14
14
# You should have received a copy of the GNU General Public License
15
15
# along with this program; if not, write to the Free Software
16
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
16
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
18
18
"""Bzrlib specific gzip tunings. We plan to feed these to the upstream gzip."""
20
from __future__ import absolute_import
22
from cStringIO import StringIO
20
24
# make GzipFile faster:
22
from gzip import U32, LOWU32, FEXTRA, FCOMMENT, FNAME, FHCRC
26
from gzip import FEXTRA, FCOMMENT, FNAME, FHCRC
27
__all__ = ["GzipFile"]
31
# we want a \n preserved, break on \n only splitlines.
32
from bzrlib import symbol_versioning
34
__all__ = ["GzipFile", "bytes_to_gzip"]
38
"""Return i as an unsigned integer, assuming it fits in 32 bits.
40
If it's >= 2GB when viewed as a 32-bit unsigned int, return a long.
48
"""Return the low-order 32 bits of an int, as a non-negative int."""
49
return i & 0xFFFFFFFFL
52
def bytes_to_gzip(bytes, factory=zlib.compressobj,
53
level=zlib.Z_DEFAULT_COMPRESSION, method=zlib.DEFLATED,
54
width=-zlib.MAX_WBITS, mem=zlib.DEF_MEM_LEVEL,
56
"""Create a gzip file containing bytes and return its content."""
57
return chunks_to_gzip([bytes])
60
def chunks_to_gzip(chunks, factory=zlib.compressobj,
61
level=zlib.Z_DEFAULT_COMPRESSION, method=zlib.DEFLATED,
62
width=-zlib.MAX_WBITS, mem=zlib.DEF_MEM_LEVEL,
64
"""Create a gzip file containing chunks and return its content.
66
:param chunks: An iterable of strings. Each string can have arbitrary
70
'\037\213' # self.fileobj.write('\037\213') # magic header
71
'\010' # self.fileobj.write('\010') # compression method
72
# fname = self.filename[:-3]
76
'\x00' # self.fileobj.write(chr(flags))
77
'\0\0\0\0' # write32u(self.fileobj, long(time.time()))
78
'\002' # self.fileobj.write('\002')
79
'\377' # self.fileobj.write('\377')
81
'' # self.fileobj.write(fname + '\000')
83
# using a compressobj avoids a small header and trailer that the compress()
84
# utility function adds.
85
compress = factory(level, method, width, mem, 0)
89
crc = crc32(chunk, crc)
90
total_len += len(chunk)
91
zbytes = compress.compress(chunk)
94
result.append(compress.flush())
95
# size may exceed 2GB, or even 4GB
96
result.append(struct.pack("<LL", LOWU32(crc), LOWU32(total_len)))
97
return ''.join(result)
30
100
class GzipFile(gzip.GzipFile):
50
120
Yes, its only 1.6 seconds, but they add up.
53
def _add_read_data(self, data):
55
# temp var for len(data) and switch to +='s.
58
self.crc = zlib.crc32(data, self.crc)
60
self.extrasize += len_data
63
def _read(self, size=1024):
64
# various optimisations:
65
# reduces lsprof count from 2500 to
66
# 8337 calls in 1272, 365 internal
67
if self.fileobj is None:
68
raise EOFError, "Reached EOF"
71
# If the _new_member flag is set, we have to
72
# jump to the next member, if there is one.
74
# First, check if we're at the end of the file;
75
# if so, it's time to stop; no more members to read.
76
next_header_bytes = self.fileobj.read(10)
77
if next_header_bytes == '':
123
def __init__(self, *args, **kwargs):
124
symbol_versioning.warn(
125
symbol_versioning.deprecated_in((2, 3, 0))
126
% 'bzrlib.tuned_gzip.GzipFile',
127
DeprecationWarning, stacklevel=2)
128
gzip.GzipFile.__init__(self, *args, **kwargs)
130
if sys.version_info >= (2, 7, 4):
131
def _add_read_data(self, data):
133
# temp var for len(data) and switch to +='s.
136
self.crc = zlib.crc32(data, self.crc) & 0xffffffffL
137
offset = self.offset - self.extrastart
138
self.extrabuf = self.extrabuf[offset:] + data
139
self.extrasize = self.extrasize + len_data
140
self.extrastart = self.offset
141
self.size = self.size + len_data
143
def _add_read_data(self, data):
145
# temp var for len(data) and switch to +='s.
148
self.crc = zlib.crc32(data, self.crc)
149
self.extrabuf += data
150
self.extrasize += len_data
151
self.size += len_data
153
def _write_gzip_header(self):
154
"""A tuned version of gzip._write_gzip_header
156
We have some extra constrains that plain Gzip does not.
157
1) We want to write the whole blob at once. rather than multiple
158
calls to fileobj.write().
159
2) We never have a filename
160
3) We don't care about the time
163
'\037\213' # self.fileobj.write('\037\213') # magic header
164
'\010' # self.fileobj.write('\010') # compression method
165
# fname = self.filename[:-3]
169
'\x00' # self.fileobj.write(chr(flags))
170
'\0\0\0\0' # write32u(self.fileobj, long(time.time()))
171
'\002' # self.fileobj.write('\002')
172
'\377' # self.fileobj.write('\377')
174
'' # self.fileobj.write(fname + '\000')
177
if sys.version_info < (2, 7, 4):
178
def _read(self, size=1024):
179
# various optimisations:
180
# reduces lsprof count from 2500 to
181
# 8337 calls in 1272, 365 internal
182
if self.fileobj is None:
78
183
raise EOFError, "Reached EOF"
81
self._read_gzip_header(next_header_bytes)
82
self.decompress = zlib.decompressobj(-zlib.MAX_WBITS)
83
self._new_member = False
85
# Read a chunk of data from the file
86
buf = self.fileobj.read(size)
88
# If the EOF has been reached, flush the decompression object
89
# and mark this object as finished.
92
self._add_read_data(self.decompress.flush())
93
assert len(self.decompress.unused_data) >= 8, "what does flush do?"
95
# tell the driving read() call we have stuffed all the data
97
raise EOFError, 'Reached EOF'
99
self._add_read_data(self.decompress.decompress(buf))
101
if self.decompress.unused_data != "":
102
# Ending case: we've come to the end of a member in the file,
103
# so seek back to the start of the data for the next member which
104
# is the length of the decompress objects unused data - the first
105
# 8 bytes for the end crc and size records.
107
# so seek back to the start of the unused data, finish up
108
# this member, and read a new gzip header.
109
# (The number of bytes to seek back is the length of the unused
110
# data, minus 8 because those 8 bytes are part of this member.
111
seek_length = len (self.decompress.unused_data) - 8
113
# we read too much data
114
self.fileobj.seek(-seek_length, 1)
115
elif seek_length < 0:
116
# we haven't read enough to check the checksum.
117
assert -8 < seek_length, "too great a seek."
118
buf = self.fileobj.read(-seek_length)
119
self.decompress.decompress(buf)
121
# Check the CRC and file size, and set the flag so we read
122
# a new member on the next call
124
self._new_member = True
127
"""tuned to reduce function calls and eliminate file seeking:
129
reduces lsprof count from 800 to 288
131
avoid U32 call by using struct format L
134
# We've read to the end of the file, so we should have 8 bytes of
135
# unused data in the decompressor. If we dont, there is a corrupt file.
136
# We use these 8 bytes to calculate the CRC and the recorded file size.
137
# We then check the that the computed CRC and size of the
138
# uncompressed data matches the stored values. Note that the size
139
# stored is the true file size mod 2**32.
140
crc32, isize = struct.unpack("<LL", self.decompress.unused_data[0:8])
141
# note that isize is unsigned - it can exceed 2GB
142
if crc32 != U32(self.crc):
143
raise IOError, "CRC check failed %d %d" % (crc32, U32(self.crc))
144
elif isize != LOWU32(self.size):
145
raise IOError, "Incorrect length of data produced"
186
# If the _new_member flag is set, we have to
187
# jump to the next member, if there is one.
189
# First, check if we're at the end of the file;
190
# if so, it's time to stop; no more members to read.
191
next_header_bytes = self.fileobj.read(10)
192
if next_header_bytes == '':
193
raise EOFError, "Reached EOF"
196
self._read_gzip_header(next_header_bytes)
197
self.decompress = zlib.decompressobj(-zlib.MAX_WBITS)
198
self._new_member = False
200
# Read a chunk of data from the file
201
buf = self.fileobj.read(size)
203
# If the EOF has been reached, flush the decompression object
204
# and mark this object as finished.
207
self._add_read_data(self.decompress.flush())
208
if len(self.decompress.unused_data) < 8:
209
raise AssertionError("what does flush do?")
210
self._gzip_tail = self.decompress.unused_data[0:8]
212
# tell the driving read() call we have stuffed all the data
214
raise EOFError, 'Reached EOF'
216
self._add_read_data(self.decompress.decompress(buf))
218
if self.decompress.unused_data != "":
219
# Ending case: we've come to the end of a member in the file,
220
# so seek back to the start of the data for the next member
221
# which is the length of the decompress objects unused data -
222
# the first 8 bytes for the end crc and size records.
224
# so seek back to the start of the unused data, finish up
225
# this member, and read a new gzip header.
226
# (The number of bytes to seek back is the length of the unused
227
# data, minus 8 because those 8 bytes are part of this member.
228
seek_length = len (self.decompress.unused_data) - 8
230
# we read too much data
231
self.fileobj.seek(-seek_length, 1)
232
self._gzip_tail = self.decompress.unused_data[0:8]
233
elif seek_length < 0:
234
# we haven't read enough to check the checksum.
235
if not (-8 < seek_length):
236
raise AssertionError("too great a seek")
237
buf = self.fileobj.read(-seek_length)
238
self._gzip_tail = self.decompress.unused_data + buf
240
self._gzip_tail = self.decompress.unused_data
242
# Check the CRC and file size, and set the flag so we read
243
# a new member on the next call
245
self._new_member = True
248
"""tuned to reduce function calls and eliminate file seeking:
250
reduces lsprof count from 800 to 288
252
avoid U32 call by using struct format L
255
# We've read to the end of the file, so we should have 8 bytes of
256
# unused data in the decompressor. If we don't, there is a corrupt
257
# file. We use these 8 bytes to calculate the CRC and the recorded
258
# file size. We then check the that the computed CRC and size of
259
# the uncompressed data matches the stored values. Note that the
260
# size stored is the true file size mod 2**32.
261
if not (len(self._gzip_tail) == 8):
262
raise AssertionError("gzip trailer is incorrect length.")
263
crc32, isize = struct.unpack("<LL", self._gzip_tail)
264
# note that isize is unsigned - it can exceed 2GB
265
if crc32 != U32(self.crc):
266
raise IOError, "CRC check failed %d %d" % (crc32, U32(self.crc))
267
elif isize != LOWU32(self.size):
268
raise IOError, "Incorrect length of data produced"
147
270
def _read_gzip_header(self, bytes=None):
148
271
"""Supply bytes if the minimum header size is already read.
150
273
:param bytes: 10 bytes of header data.
152
275
"""starting cost: 300 in 3998
added
removed
bzrlib/tuned_gzip.py
