faq: update

[rarfile.git] / rarfile.py

# rarfile.py
#
# Copyright (c) 2005-2011  Marko Kreen <markokr@gmail.com>
#
# Permission to use, copy, modify, and/or distribute this software for any
# purpose with or without fee is hereby granted, provided that the above
# copyright notice and this permission notice appear in all copies.
#
# THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
# WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
# MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
# ANY SPECIAL, DIRECT, 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.

r"""RAR archive reader.

This is Python module for Rar archive reading.  The interface
is made as zipfile like as possible.

Basic logic:
 - Parse archive structure with Python.
 - Extract non-compressed files with Python
 - Extract compressed files with unrar.
 - Optionally write compressed data to temp file to speed up unrar,
   otherwise it needs to scan whole archive on each execution.

Example::

    import rarfile

    rf = rarfile.RarFile('myarchive.rar')
    for f in rf.infolist():
        print f.filename, f.file_size
        if f.filename == 'README':
            print rf.read(f)

There are few module-level parameters to tune behaviour,
here they are with defaults, and reason to change it::

    import rarfile

    # Set to full path of unrar.exe if it is not in PATH
    rarfile.UNRAR_TOOL = "unrar"

    # Set to 0 if you don't look at comments and want to
    # avoid wasting time for parsing them
    rarfile.NEED_COMMENTS = 1

    # Set up to 1 if you don't want to deal with decoding comments
    # from unknown encoding.  rarfile will try couple of common
    # encodings in sequence.
    rarfile.UNICODE_COMMENTS = 0

    # Set to 1 if you prefer timestamps to be datetime objects
    # instead tuples
    rarfile.USE_DATETIME = 0

    # Set to '/' to be more compatible with zipfile
    rarfile.PATH_SEP = '\\'

For more details, refer to source.

"""

__version__ = '2.4'

# export only interesting items
__all__ = ['is_rarfile', 'RarInfo', 'RarFile', 'RarExtFile']

##
## Imports and compat - support both Python 2.x and 3.x
##

import sys, os, struct
from struct import pack, unpack
from binascii import crc32
from tempfile import mkstemp
from subprocess import Popen, PIPE, STDOUT
from datetime import datetime

# only needed for encryped headers
try:
    from Crypto.Cipher import AES
    try:
        from hashlib import sha1
    except ImportError:
        from sha import new as sha1
    _have_crypto = 1
except ImportError:
    _have_crypto = 0

# compat with 2.x
if sys.hexversion < 0x3000000:
    # prefer 3.x behaviour
    range = xrange
    # py2.6 has broken bytes()
    def bytes(s, enc):
        return str(s)

# see if compat bytearray() is needed
try:
    bytearray
except NameError:
    import array
    class bytearray:
        def __init__(self, val = ''):
            self.arr = array.array('B', val)
            self.append = self.arr.append
            self.__getitem__ = self.arr.__getitem__
            self.__len__ = self.arr.__len__
        def decode(self, *args):
            return self.arr.tostring().decode(*args)

# Optimized .readinto() requires memoryview
try:
    memoryview
    have_memoryview = 1
except NameError:
    have_memoryview = 0

# Struct() for older python
try:
    from struct import Struct
except ImportError:
    class Struct:
        def __init__(self, fmt):
            self.format = fmt
            self.size = struct.calcsize(fmt)
        def unpack(self, buf):
            return unpack(self.format, buf)
        def unpack_from(self, buf, ofs = 0):
            return unpack(self.format, buf[ofs : ofs + self.size])
        def pack(self, *args):
            return pack(self.format, *args)

# file object superclass
try:
    from io import RawIOBase
except ImportError:
    class RawIOBase(object):
        def close(self):
            pass


##
## Module configuration.  Can be tuned after importing.
##

# default fallback charset
DEFAULT_CHARSET = "windows-1252"

# list of encodings to try, with fallback to DEFAULT_CHARSET if none succeed
TRY_ENCODINGS = ('utf8', 'utf-16le')

# 'unrar', 'rar' or full path to either one
UNRAR_TOOL = "unrar"

# Command line args to use for opening file for reading.
OPEN_ARGS = ('p', '-inul')

# Command line args to use for extracting file to disk.
EXTRACT_ARGS = ('x', '-y', '-idq')

# args for testrar()
TEST_ARGS = ('t', '-idq')

# whether to speed up decompression by using tmp archive
USE_EXTRACT_HACK = 1

# limit the filesize for tmp archive usage
HACK_SIZE_LIMIT = 20*1024*1024

# whether to parse file/archive comments.
NEED_COMMENTS = 1

# whether to convert comments to unicode strings
UNICODE_COMMENTS = 0

# When RAR is corrupt, stopping on bad header is better
# On unknown/misparsed RAR headers reporting is better
REPORT_BAD_HEADER = 0

# Convert RAR time tuple into datetime() object
USE_DATETIME = 0

# Separator for path name components.  RAR internally uses '\\'.
# Use '/' to be similar with zipfile.
PATH_SEP = '\\'

##
## rar constants
##

# block types
RAR_BLOCK_MARK          = 0x72 # r
RAR_BLOCK_MAIN          = 0x73 # s
RAR_BLOCK_FILE          = 0x74 # t
RAR_BLOCK_OLD_COMMENT   = 0x75 # u
RAR_BLOCK_OLD_EXTRA     = 0x76 # v
RAR_BLOCK_OLD_SUB       = 0x77 # w
RAR_BLOCK_OLD_RECOVERY  = 0x78 # x
RAR_BLOCK_OLD_AUTH      = 0x79 # y
RAR_BLOCK_SUB           = 0x7a # z
RAR_BLOCK_ENDARC        = 0x7b # {

# flags for RAR_BLOCK_MAIN
RAR_MAIN_VOLUME         = 0x0001
RAR_MAIN_COMMENT        = 0x0002
RAR_MAIN_LOCK           = 0x0004
RAR_MAIN_SOLID          = 0x0008
RAR_MAIN_NEWNUMBERING   = 0x0010
RAR_MAIN_AUTH           = 0x0020
RAR_MAIN_RECOVERY       = 0x0040
RAR_MAIN_PASSWORD       = 0x0080
RAR_MAIN_FIRSTVOLUME    = 0x0100
RAR_MAIN_ENCRYPTVER     = 0x0200

# flags for RAR_BLOCK_FILE
RAR_FILE_SPLIT_BEFORE   = 0x0001
RAR_FILE_SPLIT_AFTER    = 0x0002
RAR_FILE_PASSWORD       = 0x0004
RAR_FILE_COMMENT        = 0x0008
RAR_FILE_SOLID          = 0x0010
RAR_FILE_DICTMASK       = 0x00e0
RAR_FILE_DICT64         = 0x0000
RAR_FILE_DICT128        = 0x0020
RAR_FILE_DICT256        = 0x0040
RAR_FILE_DICT512        = 0x0060
RAR_FILE_DICT1024       = 0x0080
RAR_FILE_DICT2048       = 0x00a0
RAR_FILE_DICT4096       = 0x00c0
RAR_FILE_DIRECTORY      = 0x00e0
RAR_FILE_LARGE          = 0x0100
RAR_FILE_UNICODE        = 0x0200
RAR_FILE_SALT           = 0x0400
RAR_FILE_VERSION        = 0x0800
RAR_FILE_EXTTIME        = 0x1000
RAR_FILE_EXTFLAGS       = 0x2000

# flags for RAR_BLOCK_ENDARC
RAR_ENDARC_NEXT_VOLUME  = 0x0001
RAR_ENDARC_DATACRC      = 0x0002
RAR_ENDARC_REVSPACE     = 0x0004
RAR_ENDARC_VOLNR        = 0x0008

# flags common to all blocks
RAR_SKIP_IF_UNKNOWN     = 0x4000
RAR_LONG_BLOCK          = 0x8000

# Host OS types
RAR_OS_MSDOS = 0
RAR_OS_OS2   = 1
RAR_OS_WIN32 = 2
RAR_OS_UNIX  = 3
RAR_OS_MACOS = 4
RAR_OS_BEOS  = 5

# Compression methods - '0'..'5'
RAR_M0 = 0x30
RAR_M1 = 0x31
RAR_M2 = 0x32
RAR_M3 = 0x33
RAR_M4 = 0x34
RAR_M5 = 0x35

##
## internal constants
##

RAR_ID = bytes("Rar!\x1a\x07\x00", 'ascii')
ZERO = bytes("\0", 'ascii')
EMPTY = bytes("", 'ascii')

S_BLK_HDR = Struct('<HBHH')
S_FILE_HDR = Struct('<LLBLLBBHL')
S_LONG = Struct('<L')
S_SHORT = Struct('<H')
S_BYTE = Struct('<B')
S_COMMENT_HDR = Struct('<HBBH')

##
## Public interface
##

class Error(Exception):
    """Base class for rarfile errors."""
class BadRarFile(Error):
    """Incorrect data in archive."""
class NotRarFile(Error):
    """The file is not RAR archive."""
class BadRarName(Error):
    """Cannot guess multipart name components."""
class NoRarEntry(Error):
    """File not found in RAR"""
class PasswordRequired(Error):
    """File requires password"""
class NeedFirstVolume(Error):
    """Need to start from first volume."""
class NoCrypto(Error):
    """Cannot parse encrypted headers - no crypto available."""


def is_rarfile(fn):
    '''Check quickly whether file is rar archive.'''
    buf = open(fn, "rb").read(len(RAR_ID))
    return buf == RAR_ID


class RarInfo(object):
    '''An entry in rar archive.
    
    @ivar filename:
        File name with relative path.
        Default path separator is '/', to change set rarfile.PATH_SEP.
        Always unicode string.
    @ivar date_time:
        Modification time, tuple of (year, month, day, hour, minute, second).
        Or datetime() object if USE_DATETIME is set.
    @ivar file_size:
        Uncompressed size.
    @ivar compress_size:
        Compressed size.
    @ivar compress_type:
        Compression method: 0x30 - 0x35.
    @ivar extract_version:
        Minimal Rar version needed for decompressing.
    @ivar host_os:
        Host OS type, one of RAR_OS_* constants.
    @ivar mode:
        File attributes. May be either dos-style or unix-style, depending on host_os.
    @ivar CRC:
        CRC-32 of uncompressed file, unsigned int.
    @ivar volume:
        Volume nr, starting from 0.
    @ivar volume_file:
        Volume file name, where file starts.
    @ivar type:
        One of RAR_BLOCK_* types.  Only entries with type==RAR_BLOCK_FILE are shown in .infolist().
    @ivar flags:
        For files, RAR_FILE_* bits.
    @ivar comment:
        File comment (unicode string or None).

    @ivar mtime:
        Optional time field: Modification time, with float seconds.
        Same as .date_time but with more precision.
    @ivar ctime:
        Optional time field: creation time, with float seconds.
    @ivar atime:
        Optional time field: last access time, with float seconds.
    @ivar arctime:
        Optional time field: archival time, with float seconds.
    '''

    __slots__ = (
        # zipfile-compatible fields
        'filename',
        'file_size',
        'compress_size',
        'date_time',
        'comment',
        'CRC',
        'volume',
        'orig_filename', # bytes in unknown encoding

        # rar-specific fields
        'extract_version',
        'compress_type',
        'host_os',
        'mode',
        'type',
        'flags',

        # optional extended time fields
        # tuple where the sec is float, or datetime().
        'mtime', # same as .date_time
        'ctime',
        'atime',
        'arctime',

        # RAR internals
        'name_size',
        'header_size',
        'header_crc',
        'file_offset',
        'add_size',
        'header_data',
        'header_base',
        'header_offset',
        'salt',
        'volume_file',
    )

    def isdir(self):
        '''Returns True if the entry is a directory.'''
        if self.type == RAR_BLOCK_FILE:
            return (self.flags & RAR_FILE_DIRECTORY) == RAR_FILE_DIRECTORY
        return False

    def needs_password(self):
        return self.flags & RAR_FILE_PASSWORD


class RarFile(object):
    '''Parse RAR structure, provide access to files in archive.

    @ivar comment:
        Archive comment (unicode string or None).
    '''

    def __init__(self, rarfile, mode="r", charset=None, info_callback=None, crc_check = True):
        """Open and parse a RAR archive.
        
        @param rarfile: archive file name
        @param mode: only 'r' is supported.
        @param charset: fallback charset to use, if filenames are not already Unicode-enabled.
        @param info_callback: debug callback, gets to see all archive entries.
        @param crc_check: set to False to disable CRC checks
        """
        self.rarfile = rarfile
        self.comment = None
        self._charset = charset or DEFAULT_CHARSET
        self._info_callback = info_callback

        self._info_list = []
        self._info_map = {}
        self._needs_password = False
        self._password = None
        self._crc_check = crc_check

        self._main = None

        if mode != "r":
            raise NotImplementedError("RarFile supports only mode=r")

        self._parse()

    def setpassword(self, password):
        '''Sets the password to use when extracting.'''
        self._password = password
        if not self._main:
            self._parse()

    def needs_password(self):
        '''Returns True if any archive entries require password for extraction.'''
        return self._needs_password

    def namelist(self):
        '''Return list of filenames in archive.'''
        return [f.filename for f in self._info_list]

    def infolist(self):
        '''Return RarInfo objects for all files/directories in archive.'''
        return self._info_list

    def getinfo(self, fname):
        '''Return RarInfo for file.'''

        if isinstance(fname, RarInfo):
            return fname

        # accept both ways here
        if PATH_SEP == '/':
            fname2 = fname.replace("\\", "/")
        else:
            fname2 = fname.replace("/", "\\")

        try:
            return self._info_map[fname]
        except KeyError:
            try:
                return self._info_map[fname2]
            except KeyError:
                raise NoRarEntry("No such file: "+fname)

    def open(self, fname, mode = 'r', psw = None):
        '''Return open file object, where the data can be read.
        
        The object implements io.RawIOBase interface, so it can
        be further wrapped with io.BufferedReader and io.TextIOWrapper.

        On older Python where io module is not available, it implements
        only .read(), .seek(), .tell() and .close() methods.

        The object is seekable, although the seeking is fast only on
        uncompressed files, on compressed files the seeking is implemented
        by reading ahead and/or restarting the decompression.

        @param fname: file name or RarInfo instance.
        @param mode: must be 'r'
        @param psw: password to use for extracting.
        '''

        if mode != 'r':
            raise NotImplementedError("RarFile.open() supports only mode=r")

        # entry lookup
        inf = self.getinfo(fname)
        if inf.isdir():
            raise TypeError("Directory does not have any data: " + inf.filename)

        if inf.flags & RAR_FILE_SPLIT_BEFORE:
            raise NeedFirstVolume("Partial file, please start from first volume: " + inf.filename)

        # check password
        if inf.needs_password():
            psw = psw or self._password
            if psw is None:
                raise PasswordRequired("File %s requires password" % inf.filename)
        else:
            psw = None

        # is temp write usable?
        if not USE_EXTRACT_HACK or not self._main:
            use_hack = 0
        elif self._main.flags & (RAR_MAIN_SOLID | RAR_MAIN_PASSWORD):
            use_hack = 0
        elif inf.flags & (RAR_FILE_SPLIT_BEFORE | RAR_FILE_SPLIT_AFTER):
            use_hack = 0
        elif inf.file_size > HACK_SIZE_LIMIT:
            use_hack = 0
        else:
            use_hack = 1

        # now extract
        if inf.compress_type == RAR_M0 and (inf.flags & RAR_FILE_PASSWORD) == 0:
            return self._open_clear(inf)
        elif use_hack:
            return self._open_hack(inf, psw)
        else:
            return self._open_unrar(self.rarfile, inf, psw)

    def read(self, fname, psw = None):
        """Return uncompressed data for archive entry.
        
        For longer files using .open() may be better idea.

        @param fname: filename or RarInfo instance
        @param psw: password to use for extracting.
        """

        f = self.open(fname, 'r', psw)
        try:
            return f.read()
        finally:
            f.close()

    def close(self):
        """Release open resources."""
        pass

    def printdir(self):
        """Print archive file list to stdout."""
        for f in self._info_list:
            print(f.filename)

    def extract(self, member, path=None, pwd=None):
        """Extract single file into current directory.
        
        @param member: filename or RarInfo instance
        @param path: optional destination path
        @param pwd: optional password to use
        """
        if isinstance(member, RarInfo):
            fname = member.filename
        else:
            fname = member
        self._extract([fname], path, pwd)

    def extractall(self, path=None, members=None, pwd=None):
        """Extract all files into current directory.
        
        @param path: optional destination path
        @param members: optional filename or RarInfo instance list to extract
        @param pwd: optional password to use
        """
        fnlist = []
        if members is not None:
            for m in members:
                if isinstance(m, RarInfo):
                    fnlist.append(m.filename)
                else:
                    fnlist.append(m)
        self._extract(fnlist, path, pwd)

    def testrar(self):
        """Let 'unrar' test the archive.
        """
        cmd = [UNRAR_TOOL] + list(TEST_ARGS)
        if self._password is not None:
            cmd.append('-p' + self._password)
        else:
            cmd.append('-p-')
        cmd.append(self.rarfile)
        p = custom_popen(cmd)
        p.communicate()
        if p.returncode != 0:
            raise BadRarFile("Testing failed")

    ##
    ## private methods
    ##

    # store entry
    def _process_entry(self, item):
        if item.type == RAR_BLOCK_FILE:
            # use only first part
            if (item.flags & RAR_FILE_SPLIT_BEFORE) == 0:
                self._info_map[item.filename] = item
                self._info_list.append(item)
                # remember if any items require password
                if item.needs_password():
                    self._needs_password = True
            elif len(self._info_list) > 0:
                # final crc is in last block
                old = self._info_list[-1]
                old.CRC = item.CRC
                old.compress_size += item.compress_size

        # parse new-style comment
        if item.type == RAR_BLOCK_SUB and item.filename == 'CMT':
            if not NEED_COMMENTS:
                pass
            elif item.flags & (RAR_FILE_SPLIT_BEFORE | RAR_FILE_SPLIT_AFTER):
                pass
            elif item.flags & RAR_FILE_SOLID:
                # file comment
                cmt = self._read_comment_v3(item, self._password)
                if len(self._info_list) > 0:
                    old = self._info_list[-1]
                    old.comment = cmt
            else:
                # archive comment
                cmt = self._read_comment_v3(item, self._password)
                self.comment = cmt

        if self._info_callback:
            self._info_callback(item)

    # read rar
    def _parse(self):
        self._fd = None
        try:
            self._parse_real()
        finally:
            if self._fd:
                self._fd.close()
                self._fd = None

    def _parse_real(self):
        fd = open(self.rarfile, "rb")
        self._fd = fd
        id = fd.read(len(RAR_ID))
        if id != RAR_ID:
            raise NotRarFile("Not a Rar archive: "+self.rarfile)

        volume = 0  # first vol (.rar) is 0
        more_vols = 0
        endarc = 0
        volfile = self.rarfile
        while 1:
            if endarc:
                h = None    # don't read past ENDARC
            else:
                h = self._parse_header(fd)
            if not h:
                if more_vols:
                    volume += 1
                    volfile = self._next_volname(volfile)
                    fd.close()
                    fd = open(volfile, "rb")
                    self._fd = fd
                    more_vols = 0
                    endarc = 0
                    continue
                break
            h.volume = volume
            h.volume_file = volfile

            if h.type == RAR_BLOCK_MAIN and not self._main:
                self._main = h
                if h.flags & RAR_MAIN_NEWNUMBERING:
                    # RAR 2.x does not set FIRSTVOLUME,
                    # so check it only if NEWNUMBERING is used
                    if (h.flags & RAR_MAIN_FIRSTVOLUME) == 0:
                        raise NeedFirstVolume("Need to start from first volume")
                if h.flags & RAR_MAIN_PASSWORD:
                    self._needs_password = True
                    if not self._password:
                        self._main = None
                        break
            elif h.type == RAR_BLOCK_ENDARC:
                more_vols = h.flags & RAR_ENDARC_NEXT_VOLUME
                endarc = 1
            elif h.type == RAR_BLOCK_FILE:
                # RAR 2.x does not write RAR_BLOCK_ENDARC
                if h.flags & RAR_FILE_SPLIT_AFTER:
                    more_vols = 1
                # RAR 2.x does not set RAR_MAIN_FIRSTVOLUME
                if volume == 0 and h.flags & RAR_FILE_SPLIT_BEFORE:
                    raise NeedFirstVolume("Need to start from first volume")

            # store it
            self._process_entry(h)

            # go to next header
            if h.add_size > 0:
                fd.seek(h.file_offset + h.add_size, 0)

    # AES encrypted headers
    _last_aes_key = (None, None, None) # (salt, key, iv)
    def _decrypt_header(self, fd):
        if not _have_crypto:
            raise NoCrypto('Cannot parse encrypted headers - no crypto')
        salt = fd.read(8)
        if self._last_aes_key[0] == salt:
            key, iv = self._last_aes_key[1:]
        else:
            key, iv = rar3_s2k(self._password, salt)
            self._last_aes_key = (salt, key, iv)
        return HeaderDecrypt(fd, key, iv)

    # read single header
    def _parse_header(self, fd):
        try:
            # handle encrypted headers
            if self._main and self._main.flags & RAR_MAIN_PASSWORD:
                if not self._password:
                    return
                fd = self._decrypt_header(fd)

            # now read actual header
            return self._parse_block_header(fd)
        except struct.error:
            if REPORT_BAD_HEADER:
                raise BadRarFile('Broken header in RAR file')
            return None

    # common header
    def _parse_block_header(self, fd):
        h = RarInfo()
        h.header_offset = fd.tell()
        h.comment = None

        # read and parse base header
        buf = fd.read(S_BLK_HDR.size)
        if not buf:
            return None
        t = S_BLK_HDR.unpack_from(buf)
        h.header_crc, h.type, h.flags, h.header_size = t
        h.header_base = S_BLK_HDR.size
        pos = S_BLK_HDR.size

        # read full header
        if h.header_size > S_BLK_HDR.size:
            h.header_data = buf + fd.read(h.header_size - S_BLK_HDR.size)
        else:
            h.header_data = buf
        h.file_offset = fd.tell()

        # unexpected EOF?
        if len(h.header_data) != h.header_size:
            if REPORT_BAD_HEADER:
                raise BadRarFile('Unexpected EOF when reading header')
            return None

        # block has data assiciated with it?
        if h.flags & RAR_LONG_BLOCK:
            h.add_size = S_LONG.unpack_from(h.header_data, pos)[0]
        else:
            h.add_size = 0

        # parse interesting ones, decide header boundaries for crc
        if h.type == RAR_BLOCK_MARK:
            return h
        elif h.type == RAR_BLOCK_MAIN:
            h.header_base += 6
            if h.flags & RAR_MAIN_ENCRYPTVER:
                h.header_base += 1
            if h.flags & RAR_MAIN_COMMENT:
                self._parse_subblocks(h, h.header_base)
                self.comment = h.comment
        elif h.type == RAR_BLOCK_FILE:
            self._parse_file_header(h, pos)
        elif h.type == RAR_BLOCK_SUB:
            self._parse_file_header(h, pos)
            h.header_base = h.header_size
        elif h.type == RAR_BLOCK_OLD_AUTH:
            h.header_base += 8
        elif h.type == RAR_BLOCK_OLD_EXTRA:
            h.header_base += 7
        else:
            h.header_base = h.header_size

        # check crc
        if h.type == RAR_BLOCK_OLD_SUB:
            crcdat = h.header_data[2:] + fd.read(h.add_size)
        else:
            crcdat = h.header_data[2:h.header_base]

        calc_crc = crc32(crcdat) & 0xFFFF

        # return good header
        if h.header_crc == calc_crc:
            return h

        # need to panic?
        if REPORT_BAD_HEADER:
            xlen = len(crcdat)
            crcdat = h.header_data[2:]
            msg = 'Header CRC error (%02x): exp=%x got=%x (xlen = %d)' % ( h.type, h.header_crc, calc_crc, xlen )
            xlen = len(crcdat)
            while xlen >= S_BLK_HDR.size - 2:
                crc = crc32(crcdat[:xlen]) & 0xFFFF
                if crc == h.header_crc:
                    msg += ' / crc match, xlen = %d' % xlen
                xlen -= 1
            raise BadRarFile(msg)

        # instead panicing, send eof
        return None

    # read file-specific header
    def _parse_file_header(self, h, pos):
        fld = S_FILE_HDR.unpack_from(h.header_data, pos)
        h.compress_size = fld[0]
        h.file_size = fld[1]
        h.host_os = fld[2]
        h.CRC = fld[3]
        h.date_time = parse_dos_time(fld[4])
        h.extract_version = fld[5]
        h.compress_type = fld[6]
        h.name_size = fld[7]
        h.mode = fld[8]
        pos += S_FILE_HDR.size

        if h.flags & RAR_FILE_LARGE:
            h1 = S_LONG.unpack_from(h.header_data, pos)[0]
            h2 = S_LONG.unpack_from(h.header_data, pos + 4)[0]
            h.compress_size |= h1 << 32
            h.file_size |= h2 << 32
            pos += 8
            h.add_size = h.compress_size

        name = h.header_data[pos : pos + h.name_size ]
        pos += h.name_size
        if h.flags & RAR_FILE_UNICODE:
            nul = name.find(ZERO)
            h.orig_filename = name[:nul]
            u = UnicodeFilename(h.orig_filename, name[nul + 1 : ])
            h.filename = u.decode()

            # if parsing failed fall back to simple name
            if u.failed:
                h.filename = self._decode(h.orig_filename)
        else:
            h.orig_filename = name
            h.filename = self._decode(name)

        # change separator, if requested
        if PATH_SEP != '\\':
            h.filename = h.filename.replace('\\', PATH_SEP)

        if h.flags & RAR_FILE_SALT:
            h.salt = h.header_data[pos : pos + 8]
            pos += 8
        else:
            h.salt = None

        # optional extended time stamps
        if h.flags & RAR_FILE_EXTTIME:
            pos = self._parse_ext_time(h, pos)
        else:
            h.mtime = h.atime = h.ctime = h.arctime = None

        # base header end
        h.header_base = pos

        if h.flags & RAR_FILE_COMMENT:
            self._parse_subblocks(h, pos)

        # convert timestamps
        if USE_DATETIME:
            h.date_time = to_datetime(h.date_time)
            h.mtime = to_datetime(h.mtime)
            h.atime = to_datetime(h.atime)
            h.ctime = to_datetime(h.ctime)
            h.arctime = to_datetime(h.arctime)

        # .mtime is .date_time with more precision
        if h.mtime:
            if USE_DATETIME:
                h.date_time = h.mtime
            else:
                # keep seconds int
                h.date_time = h.mtime[:5] + (int(h.mtime[5]),)

        return pos

    # find old-style comment subblock
    def _parse_subblocks(self, h, pos):
        hdata = h.header_data
        while pos < len(hdata):
            # ordinary block header
            t = S_BLK_HDR.unpack_from(hdata, pos)
            scrc, stype, sflags, slen = t
            pos_next = pos + slen
            pos += S_BLK_HDR.size

            # corrupt header
            if pos_next < pos:
                break

            # followed by block-specific header
            if stype == RAR_BLOCK_OLD_COMMENT and pos + S_COMMENT_HDR.size <= pos_next:
                declen, ver, meth, crc = S_COMMENT_HDR.unpack_from(hdata, pos)
                pos += S_COMMENT_HDR.size
                data = hdata[pos : pos_next]
                cmt = rar_decompress(ver, meth, data, declen, sflags,
                                     crc, self._password)
                if not self._crc_check:
                    h.comment = self._decode_comment(cmt)
                elif crc32(cmt) & 0xFFFF == crc:
                    h.comment = self._decode_comment(cmt)

            pos = pos_next

    def _parse_ext_time(self, h, pos):
        data = h.header_data

        # flags and rest of data can be missing
        flags = 0
        if pos + 2 <= len(data):
            flags = S_SHORT.unpack_from(data, pos)[0]
            pos += 2

        h.mtime, pos = self._parse_xtime(flags >> 3*4, data, pos, h.date_time)
        h.ctime, pos = self._parse_xtime(flags >> 2*4, data, pos)
        h.atime, pos = self._parse_xtime(flags >> 1*4, data, pos)
        h.arctime, pos = self._parse_xtime(flags >> 0*4, data, pos)
        return pos

    def _parse_xtime(self, flag, data, pos, dostime = None):
        unit = 10000000.0 # 100 ns units
        if flag & 8:
            if not dostime:
                t = S_LONG.unpack_from(data, pos)[0]
                dostime = parse_dos_time(t)
                pos += 4
            rem = 0
            cnt = flag & 3
            for i in range(cnt):
                b = S_BYTE.unpack_from(data, pos)[0]
                rem = (b << 16) | (rem >> 8)
                pos += 1
            sec = dostime[5] + rem / unit
            if flag & 4:
                sec += 1
            dostime = dostime[:5] + (sec,)
        return dostime, pos

    # given current vol name, construct next one
    def _next_volname(self, volfile):
        if self._main.flags & RAR_MAIN_NEWNUMBERING:
            return self._next_newvol(volfile)
        return self._next_oldvol(volfile)

    # new-style next volume
    def _next_newvol(self, volfile):
        i = len(volfile) - 1
        while i >= 0:
            if volfile[i] >= '0' and volfile[i] <= '9':
                return self._inc_volname(volfile, i)
            i -= 1
        raise BadRarName("Cannot construct volume name: "+volfile)

    # old-style next volume
    def _next_oldvol(self, volfile):
        # rar -> r00
        if volfile[-4:].lower() == '.rar':
            return volfile[:-2] + '00'
        return self._inc_volname(volfile, len(volfile) - 1)

    # increase digits with carry, otherwise just increment char
    def _inc_volname(self, volfile, i):
        fn = list(volfile)
        while i >= 0:
            if fn[i] != '9':
                fn[i] = chr(ord(fn[i]) + 1)
                break
            fn[i] = '0'
            i -= 1
        return ''.join(fn)

    def _open_clear(self, inf):
        return DirectReader(self, inf)

    # put file compressed data into temporary .rar archive, and run
    # unrar on that, thus avoiding unrar going over whole archive
    def _open_hack(self, inf, psw = None):
        BSIZE = 32*1024

        size = inf.compress_size + inf.header_size
        rf = open(inf.volume_file, "rb", 0)
        rf.seek(inf.header_offset)

        tmpfd, tmpname = mkstemp(suffix='.rar')
        tmpf = os.fdopen(tmpfd, "wb")

        try:
            # create main header: crc, type, flags, size, res1, res2
            mh = S_BLK_HDR.pack(0x90CF, 0x73, 0, 13) + ZERO * (2+4)
            tmpf.write(RAR_ID + mh)
            while size > 0:
                if size > BSIZE:
                    buf = rf.read(BSIZE)
                else:
                    buf = rf.read(size)
                if not buf:
                    raise BadRarFile('read failed: ' + inf.filename)
                tmpf.write(buf)
                size -= len(buf)
            tmpf.close()
            rf.close()
        except:
            rf.close()
            tmpf.close()
            os.unlink(tmpname)
            raise

        return self._open_unrar(tmpname, inf, psw, tmpname)

    def _read_comment_v3(self, inf, psw=None):

        # read data
        rf = open(inf.volume_file, "rb")
        rf.seek(inf.file_offset)
        data = rf.read(inf.compress_size)
        rf.close()

        # decompress
        cmt = rar_decompress(inf.extract_version, inf.compress_type, data,
                             inf.file_size, inf.flags, inf.CRC, psw, inf.salt)

        # check crc
        if self._crc_check:
            crc = crc32(cmt)
            if crc < 0:
                crc += (long(1) << 32)
            if crc != inf.CRC:
                return None

        return self._decode_comment(cmt)

    # extract using unrar
    def _open_unrar(self, rarfile, inf, psw = None, tmpfile = None):
        cmd = [UNRAR_TOOL] + list(OPEN_ARGS)
        if psw is not None:
            cmd.append("-p" + psw)
        cmd.append(rarfile)

        # not giving filename avoids encoding related problems
        if not tmpfile:
            fn = inf.filename
            if PATH_SEP != os.sep:
                fn = fn.replace(PATH_SEP, os.sep)
            cmd.append(fn)

        # read from unrar pipe
        return PipeReader(self, inf, cmd, tmpfile)

    def _decode(self, val):
        for c in TRY_ENCODINGS:
            try:
                return val.decode(c)
            except UnicodeError:
                pass
        return val.decode(self._charset, 'replace')

    def _decode_comment(self, val):
        if UNICODE_COMMENTS:
            return self._decode(val)
        return val

    # call unrar to extract a file
    def _extract(self, fnlist, path=None, psw=None):
        cmd = [UNRAR_TOOL] + list(EXTRACT_ARGS)

        # pasoword
        psw = psw or self._password
        if psw is not None:
            cmd.append('-p' + psw)
        else:
            cmd.append('-p-')

        # rar file
        cmd.append(self.rarfile)

        # file list
        for fn in fnlist:
            if os.sep != PATH_SEP:
                fn = fn.replace(PATH_SEP, os.sep)
            cmd.append(fn)

        # destination path
        if path is not None:
            cmd.append(path + os.sep)

        # call
        p = custom_popen(cmd)
        p.communicate()

##
## Utility classes
##

class UnicodeFilename:
    """Handle unicode filename decompression"""

    def __init__(self, name, encdata):
        self.std_name = bytearray(name)
        self.encdata = bytearray(encdata)
        self.pos = self.encpos = 0
        self.buf = bytearray()
        self.failed = 0

    def enc_byte(self):
        try:
            c = self.encdata[self.encpos]
            self.encpos += 1
            return c
        except IndexError:
            self.failed = 1
            return 0

    def std_byte(self):
        try:
            return self.std_name[self.pos]
        except IndexError:
            self.failed = 1
            return ord('?')

    def put(self, lo, hi):
        self.buf.append(lo)
        self.buf.append(hi)
        self.pos += 1

    def decode(self):
        hi = self.enc_byte()
        flagbits = 0
        while self.encpos < len(self.encdata):
            if flagbits == 0:
                flags = self.enc_byte()
                flagbits = 8
            flagbits -= 2
            t = (flags >> flagbits) & 3
            if t == 0:
                self.put(self.enc_byte(), 0)
            elif t == 1:
                self.put(self.enc_byte(), hi)
            elif t == 2:
                self.put(self.enc_byte(), self.enc_byte())
            else:
                n = self.enc_byte()
                if n & 0x80:
                    c = self.enc_byte()
                    for i in range((n & 0x7f) + 2):
                        lo = (self.std_byte() + c) & 0xFF
                        self.put(lo, hi)
                else:
                    for i in range(n + 2):
                        self.put(self.std_byte(), 0)
        return self.buf.decode("utf-16le", "replace")


class RarExtFile(RawIOBase):
    """Base class for 'file-like' object that RarFile.open() returns.

    Provides public methods and common crc checking.

    Behaviour:
     - no short reads - .read() and .readinfo() read as much as requested.
     - no internal buffer, use io.BufferedReader for that.

    @ivar name:
      filename of the archive entry.
    """

    def __init__(self, rf, inf):
        """Fill common fields"""

        RawIOBase.__init__(self)

        # standard io.* properties
        self.name = inf.filename
        self.mode = 'rb'

        self.rf = rf
        self.inf = inf
        self.crc_check = rf._crc_check
        self.fd = None
        self.CRC = 0
        self.remain = 0

        self._open()

    def _open(self):
        if self.fd:
            self.fd.close()
        self.fd = None
        self.CRC = 0
        self.remain = self.inf.file_size

    def read(self, cnt = None):
        """Read all or specified amount of data from archive entry."""

        # sanitize cnt
        if cnt is None or cnt < 0:
            cnt = self.remain
        elif cnt > self.remain:
            cnt = self.remain
        if cnt == 0:
            return EMPTY

        # actual read
        data = self._read(cnt)
        if data:
            self.CRC = crc32(data, self.CRC)
            self.remain -= len(data)
        if len(data) != cnt:
            raise BadRarFile("Failed the read enough data")

        # done?
        if not data or self.remain == 0:
            #self.close()
            self._check()
        return data

    def _check(self):
        """Check final CRC."""
        if not self.crc_check:
            return
        if self.remain != 0:
            raise BadRarFile("Failed the read enough data")
        crc = self.CRC
        if crc < 0:
            crc += (long(1) << 32)
        if crc != self.inf.CRC:
            raise BadRarFile("Corrupt file - CRC check failed: " + self.inf.filename)

    def _read(self, cnt):
        """Actual read that gets sanitized cnt."""

    def close(self):
        """Close open resources."""

        RawIOBase.close(self)

        if self.fd:
            self.fd.close()
            self.fd = None

    def __del__(self):
        """Hook delete to make sure tempfile is removed."""
        self.close()

    def readinto(self, buf):
        """Zero-copy read directly into buffer.

        Returns bytes read.
        """

        data = self.read(len(buf))
        n = len(data)
        try:
            buf[:n] = data
        except TypeError:
            import array
            if not isinstance(buf, array.array):
                raise
            buf[:n] = array.array(buf.typecode, data)
        return n

    def tell(self):
        """Return current reading position in uncompressed data."""
        return self.inf.file_size - self.remain

    def seek(self, ofs, whence = 0):
        """Seek in data."""

        # disable crc check when seeking
        self.crc_check = 0

        fsize = self.inf.file_size
        cur_ofs = self.tell()

        if whence == 0:     # seek from beginning of file
            new_ofs = ofs
        elif whence == 1:   # seek from current position
            new_ofs = cur_ofs + ofs
        elif whence == 2:   # seek from end of file
            new_ofs = fsize + ofs
        else:
            raise ValueError('Invalid value for whence')

        # sanity check
        if new_ofs < 0:
            new_ofs = 0
        elif new_ofs > fsize:
            new_ofs = fsize

        # do the actual seek
        if new_ofs >= cur_ofs:
            self._skip(new_ofs - cur_ofs)
        else:
            # process old data ?
            #self._skip(fsize - cur_ofs)
            # reopen and seek
            self._open()
            self._skip(new_ofs)
        return self.tell()

    def _skip(self, cnt):
        """Read and discard data"""
        while cnt > 0:
            if cnt > 8192:
                buf = self.read(8192)
            else:
                buf = self.read(cnt)
            if not buf:
                break
            cnt -= len(buf)

    def readable(self):
        """Returns True"""
        return True

    def seekable(self):
        """Returns True"""
        return True

    def readall(self):
        """Read all remaining data"""
        # avoid RawIOBase default impl
        return self.read()


class PipeReader(RarExtFile):
    """Read data from pipe, handle tempfile cleanup."""

    def __init__(self, rf, inf, cmd, tempfile=None):
        self.cmd = cmd
        self.proc = None
        self.tempfile = tempfile
        RarExtFile.__init__(self, rf, inf)

    def _close_proc(self):
        if not self.proc:
            return
        if self.proc.stdout:
            self.proc.stdout.close()
        if self.proc.stdin:
            self.proc.stdin.close()
        if self.proc.stderr:
            self.proc.stderr.close()
        self.proc.wait()
        self.proc = None

    def _open(self):
        RarExtFile._open(self)

        # stop old process
        self._close_proc()

        # launch new process
        self.proc = custom_popen(self.cmd)
        self.fd = self.proc.stdout

        # avoid situation where unrar waits on stdin
        if self.proc.stdin:
            self.proc.stdin.close()

    def _read(self, cnt):
        """Read from pipe."""

        # normal read is usually enough
        data = self.fd.read(cnt)
        if len(data) == cnt or not data:
            return data

        # short read, try looping
        buf = [data]
        cnt -= len(data)
        while cnt > 0:
            data = self.fd.read(cnt)
            if not data:
                break
            cnt -= len(data)
            buf.append(data)
        return EMPTY.join(buf)

    def close(self):
        """Close open resources."""

        self._close_proc()
        RarExtFile.close(self)

        if self.tempfile:
            try:
                os.unlink(self.tempfile)
            except OSError:
                pass
            self.tempfile = None

    if have_memoryview:
        def readinto(self, buf):
            """Zero-copy read directly into buffer."""
            cnt = len(buf)
            if cnt > self.remain:
                cnt = self.remain
            vbuf = memoryview(buf)
            res = got = 0
            while got < cnt:
                res = self.fd.readinto(vbuf[got : cnt])
                if not res:
                    break
                if self.crc_check:
                    self.CRC = crc32(vbuf[got : got + res], self.CRC)
                self.remain -= res
                got += res
            return got


class DirectReader(RarExtFile):
    """Read uncompressed data directly from archive."""

    def _open(self):
        RarExtFile._open(self)

        self.volfile = self.inf.volume_file
        self.fd = open(self.volfile, "rb", 0)
        self.fd.seek(self.inf.header_offset, 0)
        self.cur = self.rf._parse_header(self.fd)
        self.cur_avail = self.cur.add_size

    def _skip(self, cnt):
        """RAR Seek, skipping through rar files to get to correct position
        """

        while cnt > 0:
            # next vol needed?
            if self.cur_avail == 0:
                if not self._open_next():
                    break

            # fd is in read pos, do the read
            if cnt > self.cur_avail:
                cnt -= self.cur_avail
                self.remain -= self.cur_avail
                self.cur_avail = 0
            else:
                self.fd.seek(cnt, 1)
                self.cur_avail -= cnt
                self.remain -= cnt
                cnt = 0

    def _read(self, cnt):
        """Read from potentially multi-volume archive."""

        buf = []
        while cnt > 0:
            # next vol needed?
            if self.cur_avail == 0:
                if not self._open_next():
                    break

            # fd is in read pos, do the read
            if cnt > self.cur_avail:
                data = self.fd.read(self.cur_avail)
            else:
                data = self.fd.read(cnt)
            if not data:
                break

            # got some data
            cnt -= len(data)
            self.cur_avail -= len(data)
            buf.append(data)

        if len(buf) == 1:
            return buf[0]
        return EMPTY.join(buf)

    def _open_next(self):
        """Proceed to next volume."""

        # is the file split over archives?
        if (self.cur.flags & RAR_FILE_SPLIT_AFTER) == 0:
            return False

        if self.fd:
            self.fd.close()
            self.fd = None

        # open next part
        self.volfile = self.rf._next_volname(self.volfile)
        fd = open(self.volfile, "rb", 0)
        self.fd = fd

        # loop until first file header
        while 1:
            cur = self.rf._parse_header(fd)
            if not cur:
                raise BadRarFile("Unexpected EOF")
            if cur.type in (RAR_BLOCK_MARK, RAR_BLOCK_MAIN):
                if cur.add_size:
                    fd.seek(cur.add_size, 1)
                continue
            if cur.orig_filename != self.inf.orig_filename:
                raise BadRarFile("Did not found file entry")
            self.cur = cur
            self.cur_avail = cur.add_size
            return True

    if have_memoryview:
        def readinto(self, buf):
            """Zero-copy read directly into buffer."""
            got = 0
            vbuf = memoryview(buf)
            while got < len(buf):
                # next vol needed?
                if self.cur_avail == 0:
                    if not self._open_next():
                        break

                # lenght for next read
                cnt = len(buf) - got
                if cnt > self.cur_avail:
                    cnt = self.cur_avail

                # read into temp view
                res = self.fd.readinto(vbuf[got : got + cnt])
                if not res:
                    break
                if self.crc_check:
                    self.CRC = crc32(vbuf[got : got + res], self.CRC)
                self.cur_avail -= res
                self.remain -= res
                got += res
            return got


class HeaderDecrypt:
    """File-like object that decrypts from another file"""
    def __init__(self, f, key, iv):
        self.f = f
        self.ciph = AES.new(key, AES.MODE_CBC, iv)
        self.buf = EMPTY

    def tell(self):
        return self.f.tell()

    def read(self, cnt=None):
        if cnt > 8*1024:
            raise BadRarFile('Bad count to header decrypt - wrong password?')

        # consume old data
        if cnt <= len(self.buf):
            res = self.buf[:cnt]
            self.buf = self.buf[cnt:]
            return res
        res = self.buf
        self.buf = EMPTY
        cnt -= len(res)

        # decrypt new data
        BLK = self.ciph.block_size
        while cnt > 0:
            enc = self.f.read(BLK)
            if len(enc) < BLK:
                break
            dec = self.ciph.decrypt(enc)
            if cnt >= len(dec):
                res += dec
                cnt -= len(dec)
            else:
                res += dec[:cnt]
                self.buf = dec[cnt:]
                cnt = 0

        return res

##
## Utility functions
##

def rar3_s2k(psw, salt):
    """String-to-key hash for RAR3."""

    seed = psw.encode('utf-16le') + salt
    iv = EMPTY
    h = sha1()
    for i in range(16):
        for j in range(0x4000):
            cnt = S_LONG.pack(i*0x4000 + j)
            h.update(seed + cnt[:3])
            if j == 0:
                iv += h.digest()[19:20]
    key_be = h.digest()[:16]
    key_le = pack("<LLLL", *unpack(">LLLL", key_be))
    return key_le, iv

def rar_decompress(vers, meth, data, declen=0, flags=0, crc=0, psw=None, salt=None):
    """Decompress blob of compressed data.

    Used for data with non-standard header - eg. comments.
    """

    # already uncompressed?
    if meth == RAR_M0 and (flags & RAR_FILE_PASSWORD) == 0:
        return data

    # take only necessary flags
    flags = flags & (RAR_FILE_PASSWORD | RAR_FILE_SALT | RAR_FILE_DICTMASK)
    flags |= RAR_LONG_BLOCK

    # file header
    fname = bytes('data', 'ascii')
    date = 0
    mode = 0x20
    fhdr = S_FILE_HDR.pack(len(data), declen, RAR_OS_MSDOS, crc,
                           date, vers, meth, len(fname), mode)
    fhdr += fname
    if flags & RAR_FILE_SALT:
        if not salt:
            return EMPTY
        fhdr += salt

    # full header
    hlen = S_BLK_HDR.size + len(fhdr)
    hdr = S_BLK_HDR.pack(0, RAR_BLOCK_FILE, flags, hlen) + fhdr
    hcrc = crc32(hdr[2:]) & 0xFFFF
    hdr = S_BLK_HDR.pack(hcrc, RAR_BLOCK_FILE, flags, hlen) + fhdr

    # archive main header
    mh = S_BLK_HDR.pack(0x90CF, RAR_BLOCK_MAIN, 0, 13) + ZERO * (2+4)

    # decompress via temp rar
    tmpfd, tmpname = mkstemp(suffix='.rar')
    tmpf = os.fdopen(tmpfd, "wb")
    try:
        tmpf.write(RAR_ID + mh + hdr + data)
        tmpf.close()

        cmd = [UNRAR_TOOL] + list(OPEN_ARGS)
        if psw is not None and (flags & RAR_FILE_PASSWORD):
            cmd.append("-p" + psw)
        else:
            cmd.append("-p-")
        cmd.append(tmpname)

        p = custom_popen(cmd)
        return p.communicate()[0]
    finally:
        tmpf.close()
        os.unlink(tmpname)

def to_datetime(t):
    """Convert 6-part time tuple into datetime object."""

    if t is None:
        return None

    # extract values
    year, mon, day, h, m, xs = t
    s = int(xs)
    us = int(1000000 * (xs - s))

    # assume the values are valid
    try:
        return datetime(year, mon, day, h, m, s, us)
    except ValueError:
        pass

    # sanitize invalid values
    MDAY = (0, 31, 29, 31, 30, 31, 30, 31, 31, 30, 31, 30, 31)
    if mon < 1:  mon = 1
    if mon > 12: mon = 12
    if day < 1:  day = 1
    if day > MDAY[mon]: day = MDAY[mon]
    if h > 23:   h = 23
    if m > 59:   m = 59
    if s > 59:   s = 59
    if mon == 2 and day == 29:
        try:
            return datetime(year, mon, day, h, m, s, us)
        except ValueError:
            day = 28
    return datetime(year, mon, day, h, m, s, us)

def parse_dos_time(stamp):
    """Parse standard 32-bit DOS timestamp."""

    sec = stamp & 0x1F; stamp = stamp >> 5
    min = stamp & 0x3F; stamp = stamp >> 6
    hr  = stamp & 0x1F; stamp = stamp >> 5
    day = stamp & 0x1F; stamp = stamp >> 5
    mon = stamp & 0x0F; stamp = stamp >> 4
    yr = (stamp & 0x7F) + 1980
    return (yr, mon, day, hr, min, sec * 2)

def custom_popen(cmd):
    """Disconnect cmd from parent fds, read only from stdout."""

    # needed for py2exe
    creationflags = 0
    if sys.platform == 'win32':
        creationflags = 0x08000000 # CREATE_NO_WINDOW

    # run command
    p = Popen(cmd, bufsize = 0, stdout = PIPE, stdin = PIPE, stderr = STDOUT,
              creationflags = creationflags)
    return p