48
36
def do(self, path, *args):
49
37
"""Execute a repository request.
51
All Repository requests take a path to the repository as their first
52
argument. The repository must be at the exact path given by the
53
client - no searching is done.
39
The repository must be at the exact path - no searching is done.
55
41
The actual logic is delegated to self.do_repository_request.
57
:param client_path: The path for the repository as received from the
59
:return: A SmartServerResponse from self.do_repository_request().
43
:param path: The path for the repository.
44
:return: A smart server from self.do_repository_request().
61
transport = self.transport_from_client_path(path)
46
transport = self._backing_transport.clone(path)
62
47
bzrdir = BzrDir.open_from_transport(transport)
63
# Save the repository for use with do_body.
64
self._repository = bzrdir.open_repository()
65
return self.do_repository_request(self._repository, *args)
67
def do_repository_request(self, repository, *args):
68
"""Override to provide an implementation for a verb."""
69
# No-op for verbs that take bodies (None as a result indicates a body
73
def recreate_search(self, repository, search_bytes, discard_excess=False):
74
"""Recreate a search from its serialised form.
76
:param discard_excess: If True, and the search refers to data we don't
77
have, just silently accept that fact - the verb calling
78
recreate_search trusts that clients will look for missing things
79
they expected and get it from elsewhere.
81
lines = search_bytes.split('\n')
82
if lines[0] == 'ancestry-of':
84
search_result = graph.PendingAncestryResult(heads, repository)
85
return search_result, None
86
elif lines[0] == 'search':
87
return self.recreate_search_from_recipe(repository, lines[1:],
88
discard_excess=discard_excess)
90
return (None, FailedSmartServerResponse(('BadSearch',)))
92
def recreate_search_from_recipe(self, repository, lines,
93
discard_excess=False):
94
"""Recreate a specific revision search (vs a from-tip search).
96
:param discard_excess: If True, and the search refers to data we don't
97
have, just silently accept that fact - the verb calling
98
recreate_search trusts that clients will look for missing things
99
they expected and get it from elsewhere.
101
start_keys = set(lines[0].split(' '))
102
exclude_keys = set(lines[1].split(' '))
103
revision_count = int(lines[2])
104
repository.lock_read()
106
search = repository.get_graph()._make_breadth_first_searcher(
110
next_revs = search.next()
111
except StopIteration:
113
search.stop_searching_any(exclude_keys.intersection(next_revs))
114
search_result = search.get_result()
115
if (not discard_excess and
116
search_result.get_recipe()[3] != revision_count):
117
# we got back a different amount of data than expected, this
118
# gets reported as NoSuchRevision, because less revisions
119
# indicates missing revisions, and more should never happen as
120
# the excludes list considers ghosts and ensures that ghost
121
# filling races are not a problem.
122
return (None, FailedSmartServerResponse(('NoSuchRevision',)))
123
return (search_result, None)
128
class SmartServerRepositoryReadLocked(SmartServerRepositoryRequest):
129
"""Calls self.do_readlocked_repository_request."""
131
def do_repository_request(self, repository, *args):
132
"""Read lock a repository for do_readlocked_repository_request."""
133
repository.lock_read()
135
return self.do_readlocked_repository_request(repository, *args)
140
class SmartServerRepositoryGetParentMap(SmartServerRepositoryRequest):
141
"""Bzr 1.2+ - get parent data for revisions during a graph search."""
143
no_extra_results = False
145
def do_repository_request(self, repository, *revision_ids):
146
"""Get parent details for some revisions.
148
All the parents for revision_ids are returned. Additionally up to 64KB
149
of additional parent data found by performing a breadth first search
150
from revision_ids is returned. The verb takes a body containing the
151
current search state, see do_body for details.
153
If 'include-missing:' is in revision_ids, ghosts encountered in the
154
graph traversal for getting parent data are included in the result with
155
a prefix of 'missing:'.
157
:param repository: The repository to query in.
158
:param revision_ids: The utf8 encoded revision_id to answer for.
160
self._revision_ids = revision_ids
161
return None # Signal that we want a body.
163
def do_body(self, body_bytes):
164
"""Process the current search state and perform the parent lookup.
166
:return: A smart server response where the body contains an utf8
167
encoded flattened list of the parents of the revisions (the same
168
format as Repository.get_revision_graph) which has been bz2
171
repository = self._repository
172
repository.lock_read()
174
return self._do_repository_request(body_bytes)
178
def _do_repository_request(self, body_bytes):
179
repository = self._repository
180
revision_ids = set(self._revision_ids)
181
include_missing = 'include-missing:' in revision_ids
183
revision_ids.remove('include-missing:')
184
body_lines = body_bytes.split('\n')
185
search_result, error = self.recreate_search_from_recipe(
186
repository, body_lines)
187
if error is not None:
189
# TODO might be nice to start up the search again; but thats not
190
# written or tested yet.
191
client_seen_revs = set(search_result.get_keys())
192
# Always include the requested ids.
193
client_seen_revs.difference_update(revision_ids)
195
repo_graph = repository.get_graph()
199
next_revs = revision_ids
200
first_loop_done = False
202
queried_revs.update(next_revs)
203
parent_map = repo_graph.get_parent_map(next_revs)
204
current_revs = next_revs
206
for revision_id in current_revs:
208
parents = parent_map.get(revision_id)
209
if parents is not None:
210
# adjust for the wire
211
if parents == (_mod_revision.NULL_REVISION,):
213
# prepare the next query
214
next_revs.update(parents)
215
encoded_id = revision_id
218
encoded_id = "missing:" + revision_id
220
if (revision_id not in client_seen_revs and
221
(not missing_rev or include_missing)):
222
# Client does not have this revision, give it to it.
223
# add parents to the result
224
result[encoded_id] = parents
225
# Approximate the serialized cost of this revision_id.
226
size_so_far += 2 + len(encoded_id) + sum(map(len, parents))
227
# get all the directly asked for parents, and then flesh out to
228
# 64K (compressed) or so. We do one level of depth at a time to
229
# stay in sync with the client. The 250000 magic number is
230
# estimated compression ratio taken from bzr.dev itself.
231
if self.no_extra_results or (
232
first_loop_done and size_so_far > 250000):
235
# don't query things we've already queried
236
next_revs.difference_update(queried_revs)
237
first_loop_done = True
239
# sorting trivially puts lexographically similar revision ids together.
241
for revision, parents in sorted(result.items()):
242
lines.append(' '.join((revision, ) + tuple(parents)))
244
return SuccessfulSmartServerResponse(
245
('ok', ), bz2.compress('\n'.join(lines)))
248
class SmartServerRepositoryGetRevisionGraph(SmartServerRepositoryReadLocked):
250
def do_readlocked_repository_request(self, repository, revision_id):
48
repository = bzrdir.open_repository()
49
return self.do_repository_request(repository, *args)
52
class SmartServerRepositoryGetRevisionGraph(SmartServerRepositoryRequest):
54
def do_repository_request(self, repository, revision_id):
251
55
"""Return the result of repository.get_revision_graph(revision_id).
253
Deprecated as of bzr 1.4, but supported for older clients.
255
57
:param repository: The repository to query in.
256
58
:param revision_id: The utf8 encoded revision_id to get a graph from.
257
59
:return: A smart server response where the body contains an utf8
393
163
return FailedSmartServerResponse(('LockContention',))
394
164
except errors.UnlockableTransport:
395
165
return FailedSmartServerResponse(('UnlockableTransport',))
396
except errors.LockFailed, e:
397
return FailedSmartServerResponse(('LockFailed',
398
str(e.lock), str(e.why)))
399
if token is not None:
400
repository.leave_lock_in_place()
166
repository.leave_lock_in_place()
401
167
repository.unlock()
402
168
if token is None:
404
170
return SuccessfulSmartServerResponse(('ok', token))
407
class SmartServerRepositoryGetStream(SmartServerRepositoryRequest):
409
def do_repository_request(self, repository, to_network_name):
410
"""Get a stream for inserting into a to_format repository.
412
:param repository: The repository to stream from.
413
:param to_network_name: The network name of the format of the target
416
self._to_format = network_format_registry.get(to_network_name)
417
return None # Signal that we want a body.
419
def do_body(self, body_bytes):
420
repository = self._repository
421
repository.lock_read()
423
search_result, error = self.recreate_search(repository, body_bytes,
425
if error is not None:
428
source = repository._get_source(self._to_format)
429
stream = source.get_stream(search_result)
431
exc_info = sys.exc_info()
433
# On non-error, unlocking is done by the body stream handler.
436
raise exc_info[0], exc_info[1], exc_info[2]
437
return SuccessfulSmartServerResponse(('ok',),
438
body_stream=self.body_stream(stream, repository))
440
def body_stream(self, stream, repository):
441
byte_stream = _stream_to_byte_stream(stream, repository._format)
443
for bytes in byte_stream:
445
except errors.RevisionNotPresent, e:
446
# This shouldn't be able to happen, but as we don't buffer
447
# everything it can in theory happen.
449
yield FailedSmartServerResponse(('NoSuchRevision', e.revision_id))
454
def _stream_to_byte_stream(stream, src_format):
455
"""Convert a record stream to a self delimited byte stream."""
456
pack_writer = pack.ContainerSerialiser()
457
yield pack_writer.begin()
458
yield pack_writer.bytes_record(src_format.network_name(), '')
459
for substream_type, substream in stream:
460
for record in substream:
461
if record.storage_kind in ('chunked', 'fulltext'):
462
serialised = record_to_fulltext_bytes(record)
463
elif record.storage_kind == 'absent':
464
raise ValueError("Absent factory for %s" % (record.key,))
466
serialised = record.get_bytes_as(record.storage_kind)
468
# Some streams embed the whole stream into the wire
469
# representation of the first record, which means that
470
# later records have no wire representation: we skip them.
471
yield pack_writer.bytes_record(serialised, [(substream_type,)])
472
yield pack_writer.end()
475
def _byte_stream_to_stream(byte_stream):
476
"""Convert a byte stream into a format and a stream.
478
:param byte_stream: A bytes iterator, as output by _stream_to_byte_stream.
479
:return: (RepositoryFormat, stream_generator)
481
stream_decoder = pack.ContainerPushParser()
483
"""Closure to return the substreams."""
484
# May have fully parsed records already.
485
for record in stream_decoder.read_pending_records():
486
record_names, record_bytes = record
487
record_name, = record_names
488
substream_type = record_name[0]
489
substream = NetworkRecordStream([record_bytes])
490
yield substream_type, substream.read()
491
for bytes in byte_stream:
492
stream_decoder.accept_bytes(bytes)
493
for record in stream_decoder.read_pending_records():
494
record_names, record_bytes = record
495
record_name, = record_names
496
substream_type = record_name[0]
497
substream = NetworkRecordStream([record_bytes])
498
yield substream_type, substream.read()
499
for bytes in byte_stream:
500
stream_decoder.accept_bytes(bytes)
501
for record in stream_decoder.read_pending_records(max=1):
502
record_names, src_format_name = record
503
src_format = network_format_registry.get(src_format_name)
504
return src_format, record_stream()
507
173
class SmartServerRepositoryUnlock(SmartServerRepositoryRequest):
509
175
def do_repository_request(self, repository, token):
578
235
dirname = dirname.encode(sys.getfilesystemencoding())
579
236
# python's tarball module includes the whole path by default so
581
if not dirname.endswith('.bzr'):
582
raise ValueError(dirname)
238
assert dirname.endswith('.bzr')
583
239
tarball.add(dirname, '.bzr') # recursive by default
588
class SmartServerRepositoryInsertStreamLocked(SmartServerRepositoryRequest):
589
"""Insert a record stream from a RemoteSink into a repository.
591
This gets bytes pushed to it by the network infrastructure and turns that
592
into a bytes iterator using a thread. That is then processed by
593
_byte_stream_to_stream.
598
def do_repository_request(self, repository, resume_tokens, lock_token):
599
"""StreamSink.insert_stream for a remote repository."""
600
repository.lock_write(token=lock_token)
601
self.do_insert_stream_request(repository, resume_tokens)
603
def do_insert_stream_request(self, repository, resume_tokens):
604
tokens = [token for token in resume_tokens.split(' ') if token]
606
self.repository = repository
607
self.queue = Queue.Queue()
608
self.insert_thread = threading.Thread(target=self._inserter_thread)
609
self.insert_thread.start()
611
def do_chunk(self, body_stream_chunk):
612
self.queue.put(body_stream_chunk)
614
def _inserter_thread(self):
616
src_format, stream = _byte_stream_to_stream(
617
self.blocking_byte_stream())
618
self.insert_result = self.repository._get_sink().insert_stream(
619
stream, src_format, self.tokens)
620
self.insert_ok = True
622
self.insert_exception = sys.exc_info()
623
self.insert_ok = False
625
def blocking_byte_stream(self):
627
bytes = self.queue.get()
628
if bytes is StopIteration:
634
self.queue.put(StopIteration)
635
if self.insert_thread is not None:
636
self.insert_thread.join()
637
if not self.insert_ok:
638
exc_info = self.insert_exception
639
raise exc_info[0], exc_info[1], exc_info[2]
640
write_group_tokens, missing_keys = self.insert_result
641
if write_group_tokens or missing_keys:
642
# bzip needed? missing keys should typically be a small set.
643
# Should this be a streaming body response ?
644
missing_keys = sorted(missing_keys)
645
bytes = bencode.bencode((write_group_tokens, missing_keys))
646
self.repository.unlock()
647
return SuccessfulSmartServerResponse(('missing-basis', bytes))
649
self.repository.unlock()
650
return SuccessfulSmartServerResponse(('ok', ))
653
class SmartServerRepositoryInsertStream(SmartServerRepositoryInsertStreamLocked):
654
"""Insert a record stream from a RemoteSink into an unlocked repository.
656
This is the same as SmartServerRepositoryInsertStreamLocked, except it
657
takes no lock_tokens; i.e. it works with an unlocked (or lock-free, e.g.
658
like pack format) repository.
663
def do_repository_request(self, repository, resume_tokens):
664
"""StreamSink.insert_stream for a remote repository."""
665
repository.lock_write()
666
self.do_insert_stream_request(repository, resume_tokens)
added
removed
bzrlib/smart/repository.py
