870
846
# 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
849
######################################################################
894
class Repository(_RelockDebugMixin, bzrdir.ControlComponent):
853
class Repository(object):
895
854
"""Repository holding history for one or more branches.
897
856
The repository holds and retrieves historical information including
1333
1293
self._reconcile_does_inventory_gc = True
1334
1294
self._reconcile_fixes_text_parents = False
1335
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()
1336
1301
self._write_group = None
1337
1302
# Additional places to query for data.
1338
1303
self._fallback_repositories = []
1339
1304
# An InventoryEntry cache, used during deserialization
1340
1305
self._inventory_entry_cache = fifo_cache.FIFOCache(10*1024)
1341
# Is it safe to return inventory entries directly from the entry cache,
1342
# rather copying them?
1343
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
1307
def __repr__(self):
1354
1308
if self._fallback_repositories:
1417
1365
:raises MismatchedToken: if the specified token doesn't match the token
1418
1366
of the existing lock.
1419
1367
:seealso: start_write_group.
1420
: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
1422
1375
locked = self.is_locked()
1423
token = self.control_files.lock_write(token=token)
1376
result = self.control_files.lock_write(token=token)
1425
self._warn_if_deprecated()
1426
self._note_lock('w')
1427
1378
for repo in self._fallback_repositories:
1428
1379
# Writes don't affect fallback repos
1429
1380
repo.lock_read()
1430
1381
self._refresh_data()
1431
return RepositoryWriteLockResult(self.unlock, token)
1433
1384
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
1385
locked = self.is_locked()
1440
1386
self.control_files.lock_read()
1442
self._warn_if_deprecated()
1443
self._note_lock('r')
1444
1388
for repo in self._fallback_repositories:
1445
1389
repo.lock_read()
1446
1390
self._refresh_data()
1447
return LogicalLockResult(self.unlock)
1449
1392
def get_physical_lock_status(self):
1450
1393
return self.control_files.get_physical_lock_status()
1541
1487
except errors.NoRepositoryPresent:
1544
return False, ([], repository)
1490
return False, (None, repository)
1545
1491
self.first_call = False
1546
value = (bzrdir.list_branches(), None)
1493
value = (bzrdir.open_branch(), None)
1494
except errors.NotBranchError:
1495
value = (None, None)
1547
1496
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)
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)
1554
1503
if not using and repository is not None:
1555
ret.extend(repository.find_branches())
1504
branches.extend(repository.find_branches())
1558
1507
@needs_read_lock
1559
1508
def search_missing_revision_ids(self, other, revision_id=None, find_ghosts=True):
1668
1617
return missing_keys
1670
1619
def refresh_data(self):
1671
"""Re-read any data needed to synchronise with disk.
1620
"""Re-read any data needed to to synchronise with disk.
1673
1622
This method is intended to be called after another repository instance
1674
1623
(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.
1624
repository. It may not be called during a write group, but may be
1625
called at any other time.
1627
if self.is_in_write_group():
1628
raise errors.InternalBzrError(
1629
"May not refresh_data while in a write group.")
1681
1630
self._refresh_data()
1683
1632
def resume_write_group(self, tokens):
1937
1885
rev = self._serializer.read_revision_from_string(text)
1938
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()
1940
1905
def get_deltas_for_revisions(self, revisions, specific_fileids=None):
1941
1906
"""Produce a generator of revision deltas.
2441
2410
next_key = None
2444
def _deserialise_inventory(self, revision_id, xml):
2413
def deserialise_inventory(self, revision_id, xml):
2445
2414
"""Transform the xml into an inventory object.
2447
2416
:param revision_id: The expected revision id of the inventory.
2448
2417
:param xml: A serialised inventory.
2450
2419
result = self._serializer.read_inventory_from_string(xml, revision_id,
2451
entry_cache=self._inventory_entry_cache,
2452
return_from_cache=self._safe_to_return_from_cache)
2420
entry_cache=self._inventory_entry_cache)
2453
2421
if result.revision_id != revision_id:
2454
2422
raise AssertionError('revision id mismatch %s != %s' % (
2455
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)
2458
2432
def get_serializer_format(self):
2459
2433
return self._serializer.format_num
2461
2435
@needs_read_lock
2462
def _get_inventory_xml(self, revision_id):
2463
"""Get serialized inventory as a string."""
2436
def get_inventory_xml(self, revision_id):
2437
"""Get inventory XML as a file object."""
2464
2438
texts = self._iter_inventory_xmls([revision_id], 'unordered')
2466
2440
text, revision_id = texts.next()
2525
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)
2527
2523
def is_shared(self):
2528
2524
"""Return True if this repository is flagged as a shared repository."""
2529
2525
raise NotImplementedError(self.is_shared)
2780
2764
result.check(callback_refs)
2783
def _warn_if_deprecated(self, branch=None):
2767
def _warn_if_deprecated(self):
2784
2768
global _deprecation_warning_done
2785
2769
if _deprecation_warning_done:
2789
conf = config.GlobalConfig()
2791
conf = branch.get_config()
2792
if conf.suppress_warning('format_deprecation'):
2794
warning("Format %s for %s is deprecated -"
2795
" please use 'bzr upgrade' to get better performance"
2796
% (self._format, self.bzrdir.transport.base))
2798
_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))
2800
2775
def supports_rich_root(self):
2801
2776
return self._format.rich_root_data
3432
3396
:param revision_id: if None all content is copied, if NULL_REVISION no
3433
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.
3437
ui.ui_factory.warn_experimental_format_fetch(self)
3438
3402
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
3403
f = RepoFetcher(to_repository=self.target,
3445
3404
from_repository=self.source,
3446
3405
last_revision=revision_id,
3447
3406
fetch_spec=fetch_spec,
3448
find_ghosts=find_ghosts)
3407
pb=pb, find_ghosts=find_ghosts)
3450
3409
def _walk_to_common_revisions(self, revision_ids):
3451
3410
"""Walk out from revision_ids in source to revisions target has.
3776
def _get_trees(self, revision_ids, cache):
3778
for rev_id in revision_ids:
3780
possible_trees.append((rev_id, cache[rev_id]))
3782
# Not cached, but inventory might be present anyway.
3784
tree = self.source.revision_tree(rev_id)
3785
except errors.NoSuchRevision:
3786
# Nope, parent is ghost.
3789
cache[rev_id] = tree
3790
possible_trees.append((rev_id, tree))
3791
return possible_trees
3793
def _get_delta_for_revision(self, tree, parent_ids, possible_trees):
3735
def _get_delta_for_revision(self, tree, parent_ids, basis_id, cache):
3794
3736
"""Get the best delta and base for this revision.
3796
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]))
3799
# Generate deltas against each tree, to find the shortest.
3800
texts_possibly_new_in_tree = set()
3801
3748
for basis_id, basis_tree in possible_trees:
3802
3749
delta = tree.inventory._make_delta(basis_tree.inventory)
3803
for old_path, new_path, file_id, new_entry in delta:
3804
if new_path is None:
3805
# This file_id isn't present in the new rev, so we don't
3809
# Rich roots are handled elsewhere...
3811
kind = new_entry.kind
3812
if kind != 'directory' and kind != 'file':
3813
# No text record associated with this inventory entry.
3815
# This is a directory or file that has changed somehow.
3816
texts_possibly_new_in_tree.add((file_id, new_entry.revision))
3817
3750
deltas.append((len(delta), basis_id, delta))
3819
3752
return deltas[0][1:]
3821
def _fetch_parent_invs_for_stacking(self, parent_map, cache):
3822
"""Find all parent revisions that are absent, but for which the
3823
inventory is present, and copy those inventories.
3825
This is necessary to preserve correctness when the source is stacked
3826
without fallbacks configured. (Note that in cases like upgrade the
3827
source may be not have _fallback_repositories even though it is
3831
for parents in parent_map.values():
3832
parent_revs.update(parents)
3833
present_parents = self.source.get_parent_map(parent_revs)
3834
absent_parents = set(parent_revs).difference(present_parents)
3835
parent_invs_keys_for_stacking = self.source.inventories.get_parent_map(
3836
(rev_id,) for rev_id in absent_parents)
3837
parent_inv_ids = [key[-1] for key in parent_invs_keys_for_stacking]
3838
for parent_tree in self.source.revision_trees(parent_inv_ids):
3839
current_revision_id = parent_tree.get_revision_id()
3840
parents_parents_keys = parent_invs_keys_for_stacking[
3841
(current_revision_id,)]
3842
parents_parents = [key[-1] for key in parents_parents_keys]
3843
basis_id = _mod_revision.NULL_REVISION
3844
basis_tree = self.source.revision_tree(basis_id)
3845
delta = parent_tree.inventory._make_delta(basis_tree.inventory)
3846
self.target.add_inventory_by_delta(
3847
basis_id, delta, current_revision_id, parents_parents)
3848
cache[current_revision_id] = parent_tree
3850
def _fetch_batch(self, revision_ids, basis_id, cache, a_graph=None):
3754
def _fetch_batch(self, revision_ids, basis_id, cache):
3851
3755
"""Fetch across a few revisions.
3853
3757
:param revision_ids: The revisions to copy
3854
3758
:param basis_id: The revision_id of a tree that must be in cache, used
3855
3759
as a basis for delta when no other base is available
3856
3760
:param cache: A cache of RevisionTrees that we can use.
3857
:param a_graph: A Graph object to determine the heads() of the
3858
rich-root data stream.
3859
3761
:return: The revision_id of the last converted tree. The RevisionTree
3860
3762
for it will be in cache
3867
3769
pending_deltas = []
3868
3770
pending_revisions = []
3869
3771
parent_map = self.source.get_parent_map(revision_ids)
3870
self._fetch_parent_invs_for_stacking(parent_map, cache)
3871
self.source._safe_to_return_from_cache = True
3872
3772
for tree in self.source.revision_trees(revision_ids):
3873
# Find a inventory delta for this revision.
3874
# Find text entries that need to be copied, too.
3875
3773
current_revision_id = tree.get_revision_id()
3876
3774
parent_ids = parent_map.get(current_revision_id, ())
3877
parent_trees = self._get_trees(parent_ids, cache)
3878
possible_trees = list(parent_trees)
3879
if len(possible_trees) == 0:
3880
# There either aren't any parents, or the parents are ghosts,
3881
# so just use the last converted tree.
3882
possible_trees.append((basis_id, cache[basis_id]))
3883
3775
basis_id, delta = self._get_delta_for_revision(tree, parent_ids,
3885
revision = self.source.get_revision(current_revision_id)
3886
pending_deltas.append((basis_id, delta,
3887
current_revision_id, revision.parent_ids))
3888
3777
if self._converting_to_rich_root:
3889
3778
self._revision_id_to_root_id[current_revision_id] = \
3890
3779
tree.get_root_id()
3891
# Determine which texts are in present in this revision but not in
3892
# any of the available parents.
3893
texts_possibly_new_in_tree = set()
3780
# Find text entries that need to be copied
3894
3781
for old_path, new_path, file_id, entry in delta:
3895
if new_path is None:
3896
# This file_id isn't present in the new rev
3900
if not self.target.supports_rich_root():
3901
# The target doesn't support rich root, so we don't
3904
if self._converting_to_rich_root:
3905
# This can't be copied normally, we have to insert
3907
root_keys_to_create.add((file_id, entry.revision))
3910
texts_possibly_new_in_tree.add((file_id, entry.revision))
3911
for basis_id, basis_tree in possible_trees:
3912
basis_inv = basis_tree.inventory
3913
for file_key in list(texts_possibly_new_in_tree):
3914
file_id, file_revision = file_key
3916
entry = basis_inv[file_id]
3917
except errors.NoSuchId:
3919
if entry.revision == file_revision:
3920
texts_possibly_new_in_tree.remove(file_key)
3921
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))
3922
3798
pending_revisions.append(revision)
3923
3799
cache[current_revision_id] = tree
3924
3800
basis_id = current_revision_id
3925
self.source._safe_to_return_from_cache = False
3926
3801
# Copy file texts
3927
3802
from_texts = self.source.texts
3928
3803
to_texts = self.target.texts
3929
3804
if root_keys_to_create:
3930
root_stream = _mod_fetch._new_root_data_stream(
3805
from bzrlib.fetch import _new_root_data_stream
3806
root_stream = _new_root_data_stream(
3931
3807
root_keys_to_create, self._revision_id_to_root_id, parent_map,
3932
self.source, graph=a_graph)
3933
3809
to_texts.insert_record_stream(root_stream)
3934
3810
to_texts.insert_record_stream(from_texts.get_record_stream(
3935
3811
text_keys, self.target._format._fetch_order,
3957
3833
for parent_tree in self.source.revision_trees(parent_map):
3958
3834
current_revision_id = parent_tree.get_revision_id()
3959
3835
parents_parents = parent_map[current_revision_id]
3960
possible_trees = self._get_trees(parents_parents, cache)
3961
if len(possible_trees) == 0:
3962
# There either aren't any parents, or the parents are
3963
# ghosts, so just use the last converted tree.
3964
possible_trees.append((basis_id, cache[basis_id]))
3965
3836
basis_id, delta = self._get_delta_for_revision(parent_tree,
3966
parents_parents, possible_trees)
3837
parents_parents, basis_id, cache)
3967
3838
self.target.add_inventory_by_delta(
3968
3839
basis_id, delta, current_revision_id, parents_parents)
3969
3840
# insert signatures and revisions
3992
3863
cache[basis_id] = basis_tree
3993
3864
del basis_tree # We don't want to hang on to it here
3995
if self._converting_to_rich_root and len(revision_ids) > 100:
3996
a_graph = _mod_fetch._get_rich_root_heads_graph(self.source,
4001
3866
for offset in range(0, len(revision_ids), batch_size):
4002
3867
self.target.start_write_group()
4004
3869
pb.update('Transferring revisions', offset,
4005
3870
len(revision_ids))
4006
3871
batch = revision_ids[offset:offset+batch_size]
4007
basis_id = self._fetch_batch(batch, basis_id, cache,
3872
basis_id = self._fetch_batch(batch, basis_id, cache)
4010
self.source._safe_to_return_from_cache = False
4011
3874
self.target.abort_write_group()
4025
3888
"""See InterRepository.fetch()."""
4026
3889
if fetch_spec is not None:
4027
3890
raise AssertionError("Not implemented yet...")
4028
ui.ui_factory.warn_experimental_format_fetch(self)
4029
3891
if (not self.source.supports_rich_root()
4030
3892
and self.target.supports_rich_root()):
4031
3893
self._converting_to_rich_root = True
4032
3894
self._revision_id_to_root_id = {}
4034
3896
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
3897
revision_ids = self.target.search_missing_revision_ids(self.source,
4041
3898
revision_id, find_ghosts=find_ghosts).get_keys()
4042
3899
if not revision_ids:
4111
3968
:param to_convert: The disk object to convert.
4112
3969
:param pb: a progress bar to use for progress information.
4114
pb = ui.ui_factory.nested_progress_bar()
4117
3974
# this is only useful with metadir layouts - separated repo content.
4118
3975
# trigger an assertion if not such
4119
3976
repo._format.get_format_string()
4120
3977
self.repo_dir = repo.bzrdir
4121
pb.update('Moving repository to repository.backup')
3978
self.step('Moving repository to repository.backup')
4122
3979
self.repo_dir.transport.move('repository', 'repository.backup')
4123
3980
backup_transport = self.repo_dir.transport.clone('repository.backup')
4124
3981
repo._format.check_conversion_target(self.target_format)
4125
3982
self.source_repo = repo._format.open(self.repo_dir,
4127
3984
_override_transport=backup_transport)
4128
pb.update('Creating new repository')
3985
self.step('Creating new repository')
4129
3986
converted = self.target_format.initialize(self.repo_dir,
4130
3987
self.source_repo.is_shared())
4131
3988
converted.lock_write()
4133
pb.update('Copying content')
3990
self.step('Copying content into repository.')
4134
3991
self.source_repo.copy_content_into(converted)
4136
3993
converted.unlock()
4137
pb.update('Deleting old repository content')
3994
self.step('Deleting old repository content.')
4138
3995
self.repo_dir.transport.delete_tree('repository.backup')
4139
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)
4143
4004
_unescape_map = {
4636
4491
def _get_convertable_inventory_stream(self, revision_ids,
4637
4492
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
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
4642
4495
# convert on the target, so we need to put bytes-on-the-wire that can
4643
4496
# be converted. That means inventory deltas (if the remote is <1.19,
4644
4497
# RemoteStreamSink will fallback to VFS to insert the deltas).