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
860
857
# 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
860
######################################################################
880
class Repository(_RelockDebugMixin, bzrdir.ControlComponent):
864
class Repository(_RelockDebugMixin):
881
865
"""Repository holding history for one or more branches.
883
867
The repository holds and retrieves historical information including
1042
1026
:seealso: add_inventory, for the contract.
1044
inv_lines = self._serializer.write_inventory_to_lines(inv)
1028
inv_lines = self._serialise_inventory_to_lines(inv)
1045
1029
return self._inventory_add_lines(revision_id, parents,
1046
1030
inv_lines, check_content=False)
1254
1238
"""Check a single text from this repository."""
1255
1239
if kind == 'inventories':
1256
1240
rev_id = record.key[0]
1257
inv = self._deserialise_inventory(rev_id,
1241
inv = self.deserialise_inventory(rev_id,
1258
1242
record.get_bytes_as('fulltext'))
1259
1243
if last_object is not None:
1260
1244
delta = inv._make_delta(last_object)
1305
1289
:param _format: The format of the repository on disk.
1306
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.
1308
# In the future we will have a single api for all stores for
1309
# getting file texts, inventories and revisions, then
1310
# this construct will accept instances of those things.
1311
1296
super(Repository, self).__init__()
1312
1297
self._format = _format
1313
1298
# the following are part of the public API for Repository:
1319
1304
self._reconcile_does_inventory_gc = True
1320
1305
self._reconcile_fixes_text_parents = False
1321
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()
1322
1312
self._write_group = None
1323
1313
# Additional places to query for data.
1324
1314
self._fallback_repositories = []
1325
1315
# An InventoryEntry cache, used during deserialization
1326
1316
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
1318
def __repr__(self):
1340
1319
if self._fallback_repositories:
1389
1368
data during reads, and allows a 'write_group' to be obtained. Write
1390
1369
groups must be used for actual data insertion.
1392
A token should be passed in if you know that you have locked the object
1393
some other way, and need to synchronise this object's state with that
1396
XXX: this docstring is duplicated in many places, e.g. lockable_files.py
1398
1371
:param token: if this is already locked, then lock_write will fail
1399
1372
unless the token matches the existing lock.
1400
1373
:returns: a token if this instance supports tokens, otherwise None.
1403
1376
:raises MismatchedToken: if the specified token doesn't match the token
1404
1377
of the existing lock.
1405
1378
:seealso: start_write_group.
1406
: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
1408
1386
locked = self.is_locked()
1409
token = self.control_files.lock_write(token=token)
1387
result = self.control_files.lock_write(token=token)
1411
self._warn_if_deprecated()
1412
1389
self._note_lock('w')
1413
1390
for repo in self._fallback_repositories:
1414
1391
# Writes don't affect fallback repos
1415
1392
repo.lock_read()
1416
1393
self._refresh_data()
1417
return RepositoryWriteLockResult(self.unlock, token)
1419
1396
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
1397
locked = self.is_locked()
1426
1398
self.control_files.lock_read()
1428
self._warn_if_deprecated()
1429
1400
self._note_lock('r')
1430
1401
for repo in self._fallback_repositories:
1431
1402
repo.lock_read()
1432
1403
self._refresh_data()
1435
1405
def get_physical_lock_status(self):
1436
1406
return self.control_files.get_physical_lock_status()
1497
1467
# now gather global repository information
1498
1468
# XXX: This is available for many repos regardless of listability.
1499
if self.user_transport.listable():
1469
if self.bzrdir.root_transport.listable():
1500
1470
# XXX: do we want to __define len__() ?
1501
1471
# Maybe the versionedfiles object should provide a different
1502
1472
# method to get the number of keys.
1512
1482
:param using: If True, list only branches using this repository.
1514
1484
if using and not self.is_shared():
1515
return self.bzrdir.list_branches()
1486
return [self.bzrdir.open_branch()]
1487
except errors.NotBranchError:
1516
1489
class Evaluator(object):
1518
1491
def __init__(self):
1527
1500
except errors.NoRepositoryPresent:
1530
return False, ([], repository)
1503
return False, (None, repository)
1531
1504
self.first_call = False
1532
value = (bzrdir.list_branches(), None)
1506
value = (bzrdir.open_branch(), None)
1507
except errors.NotBranchError:
1508
value = (None, None)
1533
1509
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)
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)
1540
1516
if not using and repository is not None:
1541
ret.extend(repository.find_branches())
1517
branches.extend(repository.find_branches())
1544
1520
@needs_read_lock
1545
1521
def search_missing_revision_ids(self, other, revision_id=None, find_ghosts=True):
1923
1899
rev = self._serializer.read_revision_from_string(text)
1924
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()
1926
1919
def get_deltas_for_revisions(self, revisions, specific_fileids=None):
1927
1920
"""Produce a generator of revision deltas.
2171
2164
selected_keys = set((revid,) for revid in revision_ids)
2172
2165
w = _inv_weave or self.inventories
2173
return self._find_file_ids_from_xml_inventory_lines(
2174
w.iter_lines_added_or_present_in_keys(
2175
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),
2178
2175
def iter_files_bytes(self, desired_files):
2179
2176
"""Iterate through file versions.
2389
2386
"""single-document based inventory iteration."""
2390
2387
inv_xmls = self._iter_inventory_xmls(revision_ids, ordering)
2391
2388
for text, revision_id in inv_xmls:
2392
yield self._deserialise_inventory(revision_id, text)
2389
yield self.deserialise_inventory(revision_id, text)
2394
2391
def _iter_inventory_xmls(self, revision_ids, ordering):
2395
2392
if ordering is None:
2427
2424
next_key = None
2430
def _deserialise_inventory(self, revision_id, xml):
2427
def deserialise_inventory(self, revision_id, xml):
2431
2428
"""Transform the xml into an inventory object.
2433
2430
:param revision_id: The expected revision id of the inventory.
2434
2431
:param xml: A serialised inventory.
2436
2433
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)
2434
entry_cache=self._inventory_entry_cache)
2439
2435
if result.revision_id != revision_id:
2440
2436
raise AssertionError('revision id mismatch %s != %s' % (
2441
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)
2444
2446
def get_serializer_format(self):
2445
2447
return self._serializer.format_num
2447
2449
@needs_read_lock
2448
def _get_inventory_xml(self, revision_id):
2449
"""Get serialized inventory as a string."""
2450
def get_inventory_xml(self, revision_id):
2451
"""Get inventory XML as a file object."""
2450
2452
texts = self._iter_inventory_xmls([revision_id], 'unordered')
2452
2454
text, revision_id = texts.next()
2454
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
2457
2465
def get_rev_id_for_revno(self, revno, known_pair):
2458
2466
"""Return the revision id of a revno, given a later (revno, revid)
2459
2467
pair in the same history.
2511
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)
2513
2537
def is_shared(self):
2514
2538
"""Return True if this repository is flagged as a shared repository."""
2515
2539
raise NotImplementedError(self.is_shared)
2549
2573
return RevisionTree(self, Inventory(root_id=None),
2550
2574
_mod_revision.NULL_REVISION)
2552
inv = self.get_inventory(revision_id)
2576
inv = self.get_revision_inventory(revision_id)
2553
2577
return RevisionTree(self, inv, revision_id)
2555
2579
def revision_trees(self, revision_ids):
2608
2632
keys = tsort.topo_sort(parent_map)
2609
2633
return [None] + list(keys)
2611
def pack(self, hint=None, clean_obsolete_packs=False):
2635
def pack(self, hint=None):
2612
2636
"""Compress the data within the repository.
2614
2638
This operation only makes sense for some repository types. For other
2624
2648
obtained from the result of commit_write_group(). Out of
2625
2649
date hints are simply ignored, because concurrent operations
2626
2650
can obsolete them rapidly.
2628
:param clean_obsolete_packs: Clean obsolete packs immediately after
2632
2653
def get_transaction(self):
2657
2678
def _make_parents_provider(self):
2661
def get_known_graph_ancestry(self, revision_ids):
2662
"""Return the known graph for a set of revision ids and their ancestors.
2664
st = static_tuple.StaticTuple
2665
revision_keys = [st(r_id).intern() for r_id in revision_ids]
2666
known_graph = self.revisions.get_known_graph_ancestry(revision_keys)
2667
return graph.GraphThunkIdsToKeys(known_graph)
2669
2681
def get_graph(self, other_repository=None):
2670
2682
"""Return the graph walker for this repository format"""
2671
2683
parents_provider = self._make_parents_provider()
2766
2778
result.check(callback_refs)
2769
def _warn_if_deprecated(self, branch=None):
2781
def _warn_if_deprecated(self):
2770
2782
global _deprecation_warning_done
2771
2783
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
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))
2786
2789
def supports_rich_root(self):
2787
2790
return self._format.rich_root_data
3070
3073
pack_compresses = False
3071
3074
# Does the repository inventory storage understand references to trees?
3072
3075
supports_tree_reference = None
3073
# Is the format experimental ?
3074
experimental = False
3077
return "%s()" % self.__class__.__name__
3078
return "<%s>" % self.__class__.__name__
3079
3080
def __eq__(self, other):
3080
3081
# format objects are generally stateless
3199
3200
raise NotImplementedError(self.open)
3201
def _run_post_repo_init_hooks(self, repository, a_bzrdir, shared):
3202
from bzrlib.bzrdir import BzrDir, RepoInitHookParams
3203
hooks = BzrDir.hooks['post_repo_init']
3206
params = RepoInitHookParams(repository, self, a_bzrdir, shared)
3211
3203
class MetaDirRepositoryFormat(RepositoryFormat):
3212
3204
"""Common base class for the new repositories using the metadir layout."""
3418
3410
:param revision_id: if None all content is copied, if NULL_REVISION no
3419
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.
3423
ui.ui_factory.warn_experimental_format_fetch(self)
3424
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
f = RepoFetcher(to_repository=self.target,
3416
f = _mod_fetch.RepoFetcher(to_repository=self.target,
3431
3417
from_repository=self.source,
3432
3418
last_revision=revision_id,
3433
3419
fetch_spec=fetch_spec,
3434
find_ghosts=find_ghosts)
3420
pb=pb, find_ghosts=find_ghosts)
3436
3422
def _walk_to_common_revisions(self, revision_ids):
3437
3423
"""Walk out from revision_ids in source to revisions target has.
3854
3840
pending_revisions = []
3855
3841
parent_map = self.source.get_parent_map(revision_ids)
3856
3842
self._fetch_parent_invs_for_stacking(parent_map, cache)
3857
self.source._safe_to_return_from_cache = True
3858
3843
for tree in self.source.revision_trees(revision_ids):
3859
3844
# Find a inventory delta for this revision.
3860
3845
# Find text entries that need to be copied, too.
3908
3893
pending_revisions.append(revision)
3909
3894
cache[current_revision_id] = tree
3910
3895
basis_id = current_revision_id
3911
self.source._safe_to_return_from_cache = False
3912
3896
# Copy file texts
3913
3897
from_texts = self.source.texts
3914
3898
to_texts = self.target.texts
3993
3977
basis_id = self._fetch_batch(batch, basis_id, cache,
3994
3978
a_graph=a_graph)
3996
self.source._safe_to_return_from_cache = False
3997
3980
self.target.abort_write_group()
4011
3994
"""See InterRepository.fetch()."""
4012
3995
if fetch_spec is not None:
4013
3996
raise AssertionError("Not implemented yet...")
4014
ui.ui_factory.warn_experimental_format_fetch(self)
4015
3997
if (not self.source.supports_rich_root()
4016
3998
and self.target.supports_rich_root()):
4017
3999
self._converting_to_rich_root = True
4018
4000
self._revision_id_to_root_id = {}
4020
4002
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
4003
revision_ids = self.target.search_missing_revision_ids(self.source,
4027
4004
revision_id, find_ghosts=find_ghosts).get_keys()
4028
4005
if not revision_ids:
4097
4074
:param to_convert: The disk object to convert.
4098
4075
:param pb: a progress bar to use for progress information.
4100
pb = ui.ui_factory.nested_progress_bar()
4103
4080
# this is only useful with metadir layouts - separated repo content.
4104
4081
# trigger an assertion if not such
4105
4082
repo._format.get_format_string()
4106
4083
self.repo_dir = repo.bzrdir
4107
pb.update('Moving repository to repository.backup')
4084
self.step('Moving repository to repository.backup')
4108
4085
self.repo_dir.transport.move('repository', 'repository.backup')
4109
4086
backup_transport = self.repo_dir.transport.clone('repository.backup')
4110
4087
repo._format.check_conversion_target(self.target_format)
4111
4088
self.source_repo = repo._format.open(self.repo_dir,
4113
4090
_override_transport=backup_transport)
4114
pb.update('Creating new repository')
4091
self.step('Creating new repository')
4115
4092
converted = self.target_format.initialize(self.repo_dir,
4116
4093
self.source_repo.is_shared())
4117
4094
converted.lock_write()
4119
pb.update('Copying content')
4096
self.step('Copying content')
4120
4097
self.source_repo.copy_content_into(converted)
4122
4099
converted.unlock()
4123
pb.update('Deleting old repository content')
4100
self.step('Deleting old repository content')
4124
4101
self.repo_dir.transport.delete_tree('repository.backup')
4125
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)
4129
4110
_unescape_map = {
4622
4603
def _get_convertable_inventory_stream(self, revision_ids,
4623
4604
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
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
4628
4607
# convert on the target, so we need to put bytes-on-the-wire that can
4629
4608
# be converted. That means inventory deltas (if the remote is <1.19,
4630
4609
# RemoteStreamSink will fallback to VFS to insert the deltas).
added
removed
bzrlib/repository.py
