13
13
# You should have received a copy of the GNU General Public License
14
14
# along with this program; if not, write to the Free Software
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
15
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
17
17
"""Container format for Bazaar data.
19
"Containers" and "records" are described in doc/developers/container-format.txt.
19
"Containers" and "records" are described in
20
doc/developers/container-format.txt.
23
from __future__ import absolute_import
25
from cStringIO import StringIO
24
28
from bzrlib import errors
57
61
raise errors.InvalidRecordError(str(e))
64
class ContainerSerialiser(object):
65
"""A helper class for serialising containers.
67
It simply returns bytes from method calls to 'begin', 'end' and
68
'bytes_record'. You may find ContainerWriter to be a more convenient
73
"""Return the bytes to begin a container."""
74
return FORMAT_ONE + "\n"
77
"""Return the bytes to finish a container."""
80
def bytes_header(self, length, names):
81
"""Return the header for a Bytes record."""
85
byte_sections.append(str(length) + "\n")
87
for name_tuple in names:
88
# Make sure we're writing valid names. Note that we will leave a
89
# half-written record if a name is bad!
90
for name in name_tuple:
92
byte_sections.append('\x00'.join(name_tuple) + "\n")
94
byte_sections.append("\n")
95
return ''.join(byte_sections)
97
def bytes_record(self, bytes, names):
98
"""Return the bytes for a Bytes record with the given name and
101
If the content may be large, construct the header separately and then
102
stream out the contents.
104
return self.bytes_header(len(bytes), names) + bytes
60
107
class ContainerWriter(object):
61
"""A class for writing containers."""
108
"""A class for writing containers to a file.
110
:attribute records_written: The number of user records added to the
111
container. This does not count the prelude or suffix of the container
112
introduced by the begin() and end() methods.
115
# Join up headers with the body if writing fewer than this many bytes:
116
# trades off memory usage and copying to do less IO ops.
117
_JOIN_WRITES_THRESHOLD = 100000
63
119
def __init__(self, write_func):
66
122
:param write_func: a callable that will be called when this
67
123
ContainerWriter needs to write some bytes.
69
self.write_func = write_func
125
self._write_func = write_func
126
self.current_offset = 0
127
self.records_written = 0
128
self._serialiser = ContainerSerialiser()
72
131
"""Begin writing a container."""
73
self.write_func(FORMAT_ONE + "\n")
132
self.write_func(self._serialiser.begin())
134
def write_func(self, bytes):
135
self._write_func(bytes)
136
self.current_offset += len(bytes)
76
139
"""Finish writing a container."""
140
self.write_func(self._serialiser.end())
79
142
def add_bytes_record(self, bytes, names):
80
"""Add a Bytes record with the given names."""
84
self.write_func(str(len(bytes)) + "\n")
87
# Make sure we're writing valid names. Note that we will leave a
88
# half-written record if a name is bad!
90
self.write_func(name + "\n")
93
# Finally, the contents.
94
self.write_func(bytes)
143
"""Add a Bytes record with the given names.
145
:param bytes: The bytes to insert.
146
:param names: The names to give the inserted bytes. Each name is
147
a tuple of bytestrings. The bytestrings may not contain
149
:return: An offset, length tuple. The offset is the offset
150
of the record within the container, and the length is the
151
length of data that will need to be read to reconstitute the
152
record. These offset and length can only be used with the pack
153
interface - they might be offset by headers or other such details
154
and thus are only suitable for use by a ContainerReader.
156
current_offset = self.current_offset
158
if length < self._JOIN_WRITES_THRESHOLD:
159
self.write_func(self._serialiser.bytes_header(length, names)
162
self.write_func(self._serialiser.bytes_header(length, names))
163
self.write_func(bytes)
164
self.records_written += 1
165
# return a memo of where we wrote data to allow random access.
166
return current_offset, self.current_offset - current_offset
169
class ReadVFile(object):
170
"""Adapt a readv result iterator to a file like protocol.
172
The readv result must support the iterator protocol returning (offset,
176
# XXX: This could be a generic transport class, as other code may want to
177
# gradually consume the readv result.
179
def __init__(self, readv_result):
180
"""Construct a new ReadVFile wrapper.
182
:seealso: make_readv_reader
184
:param readv_result: the most recent readv result - list or generator
186
# readv can return a sequence or an iterator, but we require an
187
# iterator to know how much has been consumed.
188
readv_result = iter(readv_result)
189
self.readv_result = readv_result
193
if (self._string is None or
194
self._string.tell() == self._string_length):
195
offset, data = self.readv_result.next()
196
self._string_length = len(data)
197
self._string = StringIO(data)
199
def read(self, length):
201
result = self._string.read(length)
202
if len(result) < length:
203
raise errors.BzrError('wanted %d bytes but next '
204
'hunk only contains %d: %r...' %
205
(length, len(result), result[:20]))
209
"""Note that readline will not cross readv segments."""
211
result = self._string.readline()
212
if self._string.tell() == self._string_length and result[-1] != '\n':
213
raise errors.BzrError('short readline in the readvfile hunk: %r'
218
def make_readv_reader(transport, filename, requested_records):
219
"""Create a ContainerReader that will read selected records only.
221
:param transport: The transport the pack file is located on.
222
:param filename: The filename of the pack file.
223
:param requested_records: The record offset, length tuples as returned
224
by add_bytes_record for the desired records.
226
readv_blocks = [(0, len(FORMAT_ONE)+1)]
227
readv_blocks.extend(requested_records)
228
result = ContainerReader(ReadVFile(
229
transport.readv(filename, readv_blocks)))
97
233
class BaseReader(object):
124
260
is a ``list`` and bytes is a function that takes one argument,
127
You **must not** call the callable after advancing the interator to the
263
You **must not** call the callable after advancing the iterator to the
128
264
next record. That is, this code is invalid::
130
266
record_iter = container.iter_records()
131
267
names1, callable1 = record_iter.next()
132
268
names2, callable2 = record_iter.next()
133
269
bytes1 = callable1(None)
135
271
As it will give incorrect results and invalidate the state of the
138
:raises ContainerError: if any sort of containter corruption is
274
:raises ContainerError: if any sort of container corruption is
139
275
detected, e.g. UnknownContainerFormatError is the format of the
140
276
container is unrecognised.
141
277
:seealso: ContainerReader.read
143
279
self._read_format()
144
280
return self._iter_records()
146
282
def iter_record_objects(self):
147
283
"""Iterate over the container, yielding each record as it is read.
197
333
all_names = set()
198
334
for record_names, read_bytes in self.iter_records():
200
for name in record_names:
201
_check_name_encoding(name)
336
for name_tuple in record_names:
337
for name in name_tuple:
338
_check_name_encoding(name)
202
339
# Check that the name is unique. Note that Python will refuse
203
340
# to decode non-shortest forms of UTF-8 encoding, so there is no
204
341
# risk that the same unicode string has been encoded two
205
342
# different ways.
206
if name in all_names:
207
raise errors.DuplicateRecordNameError(name)
343
if name_tuple in all_names:
344
raise errors.DuplicateRecordNameError(name_tuple[0])
345
all_names.add(name_tuple)
209
346
excess_bytes = self.reader_func(1)
210
347
if excess_bytes != '':
211
348
raise errors.ContainerHasExcessDataError(excess_bytes)
263
402
:raises ContainerError: if this record is invalid.
265
404
names, read_bytes = self.read()
267
_check_name_encoding(name)
405
for name_tuple in names:
406
for name in name_tuple:
407
_check_name_encoding(name)
411
class ContainerPushParser(object):
412
"""A "push" parser for container format 1.
414
It accepts bytes via the ``accept_bytes`` method, and parses them into
415
records which can be retrieved via the ``read_pending_records`` method.
420
self._state_handler = self._state_expecting_format_line
421
self._parsed_records = []
422
self._reset_current_record()
423
self.finished = False
425
def _reset_current_record(self):
426
self._current_record_length = None
427
self._current_record_names = []
429
def accept_bytes(self, bytes):
430
self._buffer += bytes
431
# Keep iterating the state machine until it stops consuming bytes from
433
last_buffer_length = None
434
cur_buffer_length = len(self._buffer)
435
last_state_handler = None
436
while (cur_buffer_length != last_buffer_length
437
or last_state_handler != self._state_handler):
438
last_buffer_length = cur_buffer_length
439
last_state_handler = self._state_handler
440
self._state_handler()
441
cur_buffer_length = len(self._buffer)
443
def read_pending_records(self, max=None):
445
records = self._parsed_records[:max]
446
del self._parsed_records[:max]
449
records = self._parsed_records
450
self._parsed_records = []
453
def _consume_line(self):
454
"""Take a line out of the buffer, and return the line.
456
If a newline byte is not found in the buffer, the buffer is
457
unchanged and this returns None instead.
459
newline_pos = self._buffer.find('\n')
460
if newline_pos != -1:
461
line = self._buffer[:newline_pos]
462
self._buffer = self._buffer[newline_pos+1:]
467
def _state_expecting_format_line(self):
468
line = self._consume_line()
470
if line != FORMAT_ONE:
471
raise errors.UnknownContainerFormatError(line)
472
self._state_handler = self._state_expecting_record_type
474
def _state_expecting_record_type(self):
475
if len(self._buffer) >= 1:
476
record_type = self._buffer[0]
477
self._buffer = self._buffer[1:]
478
if record_type == 'B':
479
self._state_handler = self._state_expecting_length
480
elif record_type == 'E':
482
self._state_handler = self._state_expecting_nothing
484
raise errors.UnknownRecordTypeError(record_type)
486
def _state_expecting_length(self):
487
line = self._consume_line()
490
self._current_record_length = int(line)
492
raise errors.InvalidRecordError(
493
"%r is not a valid length." % (line,))
494
self._state_handler = self._state_expecting_name
496
def _state_expecting_name(self):
497
encoded_name_parts = self._consume_line()
498
if encoded_name_parts == '':
499
self._state_handler = self._state_expecting_body
500
elif encoded_name_parts:
501
name_parts = tuple(encoded_name_parts.split('\x00'))
502
for name_part in name_parts:
503
_check_name(name_part)
504
self._current_record_names.append(name_parts)
506
def _state_expecting_body(self):
507
if len(self._buffer) >= self._current_record_length:
508
body_bytes = self._buffer[:self._current_record_length]
509
self._buffer = self._buffer[self._current_record_length:]
510
record = (self._current_record_names, body_bytes)
511
self._parsed_records.append(record)
512
self._reset_current_record()
513
self._state_handler = self._state_expecting_record_type
515
def _state_expecting_nothing(self):
518
def read_size_hint(self):
520
if self._state_handler == self._state_expecting_body:
521
remaining = self._current_record_length - len(self._buffer)
524
return max(hint, remaining)
528
def iter_records_from_file(source_file):
529
parser = ContainerPushParser()
531
bytes = source_file.read(parser.read_size_hint())
532
parser.accept_bytes(bytes)
533
for record in parser.read_pending_records():
added
removed
bzrlib/pack.py
