1
# Copyright (C) 2007, 2009 Canonical Ltd
3
# This program is free software; you can redistribute it and/or modify
4
# it under the terms of the GNU General Public License as published by
5
# the Free Software Foundation; either version 2 of the License, or
6
# (at your option) any later version.
8
# This program is distributed in the hope that it will be useful,
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11
# GNU General Public License for more details.
13
# You should have received a copy of the GNU General Public License
14
# along with this program; if not, write to the Free Software
15
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
17
"""Container format for Bazaar data.
19
"Containers" and "records" are described in
20
doc/developers/container-format.txt.
23
from cStringIO import StringIO
26
from bzrlib import errors
29
FORMAT_ONE = "Bazaar pack format 1 (introduced in 0.18)"
32
_whitespace_re = re.compile('[\t\n\x0b\x0c\r ]')
35
def _check_name(name):
36
"""Do some basic checking of 'name'.
38
At the moment, this just checks that there are no whitespace characters in a
41
:raises InvalidRecordError: if name is not valid.
42
:seealso: _check_name_encoding
44
if _whitespace_re.search(name) is not None:
45
raise errors.InvalidRecordError("%r is not a valid name." % (name,))
48
def _check_name_encoding(name):
49
"""Check that 'name' is valid UTF-8.
51
This is separate from _check_name because UTF-8 decoding is relatively
52
expensive, and we usually want to avoid it.
54
:raises InvalidRecordError: if name is not valid UTF-8.
58
except UnicodeDecodeError, e:
59
raise errors.InvalidRecordError(str(e))
62
class ContainerSerialiser(object):
63
"""A helper class for serialising containers.
65
It simply returns bytes from method calls to 'begin', 'end' and
66
'bytes_record'. You may find ContainerWriter to be a more convenient
71
"""Return the bytes to begin a container."""
72
return FORMAT_ONE + "\n"
75
"""Return the bytes to finish a container."""
78
def bytes_record(self, bytes, names):
79
"""Return the bytes for a Bytes record with the given name and
85
byte_sections.append(str(len(bytes)) + "\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
# Finally, the contents.
96
byte_sections.append(bytes)
97
# XXX: This causes a memory copy of bytes in size, but is usually
98
# faster than two write calls (12 vs 13 seconds to output a gig of
99
# 1k records.) - results may differ on significantly larger records
100
# like .iso's but as they should be rare in any case and thus not
101
# likely to be the common case. The biggest issue is causing extreme
102
# memory pressure in that case. One possibly improvement here is to
103
# check the size of the content before deciding to join here vs call
105
return ''.join(byte_sections)
108
class ContainerWriter(object):
109
"""A class for writing containers to a file.
111
:attribute records_written: The number of user records added to the
112
container. This does not count the prelude or suffix of the container
113
introduced by the begin() and end() methods.
116
def __init__(self, write_func):
119
:param write_func: a callable that will be called when this
120
ContainerWriter needs to write some bytes.
122
self._write_func = write_func
123
self.current_offset = 0
124
self.records_written = 0
125
self._serialiser = ContainerSerialiser()
128
"""Begin writing a container."""
129
self.write_func(self._serialiser.begin())
131
def write_func(self, bytes):
132
self._write_func(bytes)
133
self.current_offset += len(bytes)
136
"""Finish writing a container."""
137
self.write_func(self._serialiser.end())
139
def add_bytes_record(self, bytes, names):
140
"""Add a Bytes record with the given names.
142
:param bytes: The bytes to insert.
143
:param names: The names to give the inserted bytes. Each name is
144
a tuple of bytestrings. The bytestrings may not contain
146
:return: An offset, length tuple. The offset is the offset
147
of the record within the container, and the length is the
148
length of data that will need to be read to reconstitute the
149
record. These offset and length can only be used with the pack
150
interface - they might be offset by headers or other such details
151
and thus are only suitable for use by a ContainerReader.
153
current_offset = self.current_offset
154
serialised_record = self._serialiser.bytes_record(bytes, names)
155
self.write_func(serialised_record)
156
self.records_written += 1
157
# return a memo of where we wrote data to allow random access.
158
return current_offset, self.current_offset - current_offset
161
class ReadVFile(object):
162
"""Adapt a readv result iterator to a file like protocol.
164
The readv result must support the iterator protocol returning (offset,
168
# XXX: This could be a generic transport class, as other code may want to
169
# gradually consume the readv result.
171
def __init__(self, readv_result):
172
self.readv_result = readv_result
173
# the most recent readv result block
177
if (self._string is None or
178
self._string.tell() == self._string_length):
179
offset, data = self.readv_result.next()
180
self._string_length = len(data)
181
self._string = StringIO(data)
183
def read(self, length):
185
result = self._string.read(length)
186
if len(result) < length:
187
raise errors.BzrError('request for too much data from a readv hunk.')
191
"""Note that readline will not cross readv segments."""
193
result = self._string.readline()
194
if self._string.tell() == self._string_length and result[-1] != '\n':
195
raise errors.BzrError('short readline in the readvfile hunk.')
199
def make_readv_reader(transport, filename, requested_records):
200
"""Create a ContainerReader that will read selected records only.
202
:param transport: The transport the pack file is located on.
203
:param filename: The filename of the pack file.
204
:param requested_records: The record offset, length tuples as returned
205
by add_bytes_record for the desired records.
207
readv_blocks = [(0, len(FORMAT_ONE)+1)]
208
readv_blocks.extend(requested_records)
209
result = ContainerReader(ReadVFile(
210
transport.readv(filename, readv_blocks)))
214
class BaseReader(object):
216
def __init__(self, source_file):
219
:param source_file: a file-like object with `read` and `readline`
222
self._source = source_file
224
def reader_func(self, length=None):
225
return self._source.read(length)
227
def _read_line(self):
228
line = self._source.readline()
229
if not line.endswith('\n'):
230
raise errors.UnexpectedEndOfContainerError()
231
return line.rstrip('\n')
234
class ContainerReader(BaseReader):
235
"""A class for reading Bazaar's container format."""
237
def iter_records(self):
238
"""Iterate over the container, yielding each record as it is read.
240
Each yielded record will be a 2-tuple of (names, callable), where names
241
is a ``list`` and bytes is a function that takes one argument,
244
You **must not** call the callable after advancing the iterator to the
245
next record. That is, this code is invalid::
247
record_iter = container.iter_records()
248
names1, callable1 = record_iter.next()
249
names2, callable2 = record_iter.next()
250
bytes1 = callable1(None)
252
As it will give incorrect results and invalidate the state of the
255
:raises ContainerError: if any sort of container corruption is
256
detected, e.g. UnknownContainerFormatError is the format of the
257
container is unrecognised.
258
:seealso: ContainerReader.read
261
return self._iter_records()
263
def iter_record_objects(self):
264
"""Iterate over the container, yielding each record as it is read.
266
Each yielded record will be an object with ``read`` and ``validate``
267
methods. Like with iter_records, it is not safe to use a record object
268
after advancing the iterator to yield next record.
270
:raises ContainerError: if any sort of container corruption is
271
detected, e.g. UnknownContainerFormatError is the format of the
272
container is unrecognised.
273
:seealso: iter_records
276
return self._iter_record_objects()
278
def _iter_records(self):
279
for record in self._iter_record_objects():
282
def _iter_record_objects(self):
284
record_kind = self.reader_func(1)
285
if record_kind == 'B':
287
reader = BytesRecordReader(self._source)
289
elif record_kind == 'E':
290
# End marker. There are no more records.
292
elif record_kind == '':
293
# End of stream encountered, but no End Marker record seen, so
294
# this container is incomplete.
295
raise errors.UnexpectedEndOfContainerError()
297
# Unknown record type.
298
raise errors.UnknownRecordTypeError(record_kind)
300
def _read_format(self):
301
format = self._read_line()
302
if format != FORMAT_ONE:
303
raise errors.UnknownContainerFormatError(format)
306
"""Validate this container and its records.
308
Validating consumes the data stream just like iter_records and
309
iter_record_objects, so you cannot call it after
310
iter_records/iter_record_objects.
312
:raises ContainerError: if something is invalid.
315
for record_names, read_bytes in self.iter_records():
317
for name_tuple in record_names:
318
for name in name_tuple:
319
_check_name_encoding(name)
320
# Check that the name is unique. Note that Python will refuse
321
# to decode non-shortest forms of UTF-8 encoding, so there is no
322
# risk that the same unicode string has been encoded two
324
if name_tuple in all_names:
325
raise errors.DuplicateRecordNameError(name_tuple)
326
all_names.add(name_tuple)
327
excess_bytes = self.reader_func(1)
328
if excess_bytes != '':
329
raise errors.ContainerHasExcessDataError(excess_bytes)
332
class BytesRecordReader(BaseReader):
337
You can either validate or read a record, you can't do both.
339
:returns: A tuple of (names, callable). The callable can be called
340
repeatedly to obtain the bytes for the record, with a max_length
341
argument. If max_length is None, returns all the bytes. Because
342
records can be arbitrarily large, using None is not recommended
343
unless you have reason to believe the content will fit in memory.
345
# Read the content length.
346
length_line = self._read_line()
348
length = int(length_line)
350
raise errors.InvalidRecordError(
351
"%r is not a valid length." % (length_line,))
353
# Read the list of names.
356
name_line = self._read_line()
359
name_tuple = tuple(name_line.split('\x00'))
360
for name in name_tuple:
362
names.append(name_tuple)
364
self._remaining_length = length
365
return names, self._content_reader
367
def _content_reader(self, max_length):
368
if max_length is None:
369
length_to_read = self._remaining_length
371
length_to_read = min(max_length, self._remaining_length)
372
self._remaining_length -= length_to_read
373
bytes = self.reader_func(length_to_read)
374
if len(bytes) != length_to_read:
375
raise errors.UnexpectedEndOfContainerError()
379
"""Validate this record.
381
You can either validate or read, you can't do both.
383
:raises ContainerError: if this record is invalid.
385
names, read_bytes = self.read()
386
for name_tuple in names:
387
for name in name_tuple:
388
_check_name_encoding(name)
392
class ContainerPushParser(object):
393
"""A "push" parser for container format 1.
395
It accepts bytes via the ``accept_bytes`` method, and parses them into
396
records which can be retrieved via the ``read_pending_records`` method.
401
self._state_handler = self._state_expecting_format_line
402
self._parsed_records = []
403
self._reset_current_record()
404
self.finished = False
406
def _reset_current_record(self):
407
self._current_record_length = None
408
self._current_record_names = []
410
def accept_bytes(self, bytes):
411
self._buffer += bytes
412
# Keep iterating the state machine until it stops consuming bytes from
414
last_buffer_length = None
415
cur_buffer_length = len(self._buffer)
416
last_state_handler = None
417
while (cur_buffer_length != last_buffer_length
418
or last_state_handler != self._state_handler):
419
last_buffer_length = cur_buffer_length
420
last_state_handler = self._state_handler
421
self._state_handler()
422
cur_buffer_length = len(self._buffer)
424
def read_pending_records(self, max=None):
426
records = self._parsed_records[:max]
427
del self._parsed_records[:max]
430
records = self._parsed_records
431
self._parsed_records = []
434
def _consume_line(self):
435
"""Take a line out of the buffer, and return the line.
437
If a newline byte is not found in the buffer, the buffer is
438
unchanged and this returns None instead.
440
newline_pos = self._buffer.find('\n')
441
if newline_pos != -1:
442
line = self._buffer[:newline_pos]
443
self._buffer = self._buffer[newline_pos+1:]
448
def _state_expecting_format_line(self):
449
line = self._consume_line()
451
if line != FORMAT_ONE:
452
raise errors.UnknownContainerFormatError(line)
453
self._state_handler = self._state_expecting_record_type
455
def _state_expecting_record_type(self):
456
if len(self._buffer) >= 1:
457
record_type = self._buffer[0]
458
self._buffer = self._buffer[1:]
459
if record_type == 'B':
460
self._state_handler = self._state_expecting_length
461
elif record_type == 'E':
463
self._state_handler = self._state_expecting_nothing
465
raise errors.UnknownRecordTypeError(record_type)
467
def _state_expecting_length(self):
468
line = self._consume_line()
471
self._current_record_length = int(line)
473
raise errors.InvalidRecordError(
474
"%r is not a valid length." % (line,))
475
self._state_handler = self._state_expecting_name
477
def _state_expecting_name(self):
478
encoded_name_parts = self._consume_line()
479
if encoded_name_parts == '':
480
self._state_handler = self._state_expecting_body
481
elif encoded_name_parts:
482
name_parts = tuple(encoded_name_parts.split('\x00'))
483
for name_part in name_parts:
484
_check_name(name_part)
485
self._current_record_names.append(name_parts)
487
def _state_expecting_body(self):
488
if len(self._buffer) >= self._current_record_length:
489
body_bytes = self._buffer[:self._current_record_length]
490
self._buffer = self._buffer[self._current_record_length:]
491
record = (self._current_record_names, body_bytes)
492
self._parsed_records.append(record)
493
self._reset_current_record()
494
self._state_handler = self._state_expecting_record_type
496
def _state_expecting_nothing(self):
499
def read_size_hint(self):
501
if self._state_handler == self._state_expecting_body:
502
remaining = self._current_record_length - len(self._buffer)
505
return max(hint, remaining)
509
def iter_records_from_file(source_file):
510
parser = ContainerPushParser()
512
bytes = source_file.read(parser.read_size_hint())
513
parser.accept_bytes(bytes)
514
for record in parser.read_pending_records():
added
removed
bzrlib/pack.py
