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
56
52
from bzrlib.decorators import needs_read_lock, needs_write_lock, only_raises
53
from bzrlib.lock import _RelockDebugMixin
57
54
from bzrlib.inter import InterObject
58
55
from bzrlib.inventory import (
64
from bzrlib.lock import _RelockDebugMixin
65
61
from bzrlib import registry
66
62
from bzrlib.trace import (
67
63
log_exception_quietly, note, mutter, mutter_callsite, warning)
210
206
# an inventory delta was accumulated without creating a new
212
208
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(
209
self.inv_sha1 = self.repository.add_inventory_by_delta(
217
210
basis_id, self._basis_delta, self._new_revision_id,
867
class Repository(_RelockDebugMixin, bzrdir.ControlComponent):
860
class Repository(_RelockDebugMixin):
868
861
"""Repository holding history for one or more branches.
870
863
The repository holds and retrieves historical information including
1029
1022
:seealso: add_inventory, for the contract.
1031
inv_lines = self._serializer.write_inventory_to_lines(inv)
1024
inv_lines = self._serialise_inventory_to_lines(inv)
1032
1025
return self._inventory_add_lines(revision_id, parents,
1033
1026
inv_lines, check_content=False)
1241
1234
"""Check a single text from this repository."""
1242
1235
if kind == 'inventories':
1243
1236
rev_id = record.key[0]
1244
inv = self._deserialise_inventory(rev_id,
1237
inv = self.deserialise_inventory(rev_id,
1245
1238
record.get_bytes_as('fulltext'))
1246
1239
if last_object is not None:
1247
1240
delta = inv._make_delta(last_object)
1292
1285
:param _format: The format of the repository on disk.
1293
1286
:param a_bzrdir: The BzrDir of the repository.
1288
In the future we will have a single api for all stores for
1289
getting file texts, inventories and revisions, then
1290
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
1292
super(Repository, self).__init__()
1299
1293
self._format = _format
1300
1294
# the following are part of the public API for Repository:
1306
1300
self._reconcile_does_inventory_gc = True
1307
1301
self._reconcile_fixes_text_parents = False
1308
1302
self._reconcile_backsup_inventory = True
1303
# not right yet - should be more semantically clear ?
1305
# TODO: make sure to construct the right store classes, etc, depending
1306
# on whether escaping is required.
1307
self._warn_if_deprecated()
1309
1308
self._write_group = None
1310
1309
# Additional places to query for data.
1311
1310
self._fallback_repositories = []
1312
1311
# An InventoryEntry cache, used during deserialization
1313
1312
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
1314
def __repr__(self):
1327
1315
if self._fallback_repositories:
1394
1382
locked = self.is_locked()
1395
1383
result = self.control_files.lock_write(token=token)
1397
self._warn_if_deprecated()
1398
1385
self._note_lock('w')
1399
1386
for repo in self._fallback_repositories:
1400
1387
# Writes don't affect fallback repos
1477
1463
# now gather global repository information
1478
1464
# XXX: This is available for many repos regardless of listability.
1479
if self.user_transport.listable():
1465
if self.bzrdir.root_transport.listable():
1480
1466
# XXX: do we want to __define len__() ?
1481
1467
# Maybe the versionedfiles object should provide a different
1482
1468
# method to get the number of keys.
1492
1478
:param using: If True, list only branches using this repository.
1494
1480
if using and not self.is_shared():
1495
return self.bzrdir.list_branches()
1482
return [self.bzrdir.open_branch()]
1483
except errors.NotBranchError:
1496
1485
class Evaluator(object):
1498
1487
def __init__(self):
1507
1496
except errors.NoRepositoryPresent:
1510
return False, ([], repository)
1499
return False, (None, repository)
1511
1500
self.first_call = False
1512
value = (bzrdir.list_branches(), None)
1502
value = (bzrdir.open_branch(), None)
1503
except errors.NotBranchError:
1504
value = (None, None)
1513
1505
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)
1508
for branch, repository in bzrdir.BzrDir.find_bzrdirs(
1509
self.bzrdir.root_transport, evaluate=Evaluator()):
1510
if branch is not None:
1511
branches.append(branch)
1520
1512
if not using and repository is not None:
1521
ret.extend(repository.find_branches())
1513
branches.extend(repository.find_branches())
1524
1516
@needs_read_lock
1525
1517
def search_missing_revision_ids(self, other, revision_id=None, find_ghosts=True):
1903
1895
rev = self._serializer.read_revision_from_string(text)
1904
1896
yield (revid, rev)
1899
def get_revision_xml(self, revision_id):
1900
# TODO: jam 20070210 This shouldn't be necessary since get_revision
1901
# would have already do it.
1902
# TODO: jam 20070210 Just use _serializer.write_revision_to_string()
1903
# TODO: this can't just be replaced by:
1904
# return self._serializer.write_revision_to_string(
1905
# self.get_revision(revision_id))
1906
# as cStringIO preservers the encoding unlike write_revision_to_string
1907
# or some other call down the path.
1908
rev = self.get_revision(revision_id)
1909
rev_tmp = cStringIO.StringIO()
1910
# the current serializer..
1911
self._serializer.write_revision(rev, rev_tmp)
1913
return rev_tmp.getvalue()
1906
1915
def get_deltas_for_revisions(self, revisions, specific_fileids=None):
1907
1916
"""Produce a generator of revision deltas.
2151
2160
selected_keys = set((revid,) for revid in revision_ids)
2152
2161
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),
2162
pb = ui.ui_factory.nested_progress_bar()
2164
return self._find_file_ids_from_xml_inventory_lines(
2165
w.iter_lines_added_or_present_in_keys(
2166
selected_keys, pb=pb),
2158
2171
def iter_files_bytes(self, desired_files):
2159
2172
"""Iterate through file versions.
2369
2382
"""single-document based inventory iteration."""
2370
2383
inv_xmls = self._iter_inventory_xmls(revision_ids, ordering)
2371
2384
for text, revision_id in inv_xmls:
2372
yield self._deserialise_inventory(revision_id, text)
2385
yield self.deserialise_inventory(revision_id, text)
2374
2387
def _iter_inventory_xmls(self, revision_ids, ordering):
2375
2388
if ordering is None:
2407
2420
next_key = None
2410
def _deserialise_inventory(self, revision_id, xml):
2423
def deserialise_inventory(self, revision_id, xml):
2411
2424
"""Transform the xml into an inventory object.
2413
2426
:param revision_id: The expected revision id of the inventory.
2414
2427
:param xml: A serialised inventory.
2416
2429
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)
2430
entry_cache=self._inventory_entry_cache)
2419
2431
if result.revision_id != revision_id:
2420
2432
raise AssertionError('revision id mismatch %s != %s' % (
2421
2433
result.revision_id, revision_id))
2436
def serialise_inventory(self, inv):
2437
return self._serializer.write_inventory_to_string(inv)
2439
def _serialise_inventory_to_lines(self, inv):
2440
return self._serializer.write_inventory_to_lines(inv)
2424
2442
def get_serializer_format(self):
2425
2443
return self._serializer.format_num
2427
2445
@needs_read_lock
2428
def _get_inventory_xml(self, revision_id):
2429
"""Get serialized inventory as a string."""
2446
def get_inventory_xml(self, revision_id):
2447
"""Get inventory XML as a file object."""
2430
2448
texts = self._iter_inventory_xmls([revision_id], 'unordered')
2432
2450
text, revision_id = texts.next()
2434
2452
raise errors.HistoryMissing(self, 'inventory', revision_id)
2456
def get_inventory_sha1(self, revision_id):
2457
"""Return the sha1 hash of the inventory entry
2459
return self.get_revision(revision_id).inventory_sha1
2437
2461
def get_rev_id_for_revno(self, revno, known_pair):
2438
2462
"""Return the revision id of a revno, given a later (revno, revid)
2439
2463
pair in the same history.
2491
2515
next_id = parents[0]
2518
def get_revision_inventory(self, revision_id):
2519
"""Return inventory of a past revision."""
2520
# TODO: Unify this with get_inventory()
2521
# bzr 0.0.6 and later imposes the constraint that the inventory_id
2522
# must be the same as its revision, so this is trivial.
2523
if revision_id is None:
2524
# This does not make sense: if there is no revision,
2525
# then it is the current tree inventory surely ?!
2526
# and thus get_root_id() is something that looks at the last
2527
# commit on the branch, and the get_root_id is an inventory check.
2528
raise NotImplementedError
2529
# return Inventory(self.get_root_id())
2531
return self.get_inventory(revision_id)
2493
2533
def is_shared(self):
2494
2534
"""Return True if this repository is flagged as a shared repository."""
2495
2535
raise NotImplementedError(self.is_shared)
2529
2569
return RevisionTree(self, Inventory(root_id=None),
2530
2570
_mod_revision.NULL_REVISION)
2532
inv = self.get_inventory(revision_id)
2572
inv = self.get_revision_inventory(revision_id)
2533
2573
return RevisionTree(self, inv, revision_id)
2535
2575
def revision_trees(self, revision_ids):
2588
2628
keys = tsort.topo_sort(parent_map)
2589
2629
return [None] + list(keys)
2591
def pack(self, hint=None, clean_obsolete_packs=False):
2631
def pack(self, hint=None):
2592
2632
"""Compress the data within the repository.
2594
2634
This operation only makes sense for some repository types. For other
2604
2644
obtained from the result of commit_write_group(). Out of
2605
2645
date hints are simply ignored, because concurrent operations
2606
2646
can obsolete them rapidly.
2608
:param clean_obsolete_packs: Clean obsolete packs immediately after
2612
2649
def get_transaction(self):
2628
2665
for ((revision_id,), parent_keys) in \
2629
2666
self.revisions.get_parent_map(query_keys).iteritems():
2630
2667
if parent_keys:
2631
result[revision_id] = tuple([parent_revid
2632
for (parent_revid,) in parent_keys])
2668
result[revision_id] = tuple(parent_revid
2669
for (parent_revid,) in parent_keys)
2634
2671
result[revision_id] = (_mod_revision.NULL_REVISION,)
2637
2674
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
2677
def get_graph(self, other_repository=None):
2650
2678
"""Return the graph walker for this repository format"""
2651
2679
parents_provider = self._make_parents_provider()
2746
2774
result.check(callback_refs)
2749
def _warn_if_deprecated(self, branch=None):
2777
def _warn_if_deprecated(self):
2750
2778
global _deprecation_warning_done
2751
2779
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
2781
_deprecation_warning_done = True
2782
warning("Format %s for %s is deprecated - please use 'bzr upgrade' to get better performance"
2783
% (self._format, self.bzrdir.transport.base))
2766
2785
def supports_rich_root(self):
2767
2786
return self._format.rich_root_data
3050
3069
pack_compresses = False
3051
3070
# Does the repository inventory storage understand references to trees?
3052
3071
supports_tree_reference = None
3053
# Is the format experimental ?
3054
experimental = False
3057
return "%s()" % self.__class__.__name__
3074
return "<%s>" % self.__class__.__name__
3059
3076
def __eq__(self, other):
3060
3077
# format objects are generally stateless
3075
3092
transport = a_bzrdir.get_repository_transport(None)
3076
format_string = transport.get_bytes("format")
3093
format_string = transport.get("format").read()
3077
3094
return format_registry.get(format_string)
3078
3095
except errors.NoSuchFile:
3079
3096
raise errors.NoRepositoryPresent(a_bzrdir)
3179
3196
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
3199
class MetaDirRepositoryFormat(RepositoryFormat):
3192
3200
"""Common base class for the new repositories using the metadir layout."""
3398
3406
:param revision_id: if None all content is copied, if NULL_REVISION no
3399
3407
content is copied.
3408
:param pb: optional progress bar to use for progress reports. If not
3409
provided a default one will be created.
3403
ui.ui_factory.warn_experimental_format_fetch(self)
3404
3412
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
3413
f = RepoFetcher(to_repository=self.target,
3411
3414
from_repository=self.source,
3412
3415
last_revision=revision_id,
3413
3416
fetch_spec=fetch_spec,
3414
find_ghosts=find_ghosts)
3417
pb=pb, find_ghosts=find_ghosts)
3416
3419
def _walk_to_common_revisions(self, revision_ids):
3417
3420
"""Walk out from revision_ids in source to revisions target has.
3813
3816
basis_id, delta, current_revision_id, parents_parents)
3814
3817
cache[current_revision_id] = parent_tree
3816
def _fetch_batch(self, revision_ids, basis_id, cache, a_graph=None):
3819
def _fetch_batch(self, revision_ids, basis_id, cache):
3817
3820
"""Fetch across a few revisions.
3819
3822
:param revision_ids: The revisions to copy
3820
3823
:param basis_id: The revision_id of a tree that must be in cache, used
3821
3824
as a basis for delta when no other base is available
3822
3825
: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
3826
:return: The revision_id of the last converted tree. The RevisionTree
3826
3827
for it will be in cache
3834
3835
pending_revisions = []
3835
3836
parent_map = self.source.get_parent_map(revision_ids)
3836
3837
self._fetch_parent_invs_for_stacking(parent_map, cache)
3837
self.source._safe_to_return_from_cache = True
3838
3838
for tree in self.source.revision_trees(revision_ids):
3839
3839
# Find a inventory delta for this revision.
3840
3840
# Find text entries that need to be copied, too.
3888
3888
pending_revisions.append(revision)
3889
3889
cache[current_revision_id] = tree
3890
3890
basis_id = current_revision_id
3891
self.source._safe_to_return_from_cache = False
3892
3891
# Copy file texts
3893
3892
from_texts = self.source.texts
3894
3893
to_texts = self.target.texts
3895
3894
if root_keys_to_create:
3896
root_stream = _mod_fetch._new_root_data_stream(
3895
from bzrlib.fetch import _new_root_data_stream
3896
root_stream = _new_root_data_stream(
3897
3897
root_keys_to_create, self._revision_id_to_root_id, parent_map,
3898
self.source, graph=a_graph)
3899
3899
to_texts.insert_record_stream(root_stream)
3900
3900
to_texts.insert_record_stream(from_texts.get_record_stream(
3901
3901
text_keys, self.target._format._fetch_order,
3958
3958
cache[basis_id] = basis_tree
3959
3959
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
3961
for offset in range(0, len(revision_ids), batch_size):
3968
3962
self.target.start_write_group()
3970
3964
pb.update('Transferring revisions', offset,
3971
3965
len(revision_ids))
3972
3966
batch = revision_ids[offset:offset+batch_size]
3973
basis_id = self._fetch_batch(batch, basis_id, cache,
3967
basis_id = self._fetch_batch(batch, basis_id, cache)
3976
self.source._safe_to_return_from_cache = False
3977
3969
self.target.abort_write_group()
3991
3983
"""See InterRepository.fetch()."""
3992
3984
if fetch_spec is not None:
3993
3985
raise AssertionError("Not implemented yet...")
3994
ui.ui_factory.warn_experimental_format_fetch(self)
3995
3986
if (not self.source.supports_rich_root()
3996
3987
and self.target.supports_rich_root()):
3997
3988
self._converting_to_rich_root = True
3998
3989
self._revision_id_to_root_id = {}
4000
3991
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
3992
revision_ids = self.target.search_missing_revision_ids(self.source,
4007
3993
revision_id, find_ghosts=find_ghosts).get_keys()
4008
3994
if not revision_ids:
4077
4063
:param to_convert: The disk object to convert.
4078
4064
:param pb: a progress bar to use for progress information.
4080
pb = ui.ui_factory.nested_progress_bar()
4083
4069
# this is only useful with metadir layouts - separated repo content.
4084
4070
# trigger an assertion if not such
4085
4071
repo._format.get_format_string()
4086
4072
self.repo_dir = repo.bzrdir
4087
pb.update('Moving repository to repository.backup')
4073
self.step('Moving repository to repository.backup')
4088
4074
self.repo_dir.transport.move('repository', 'repository.backup')
4089
4075
backup_transport = self.repo_dir.transport.clone('repository.backup')
4090
4076
repo._format.check_conversion_target(self.target_format)
4091
4077
self.source_repo = repo._format.open(self.repo_dir,
4093
4079
_override_transport=backup_transport)
4094
pb.update('Creating new repository')
4080
self.step('Creating new repository')
4095
4081
converted = self.target_format.initialize(self.repo_dir,
4096
4082
self.source_repo.is_shared())
4097
4083
converted.lock_write()
4099
pb.update('Copying content')
4085
self.step('Copying content')
4100
4086
self.source_repo.copy_content_into(converted)
4102
4088
converted.unlock()
4103
pb.update('Deleting old repository content')
4089
self.step('Deleting old repository content')
4104
4090
self.repo_dir.transport.delete_tree('repository.backup')
4105
4091
ui.ui_factory.note('repository converted')
4093
def step(self, message):
4094
"""Update the pb by a step."""
4096
self.pb.update(message, self.count, self.total)
4109
4099
_unescape_map = {
4330
4320
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
4322
missing_keys.update((prefix,) + key for key in
4340
4323
versioned_file.get_missing_compression_parent_keys())
4341
4324
except NotImplementedError:
4453
4436
fetching the inventory weave.
4455
4438
if self._rich_root_upgrade():
4456
return _mod_fetch.Inter1and2Helper(
4440
return bzrlib.fetch.Inter1and2Helper(
4457
4441
self.from_repository).generate_root_texts(revs)
4602
4586
def _get_convertable_inventory_stream(self, revision_ids,
4603
4587
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
4588
# The source is using CHKs, but the target either doesn't or it has a
4589
# different serializer. The StreamSink code expects to be able to
4608
4590
# convert on the target, so we need to put bytes-on-the-wire that can
4609
4591
# be converted. That means inventory deltas (if the remote is <1.19,
4610
4592
# RemoteStreamSink will fallback to VFS to insert the deltas).
added
removed
bzrlib/repository.py
