100
120
Yes, its only 1.6 seconds, but they add up.
103
def _add_read_data(self, data):
105
# temp var for len(data) and switch to +='s.
108
self.crc = zlib.crc32(data, self.crc)
109
self.extrabuf += data
110
self.extrasize += len_data
111
self.size += len_data
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
113
153
def _write_gzip_header(self):
114
154
"""A tuned version of gzip._write_gzip_header
116
156
We have some extra constrains that plain Gzip does not.
117
1) We want to write the whole blob at once. rather than multiple
157
1) We want to write the whole blob at once. rather than multiple
118
158
calls to fileobj.write().
119
159
2) We never have a filename
120
160
3) We don't care about the time
134
174
'' # self.fileobj.write(fname + '\000')
137
def _read(self, size=1024):
138
# various optimisations:
139
# reduces lsprof count from 2500 to
140
# 8337 calls in 1272, 365 internal
141
if self.fileobj is None:
142
raise EOFError, "Reached EOF"
145
# If the _new_member flag is set, we have to
146
# jump to the next member, if there is one.
148
# First, check if we're at the end of the file;
149
# if so, it's time to stop; no more members to read.
150
next_header_bytes = self.fileobj.read(10)
151
if next_header_bytes == '':
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:
152
183
raise EOFError, "Reached EOF"
155
self._read_gzip_header(next_header_bytes)
156
self.decompress = zlib.decompressobj(-zlib.MAX_WBITS)
157
self._new_member = False
159
# Read a chunk of data from the file
160
buf = self.fileobj.read(size)
162
# If the EOF has been reached, flush the decompression object
163
# and mark this object as finished.
166
self._add_read_data(self.decompress.flush())
167
if len(self.decompress.unused_data) < 8:
168
raise AssertionError("what does flush do?")
169
self._gzip_tail = self.decompress.unused_data[0:8]
171
# tell the driving read() call we have stuffed all the data
173
raise EOFError, 'Reached EOF'
175
self._add_read_data(self.decompress.decompress(buf))
177
if self.decompress.unused_data != "":
178
# Ending case: we've come to the end of a member in the file,
179
# so seek back to the start of the data for the next member which
180
# is the length of the decompress objects unused data - the first
181
# 8 bytes for the end crc and size records.
183
# so seek back to the start of the unused data, finish up
184
# this member, and read a new gzip header.
185
# (The number of bytes to seek back is the length of the unused
186
# data, minus 8 because those 8 bytes are part of this member.
187
seek_length = len (self.decompress.unused_data) - 8
189
# we read too much data
190
self.fileobj.seek(-seek_length, 1)
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?")
191
210
self._gzip_tail = self.decompress.unused_data[0:8]
192
elif seek_length < 0:
193
# we haven't read enough to check the checksum.
194
if not (-8 < seek_length):
195
raise AssertionError("too great a seek")
196
buf = self.fileobj.read(-seek_length)
197
self._gzip_tail = self.decompress.unused_data + buf
199
self._gzip_tail = self.decompress.unused_data
201
# Check the CRC and file size, and set the flag so we read
202
# a new member on the next call
204
self._new_member = True
207
"""tuned to reduce function calls and eliminate file seeking:
209
reduces lsprof count from 800 to 288
211
avoid U32 call by using struct format L
214
# We've read to the end of the file, so we should have 8 bytes of
215
# unused data in the decompressor. If we don't, there is a corrupt file.
216
# We use these 8 bytes to calculate the CRC and the recorded file size.
217
# We then check the that the computed CRC and size of the
218
# uncompressed data matches the stored values. Note that the size
219
# stored is the true file size mod 2**32.
220
if not (len(self._gzip_tail) == 8):
221
raise AssertionError("gzip trailer is incorrect length.")
222
crc32, isize = struct.unpack("<LL", self._gzip_tail)
223
# note that isize is unsigned - it can exceed 2GB
224
if crc32 != U32(self.crc):
225
raise IOError, "CRC check failed %d %d" % (crc32, U32(self.crc))
226
elif isize != LOWU32(self.size):
227
raise IOError, "Incorrect length of data produced"
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"
229
270
def _read_gzip_header(self, bytes=None):
230
271
"""Supply bytes if the minimum header size is already read.
232
273
:param bytes: 10 bytes of header data.
234
275
"""starting cost: 300 in 3998