1306
1293
self._reconcile_does_inventory_gc = True
1307
1294
self._reconcile_fixes_text_parents = False
1308
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()
1309
1301
self._write_group = None
1310
1302
# Additional places to query for data.
1311
1303
self._fallback_repositories = []
1312
1304
# An InventoryEntry cache, used during deserialization
1313
1305
self._inventory_entry_cache = fifo_cache.FIFOCache(10*1024)
1314
# Is it safe to return inventory entries directly from the entry cache,
1315
# rather copying them?
1316
self._safe_to_return_from_cache = False
1319
def user_transport(self):
1320
return self.bzrdir.user_transport
1323
def control_transport(self):
1324
return self._transport
1326
1307
def __repr__(self):
1327
1308
if self._fallback_repositories:
1507
1487
except errors.NoRepositoryPresent:
1510
return False, ([], repository)
1490
return False, (None, repository)
1511
1491
self.first_call = False
1512
value = (bzrdir.list_branches(), None)
1493
value = (bzrdir.open_branch(), None)
1494
except errors.NotBranchError:
1495
value = (None, None)
1513
1496
return True, value
1516
for branches, repository in bzrdir.BzrDir.find_bzrdirs(
1517
self.user_transport, evaluate=Evaluator()):
1518
if branches is not None:
1519
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)
1520
1503
if not using and repository is not None:
1521
ret.extend(repository.find_branches())
1504
branches.extend(repository.find_branches())
1524
1507
@needs_read_lock
1525
1508
def search_missing_revision_ids(self, other, revision_id=None, find_ghosts=True):
1903
1885
rev = self._serializer.read_revision_from_string(text)
1904
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()
1906
1905
def get_deltas_for_revisions(self, revisions, specific_fileids=None):
1907
1906
"""Produce a generator of revision deltas.
2407
2410
next_key = None
2410
def _deserialise_inventory(self, revision_id, xml):
2413
def deserialise_inventory(self, revision_id, xml):
2411
2414
"""Transform the xml into an inventory object.
2413
2416
:param revision_id: The expected revision id of the inventory.
2414
2417
:param xml: A serialised inventory.
2416
2419
result = self._serializer.read_inventory_from_string(xml, revision_id,
2417
entry_cache=self._inventory_entry_cache,
2418
return_from_cache=self._safe_to_return_from_cache)
2420
entry_cache=self._inventory_entry_cache)
2419
2421
if result.revision_id != revision_id:
2420
2422
raise AssertionError('revision id mismatch %s != %s' % (
2421
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)
2424
2432
def get_serializer_format(self):
2425
2433
return self._serializer.format_num
2427
2435
@needs_read_lock
2428
def _get_inventory_xml(self, revision_id):
2429
"""Get serialized inventory as a string."""
2436
def get_inventory_xml(self, revision_id):
2437
"""Get inventory XML as a file object."""
2430
2438
texts = self._iter_inventory_xmls([revision_id], 'unordered')
2432
2440
text, revision_id = texts.next()
2491
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)
2493
2523
def is_shared(self):
2494
2524
"""Return True if this repository is flagged as a shared repository."""
2495
2525
raise NotImplementedError(self.is_shared)
2746
2764
result.check(callback_refs)
2749
def _warn_if_deprecated(self, branch=None):
2767
def _warn_if_deprecated(self):
2750
2768
global _deprecation_warning_done
2751
2769
if _deprecation_warning_done:
2755
conf = config.GlobalConfig()
2757
conf = branch.get_config()
2758
if conf.suppress_warning('format_deprecation'):
2760
warning("Format %s for %s is deprecated -"
2761
" please use 'bzr upgrade' to get better performance"
2762
% (self._format, self.bzrdir.transport.base))
2764
_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))
2766
2775
def supports_rich_root(self):
2767
2776
return self._format.rich_root_data
3161
3166
raise NotImplementedError(self.network_name)
3163
3168
def check_conversion_target(self, target_format):
3164
if self.rich_root_data and not target_format.rich_root_data:
3165
raise errors.BadConversionTarget(
3166
'Does not support rich root data.', target_format,
3168
if (self.supports_tree_reference and
3169
not getattr(target_format, 'supports_tree_reference', False)):
3170
raise errors.BadConversionTarget(
3171
'Does not support nested trees', target_format,
3169
raise NotImplementedError(self.check_conversion_target)
3174
3171
def open(self, a_bzrdir, _found=False):
3175
3172
"""Return an instance of this format for the bzrdir a_bzrdir.
3398
3386
:param revision_id: if None all content is copied, if NULL_REVISION no
3399
3387
content is copied.
3388
:param pb: optional progress bar to use for progress reports. If not
3389
provided a default one will be created.
3403
ui.ui_factory.warn_experimental_format_fetch(self)
3404
3392
from bzrlib.fetch import RepoFetcher
3405
# See <https://launchpad.net/bugs/456077> asking for a warning here
3406
if self.source._format.network_name() != self.target._format.network_name():
3407
ui.ui_factory.show_user_warning('cross_format_fetch',
3408
from_format=self.source._format,
3409
to_format=self.target._format)
3410
3393
f = RepoFetcher(to_repository=self.target,
3411
3394
from_repository=self.source,
3412
3395
last_revision=revision_id,
3413
3396
fetch_spec=fetch_spec,
3414
find_ghosts=find_ghosts)
3397
pb=pb, find_ghosts=find_ghosts)
3416
3399
def _walk_to_common_revisions(self, revision_ids):
3417
3400
"""Walk out from revision_ids in source to revisions target has.
3742
def _get_trees(self, revision_ids, cache):
3744
for rev_id in revision_ids:
3746
possible_trees.append((rev_id, cache[rev_id]))
3748
# Not cached, but inventory might be present anyway.
3750
tree = self.source.revision_tree(rev_id)
3751
except errors.NoSuchRevision:
3752
# Nope, parent is ghost.
3755
cache[rev_id] = tree
3756
possible_trees.append((rev_id, tree))
3757
return possible_trees
3759
def _get_delta_for_revision(self, tree, parent_ids, possible_trees):
3725
def _get_delta_for_revision(self, tree, parent_ids, basis_id, cache):
3760
3726
"""Get the best delta and base for this revision.
3762
3728
:return: (basis_id, delta)
3730
possible_trees = [(parent_id, cache[parent_id])
3731
for parent_id in parent_ids
3732
if parent_id in cache]
3733
if len(possible_trees) == 0:
3734
# There either aren't any parents, or the parents aren't in the
3735
# cache, so just use the last converted tree
3736
possible_trees.append((basis_id, cache[basis_id]))
3765
# Generate deltas against each tree, to find the shortest.
3766
texts_possibly_new_in_tree = set()
3767
3738
for basis_id, basis_tree in possible_trees:
3768
3739
delta = tree.inventory._make_delta(basis_tree.inventory)
3769
for old_path, new_path, file_id, new_entry in delta:
3770
if new_path is None:
3771
# This file_id isn't present in the new rev, so we don't
3775
# Rich roots are handled elsewhere...
3777
kind = new_entry.kind
3778
if kind != 'directory' and kind != 'file':
3779
# No text record associated with this inventory entry.
3781
# This is a directory or file that has changed somehow.
3782
texts_possibly_new_in_tree.add((file_id, new_entry.revision))
3783
3740
deltas.append((len(delta), basis_id, delta))
3785
3742
return deltas[0][1:]
3787
def _fetch_parent_invs_for_stacking(self, parent_map, cache):
3788
"""Find all parent revisions that are absent, but for which the
3789
inventory is present, and copy those inventories.
3791
This is necessary to preserve correctness when the source is stacked
3792
without fallbacks configured. (Note that in cases like upgrade the
3793
source may be not have _fallback_repositories even though it is
3797
for parents in parent_map.values():
3798
parent_revs.update(parents)
3799
present_parents = self.source.get_parent_map(parent_revs)
3800
absent_parents = set(parent_revs).difference(present_parents)
3801
parent_invs_keys_for_stacking = self.source.inventories.get_parent_map(
3802
(rev_id,) for rev_id in absent_parents)
3803
parent_inv_ids = [key[-1] for key in parent_invs_keys_for_stacking]
3804
for parent_tree in self.source.revision_trees(parent_inv_ids):
3805
current_revision_id = parent_tree.get_revision_id()
3806
parents_parents_keys = parent_invs_keys_for_stacking[
3807
(current_revision_id,)]
3808
parents_parents = [key[-1] for key in parents_parents_keys]
3809
basis_id = _mod_revision.NULL_REVISION
3810
basis_tree = self.source.revision_tree(basis_id)
3811
delta = parent_tree.inventory._make_delta(basis_tree.inventory)
3812
self.target.add_inventory_by_delta(
3813
basis_id, delta, current_revision_id, parents_parents)
3814
cache[current_revision_id] = parent_tree
3816
def _fetch_batch(self, revision_ids, basis_id, cache, a_graph=None):
3744
def _fetch_batch(self, revision_ids, basis_id, cache):
3817
3745
"""Fetch across a few revisions.
3819
3747
:param revision_ids: The revisions to copy
3820
3748
:param basis_id: The revision_id of a tree that must be in cache, used
3821
3749
as a basis for delta when no other base is available
3822
3750
:param cache: A cache of RevisionTrees that we can use.
3823
:param a_graph: A Graph object to determine the heads() of the
3824
rich-root data stream.
3825
3751
:return: The revision_id of the last converted tree. The RevisionTree
3826
3752
for it will be in cache
3833
3759
pending_deltas = []
3834
3760
pending_revisions = []
3835
3761
parent_map = self.source.get_parent_map(revision_ids)
3836
self._fetch_parent_invs_for_stacking(parent_map, cache)
3837
self.source._safe_to_return_from_cache = True
3838
3762
for tree in self.source.revision_trees(revision_ids):
3839
# Find a inventory delta for this revision.
3840
# Find text entries that need to be copied, too.
3841
3763
current_revision_id = tree.get_revision_id()
3842
3764
parent_ids = parent_map.get(current_revision_id, ())
3843
parent_trees = self._get_trees(parent_ids, cache)
3844
possible_trees = list(parent_trees)
3845
if len(possible_trees) == 0:
3846
# There either aren't any parents, or the parents are ghosts,
3847
# so just use the last converted tree.
3848
possible_trees.append((basis_id, cache[basis_id]))
3849
3765
basis_id, delta = self._get_delta_for_revision(tree, parent_ids,
3851
revision = self.source.get_revision(current_revision_id)
3852
pending_deltas.append((basis_id, delta,
3853
current_revision_id, revision.parent_ids))
3854
3767
if self._converting_to_rich_root:
3855
3768
self._revision_id_to_root_id[current_revision_id] = \
3856
3769
tree.get_root_id()
3857
# Determine which texts are in present in this revision but not in
3858
# any of the available parents.
3859
texts_possibly_new_in_tree = set()
3770
# Find text entries that need to be copied
3860
3771
for old_path, new_path, file_id, entry in delta:
3861
if new_path is None:
3862
# This file_id isn't present in the new rev
3866
if not self.target.supports_rich_root():
3867
# The target doesn't support rich root, so we don't
3870
if self._converting_to_rich_root:
3871
# This can't be copied normally, we have to insert
3873
root_keys_to_create.add((file_id, entry.revision))
3876
texts_possibly_new_in_tree.add((file_id, entry.revision))
3877
for basis_id, basis_tree in possible_trees:
3878
basis_inv = basis_tree.inventory
3879
for file_key in list(texts_possibly_new_in_tree):
3880
file_id, file_revision = file_key
3882
entry = basis_inv[file_id]
3883
except errors.NoSuchId:
3885
if entry.revision == file_revision:
3886
texts_possibly_new_in_tree.remove(file_key)
3887
text_keys.update(texts_possibly_new_in_tree)
3772
if new_path is not None:
3775
if not self.target.supports_rich_root():
3776
# The target doesn't support rich root, so we don't
3779
if self._converting_to_rich_root:
3780
# This can't be copied normally, we have to insert
3782
root_keys_to_create.add((file_id, entry.revision))
3784
text_keys.add((file_id, entry.revision))
3785
revision = self.source.get_revision(current_revision_id)
3786
pending_deltas.append((basis_id, delta,
3787
current_revision_id, revision.parent_ids))
3888
3788
pending_revisions.append(revision)
3889
3789
cache[current_revision_id] = tree
3890
3790
basis_id = current_revision_id
3891
self.source._safe_to_return_from_cache = False
3892
3791
# Copy file texts
3893
3792
from_texts = self.source.texts
3894
3793
to_texts = self.target.texts
3895
3794
if root_keys_to_create:
3896
root_stream = _mod_fetch._new_root_data_stream(
3795
from bzrlib.fetch import _new_root_data_stream
3796
root_stream = _new_root_data_stream(
3897
3797
root_keys_to_create, self._revision_id_to_root_id, parent_map,
3898
self.source, graph=a_graph)
3899
3799
to_texts.insert_record_stream(root_stream)
3900
3800
to_texts.insert_record_stream(from_texts.get_record_stream(
3901
3801
text_keys, self.target._format._fetch_order,
3923
3823
for parent_tree in self.source.revision_trees(parent_map):
3924
3824
current_revision_id = parent_tree.get_revision_id()
3925
3825
parents_parents = parent_map[current_revision_id]
3926
possible_trees = self._get_trees(parents_parents, cache)
3927
if len(possible_trees) == 0:
3928
# There either aren't any parents, or the parents are
3929
# ghosts, so just use the last converted tree.
3930
possible_trees.append((basis_id, cache[basis_id]))
3931
3826
basis_id, delta = self._get_delta_for_revision(parent_tree,
3932
parents_parents, possible_trees)
3827
parents_parents, basis_id, cache)
3933
3828
self.target.add_inventory_by_delta(
3934
3829
basis_id, delta, current_revision_id, parents_parents)
3935
3830
# insert signatures and revisions
3958
3853
cache[basis_id] = basis_tree
3959
3854
del basis_tree # We don't want to hang on to it here
3961
if self._converting_to_rich_root and len(revision_ids) > 100:
3962
a_graph = _mod_fetch._get_rich_root_heads_graph(self.source,
3967
3856
for offset in range(0, len(revision_ids), batch_size):
3968
3857
self.target.start_write_group()
3970
3859
pb.update('Transferring revisions', offset,
3971
3860
len(revision_ids))
3972
3861
batch = revision_ids[offset:offset+batch_size]
3973
basis_id = self._fetch_batch(batch, basis_id, cache,
3862
basis_id = self._fetch_batch(batch, basis_id, cache)
3976
self.source._safe_to_return_from_cache = False
3977
3864
self.target.abort_write_group()
3991
3878
"""See InterRepository.fetch()."""
3992
3879
if fetch_spec is not None:
3993
3880
raise AssertionError("Not implemented yet...")
3994
ui.ui_factory.warn_experimental_format_fetch(self)
3995
3881
if (not self.source.supports_rich_root()
3996
3882
and self.target.supports_rich_root()):
3997
3883
self._converting_to_rich_root = True
3998
3884
self._revision_id_to_root_id = {}
4000
3886
self._converting_to_rich_root = False
4001
# See <https://launchpad.net/bugs/456077> asking for a warning here
4002
if self.source._format.network_name() != self.target._format.network_name():
4003
ui.ui_factory.show_user_warning('cross_format_fetch',
4004
from_format=self.source._format,
4005
to_format=self.target._format)
4006
3887
revision_ids = self.target.search_missing_revision_ids(self.source,
4007
3888
revision_id, find_ghosts=find_ghosts).get_keys()
4008
3889
if not revision_ids:
4077
3958
:param to_convert: The disk object to convert.
4078
3959
:param pb: a progress bar to use for progress information.
4080
pb = ui.ui_factory.nested_progress_bar()
4083
3964
# this is only useful with metadir layouts - separated repo content.
4084
3965
# trigger an assertion if not such
4085
3966
repo._format.get_format_string()
4086
3967
self.repo_dir = repo.bzrdir
4087
pb.update('Moving repository to repository.backup')
3968
self.step('Moving repository to repository.backup')
4088
3969
self.repo_dir.transport.move('repository', 'repository.backup')
4089
3970
backup_transport = self.repo_dir.transport.clone('repository.backup')
4090
3971
repo._format.check_conversion_target(self.target_format)
4091
3972
self.source_repo = repo._format.open(self.repo_dir,
4093
3974
_override_transport=backup_transport)
4094
pb.update('Creating new repository')
3975
self.step('Creating new repository')
4095
3976
converted = self.target_format.initialize(self.repo_dir,
4096
3977
self.source_repo.is_shared())
4097
3978
converted.lock_write()
4099
pb.update('Copying content')
3980
self.step('Copying content into repository.')
4100
3981
self.source_repo.copy_content_into(converted)
4102
3983
converted.unlock()
4103
pb.update('Deleting old repository content')
3984
self.step('Deleting old repository content.')
4104
3985
self.repo_dir.transport.delete_tree('repository.backup')
4105
ui.ui_factory.note('repository converted')
3986
self.pb.note('repository converted')
3988
def step(self, message):
3989
"""Update the pb by a step."""
3991
self.pb.update(message, self.count, self.total)
4109
3994
_unescape_map = {
4602
4481
def _get_convertable_inventory_stream(self, revision_ids,
4603
4482
delta_versus_null=False):
4604
# The two formats are sufficiently different that there is no fast
4605
# path, so we need to send just inventorydeltas, which any
4606
# sufficiently modern client can insert into any repository.
4607
# The StreamSink code expects to be able to
4483
# The source is using CHKs, but the target either doesn't or it has a
4484
# different serializer. The StreamSink code expects to be able to
4608
4485
# convert on the target, so we need to put bytes-on-the-wire that can
4609
4486
# be converted. That means inventory deltas (if the remote is <1.19,
4610
4487
# RemoteStreamSink will fallback to VFS to insert the deltas).