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
return "RepositoryWriteLockResult(%s, %s)" % (self.repository_token,
880
849
######################################################################
884
class Repository(_RelockDebugMixin, bzrdir.ControlComponent):
853
class Repository(object):
885
854
"""Repository holding history for one or more branches.
887
856
The repository holds and retrieves historical information including
1323
1293
self._reconcile_does_inventory_gc = True
1324
1294
self._reconcile_fixes_text_parents = False
1325
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()
1326
1301
self._write_group = None
1327
1302
# Additional places to query for data.
1328
1303
self._fallback_repositories = []
1329
1304
# An InventoryEntry cache, used during deserialization
1330
1305
self._inventory_entry_cache = fifo_cache.FIFOCache(10*1024)
1331
# Is it safe to return inventory entries directly from the entry cache,
1332
# rather copying them?
1333
self._safe_to_return_from_cache = False
1336
def user_transport(self):
1337
return self.bzrdir.user_transport
1340
def control_transport(self):
1341
return self._transport
1343
1307
def __repr__(self):
1344
1308
if self._fallback_repositories:
1407
1365
:raises MismatchedToken: if the specified token doesn't match the token
1408
1366
of the existing lock.
1409
1367
:seealso: start_write_group.
1410
: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
1412
1375
locked = self.is_locked()
1413
token = self.control_files.lock_write(token=token)
1376
result = self.control_files.lock_write(token=token)
1415
self._warn_if_deprecated()
1416
self._note_lock('w')
1417
1378
for repo in self._fallback_repositories:
1418
1379
# Writes don't affect fallback repos
1419
1380
repo.lock_read()
1420
1381
self._refresh_data()
1421
return RepositoryWriteLockResult(self.unlock, token)
1423
1384
def lock_read(self):
1424
"""Lock the repository for read operations.
1426
:return: An object with an unlock method which will release the lock
1429
1385
locked = self.is_locked()
1430
1386
self.control_files.lock_read()
1432
self._warn_if_deprecated()
1433
self._note_lock('r')
1434
1388
for repo in self._fallback_repositories:
1435
1389
repo.lock_read()
1436
1390
self._refresh_data()
1439
1392
def get_physical_lock_status(self):
1440
1393
return self.control_files.get_physical_lock_status()
1531
1487
except errors.NoRepositoryPresent:
1534
return False, ([], repository)
1490
return False, (None, repository)
1535
1491
self.first_call = False
1536
value = (bzrdir.list_branches(), None)
1493
value = (bzrdir.open_branch(), None)
1494
except errors.NotBranchError:
1495
value = (None, None)
1537
1496
return True, value
1540
for branches, repository in bzrdir.BzrDir.find_bzrdirs(
1541
self.user_transport, evaluate=Evaluator()):
1542
if branches is not None:
1543
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)
1544
1503
if not using and repository is not None:
1545
ret.extend(repository.find_branches())
1504
branches.extend(repository.find_branches())
1548
1507
@needs_read_lock
1549
1508
def search_missing_revision_ids(self, other, revision_id=None, find_ghosts=True):
1927
1885
rev = self._serializer.read_revision_from_string(text)
1928
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()
1930
1905
def get_deltas_for_revisions(self, revisions, specific_fileids=None):
1931
1906
"""Produce a generator of revision deltas.
2431
2410
next_key = None
2434
def _deserialise_inventory(self, revision_id, xml):
2413
def deserialise_inventory(self, revision_id, xml):
2435
2414
"""Transform the xml into an inventory object.
2437
2416
:param revision_id: The expected revision id of the inventory.
2438
2417
:param xml: A serialised inventory.
2440
2419
result = self._serializer.read_inventory_from_string(xml, revision_id,
2441
entry_cache=self._inventory_entry_cache,
2442
return_from_cache=self._safe_to_return_from_cache)
2420
entry_cache=self._inventory_entry_cache)
2443
2421
if result.revision_id != revision_id:
2444
2422
raise AssertionError('revision id mismatch %s != %s' % (
2445
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)
2448
2432
def get_serializer_format(self):
2449
2433
return self._serializer.format_num
2451
2435
@needs_read_lock
2452
def _get_inventory_xml(self, revision_id):
2453
"""Get serialized inventory as a string."""
2436
def get_inventory_xml(self, revision_id):
2437
"""Get inventory XML as a file object."""
2454
2438
texts = self._iter_inventory_xmls([revision_id], 'unordered')
2456
2440
text, revision_id = texts.next()
2515
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)
2517
2523
def is_shared(self):
2518
2524
"""Return True if this repository is flagged as a shared repository."""
2519
2525
raise NotImplementedError(self.is_shared)
2770
2764
result.check(callback_refs)
2773
def _warn_if_deprecated(self, branch=None):
2767
def _warn_if_deprecated(self):
2774
2768
global _deprecation_warning_done
2775
2769
if _deprecation_warning_done:
2779
conf = config.GlobalConfig()
2781
conf = branch.get_config()
2782
if conf.suppress_warning('format_deprecation'):
2784
warning("Format %s for %s is deprecated -"
2785
" please use 'bzr upgrade' to get better performance"
2786
% (self._format, self.bzrdir.transport.base))
2788
_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))
2790
2775
def supports_rich_root(self):
2791
2776
return self._format.rich_root_data
3422
3396
:param revision_id: if None all content is copied, if NULL_REVISION no
3423
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.
3427
ui.ui_factory.warn_experimental_format_fetch(self)
3428
3402
from bzrlib.fetch import RepoFetcher
3429
# See <https://launchpad.net/bugs/456077> asking for a warning here
3430
if self.source._format.network_name() != self.target._format.network_name():
3431
ui.ui_factory.show_user_warning('cross_format_fetch',
3432
from_format=self.source._format,
3433
to_format=self.target._format)
3434
3403
f = RepoFetcher(to_repository=self.target,
3435
3404
from_repository=self.source,
3436
3405
last_revision=revision_id,
3437
3406
fetch_spec=fetch_spec,
3438
find_ghosts=find_ghosts)
3407
pb=pb, find_ghosts=find_ghosts)
3440
3409
def _walk_to_common_revisions(self, revision_ids):
3441
3410
"""Walk out from revision_ids in source to revisions target has.
3766
def _get_trees(self, revision_ids, cache):
3768
for rev_id in revision_ids:
3770
possible_trees.append((rev_id, cache[rev_id]))
3772
# Not cached, but inventory might be present anyway.
3774
tree = self.source.revision_tree(rev_id)
3775
except errors.NoSuchRevision:
3776
# Nope, parent is ghost.
3779
cache[rev_id] = tree
3780
possible_trees.append((rev_id, tree))
3781
return possible_trees
3783
def _get_delta_for_revision(self, tree, parent_ids, possible_trees):
3735
def _get_delta_for_revision(self, tree, parent_ids, basis_id, cache):
3784
3736
"""Get the best delta and base for this revision.
3786
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]))
3789
# Generate deltas against each tree, to find the shortest.
3790
texts_possibly_new_in_tree = set()
3791
3748
for basis_id, basis_tree in possible_trees:
3792
3749
delta = tree.inventory._make_delta(basis_tree.inventory)
3793
for old_path, new_path, file_id, new_entry in delta:
3794
if new_path is None:
3795
# This file_id isn't present in the new rev, so we don't
3799
# Rich roots are handled elsewhere...
3801
kind = new_entry.kind
3802
if kind != 'directory' and kind != 'file':
3803
# No text record associated with this inventory entry.
3805
# This is a directory or file that has changed somehow.
3806
texts_possibly_new_in_tree.add((file_id, new_entry.revision))
3807
3750
deltas.append((len(delta), basis_id, delta))
3809
3752
return deltas[0][1:]
3811
def _fetch_parent_invs_for_stacking(self, parent_map, cache):
3812
"""Find all parent revisions that are absent, but for which the
3813
inventory is present, and copy those inventories.
3815
This is necessary to preserve correctness when the source is stacked
3816
without fallbacks configured. (Note that in cases like upgrade the
3817
source may be not have _fallback_repositories even though it is
3821
for parents in parent_map.values():
3822
parent_revs.update(parents)
3823
present_parents = self.source.get_parent_map(parent_revs)
3824
absent_parents = set(parent_revs).difference(present_parents)
3825
parent_invs_keys_for_stacking = self.source.inventories.get_parent_map(
3826
(rev_id,) for rev_id in absent_parents)
3827
parent_inv_ids = [key[-1] for key in parent_invs_keys_for_stacking]
3828
for parent_tree in self.source.revision_trees(parent_inv_ids):
3829
current_revision_id = parent_tree.get_revision_id()
3830
parents_parents_keys = parent_invs_keys_for_stacking[
3831
(current_revision_id,)]
3832
parents_parents = [key[-1] for key in parents_parents_keys]
3833
basis_id = _mod_revision.NULL_REVISION
3834
basis_tree = self.source.revision_tree(basis_id)
3835
delta = parent_tree.inventory._make_delta(basis_tree.inventory)
3836
self.target.add_inventory_by_delta(
3837
basis_id, delta, current_revision_id, parents_parents)
3838
cache[current_revision_id] = parent_tree
3840
def _fetch_batch(self, revision_ids, basis_id, cache, a_graph=None):
3754
def _fetch_batch(self, revision_ids, basis_id, cache):
3841
3755
"""Fetch across a few revisions.
3843
3757
:param revision_ids: The revisions to copy
3844
3758
:param basis_id: The revision_id of a tree that must be in cache, used
3845
3759
as a basis for delta when no other base is available
3846
3760
:param cache: A cache of RevisionTrees that we can use.
3847
:param a_graph: A Graph object to determine the heads() of the
3848
rich-root data stream.
3849
3761
:return: The revision_id of the last converted tree. The RevisionTree
3850
3762
for it will be in cache
3857
3769
pending_deltas = []
3858
3770
pending_revisions = []
3859
3771
parent_map = self.source.get_parent_map(revision_ids)
3860
self._fetch_parent_invs_for_stacking(parent_map, cache)
3861
self.source._safe_to_return_from_cache = True
3862
3772
for tree in self.source.revision_trees(revision_ids):
3863
# Find a inventory delta for this revision.
3864
# Find text entries that need to be copied, too.
3865
3773
current_revision_id = tree.get_revision_id()
3866
3774
parent_ids = parent_map.get(current_revision_id, ())
3867
parent_trees = self._get_trees(parent_ids, cache)
3868
possible_trees = list(parent_trees)
3869
if len(possible_trees) == 0:
3870
# There either aren't any parents, or the parents are ghosts,
3871
# so just use the last converted tree.
3872
possible_trees.append((basis_id, cache[basis_id]))
3873
3775
basis_id, delta = self._get_delta_for_revision(tree, parent_ids,
3875
revision = self.source.get_revision(current_revision_id)
3876
pending_deltas.append((basis_id, delta,
3877
current_revision_id, revision.parent_ids))
3878
3777
if self._converting_to_rich_root:
3879
3778
self._revision_id_to_root_id[current_revision_id] = \
3880
3779
tree.get_root_id()
3881
# Determine which texts are in present in this revision but not in
3882
# any of the available parents.
3883
texts_possibly_new_in_tree = set()
3780
# Find text entries that need to be copied
3884
3781
for old_path, new_path, file_id, entry in delta:
3885
if new_path is None:
3886
# This file_id isn't present in the new rev
3890
if not self.target.supports_rich_root():
3891
# The target doesn't support rich root, so we don't
3894
if self._converting_to_rich_root:
3895
# This can't be copied normally, we have to insert
3897
root_keys_to_create.add((file_id, entry.revision))
3900
texts_possibly_new_in_tree.add((file_id, entry.revision))
3901
for basis_id, basis_tree in possible_trees:
3902
basis_inv = basis_tree.inventory
3903
for file_key in list(texts_possibly_new_in_tree):
3904
file_id, file_revision = file_key
3906
entry = basis_inv[file_id]
3907
except errors.NoSuchId:
3909
if entry.revision == file_revision:
3910
texts_possibly_new_in_tree.remove(file_key)
3911
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))
3912
3798
pending_revisions.append(revision)
3913
3799
cache[current_revision_id] = tree
3914
3800
basis_id = current_revision_id
3915
self.source._safe_to_return_from_cache = False
3916
3801
# Copy file texts
3917
3802
from_texts = self.source.texts
3918
3803
to_texts = self.target.texts
3919
3804
if root_keys_to_create:
3920
root_stream = _mod_fetch._new_root_data_stream(
3805
from bzrlib.fetch import _new_root_data_stream
3806
root_stream = _new_root_data_stream(
3921
3807
root_keys_to_create, self._revision_id_to_root_id, parent_map,
3922
self.source, graph=a_graph)
3923
3809
to_texts.insert_record_stream(root_stream)
3924
3810
to_texts.insert_record_stream(from_texts.get_record_stream(
3925
3811
text_keys, self.target._format._fetch_order,
3947
3833
for parent_tree in self.source.revision_trees(parent_map):
3948
3834
current_revision_id = parent_tree.get_revision_id()
3949
3835
parents_parents = parent_map[current_revision_id]
3950
possible_trees = self._get_trees(parents_parents, cache)
3951
if len(possible_trees) == 0:
3952
# There either aren't any parents, or the parents are
3953
# ghosts, so just use the last converted tree.
3954
possible_trees.append((basis_id, cache[basis_id]))
3955
3836
basis_id, delta = self._get_delta_for_revision(parent_tree,
3956
parents_parents, possible_trees)
3837
parents_parents, basis_id, cache)
3957
3838
self.target.add_inventory_by_delta(
3958
3839
basis_id, delta, current_revision_id, parents_parents)
3959
3840
# insert signatures and revisions
3982
3863
cache[basis_id] = basis_tree
3983
3864
del basis_tree # We don't want to hang on to it here
3985
if self._converting_to_rich_root and len(revision_ids) > 100:
3986
a_graph = _mod_fetch._get_rich_root_heads_graph(self.source,
3991
3866
for offset in range(0, len(revision_ids), batch_size):
3992
3867
self.target.start_write_group()
3994
3869
pb.update('Transferring revisions', offset,
3995
3870
len(revision_ids))
3996
3871
batch = revision_ids[offset:offset+batch_size]
3997
basis_id = self._fetch_batch(batch, basis_id, cache,
3872
basis_id = self._fetch_batch(batch, basis_id, cache)
4000
self.source._safe_to_return_from_cache = False
4001
3874
self.target.abort_write_group()
4015
3888
"""See InterRepository.fetch()."""
4016
3889
if fetch_spec is not None:
4017
3890
raise AssertionError("Not implemented yet...")
4018
ui.ui_factory.warn_experimental_format_fetch(self)
4019
3891
if (not self.source.supports_rich_root()
4020
3892
and self.target.supports_rich_root()):
4021
3893
self._converting_to_rich_root = True
4022
3894
self._revision_id_to_root_id = {}
4024
3896
self._converting_to_rich_root = False
4025
# See <https://launchpad.net/bugs/456077> asking for a warning here
4026
if self.source._format.network_name() != self.target._format.network_name():
4027
ui.ui_factory.show_user_warning('cross_format_fetch',
4028
from_format=self.source._format,
4029
to_format=self.target._format)
4030
3897
revision_ids = self.target.search_missing_revision_ids(self.source,
4031
3898
revision_id, find_ghosts=find_ghosts).get_keys()
4032
3899
if not revision_ids:
4101
3968
:param to_convert: The disk object to convert.
4102
3969
:param pb: a progress bar to use for progress information.
4104
pb = ui.ui_factory.nested_progress_bar()
4107
3974
# this is only useful with metadir layouts - separated repo content.
4108
3975
# trigger an assertion if not such
4109
3976
repo._format.get_format_string()
4110
3977
self.repo_dir = repo.bzrdir
4111
pb.update('Moving repository to repository.backup')
3978
self.step('Moving repository to repository.backup')
4112
3979
self.repo_dir.transport.move('repository', 'repository.backup')
4113
3980
backup_transport = self.repo_dir.transport.clone('repository.backup')
4114
3981
repo._format.check_conversion_target(self.target_format)
4115
3982
self.source_repo = repo._format.open(self.repo_dir,
4117
3984
_override_transport=backup_transport)
4118
pb.update('Creating new repository')
3985
self.step('Creating new repository')
4119
3986
converted = self.target_format.initialize(self.repo_dir,
4120
3987
self.source_repo.is_shared())
4121
3988
converted.lock_write()
4123
pb.update('Copying content')
3990
self.step('Copying content into repository.')
4124
3991
self.source_repo.copy_content_into(converted)
4126
3993
converted.unlock()
4127
pb.update('Deleting old repository content')
3994
self.step('Deleting old repository content.')
4128
3995
self.repo_dir.transport.delete_tree('repository.backup')
4129
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)
4133
4004
_unescape_map = {
4626
4491
def _get_convertable_inventory_stream(self, revision_ids,
4627
4492
delta_versus_null=False):
4628
# The two formats are sufficiently different that there is no fast
4629
# path, so we need to send just inventorydeltas, which any
4630
# sufficiently modern client can insert into any repository.
4631
# 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
4632
4495
# convert on the target, so we need to put bytes-on-the-wire that can
4633
4496
# be converted. That means inventory deltas (if the remote is <1.19,
4634
4497
# RemoteStreamSink will fallback to VFS to insert the deltas).