860
846
# versioned roots do not change unless the tree found a change.
863
class RepositoryWriteLockResult(object):
864
"""The result of write locking a repository.
866
:ivar repository_token: The token obtained from the underlying lock, or
868
:ivar unlock: A callable which will unlock the lock.
871
def __init__(self, unlock, repository_token):
872
self.repository_token = repository_token
876
849
######################################################################
880
class Repository(_RelockDebugMixin, bzrdir.ControlComponent):
853
class Repository(object):
881
854
"""Repository holding history for one or more branches.
883
856
The repository holds and retrieves historical information including
1319
1293
self._reconcile_does_inventory_gc = True
1320
1294
self._reconcile_fixes_text_parents = False
1321
1295
self._reconcile_backsup_inventory = True
1296
# not right yet - should be more semantically clear ?
1298
# TODO: make sure to construct the right store classes, etc, depending
1299
# on whether escaping is required.
1300
self._warn_if_deprecated()
1322
1301
self._write_group = None
1323
1302
# Additional places to query for data.
1324
1303
self._fallback_repositories = []
1325
1304
# An InventoryEntry cache, used during deserialization
1326
1305
self._inventory_entry_cache = fifo_cache.FIFOCache(10*1024)
1327
# Is it safe to return inventory entries directly from the entry cache,
1328
# rather copying them?
1329
self._safe_to_return_from_cache = False
1332
def user_transport(self):
1333
return self.bzrdir.user_transport
1336
def control_transport(self):
1337
return self._transport
1339
1307
def __repr__(self):
1340
1308
if self._fallback_repositories:
1403
1365
:raises MismatchedToken: if the specified token doesn't match the token
1404
1366
of the existing lock.
1405
1367
:seealso: start_write_group.
1406
:return: A RepositoryWriteLockResult.
1369
A token should be passed in if you know that you have locked the object
1370
some other way, and need to synchronise this object's state with that
1373
XXX: this docstring is duplicated in many places, e.g. lockable_files.py
1408
1375
locked = self.is_locked()
1409
token = self.control_files.lock_write(token=token)
1376
result = self.control_files.lock_write(token=token)
1411
self._warn_if_deprecated()
1412
self._note_lock('w')
1413
1378
for repo in self._fallback_repositories:
1414
1379
# Writes don't affect fallback repos
1415
1380
repo.lock_read()
1416
1381
self._refresh_data()
1417
return RepositoryWriteLockResult(self.unlock, token)
1419
1384
def lock_read(self):
1420
"""Lock the repository for read operations.
1422
:return: An object with an unlock method which will release the lock
1425
1385
locked = self.is_locked()
1426
1386
self.control_files.lock_read()
1428
self._warn_if_deprecated()
1429
self._note_lock('r')
1430
1388
for repo in self._fallback_repositories:
1431
1389
repo.lock_read()
1432
1390
self._refresh_data()
1435
1392
def get_physical_lock_status(self):
1436
1393
return self.control_files.get_physical_lock_status()
1527
1487
except errors.NoRepositoryPresent:
1530
return False, ([], repository)
1490
return False, (None, repository)
1531
1491
self.first_call = False
1532
value = (bzrdir.list_branches(), None)
1493
value = (bzrdir.open_branch(), None)
1494
except errors.NotBranchError:
1495
value = (None, None)
1533
1496
return True, value
1536
for branches, repository in bzrdir.BzrDir.find_bzrdirs(
1537
self.user_transport, evaluate=Evaluator()):
1538
if branches is not None:
1539
ret.extend(branches)
1499
for branch, repository in bzrdir.BzrDir.find_bzrdirs(
1500
self.bzrdir.root_transport, evaluate=Evaluator()):
1501
if branch is not None:
1502
branches.append(branch)
1540
1503
if not using and repository is not None:
1541
ret.extend(repository.find_branches())
1504
branches.extend(repository.find_branches())
1544
1507
@needs_read_lock
1545
1508
def search_missing_revision_ids(self, other, revision_id=None, find_ghosts=True):
1923
1885
rev = self._serializer.read_revision_from_string(text)
1924
1886
yield (revid, rev)
1889
def get_revision_xml(self, revision_id):
1890
# TODO: jam 20070210 This shouldn't be necessary since get_revision
1891
# would have already do it.
1892
# TODO: jam 20070210 Just use _serializer.write_revision_to_string()
1893
# TODO: this can't just be replaced by:
1894
# return self._serializer.write_revision_to_string(
1895
# self.get_revision(revision_id))
1896
# as cStringIO preservers the encoding unlike write_revision_to_string
1897
# or some other call down the path.
1898
rev = self.get_revision(revision_id)
1899
rev_tmp = cStringIO.StringIO()
1900
# the current serializer..
1901
self._serializer.write_revision(rev, rev_tmp)
1903
return rev_tmp.getvalue()
1926
1905
def get_deltas_for_revisions(self, revisions, specific_fileids=None):
1927
1906
"""Produce a generator of revision deltas.
2427
2410
next_key = None
2430
def _deserialise_inventory(self, revision_id, xml):
2413
def deserialise_inventory(self, revision_id, xml):
2431
2414
"""Transform the xml into an inventory object.
2433
2416
:param revision_id: The expected revision id of the inventory.
2434
2417
:param xml: A serialised inventory.
2436
2419
result = self._serializer.read_inventory_from_string(xml, revision_id,
2437
entry_cache=self._inventory_entry_cache,
2438
return_from_cache=self._safe_to_return_from_cache)
2420
entry_cache=self._inventory_entry_cache)
2439
2421
if result.revision_id != revision_id:
2440
2422
raise AssertionError('revision id mismatch %s != %s' % (
2441
2423
result.revision_id, revision_id))
2426
def serialise_inventory(self, inv):
2427
return self._serializer.write_inventory_to_string(inv)
2429
def _serialise_inventory_to_lines(self, inv):
2430
return self._serializer.write_inventory_to_lines(inv)
2444
2432
def get_serializer_format(self):
2445
2433
return self._serializer.format_num
2447
2435
@needs_read_lock
2448
def _get_inventory_xml(self, revision_id):
2449
"""Get serialized inventory as a string."""
2436
def get_inventory_xml(self, revision_id):
2437
"""Get inventory XML as a file object."""
2450
2438
texts = self._iter_inventory_xmls([revision_id], 'unordered')
2452
2440
text, revision_id = texts.next()
2511
2505
next_id = parents[0]
2508
def get_revision_inventory(self, revision_id):
2509
"""Return inventory of a past revision."""
2510
# TODO: Unify this with get_inventory()
2511
# bzr 0.0.6 and later imposes the constraint that the inventory_id
2512
# must be the same as its revision, so this is trivial.
2513
if revision_id is None:
2514
# This does not make sense: if there is no revision,
2515
# then it is the current tree inventory surely ?!
2516
# and thus get_root_id() is something that looks at the last
2517
# commit on the branch, and the get_root_id is an inventory check.
2518
raise NotImplementedError
2519
# return Inventory(self.get_root_id())
2521
return self.get_inventory(revision_id)
2513
2523
def is_shared(self):
2514
2524
"""Return True if this repository is flagged as a shared repository."""
2515
2525
raise NotImplementedError(self.is_shared)
2766
2764
result.check(callback_refs)
2769
def _warn_if_deprecated(self, branch=None):
2767
def _warn_if_deprecated(self):
2770
2768
global _deprecation_warning_done
2771
2769
if _deprecation_warning_done:
2775
conf = config.GlobalConfig()
2777
conf = branch.get_config()
2778
if conf.suppress_warning('format_deprecation'):
2780
warning("Format %s for %s is deprecated -"
2781
" please use 'bzr upgrade' to get better performance"
2782
% (self._format, self.bzrdir.transport.base))
2784
_deprecation_warning_done = True
2771
_deprecation_warning_done = True
2772
warning("Format %s for %s is deprecated - please use 'bzr upgrade' to get better performance"
2773
% (self._format, self.bzrdir.transport.base))
2786
2775
def supports_rich_root(self):
2787
2776
return self._format.rich_root_data
3418
3396
:param revision_id: if None all content is copied, if NULL_REVISION no
3419
3397
content is copied.
3398
:param pb: optional progress bar to use for progress reports. If not
3399
provided a default one will be created.
3423
ui.ui_factory.warn_experimental_format_fetch(self)
3424
3402
from bzrlib.fetch import RepoFetcher
3425
# See <https://launchpad.net/bugs/456077> asking for a warning here
3426
if self.source._format.network_name() != self.target._format.network_name():
3427
ui.ui_factory.show_user_warning('cross_format_fetch',
3428
from_format=self.source._format,
3429
to_format=self.target._format)
3430
3403
f = RepoFetcher(to_repository=self.target,
3431
3404
from_repository=self.source,
3432
3405
last_revision=revision_id,
3433
3406
fetch_spec=fetch_spec,
3434
find_ghosts=find_ghosts)
3407
pb=pb, find_ghosts=find_ghosts)
3436
3409
def _walk_to_common_revisions(self, revision_ids):
3437
3410
"""Walk out from revision_ids in source to revisions target has.
3762
def _get_trees(self, revision_ids, cache):
3764
for rev_id in revision_ids:
3766
possible_trees.append((rev_id, cache[rev_id]))
3768
# Not cached, but inventory might be present anyway.
3770
tree = self.source.revision_tree(rev_id)
3771
except errors.NoSuchRevision:
3772
# Nope, parent is ghost.
3775
cache[rev_id] = tree
3776
possible_trees.append((rev_id, tree))
3777
return possible_trees
3779
def _get_delta_for_revision(self, tree, parent_ids, possible_trees):
3735
def _get_delta_for_revision(self, tree, parent_ids, basis_id, cache):
3780
3736
"""Get the best delta and base for this revision.
3782
3738
:return: (basis_id, delta)
3740
possible_trees = [(parent_id, cache[parent_id])
3741
for parent_id in parent_ids
3742
if parent_id in cache]
3743
if len(possible_trees) == 0:
3744
# There either aren't any parents, or the parents aren't in the
3745
# cache, so just use the last converted tree
3746
possible_trees.append((basis_id, cache[basis_id]))
3785
# Generate deltas against each tree, to find the shortest.
3786
texts_possibly_new_in_tree = set()
3787
3748
for basis_id, basis_tree in possible_trees:
3788
3749
delta = tree.inventory._make_delta(basis_tree.inventory)
3789
for old_path, new_path, file_id, new_entry in delta:
3790
if new_path is None:
3791
# This file_id isn't present in the new rev, so we don't
3795
# Rich roots are handled elsewhere...
3797
kind = new_entry.kind
3798
if kind != 'directory' and kind != 'file':
3799
# No text record associated with this inventory entry.
3801
# This is a directory or file that has changed somehow.
3802
texts_possibly_new_in_tree.add((file_id, new_entry.revision))
3803
3750
deltas.append((len(delta), basis_id, delta))
3805
3752
return deltas[0][1:]
3807
def _fetch_parent_invs_for_stacking(self, parent_map, cache):
3808
"""Find all parent revisions that are absent, but for which the
3809
inventory is present, and copy those inventories.
3811
This is necessary to preserve correctness when the source is stacked
3812
without fallbacks configured. (Note that in cases like upgrade the
3813
source may be not have _fallback_repositories even though it is
3817
for parents in parent_map.values():
3818
parent_revs.update(parents)
3819
present_parents = self.source.get_parent_map(parent_revs)
3820
absent_parents = set(parent_revs).difference(present_parents)
3821
parent_invs_keys_for_stacking = self.source.inventories.get_parent_map(
3822
(rev_id,) for rev_id in absent_parents)
3823
parent_inv_ids = [key[-1] for key in parent_invs_keys_for_stacking]
3824
for parent_tree in self.source.revision_trees(parent_inv_ids):
3825
current_revision_id = parent_tree.get_revision_id()
3826
parents_parents_keys = parent_invs_keys_for_stacking[
3827
(current_revision_id,)]
3828
parents_parents = [key[-1] for key in parents_parents_keys]
3829
basis_id = _mod_revision.NULL_REVISION
3830
basis_tree = self.source.revision_tree(basis_id)
3831
delta = parent_tree.inventory._make_delta(basis_tree.inventory)
3832
self.target.add_inventory_by_delta(
3833
basis_id, delta, current_revision_id, parents_parents)
3834
cache[current_revision_id] = parent_tree
3836
def _fetch_batch(self, revision_ids, basis_id, cache, a_graph=None):
3754
def _fetch_batch(self, revision_ids, basis_id, cache):
3837
3755
"""Fetch across a few revisions.
3839
3757
:param revision_ids: The revisions to copy
3840
3758
:param basis_id: The revision_id of a tree that must be in cache, used
3841
3759
as a basis for delta when no other base is available
3842
3760
:param cache: A cache of RevisionTrees that we can use.
3843
:param a_graph: A Graph object to determine the heads() of the
3844
rich-root data stream.
3845
3761
:return: The revision_id of the last converted tree. The RevisionTree
3846
3762
for it will be in cache
3853
3769
pending_deltas = []
3854
3770
pending_revisions = []
3855
3771
parent_map = self.source.get_parent_map(revision_ids)
3856
self._fetch_parent_invs_for_stacking(parent_map, cache)
3857
self.source._safe_to_return_from_cache = True
3858
3772
for tree in self.source.revision_trees(revision_ids):
3859
# Find a inventory delta for this revision.
3860
# Find text entries that need to be copied, too.
3861
3773
current_revision_id = tree.get_revision_id()
3862
3774
parent_ids = parent_map.get(current_revision_id, ())
3863
parent_trees = self._get_trees(parent_ids, cache)
3864
possible_trees = list(parent_trees)
3865
if len(possible_trees) == 0:
3866
# There either aren't any parents, or the parents are ghosts,
3867
# so just use the last converted tree.
3868
possible_trees.append((basis_id, cache[basis_id]))
3869
3775
basis_id, delta = self._get_delta_for_revision(tree, parent_ids,
3871
revision = self.source.get_revision(current_revision_id)
3872
pending_deltas.append((basis_id, delta,
3873
current_revision_id, revision.parent_ids))
3874
3777
if self._converting_to_rich_root:
3875
3778
self._revision_id_to_root_id[current_revision_id] = \
3876
3779
tree.get_root_id()
3877
# Determine which texts are in present in this revision but not in
3878
# any of the available parents.
3879
texts_possibly_new_in_tree = set()
3780
# Find text entries that need to be copied
3880
3781
for old_path, new_path, file_id, entry in delta:
3881
if new_path is None:
3882
# This file_id isn't present in the new rev
3886
if not self.target.supports_rich_root():
3887
# The target doesn't support rich root, so we don't
3890
if self._converting_to_rich_root:
3891
# This can't be copied normally, we have to insert
3893
root_keys_to_create.add((file_id, entry.revision))
3896
texts_possibly_new_in_tree.add((file_id, entry.revision))
3897
for basis_id, basis_tree in possible_trees:
3898
basis_inv = basis_tree.inventory
3899
for file_key in list(texts_possibly_new_in_tree):
3900
file_id, file_revision = file_key
3902
entry = basis_inv[file_id]
3903
except errors.NoSuchId:
3905
if entry.revision == file_revision:
3906
texts_possibly_new_in_tree.remove(file_key)
3907
text_keys.update(texts_possibly_new_in_tree)
3782
if new_path is not None:
3785
if not self.target.supports_rich_root():
3786
# The target doesn't support rich root, so we don't
3789
if self._converting_to_rich_root:
3790
# This can't be copied normally, we have to insert
3792
root_keys_to_create.add((file_id, entry.revision))
3794
text_keys.add((file_id, entry.revision))
3795
revision = self.source.get_revision(current_revision_id)
3796
pending_deltas.append((basis_id, delta,
3797
current_revision_id, revision.parent_ids))
3908
3798
pending_revisions.append(revision)
3909
3799
cache[current_revision_id] = tree
3910
3800
basis_id = current_revision_id
3911
self.source._safe_to_return_from_cache = False
3912
3801
# Copy file texts
3913
3802
from_texts = self.source.texts
3914
3803
to_texts = self.target.texts
3915
3804
if root_keys_to_create:
3916
root_stream = _mod_fetch._new_root_data_stream(
3805
from bzrlib.fetch import _new_root_data_stream
3806
root_stream = _new_root_data_stream(
3917
3807
root_keys_to_create, self._revision_id_to_root_id, parent_map,
3918
self.source, graph=a_graph)
3919
3809
to_texts.insert_record_stream(root_stream)
3920
3810
to_texts.insert_record_stream(from_texts.get_record_stream(
3921
3811
text_keys, self.target._format._fetch_order,
3943
3833
for parent_tree in self.source.revision_trees(parent_map):
3944
3834
current_revision_id = parent_tree.get_revision_id()
3945
3835
parents_parents = parent_map[current_revision_id]
3946
possible_trees = self._get_trees(parents_parents, cache)
3947
if len(possible_trees) == 0:
3948
# There either aren't any parents, or the parents are
3949
# ghosts, so just use the last converted tree.
3950
possible_trees.append((basis_id, cache[basis_id]))
3951
3836
basis_id, delta = self._get_delta_for_revision(parent_tree,
3952
parents_parents, possible_trees)
3837
parents_parents, basis_id, cache)
3953
3838
self.target.add_inventory_by_delta(
3954
3839
basis_id, delta, current_revision_id, parents_parents)
3955
3840
# insert signatures and revisions
3978
3863
cache[basis_id] = basis_tree
3979
3864
del basis_tree # We don't want to hang on to it here
3981
if self._converting_to_rich_root and len(revision_ids) > 100:
3982
a_graph = _mod_fetch._get_rich_root_heads_graph(self.source,
3987
3866
for offset in range(0, len(revision_ids), batch_size):
3988
3867
self.target.start_write_group()
3990
3869
pb.update('Transferring revisions', offset,
3991
3870
len(revision_ids))
3992
3871
batch = revision_ids[offset:offset+batch_size]
3993
basis_id = self._fetch_batch(batch, basis_id, cache,
3872
basis_id = self._fetch_batch(batch, basis_id, cache)
3996
self.source._safe_to_return_from_cache = False
3997
3874
self.target.abort_write_group()
4011
3888
"""See InterRepository.fetch()."""
4012
3889
if fetch_spec is not None:
4013
3890
raise AssertionError("Not implemented yet...")
4014
ui.ui_factory.warn_experimental_format_fetch(self)
4015
3891
if (not self.source.supports_rich_root()
4016
3892
and self.target.supports_rich_root()):
4017
3893
self._converting_to_rich_root = True
4018
3894
self._revision_id_to_root_id = {}
4020
3896
self._converting_to_rich_root = False
4021
# See <https://launchpad.net/bugs/456077> asking for a warning here
4022
if self.source._format.network_name() != self.target._format.network_name():
4023
ui.ui_factory.show_user_warning('cross_format_fetch',
4024
from_format=self.source._format,
4025
to_format=self.target._format)
4026
3897
revision_ids = self.target.search_missing_revision_ids(self.source,
4027
3898
revision_id, find_ghosts=find_ghosts).get_keys()
4028
3899
if not revision_ids:
4097
3968
:param to_convert: The disk object to convert.
4098
3969
:param pb: a progress bar to use for progress information.
4100
pb = ui.ui_factory.nested_progress_bar()
4103
3974
# this is only useful with metadir layouts - separated repo content.
4104
3975
# trigger an assertion if not such
4105
3976
repo._format.get_format_string()
4106
3977
self.repo_dir = repo.bzrdir
4107
pb.update('Moving repository to repository.backup')
3978
self.step('Moving repository to repository.backup')
4108
3979
self.repo_dir.transport.move('repository', 'repository.backup')
4109
3980
backup_transport = self.repo_dir.transport.clone('repository.backup')
4110
3981
repo._format.check_conversion_target(self.target_format)
4111
3982
self.source_repo = repo._format.open(self.repo_dir,
4113
3984
_override_transport=backup_transport)
4114
pb.update('Creating new repository')
3985
self.step('Creating new repository')
4115
3986
converted = self.target_format.initialize(self.repo_dir,
4116
3987
self.source_repo.is_shared())
4117
3988
converted.lock_write()
4119
pb.update('Copying content')
3990
self.step('Copying content into repository.')
4120
3991
self.source_repo.copy_content_into(converted)
4122
3993
converted.unlock()
4123
pb.update('Deleting old repository content')
3994
self.step('Deleting old repository content.')
4124
3995
self.repo_dir.transport.delete_tree('repository.backup')
4125
ui.ui_factory.note('repository converted')
3996
self.pb.note('repository converted')
3998
def step(self, message):
3999
"""Update the pb by a step."""
4001
self.pb.update(message, self.count, self.total)
4129
4004
_unescape_map = {
4622
4491
def _get_convertable_inventory_stream(self, revision_ids,
4623
4492
delta_versus_null=False):
4624
# The two formats are sufficiently different that there is no fast
4625
# path, so we need to send just inventorydeltas, which any
4626
# sufficiently modern client can insert into any repository.
4627
# The StreamSink code expects to be able to
4493
# The source is using CHKs, but the target either doesn't or it has a
4494
# different serializer. The StreamSink code expects to be able to
4628
4495
# convert on the target, so we need to put bytes-on-the-wire that can
4629
4496
# be converted. That means inventory deltas (if the remote is <1.19,
4630
4497
# RemoteStreamSink will fallback to VFS to insert the deltas).