'\xD0\xCF\x11\xE0\xA1\xB1\x1A\xE1'

# [PL]: added constants for Sector IDs (from AAF specifications)
MAXREGSECT = 0xFFFFFFFA  # (-6) maximum SECT
DIFSECT    = 0xFFFFFFFC  # (-4) denotes a DIFAT sector in a FAT
FATSECT    = 0xFFFFFFFD  # (-3) denotes a FAT sector in a FAT
ENDOFCHAIN = 0xFFFFFFFE  # (-2) end of a virtual stream chain
FREESECT   = 0xFFFFFFFF  # (-1) unallocated sector

# [PL]: added constants for Directory Entry IDs (from AAF specifications)
MAXREGSID  = 0xFFFFFFFA  # (-6) maximum directory entry ID
NOSTREAM   = 0xFFFFFFFF  # (-1) unallocated directory entry

# [PL] object types in storage (from AAF specifications)
STGTY_EMPTY     = 0  # empty directory entry (according to OpenOffice.org doc)
STGTY_STORAGE   = 1  # element is a storage object
STGTY_STREAM    = 2  # element is a stream object
STGTY_LOCKBYTES = 3  # element is an ILockBytes object
STGTY_PROPERTY  = 4  # element is an IPropertyStorage object
STGTY_ROOT      = 5  # element is a root storage


#
# --------------------------------------------------------------------
# property types

VT_EMPTY = 0; VT_NULL = 1; VT_I2 = 2; VT_I4 = 3; VT_R4 = 4; VT_R8 = 5; VT_CY = 6;
VT_DATE = 7; VT_BSTR = 8; VT_DISPATCH = 9; VT_ERROR = 10; VT_BOOL = 11;
VT_VARIANT = 12; VT_UNKNOWN = 13; VT_DECIMAL = 14; VT_I1 = 16; VT_UI1 = 17;
VT_UI2 = 18; VT_UI4 = 19; VT_I8 = 20; VT_UI8 = 21; VT_INT = 22; VT_UINT = 23;
VT_VOID = 24; VT_HRESULT = 25; VT_PTR = 26; VT_SAFEARRAY = 27; VT_CARRAY = 28;
VT_USERDEFINED = 29; VT_LPSTR = 30; VT_LPWSTR = 31; VT_FILETIME = 64;
VT_BLOB = 65; VT_STREAM = 66; VT_STORAGE = 67; VT_STREAMED_OBJECT = 68;
VT_STORED_OBJECT = 69; VT_BLOB_OBJECT = 70; VT_CF = 71; VT_CLSID = 72;
VT_VECTOR = 0x1000;

# map property id to name (for debugging purposes)

VT = {}
for keyword, var in list(vars().items()):
    if keyword[:3] == "VT_":
        VT[var] = keyword

#
# --------------------------------------------------------------------
# Some common document types (root.clsid fields)

WORD_CLSID = "00020900-0000-0000-C000-000000000046"
#TODO: check Excel, PPT, ...

# [PL]: Defect levels to classify parsing errors - see OleFileIO._raise_defect()
DEFECT_UNSURE =    10   # a case which looks weird, but not sure it's a defect
DEFECT_POTENTIAL = 20   # a potential defect
DEFECT_INCORRECT = 30   # an error according to specifications, but parsing
                        # can go on
DEFECT_FATAL =     40   # an error which cannot be ignored, parsing is
                        # impossible

# Minimal size of an empty OLE file, with 512-bytes sectors = 1536 bytes
# (this is used in isOleFile and OleFile.open)
MINIMAL_OLEFILE_SIZE = 1536

# [PL] add useful constants to __all__:
# for key in list(vars().keys()):
#     if key.startswith('STGTY_') or key.startswith('DEFECT_'):
#         __all__.append(key)


#=== FUNCTIONS ===============================================================

def isOleFile(filename):
    """
    Test if a file is an OLE container (according to the magic bytes in its header).

    :param filename: string-like or file-like object, OLE file to parse

        - if filename is a string smaller than 1536 bytes, it is the path
          of the file to open. (bytes or unicode string)
        - if filename is a string longer than 1535 bytes, it is parsed
          as the content of an OLE file in memory. (bytes type only)
        - if filename is a file-like object (with read and seek methods),
          it is parsed as-is.

    :returns: True if OLE, False otherwise.
    """
    # check if filename is a string-like or file-like object:
    if hasattr(filename, 'read'):
        # file-like object: use it directly
        header = filename.read(len(MAGIC))
        # just in case, seek back to start of file:
        filename.seek(0)
    elif isinstance(filename, bytes) and len(filename) >= MINIMAL_OLEFILE_SIZE:
        # filename is a bytes string containing the OLE file to be parsed:
        header = filename[:len(MAGIC)]
    else:
        # string-like object: filename of file on disk
        header = open(filename, 'rb').read(len(MAGIC))
    if header == MAGIC:
        return True
    else:
        return False


if bytes is str:
    # version for Python 2.x
    def i8(c):
        return ord(c)
else:
    # version for Python 3.x
    def i8(c):
        return c if c.__class__ is int else c[0]


#TODO: replace i16 and i32 with more readable struct.unpack equivalent?

def i16(c, o = 0):
    """
    Converts a 2-bytes (16 bits) string to an integer.

    c: string containing bytes to convert
    o: offset of bytes to convert in string
    """
    return struct.unpack("<H", c[o:o+2])[0]


def i32(c, o = 0):
    """
    Converts a 4-bytes (32 bits) string to an integer.

    c: string containing bytes to convert
    o: offset of bytes to convert in string
    """
    return struct.unpack("<I", c[o:o+4])[0]


def _clsid(clsid):
    """
    Converts a CLSID to a human-readable string.

    :param clsid: string of length 16.
    """
    assert len(clsid) == 16
    # if clsid is only made of null bytes, return an empty string:
    # (PL: why not simply return the string with zeroes?)
    if not clsid.strip(b"\0"):
        return ""
    return (("%08X-%04X-%04X-%02X%02X-" + "%02X" * 6) %
            ((i32(clsid, 0), i16(clsid, 4), i16(clsid, 6)) +
            tuple(map(i8, clsid[8:16]))))


def filetime2datetime(filetime):
    """
    convert FILETIME (64 bits int) to Python datetime.datetime
    """
    # TODO: manage exception when microseconds is too large
    # inspired from http://code.activestate.com/recipes/511425-filetime-to-datetime/
    _FILETIME_null_date = datetime.datetime(1601, 1, 1, 0, 0, 0)
    #debug('timedelta days=%d' % (filetime//(10*1000000*3600*24)))
    return _FILETIME_null_date + datetime.timedelta(microseconds=filetime//10)


#=== CLASSES ==================================================================

class OleMetadata(object):
    """
    class to parse and store metadata from standard properties of OLE files.

    Available attributes:
    codepage, title, subject, author, keywords, comments, template,
    last_saved_by, revision_number, total_edit_time, last_printed, create_time,
    last_saved_time, num_pages, num_words, num_chars, thumbnail,
    creating_application, security, codepage_doc, category, presentation_target,
    bytes, lines, paragraphs, slides, notes, hidden_slides, mm_clips,
    scale_crop, heading_pairs, titles_of_parts, manager, company, links_dirty,
    chars_with_spaces, unused, shared_doc, link_base, hlinks, hlinks_changed,
    version, dig_sig, content_type, content_status, language, doc_version

    Note: an attribute is set to None when not present in the properties of the
    OLE file.

    References for SummaryInformation stream:
    - http://msdn.microsoft.com/en-us/library/dd942545.aspx
    - http://msdn.microsoft.com/en-us/library/dd925819%28v=office.12%29.aspx
    - http://msdn.microsoft.com/en-us/library/windows/desktop/aa380376%28v=vs.85%29.aspx
    - http://msdn.microsoft.com/en-us/library/aa372045.aspx
    - http://sedna-soft.de/summary-information-stream/
    - http://poi.apache.org/apidocs/org/apache/poi/hpsf/SummaryInformation.html

    References for DocumentSummaryInformation stream:
    - http://msdn.microsoft.com/en-us/library/dd945671%28v=office.12%29.aspx
    - http://msdn.microsoft.com/en-us/library/windows/desktop/aa380374%28v=vs.85%29.aspx
    - http://poi.apache.org/apidocs/org/apache/poi/hpsf/DocumentSummaryInformation.html

    new in version 0.25
    """

    # attribute names for SummaryInformation stream properties:
    # (ordered by property id, starting at 1)
    SUMMARY_ATTRIBS = ['codepage', 'title', 'subject', 'author', 'keywords', 'comments',
        'template', 'last_saved_by', 'revision_number', 'total_edit_time',
        'last_printed', 'create_time', 'last_saved_time', 'num_pages',
        'num_words', 'num_chars', 'thumbnail', 'creating_application',
        'security']

    # attribute names for DocumentSummaryInformation stream properties:
    # (ordered by property id, starting at 1)
    DOCSUM_ATTRIBS = ['codepage_doc', 'category', 'presentation_target', 'bytes', 'lines', 'paragraphs',
        'slides', 'notes', 'hidden_slides', 'mm_clips',
        'scale_crop', 'heading_pairs', 'titles_of_parts', 'manager',
        'company', 'links_dirty', 'chars_with_spaces', 'unused', 'shared_doc',
        'link_base', 'hlinks', 'hlinks_changed', 'version', 'dig_sig',
        'content_type', 'content_status', 'language', 'doc_version']

    def __init__(self):
        """
        Constructor for OleMetadata
        All attributes are set to None by default
        """
        # properties from SummaryInformation stream
        self.codepage = None
        self.title = None
        self.subject = None
        self.author = None
        self.keywords = None
        self.comments = None
        self.template = None
        self.last_saved_by = None
        self.revision_number = None
        self.total_edit_time = None
        self.last_printed = None
        self.create_time = None
        self.last_saved_time = None
        self.num_pages = None
        self.num_words = None
        self.num_chars = None
        self.thumbnail = None
        self.creating_application = None
        self.security = None
        # properties from DocumentSummaryInformation stream
        self.codepage_doc = None
        self.category = None
        self.presentation_target = None
        self.bytes = None
        self.lines = None
        self.paragraphs = None
        self.slides = None
        self.notes = None
        self.hidden_slides = None
        self.mm_clips = None
        self.scale_crop = None
        self.heading_pairs = None
        self.titles_of_parts = None
        self.manager = None
        self.company = None
        self.links_dirty = None
        self.chars_with_spaces = None
        self.unused = None
        self.shared_doc = None
        self.link_base = None
        self.hlinks = None
        self.hlinks_changed = None
        self.version = None
        self.dig_sig = None
        self.content_type = None
        self.content_status = None
        self.language = None
        self.doc_version = None

    def parse_properties(self, olefile):
        """
        Parse standard properties of an OLE file, from the streams
        "\x05SummaryInformation" and "\x05DocumentSummaryInformation",
        if present.
        Properties are converted to strings, integers or python datetime objects.
        If a property is not present, its value is set to None.
        """
        # first set all attributes to None:
        for attrib in (self.SUMMARY_ATTRIBS + self.DOCSUM_ATTRIBS):
            setattr(self, attrib, None)
        if olefile.exists("\x05SummaryInformation"):
            # get properties from the stream:
            # (converting timestamps to python datetime, except total_edit_time,
            # which is property #10)
            props = olefile.getproperties("\x05SummaryInformation",
                convert_time=True, no_conversion=[10])
            # store them into this object's attributes:
            for i in range(len(self.SUMMARY_ATTRIBS)):
                # ids for standards properties start at 0x01, until 0x13
                value = props.get(i+1, None)
                setattr(self, self.SUMMARY_ATTRIBS[i], value)
        if olefile.exists("\x05DocumentSummaryInformation"):
            # get properties from the stream:
            props = olefile.getproperties("\x05DocumentSummaryInformation",
                convert_time=True)
            # store them into this object's attributes:
            for i in range(len(self.DOCSUM_ATTRIBS)):
                # ids for standards properties start at 0x01, until 0x13
                value = props.get(i+1, None)
                setattr(self, self.DOCSUM_ATTRIBS[i], value)

    def dump(self):
        """
        Dump all metadata, for debugging purposes.
        """
        print('Properties from SummaryInformation stream:')
        for prop in self.SUMMARY_ATTRIBS:
            value = getattr(self, prop)
            print('- %s: %s' % (prop, repr(value)))
        print('Properties from DocumentSummaryInformation stream:')
        for prop in self.DOCSUM_ATTRIBS:
            value = getattr(self, prop)
            print('- %s: %s' % (prop, repr(value)))


#--- _OleStream ---------------------------------------------------------------

class _OleStream(io.BytesIO):
    """
    OLE2 Stream

    Returns a read-only file object which can be used to read
    the contents of a OLE stream (instance of the BytesIO class).
    To open a stream, use the openstream method in the OleFile class.

    This function can be used with either ordinary streams,
    or ministreams, depending on the offset, sectorsize, and
    fat table arguments.

    Attributes:

        - size: actual size of data stream, after it was opened.
    """

    # FIXME: should store the list of sects obtained by following
    # the fat chain, and load new sectors on demand instead of
    # loading it all in one go.

    def __init__(self, fp, sect, size, offset, sectorsize, fat, filesize):
        """
        Constructor for _OleStream class.

        :param fp: file object, the OLE container or the MiniFAT stream
        :param sect: sector index of first sector in the stream
        :param size: total size of the stream
        :param offset: offset in bytes for the first FAT or MiniFAT sector
        :param sectorsize: size of one sector
        :param fat: array/list of sector indexes (FAT or MiniFAT)
        :param filesize: size of OLE file (for debugging)
        :returns: a BytesIO instance containing the OLE stream
        """
        debug('_OleStream.__init__:')
        debug('  sect=%d (%X), size=%d, offset=%d, sectorsize=%d, len(fat)=%d, fp=%s'
            % (sect, sect, size, offset, sectorsize, len(fat), repr(fp)))
        # [PL] To detect malformed documents with FAT loops, we compute the
        # expected number of sectors in the stream:
        unknown_size = False
        if size == 0x7FFFFFFF:
            # this is the case when called from OleFileIO._open(), and stream
            # size is not known in advance (for example when reading the
            # Directory stream). Then we can only guess maximum size:
            size = len(fat)*sectorsize
            # and we keep a record that size was unknown:
            unknown_size = True
            debug('  stream with UNKNOWN SIZE')
        nb_sectors = (size + (sectorsize-1)) // sectorsize
        debug('nb_sectors = %d' % nb_sectors)
        # This number should (at least) be less than the total number of
        # sectors in the given FAT:
        if nb_sectors > len(fat):
            raise IOError('malformed OLE document, stream too large')
        # optimization(?): data is first a list of strings, and join() is called
        # at the end to concatenate all in one string.
        # (this may not be really useful with recent Python versions)
        data = []
        # if size is zero, then first sector index should be ENDOFCHAIN:
        if size == 0 and sect != ENDOFCHAIN:
            debug('size == 0 and sect != ENDOFCHAIN:')
            raise IOError('incorrect OLE sector index for empty stream')
        # [PL] A fixed-length for loop is used instead of an undefined while
        # loop to avoid DoS attacks:
        for i in range(nb_sectors):
            # Sector index may be ENDOFCHAIN, but only if size was unknown
            if sect == ENDOFCHAIN:
                if unknown_size:
                    break
                else:
                    # else this means that the stream is smaller than declared:
                    debug('sect=ENDOFCHAIN before expected size')
                    raise IOError('incomplete OLE stream')
            # sector index should be within FAT:
            if sect < 0 or sect >= len(fat):
                debug('sect=%d (%X) / len(fat)=%d' % (sect, sect, len(fat)))
                debug('i=%d / nb_sectors=%d' % (i, nb_sectors))
##                tmp_data = b"".join(data)
##                f = open('test_debug.bin', 'wb')
##                f.write(tmp_data)
##                f.close()
##                debug('data read so far: %d bytes' % len(tmp_data))
                raise IOError('incorrect OLE FAT, sector index out of range')
            #TODO: merge this code with OleFileIO.getsect() ?
            #TODO: check if this works with 4K sectors:
            try:
                fp.seek(offset + sectorsize * sect)
            except:
                debug('sect=%d, seek=%d, filesize=%d' %
                    (sect, offset+sectorsize*sect, filesize))
                raise IOError('OLE sector index out of range')
            sector_data = fp.read(sectorsize)
            # [PL] check if there was enough data:
            # Note: if sector is the last of the file, sometimes it is not a
            # complete sector (of 512 or 4K), so we may read less than
            # sectorsize.
            if len(sector_data) != sectorsize and sect != (len(fat)-1):
                debug('sect=%d / len(fat)=%d, seek=%d / filesize=%d, len read=%d' %
                    (sect, len(fat), offset+sectorsize*sect, filesize, len(sector_data)))
                debug('seek+len(read)=%d' % (offset+sectorsize*sect+len(sector_data)))
                raise IOError('incomplete OLE sector')
            data.append(sector_data)
            # jump to next sector in the FAT:
            try:
                sect = fat[sect] & 0xFFFFFFFF  # JYTHON-WORKAROUND
            except IndexError:
                # [PL] if pointer is out of the FAT an exception is raised
                raise IOError('incorrect OLE FAT, sector index out of range')
        # [PL] Last sector should be a "end of chain" marker:
        if sect != ENDOFCHAIN:
            raise IOError('incorrect last sector index in OLE stream')
        data = b"".join(data)
        # Data is truncated to the actual stream size:
        if len(data) >= size:
            data = data[:size]
            # actual stream size is stored for future use:
            self.size = size
        elif unknown_size:
            # actual stream size was not known, now we know the size of read
            # data:
            self.size = len(data)
        else:
            # read data is less than expected:
            debug('len(data)=%d, size=%d' % (len(data), size))
            raise IOError('OLE stream size is less than declared')
        # when all data is read in memory, BytesIO constructor is called
        io.BytesIO.__init__(self, data)
        # Then the _OleStream object can be used as a read-only file object.


#--- _OleDirectoryEntry -------------------------------------------------------

class _OleDirectoryEntry(object):

    """
    OLE2 Directory Entry
    """
    # [PL] parsing code moved from OleFileIO.loaddirectory

    # struct to parse directory entries:
    # <: little-endian byte order, standard sizes
    #    (note: this should guarantee that Q returns a 64 bits int)
    # 64s: string containing entry name in unicode (max 31 chars) + null char
    # H: uint16, number of bytes used in name buffer, including null = (len+1)*2
    # B: uint8, dir entry type (between 0 and 5)
    # B: uint8, color: 0=black, 1=red
    # I: uint32, index of left child node in the red-black tree, NOSTREAM if none
    # I: uint32, index of right child node in the red-black tree, NOSTREAM if none
    # I: uint32, index of child root node if it is a storage, else NOSTREAM
    # 16s: CLSID, unique identifier (only used if it is a storage)
    # I: uint32, user flags
    # Q (was 8s): uint64, creation timestamp or zero
    # Q (was 8s): uint64, modification timestamp or zero
    # I: uint32, SID of first sector if stream or ministream, SID of 1st sector
    #    of stream containing ministreams if root entry, 0 otherwise
    # I: uint32, total stream size in bytes if stream (low 32 bits), 0 otherwise
    # I: uint32, total stream size in bytes if stream (high 32 bits), 0 otherwise
    STRUCT_DIRENTRY = '<64sHBBIII16sIQQIII'
    # size of a directory entry: 128 bytes
    DIRENTRY_SIZE = 128
    assert struct.calcsize(STRUCT_DIRENTRY) == DIRENTRY_SIZE

    def __init__(self, entry, sid, olefile):
        """
        Constructor for an _OleDirectoryEntry object.
        Parses a 128-bytes entry from the OLE Directory stream.

        :param entry  : string (must be 128 bytes long)
        :param sid    : index of this directory entry in the OLE file directory
        :param olefile: OleFileIO containing this directory entry
        """
        self.sid = sid
        # ref to olefile is stored for future use
        self.olefile = olefile
        # kids is a list of children entries, if this entry is a storage:
        # (list of _OleDirectoryEntry objects)
        self.kids = []
        # kids_dict is a dictionary of children entries, indexed by their
        # name in lowercase: used to quickly find an entry, and to detect
        # duplicates
        self.kids_dict = {}
        # flag used to detect if the entry is referenced more than once in
        # directory:
        self.used = False
        # decode DirEntry
        (
            name,
            namelength,
            self.entry_type,
            self.color,
            self.sid_left,
            self.sid_right,
            self.sid_child,
            clsid,
            self.dwUserFlags,
            self.createTime,
            self.modifyTime,
            self.isectStart,
            sizeLow,
            sizeHigh
        ) = struct.unpack(_OleDirectoryEntry.STRUCT_DIRENTRY, entry)
        if self.entry_type not in [STGTY_ROOT, STGTY_STORAGE, STGTY_STREAM, STGTY_EMPTY]:
            olefile.raise_defect(DEFECT_INCORRECT, 'unhandled OLE storage type')
        # only first directory entry can (and should) be root:
        if self.entry_type == STGTY_ROOT and sid != 0:
            olefile.raise_defect(DEFECT_INCORRECT, 'duplicate OLE root entry')
        if sid == 0 and self.entry_type != STGTY_ROOT:
            olefile.raise_defect(DEFECT_INCORRECT, 'incorrect OLE root entry')
        #debug (struct.unpack(fmt_entry, entry[:len_entry]))
        # name should be at most 31 unicode characters + null character,
        # so 64 bytes in total (31*2 + 2):
        if namelength > 64:
            olefile.raise_defect(DEFECT_INCORRECT, 'incorrect DirEntry name length')
            # if exception not raised, namelength is set to the maximum value:
            namelength = 64
        # only characters without ending null char are kept:
        name = name[:(namelength-2)]
        #TODO: check if the name is actually followed by a null unicode character ([MS-CFB] 2.6.1)
        #TODO: check if the name does not contain forbidden characters:
        # [MS-CFB] 2.6.1: "The following characters are illegal and MUST NOT be part of the name: '/', '\', ':', '!'."
        # name is converted from UTF-16LE to the path encoding specified in the OleFileIO:
        self.name = olefile._decode_utf16_str(name)

        debug('DirEntry SID=%d: %s' % (self.sid, repr(self.name)))
        debug(' - type: %d' % self.entry_type)
        debug(' - sect: %d' % self.isectStart)
        debug(' - SID left: %d, right: %d, child: %d' % (self.sid_left,
            self.sid_right, self.sid_child))

        # sizeHigh is only used for 4K sectors, it should be zero for 512 bytes
        # sectors, BUT apparently some implementations set it as 0xFFFFFFFF, 1
        # or some other value so it cannot be raised as a defect in general:
        if olefile.sectorsize == 512:
            if sizeHigh != 0 and sizeHigh != 0xFFFFFFFF:
                debug('sectorsize=%d, sizeLow=%d, sizeHigh=%d (%X)' %
                    (olefile.sectorsize, sizeLow, sizeHigh, sizeHigh))
                olefile.raise_defect(DEFECT_UNSURE, 'incorrect OLE stream size')
            self.size = sizeLow
        else:
            self.size = sizeLow + (long(sizeHigh) << 32)
        debug(' - size: %d (sizeLow=%d, sizeHigh=%d)' % (self.size, sizeLow, sizeHigh))

        self.clsid = _clsid(clsid)
        # a storage should have a null size, BUT some implementations such as
        # Word 8 for Mac seem to allow non-null values => Potential defect:
        if self.entry_type == STGTY_STORAGE and self.size != 0:
            olefile.raise_defect(DEFECT_POTENTIAL, 'OLE storage with size>0')
        # check if stream is not already referenced elsewhere:
        if self.entry_type in (STGTY_ROOT, STGTY_STREAM) and self.size > 0:
            if self.size < olefile.minisectorcutoff \
            and self.entry_type == STGTY_STREAM: # only streams can be in MiniFAT
                # ministream object
                minifat = True
            else:
                minifat = False
            olefile._check_duplicate_stream(self.isectStart, minifat)

    def build_storage_tree(self):
        """
        Read and build the red-black tree attached to this _OleDirectoryEntry
        object, if it is a storage.
        Note that this method builds a tree of all subentries, so it should
        only be called for the root object once.
        """
        debug('build_storage_tree: SID=%d - %s - sid_child=%d'
            % (self.sid, repr(self.name), self.sid_child))
        if self.sid_child != NOSTREAM:
            # if child SID is not NOSTREAM, then this entry is a storage.
            # Let's walk through the tree of children to fill the kids list:
            self.append_kids(self.sid_child)

            # Note from OpenOffice documentation: the safest way is to
            # recreate the tree because some implementations may store broken
            # red-black trees...

            # in the OLE file, entries are sorted on (length, name).
            # for convenience, we sort them on name instead:
            # (see rich comparison methods in this class)
            self.kids.sort()

    def append_kids(self, child_sid):
        """
        Walk through red-black tree of children of this directory entry to add
        all of them to the kids list. (recursive method)

        :param child_sid : index of child directory entry to use, or None when called
            first time for the root. (only used during recursion)
        """
        # [PL] this method was added to use simple recursion instead of a complex
        # algorithm.
        # if this is not a storage or a leaf of the tree, nothing to do:
        if child_sid == NOSTREAM:
            return
        # check if child SID is in the proper range:
        if child_sid < 0 or child_sid >= len(self.olefile.direntries):
            self.olefile.raise_defect(DEFECT_FATAL, 'OLE DirEntry index out of range')
        # get child direntry:
        child = self.olefile._load_direntry(child_sid) #direntries[child_sid]
        debug('append_kids: child_sid=%d - %s - sid_left=%d, sid_right=%d, sid_child=%d'
            % (child.sid, repr(child.name), child.sid_left, child.sid_right, child.sid_child))
        # the directory entries are organized as a red-black tree.
        # (cf. Wikipedia for details)
        # First walk through left side of the tree:
        self.append_kids(child.sid_left)
        # Check if its name is not already used (case-insensitive):
        name_lower = child.name.lower()
        if name_lower in self.kids_dict:
            self.olefile.raise_defect(DEFECT_INCORRECT,
                "Duplicate filename in OLE storage")
        # Then the child_sid _OleDirectoryEntry object is appended to the
        # kids list and dictionary:
        self.kids.append(child)
        self.kids_dict[name_lower] = child
        # Check if kid was not already referenced in a storage:
        if child.used:
            self.olefile.raise_defect(DEFECT_INCORRECT,
                'OLE Entry referenced more than once')
        child.used = True
        # Finally walk through right side of the tree:
        self.append_kids(child.sid_right)
        # Afterwards build kid's own tree if it's also a storage:
        child.build_storage_tree()

    def __eq__(self, other):
        "Compare entries by name"
        return self.name == other.name

    def __lt__(self, other):
        "Compare entries by name"
        return self.name < other.name

    def __ne__(self, other):
        return not self.__eq__(other)

    def __le__(self, other):
        return self.__eq__(other) or self.__lt__(other)

    # Reflected __lt__() and __le__() will be used for __gt__() and __ge__()

    #TODO: replace by the same function as MS implementation ?
    # (order by name length first, then case-insensitive order)

    def dump(self, tab = 0):
        "Dump this entry, and all its subentries (for debug purposes only)"
        TYPES = ["(invalid)", "(storage)", "(stream)", "(lockbytes)",
                 "(property)", "(root)"]
        print(" "*tab + repr(self.name), TYPES[self.entry_type], end=' ')
        if self.entry_type in (STGTY_STREAM, STGTY_ROOT):
            print(self.size, "bytes", end=' ')
        print()
        if self.entry_type in (STGTY_STORAGE, STGTY_ROOT) and self.clsid:
            print(" "*tab + "{%s}" % self.clsid)

        for kid in self.kids:
            kid.dump(tab + 2)

    def getmtime(self):
        """
        Return modification time of a directory entry.

        :returns: None if modification time is null, a python datetime object
            otherwise (UTC timezone)

        new in version 0.26
        """
        if self.modifyTime == 0:
            return None
        return filetime2datetime(self.modifyTime)

    def getctime(self):
        """
        Return creation time of a directory entry.

        :returns: None if modification time is null, a python datetime object
            otherwise (UTC timezone)

        new in version 0.26
        """
        if self.createTime == 0:
            return None
        return filetime2datetime(self.createTime)


#--- OleFileIO ----------------------------------------------------------------

class OleFileIO(object):
    """
    OLE container object

    This class encapsulates the interface to an OLE 2 structured
    storage file.  Use the :py:meth:`~PIL.OleFileIO.OleFileIO.listdir` and
    :py:meth:`~PIL.OleFileIO.OleFileIO.openstream` methods to
    access the contents of this file.

    Object names are given as a list of strings, one for each subentry
    level.  The root entry should be omitted.  For example, the following
    code extracts all image streams from a Microsoft Image Composer file::

        ole = OleFileIO("fan.mic")

        for entry in ole.listdir():
            if entry[1:2] == "Image":
                fin = ole.openstream(entry)
                fout = open(entry[0:1], "wb")
                while True:
                    s = fin.read(8192)
                    if not s:
                        break
                    fout.write(s)

    You can use the viewer application provided with the Python Imaging
    Library to view the resulting files (which happens to be standard
    TIFF files).
    """

    def __init__(self, filename=None, raise_defects=DEFECT_FATAL,
                 write_mode=False, debug=False, path_encoding=DEFAULT_PATH_ENCODING):
        """
        Constructor for the OleFileIO class.

        :param filename: file to open.

            - if filename is a string smaller than 1536 bytes, it is the path
              of the file to open. (bytes or unicode string)
            - if filename is a string longer than 1535 bytes, it is parsed
              as the content of an OLE file in memory. (bytes type only)
            - if filename is a file-like object (with read, seek and tell methods),
              it is parsed as-is.

        :param raise_defects: minimal level for defects to be raised as exceptions.
            (use DEFECT_FATAL for a typical application, DEFECT_INCORRECT for a
            security-oriented application, see source code for details)

        :param write_mode: bool, if True the file is opened in read/write mode instead
            of read-only by default.

        :param debug: bool, set debug mode

        :param path_encoding: None or str, name of the codec to use for path
            names (streams and storages), or None for Unicode.
            Unicode by default on Python 3+, UTF-8 on Python 2.x.
            (new in olefile 0.42, was hardcoded to Latin-1 until olefile v0.41)
        """
        set_debug_mode(debug)
        # minimal level for defects to be raised as exceptions:
        self._raise_defects_level = raise_defects
        # list of defects/issues not raised as exceptions:
        # tuples of (exception type, message)
        self.parsing_issues = []
        self.write_mode = write_mode
        self.path_encoding = path_encoding
        self._filesize = None
        self.fp = None
        if filename:
            self.open(filename, write_mode=write_mode)

    def raise_defect(self, defect_level, message, exception_type=IOError):
        """
        This method should be called for any defect found during file parsing.
        It may raise an IOError exception according to the minimal level chosen
        for the OleFileIO object.

        :param defect_level: defect level, possible values are:

            - DEFECT_UNSURE    : a case which looks weird, but not sure it's a defect
            - DEFECT_POTENTIAL : a potential defect
            - DEFECT_INCORRECT : an error according to specifications, but parsing can go on
            - DEFECT_FATAL     : an error which cannot be ignored, parsing is impossible

        :param message: string describing the defect, used with raised exception.
        :param exception_type: exception class to be raised, IOError by default
        """
        # added by [PL]
        if defect_level >= self._raise_defects_level:
            raise exception_type(message)
        else:
            # just record the issue, no exception raised:
            self.parsing_issues.append((exception_type, message))

    def _decode_utf16_str(self, utf16_str, errors='replace'):
        """
        Decode a string encoded in UTF-16 LE format, as found in the OLE
        directory or in property streams. Return a string encoded
        according to the path_encoding specified for the OleFileIO object.

        :param utf16_str: bytes string encoded in UTF-16 LE format
        :param errors: str, see python documentation for str.decode()
        :return: str, encoded according to path_encoding
        """
        unicode_str = utf16_str.decode('UTF-16LE', errors)
        if self.path_encoding:
            # an encoding has been specified for path names:
            return unicode_str.encode(self.path_encoding, errors)
        else:
            # path_encoding=None, return the Unicode string as-is:
            return unicode_str

    def open(self, filename, write_mode=False):
        """
        Open an OLE2 file in read-only or read/write mode.
        Read and parse the header, FAT and directory.

        :param filename: string-like or file-like object, OLE file to parse

            - if filename is a string smaller than 1536 bytes, it is the path
              of the file to open. (bytes or unicode string)
            - if filename is a string longer than 1535 bytes, it is parsed
              as the content of an OLE file in memory. (bytes type only)
            - if filename is a file-like object (with read, seek and tell methods),
              it is parsed as-is.

        :param write_mode: bool, if True the file is opened in read/write mode instead
            of read-only by default. (ignored if filename is not a path)
        """
        self.write_mode = write_mode
        # [PL] check if filename is a string-like or file-like object:
        # (it is better to check for a read() method)
        if hasattr(filename, 'read'):
            #TODO: also check seek and tell methods?
            # file-like object: use it directly
            self.fp = filename
        elif isinstance(filename, bytes) and len(filename) >= MINIMAL_OLEFILE_SIZE:
            # filename is a bytes string containing the OLE file to be parsed:
            # convert it to BytesIO
            self.fp = io.BytesIO(filename)
        else:
            # string-like object: filename of file on disk
            if self.write_mode:
                # open file in mode 'read with update, binary'
                # According to https://docs.python.org/2/library/functions.html#open
                # 'w' would truncate the file, 'a' may only append on some Unixes
                mode = 'r+b'
            else:
                # read-only mode by default
                mode = 'rb'
            self.fp = open(filename, mode)
        # obtain the filesize by using seek and tell, which should work on most
        # file-like objects:
        #TODO: do it above, using getsize with filename when possible?
        #TODO: fix code to fail with clear exception when filesize cannot be obtained
        filesize = 0
        self.fp.seek(0, os.SEEK_END)
        try:
            filesize = self.fp.tell()
        finally:
            self.fp.seek(0)
        self._filesize = filesize

        # lists of streams in FAT and MiniFAT, to detect duplicate references
        # (list of indexes of first sectors of each stream)
        self._used_streams_fat = []
        self._used_streams_minifat = []

        header = self.fp.read(512)

        if len(header) != 512 or header[:8] != MAGIC:
            self.raise_defect(DEFECT_FATAL, "not an OLE2 structured storage file")

        # [PL] header structure according to AAF specifications:
        ##Header
        ##struct StructuredStorageHeader { // [offset from start (bytes), length (bytes)]
        ##BYTE _abSig[8]; // [00H,08] {0xd0, 0xcf, 0x11, 0xe0, 0xa1, 0xb1,
        ##                // 0x1a, 0xe1} for current version
        ##CLSID _clsid;   // [08H,16] reserved must be zero (WriteClassStg/
        ##                // GetClassFile uses root directory class id)
        ##USHORT _uMinorVersion; // [18H,02] minor version of the format: 33 is
        ##                       // written by reference implementation
        ##USHORT _uDllVersion;   // [1AH,02] major version of the dll/format: 3 for
        ##                       // 512-byte sectors, 4 for 4 KB sectors
        ##USHORT _uByteOrder;    // [1CH,02] 0xFFFE: indicates Intel byte-ordering
        ##USHORT _uSectorShift;  // [1EH,02] size of sectors in power-of-two;
        ##                       // typically 9 indicating 512-byte sectors
        ##USHORT _uMiniSectorShift; // [20H,02] size of mini-sectors in power-of-two;
        ##                          // typically 6 indicating 64-byte mini-sectors
        ##USHORT _usReserved; // [22H,02] reserved, must be zero
        ##ULONG _ulReserved1; // [24H,04] reserved, must be zero
        ##FSINDEX _csectDir; // [28H,04] must be zero for 512-byte sectors,
        ##                   // number of SECTs in directory chain for 4 KB
        ##                   // sectors
        ##FSINDEX _csectFat; // [2CH,04] number of SECTs in the FAT chain
        ##SECT _sectDirStart; // [30H,04] first SECT in the directory chain
        ##DFSIGNATURE _signature; // [34H,04] signature used for transactions; must
        ##                        // be zero. The reference implementation
        ##                        // does not support transactions
        ##ULONG _ulMiniSectorCutoff; // [38H,04] maximum size for a mini stream;
        ##                           // typically 4096 bytes
        ##SECT _sectMiniFatStart; // [3CH,04] first SECT in the MiniFAT chain
        ##FSINDEX _csectMiniFat; // [40H,04] number of SECTs in the MiniFAT chain
        ##SECT _sectDifStart; // [44H,04] first SECT in the DIFAT chain
        ##FSINDEX _csectDif; // [48H,04] number of SECTs in the DIFAT chain
        ##SECT _sectFat[109]; // [4CH,436] the SECTs of first 109 FAT sectors
        ##};

        # [PL] header decoding:
        # '<' indicates little-endian byte ordering for Intel (cf. struct module help)
        fmt_header = '<8s16sHHHHHHLLLLLLLLLL'
        header_size = struct.calcsize(fmt_header)
        debug("fmt_header size = %d, +FAT = %d" % (header_size, header_size + 109*4))
        header1 = header[:header_size]
        (
            self.Sig,
            self.clsid,
            self.MinorVersion,
            self.DllVersion,
            self.ByteOrder,
            self.SectorShift,
            self.MiniSectorShift,
            self.Reserved, self.Reserved1,
            self.csectDir,
            self.csectFat,
            self.sectDirStart,
            self.signature,
            self.MiniSectorCutoff,
            self.MiniFatStart,
            self.csectMiniFat,
            self.sectDifStart,
            self.csectDif
        ) = struct.unpack(fmt_header, header1)
        debug(struct.unpack(fmt_header,    header1))

        if self.Sig != MAGIC:
            # OLE signature should always be present
            self.raise_defect(DEFECT_FATAL, "incorrect OLE signature")
        if self.clsid != bytearray(16):
            # according to AAF specs, CLSID should always be zero
            self.raise_defect(DEFECT_INCORRECT, "incorrect CLSID in OLE header")
        debug("MinorVersion = %d" % self.MinorVersion)
        debug("DllVersion   = %d" % self.DllVersion)
        if self.DllVersion not in [3, 4]:
            # version 3: usual format, 512 bytes per sector
            # version 4: large format, 4K per sector
            self.raise_defect(DEFECT_INCORRECT, "incorrect DllVersion in OLE header")
        debug("ByteOrder    = %X" % self.ByteOrder)
        if self.ByteOrder != 0xFFFE:
            # For now only common little-endian documents are handled correctly
            self.raise_defect(DEFECT_FATAL, "incorrect ByteOrder in OLE header")
            # TODO: add big-endian support for documents created on Mac ?
            # But according to [MS-CFB] ? v20140502, ByteOrder MUST be 0xFFFE.
        self.SectorSize = 2**self.SectorShift
        debug("SectorSize   = %d" % self.SectorSize)
        if self.SectorSize not in [512, 4096]:
            self.raise_defect(DEFECT_INCORRECT, "incorrect SectorSize in OLE header")
        if (self.DllVersion == 3 and self.SectorSize != 512) \
        or (self.DllVersion == 4 and self.SectorSize != 4096):
            self.raise_defect(DEFECT_INCORRECT, "SectorSize does not match DllVersion in OLE header")
        self.MiniSectorSize = 2**self.MiniSectorShift
        debug("MiniSectorSize   = %d" % self.MiniSectorSize)
        if self.MiniSectorSize not in [64]:
            self.raise_defect(DEFECT_INCORRECT, "incorrect MiniSectorSize in OLE header")
        if self.Reserved != 0 or self.Reserved1 != 0:
            self.raise_defect(DEFECT_INCORRECT, "incorrect OLE header (non-null reserved bytes)")
        debug("csectDir     = %d" % self.csectDir)
        # Number of directory sectors (only allowed if DllVersion != 3)
        if self.SectorSize == 512 and self.csectDir != 0:
            self.raise_defect(DEFECT_INCORRECT, "incorrect csectDir in OLE header")
        debug("csectFat     = %d" % self.csectFat)
        # csectFat = number of FAT sectors in the file
        debug("sectDirStart = %X" % self.sectDirStart)
        # sectDirStart = 1st sector containing the directory
        debug("signature    = %d" % self.signature)
        # Signature should be zero, BUT some implementations do not follow this
        # rule => only a potential defect:
        # (according to MS-CFB, may be != 0 for applications supporting file
        # transactions)
        if self.signature != 0:
            self.raise_defect(DEFECT_POTENTIAL, "incorrect OLE header (signature>0)")
        debug("MiniSectorCutoff = %d" % self.MiniSectorCutoff)
        # MS-CFB: This integer field MUST be set to 0x00001000. This field
        # specifies the maximum size of a user-defined data stream allocated
        # from the mini FAT and mini stream, and that cutoff is 4096 bytes.
        # Any user-defined data stream larger than or equal to this cutoff size
        # must be allocated as normal sectors from the FAT.
        if self.MiniSectorCutoff != 0x1000:
            self.raise_defect(DEFECT_INCORRECT, "incorrect MiniSectorCutoff in OLE header")
        debug("MiniFatStart     = %X" % self.MiniFatStart)
        debug("csectMiniFat     = %d" % self.csectMiniFat)
        debug("sectDifStart     = %X" % self.sectDifStart)
        debug("csectDif         = %d" % self.csectDif)

        # calculate the number of sectors in the file
        # (-1 because header doesn't count)
        self.nb_sect = ((filesize + self.SectorSize-1) // self.SectorSize) - 1
        debug("Number of sectors in the file: %d" % self.nb_sect)
        #TODO: change this test, because an OLE file MAY contain other data
        # after the last sector.

        # file clsid
        self.clsid = _clsid(header[8:24])

        #TODO: remove redundant attributes, and fix the code which uses them?
        self.sectorsize = self.SectorSize #1 << i16(header, 30)
        self.minisectorsize = self.MiniSectorSize  #1 << i16(header, 32)
        self.minisectorcutoff = self.MiniSectorCutoff # i32(header, 56)

        # check known streams for duplicate references (these are always in FAT,
        # never in MiniFAT):
        self._check_duplicate_stream(self.sectDirStart)
        # check MiniFAT only if it is not empty:
        if self.csectMiniFat:
            self._check_duplicate_stream(self.MiniFatStart)
        # check DIFAT only if it is not empty:
        if self.csectDif:
            self._check_duplicate_stream(self.sectDifStart)

        # Load file allocation tables
        self.loadfat(header)
        # Load directory.  This sets both the direntries list (ordered by sid)
        # and the root (ordered by hierarchy) members.
        self.loaddirectory(self.sectDirStart)#i32(header, 48))
        self.ministream = None
        self.minifatsect = self.MiniFatStart #i32(header, 60)

    def close(self):
        """
        close the OLE file, to release the file object
        """
        self.fp.close()

    def _check_duplicate_stream(self, first_sect, minifat=False):
        """
        Checks if a stream has not been already referenced elsewhere.
        This method should only be called once for each known stream, and only
        if stream size is not null.

        :param first_sect: int, index of first sector of the stream in FAT
        :param minifat: bool, if True, stream is located in the MiniFAT, else in the FAT
        """
        if minifat:
            debug('_check_duplicate_stream: sect=%d in MiniFAT' % first_sect)
            used_streams = self._used_streams_minifat
        else:
            debug('_check_duplicate_stream: sect=%d in FAT' % first_sect)
            # some values can be safely ignored (not a real stream):
            if first_sect in (DIFSECT, FATSECT, ENDOFCHAIN, FREESECT):
                return
            used_streams = self._used_streams_fat
        #TODO: would it be more efficient using a dict or hash values, instead
        #      of a list of long ?
        if first_sect in used_streams:
            self.raise_defect(DEFECT_INCORRECT, 'Stream referenced twice')
        else:
            used_streams.append(first_sect)

    def dumpfat(self, fat, firstindex=0):
        "Displays a part of FAT in human-readable form for debugging purpose"
        # [PL] added only for debug
        if not DEBUG_MODE:
            return
        # dictionary to convert special FAT values in human-readable strings
        VPL = 8 # values per line (8+1 * 8+1 = 81)
        fatnames = {
            FREESECT:   "..free..",
            ENDOFCHAIN: "[ END. ]",
            FATSECT:    "FATSECT ",
            DIFSECT:    "DIFSECT "
            }
        nbsect = len(fat)
        nlines = (nbsect+VPL-1)//VPL
        print("index", end=" ")
        for i in range(VPL):
            print("%8X" % i, end=" ")
        print()
        for l in range(nlines):
            index = l*VPL
            print("%8X:" % (firstindex+index), end=" ")
            for i in range(index, index+VPL):
                if i >= nbsect:
                    break
                sect = fat[i]
                aux = sect & 0xFFFFFFFF  # JYTHON-WORKAROUND
                if aux in fatnames:
                    name = fatnames[aux]
                else:
                    if sect == i+1:
                        name = "    --->"
                    else:
                        name = "%8X" % sect
                print(name, end=" ")
            print()

    def dumpsect(self, sector, firstindex=0):
        "Displays a sector in a human-readable form, for debugging purpose."
        if not DEBUG_MODE:
            return
        VPL = 8  # number of values per line (8+1 * 8+1 = 81)
        tab = array.array(UINT32, sector)
        if sys.byteorder == 'big':
            tab.byteswap()
        nbsect = len(tab)
        nlines = (nbsect+VPL-1)//VPL
        print("index", end=" ")
        for i in range(VPL):
            print("%8X" % i, end=" ")
        print()
        for l in range(nlines):
            index = l*VPL
            print("%8X:" % (firstindex+index), end=" ")
            for i in range(index, index+VPL):
                if i >= nbsect:
                    break
                sect = tab[i]
                name = "%8X" % sect
                print(name, end=" ")
            print()

    def sect2array(self, sect):
        """
        convert a sector to an array of 32 bits unsigned integers,
        swapping bytes on big endian CPUs such as PowerPC (old Macs)
        """
        a = array.array(UINT32, sect)
        # if CPU is big endian, swap bytes:
        if sys.byteorder == 'big':
            a.byteswap()
        return a

    def loadfat_sect(self, sect):
        """
        Adds the indexes of the given sector to the FAT

        :param sect: string containing the first FAT sector, or array of long integers
        :returns: index of last FAT sector.
        """
        # a FAT sector is an array of ulong integers.
        if isinstance(sect, array.array):
            # if sect is already an array it is directly used
            fat1 = sect
        else:
            # if it's a raw sector, it is parsed in an array
            fat1 = self.sect2array(sect)
            self.dumpsect(sect)
        # The FAT is a sector chain starting at the first index of itself.
        for isect in fat1:
            isect = isect & 0xFFFFFFFF  # JYTHON-WORKAROUND
            debug("isect = %X" % isect)
            if isect == ENDOFCHAIN or isect == FREESECT:
                # the end of the sector chain has been reached
                debug("found end of sector chain")
                break
            # read the FAT sector
            s = self.getsect(isect)
            # parse it as an array of 32 bits integers, and add it to the
            # global FAT array
            nextfat = self.sect2array(s)
            self.fat = self.fat + nextfat
        return isect

    def loadfat(self, header):
        """
        Load the FAT table.
        """
        # The 1st sector of the file contains sector numbers for the first 109
        # FAT sectors, right after the header which is 76 bytes long.
        # (always 109, whatever the sector size: 512 bytes = 76+4*109)
        # Additional sectors are described by DIF blocks

        sect = header[76:512]
        debug("len(sect)=%d, so %d integers" % (len(sect), len(sect)//4))
        #fat    = []
        # [PL] FAT is an array of 32 bits unsigned ints, it's more effective
        # to use an array than a list in Python.
        # It's initialized as empty first:
        self.fat = array.array(UINT32)
        self.loadfat_sect(sect)
        #self.dumpfat(self.fat)
##      for i in range(0, len(sect), 4):
##          ix = i32(sect, i)
##          # [PL] if ix == -2 or ix == -1: # ix == 0xFFFFFFFE or ix == 0xFFFFFFFF:
##          if ix == 0xFFFFFFFE or ix == 0xFFFFFFFF:
##              break
##          s = self.getsect(ix)
##          #fat    = fat + [i32(s, i) for i in range(0, len(s), 4)]
##          fat = fat + array.array(UINT32, s)
        if self.csectDif != 0:
            # [PL] There's a DIFAT because file is larger than 6.8MB
            # some checks just in case:
            if self.csectFat <= 109:
                # there must be at least 109 blocks in header and the rest in
                # DIFAT, so number of sectors must be >109.
                self.raise_defect(DEFECT_INCORRECT, 'incorrect DIFAT, not enough sectors')
            if self.sectDifStart >= self.nb_sect:
                # initial DIFAT block index must be valid
                self.raise_defect(DEFECT_FATAL, 'incorrect DIFAT, first index out of range')
            debug("DIFAT analysis...")
            # We compute the necessary number of DIFAT sectors :
            # Number of pointers per DIFAT sector = (sectorsize/4)-1
            # (-1 because the last pointer is the next DIFAT sector number)
            nb_difat_sectors = (self.sectorsize//4)-1
            # (if 512 bytes: each DIFAT sector = 127 pointers + 1 towards next DIFAT sector)
            nb_difat = (self.csectFat-109 + nb_difat_sectors-1)//nb_difat_sectors
            debug("nb_difat = %d" % nb_difat)
            if self.csectDif != nb_difat:
                raise IOError('incorrect DIFAT')
            isect_difat = self.sectDifStart
            for i in iterrange(nb_difat):
                debug("DIFAT block %d, sector %X" % (i, isect_difat))
                #TODO: check if corresponding FAT SID = DIFSECT
                sector_difat = self.getsect(isect_difat)
                difat = self.sect2array(sector_difat)
                self.dumpsect(sector_difat)
                self.loadfat_sect(difat[:nb_difat_sectors])
                # last DIFAT pointer is next DIFAT sector:
                isect_difat = difat[nb_difat_sectors]
                debug("next DIFAT sector: %X" % isect_difat)
            # checks:
            if isect_difat not in [ENDOFCHAIN, FREESECT]:
                # last DIFAT pointer value must be ENDOFCHAIN or FREESECT
                raise IOError('incorrect end of DIFAT')
##          if len(self.fat) != self.csectFat:
##              # FAT should contain csectFat blocks
##              print("FAT length: %d instead of %d" % (len(self.fat), self.csectFat))
##              raise IOError('incorrect DIFAT')
        # since FAT is read from fixed-size sectors, it may contain more values
        # than the actual number of sectors in the file.
        # Keep only the relevant sector indexes:
        if len(self.fat) > self.nb_sect:
            debug('len(fat)=%d, shrunk to nb_sect=%d' % (len(self.fat), self.nb_sect))
            self.fat = self.fat[:self.nb_sect]
        debug('\nFAT:')
        self.dumpfat(self.fat)

    def loadminifat(self):
        """
        Load the MiniFAT table.
        """
        # MiniFAT is stored in a standard  sub-stream, pointed to by a header
        # field.
        # NOTE: there are two sizes to take into account for this stream:
        # 1) Stream size is calculated according to the number of sectors
        #    declared in the OLE header. This allocated stream may be more than
        #    needed to store the actual sector indexes.
        # (self.csectMiniFat is the number of sectors of size self.SectorSize)
        stream_size = self.csectMiniFat * self.SectorSize
        # 2) Actually used size is calculated by dividing the MiniStream size
        #    (given by root entry size) by the size of mini sectors, *4 for
        #    32 bits indexes:
        nb_minisectors = (self.root.size + self.MiniSectorSize-1) // self.MiniSectorSize
        used_size = nb_minisectors * 4
        debug('loadminifat(): minifatsect=%d, nb FAT sectors=%d, used_size=%d, stream_size=%d, nb MiniSectors=%d' %
            (self.minifatsect, self.csectMiniFat, used_size, stream_size, nb_minisectors))
        if used_size > stream_size:
            # This is not really a problem, but may indicate a wrong implementation:
            self.raise_defect(DEFECT_INCORRECT, 'OLE MiniStream is larger than MiniFAT')
        # In any case, first read stream_size:
        s = self._open(self.minifatsect, stream_size, force_FAT=True).read()
        # [PL] Old code replaced by an array:
        # self.minifat = [i32(s, i) for i in range(0, len(s), 4)]
        self.minifat = self.sect2array(s)
        # Then shrink the array to used size, to avoid indexes out of MiniStream:
        debug('MiniFAT shrunk from %d to %d sectors' % (len(self.minifat), nb_minisectors))
        self.minifat = self.minifat[:nb_minisectors]
        debug('loadminifat(): len=%d' % len(self.minifat))
        debug('\nMiniFAT:')
        self.dumpfat(self.minifat)

    def getsect(self, sect):
        """
        Read given sector from file on disk.

        :param sect: int, sector index
        :returns: a string containing the sector data.
        """
        # From [MS-CFB]: A sector number can be converted into a byte offset
        # into the file by using the following formula:
        # (sector number + 1) x Sector Size.
        # This implies that sector #0 of the file begins at byte offset Sector
        # Size, not at 0.

        # [PL] the original code in PIL was wrong when sectors are 4KB instead of
        # 512 bytes:
        # self.fp.seek(512 + self.sectorsize * sect)
        # [PL]: added safety checks:
        # print("getsect(%X)" % sect)
        try:
            self.fp.seek(self.sectorsize * (sect+1))
        except:
            debug('getsect(): sect=%X, seek=%d, filesize=%d' %
                (sect, self.sectorsize*(sect+1), self._filesize))
            self.raise_defect(DEFECT_FATAL, 'OLE sector index out of range')
        sector = self.fp.read(self.sectorsize)
        if len(sector) != self.sectorsize:
            debug('getsect(): sect=%X, read=%d, sectorsize=%d' %
                (sect, len(sector), self.sectorsize))
            self.raise_defect(DEFECT_FATAL, 'incomplete OLE sector')
        return sector

    def write_sect(self, sect, data, padding=b'\x00'):
        """
        Write given sector to file on disk.

        :param sect: int, sector index
        :param data: bytes, sector data
        :param padding: single byte, padding character if data < sector size
        """
        if not isinstance(data, bytes):
            raise TypeError("write_sect: data must be a bytes string")
        if not isinstance(padding, bytes) or len(padding) != 1:
            raise TypeError("write_sect: padding must be a bytes string of 1 char")
        #TODO: we could allow padding=None for no padding at all
        try:
            self.fp.seek(self.sectorsize * (sect+1))
        except:
            debug('write_sect(): sect=%X, seek=%d, filesize=%d' %
                (sect, self.sectorsize*(sect+1), self._filesize))
            self.raise_defect(DEFECT_FATAL, 'OLE sector index out of range')
        if len(data) < self.sectorsize:
            # add padding
            data += padding * (self.sectorsize - len(data))
        elif len(data) < self.sectorsize:
            raise ValueError("Data is larger than sector size")
        self.fp.write(data)

    def loaddirectory(self, sect):
        """
        Load the directory.

        :param sect: sector index of directory stream.
        """
        # The directory is  stored in a standard
        # substream, independent of its size.

        # open directory stream as a read-only file:
        # (stream size is not known in advance)
        self.directory_fp = self._open(sect)

        # [PL] to detect malformed documents and avoid DoS attacks, the maximum
        # number of directory entries can be calculated:
        max_entries = self.directory_fp.size // 128
        debug('loaddirectory: size=%d, max_entries=%d' %
            (self.directory_fp.size, max_entries))

        # Create list of directory entries
        # self.direntries = []
        # We start with a list of "None" object
        self.direntries = [None] * max_entries
##        for sid in iterrange(max_entries):
##            entry = fp.read(128)
##            if not entry:
##                break
##            self.direntries.append(_OleDirectoryEntry(entry, sid, self))
        # load root entry:
        root_entry = self._load_direntry(0)
        # Root entry is the first entry:
        self.root = self.direntries[0]
        # read and build all storage trees, starting from the root:
        self.root.build_storage_tree()

    def _load_direntry(self, sid):
        """
        Load a directory entry from the directory.
        This method should only be called once for each storage/stream when
        loading the directory.

        :param sid: index of storage/stream in the directory.
        :returns: a _OleDirectoryEntry object

        :exception IOError: if the entry has always been referenced.
        """
        # check if SID is OK:
        if sid < 0 or sid >= len(self.direntries):
            self.raise_defect(DEFECT_FATAL, "OLE directory index out of range")
        # check if entry was already referenced:
        if self.direntries[sid] is not None:
            self.raise_defect(DEFECT_INCORRECT,
                "double reference for OLE stream/storage")
            # if exception not raised, return the object
            return self.direntries[sid]
        self.directory_fp.seek(sid * 128)
        entry = self.directory_fp.read(128)
        self.direntries[sid] = _OleDirectoryEntry(entry, sid, self)
        return self.direntries[sid]

    def dumpdirectory(self):
        """
        Dump directory (for debugging only)
        """
        self.root.dump()

    def _open(self, start, size = 0x7FFFFFFF, force_FAT=False):
        """
        Open a stream, either in FAT or MiniFAT according to its size.
        (openstream helper)

        :param start: index of first sector
        :param size: size of stream (or nothing if size is unknown)
        :param force_FAT: if False (default), stream will be opened in FAT or MiniFAT
            according to size. If True, it will always be opened in FAT.
        """
        debug('OleFileIO.open(): sect=%d, size=%d, force_FAT=%s' %
            (start, size, str(force_FAT)))
        # stream size is compared to the MiniSectorCutoff threshold:
        if size < self.minisectorcutoff and not force_FAT:
            # ministream object
            if not self.ministream:
                # load MiniFAT if it wasn't already done:
                self.loadminifat()
                # The first sector index of the miniFAT stream is stored in the
                # root directory entry:
                size_ministream = self.root.size
                debug('Opening MiniStream: sect=%d, size=%d' %
                    (self.root.isectStart, size_ministream))
                self.ministream = self._open(self.root.isectStart,
                    size_ministream, force_FAT=True)
            return _OleStream(fp=self.ministream, sect=start, size=size,
                              offset=0, sectorsize=self.minisectorsize,
                              fat=self.minifat, filesize=self.ministream.size)
        else:
            # standard stream
            return _OleStream(fp=self.fp, sect=start, size=size,
                              offset=self.sectorsize,
                              sectorsize=self.sectorsize, fat=self.fat,
                              filesize=self._filesize)

    def _list(self, files, prefix, node, streams=True, storages=False):
        """
        listdir helper

        :param files: list of files to fill in
        :param prefix: current location in storage tree (list of names)
        :param node: current node (_OleDirectoryEntry object)
        :param streams: bool, include streams if True (True by default) - new in v0.26
        :param storages: bool, include storages if True (False by default) - new in v0.26
            (note: the root storage is never included)
        """
        prefix = prefix + [node.name]
        for entry in node.kids:
            if entry.entry_type == STGTY_STORAGE:
                # this is a storage
                if storages:
                    # add it to the list
                    files.append(prefix[1:] + [entry.name])
                # check its kids
                self._list(files, prefix, entry, streams, storages)
            elif entry.entry_type == STGTY_STREAM:
                # this is a stream
                if streams:
                    # add it to the list
                    files.append(prefix[1:] + [entry.name])
            else:
                self.raise_defect(DEFECT_INCORRECT, 'The directory tree contains an entry which is not a stream nor a storage.')

    def listdir(self, streams=True, storages=False):
        """
        Return a list of streams and/or storages stored in this file

        :param streams: bool, include streams if True (True by default) - new in v0.26
        :param storages: bool, include storages if True (False by default) - new in v0.26
            (note: the root storage is never included)
        :returns: list of stream and/or storage paths
        """
        files = []
        self._list(files, [], self.root, streams, storages)
        return files

    def _find(self, filename):
        """
        Returns directory entry of given filename. (openstream helper)
        Note: this method is case-insensitive.

        :param filename: path of stream in storage tree (except root entry), either:

            - a string using Unix path syntax, for example:
              'storage_1/storage_1.2/stream'
            - or a list of storage filenames, path to the desired stream/storage.
              Example: ['storage_1', 'storage_1.2', 'stream']

        :returns: sid of requested filename
        :exception IOError: if file not found
        """

        # if filename is a string instead of a list, split it on slashes to
        # convert to a list:
        if isinstance(filename, basestring):
            filename = filename.split('/')
        # walk across storage tree, following given path:
        node = self.root
        for name in filename:
            for kid in node.kids:
                if kid.name.lower() == name.lower():
                    break
            else:
                raise IOError("file not found")
            node = kid
        return node.sid

    def openstream(self, filename):
        """
        Open a stream as a read-only file object (BytesIO).
        Note: filename is case-insensitive.

        :param filename: path of stream in storage tree (except root entry), either:

            - a string using Unix path syntax, for example:
              'storage_1/storage_1.2/stream'
            - or a list of storage filenames, path to the desired stream/storage.
              Example: ['storage_1', 'storage_1.2', 'stream']

        :returns: file object (read-only)
        :exception IOError: if filename not found, or if this is not a stream.
        """
        sid = self._find(filename)
        entry = self.direntries[sid]
        if entry.entry_type != STGTY_STREAM:
            raise IOError("this file is not a stream")
        return self._open(entry.isectStart, entry.size)

    def write_stream(self, stream_name, data):
        """
        Write a stream to disk. For now, it is only possible to replace an
        existing stream by data of the same size.

        :param stream_name: path of stream in storage tree (except root entry), either:

            - a string using Unix path syntax, for example:
              'storage_1/storage_1.2/stream'
            - or a list of storage filenames, path to the desired stream/storage.
              Example: ['storage_1', 'storage_1.2', 'stream']

        :param data: bytes, data to be written, must be the same size as the original
            stream.
        """
        if not isinstance(data, bytes):
            raise TypeError("write_stream: data must be a bytes string")
        sid = self._find(stream_name)
        entry = self.direntries[sid]
        if entry.entry_type != STGTY_STREAM:
            raise IOError("this is not a stream")
        size = entry.size
        if size != len(data):
            raise ValueError("write_stream: data must be the same size as the existing stream")
        if size < self.minisectorcutoff:
            raise NotImplementedError("Writing a stream in MiniFAT is not implemented yet")
        sect = entry.isectStart
        # number of sectors to write
        nb_sectors = (size + (self.sectorsize-1)) // self.sectorsize
        debug('nb_sectors = %d' % nb_sectors)
        for i in range(nb_sectors):
            # try:
            #     self.fp.seek(offset + self.sectorsize * sect)
            # except:
            #     debug('sect=%d, seek=%d' %
            #         (sect, offset+self.sectorsize*sect))
            #     raise IOError('OLE sector index out of range')
            # extract one sector from data, the last one being smaller:
            if i < (nb_sectors-1):
                data_sector = data[i*self.sectorsize:(i+1)*self.sectorsize]
                #TODO: comment this if it works
                assert(len(data_sector) == self.sectorsize)
            else:
                data_sector = data[i*self.sectorsize:]
                # TODO: comment this if it works
                debug('write_stream: size=%d sectorsize=%d data_sector=%d size%%sectorsize=%d'
                    % (size, self.sectorsize, len(data_sector), size % self.sectorsize))
                assert(len(data_sector) % self.sectorsize == size % self.sectorsize)
            self.write_sect(sect, data_sector)
#            self.fp.write(data_sector)
            # jump to next sector in the FAT:
            try:
                sect = self.fat[sect]
            except IndexError:
                # [PL] if pointer is out of the FAT an exception is raised
                raise IOError('incorrect OLE FAT, sector index out of range')
        # [PL] Last sector should be a "end of chain" marker:
        if sect != ENDOFCHAIN:
            raise IOError('incorrect last sector index in OLE stream')

    def get_type(self, filename):
        """
        Test if given filename exists as a stream or a storage in the OLE
        container, and return its type.

        :param filename: path of stream in storage tree. (see openstream for syntax)
        :returns: False if object does not exist, its entry type (>0) otherwise:

            - STGTY_STREAM: a stream
            - STGTY_STORAGE: a storage
            - STGTY_ROOT: the root entry
        """
        try:
            sid = self._find(filename)
            entry = self.direntries[sid]
            return entry.entry_type
        except:
            return False

    def getmtime(self, filename):
        """
        Return modification time of a stream/storage.

        :param filename: path of stream/storage in storage tree. (see openstream for
            syntax)
        :returns: None if modification time is null, a python datetime object
            otherwise (UTC timezone)

        new in version 0.26
        """
        sid = self._find(filename)
        entry = self.direntries[sid]
        return entry.getmtime()

    def getctime(self, filename):
        """
        Return creation time of a stream/storage.

        :param filename: path of stream/storage in storage tree. (see openstream for
            syntax)
        :returns: None if creation time is null, a python datetime object
            otherwise (UTC timezone)

        new in version 0.26
        """
        sid = self._find(filename)
        entry = self.direntries[sid]
        return entry.getctime()

    def exists(self, filename):
        """
        Test if given filename exists as a stream or a storage in the OLE
        container.
        Note: filename is case-insensitive.

        :param filename: path of stream in storage tree. (see openstream for syntax)
        :returns: True if object exist, else False.
        """
        try:
            sid = self._find(filename)
            return True
        except:
            return False

    def get_size(self, filename):
        """
        Return size of a stream in the OLE container, in bytes.

        :param filename: path of stream in storage tree (see openstream for syntax)
        :returns: size in bytes (long integer)
        :exception IOError: if file not found
        :exception TypeError: if this is not a stream.
        """
        sid = self._find(filename)
        entry = self.direntries[sid]
        if entry.entry_type != STGTY_STREAM:
            #TODO: Should it return zero instead of raising an exception ?
            raise TypeError('object is not an OLE stream')
        return entry.size

    def get_rootentry_name(self):
        """
        Return root entry name. Should usually be 'Root Entry' or 'R' in most
        implementations.
        """
        return self.root.name

    def getproperties(self, filename, convert_time=False, no_conversion=None):
        """
        Return properties described in substream.

        :param filename: path of stream in storage tree (see openstream for syntax)
        :param convert_time: bool, if True timestamps will be converted to Python datetime
        :param no_conversion: None or list of int, timestamps not to be converted
                              (for example total editing time is not a real timestamp)

        :returns: a dictionary of values indexed by id (integer)
        """
        # REFERENCE: [MS-OLEPS] https://msdn.microsoft.com/en-us/library/dd942421.aspx
        # make sure no_conversion is a list, just to simplify code below:
        if no_conversion is None:
            no_conversion = []
        # stream path as a string to report exceptions:
        streampath = filename
        if not isinstance(streampath, str):
            streampath = '/'.join(streampath)

        fp = self.openstream(filename)

        data = {}

        try:
            # header
            s = fp.read(28)
            clsid = _clsid(s[8:24])

            # format id
            s = fp.read(20)
            fmtid = _clsid(s[:16])
            fp.seek(i32(s, 16))

            # get section
            s = b"****" + fp.read(i32(fp.read(4))-4)
            # number of properties:
            num_props = i32(s, 4)
        except BaseException as exc:
            # catch exception while parsing property header, and only raise
            # a DEFECT_INCORRECT then return an empty dict, because this is not
            # a fatal error when parsing the whole file
            msg = 'Error while parsing properties header in stream %s: %s' % (
                repr(streampath), exc)
            self.raise_defect(DEFECT_INCORRECT, msg, type(exc))
            return data

        for i in range(num_props):
            try:
                id = 0 # just in case of an exception
                id = i32(s, 8+i*8)
                offset = i32(s, 12+i*8)
                type = i32(s, offset)

                debug('property id=%d: type=%d offset=%X' % (id, type, offset))

                # test for common types first (should perhaps use
                # a dictionary instead?)

                if type == VT_I2: # 16-bit signed integer
                    value = i16(s, offset+4)
                    if value >= 32768:
                        value = value - 65536
                elif type == VT_UI2:  # 2-byte unsigned integer
                    value = i16(s, offset+4)
                elif type in (VT_I4, VT_INT, VT_ERROR):
                    # VT_I4: 32-bit signed integer
                    # VT_ERROR: HRESULT, similar to 32-bit signed integer,
                    # see http://msdn.microsoft.com/en-us/library/cc230330.aspx
                    value = i32(s, offset+4)
                elif type in (VT_UI4, VT_UINT):  # 4-byte unsigned integer
                    value = i32(s, offset+4) # FIXME
                elif type in (VT_BSTR, VT_LPSTR):
                    # CodePageString, see http://msdn.microsoft.com/en-us/library/dd942354.aspx
                    # size is a 32 bits integer, including the null terminator, and
                    # possibly trailing or embedded null chars
                    #TODO: if codepage is unicode, the string should be converted as such
                    count = i32(s, offset+4)
                    value = s[offset+8:offset+8+count-1]
                    # remove all null chars:
                    value = value.replace(b'\x00', b'')
                elif type == VT_BLOB:
                    # binary large object (BLOB)
                    # see http://msdn.microsoft.com/en-us/library/dd942282.aspx
                    count = i32(s, offset+4)
                    value = s[offset+8:offset+8+count]
                elif type == VT_LPWSTR:
                    # UnicodeString
                    # see http://msdn.microsoft.com/en-us/library/dd942313.aspx
                    # "the string should NOT contain embedded or additional trailing
                    # null characters."
                    count = i32(s, offset+4)
                    value = self._decode_utf16_str(s[offset+8:offset+8+count*2])
                elif type == VT_FILETIME:
                    value = long(i32(s, offset+4)) + (long(i32(s, offset+8)) << 32)
                    # FILETIME is a 64-bit int: "number of 100ns periods
                    # since Jan 1,1601".
                    if convert_time and id not in no_conversion:
                        debug('Converting property #%d to python datetime, value=%d=%fs'
                                % (id, value, float(value) / 10000000))
                        # convert FILETIME to Python datetime.datetime
                        # inspired from http://code.activestate.com/recipes/511425-filetime-to-datetime/
                        _FILETIME_null_date = datetime.datetime(1601, 1, 1, 0, 0, 0)
                        debug('timedelta days=%d' % (value//(10*1000000*3600*24)))
                        value = _FILETIME_null_date + datetime.timedelta(microseconds=value//10)
                    else:
                        # legacy code kept for backward compatibility: returns a
                        # number of seconds since Jan 1,1601
                        value = value // 10000000 # seconds
                elif type == VT_UI1: # 1-byte unsigned integer
                    value = i8(s[offset+4])
                elif type == VT_CLSID:
                    value = _clsid(s[offset+4:offset+20])
                elif type == VT_CF:
                    # PropertyIdentifier or ClipboardData??
                    # see http://msdn.microsoft.com/en-us/library/dd941945.aspx
                    count = i32(s, offset+4)
                    value = s[offset+8:offset+8+count]
                elif type == VT_BOOL:
                    # VARIANT_BOOL, 16 bits bool, 0x0000=Fals, 0xFFFF=True
                    # see http://msdn.microsoft.com/en-us/library/cc237864.aspx
                    value = bool(i16(s, offset+4))
                else:
                    value = None # everything else yields "None"
                    debug('property id=%d: type=%d not implemented in parser yet' % (id, type))

                # missing: VT_EMPTY, VT_NULL, VT_R4, VT_R8, VT_CY, VT_DATE,
                # VT_DECIMAL, VT_I1, VT_I8, VT_UI8,
                # see http://msdn.microsoft.com/en-us/library/dd942033.aspx

                # FIXME: add support for VT_VECTOR
                # VT_VECTOR is a 32 uint giving the number of items, followed by
                # the items in sequence. The VT_VECTOR value is combined with the
                # type of items, e.g. VT_VECTOR|VT_BSTR
                # see http://msdn.microsoft.com/en-us/library/dd942011.aspx

                # print("%08x" % id, repr(value), end=" ")
                # print("(%s)" % VT[i32(s, offset) & 0xFFF])

                data[id] = value
            except BaseException as exc:
                # catch exception while parsing each property, and only raise
                # a DEFECT_INCORRECT, because parsing can go on
                msg = 'Error while parsing property id %d in stream %s: %s' % (
                    id, repr(streampath), exc)
                self.raise_defect(DEFECT_INCORRECT, msg, type(exc))

        return data

    def get_metadata(self):
        """
        Parse standard properties streams, return an OleMetadata object
        containing all the available metadata.
        (also stored in the metadata attribute of the OleFileIO object)

        new in version 0.25
        """
        self.metadata = OleMetadata()
        self.metadata.parse_properties(self)
        return self.metadata

#
# --------------------------------------------------------------------
# This script can be used to dump the directory of any OLE2 structured
# storage file.

if __name__ == "__main__":

    # [PL] display quick usage info if launched from command-line
    if len(sys.argv) <= 1:
        print('olefile version %s %s - %s' % (__version__, __date__, __author__))
        print(
"""
Launched from the command line, this script parses OLE files and prints info.

Usage: olefile.py [-d] [-c] <file> [file2 ...]

Options:
-d : debug mode (displays a lot of debug information, for developers only)
-c : check all streams (for debugging purposes)

For more information, see http://www.decalage.info/olefile
""")
        sys.exit()

    check_streams = False
    for filename in sys.argv[1:]:
        # try:
        # OPTIONS:
        if filename == '-d':
            # option to switch debug mode on:
            set_debug_mode(True)
            continue
        if filename == '-c':
            # option to switch check streams mode on:
            check_streams = True
            continue

        ole = OleFileIO(filename)#, raise_defects=DEFECT_INCORRECT)
        print("-" * 68)
        print(filename)
        print("-" * 68)
        ole.dumpdirectory()
        for streamname in ole.listdir():
            if streamname[-1][0] == "\005":
                print(streamname, ": properties")
                props = ole.getproperties(streamname, convert_time=True)
                props = sorted(props.items())
                for k, v in props:
                    # [PL]: avoid to display too large or binary values:
                    if isinstance(v, (basestring, bytes)):
                        if len(v) > 50:
                            v = v[:50]
                    if isinstance(v, bytes):
                        # quick and dirty binary check:
                        for c in (1, 2, 3, 4, 5, 6, 7, 11, 12, 14, 15, 16, 17, 18, 19, 20,
                                  21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31):
                            if c in bytearray(v):
                                v = '(binary data)'
                                break
                    print("   ", k, v)

        if check_streams:
            # Read all streams to check if there are errors:
            print('\nChecking streams...')
            for streamname in ole.listdir():
                # print name using repr() to convert binary chars to \xNN:
                print('-', repr('/'.join(streamname)), '-', end=' ')
                st_type = ole.get_type(streamname)
                if st_type == STGTY_STREAM:
                    print('size %d' % ole.get_size(streamname))
                    # just try to read stream in memory:
                    ole.openstream(streamname)
                else:
                    print('NOT a stream : type=%d' % st_type)
            print()

#            for streamname in ole.listdir():
#                # print name using repr() to convert binary chars to \xNN:
#                print('-', repr('/'.join(streamname)),'-', end=' ')
#                print(ole.getmtime(streamname))
#            print()

        print('Modification/Creation times of all directory entries:')
        for entry in ole.direntries:
            if entry is not None:
                print('- %s: mtime=%s ctime=%s' % (entry.name,
                    entry.getmtime(), entry.getctime()))
        print()

        # parse and display metadata:
        meta = ole.get_metadata()
        meta.dump()
        print()
        # [PL] Test a few new methods:
        root = ole.get_rootentry_name()
        print('Root entry name: "%s"' % root)
        if ole.exists('worddocument'):
            print("This is a Word document.")
            print("type of stream 'WordDocument':", ole.get_type('worddocument'))
            print("size :", ole.get_size('worddocument'))
            if ole.exists('macros/vba'):
                print("This document may contain VBA macros.")

        # print parsing issues:
        print('\nNon-fatal issues raised during parsing:')
        if ole.parsing_issues:
            for exctype, msg in ole.parsing_issues:
                print('- %s: %s' % (exctype.__name__, msg))
        else:
            print('None')
##      except IOError as v:
##          print("***", "cannot read", file, "-", v)

# this code was developed while listening to The Wedding Present "Sea Monsters"