1
# Copyright (C) 2005-2010 Canonical Ltd
1
# Copyright (C) 2005, 2006, 2007, 2008, 2009 Canonical Ltd
3
3
# This program is free software; you can redistribute it and/or modify
4
4
# it under the terms of the GNU General Public License as published by
52
51
from bzrlib.testament import Testament
59
54
from bzrlib.decorators import needs_read_lock, needs_write_lock, only_raises
60
55
from bzrlib.inter import InterObject
61
56
from bzrlib.inventory import (
67
from bzrlib.lock import _RelockDebugMixin, LogicalLockResult
62
from bzrlib.lock import _RelockDebugMixin
63
from bzrlib import registry
68
64
from bzrlib.trace import (
69
65
log_exception_quietly, note, mutter, mutter_callsite, warning)
73
69
_deprecation_warning_done = False
76
class IsInWriteGroupError(errors.InternalBzrError):
78
_fmt = "May not refresh_data of repo %(repo)s while in a write group."
80
def __init__(self, repo):
81
errors.InternalBzrError.__init__(self, repo=repo)
84
72
class CommitBuilder(object):
85
73
"""Provides an interface to build up a commit.
870
858
# versioned roots do not change unless the tree found a change.
873
class RepositoryWriteLockResult(LogicalLockResult):
874
"""The result of write locking a repository.
876
:ivar repository_token: The token obtained from the underlying lock, or
878
:ivar unlock: A callable which will unlock the lock.
881
def __init__(self, unlock, repository_token):
882
LogicalLockResult.__init__(self, unlock)
883
self.repository_token = repository_token
886
return "RepositoryWriteLockResult(%s, %s)" % (self.repository_token,
890
861
######################################################################
894
class Repository(_RelockDebugMixin, bzrdir.ControlComponent):
865
class Repository(_RelockDebugMixin):
895
866
"""Repository holding history for one or more branches.
897
868
The repository holds and retrieves historical information including
1056
1027
:seealso: add_inventory, for the contract.
1058
inv_lines = self._serializer.write_inventory_to_lines(inv)
1029
inv_lines = self._serialise_inventory_to_lines(inv)
1059
1030
return self._inventory_add_lines(revision_id, parents,
1060
1031
inv_lines, check_content=False)
1268
1239
"""Check a single text from this repository."""
1269
1240
if kind == 'inventories':
1270
1241
rev_id = record.key[0]
1271
inv = self._deserialise_inventory(rev_id,
1242
inv = self.deserialise_inventory(rev_id,
1272
1243
record.get_bytes_as('fulltext'))
1273
1244
if last_object is not None:
1274
1245
delta = inv._make_delta(last_object)
1319
1290
:param _format: The format of the repository on disk.
1320
1291
:param a_bzrdir: The BzrDir of the repository.
1293
In the future we will have a single api for all stores for
1294
getting file texts, inventories and revisions, then
1295
this construct will accept instances of those things.
1322
# In the future we will have a single api for all stores for
1323
# getting file texts, inventories and revisions, then
1324
# this construct will accept instances of those things.
1325
1297
super(Repository, self).__init__()
1326
1298
self._format = _format
1327
1299
# the following are part of the public API for Repository:
1342
1314
# rather copying them?
1343
1315
self._safe_to_return_from_cache = False
1346
def user_transport(self):
1347
return self.bzrdir.user_transport
1350
def control_transport(self):
1351
return self._transport
1353
1317
def __repr__(self):
1354
1318
if self._fallback_repositories:
1355
1319
return '%s(%r, fallback_repositories=%r)' % (
1403
1367
data during reads, and allows a 'write_group' to be obtained. Write
1404
1368
groups must be used for actual data insertion.
1406
A token should be passed in if you know that you have locked the object
1407
some other way, and need to synchronise this object's state with that
1410
XXX: this docstring is duplicated in many places, e.g. lockable_files.py
1412
1370
:param token: if this is already locked, then lock_write will fail
1413
1371
unless the token matches the existing lock.
1414
1372
:returns: a token if this instance supports tokens, otherwise None.
1417
1375
:raises MismatchedToken: if the specified token doesn't match the token
1418
1376
of the existing lock.
1419
1377
:seealso: start_write_group.
1420
:return: A RepositoryWriteLockResult.
1379
A token should be passed in if you know that you have locked the object
1380
some other way, and need to synchronise this object's state with that
1383
XXX: this docstring is duplicated in many places, e.g. lockable_files.py
1422
1385
locked = self.is_locked()
1423
token = self.control_files.lock_write(token=token)
1386
result = self.control_files.lock_write(token=token)
1425
1388
self._warn_if_deprecated()
1426
1389
self._note_lock('w')
1428
1391
# Writes don't affect fallback repos
1429
1392
repo.lock_read()
1430
1393
self._refresh_data()
1431
return RepositoryWriteLockResult(self.unlock, token)
1433
1396
def lock_read(self):
1434
"""Lock the repository for read operations.
1436
:return: An object with an unlock method which will release the lock
1439
1397
locked = self.is_locked()
1440
1398
self.control_files.lock_read()
1511
1468
# now gather global repository information
1512
1469
# XXX: This is available for many repos regardless of listability.
1513
if self.user_transport.listable():
1470
if self.bzrdir.root_transport.listable():
1514
1471
# XXX: do we want to __define len__() ?
1515
1472
# Maybe the versionedfiles object should provide a different
1516
1473
# method to get the number of keys.
1526
1483
:param using: If True, list only branches using this repository.
1528
1485
if using and not self.is_shared():
1529
return self.bzrdir.list_branches()
1487
return [self.bzrdir.open_branch()]
1488
except errors.NotBranchError:
1530
1490
class Evaluator(object):
1532
1492
def __init__(self):
1541
1501
except errors.NoRepositoryPresent:
1544
return False, ([], repository)
1504
return False, (None, repository)
1545
1505
self.first_call = False
1546
value = (bzrdir.list_branches(), None)
1507
value = (bzrdir.open_branch(), None)
1508
except errors.NotBranchError:
1509
value = (None, None)
1547
1510
return True, value
1550
for branches, repository in bzrdir.BzrDir.find_bzrdirs(
1551
self.user_transport, evaluate=Evaluator()):
1552
if branches is not None:
1553
ret.extend(branches)
1513
for branch, repository in bzrdir.BzrDir.find_bzrdirs(
1514
self.bzrdir.root_transport, evaluate=Evaluator()):
1515
if branch is not None:
1516
branches.append(branch)
1554
1517
if not using and repository is not None:
1555
ret.extend(repository.find_branches())
1518
branches.extend(repository.find_branches())
1558
1521
@needs_read_lock
1559
1522
def search_missing_revision_ids(self, other, revision_id=None, find_ghosts=True):
1668
1631
return missing_keys
1670
1633
def refresh_data(self):
1671
"""Re-read any data needed to synchronise with disk.
1634
"""Re-read any data needed to to synchronise with disk.
1673
1636
This method is intended to be called after another repository instance
1674
1637
(such as one used by a smart server) has inserted data into the
1675
repository. On all repositories this will work outside of write groups.
1676
Some repository formats (pack and newer for bzrlib native formats)
1677
support refresh_data inside write groups. If called inside a write
1678
group on a repository that does not support refreshing in a write group
1679
IsInWriteGroupError will be raised.
1638
repository. It may not be called during a write group, but may be
1639
called at any other time.
1641
if self.is_in_write_group():
1642
raise errors.InternalBzrError(
1643
"May not refresh_data while in a write group.")
1681
1644
self._refresh_data()
1683
1646
def resume_write_group(self, tokens):
1937
1900
rev = self._serializer.read_revision_from_string(text)
1938
1901
yield (revid, rev)
1904
def get_revision_xml(self, revision_id):
1905
# TODO: jam 20070210 This shouldn't be necessary since get_revision
1906
# would have already do it.
1907
# TODO: jam 20070210 Just use _serializer.write_revision_to_string()
1908
# TODO: this can't just be replaced by:
1909
# return self._serializer.write_revision_to_string(
1910
# self.get_revision(revision_id))
1911
# as cStringIO preservers the encoding unlike write_revision_to_string
1912
# or some other call down the path.
1913
rev = self.get_revision(revision_id)
1914
rev_tmp = cStringIO.StringIO()
1915
# the current serializer..
1916
self._serializer.write_revision(rev, rev_tmp)
1918
return rev_tmp.getvalue()
1940
1920
def get_deltas_for_revisions(self, revisions, specific_fileids=None):
1941
1921
"""Produce a generator of revision deltas.
2185
2165
selected_keys = set((revid,) for revid in revision_ids)
2186
2166
w = _inv_weave or self.inventories
2187
return self._find_file_ids_from_xml_inventory_lines(
2188
w.iter_lines_added_or_present_in_keys(
2189
selected_keys, pb=None),
2167
pb = ui.ui_factory.nested_progress_bar()
2169
return self._find_file_ids_from_xml_inventory_lines(
2170
w.iter_lines_added_or_present_in_keys(
2171
selected_keys, pb=pb),
2192
2176
def iter_files_bytes(self, desired_files):
2193
2177
"""Iterate through file versions.
2403
2387
"""single-document based inventory iteration."""
2404
2388
inv_xmls = self._iter_inventory_xmls(revision_ids, ordering)
2405
2389
for text, revision_id in inv_xmls:
2406
yield self._deserialise_inventory(revision_id, text)
2390
yield self.deserialise_inventory(revision_id, text)
2408
2392
def _iter_inventory_xmls(self, revision_ids, ordering):
2409
2393
if ordering is None:
2441
2425
next_key = None
2444
def _deserialise_inventory(self, revision_id, xml):
2428
def deserialise_inventory(self, revision_id, xml):
2445
2429
"""Transform the xml into an inventory object.
2447
2431
:param revision_id: The expected revision id of the inventory.
2455
2439
result.revision_id, revision_id))
2442
def serialise_inventory(self, inv):
2443
return self._serializer.write_inventory_to_string(inv)
2445
def _serialise_inventory_to_lines(self, inv):
2446
return self._serializer.write_inventory_to_lines(inv)
2458
2448
def get_serializer_format(self):
2459
2449
return self._serializer.format_num
2461
2451
@needs_read_lock
2462
def _get_inventory_xml(self, revision_id):
2463
"""Get serialized inventory as a string."""
2452
def get_inventory_xml(self, revision_id):
2453
"""Get inventory XML as a file object."""
2464
2454
texts = self._iter_inventory_xmls([revision_id], 'unordered')
2466
2456
text, revision_id = texts.next()
2468
2458
raise errors.HistoryMissing(self, 'inventory', revision_id)
2462
def get_inventory_sha1(self, revision_id):
2463
"""Return the sha1 hash of the inventory entry
2465
return self.get_revision(revision_id).inventory_sha1
2471
2467
def get_rev_id_for_revno(self, revno, known_pair):
2472
2468
"""Return the revision id of a revno, given a later (revno, revid)
2473
2469
pair in the same history.
2525
2521
next_id = parents[0]
2524
def get_revision_inventory(self, revision_id):
2525
"""Return inventory of a past revision."""
2526
# TODO: Unify this with get_inventory()
2527
# bzr 0.0.6 and later imposes the constraint that the inventory_id
2528
# must be the same as its revision, so this is trivial.
2529
if revision_id is None:
2530
# This does not make sense: if there is no revision,
2531
# then it is the current tree inventory surely ?!
2532
# and thus get_root_id() is something that looks at the last
2533
# commit on the branch, and the get_root_id is an inventory check.
2534
raise NotImplementedError
2535
# return Inventory(self.get_root_id())
2537
return self.get_inventory(revision_id)
2527
2539
def is_shared(self):
2528
2540
"""Return True if this repository is flagged as a shared repository."""
2529
2541
raise NotImplementedError(self.is_shared)
2563
2575
return RevisionTree(self, Inventory(root_id=None),
2564
2576
_mod_revision.NULL_REVISION)
2566
inv = self.get_inventory(revision_id)
2578
inv = self.get_revision_inventory(revision_id)
2567
2579
return RevisionTree(self, inv, revision_id)
2569
2581
def revision_trees(self, revision_ids):
2622
2634
keys = tsort.topo_sort(parent_map)
2623
2635
return [None] + list(keys)
2625
def pack(self, hint=None, clean_obsolete_packs=False):
2637
def pack(self, hint=None):
2626
2638
"""Compress the data within the repository.
2628
2640
This operation only makes sense for some repository types. For other
2638
2650
obtained from the result of commit_write_group(). Out of
2639
2651
date hints are simply ignored, because concurrent operations
2640
2652
can obsolete them rapidly.
2642
:param clean_obsolete_packs: Clean obsolete packs immediately after
2646
2655
def get_transaction(self):
2671
2680
def _make_parents_provider(self):
2675
def get_known_graph_ancestry(self, revision_ids):
2676
"""Return the known graph for a set of revision ids and their ancestors.
2678
st = static_tuple.StaticTuple
2679
revision_keys = [st(r_id).intern() for r_id in revision_ids]
2680
known_graph = self.revisions.get_known_graph_ancestry(revision_keys)
2681
return graph.GraphThunkIdsToKeys(known_graph)
2683
2683
def get_graph(self, other_repository=None):
2684
2684
"""Return the graph walker for this repository format"""
2685
2685
parents_provider = self._make_parents_provider()
3084
3084
pack_compresses = False
3085
3085
# Does the repository inventory storage understand references to trees?
3086
3086
supports_tree_reference = None
3087
# Is the format experimental ?
3088
experimental = False
3091
return "%s()" % self.__class__.__name__
3089
return "<%s>" % self.__class__.__name__
3093
3091
def __eq__(self, other):
3094
3092
# format objects are generally stateless
3213
3211
raise NotImplementedError(self.open)
3215
def _run_post_repo_init_hooks(self, repository, a_bzrdir, shared):
3216
from bzrlib.bzrdir import BzrDir, RepoInitHookParams
3217
hooks = BzrDir.hooks['post_repo_init']
3220
params = RepoInitHookParams(repository, self, a_bzrdir, shared)
3225
3214
class MetaDirRepositoryFormat(RepositoryFormat):
3226
3215
"""Common base class for the new repositories using the metadir layout."""
3432
3421
:param revision_id: if None all content is copied, if NULL_REVISION no
3433
3422
content is copied.
3423
:param pb: optional progress bar to use for progress reports. If not
3424
provided a default one will be created.
3437
ui.ui_factory.warn_experimental_format_fetch(self)
3438
from bzrlib.fetch import RepoFetcher
3439
# See <https://launchpad.net/bugs/456077> asking for a warning here
3440
if self.source._format.network_name() != self.target._format.network_name():
3441
ui.ui_factory.show_user_warning('cross_format_fetch',
3442
from_format=self.source._format,
3443
to_format=self.target._format)
3444
f = RepoFetcher(to_repository=self.target,
3427
f = _mod_fetch.RepoFetcher(to_repository=self.target,
3445
3428
from_repository=self.source,
3446
3429
last_revision=revision_id,
3447
3430
fetch_spec=fetch_spec,
3448
find_ghosts=find_ghosts)
3431
pb=pb, find_ghosts=find_ghosts)
3450
3433
def _walk_to_common_revisions(self, revision_ids):
3451
3434
"""Walk out from revision_ids in source to revisions target has.
4025
4008
"""See InterRepository.fetch()."""
4026
4009
if fetch_spec is not None:
4027
4010
raise AssertionError("Not implemented yet...")
4028
ui.ui_factory.warn_experimental_format_fetch(self)
4029
4011
if (not self.source.supports_rich_root()
4030
4012
and self.target.supports_rich_root()):
4031
4013
self._converting_to_rich_root = True
4032
4014
self._revision_id_to_root_id = {}
4034
4016
self._converting_to_rich_root = False
4035
# See <https://launchpad.net/bugs/456077> asking for a warning here
4036
if self.source._format.network_name() != self.target._format.network_name():
4037
ui.ui_factory.show_user_warning('cross_format_fetch',
4038
from_format=self.source._format,
4039
to_format=self.target._format)
4040
4017
revision_ids = self.target.search_missing_revision_ids(self.source,
4041
4018
revision_id, find_ghosts=find_ghosts).get_keys()
4042
4019
if not revision_ids:
4111
4088
:param to_convert: The disk object to convert.
4112
4089
:param pb: a progress bar to use for progress information.
4114
pb = ui.ui_factory.nested_progress_bar()
4117
4094
# this is only useful with metadir layouts - separated repo content.
4118
4095
# trigger an assertion if not such
4119
4096
repo._format.get_format_string()
4120
4097
self.repo_dir = repo.bzrdir
4121
pb.update('Moving repository to repository.backup')
4098
self.step('Moving repository to repository.backup')
4122
4099
self.repo_dir.transport.move('repository', 'repository.backup')
4123
4100
backup_transport = self.repo_dir.transport.clone('repository.backup')
4124
4101
repo._format.check_conversion_target(self.target_format)
4125
4102
self.source_repo = repo._format.open(self.repo_dir,
4127
4104
_override_transport=backup_transport)
4128
pb.update('Creating new repository')
4105
self.step('Creating new repository')
4129
4106
converted = self.target_format.initialize(self.repo_dir,
4130
4107
self.source_repo.is_shared())
4131
4108
converted.lock_write()
4133
pb.update('Copying content')
4110
self.step('Copying content')
4134
4111
self.source_repo.copy_content_into(converted)
4136
4113
converted.unlock()
4137
pb.update('Deleting old repository content')
4114
self.step('Deleting old repository content')
4138
4115
self.repo_dir.transport.delete_tree('repository.backup')
4139
4116
ui.ui_factory.note('repository converted')
4118
def step(self, message):
4119
"""Update the pb by a step."""
4121
self.pb.update(message, self.count, self.total)
4143
4124
_unescape_map = {
4636
4617
def _get_convertable_inventory_stream(self, revision_ids,
4637
4618
delta_versus_null=False):
4638
# The two formats are sufficiently different that there is no fast
4639
# path, so we need to send just inventorydeltas, which any
4640
# sufficiently modern client can insert into any repository.
4641
# The StreamSink code expects to be able to
4619
# The source is using CHKs, but the target either doesn't or it has a
4620
# different serializer. The StreamSink code expects to be able to
4642
4621
# convert on the target, so we need to put bytes-on-the-wire that can
4643
4622
# be converted. That means inventory deltas (if the remote is <1.19,
4644
4623
# RemoteStreamSink will fallback to VFS to insert the deltas).
added
removed
bzrlib/repository.py
