1
# pack.py -- For dealing wih packed git objects.
2
# Copyright (C) 2007 James Westby <jw+debian@jameswestby.net>
3
# Copryight (C) 2008 Jelmer Vernooij <jelmer@samba.org>
4
# The code is loosely based on that in the sha1_file.c file from git itself,
5
# which is Copyright (C) Linus Torvalds, 2005 and distributed under the
8
# This program is free software; you can redistribute it and/or
9
# modify it under the terms of the GNU General Public License
10
# as published by the Free Software Foundation; version 2
13
# This program is distributed in the hope that it will be useful,
14
# but WITHOUT ANY WARRANTY; without even the implied warranty of
15
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
16
# GNU General Public License for more details.
18
# You should have received a copy of the GNU General Public License
19
# along with this program; if not, write to the Free Software
20
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston,
23
"""Classes for dealing with packed git objects.
25
A pack is a compact representation of a bunch of objects, stored
26
using deltas where possible.
28
They have two parts, the pack file, which stores the data, and an index
29
that tells you where the data is.
31
To find an object you look in all of the index files 'til you find a
32
match for the object name. You then use the pointer got from this as
33
a pointer in to the corresponding packfile.
40
supports_mmap_offset = (sys.version_info[0] >= 3 or
41
(sys.version_info[0] == 2 and sys.version_info[1] >= 6))
43
from objects import (ShaFile,
47
hex_to_sha = lambda hex: int(hex, 16)
49
MAX_MMAP_SIZE = 256 * 1024 * 1024
51
def simple_mmap(f, offset, size, access=mmap.ACCESS_READ):
52
if offset+size > MAX_MMAP_SIZE and not supports_mmap_offset:
53
raise AssertionError("%s is larger than 256 meg, and this version "
54
"of Python does not support the offset argument to mmap().")
55
if supports_mmap_offset:
56
return mmap.mmap(f.fileno(), size, access=access, offset=offset)
58
class ArraySkipper(object):
60
def __init__(self, array, offset):
64
def __getslice__(self, i, j):
65
return self.array[i+self.offset:j+self.offset]
67
def __getitem__(self, i):
68
return self.array[i+self.offset]
70
mem = mmap.mmap(f.fileno(), size, access=access)
73
return ArraySkipper(mem, offset)
76
def multi_ord(map, start, count):
78
for i in range(count):
79
value = value * 0x100 + ord(map[start+i])
82
class PackIndex(object):
83
"""An index in to a packfile.
85
Given a sha id of an object a pack index can tell you the location in the
86
packfile of that object if it has it.
88
To do the loop it opens the file, and indexes first 256 4 byte groups
89
with the first byte of the sha id. The value in the four byte group indexed
90
is the end of the group that shares the same starting byte. Subtract one
91
from the starting byte and index again to find the start of the group.
92
The values are sorted by sha id within the group, so do the math to find
93
the start and end offset and then bisect in to find if the value is present.
96
header_record_size = 4
97
header_size = 256 * header_record_size
100
record_size = sha_bytes + index_size
102
def __init__(self, filename):
103
"""Create a pack index object.
105
Provide it with the name of the index file to consider, and it will map
106
it whenever required.
108
self._filename = filename
109
assert os.path.exists(filename), "%s is not a pack index" % filename
110
# Take the size now, so it can be checked each time we map the file to
111
# ensure that it hasn't changed.
112
self._size = os.path.getsize(filename)
113
assert self._size > self.header_size, "%s is too small to be a packfile" % \
116
def object_index(self, sha):
117
"""Return the index in to the corresponding packfile for the object.
119
Given the name of an object it will return the offset that object lives
120
at within the corresponding pack file. If the pack file doesn't have the
121
object then None will be returned.
123
size = os.path.getsize(self._filename)
124
assert size == self._size, "Pack index %s has changed size, I don't " \
125
"like that" % self._filename
126
f = open(self._filename, 'rb')
128
map = simple_mmap(f, 0, size)
129
return self._object_index(map, sha)
133
def _object_index(self, map, hexsha):
134
"""See object_index"""
135
first_byte = hex_to_sha(hexsha[:2])
136
header_offset = self.header_record_size * first_byte
137
start = multi_ord(map, header_offset-self.header_record_size, self.header_record_size)
138
end = multi_ord(map, header_offset, self.header_record_size)
139
sha = hex_to_sha(hexsha)
142
offset = self.header_size + (i * self.record_size)
143
file_sha = multi_ord(map, offset + self.index_size, self.sha_bytes)
145
return multi_ord(map, offset, self.index_size)
153
class PackData(object):
154
"""The data contained in a packfile.
156
Pack files can be accessed both sequentially for exploding a pack, and
157
directly with the help of an index to retrieve a specific object.
159
The objects within are either complete or a delta aginst another.
161
The header is variable length. If the MSB of each byte is set then it
162
indicates that the subsequent byte is still part of the header.
163
For the first byte the next MS bits are the type, which tells you the type
164
of object, and whether it is a delta. The LS byte is the lowest bits of the
165
size. For each subsequent byte the LS 7 bits are the next MS bits of the
166
size, i.e. the last byte of the header contains the MS bits of the size.
168
For the complete objects the data is stored as zlib deflated data.
169
The size in the header is the uncompressed object size, so to uncompress
170
you need to just keep feeding data to zlib until you get an object back,
171
or it errors on bad data. This is done here by just giving the complete
172
buffer from the start of the deflated object on. This is bad, but until I
173
get mmap sorted out it will have to do.
175
Currently there are no integrity checks done. Also no attempt is made to try
176
and detect the delta case, or a request for an object at the wrong position.
177
It will all just throw a zlib or KeyError.
180
def __init__(self, filename):
181
"""Create a PackData object that represents the pack in the given filename.
183
The file must exist and stay readable until the object is disposed of. It
184
must also stay the same size. It will be mapped whenever needed.
186
Currently there is a restriction on the size of the pack as the python
187
mmap implementation is flawed.
189
self._filename = filename
190
assert os.path.exists(filename), "%s is not a packfile" % filename
191
self._size = os.path.getsize(filename)
193
def get_object_at(self, offset):
194
"""Given an offset in to the packfile return the object that is there.
196
Using the associated index the location of an object can be looked up, and
197
then the packfile can be asked directly for that object using this
200
Currently only non-delta objects are supported.
202
assert isinstance(offset, long) or isinstance(offset, int)
203
size = os.path.getsize(self._filename)
204
assert size == self._size, "Pack data %s has changed size, I don't " \
205
"like that" % self._filename
206
f = open(self._filename, 'rb')
208
map = simple_mmap(f, offset, size)
209
return self._get_object_at(map)
213
def _get_object_at(self, map):
214
first_byte = ord(map[0])
215
sign_extend = first_byte & 0x80
216
type = (first_byte >> 4) & 0x07
217
size = first_byte & 0x0f
219
while sign_extend > 0:
220
byte = ord(map[cur_offset+1])
221
sign_extend = byte & 0x80
222
size_part = byte & 0x7f
223
size += size_part << ((cur_offset * 7) + 4)
225
raw_base = cur_offset+1
226
# The size is the inflated size, so we have no idea what the deflated size
227
# is, so for now give it as much as we have. It should really iterate
228
# feeding it more data if it doesn't decompress, but as we have the whole
229
# thing then just use it.
231
uncomp = _decompress(raw)
232
obj = ShaFile.from_raw_string(type, uncomp)
added
removed
dulwich/pack.py
