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
53
49
from bzrlib.testament import Testament
56
from bzrlib.decorators import needs_read_lock, needs_write_lock, only_raises
52
from bzrlib.decorators import needs_read_lock, needs_write_lock
57
53
from bzrlib.inter import InterObject
58
54
from bzrlib.inventory import (
64
from bzrlib.lock import _RelockDebugMixin
65
60
from bzrlib import registry
66
61
from bzrlib.trace import (
67
62
log_exception_quietly, note, mutter, mutter_callsite, warning)
210
205
# an inventory delta was accumulated without creating a new
212
207
basis_id = self.basis_delta_revision
213
# We ignore the 'inventory' returned by add_inventory_by_delta
214
# because self.new_inventory is used to hint to the rest of the
215
# system what code path was taken
216
self.inv_sha1, _ = self.repository.add_inventory_by_delta(
208
self.inv_sha1 = self.repository.add_inventory_by_delta(
217
209
basis_id, self._basis_delta, self._new_revision_id,
867
class Repository(_RelockDebugMixin, bzrdir.ControlComponent):
859
class Repository(object):
868
860
"""Repository holding history for one or more branches.
870
862
The repository holds and retrieves historical information including
1029
1021
:seealso: add_inventory, for the contract.
1031
inv_lines = self._serializer.write_inventory_to_lines(inv)
1023
inv_lines = self._serialise_inventory_to_lines(inv)
1032
1024
return self._inventory_add_lines(revision_id, parents,
1033
1025
inv_lines, check_content=False)
1241
1233
"""Check a single text from this repository."""
1242
1234
if kind == 'inventories':
1243
1235
rev_id = record.key[0]
1244
inv = self._deserialise_inventory(rev_id,
1236
inv = self.deserialise_inventory(rev_id,
1245
1237
record.get_bytes_as('fulltext'))
1246
1238
if last_object is not None:
1247
1239
delta = inv._make_delta(last_object)
1292
1284
:param _format: The format of the repository on disk.
1293
1285
:param a_bzrdir: The BzrDir of the repository.
1287
In the future we will have a single api for all stores for
1288
getting file texts, inventories and revisions, then
1289
this construct will accept instances of those things.
1295
# In the future we will have a single api for all stores for
1296
# getting file texts, inventories and revisions, then
1297
# this construct will accept instances of those things.
1298
1291
super(Repository, self).__init__()
1299
1292
self._format = _format
1300
1293
# the following are part of the public API for Repository:
1306
1299
self._reconcile_does_inventory_gc = True
1307
1300
self._reconcile_fixes_text_parents = False
1308
1301
self._reconcile_backsup_inventory = True
1302
# not right yet - should be more semantically clear ?
1304
# TODO: make sure to construct the right store classes, etc, depending
1305
# on whether escaping is required.
1306
self._warn_if_deprecated()
1309
1307
self._write_group = None
1310
1308
# Additional places to query for data.
1311
1309
self._fallback_repositories = []
1312
1310
# An InventoryEntry cache, used during deserialization
1313
1311
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
1313
def __repr__(self):
1327
1314
if self._fallback_repositories:
1394
1381
locked = self.is_locked()
1395
1382
result = self.control_files.lock_write(token=token)
1397
self._warn_if_deprecated()
1398
self._note_lock('w')
1399
1384
for repo in self._fallback_repositories:
1400
1385
# Writes don't affect fallback repos
1401
1386
repo.lock_read()
1477
1460
# now gather global repository information
1478
1461
# XXX: This is available for many repos regardless of listability.
1479
if self.user_transport.listable():
1462
if self.bzrdir.root_transport.listable():
1480
1463
# XXX: do we want to __define len__() ?
1481
1464
# Maybe the versionedfiles object should provide a different
1482
1465
# method to get the number of keys.
1492
1475
:param using: If True, list only branches using this repository.
1494
1477
if using and not self.is_shared():
1495
return self.bzrdir.list_branches()
1479
return [self.bzrdir.open_branch()]
1480
except errors.NotBranchError:
1496
1482
class Evaluator(object):
1498
1484
def __init__(self):
1507
1493
except errors.NoRepositoryPresent:
1510
return False, ([], repository)
1496
return False, (None, repository)
1511
1497
self.first_call = False
1512
value = (bzrdir.list_branches(), None)
1499
value = (bzrdir.open_branch(), None)
1500
except errors.NotBranchError:
1501
value = (None, None)
1513
1502
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)
1505
for branch, repository in bzrdir.BzrDir.find_bzrdirs(
1506
self.bzrdir.root_transport, evaluate=Evaluator()):
1507
if branch is not None:
1508
branches.append(branch)
1520
1509
if not using and repository is not None:
1521
ret.extend(repository.find_branches())
1510
branches.extend(repository.find_branches())
1524
1513
@needs_read_lock
1525
1514
def search_missing_revision_ids(self, other, revision_id=None, find_ghosts=True):
1903
1891
rev = self._serializer.read_revision_from_string(text)
1904
1892
yield (revid, rev)
1895
def get_revision_xml(self, revision_id):
1896
# TODO: jam 20070210 This shouldn't be necessary since get_revision
1897
# would have already do it.
1898
# TODO: jam 20070210 Just use _serializer.write_revision_to_string()
1899
# TODO: this can't just be replaced by:
1900
# return self._serializer.write_revision_to_string(
1901
# self.get_revision(revision_id))
1902
# as cStringIO preservers the encoding unlike write_revision_to_string
1903
# or some other call down the path.
1904
rev = self.get_revision(revision_id)
1905
rev_tmp = cStringIO.StringIO()
1906
# the current serializer..
1907
self._serializer.write_revision(rev, rev_tmp)
1909
return rev_tmp.getvalue()
1906
1911
def get_deltas_for_revisions(self, revisions, specific_fileids=None):
1907
1912
"""Produce a generator of revision deltas.
2151
2156
selected_keys = set((revid,) for revid in revision_ids)
2152
2157
w = _inv_weave or self.inventories
2153
return self._find_file_ids_from_xml_inventory_lines(
2154
w.iter_lines_added_or_present_in_keys(
2155
selected_keys, pb=None),
2158
pb = ui.ui_factory.nested_progress_bar()
2160
return self._find_file_ids_from_xml_inventory_lines(
2161
w.iter_lines_added_or_present_in_keys(
2162
selected_keys, pb=pb),
2158
2167
def iter_files_bytes(self, desired_files):
2159
2168
"""Iterate through file versions.
2320
2329
num_file_ids = len(file_ids)
2321
2330
for file_id, altered_versions in file_ids.iteritems():
2322
2331
if pb is not None:
2323
pb.update("Fetch texts", count, num_file_ids)
2332
pb.update("fetch texts", count, num_file_ids)
2325
2334
yield ("file", file_id, altered_versions)
2369
2378
"""single-document based inventory iteration."""
2370
2379
inv_xmls = self._iter_inventory_xmls(revision_ids, ordering)
2371
2380
for text, revision_id in inv_xmls:
2372
yield self._deserialise_inventory(revision_id, text)
2381
yield self.deserialise_inventory(revision_id, text)
2374
2383
def _iter_inventory_xmls(self, revision_ids, ordering):
2375
2384
if ordering is None:
2407
2416
next_key = None
2410
def _deserialise_inventory(self, revision_id, xml):
2419
def deserialise_inventory(self, revision_id, xml):
2411
2420
"""Transform the xml into an inventory object.
2413
2422
:param revision_id: The expected revision id of the inventory.
2414
2423
:param xml: A serialised inventory.
2416
2425
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)
2426
entry_cache=self._inventory_entry_cache)
2419
2427
if result.revision_id != revision_id:
2420
2428
raise AssertionError('revision id mismatch %s != %s' % (
2421
2429
result.revision_id, revision_id))
2432
def serialise_inventory(self, inv):
2433
return self._serializer.write_inventory_to_string(inv)
2435
def _serialise_inventory_to_lines(self, inv):
2436
return self._serializer.write_inventory_to_lines(inv)
2424
2438
def get_serializer_format(self):
2425
2439
return self._serializer.format_num
2427
2441
@needs_read_lock
2428
def _get_inventory_xml(self, revision_id):
2429
"""Get serialized inventory as a string."""
2442
def get_inventory_xml(self, revision_id):
2443
"""Get inventory XML as a file object."""
2430
2444
texts = self._iter_inventory_xmls([revision_id], 'unordered')
2432
2446
text, revision_id = texts.next()
2434
2448
raise errors.HistoryMissing(self, 'inventory', revision_id)
2452
def get_inventory_sha1(self, revision_id):
2453
"""Return the sha1 hash of the inventory entry
2455
return self.get_revision(revision_id).inventory_sha1
2437
2457
def get_rev_id_for_revno(self, revno, known_pair):
2438
2458
"""Return the revision id of a revno, given a later (revno, revid)
2439
2459
pair in the same history.
2491
2511
next_id = parents[0]
2514
def get_revision_inventory(self, revision_id):
2515
"""Return inventory of a past revision."""
2516
# TODO: Unify this with get_inventory()
2517
# bzr 0.0.6 and later imposes the constraint that the inventory_id
2518
# must be the same as its revision, so this is trivial.
2519
if revision_id is None:
2520
# This does not make sense: if there is no revision,
2521
# then it is the current tree inventory surely ?!
2522
# and thus get_root_id() is something that looks at the last
2523
# commit on the branch, and the get_root_id is an inventory check.
2524
raise NotImplementedError
2525
# return Inventory(self.get_root_id())
2527
return self.get_inventory(revision_id)
2493
2529
def is_shared(self):
2494
2530
"""Return True if this repository is flagged as a shared repository."""
2495
2531
raise NotImplementedError(self.is_shared)
2529
2565
return RevisionTree(self, Inventory(root_id=None),
2530
2566
_mod_revision.NULL_REVISION)
2532
inv = self.get_inventory(revision_id)
2568
inv = self.get_revision_inventory(revision_id)
2533
2569
return RevisionTree(self, inv, revision_id)
2535
2571
def revision_trees(self, revision_ids):
2588
2624
keys = tsort.topo_sort(parent_map)
2589
2625
return [None] + list(keys)
2591
def pack(self, hint=None, clean_obsolete_packs=False):
2627
def pack(self, hint=None):
2592
2628
"""Compress the data within the repository.
2594
2630
This operation only makes sense for some repository types. For other
2604
2640
obtained from the result of commit_write_group(). Out of
2605
2641
date hints are simply ignored, because concurrent operations
2606
2642
can obsolete them rapidly.
2608
:param clean_obsolete_packs: Clean obsolete packs immediately after
2612
2645
def get_transaction(self):
2628
2661
for ((revision_id,), parent_keys) in \
2629
2662
self.revisions.get_parent_map(query_keys).iteritems():
2630
2663
if parent_keys:
2631
result[revision_id] = tuple([parent_revid
2632
for (parent_revid,) in parent_keys])
2664
result[revision_id] = tuple(parent_revid
2665
for (parent_revid,) in parent_keys)
2634
2667
result[revision_id] = (_mod_revision.NULL_REVISION,)
2637
2670
def _make_parents_provider(self):
2641
def get_known_graph_ancestry(self, revision_ids):
2642
"""Return the known graph for a set of revision ids and their ancestors.
2644
st = static_tuple.StaticTuple
2645
revision_keys = [st(r_id).intern() for r_id in revision_ids]
2646
known_graph = self.revisions.get_known_graph_ancestry(revision_keys)
2647
return graph.GraphThunkIdsToKeys(known_graph)
2649
2673
def get_graph(self, other_repository=None):
2650
2674
"""Return the graph walker for this repository format"""
2651
2675
parents_provider = self._make_parents_provider()
2746
2770
result.check(callback_refs)
2749
def _warn_if_deprecated(self, branch=None):
2773
def _warn_if_deprecated(self):
2750
2774
global _deprecation_warning_done
2751
2775
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
2777
_deprecation_warning_done = True
2778
warning("Format %s for %s is deprecated - please use 'bzr upgrade' to get better performance"
2779
% (self._format, self.bzrdir.transport.base))
2766
2781
def supports_rich_root(self):
2767
2782
return self._format.rich_root_data
3050
3065
pack_compresses = False
3051
3066
# Does the repository inventory storage understand references to trees?
3052
3067
supports_tree_reference = None
3053
# Is the format experimental ?
3054
experimental = False
3057
return "%s()" % self.__class__.__name__
3070
return "<%s>" % self.__class__.__name__
3059
3072
def __eq__(self, other):
3060
3073
# format objects are generally stateless
3075
3088
transport = a_bzrdir.get_repository_transport(None)
3076
format_string = transport.get_bytes("format")
3089
format_string = transport.get("format").read()
3077
3090
return format_registry.get(format_string)
3078
3091
except errors.NoSuchFile:
3079
3092
raise errors.NoRepositoryPresent(a_bzrdir)
3179
3192
raise NotImplementedError(self.open)
3181
def _run_post_repo_init_hooks(self, repository, a_bzrdir, shared):
3182
from bzrlib.bzrdir import BzrDir, RepoInitHookParams
3183
hooks = BzrDir.hooks['post_repo_init']
3186
params = RepoInitHookParams(repository, self, a_bzrdir, shared)
3191
3195
class MetaDirRepositoryFormat(RepositoryFormat):
3192
3196
"""Common base class for the new repositories using the metadir layout."""
3398
3402
:param revision_id: if None all content is copied, if NULL_REVISION no
3399
3403
content is copied.
3404
:param pb: optional progress bar to use for progress reports. If not
3405
provided a default one will be created.
3403
ui.ui_factory.warn_experimental_format_fetch(self)
3404
3408
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
3409
f = RepoFetcher(to_repository=self.target,
3411
3410
from_repository=self.source,
3412
3411
last_revision=revision_id,
3413
3412
fetch_spec=fetch_spec,
3414
find_ghosts=find_ghosts)
3413
pb=pb, find_ghosts=find_ghosts)
3416
3415
def _walk_to_common_revisions(self, revision_ids):
3417
3416
"""Walk out from revision_ids in source to revisions target has.
3586
3585
self.target.texts.insert_record_stream(
3587
3586
self.source.texts.get_record_stream(
3588
3587
self.source.texts.keys(), 'topological', False))
3589
pb.update('Copying inventory', 0, 1)
3588
pb.update('copying inventory', 0, 1)
3590
3589
self.target.inventories.insert_record_stream(
3591
3590
self.source.inventories.get_record_stream(
3592
3591
self.source.inventories.keys(), 'topological', False))
3813
3812
basis_id, delta, current_revision_id, parents_parents)
3814
3813
cache[current_revision_id] = parent_tree
3816
def _fetch_batch(self, revision_ids, basis_id, cache, a_graph=None):
3815
def _fetch_batch(self, revision_ids, basis_id, cache):
3817
3816
"""Fetch across a few revisions.
3819
3818
:param revision_ids: The revisions to copy
3820
3819
:param basis_id: The revision_id of a tree that must be in cache, used
3821
3820
as a basis for delta when no other base is available
3822
3821
: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
3822
:return: The revision_id of the last converted tree. The RevisionTree
3826
3823
for it will be in cache
3834
3831
pending_revisions = []
3835
3832
parent_map = self.source.get_parent_map(revision_ids)
3836
3833
self._fetch_parent_invs_for_stacking(parent_map, cache)
3837
self.source._safe_to_return_from_cache = True
3838
3834
for tree in self.source.revision_trees(revision_ids):
3839
3835
# Find a inventory delta for this revision.
3840
3836
# Find text entries that need to be copied, too.
3888
3884
pending_revisions.append(revision)
3889
3885
cache[current_revision_id] = tree
3890
3886
basis_id = current_revision_id
3891
self.source._safe_to_return_from_cache = False
3892
3887
# Copy file texts
3893
3888
from_texts = self.source.texts
3894
3889
to_texts = self.target.texts
3895
3890
if root_keys_to_create:
3896
root_stream = _mod_fetch._new_root_data_stream(
3891
from bzrlib.fetch import _new_root_data_stream
3892
root_stream = _new_root_data_stream(
3897
3893
root_keys_to_create, self._revision_id_to_root_id, parent_map,
3898
self.source, graph=a_graph)
3899
3895
to_texts.insert_record_stream(root_stream)
3900
3896
to_texts.insert_record_stream(from_texts.get_record_stream(
3901
3897
text_keys, self.target._format._fetch_order,
3958
3954
cache[basis_id] = basis_tree
3959
3955
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
3957
for offset in range(0, len(revision_ids), batch_size):
3968
3958
self.target.start_write_group()
3970
3960
pb.update('Transferring revisions', offset,
3971
3961
len(revision_ids))
3972
3962
batch = revision_ids[offset:offset+batch_size]
3973
basis_id = self._fetch_batch(batch, basis_id, cache,
3963
basis_id = self._fetch_batch(batch, basis_id, cache)
3976
self.source._safe_to_return_from_cache = False
3977
3965
self.target.abort_write_group()
3991
3979
"""See InterRepository.fetch()."""
3992
3980
if fetch_spec is not None:
3993
3981
raise AssertionError("Not implemented yet...")
3994
ui.ui_factory.warn_experimental_format_fetch(self)
3995
3982
if (not self.source.supports_rich_root()
3996
3983
and self.target.supports_rich_root()):
3997
3984
self._converting_to_rich_root = True
3998
3985
self._revision_id_to_root_id = {}
4000
3987
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
3988
revision_ids = self.target.search_missing_revision_ids(self.source,
4007
3989
revision_id, find_ghosts=find_ghosts).get_keys()
4008
3990
if not revision_ids:
4077
4059
:param to_convert: The disk object to convert.
4078
4060
:param pb: a progress bar to use for progress information.
4080
pb = ui.ui_factory.nested_progress_bar()
4083
4065
# this is only useful with metadir layouts - separated repo content.
4084
4066
# trigger an assertion if not such
4085
4067
repo._format.get_format_string()
4086
4068
self.repo_dir = repo.bzrdir
4087
pb.update('Moving repository to repository.backup')
4069
self.step('Moving repository to repository.backup')
4088
4070
self.repo_dir.transport.move('repository', 'repository.backup')
4089
4071
backup_transport = self.repo_dir.transport.clone('repository.backup')
4090
4072
repo._format.check_conversion_target(self.target_format)
4091
4073
self.source_repo = repo._format.open(self.repo_dir,
4093
4075
_override_transport=backup_transport)
4094
pb.update('Creating new repository')
4076
self.step('Creating new repository')
4095
4077
converted = self.target_format.initialize(self.repo_dir,
4096
4078
self.source_repo.is_shared())
4097
4079
converted.lock_write()
4099
pb.update('Copying content')
4081
self.step('Copying content into repository.')
4100
4082
self.source_repo.copy_content_into(converted)
4102
4084
converted.unlock()
4103
pb.update('Deleting old repository content')
4085
self.step('Deleting old repository content.')
4104
4086
self.repo_dir.transport.delete_tree('repository.backup')
4105
ui.ui_factory.note('repository converted')
4087
self.pb.note('repository converted')
4089
def step(self, message):
4090
"""Update the pb by a step."""
4092
self.pb.update(message, self.count, self.total)
4109
4095
_unescape_map = {
4330
4316
if versioned_file is None:
4332
# TODO: key is often going to be a StaticTuple object
4333
# I don't believe we can define a method by which
4334
# (prefix,) + StaticTuple will work, though we could
4335
# define a StaticTuple.sq_concat that would allow you to
4336
# pass in either a tuple or a StaticTuple as the second
4337
# object, so instead we could have:
4338
# StaticTuple(prefix) + key here...
4339
4318
missing_keys.update((prefix,) + key for key in
4340
4319
versioned_file.get_missing_compression_parent_keys())
4341
4320
except NotImplementedError:
4453
4432
fetching the inventory weave.
4455
4434
if self._rich_root_upgrade():
4456
return _mod_fetch.Inter1and2Helper(
4436
return bzrlib.fetch.Inter1and2Helper(
4457
4437
self.from_repository).generate_root_texts(revs)
4602
4582
def _get_convertable_inventory_stream(self, revision_ids,
4603
4583
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
4584
# The source is using CHKs, but the target either doesn't or it has a
4585
# different serializer. The StreamSink code expects to be able to
4608
4586
# convert on the target, so we need to put bytes-on-the-wire that can
4609
4587
# be converted. That means inventory deltas (if the remote is <1.19,
4610
4588
# RemoteStreamSink will fallback to VFS to insert the deltas).
added
removed
bzrlib/repository.py
