59
57
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
60
class ContainerWriter(object):
61
"""A class for writing containers."""
63
def __init__(self, write_func):
66
:param write_func: a callable that will be called when this
67
ContainerWriter needs to write some bytes.
69
self.write_func = write_func
71
"""Return the bytes to begin a container."""
72
return FORMAT_ONE + "\n"
72
"""Begin writing a container."""
73
self.write_func(FORMAT_ONE + "\n")
75
"""Return the bytes to finish a container."""
76
"""Finish writing a container."""
78
def bytes_record(self, bytes, names):
79
"""Return the bytes for a Bytes record with the given name and
79
def add_bytes_record(self, bytes, names):
80
"""Add a Bytes record with the given names."""
85
byte_sections.append(str(len(bytes)) + "\n")
84
self.write_func(str(len(bytes)) + "\n")
87
for name_tuple in names:
88
87
# Make sure we're writing valid names. Note that we will leave a
89
88
# half-written record if a name is bad!
90
for name in name_tuple:
92
byte_sections.append('\x00'.join(name_tuple) + "\n")
90
self.write_func(name + "\n")
94
byte_sections.append("\n")
95
93
# 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
def __init__(self, readv_result):
165
self.readv_result = readv_result
166
# the most recent readv result block
170
if (self._string is None or
171
self._string.tell() == self._string_length):
172
length, data = self.readv_result.next()
173
self._string_length = len(data)
174
self._string = StringIO(data)
176
def read(self, length):
178
result = self._string.read(length)
179
if len(result) < length:
180
raise errors.BzrError('request for too much data from a readv hunk.')
184
"""Note that readline will not cross readv segments."""
186
result = self._string.readline()
187
if self._string.tell() == self._string_length and result[-1] != '\n':
188
raise errors.BzrError('short readline in the readvfile hunk.')
192
def make_readv_reader(transport, filename, requested_records):
193
"""Create a ContainerReader that will read selected records only.
195
:param transport: The transport the pack file is located on.
196
:param filename: The filename of the pack file.
197
:param requested_records: The record offset, length tuples as returned
198
by add_bytes_record for the desired records.
200
readv_blocks = [(0, len(FORMAT_ONE)+1)]
201
readv_blocks.extend(requested_records)
202
result = ContainerReader(ReadVFile(
203
transport.readv(filename, readv_blocks)))
94
self.write_func(bytes)
207
97
class BaseReader(object):
234
124
is a ``list`` and bytes is a function that takes one argument,
237
You **must not** call the callable after advancing the iterator to the
127
You **must not** call the callable after advancing the interator to the
238
128
next record. That is, this code is invalid::
240
130
record_iter = container.iter_records()
241
131
names1, callable1 = record_iter.next()
242
132
names2, callable2 = record_iter.next()
243
133
bytes1 = callable1(None)
245
135
As it will give incorrect results and invalidate the state of the
248
:raises ContainerError: if any sort of container corruption is
138
:raises ContainerError: if any sort of containter corruption is
249
139
detected, e.g. UnknownContainerFormatError is the format of the
250
140
container is unrecognised.
251
141
:seealso: ContainerReader.read
253
143
self._read_format()
254
144
return self._iter_records()
256
146
def iter_record_objects(self):
257
147
"""Iterate over the container, yielding each record as it is read.
376
263
:raises ContainerError: if this record is invalid.
378
265
names, read_bytes = self.read()
379
for name_tuple in names:
380
for name in name_tuple:
381
_check_name_encoding(name)
267
_check_name_encoding(name)
385
class ContainerPushParser(object):
386
"""A "push" parser for container format 1.
388
It accepts bytes via the ``accept_bytes`` method, and parses them into
389
records which can be retrieved via the ``read_pending_records`` method.
394
self._state_handler = self._state_expecting_format_line
395
self._parsed_records = []
396
self._reset_current_record()
397
self.finished = False
399
def _reset_current_record(self):
400
self._current_record_length = None
401
self._current_record_names = []
403
def accept_bytes(self, bytes):
404
self._buffer += bytes
405
# Keep iterating the state machine until it stops consuming bytes from
407
last_buffer_length = None
408
cur_buffer_length = len(self._buffer)
409
while cur_buffer_length != last_buffer_length:
410
last_buffer_length = cur_buffer_length
411
self._state_handler()
412
cur_buffer_length = len(self._buffer)
414
def read_pending_records(self, max=None):
416
records = self._parsed_records[:max]
417
del self._parsed_records[:max]
420
records = self._parsed_records
421
self._parsed_records = []
424
def _consume_line(self):
425
"""Take a line out of the buffer, and return the line.
427
If a newline byte is not found in the buffer, the buffer is
428
unchanged and this returns None instead.
430
newline_pos = self._buffer.find('\n')
431
if newline_pos != -1:
432
line = self._buffer[:newline_pos]
433
self._buffer = self._buffer[newline_pos+1:]
438
def _state_expecting_format_line(self):
439
line = self._consume_line()
441
if line != FORMAT_ONE:
442
raise errors.UnknownContainerFormatError(line)
443
self._state_handler = self._state_expecting_record_type
445
def _state_expecting_record_type(self):
446
if len(self._buffer) >= 1:
447
record_type = self._buffer[0]
448
self._buffer = self._buffer[1:]
449
if record_type == 'B':
450
self._state_handler = self._state_expecting_length
451
elif record_type == 'E':
453
self._state_handler = self._state_expecting_nothing
455
raise errors.UnknownRecordTypeError(record_type)
457
def _state_expecting_length(self):
458
line = self._consume_line()
461
self._current_record_length = int(line)
463
raise errors.InvalidRecordError(
464
"%r is not a valid length." % (line,))
465
self._state_handler = self._state_expecting_name
467
def _state_expecting_name(self):
468
encoded_name_parts = self._consume_line()
469
if encoded_name_parts == '':
470
self._state_handler = self._state_expecting_body
471
elif encoded_name_parts:
472
name_parts = tuple(encoded_name_parts.split('\x00'))
473
for name_part in name_parts:
474
_check_name(name_part)
475
self._current_record_names.append(name_parts)
477
def _state_expecting_body(self):
478
if len(self._buffer) >= self._current_record_length:
479
body_bytes = self._buffer[:self._current_record_length]
480
self._buffer = self._buffer[self._current_record_length:]
481
record = (self._current_record_names, body_bytes)
482
self._parsed_records.append(record)
483
self._reset_current_record()
484
self._state_handler = self._state_expecting_record_type
486
def _state_expecting_nothing(self):
489
def read_size_hint(self):
491
if self._state_handler == self._state_expecting_body:
492
remaining = self._current_record_length - len(self._buffer)
495
return max(hint, remaining)
499
def iter_records_from_file(source_file):
500
parser = ContainerPushParser()
502
bytes = source_file.read(parser.read_size_hint())
503
parser.accept_bytes(bytes)
504
for record in parser.read_pending_records():
added
removed
bzrlib/pack.py
