Application: gvSIG desktop » gvSIG scripting

# pack.py -- For dealing with packed git objects.

# Copyright (C) 2007 James Westby <jw+debian@jameswestby.net>

# Copyright (C) 2008-2013 Jelmer Vernooij <jelmer@samba.org>

#

# Dulwich is dual-licensed under the Apache License, Version 2.0 and the GNU

# General Public License as public by the Free Software Foundation; version 2.0

# or (at your option) any later version. You can redistribute it and/or

# modify it under the terms of either of these two licenses.

#

# Unless required by applicable law or agreed to in writing, software

# distributed under the License is distributed on an "AS IS" BASIS,

# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.

# See the License for the specific language governing permissions and

# limitations under the License.

#

# You should have received a copy of the licenses; if not, see

# <http://www.gnu.org/licenses/> for a copy of the GNU General Public License

# and <http://www.apache.org/licenses/LICENSE-2.0> for a copy of the Apache

# License, Version 2.0.

#

"""Classes for dealing with packed git objects.

A pack is a compact representation of a bunch of objects, stored

using deltas where possible.

They have two parts, the pack file, which stores the data, and an index

that tells you where the data is.

To find an object you look in all of the index files 'til you find a

match for the object name. You then use the pointer got from this as

a pointer in to the corresponding packfile.

"""

from collections import defaultdict

import binascii

from io import BytesIO, UnsupportedOperation

from collections import (

    deque,

    )

import difflib

import struct

from itertools import chain

try:

    from itertools import imap, izip

except ImportError:

    # Python3

    imap = map

    izip = zip

import os

import sys

try:

    import mmap

except ImportError:

    has_mmap = False

else:

    has_mmap = True

# For some reason the above try, except fails to set has_mmap = False for plan9

if sys.platform == 'Plan9':

    has_mmap = False

from hashlib import sha1

from os import (

    SEEK_CUR,

    SEEK_END,

    )

from struct import unpack_from

import zlib

from dulwich.errors import (

    ApplyDeltaError,

    ChecksumMismatch,

    )

from dulwich.file import GitFile

from dulwich.lru_cache import (

    LRUSizeCache,

    )

from dulwich.objects import (

    ShaFile,

    hex_to_sha,

    sha_to_hex,

    object_header,

    )

OFS_DELTA = 6

REF_DELTA = 7

DELTA_TYPES = (OFS_DELTA, REF_DELTA)

DEFAULT_PACK_DELTA_WINDOW_SIZE = 10

def take_msb_bytes(read, crc32=None):

    """Read bytes marked with most significant bit.

    :param read: Read function

    """

    ret = []

    while len(ret) == 0 or ret[-1] & 0x80:

        b = read(1)

        if crc32 is not None:

            crc32 = binascii.crc32(b, crc32)

        ret.append(ord(b[:1]))

    return ret, crc32

class UnpackedObject(object):

    """Class encapsulating an object unpacked from a pack file.

    These objects should only be created from within unpack_object. Most

    members start out as empty and are filled in at various points by

    read_zlib_chunks, unpack_object, DeltaChainIterator, etc.

    End users of this object should take care that the function they're getting

    this object from is guaranteed to set the members they need.

    """

    __slots__ = [

      'offset',         # Offset in its pack.

      '_sha',           # Cached binary SHA.

      'obj_type_num',   # Type of this object.

      'obj_chunks',     # Decompressed and delta-resolved chunks.

      'pack_type_num',  # Type of this object in the pack (may be a delta).

      'delta_base',     # Delta base offset or SHA.

      'comp_chunks',    # Compressed object chunks.

      'decomp_chunks',  # Decompressed object chunks.

      'decomp_len',     # Decompressed length of this object.

      'crc32',          # CRC32.

      ]

    # TODO(dborowitz): read_zlib_chunks and unpack_object could very well be

    # methods of this object.

    def __init__(self, pack_type_num, delta_base, decomp_len, crc32):

        self.offset = None

        self._sha = None

        self.pack_type_num = pack_type_num

        self.delta_base = delta_base

        self.comp_chunks = None

        self.decomp_chunks = []

        self.decomp_len = decomp_len

        self.crc32 = crc32

        if pack_type_num in DELTA_TYPES:

            self.obj_type_num = None

            self.obj_chunks = None

        else:

            self.obj_type_num = pack_type_num

            self.obj_chunks = self.decomp_chunks

            self.delta_base = delta_base

    def sha(self):

        """Return the binary SHA of this object."""

        if self._sha is None:

            self._sha = obj_sha(self.obj_type_num, self.obj_chunks)

        return self._sha

    def sha_file(self):

        """Return a ShaFile from this object."""

        return ShaFile.from_raw_chunks(self.obj_type_num, self.obj_chunks)

    # Only provided for backwards compatibility with code that expects either

    # chunks or a delta tuple.

    def _obj(self):

        """Return the decompressed chunks, or (delta base, delta chunks)."""

        if self.pack_type_num in DELTA_TYPES:

            return (self.delta_base, self.decomp_chunks)

        else:

            return self.decomp_chunks

    def __eq__(self, other):

        if not isinstance(other, UnpackedObject):

            return False

        for slot in self.__slots__:

            if getattr(self, slot) != getattr(other, slot):

                return False

        return True

    def __ne__(self, other):

        return not (self == other)

    def __repr__(self):

        data = ['%s=%r' % (s, getattr(self, s)) for s in self.__slots__]

        return '%s(%s)' % (self.__class__.__name__, ', '.join(data))

_ZLIB_BUFSIZE = 4096

def read_zlib_chunks(read_some, unpacked, include_comp=False,

                     buffer_size=_ZLIB_BUFSIZE):

    """Read zlib data from a buffer.

    This function requires that the buffer have additional data following the

    compressed data, which is guaranteed to be the case for git pack files.

    :param read_some: Read function that returns at least one byte, but may

        return less than the requested size.

    :param unpacked: An UnpackedObject to write result data to. If its crc32

        attr is not None, the CRC32 of the compressed bytes will be computed

        using this starting CRC32.

        After this function, will have the following attrs set:

        * comp_chunks    (if include_comp is True)

        * decomp_chunks

        * decomp_len

        * crc32

    :param include_comp: If True, include compressed data in the result.

    :param buffer_size: Size of the read buffer.

    :return: Leftover unused data from the decompression.

    :raise zlib.error: if a decompression error occurred.

    """

    if unpacked.decomp_len <= -1:

        raise ValueError('non-negative zlib data stream size expected')

    decomp_obj = zlib.decompressobj()

    comp_chunks = []

    decomp_chunks = unpacked.decomp_chunks

    decomp_len = 0

    crc32 = unpacked.crc32

    while True:

        add = read_some(buffer_size)

        if not add:

            raise zlib.error('EOF before end of zlib stream')

        comp_chunks.append(add)

        decomp = decomp_obj.decompress(add)

        decomp_len += len(decomp)

        decomp_chunks.append(decomp)

        unused = decomp_obj.unused_data

        if unused:

            left = len(unused)

            if crc32 is not None:

                crc32 = binascii.crc32(add[:-left], crc32)

            if include_comp:

                comp_chunks[-1] = add[:-left]

            break

        elif crc32 is not None:

            crc32 = binascii.crc32(add, crc32)

    if crc32 is not None:

        crc32 &= 0xffffffff

    if decomp_len != unpacked.decomp_len:

        raise zlib.error('decompressed data does not match expected size')

    unpacked.crc32 = crc32

    if include_comp:

        unpacked.comp_chunks = comp_chunks

    return unused

def iter_sha1(iter):

    """Return the hexdigest of the SHA1 over a set of names.

    :param iter: Iterator over string objects

    :return: 40-byte hex sha1 digest

    """

    sha = sha1()

    for name in iter:

        sha.update(name)

    return sha.hexdigest().encode('ascii')

def load_pack_index(path):

    """Load an index file by path.

    :param filename: Path to the index file

    :return: A PackIndex loaded from the given path

    """

    with GitFile(path, 'rb') as f:

        return load_pack_index_file(path, f)

def _load_file_contents(f, size=None):

    try:

        fd = f.fileno()

    except (UnsupportedOperation, AttributeError):

        fd = None

    # Attempt to use mmap if possible

    if fd is not None:

        if size is None:

            size = os.fstat(fd).st_size

        if has_mmap:

            try:

                contents = mmap.mmap(fd, size, access=mmap.ACCESS_READ)

            except mmap.error:

                # Perhaps a socket?

                pass

            else:

                return contents, size

    contents = f.read()

    size = len(contents)

    return contents, size

def load_pack_index_file(path, f):

    """Load an index file from a file-like object.

    :param path: Path for the index file

    :param f: File-like object

    :return: A PackIndex loaded from the given file

    """

    contents, size = _load_file_contents(f)

    if contents[:4] == b'\377tOc':

        version = struct.unpack(b'>L', contents[4:8])[0]

        if version == 2:

            return PackIndex2(path, file=f, contents=contents,

                size=size)

        else:

            raise KeyError('Unknown pack index format %d' % version)

    else:

        return PackIndex1(path, file=f, contents=contents, size=size)

def bisect_find_sha(start, end, sha, unpack_name):

    """Find a SHA in a data blob with sorted SHAs.

    :param start: Start index of range to search

    :param end: End index of range to search

    :param sha: Sha to find

    :param unpack_name: Callback to retrieve SHA by index

    :return: Index of the SHA, or None if it wasn't found

    """

    assert start <= end

    while start <= end:

        i = (start + end) // 2

        file_sha = unpack_name(i)

        if file_sha < sha:

            start = i + 1

        elif file_sha > sha:

            end = i - 1

        else:

            return i

    return None

class PackIndex(object):

    """An index in to a packfile.

    Given a sha id of an object a pack index can tell you the location in the

    packfile of that object if it has it.

    """

    def __eq__(self, other):

        if not isinstance(other, PackIndex):

            return False

        for (name1, _, _), (name2, _, _) in izip(self.iterentries(),

                                                 other.iterentries()):

            if name1 != name2:

                return False

        return True

    def __ne__(self, other):

        return not self.__eq__(other)

    def __len__(self):

        """Return the number of entries in this pack index."""

        raise NotImplementedError(self.__len__)

    def __iter__(self):

        """Iterate over the SHAs in this pack."""

        return imap(sha_to_hex, self._itersha())

    def iterentries(self):

        """Iterate over the entries in this pack index.

        :return: iterator over tuples with object name, offset in packfile and

            crc32 checksum.

        """

        raise NotImplementedError(self.iterentries)

    def get_pack_checksum(self):

        """Return the SHA1 checksum stored for the corresponding packfile.

        :return: 20-byte binary digest

        """

        raise NotImplementedError(self.get_pack_checksum)

    def object_index(self, sha):

        """Return the index in to the corresponding packfile for the object.

        Given the name of an object it will return the offset that object

        lives at within the corresponding pack file. If the pack file doesn't

        have the object then None will be returned.

        """

        if len(sha) == 40:

            sha = hex_to_sha(sha)

        return self._object_index(sha)

    def _object_index(self, sha):

        """See object_index.

        :param sha: A *binary* SHA string. (20 characters long)_

        """

        raise NotImplementedError(self._object_index)

    def objects_sha1(self):

        """Return the hex SHA1 over all the shas of all objects in this pack.

        :note: This is used for the filename of the pack.

        """

        return iter_sha1(self._itersha())

    def _itersha(self):

        """Yield all the SHA1's of the objects in the index, sorted."""

        raise NotImplementedError(self._itersha)

class MemoryPackIndex(PackIndex):

    """Pack index that is stored entirely in memory."""

    def __init__(self, entries, pack_checksum=None):

        """Create a new MemoryPackIndex.

        :param entries: Sequence of name, idx, crc32 (sorted)

        :param pack_checksum: Optional pack checksum

        """

        self._by_sha = {}

        for name, idx, crc32 in entries:

            self._by_sha[name] = idx

        self._entries = entries

        self._pack_checksum = pack_checksum

    def get_pack_checksum(self):

        return self._pack_checksum

    def __len__(self):

        return len(self._entries)

    def _object_index(self, sha):

        return self._by_sha[sha][0]

    def _itersha(self):

        return iter(self._by_sha)

    def iterentries(self):

        return iter(self._entries)

class FilePackIndex(PackIndex):

    """Pack index that is based on a file.

    To do the loop it opens the file, and indexes first 256 4 byte groups

    with the first byte of the sha id. The value in the four byte group indexed

    is the end of the group that shares the same starting byte. Subtract one

    from the starting byte and index again to find the start of the group.

    The values are sorted by sha id within the group, so do the math to find

    the start and end offset and then bisect in to find if the value is present.

    """

    def __init__(self, filename, file=None, contents=None, size=None):

        """Create a pack index object.

        Provide it with the name of the index file to consider, and it will map

        it whenever required.

        """

        self._filename = filename

        # Take the size now, so it can be checked each time we map the file to

        # ensure that it hasn't changed.

        if file is None:

            self._file = GitFile(filename, 'rb')

        else:

            self._file = file

        if contents is None:

            self._contents, self._size = _load_file_contents(self._file, size)

        else:

            self._contents, self._size = (contents, size)

    def __eq__(self, other):

        # Quick optimization:

        if (isinstance(other, FilePackIndex) and

            self._fan_out_table != other._fan_out_table):

            return False

        return super(FilePackIndex, self).__eq__(other)

    def close(self):

        self._file.close()

        if getattr(self._contents, "close", None) is not None:

            self._contents.close()

    def __len__(self):

        """Return the number of entries in this pack index."""

        return self._fan_out_table[-1]

    def _unpack_entry(self, i):

        """Unpack the i-th entry in the index file.

        :return: Tuple with object name (SHA), offset in pack file and CRC32

            checksum (if known).

        """

        raise NotImplementedError(self._unpack_entry)

    def _unpack_name(self, i):

        """Unpack the i-th name from the index file."""

        raise NotImplementedError(self._unpack_name)

    def _unpack_offset(self, i):

        """Unpack the i-th object offset from the index file."""

        raise NotImplementedError(self._unpack_offset)

    def _unpack_crc32_checksum(self, i):

        """Unpack the crc32 checksum for the i-th object from the index file."""

        raise NotImplementedError(self._unpack_crc32_checksum)

    def _itersha(self):

        for i in range(len(self)):

            yield self._unpack_name(i)

    def iterentries(self):

        """Iterate over the entries in this pack index.

        :return: iterator over tuples with object name, offset in packfile and

            crc32 checksum.

        """

        for i in range(len(self)):

            yield self._unpack_entry(i)

    def _read_fan_out_table(self, start_offset):

        ret = []

        for i in range(0x100):

            fanout_entry = self._contents[start_offset+i*4:start_offset+(i+1)*4]

            ret.append(struct.unpack('>L', fanout_entry)[0])

        return ret

    def check(self):

        """Check that the stored checksum matches the actual checksum."""

        actual = self.calculate_checksum()

        stored = self.get_stored_checksum()

        if actual != stored:

            raise ChecksumMismatch(stored, actual)

    def calculate_checksum(self):

        """Calculate the SHA1 checksum over this pack index.

        :return: This is a 20-byte binary digest

        """

        return sha1(self._contents[:-20]).digest()

    def get_pack_checksum(self):

        """Return the SHA1 checksum stored for the corresponding packfile.

        :return: 20-byte binary digest

        """

        return bytes(self._contents[-40:-20])

    def get_stored_checksum(self):

        """Return the SHA1 checksum stored for this index.

        :return: 20-byte binary digest

        """

        return bytes(self._contents[-20:])

    def _object_index(self, sha):

        """See object_index.

        :param sha: A *binary* SHA string. (20 characters long)_

        """

        assert len(sha) == 20

        idx = ord(sha[:1])

        if idx == 0:

            start = 0

        else:

            start = self._fan_out_table[idx-1]

        end = self._fan_out_table[idx]

        i = bisect_find_sha(start, end, sha, self._unpack_name)

        if i is None:

            raise KeyError(sha)

        return self._unpack_offset(i)

class PackIndex1(FilePackIndex):

    """Version 1 Pack Index file."""

    def __init__(self, filename, file=None, contents=None, size=None):

        super(PackIndex1, self).__init__(filename, file, contents, size)

        self.version = 1

        self._fan_out_table = self._read_fan_out_table(0)

    def _unpack_entry(self, i):

        (offset, name) = unpack_from('>L20s', self._contents,

                                     (0x100 * 4) + (i * 24))

        return (name, offset, None)

    def _unpack_name(self, i):

        offset = (0x100 * 4) + (i * 24) + 4

        return self._contents[offset:offset+20]

    def _unpack_offset(self, i):

        offset = (0x100 * 4) + (i * 24)

        return unpack_from('>L', self._contents, offset)[0]

    def _unpack_crc32_checksum(self, i):

        # Not stored in v1 index files

        return None

class PackIndex2(FilePackIndex):

    """Version 2 Pack Index file."""

    def __init__(self, filename, file=None, contents=None, size=None):

        super(PackIndex2, self).__init__(filename, file, contents, size)

        if self._contents[:4] != b'\377tOc':

            raise AssertionError('Not a v2 pack index file')

        (self.version, ) = unpack_from(b'>L', self._contents, 4)

        if self.version != 2:

            raise AssertionError('Version was %d' % self.version)

        self._fan_out_table = self._read_fan_out_table(8)

        self._name_table_offset = 8 + 0x100 * 4

        self._crc32_table_offset = self._name_table_offset + 20 * len(self)

        self._pack_offset_table_offset = (self._crc32_table_offset +

                                          4 * len(self))

        self._pack_offset_largetable_offset = (self._pack_offset_table_offset +

                                          4 * len(self))

    def _unpack_entry(self, i):

        return (self._unpack_name(i), self._unpack_offset(i),

                self._unpack_crc32_checksum(i))

    def _unpack_name(self, i):

        offset = self._name_table_offset + i * 20

        return self._contents[offset:offset+20]

    def _unpack_offset(self, i):

        offset = self._pack_offset_table_offset + i * 4

        offset = unpack_from('>L', self._contents, offset)[0]

        if offset & (2**31):

            offset = self._pack_offset_largetable_offset + (offset&(2**31-1)) * 8

            offset = unpack_from('>Q', self._contents, offset)[0]

        return offset

    def _unpack_crc32_checksum(self, i):

        return unpack_from('>L', self._contents,

                          self._crc32_table_offset + i * 4)[0]

def read_pack_header(read):

    """Read the header of a pack file.

    :param read: Read function

    :return: Tuple of (pack version, number of objects). If no data is available

        to read, returns (None, None).

    """

    header = read(12)

    if not header:

        return None, None

    if header[:4] != b'PACK':

        raise AssertionError('Invalid pack header %r' % header)

    (version,) = unpack_from(b'>L', header, 4)

    if version not in (2, 3):

        raise AssertionError('Version was %d' % version)

    (num_objects,) = unpack_from(b'>L', header, 8)

    return (version, num_objects)

def chunks_length(chunks):

    if isinstance(chunks, bytes):

        return len(chunks)

    else:

        return sum(imap(len, chunks))

def unpack_object(read_all, read_some=None, compute_crc32=False,

                  include_comp=False, zlib_bufsize=_ZLIB_BUFSIZE):

    """Unpack a Git object.

    :param read_all: Read function that blocks until the number of requested

        bytes are read.

    :param read_some: Read function that returns at least one byte, but may not

        return the number of bytes requested.

    :param compute_crc32: If True, compute the CRC32 of the compressed data. If

        False, the returned CRC32 will be None.

    :param include_comp: If True, include compressed data in the result.

    :param zlib_bufsize: An optional buffer size for zlib operations.

    :return: A tuple of (unpacked, unused), where unused is the unused data

        leftover from decompression, and unpacked in an UnpackedObject with

        the following attrs set:

        * obj_chunks     (for non-delta types)

        * pack_type_num

        * delta_base     (for delta types)

        * comp_chunks    (if include_comp is True)

        * decomp_chunks

        * decomp_len

        * crc32          (if compute_crc32 is True)

    """

    if read_some is None:

        read_some = read_all

    if compute_crc32:

        crc32 = 0

    else:

        crc32 = None

    bytes, crc32 = take_msb_bytes(read_all, crc32=crc32)

    type_num = (bytes[0] >> 4) & 0x07

    size = bytes[0] & 0x0f

    for i, byte in enumerate(bytes[1:]):

        size += (byte & 0x7f) << ((i * 7) + 4)

    raw_base = len(bytes)

    if type_num == OFS_DELTA:

        bytes, crc32 = take_msb_bytes(read_all, crc32=crc32)

        raw_base += len(bytes)

        if bytes[-1] & 0x80:

            raise AssertionError

        delta_base_offset = bytes[0] & 0x7f

        for byte in bytes[1:]:

            delta_base_offset += 1

            delta_base_offset <<= 7

            delta_base_offset += (byte & 0x7f)

        delta_base = delta_base_offset

    elif type_num == REF_DELTA:

        delta_base = read_all(20)

        if compute_crc32:

            crc32 = binascii.crc32(delta_base, crc32)

        raw_base += 20

    else:

        delta_base = None

    unpacked = UnpackedObject(type_num, delta_base, size, crc32)

    unused = read_zlib_chunks(read_some, unpacked, buffer_size=zlib_bufsize,

                              include_comp=include_comp)

    return unpacked, unused

def _compute_object_size(value):

    """Compute the size of a unresolved object for use with LRUSizeCache."""

    (num, obj) = value

    if num in DELTA_TYPES:

        return chunks_length(obj[1])

    return chunks_length(obj)

class PackStreamReader(object):

    """Class to read a pack stream.

    The pack is read from a ReceivableProtocol using read() or recv() as

    appropriate.

    """

    def __init__(self, read_all, read_some=None, zlib_bufsize=_ZLIB_BUFSIZE):

        self.read_all = read_all

        if read_some is None:

            self.read_some = read_all

        else:

            self.read_some = read_some

        self.sha = sha1()

        self._offset = 0

        self._rbuf = BytesIO()

        # trailer is a deque to avoid memory allocation on small reads

        self._trailer = deque()

        self._zlib_bufsize = zlib_bufsize

    def _read(self, read, size):

        """Read up to size bytes using the given callback.

        As a side effect, update the verifier's hash (excluding the last 20

        bytes read).

        :param read: The read callback to read from.

        :param size: The maximum number of bytes to read; the particular

            behavior is callback-specific.

        """

        data = read(size)

        # maintain a trailer of the last 20 bytes we've read

        n = len(data)

        self._offset += n

        tn = len(self._trailer)

        if n >= 20:

            to_pop = tn

            to_add = 20

        else:

            to_pop = max(n + tn - 20, 0)

            to_add = n

        self.sha.update(bytes(bytearray([self._trailer.popleft() for _ in range(to_pop)])))

        self._trailer.extend(data[-to_add:])

        # hash everything but the trailer

        self.sha.update(data[:-to_add])

        return data

    def _buf_len(self):

        buf = self._rbuf

        start = buf.tell()

        buf.seek(0, SEEK_END)

        end = buf.tell()

        buf.seek(start)

        return end - start

    @property

    def offset(self):

        return self._offset - self._buf_len()

    def read(self, size):

        """Read, blocking until size bytes are read."""

        buf_len = self._buf_len()

        if buf_len >= size:

            return self._rbuf.read(size)

        buf_data = self._rbuf.read()

        self._rbuf = BytesIO()

        return buf_data + self._read(self.read_all, size - buf_len)

    def recv(self, size):

        """Read up to size bytes, blocking until one byte is read."""

        buf_len = self._buf_len()

        if buf_len:

            data = self._rbuf.read(size)

            if size >= buf_len:

                self._rbuf = BytesIO()

            return data

        return self._read(self.read_some, size)

    def __len__(self):

        return self._num_objects

    def read_objects(self, compute_crc32=False):

        """Read the objects in this pack file.

        :param compute_crc32: If True, compute the CRC32 of the compressed

            data. If False, the returned CRC32 will be None.

        :return: Iterator over UnpackedObjects with the following members set:

            offset

            obj_type_num

            obj_chunks (for non-delta types)

            delta_base (for delta types)

            decomp_chunks

            decomp_len

            crc32 (if compute_crc32 is True)

        :raise ChecksumMismatch: if the checksum of the pack contents does not

            match the checksum in the pack trailer.

        :raise zlib.error: if an error occurred during zlib decompression.

        :raise IOError: if an error occurred writing to the output file.

        """

        pack_version, self._num_objects = read_pack_header(self.read)

        if pack_version is None:

            return

        for i in range(self._num_objects):

            offset = self.offset

            unpacked, unused = unpack_object(

              self.read, read_some=self.recv, compute_crc32=compute_crc32,

              zlib_bufsize=self._zlib_bufsize)

            unpacked.offset = offset

            # prepend any unused data to current read buffer

            buf = BytesIO()

            buf.write(unused)

            buf.write(self._rbuf.read())

            buf.seek(0)

            self._rbuf = buf

            yield unpacked

        if self._buf_len() < 20:

            # If the read buffer is full, then the last read() got the whole

            # trailer off the wire. If not, it means there is still some of the

            # trailer to read. We need to read() all 20 bytes; N come from the

            # read buffer and (20 - N) come from the wire.

            self.read(20)

        pack_sha = bytearray(self._trailer)

        if pack_sha != self.sha.digest():

            raise ChecksumMismatch(sha_to_hex(pack_sha), self.sha.hexdigest())

class PackStreamCopier(PackStreamReader):

    """Class to verify a pack stream as it is being read.

    The pack is read from a ReceivableProtocol using read() or recv() as

    appropriate and written out to the given file-like object.

    """

    def __init__(self, read_all, read_some, outfile, delta_iter=None):

        """Initialize the copier.

        :param read_all: Read function that blocks until the number of requested

            bytes are read.

        :param read_some: Read function that returns at least one byte, but may

            not return the number of bytes requested.

        :param outfile: File-like object to write output through.

        :param delta_iter: Optional DeltaChainIterator to record deltas as we

            read them.

        """

        super(PackStreamCopier, self).__init__(read_all, read_some=read_some)

        self.outfile = outfile

        self._delta_iter = delta_iter

    def _read(self, read, size):

        """Read data from the read callback and write it to the file."""

        data = super(PackStreamCopier, self)._read(read, size)

        self.outfile.write(data)

        return data

    def verify(self):

        """Verify a pack stream and write it to the output file.

        See PackStreamReader.iterobjects for a list of exceptions this may

        throw.

        """

        if self._delta_iter:

            for unpacked in self.read_objects():

                self._delta_iter.record(unpacked)

        else:

            for _ in self.read_objects():

                pass

def obj_sha(type, chunks):

    """Compute the SHA for a numeric type and object chunks."""

    sha = sha1()

    sha.update(object_header(type, chunks_length(chunks)))

    if isinstance(chunks, bytes):

        sha.update(chunks)

    else:

        for chunk in chunks:

            sha.update(chunk)

    return sha.digest()

def compute_file_sha(f, start_ofs=0, end_ofs=0, buffer_size=1<<16):

    """Hash a portion of a file into a new SHA.

    :param f: A file-like object to read from that supports seek().

    :param start_ofs: The offset in the file to start reading at.

    :param end_ofs: The offset in the file to end reading at, relative to the

        end of the file.

    :param buffer_size: A buffer size for reading.

    :return: A new SHA object updated with data read from the file.

    """

    sha = sha1()

    f.seek(0, SEEK_END)

    length = f.tell()

    if (end_ofs < 0 and length + end_ofs < start_ofs) or end_ofs > length:

        raise AssertionError(

            "Attempt to read beyond file length. "

            "start_ofs: %d, end_ofs: %d, file length: %d" % (

                start_ofs, end_ofs, length))

    todo = length + end_ofs - start_ofs

    f.seek(start_ofs)

    while todo:

        data = f.read(min(todo, buffer_size))

        sha.update(data)

        todo -= len(data)

    return sha

class PackData(object):

    """The data contained in a packfile.

    Pack files can be accessed both sequentially for exploding a pack, and

    directly with the help of an index to retrieve a specific object.

    The objects within are either complete or a delta against another.

    The header is variable length. If the MSB of each byte is set then it

    indicates that the subsequent byte is still part of the header.

    For the first byte the next MS bits are the type, which tells you the type

    of object, and whether it is a delta. The LS byte is the lowest bits of the

    size. For each subsequent byte the LS 7 bits are the next MS bits of the

    size, i.e. the last byte of the header contains the MS bits of the size.

    For the complete objects the data is stored as zlib deflated data.

    The size in the header is the uncompressed object size, so to uncompress

    you need to just keep feeding data to zlib until you get an object back,

    or it errors on bad data. This is done here by just giving the complete

    buffer from the start of the deflated object on. This is bad, but until I

    get mmap sorted out it will have to do.

    Currently there are no integrity checks done. Also no attempt is made to

    try and detect the delta case, or a request for an object at the wrong

    position.  It will all just throw a zlib or KeyError.

    """

    def __init__(self, filename, file=None, size=None):

        """Create a PackData object representing the pack in the given filename.

        The file must exist and stay readable until the object is disposed of. It

        must also stay the same size. It will be mapped whenever needed.

        Currently there is a restriction on the size of the pack as the python

        mmap implementation is flawed.

        """

        self._filename = filename

        self._size = size

        self._header_size = 12

        if file is None:

            self._file = GitFile(self._filename, 'rb')

        else:

            self._file = file

        (version, self._num_objects) = read_pack_header(self._file.read)

        self._offset_cache = LRUSizeCache(1024*1024*20,

            compute_size=_compute_object_size)

        self.pack = None

    @property

    def filename(self):

        return os.path.basename(self._filename)

    @classmethod

    def from_file(cls, file, size):

        return cls(str(file), file=file, size=size)

    @classmethod

    def from_path(cls, path):

        return cls(filename=path)

    def close(self):

        self._file.close()

    def __enter__(self):

        return self

    def __exit__(self, exc_type, exc_val, exc_tb):

        self.close()

    def _get_size(self):

        if self._size is not None:

            return self._size

        self._size = os.path.getsize(self._filename)

        if self._size < self._header_size:

            errmsg = ('%s is too small for a packfile (%d < %d)' %

                      (self._filename, self._size, self._header_size))

            raise AssertionError(errmsg)

        return self._size

    def __len__(self):

        """Returns the number of objects in this pack."""

        return self._num_objects

    def calculate_checksum(self):

        """Calculate the checksum for this pack.

        :return: 20-byte binary SHA1 digest

        """

        return compute_file_sha(self._file, end_ofs=-20).digest()

    def get_ref(self, sha):

        """Get the object for a ref SHA, only looking in this pack."""

        # TODO: cache these results

        if self.pack is None:

            raise KeyError(sha)

        try:

            offset = self.pack.index.object_index(sha)

        except KeyError:

            offset = None

        if offset:

            type, obj = self.get_object_at(offset)

        elif self.pack is not None and self.pack.resolve_ext_ref:

            type, obj = self.pack.resolve_ext_ref(sha)

        else:

            raise KeyError(sha)

        return offset, type, obj

    def resolve_object(self, offset, type, obj, get_ref=None):

        """Resolve an object, possibly resolving deltas when necessary.

        :return: Tuple with object type and contents.

        """

        # Walk down the delta chain, building a stack of deltas to reach

        # the requested object.

        base_offset = offset

        base_type = type

        base_obj = obj

        delta_stack = []

        while base_type in DELTA_TYPES:

            prev_offset = base_offset

            if get_ref is None:

                get_ref = self.get_ref

            if base_type == OFS_DELTA:

                (delta_offset, delta) = base_obj

                # TODO: clean up asserts and replace with nicer error messages

                assert (

                    isinstance(base_offset, int)

                    or isinstance(base_offset, long))

                assert (

                    isinstance(delta_offset, int)

                    or isinstance(base_offset, long))

                base_offset = base_offset - delta_offset

                base_type, base_obj = self.get_object_at(base_offset)

                assert isinstance(base_type, int)

            elif base_type == REF_DELTA:

                (basename, delta) = base_obj