1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
|
# Copyright (C) 2005, 2006 Canonical Ltd
# Written by Robert Collins <robert.collins@canonical.com>
#
# This program is free software; you can redistribute it and/or modify
# it under the terms of the GNU General Public License as published by
# the Free Software Foundation; either version 2 of the License, or
# (at your option) any later version.
#
# This program is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
# GNU General Public License for more details.
#
# You should have received a copy of the GNU General Public License
# along with this program; if not, write to the Free Software
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
"""Bzrlib specific gzip tunings. We plan to feed these to the upstream gzip."""
from cStringIO import StringIO
# make GzipFile faster:
import gzip
from gzip import U32, LOWU32, FEXTRA, FCOMMENT, FNAME, FHCRC
import sys
import struct
import zlib
# we want a \n preserved, break on \n only splitlines.
import bzrlib
__all__ = ["GzipFile"]
class GzipFile(gzip.GzipFile):
"""Knit tuned version of GzipFile.
This is based on the following lsprof stats:
python 2.4 stock GzipFile write:
58971 0 5644.3090 2721.4730 gzip:193(write)
+58971 0 1159.5530 1159.5530 +<built-in method compress>
+176913 0 987.0320 987.0320 +<len>
+58971 0 423.1450 423.1450 +<zlib.crc32>
+58971 0 353.1060 353.1060 +<method 'write' of 'cStringIO.
StringO' objects>
tuned GzipFile write:
58971 0 4477.2590 2103.1120 bzrlib.knit:1250(write)
+58971 0 1297.7620 1297.7620 +<built-in method compress>
+58971 0 406.2160 406.2160 +<zlib.crc32>
+58971 0 341.9020 341.9020 +<method 'write' of 'cStringIO.
StringO' objects>
+58971 0 328.2670 328.2670 +<len>
Yes, its only 1.6 seconds, but they add up.
"""
def _add_read_data(self, data):
# 4169 calls in 183
# temp var for len(data) and switch to +='s.
# 4169 in 139
len_data = len(data)
self.crc = zlib.crc32(data, self.crc)
self.extrabuf += data
self.extrasize += len_data
self.size += len_data
def _write_gzip_header(self):
"""A tuned version of gzip._write_gzip_header
We have some extra constrains that plain Gzip does not.
1) We want to write the whole blob at once. rather than multiple
calls to fileobj.write().
2) We never have a filename
3) We don't care about the time
"""
self.fileobj.write(
'\037\213' # self.fileobj.write('\037\213') # magic header
'\010' # self.fileobj.write('\010') # compression method
# fname = self.filename[:-3]
# flags = 0
# if fname:
# flags = FNAME
'\x00' # self.fileobj.write(chr(flags))
'\0\0\0\0' # write32u(self.fileobj, long(time.time()))
'\002' # self.fileobj.write('\002')
'\377' # self.fileobj.write('\377')
# if fname:
'' # self.fileobj.write(fname + '\000')
)
def _read(self, size=1024):
# various optimisations:
# reduces lsprof count from 2500 to
# 8337 calls in 1272, 365 internal
if self.fileobj is None:
raise EOFError, "Reached EOF"
if self._new_member:
# If the _new_member flag is set, we have to
# jump to the next member, if there is one.
#
# First, check if we're at the end of the file;
# if so, it's time to stop; no more members to read.
next_header_bytes = self.fileobj.read(10)
if next_header_bytes == '':
raise EOFError, "Reached EOF"
self._init_read()
self._read_gzip_header(next_header_bytes)
self.decompress = zlib.decompressobj(-zlib.MAX_WBITS)
self._new_member = False
# Read a chunk of data from the file
buf = self.fileobj.read(size)
# If the EOF has been reached, flush the decompression object
# and mark this object as finished.
if buf == "":
self._add_read_data(self.decompress.flush())
assert len(self.decompress.unused_data) >= 8, "what does flush do?"
self._gzip_tail = self.decompress.unused_data[0:8]
self._read_eof()
# tell the driving read() call we have stuffed all the data
# in self.extrabuf
raise EOFError, 'Reached EOF'
self._add_read_data(self.decompress.decompress(buf))
if self.decompress.unused_data != "":
# Ending case: we've come to the end of a member in the file,
# so seek back to the start of the data for the next member which
# is the length of the decompress objects unused data - the first
# 8 bytes for the end crc and size records.
#
# so seek back to the start of the unused data, finish up
# this member, and read a new gzip header.
# (The number of bytes to seek back is the length of the unused
# data, minus 8 because those 8 bytes are part of this member.
seek_length = len (self.decompress.unused_data) - 8
if seek_length > 0:
# we read too much data
self.fileobj.seek(-seek_length, 1)
self._gzip_tail = self.decompress.unused_data[0:8]
elif seek_length < 0:
# we haven't read enough to check the checksum.
assert -8 < seek_length, "too great a seek."
buf = self.fileobj.read(-seek_length)
self._gzip_tail = self.decompress.unused_data + buf
else:
self._gzip_tail = self.decompress.unused_data
# Check the CRC and file size, and set the flag so we read
# a new member on the next call
self._read_eof()
self._new_member = True
def _read_eof(self):
"""tuned to reduce function calls and eliminate file seeking:
pass 1:
reduces lsprof count from 800 to 288
4168 in 296
avoid U32 call by using struct format L
4168 in 200
"""
# We've read to the end of the file, so we should have 8 bytes of
# unused data in the decompressor. If we don't, there is a corrupt file.
# We use these 8 bytes to calculate the CRC and the recorded file size.
# We then check the that the computed CRC and size of the
# uncompressed data matches the stored values. Note that the size
# stored is the true file size mod 2**32.
assert len(self._gzip_tail) == 8, "gzip trailer is incorrect length."
crc32, isize = struct.unpack("<LL", self._gzip_tail)
# note that isize is unsigned - it can exceed 2GB
if crc32 != U32(self.crc):
raise IOError, "CRC check failed %d %d" % (crc32, U32(self.crc))
elif isize != LOWU32(self.size):
raise IOError, "Incorrect length of data produced"
def _read_gzip_header(self, bytes=None):
"""Supply bytes if the minimum header size is already read.
:param bytes: 10 bytes of header data.
"""
"""starting cost: 300 in 3998
15998 reads from 3998 calls
final cost 168
"""
if bytes is None:
bytes = self.fileobj.read(10)
magic = bytes[0:2]
if magic != '\037\213':
raise IOError, 'Not a gzipped file'
method = ord(bytes[2:3])
if method != 8:
raise IOError, 'Unknown compression method'
flag = ord(bytes[3:4])
# modtime = self.fileobj.read(4) (bytes [4:8])
# extraflag = self.fileobj.read(1) (bytes[8:9])
# os = self.fileobj.read(1) (bytes[9:10])
# self.fileobj.read(6)
if flag & FEXTRA:
# Read & discard the extra field, if present
xlen = ord(self.fileobj.read(1))
xlen = xlen + 256*ord(self.fileobj.read(1))
self.fileobj.read(xlen)
if flag & FNAME:
# Read and discard a null-terminated string containing the filename
while True:
s = self.fileobj.read(1)
if not s or s=='\000':
break
if flag & FCOMMENT:
# Read and discard a null-terminated string containing a comment
while True:
s = self.fileobj.read(1)
if not s or s=='\000':
break
if flag & FHCRC:
self.fileobj.read(2) # Read & discard the 16-bit header CRC
def readline(self, size=-1):
"""Tuned to remove buffer length calls in _unread and...
also removes multiple len(c) calls, inlines _unread,
total savings - lsprof 5800 to 5300
phase 2:
4168 calls in 2233
8176 calls to read() in 1684
changing the min chunk size to 200 halved all the cache misses
leading to a drop to:
4168 calls in 1977
4168 call to read() in 1646
- i.e. just reduced the function call overhead. May be worth
keeping.
"""
if size < 0: size = sys.maxint
bufs = []
readsize = min(200, size) # Read from the file in small chunks
while True:
if size == 0:
return "".join(bufs) # Return resulting line
# c is the chunk
c = self.read(readsize)
# number of bytes read
len_c = len(c)
i = c.find('\n')
if size is not None:
# We set i=size to break out of the loop under two
# conditions: 1) there's no newline, and the chunk is
# larger than size, or 2) there is a newline, but the
# resulting line would be longer than 'size'.
if i==-1 and len_c > size: i=size-1
elif size <= i: i = size -1
if i >= 0 or c == '':
# if i>= 0 we have a newline or have triggered the above
# if size is not None condition.
# if c == '' its EOF.
bufs.append(c[:i+1]) # Add portion of last chunk
# -- inlined self._unread --
## self._unread(c[i+1:], len_c - i) # Push back rest of chunk
self.extrabuf = c[i+1:] + self.extrabuf
self.extrasize = len_c - i + self.extrasize
self.offset -= len_c - i
# -- end inlined self._unread --
return ''.join(bufs) # Return resulting line
# Append chunk to list, decrease 'size',
bufs.append(c)
size = size - len_c
readsize = min(size, readsize * 2)
def readlines(self, sizehint=0):
# optimise to avoid all the buffer manipulation
# lsprof changed from:
# 4168 calls in 5472 with 32000 calls to readline()
# to :
# 4168 calls in 417.
# Negative numbers result in reading all the lines
# python's gzip routine uses sizehint. This is a more efficient way
# than python uses to honor it. But it is even more efficient to
# just read the entire thing and use cStringIO to split into lines.
# if sizehint <= 0:
# sizehint = -1
# content = self.read(sizehint)
# return bzrlib.osutils.split_lines(content)
content = StringIO(self.read(-1))
return content.readlines()
def _unread(self, buf, len_buf=None):
"""tuned to remove unneeded len calls.
because this is such an inner routine in readline, and readline is
in many inner loops, this has been inlined into readline().
The len_buf parameter combined with the reduction in len calls dropped
the lsprof ms count for this routine on my test data from 800 to 200 -
a 75% saving.
"""
if len_buf is None:
len_buf = len(buf)
self.extrabuf = buf + self.extrabuf
self.extrasize = len_buf + self.extrasize
self.offset -= len_buf
def write(self, data):
if self.mode != gzip.WRITE:
import errno
raise IOError(errno.EBADF, "write() on read-only GzipFile object")
if self.fileobj is None:
raise ValueError, "write() on closed GzipFile object"
data_len = len(data)
if data_len > 0:
self.size = self.size + data_len
self.crc = zlib.crc32(data, self.crc)
self.fileobj.write( self.compress.compress(data) )
self.offset += data_len
def writelines(self, lines):
# profiling indicated a significant overhead
# calling write for each line.
# this batch call is a lot faster :).
# (4 seconds to 1 seconds for the sample upgrades I was testing).
self.write(''.join(lines))
|