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
51
50
from bzrlib.testament import Testament
59
53
from bzrlib.decorators import needs_read_lock, needs_write_lock, only_raises
60
54
from bzrlib.inter import InterObject
61
55
from bzrlib.inventory import (
67
from bzrlib.recordcounter import RecordCounter
68
from bzrlib.lock import _RelockDebugMixin, LogicalLockResult
61
from bzrlib.lock import _RelockDebugMixin
62
from bzrlib import registry
69
63
from bzrlib.trace import (
70
64
log_exception_quietly, note, mutter, mutter_callsite, warning)
74
68
_deprecation_warning_done = False
77
class IsInWriteGroupError(errors.InternalBzrError):
79
_fmt = "May not refresh_data of repo %(repo)s while in a write group."
81
def __init__(self, repo):
82
errors.InternalBzrError.__init__(self, repo=repo)
85
71
class CommitBuilder(object):
86
72
"""Provides an interface to build up a commit.
290
276
:param tree: The tree which is being committed.
292
if len(self.parents) == 0:
293
raise errors.RootMissing()
278
# NB: if there are no parents then this method is not called, so no
279
# need to guard on parents having length.
294
280
entry = entry_factory['directory'](tree.path2id(''), '',
296
282
entry.revision = self._new_revision_id
871
857
# versioned roots do not change unless the tree found a change.
874
class RepositoryWriteLockResult(LogicalLockResult):
875
"""The result of write locking a repository.
877
:ivar repository_token: The token obtained from the underlying lock, or
879
:ivar unlock: A callable which will unlock the lock.
882
def __init__(self, unlock, repository_token):
883
LogicalLockResult.__init__(self, unlock)
884
self.repository_token = repository_token
887
return "RepositoryWriteLockResult(%s, %s)" % (self.repository_token,
891
860
######################################################################
895
class Repository(_RelockDebugMixin, bzrdir.ControlComponent):
864
class Repository(_RelockDebugMixin):
896
865
"""Repository holding history for one or more branches.
898
867
The repository holds and retrieves historical information including
1046
1015
" id and insertion revid (%r, %r)"
1047
1016
% (inv.revision_id, revision_id))
1048
1017
if inv.root is None:
1049
raise errors.RootMissing()
1018
raise AssertionError()
1050
1019
return self._add_inventory_checked(revision_id, inv, parents)
1052
1021
def _add_inventory_checked(self, revision_id, inv, parents):
1057
1026
:seealso: add_inventory, for the contract.
1059
inv_lines = self._serializer.write_inventory_to_lines(inv)
1028
inv_lines = self._serialise_inventory_to_lines(inv)
1060
1029
return self._inventory_add_lines(revision_id, parents,
1061
1030
inv_lines, check_content=False)
1269
1238
"""Check a single text from this repository."""
1270
1239
if kind == 'inventories':
1271
1240
rev_id = record.key[0]
1272
inv = self._deserialise_inventory(rev_id,
1241
inv = self.deserialise_inventory(rev_id,
1273
1242
record.get_bytes_as('fulltext'))
1274
1243
if last_object is not None:
1275
1244
delta = inv._make_delta(last_object)
1320
1289
:param _format: The format of the repository on disk.
1321
1290
:param a_bzrdir: The BzrDir of the repository.
1292
In the future we will have a single api for all stores for
1293
getting file texts, inventories and revisions, then
1294
this construct will accept instances of those things.
1323
# In the future we will have a single api for all stores for
1324
# getting file texts, inventories and revisions, then
1325
# this construct will accept instances of those things.
1326
1296
super(Repository, self).__init__()
1327
1297
self._format = _format
1328
1298
# the following are part of the public API for Repository:
1334
1304
self._reconcile_does_inventory_gc = True
1335
1305
self._reconcile_fixes_text_parents = False
1336
1306
self._reconcile_backsup_inventory = True
1307
# not right yet - should be more semantically clear ?
1309
# TODO: make sure to construct the right store classes, etc, depending
1310
# on whether escaping is required.
1311
self._warn_if_deprecated()
1337
1312
self._write_group = None
1338
1313
# Additional places to query for data.
1339
1314
self._fallback_repositories = []
1340
1315
# An InventoryEntry cache, used during deserialization
1341
1316
self._inventory_entry_cache = fifo_cache.FIFOCache(10*1024)
1342
# Is it safe to return inventory entries directly from the entry cache,
1343
# rather copying them?
1344
self._safe_to_return_from_cache = False
1347
def user_transport(self):
1348
return self.bzrdir.user_transport
1351
def control_transport(self):
1352
return self._transport
1354
1318
def __repr__(self):
1355
1319
if self._fallback_repositories:
1404
1368
data during reads, and allows a 'write_group' to be obtained. Write
1405
1369
groups must be used for actual data insertion.
1407
A token should be passed in if you know that you have locked the object
1408
some other way, and need to synchronise this object's state with that
1411
XXX: this docstring is duplicated in many places, e.g. lockable_files.py
1413
1371
:param token: if this is already locked, then lock_write will fail
1414
1372
unless the token matches the existing lock.
1415
1373
:returns: a token if this instance supports tokens, otherwise None.
1418
1376
:raises MismatchedToken: if the specified token doesn't match the token
1419
1377
of the existing lock.
1420
1378
:seealso: start_write_group.
1421
:return: A RepositoryWriteLockResult.
1380
A token should be passed in if you know that you have locked the object
1381
some other way, and need to synchronise this object's state with that
1384
XXX: this docstring is duplicated in many places, e.g. lockable_files.py
1423
1386
locked = self.is_locked()
1424
token = self.control_files.lock_write(token=token)
1387
result = self.control_files.lock_write(token=token)
1426
self._warn_if_deprecated()
1427
1389
self._note_lock('w')
1428
1390
for repo in self._fallback_repositories:
1429
1391
# Writes don't affect fallback repos
1430
1392
repo.lock_read()
1431
1393
self._refresh_data()
1432
return RepositoryWriteLockResult(self.unlock, token)
1434
1396
def lock_read(self):
1435
"""Lock the repository for read operations.
1437
:return: An object with an unlock method which will release the lock
1440
1397
locked = self.is_locked()
1441
1398
self.control_files.lock_read()
1443
self._warn_if_deprecated()
1444
1400
self._note_lock('r')
1445
1401
for repo in self._fallback_repositories:
1446
1402
repo.lock_read()
1447
1403
self._refresh_data()
1448
return LogicalLockResult(self.unlock)
1450
1405
def get_physical_lock_status(self):
1451
1406
return self.control_files.get_physical_lock_status()
1512
1467
# now gather global repository information
1513
1468
# XXX: This is available for many repos regardless of listability.
1514
if self.user_transport.listable():
1469
if self.bzrdir.root_transport.listable():
1515
1470
# XXX: do we want to __define len__() ?
1516
1471
# Maybe the versionedfiles object should provide a different
1517
1472
# method to get the number of keys.
1527
1482
:param using: If True, list only branches using this repository.
1529
1484
if using and not self.is_shared():
1530
return self.bzrdir.list_branches()
1486
return [self.bzrdir.open_branch()]
1487
except errors.NotBranchError:
1531
1489
class Evaluator(object):
1533
1491
def __init__(self):
1542
1500
except errors.NoRepositoryPresent:
1545
return False, ([], repository)
1503
return False, (None, repository)
1546
1504
self.first_call = False
1547
value = (bzrdir.list_branches(), None)
1506
value = (bzrdir.open_branch(), None)
1507
except errors.NotBranchError:
1508
value = (None, None)
1548
1509
return True, value
1551
for branches, repository in bzrdir.BzrDir.find_bzrdirs(
1552
self.user_transport, evaluate=Evaluator()):
1553
if branches is not None:
1554
ret.extend(branches)
1512
for branch, repository in bzrdir.BzrDir.find_bzrdirs(
1513
self.bzrdir.root_transport, evaluate=Evaluator()):
1514
if branch is not None:
1515
branches.append(branch)
1555
1516
if not using and repository is not None:
1556
ret.extend(repository.find_branches())
1517
branches.extend(repository.find_branches())
1559
1520
@needs_read_lock
1560
1521
def search_missing_revision_ids(self, other, revision_id=None, find_ghosts=True):
1669
1630
return missing_keys
1671
1632
def refresh_data(self):
1672
"""Re-read any data needed to synchronise with disk.
1633
"""Re-read any data needed to to synchronise with disk.
1674
1635
This method is intended to be called after another repository instance
1675
1636
(such as one used by a smart server) has inserted data into the
1676
repository. On all repositories this will work outside of write groups.
1677
Some repository formats (pack and newer for bzrlib native formats)
1678
support refresh_data inside write groups. If called inside a write
1679
group on a repository that does not support refreshing in a write group
1680
IsInWriteGroupError will be raised.
1637
repository. It may not be called during a write group, but may be
1638
called at any other time.
1640
if self.is_in_write_group():
1641
raise errors.InternalBzrError(
1642
"May not refresh_data while in a write group.")
1682
1643
self._refresh_data()
1684
1645
def resume_write_group(self, tokens):
1723
1684
"May not fetch while in a write group.")
1724
1685
# fast path same-url fetch operations
1725
1686
# TODO: lift out to somewhere common with RemoteRepository
1726
# <https://bugs.launchpad.net/bzr/+bug/401646>
1687
# <https://bugs.edge.launchpad.net/bzr/+bug/401646>
1727
1688
if (self.has_same_location(source)
1728
1689
and fetch_spec is None
1729
1690
and self._has_same_fallbacks(source)):
1938
1899
rev = self._serializer.read_revision_from_string(text)
1939
1900
yield (revid, rev)
1903
def get_revision_xml(self, revision_id):
1904
# TODO: jam 20070210 This shouldn't be necessary since get_revision
1905
# would have already do it.
1906
# TODO: jam 20070210 Just use _serializer.write_revision_to_string()
1907
# TODO: this can't just be replaced by:
1908
# return self._serializer.write_revision_to_string(
1909
# self.get_revision(revision_id))
1910
# as cStringIO preservers the encoding unlike write_revision_to_string
1911
# or some other call down the path.
1912
rev = self.get_revision(revision_id)
1913
rev_tmp = cStringIO.StringIO()
1914
# the current serializer..
1915
self._serializer.write_revision(rev, rev_tmp)
1917
return rev_tmp.getvalue()
1941
1919
def get_deltas_for_revisions(self, revisions, specific_fileids=None):
1942
1920
"""Produce a generator of revision deltas.
2186
2164
selected_keys = set((revid,) for revid in revision_ids)
2187
2165
w = _inv_weave or self.inventories
2188
return self._find_file_ids_from_xml_inventory_lines(
2189
w.iter_lines_added_or_present_in_keys(
2190
selected_keys, pb=None),
2166
pb = ui.ui_factory.nested_progress_bar()
2168
return self._find_file_ids_from_xml_inventory_lines(
2169
w.iter_lines_added_or_present_in_keys(
2170
selected_keys, pb=pb),
2193
2175
def iter_files_bytes(self, desired_files):
2194
2176
"""Iterate through file versions.
2404
2386
"""single-document based inventory iteration."""
2405
2387
inv_xmls = self._iter_inventory_xmls(revision_ids, ordering)
2406
2388
for text, revision_id in inv_xmls:
2407
yield self._deserialise_inventory(revision_id, text)
2389
yield self.deserialise_inventory(revision_id, text)
2409
2391
def _iter_inventory_xmls(self, revision_ids, ordering):
2410
2392
if ordering is None:
2442
2424
next_key = None
2445
def _deserialise_inventory(self, revision_id, xml):
2427
def deserialise_inventory(self, revision_id, xml):
2446
2428
"""Transform the xml into an inventory object.
2448
2430
:param revision_id: The expected revision id of the inventory.
2449
2431
:param xml: A serialised inventory.
2451
2433
result = self._serializer.read_inventory_from_string(xml, revision_id,
2452
entry_cache=self._inventory_entry_cache,
2453
return_from_cache=self._safe_to_return_from_cache)
2434
entry_cache=self._inventory_entry_cache)
2454
2435
if result.revision_id != revision_id:
2455
2436
raise AssertionError('revision id mismatch %s != %s' % (
2456
2437
result.revision_id, revision_id))
2440
def serialise_inventory(self, inv):
2441
return self._serializer.write_inventory_to_string(inv)
2443
def _serialise_inventory_to_lines(self, inv):
2444
return self._serializer.write_inventory_to_lines(inv)
2459
2446
def get_serializer_format(self):
2460
2447
return self._serializer.format_num
2462
2449
@needs_read_lock
2463
def _get_inventory_xml(self, revision_id):
2464
"""Get serialized inventory as a string."""
2450
def get_inventory_xml(self, revision_id):
2451
"""Get inventory XML as a file object."""
2465
2452
texts = self._iter_inventory_xmls([revision_id], 'unordered')
2467
2454
text, revision_id = texts.next()
2469
2456
raise errors.HistoryMissing(self, 'inventory', revision_id)
2460
def get_inventory_sha1(self, revision_id):
2461
"""Return the sha1 hash of the inventory entry
2463
return self.get_revision(revision_id).inventory_sha1
2472
2465
def get_rev_id_for_revno(self, revno, known_pair):
2473
2466
"""Return the revision id of a revno, given a later (revno, revid)
2474
2467
pair in the same history.
2526
2519
next_id = parents[0]
2522
def get_revision_inventory(self, revision_id):
2523
"""Return inventory of a past revision."""
2524
# TODO: Unify this with get_inventory()
2525
# bzr 0.0.6 and later imposes the constraint that the inventory_id
2526
# must be the same as its revision, so this is trivial.
2527
if revision_id is None:
2528
# This does not make sense: if there is no revision,
2529
# then it is the current tree inventory surely ?!
2530
# and thus get_root_id() is something that looks at the last
2531
# commit on the branch, and the get_root_id is an inventory check.
2532
raise NotImplementedError
2533
# return Inventory(self.get_root_id())
2535
return self.get_inventory(revision_id)
2528
2537
def is_shared(self):
2529
2538
"""Return True if this repository is flagged as a shared repository."""
2530
2539
raise NotImplementedError(self.is_shared)
2564
2573
return RevisionTree(self, Inventory(root_id=None),
2565
2574
_mod_revision.NULL_REVISION)
2567
inv = self.get_inventory(revision_id)
2576
inv = self.get_revision_inventory(revision_id)
2568
2577
return RevisionTree(self, inv, revision_id)
2570
2579
def revision_trees(self, revision_ids):
2623
2632
keys = tsort.topo_sort(parent_map)
2624
2633
return [None] + list(keys)
2626
def pack(self, hint=None, clean_obsolete_packs=False):
2635
def pack(self, hint=None):
2627
2636
"""Compress the data within the repository.
2629
2638
This operation only makes sense for some repository types. For other
2639
2648
obtained from the result of commit_write_group(). Out of
2640
2649
date hints are simply ignored, because concurrent operations
2641
2650
can obsolete them rapidly.
2643
:param clean_obsolete_packs: Clean obsolete packs immediately after
2647
2653
def get_transaction(self):
2672
2678
def _make_parents_provider(self):
2676
def get_known_graph_ancestry(self, revision_ids):
2677
"""Return the known graph for a set of revision ids and their ancestors.
2679
st = static_tuple.StaticTuple
2680
revision_keys = [st(r_id).intern() for r_id in revision_ids]
2681
known_graph = self.revisions.get_known_graph_ancestry(revision_keys)
2682
return graph.GraphThunkIdsToKeys(known_graph)
2684
2681
def get_graph(self, other_repository=None):
2685
2682
"""Return the graph walker for this repository format"""
2686
2683
parents_provider = self._make_parents_provider()
2781
2778
result.check(callback_refs)
2784
def _warn_if_deprecated(self, branch=None):
2781
def _warn_if_deprecated(self):
2785
2782
global _deprecation_warning_done
2786
2783
if _deprecation_warning_done:
2790
conf = config.GlobalConfig()
2792
conf = branch.get_config()
2793
if conf.suppress_warning('format_deprecation'):
2795
warning("Format %s for %s is deprecated -"
2796
" please use 'bzr upgrade' to get better performance"
2797
% (self._format, self.bzrdir.transport.base))
2799
_deprecation_warning_done = True
2785
_deprecation_warning_done = True
2786
warning("Format %s for %s is deprecated - please use 'bzr upgrade' to get better performance"
2787
% (self._format, self.bzrdir.transport.base))
2801
2789
def supports_rich_root(self):
2802
2790
return self._format.rich_root_data
3085
3073
pack_compresses = False
3086
3074
# Does the repository inventory storage understand references to trees?
3087
3075
supports_tree_reference = None
3088
# Is the format experimental ?
3089
experimental = False
3092
return "%s()" % self.__class__.__name__
3078
return "<%s>" % self.__class__.__name__
3094
3080
def __eq__(self, other):
3095
3081
# format objects are generally stateless
3214
3200
raise NotImplementedError(self.open)
3216
def _run_post_repo_init_hooks(self, repository, a_bzrdir, shared):
3217
from bzrlib.bzrdir import BzrDir, RepoInitHookParams
3218
hooks = BzrDir.hooks['post_repo_init']
3221
params = RepoInitHookParams(repository, self, a_bzrdir, shared)
3226
3203
class MetaDirRepositoryFormat(RepositoryFormat):
3227
3204
"""Common base class for the new repositories using the metadir layout."""
3433
3410
:param revision_id: if None all content is copied, if NULL_REVISION no
3434
3411
content is copied.
3412
:param pb: optional progress bar to use for progress reports. If not
3413
provided a default one will be created.
3438
ui.ui_factory.warn_experimental_format_fetch(self)
3439
from bzrlib.fetch import RepoFetcher
3440
# See <https://launchpad.net/bugs/456077> asking for a warning here
3441
if self.source._format.network_name() != self.target._format.network_name():
3442
ui.ui_factory.show_user_warning('cross_format_fetch',
3443
from_format=self.source._format,
3444
to_format=self.target._format)
3445
f = RepoFetcher(to_repository=self.target,
3416
f = _mod_fetch.RepoFetcher(to_repository=self.target,
3446
3417
from_repository=self.source,
3447
3418
last_revision=revision_id,
3448
3419
fetch_spec=fetch_spec,
3449
find_ghosts=find_ghosts)
3420
pb=pb, find_ghosts=find_ghosts)
3451
3422
def _walk_to_common_revisions(self, revision_ids):
3452
3423
"""Walk out from revision_ids in source to revisions target has.
3869
3840
pending_revisions = []
3870
3841
parent_map = self.source.get_parent_map(revision_ids)
3871
3842
self._fetch_parent_invs_for_stacking(parent_map, cache)
3872
self.source._safe_to_return_from_cache = True
3873
3843
for tree in self.source.revision_trees(revision_ids):
3874
3844
# Find a inventory delta for this revision.
3875
3845
# Find text entries that need to be copied, too.
3923
3893
pending_revisions.append(revision)
3924
3894
cache[current_revision_id] = tree
3925
3895
basis_id = current_revision_id
3926
self.source._safe_to_return_from_cache = False
3927
3896
# Copy file texts
3928
3897
from_texts = self.source.texts
3929
3898
to_texts = self.target.texts
4008
3977
basis_id = self._fetch_batch(batch, basis_id, cache,
4009
3978
a_graph=a_graph)
4011
self.source._safe_to_return_from_cache = False
4012
3980
self.target.abort_write_group()
4026
3994
"""See InterRepository.fetch()."""
4027
3995
if fetch_spec is not None:
4028
3996
raise AssertionError("Not implemented yet...")
4029
ui.ui_factory.warn_experimental_format_fetch(self)
4030
3997
if (not self.source.supports_rich_root()
4031
3998
and self.target.supports_rich_root()):
4032
3999
self._converting_to_rich_root = True
4033
4000
self._revision_id_to_root_id = {}
4035
4002
self._converting_to_rich_root = False
4036
# See <https://launchpad.net/bugs/456077> asking for a warning here
4037
if self.source._format.network_name() != self.target._format.network_name():
4038
ui.ui_factory.show_user_warning('cross_format_fetch',
4039
from_format=self.source._format,
4040
to_format=self.target._format)
4041
4003
revision_ids = self.target.search_missing_revision_ids(self.source,
4042
4004
revision_id, find_ghosts=find_ghosts).get_keys()
4043
4005
if not revision_ids:
4112
4074
:param to_convert: The disk object to convert.
4113
4075
:param pb: a progress bar to use for progress information.
4115
pb = ui.ui_factory.nested_progress_bar()
4118
4080
# this is only useful with metadir layouts - separated repo content.
4119
4081
# trigger an assertion if not such
4120
4082
repo._format.get_format_string()
4121
4083
self.repo_dir = repo.bzrdir
4122
pb.update('Moving repository to repository.backup')
4084
self.step('Moving repository to repository.backup')
4123
4085
self.repo_dir.transport.move('repository', 'repository.backup')
4124
4086
backup_transport = self.repo_dir.transport.clone('repository.backup')
4125
4087
repo._format.check_conversion_target(self.target_format)
4126
4088
self.source_repo = repo._format.open(self.repo_dir,
4128
4090
_override_transport=backup_transport)
4129
pb.update('Creating new repository')
4091
self.step('Creating new repository')
4130
4092
converted = self.target_format.initialize(self.repo_dir,
4131
4093
self.source_repo.is_shared())
4132
4094
converted.lock_write()
4134
pb.update('Copying content')
4096
self.step('Copying content')
4135
4097
self.source_repo.copy_content_into(converted)
4137
4099
converted.unlock()
4138
pb.update('Deleting old repository content')
4100
self.step('Deleting old repository content')
4139
4101
self.repo_dir.transport.delete_tree('repository.backup')
4140
4102
ui.ui_factory.note('repository converted')
4104
def step(self, message):
4105
"""Update the pb by a step."""
4107
self.pb.update(message, self.count, self.total)
4144
4110
_unescape_map = {
4284
4250
is_resume = False
4286
4252
# locked_insert_stream performs a commit|suspend.
4287
return self._locked_insert_stream(stream, src_format,
4253
return self._locked_insert_stream(stream, src_format, is_resume)
4290
4255
self.target_repo.abort_write_group(suppress_errors=True)
4338
4303
# required if the serializers are different only in terms of
4339
4304
# the inventory.
4340
4305
if src_serializer == to_serializer:
4341
self.target_repo.revisions.insert_record_stream(substream)
4306
self.target_repo.revisions.insert_record_stream(
4343
4309
self._extract_and_insert_revisions(substream,
4344
4310
src_serializer)
4452
4418
"""Create a StreamSource streaming from from_repository."""
4453
4419
self.from_repository = from_repository
4454
4420
self.to_format = to_format
4455
self._record_counter = RecordCounter()
4457
4422
def delta_on_metadata(self):
4458
4423
"""Return True if delta's are permitted on metadata streams.
4638
4603
def _get_convertable_inventory_stream(self, revision_ids,
4639
4604
delta_versus_null=False):
4640
# The two formats are sufficiently different that there is no fast
4641
# path, so we need to send just inventorydeltas, which any
4642
# sufficiently modern client can insert into any repository.
4643
# The StreamSink code expects to be able to
4605
# The source is using CHKs, but the target either doesn't or it has a
4606
# different serializer. The StreamSink code expects to be able to
4644
4607
# convert on the target, so we need to put bytes-on-the-wire that can
4645
4608
# be converted. That means inventory deltas (if the remote is <1.19,
4646
4609
# RemoteStreamSink will fallback to VFS to insert the deltas).
added
removed
bzrlib/repository.py
