160
174
self._validate_unicode_text(value,
161
175
'revision property (%s)' % (key,))
177
def _ensure_fallback_inventories(self):
178
"""Ensure that appropriate inventories are available.
180
This only applies to repositories that are stacked, and is about
181
enusring the stacking invariants. Namely, that for any revision that is
182
present, we either have all of the file content, or we have the parent
183
inventory and the delta file content.
185
if not self.repository._fallback_repositories:
187
if not self.repository._format.supports_chks:
188
raise errors.BzrError("Cannot commit directly to a stacked branch"
189
" in pre-2a formats. See "
190
"https://bugs.launchpad.net/bzr/+bug/375013 for details.")
191
# This is a stacked repo, we need to make sure we have the parent
192
# inventories for the parents.
193
parent_keys = [(p,) for p in self.parents]
194
parent_map = self.repository.inventories._index.get_parent_map(parent_keys)
195
missing_parent_keys = set([pk for pk in parent_keys
196
if pk not in parent_map])
197
fallback_repos = list(reversed(self.repository._fallback_repositories))
198
missing_keys = [('inventories', pk[0])
199
for pk in missing_parent_keys]
201
while missing_keys and fallback_repos:
202
fallback_repo = fallback_repos.pop()
203
source = fallback_repo._get_source(self.repository._format)
204
sink = self.repository._get_sink()
205
stream = source.get_stream_for_missing_keys(missing_keys)
206
missing_keys = sink.insert_stream_without_locking(stream,
207
self.repository._format)
209
raise errors.BzrError('Unable to fill in parent inventories for a'
163
212
def commit(self, message):
164
213
"""Make the actual commit.
1524
1591
@needs_read_lock
1525
def search_missing_revision_ids(self, other, revision_id=None, find_ghosts=True):
1592
def search_missing_revision_ids(self, other,
1593
revision_id=symbol_versioning.DEPRECATED_PARAMETER,
1594
find_ghosts=True, revision_ids=None, if_present_ids=None):
1526
1595
"""Return the revision ids that other has that this does not.
1528
1597
These are returned in topological order.
1530
1599
revision_id: only return revision ids included by revision_id.
1601
if symbol_versioning.deprecated_passed(revision_id):
1602
symbol_versioning.warn(
1603
'search_missing_revision_ids(revision_id=...) was '
1604
'deprecated in 2.4. Use revision_ids=[...] instead.',
1605
DeprecationWarning, stacklevel=3)
1606
if revision_ids is not None:
1607
raise AssertionError(
1608
'revision_ids is mutually exclusive with revision_id')
1609
if revision_id is not None:
1610
revision_ids = [revision_id]
1532
1611
return InterRepository.get(other, self).search_missing_revision_ids(
1533
revision_id, find_ghosts)
1612
find_ghosts=find_ghosts, revision_ids=revision_ids,
1613
if_present_ids=if_present_ids)
1536
1616
def open(base):
1979
2056
w = self.inventories
1980
2057
pb = ui.ui_factory.nested_progress_bar()
1982
return self._find_text_key_references_from_xml_inventory_lines(
2059
return self._serializer._find_text_key_references(
1983
2060
w.iter_lines_added_or_present_in_keys(revision_keys, pb=pb))
1987
def _find_text_key_references_from_xml_inventory_lines(self,
1989
"""Core routine for extracting references to texts from inventories.
1991
This performs the translation of xml lines to revision ids.
1993
:param line_iterator: An iterator of lines, origin_version_id
1994
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1995
to whether they were referred to by the inventory of the
1996
revision_id that they contain. Note that if that revision_id was
1997
not part of the line_iterator's output then False will be given -
1998
even though it may actually refer to that key.
2000
if not self._serializer.support_altered_by_hack:
2001
raise AssertionError(
2002
"_find_text_key_references_from_xml_inventory_lines only "
2003
"supported for branches which store inventory as unnested xml"
2004
", not on %r" % self)
2007
# this code needs to read every new line in every inventory for the
2008
# inventories [revision_ids]. Seeing a line twice is ok. Seeing a line
2009
# not present in one of those inventories is unnecessary but not
2010
# harmful because we are filtering by the revision id marker in the
2011
# inventory lines : we only select file ids altered in one of those
2012
# revisions. We don't need to see all lines in the inventory because
2013
# only those added in an inventory in rev X can contain a revision=X
2015
unescape_revid_cache = {}
2016
unescape_fileid_cache = {}
2018
# jam 20061218 In a big fetch, this handles hundreds of thousands
2019
# of lines, so it has had a lot of inlining and optimizing done.
2020
# Sorry that it is a little bit messy.
2021
# Move several functions to be local variables, since this is a long
2023
search = self._file_ids_altered_regex.search
2024
unescape = _unescape_xml
2025
setdefault = result.setdefault
2026
for line, line_key in line_iterator:
2027
match = search(line)
2030
# One call to match.group() returning multiple items is quite a
2031
# bit faster than 2 calls to match.group() each returning 1
2032
file_id, revision_id = match.group('file_id', 'revision_id')
2034
# Inlining the cache lookups helps a lot when you make 170,000
2035
# lines and 350k ids, versus 8.4 unique ids.
2036
# Using a cache helps in 2 ways:
2037
# 1) Avoids unnecessary decoding calls
2038
# 2) Re-uses cached strings, which helps in future set and
2040
# (2) is enough that removing encoding entirely along with
2041
# the cache (so we are using plain strings) results in no
2042
# performance improvement.
2044
revision_id = unescape_revid_cache[revision_id]
2046
unescaped = unescape(revision_id)
2047
unescape_revid_cache[revision_id] = unescaped
2048
revision_id = unescaped
2050
# Note that unconditionally unescaping means that we deserialise
2051
# every fileid, which for general 'pull' is not great, but we don't
2052
# really want to have some many fulltexts that this matters anyway.
2055
file_id = unescape_fileid_cache[file_id]
2057
unescaped = unescape(file_id)
2058
unescape_fileid_cache[file_id] = unescaped
2061
key = (file_id, revision_id)
2062
setdefault(key, False)
2063
if revision_id == line_key[-1]:
2067
2064
def _inventory_xml_lines_for_keys(self, keys):
2068
2065
"""Get a line iterator of the sort needed for findind references.
2920
class RepositoryFormatRegistry(registry.FormatRegistry):
2921
"""Repository format registry."""
2923
def __init__(self, other_registry=None):
2924
super(RepositoryFormatRegistry, self).__init__(other_registry)
2925
self._extra_formats = []
2927
def register(self, format):
2928
"""Register a new repository format."""
2929
super(RepositoryFormatRegistry, self).register(
2930
format.get_format_string(), format)
2932
def remove(self, format):
2933
"""Remove a registered repository format."""
2934
super(RepositoryFormatRegistry, self).remove(
2935
format.get_format_string())
2937
def register_extra(self, format):
2938
"""Register a repository format that can not be used in a metadir.
2940
This is mainly useful to allow custom repository formats, such as older
2941
Bazaar formats and foreign formats, to be tested.
2943
self._extra_formats.append(registry._ObjectGetter(format))
2945
def remove_extra(self, format):
2946
"""Remove an extra repository format.
2948
self._extra_formats.remove(registry._ObjectGetter(format))
2950
def register_extra_lazy(self, module_name, member_name):
2951
"""Register a repository format lazily.
2953
self._extra_formats.append(
2954
registry._LazyObjectGetter(module_name, member_name))
2956
def get_default(self):
2957
"""Return the current default format."""
2958
from bzrlib import bzrdir
2959
return bzrdir.format_registry.make_bzrdir('default').repository_format
2961
def _get_extra(self):
2963
for getter in self._extra_formats:
2964
f = getter.get_obj()
2971
"""Return all repository formats, even those not usable in metadirs.
2973
return [self.get(k) for k in self.keys()] + self._get_extra()
2968
2976
network_format_registry = registry.FormatRegistry()
2969
2977
"""Registry of formats indexed by their network name.
3325
3348
'bzrlib.repofmt.pack_repo',
3326
3349
'RepositoryFormatKnitPack6RichRoot',
3351
format_registry.register_lazy(
3352
'Bazaar repository format 2a (needs bzr 1.16 or later)\n',
3353
'bzrlib.repofmt.groupcompress_repo',
3354
'RepositoryFormat2a',
3329
3357
# Development formats.
3330
# Obsolete but kept pending a CHK based subtree format.
3358
# Check their docstrings to see if/when they are obsolete.
3331
3359
format_registry.register_lazy(
3332
3360
("Bazaar development format 2 with subtree support "
3333
3361
"(needs bzr.dev from before 1.8)\n"),
3334
3362
'bzrlib.repofmt.pack_repo',
3335
3363
'RepositoryFormatPackDevelopment2Subtree',
3338
# 1.14->1.16 go below here
3339
format_registry.register_lazy(
3340
'Bazaar development format - group compression and chk inventory'
3341
' (needs bzr.dev from 1.14)\n',
3342
'bzrlib.repofmt.groupcompress_repo',
3343
'RepositoryFormatCHK1',
3346
format_registry.register_lazy(
3347
'Bazaar development format - chk repository with bencode revision '
3348
'serialization (needs bzr.dev from 1.16)\n',
3349
'bzrlib.repofmt.groupcompress_repo',
3350
'RepositoryFormatCHK2',
3352
format_registry.register_lazy(
3353
'Bazaar repository format 2a (needs bzr 1.16 or later)\n',
3354
'bzrlib.repofmt.groupcompress_repo',
3355
'RepositoryFormat2a',
3365
format_registry.register_lazy(
3366
'Bazaar development format 8\n',
3367
'bzrlib.repofmt.groupcompress_repo',
3368
'RepositoryFormat2aSubtree',
3466
3483
return searcher.get_result()
3468
3485
@needs_read_lock
3469
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
3486
def search_missing_revision_ids(self,
3487
revision_id=symbol_versioning.DEPRECATED_PARAMETER,
3488
find_ghosts=True, revision_ids=None, if_present_ids=None):
3470
3489
"""Return the revision ids that source has that target does not.
3472
3491
:param revision_id: only return revision ids included by this
3493
:param revision_ids: return revision ids included by these
3494
revision_ids. NoSuchRevision will be raised if any of these
3495
revisions are not present.
3496
:param if_present_ids: like revision_ids, but will not cause
3497
NoSuchRevision if any of these are absent, instead they will simply
3498
not be in the result. This is useful for e.g. finding revisions
3499
to fetch for tags, which may reference absent revisions.
3474
3500
:param find_ghosts: If True find missing revisions in deep history
3475
3501
rather than just finding the surface difference.
3476
3502
:return: A bzrlib.graph.SearchResult.
3504
if symbol_versioning.deprecated_passed(revision_id):
3505
symbol_versioning.warn(
3506
'search_missing_revision_ids(revision_id=...) was '
3507
'deprecated in 2.4. Use revision_ids=[...] instead.',
3508
DeprecationWarning, stacklevel=2)
3509
if revision_ids is not None:
3510
raise AssertionError(
3511
'revision_ids is mutually exclusive with revision_id')
3512
if revision_id is not None:
3513
revision_ids = [revision_id]
3478
3515
# stop searching at found target revisions.
3479
if not find_ghosts and revision_id is not None:
3480
return self._walk_to_common_revisions([revision_id])
3516
if not find_ghosts and (revision_ids is not None or if_present_ids is
3518
return self._walk_to_common_revisions(revision_ids,
3519
if_present_ids=if_present_ids)
3481
3520
# generic, possibly worst case, slow code path.
3482
3521
target_ids = set(self.target.all_revision_ids())
3483
if revision_id is not None:
3484
source_ids = self.source.get_ancestry(revision_id)
3485
if source_ids[0] is not None:
3486
raise AssertionError()
3489
source_ids = self.source.all_revision_ids()
3522
source_ids = self._present_source_revisions_for(
3523
revision_ids, if_present_ids)
3490
3524
result_set = set(source_ids).difference(target_ids)
3491
3525
return self.source.revision_ids_to_search_result(result_set)
3527
def _present_source_revisions_for(self, revision_ids, if_present_ids=None):
3528
"""Returns set of all revisions in ancestry of revision_ids present in
3531
:param revision_ids: if None, all revisions in source are returned.
3532
:param if_present_ids: like revision_ids, but if any/all of these are
3533
absent no error is raised.
3535
if revision_ids is not None or if_present_ids is not None:
3536
# First, ensure all specified revisions exist. Callers expect
3537
# NoSuchRevision when they pass absent revision_ids here.
3538
if revision_ids is None:
3539
revision_ids = set()
3540
if if_present_ids is None:
3541
if_present_ids = set()
3542
revision_ids = set(revision_ids)
3543
if_present_ids = set(if_present_ids)
3544
all_wanted_ids = revision_ids.union(if_present_ids)
3545
graph = self.source.get_graph()
3546
present_revs = set(graph.get_parent_map(all_wanted_ids))
3547
missing = revision_ids.difference(present_revs)
3549
raise errors.NoSuchRevision(self.source, missing.pop())
3550
found_ids = all_wanted_ids.intersection(present_revs)
3551
source_ids = [rev_id for (rev_id, parents) in
3552
graph.iter_ancestry(found_ids)
3553
if rev_id != _mod_revision.NULL_REVISION
3554
and parents is not None]
3556
source_ids = self.source.all_revision_ids()
3557
return set(source_ids)
3494
3560
def _same_model(source, target):
3495
3561
"""True if source and target have the same data representation.
3536
3602
return InterRepository._same_model(source, target)
3539
class InterWeaveRepo(InterSameDataRepository):
3540
"""Optimised code paths between Weave based repositories.
3542
This should be in bzrlib/repofmt/weaverepo.py but we have not yet
3543
implemented lazy inter-object optimisation.
3547
def _get_repo_format_to_test(self):
3548
from bzrlib.repofmt import weaverepo
3549
return weaverepo.RepositoryFormat7()
3552
def is_compatible(source, target):
3553
"""Be compatible with known Weave formats.
3555
We don't test for the stores being of specific types because that
3556
could lead to confusing results, and there is no need to be
3559
from bzrlib.repofmt.weaverepo import (
3565
return (isinstance(source._format, (RepositoryFormat5,
3567
RepositoryFormat7)) and
3568
isinstance(target._format, (RepositoryFormat5,
3570
RepositoryFormat7)))
3571
except AttributeError:
3575
def copy_content(self, revision_id=None):
3576
"""See InterRepository.copy_content()."""
3577
# weave specific optimised path:
3579
self.target.set_make_working_trees(self.source.make_working_trees())
3580
except (errors.RepositoryUpgradeRequired, NotImplemented):
3582
# FIXME do not peek!
3583
if self.source._transport.listable():
3584
pb = ui.ui_factory.nested_progress_bar()
3586
self.target.texts.insert_record_stream(
3587
self.source.texts.get_record_stream(
3588
self.source.texts.keys(), 'topological', False))
3589
pb.update('Copying inventory', 0, 1)
3590
self.target.inventories.insert_record_stream(
3591
self.source.inventories.get_record_stream(
3592
self.source.inventories.keys(), 'topological', False))
3593
self.target.signatures.insert_record_stream(
3594
self.source.signatures.get_record_stream(
3595
self.source.signatures.keys(),
3597
self.target.revisions.insert_record_stream(
3598
self.source.revisions.get_record_stream(
3599
self.source.revisions.keys(),
3600
'topological', True))
3604
self.target.fetch(self.source, revision_id=revision_id)
3607
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
3608
"""See InterRepository.missing_revision_ids()."""
3609
# we want all revisions to satisfy revision_id in source.
3610
# but we don't want to stat every file here and there.
3611
# we want then, all revisions other needs to satisfy revision_id
3612
# checked, but not those that we have locally.
3613
# so the first thing is to get a subset of the revisions to
3614
# satisfy revision_id in source, and then eliminate those that
3615
# we do already have.
3616
# this is slow on high latency connection to self, but as this
3617
# disk format scales terribly for push anyway due to rewriting
3618
# inventory.weave, this is considered acceptable.
3620
if revision_id is not None:
3621
source_ids = self.source.get_ancestry(revision_id)
3622
if source_ids[0] is not None:
3623
raise AssertionError()
3626
source_ids = self.source._all_possible_ids()
3627
source_ids_set = set(source_ids)
3628
# source_ids is the worst possible case we may need to pull.
3629
# now we want to filter source_ids against what we actually
3630
# have in target, but don't try to check for existence where we know
3631
# we do not have a revision as that would be pointless.
3632
target_ids = set(self.target._all_possible_ids())
3633
possibly_present_revisions = target_ids.intersection(source_ids_set)
3634
actually_present_revisions = set(
3635
self.target._eliminate_revisions_not_present(possibly_present_revisions))
3636
required_revisions = source_ids_set.difference(actually_present_revisions)
3637
if revision_id is not None:
3638
# we used get_ancestry to determine source_ids then we are assured all
3639
# revisions referenced are present as they are installed in topological order.
3640
# and the tip revision was validated by get_ancestry.
3641
result_set = required_revisions
3643
# if we just grabbed the possibly available ids, then
3644
# we only have an estimate of whats available and need to validate
3645
# that against the revision records.
3647
self.source._eliminate_revisions_not_present(required_revisions))
3648
return self.source.revision_ids_to_search_result(result_set)
3651
class InterKnitRepo(InterSameDataRepository):
3652
"""Optimised code paths between Knit based repositories."""
3655
def _get_repo_format_to_test(self):
3656
from bzrlib.repofmt import knitrepo
3657
return knitrepo.RepositoryFormatKnit1()
3660
def is_compatible(source, target):
3661
"""Be compatible with known Knit formats.
3663
We don't test for the stores being of specific types because that
3664
could lead to confusing results, and there is no need to be
3667
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit
3669
are_knits = (isinstance(source._format, RepositoryFormatKnit) and
3670
isinstance(target._format, RepositoryFormatKnit))
3671
except AttributeError:
3673
return are_knits and InterRepository._same_model(source, target)
3676
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
3677
"""See InterRepository.missing_revision_ids()."""
3678
if revision_id is not None:
3679
source_ids = self.source.get_ancestry(revision_id)
3680
if source_ids[0] is not None:
3681
raise AssertionError()
3684
source_ids = self.source.all_revision_ids()
3685
source_ids_set = set(source_ids)
3686
# source_ids is the worst possible case we may need to pull.
3687
# now we want to filter source_ids against what we actually
3688
# have in target, but don't try to check for existence where we know
3689
# we do not have a revision as that would be pointless.
3690
target_ids = set(self.target.all_revision_ids())
3691
possibly_present_revisions = target_ids.intersection(source_ids_set)
3692
actually_present_revisions = set(
3693
self.target._eliminate_revisions_not_present(possibly_present_revisions))
3694
required_revisions = source_ids_set.difference(actually_present_revisions)
3695
if revision_id is not None:
3696
# we used get_ancestry to determine source_ids then we are assured all
3697
# revisions referenced are present as they are installed in topological order.
3698
# and the tip revision was validated by get_ancestry.
3699
result_set = required_revisions
3701
# if we just grabbed the possibly available ids, then
3702
# we only have an estimate of whats available and need to validate
3703
# that against the revision records.
3705
self.source._eliminate_revisions_not_present(required_revisions))
3706
return self.source.revision_ids_to_search_result(result_set)
3709
3605
class InterDifferingSerializer(InterRepository):
4249
4114
is_resume = False
4251
4116
# locked_insert_stream performs a commit|suspend.
4252
return self._locked_insert_stream(stream, src_format, is_resume)
4117
missing_keys = self.insert_stream_without_locking(stream,
4118
src_format, is_resume)
4120
# suspend the write group and tell the caller what we is
4121
# missing. We know we can suspend or else we would not have
4122
# entered this code path. (All repositories that can handle
4123
# missing keys can handle suspending a write group).
4124
write_group_tokens = self.target_repo.suspend_write_group()
4125
return write_group_tokens, missing_keys
4126
hint = self.target_repo.commit_write_group()
4127
to_serializer = self.target_repo._format._serializer
4128
src_serializer = src_format._serializer
4129
if (to_serializer != src_serializer and
4130
self.target_repo._format.pack_compresses):
4131
self.target_repo.pack(hint=hint)
4254
4134
self.target_repo.abort_write_group(suppress_errors=True)
4257
4137
self.target_repo.unlock()
4259
def _locked_insert_stream(self, stream, src_format, is_resume):
4139
def insert_stream_without_locking(self, stream, src_format,
4141
"""Insert a stream's content into the target repository.
4143
This assumes that you already have a locked repository and an active
4146
:param src_format: a bzr repository format.
4147
:param is_resume: Passed down to get_missing_parent_inventories to
4148
indicate if we should be checking for missing texts at the same
4151
:return: A set of keys that are missing.
4153
if not self.target_repo.is_write_locked():
4154
raise errors.ObjectNotLocked(self)
4155
if not self.target_repo.is_in_write_group():
4156
raise errors.BzrError('you must already be in a write group')
4260
4157
to_serializer = self.target_repo._format._serializer
4261
4158
src_serializer = src_format._serializer
4262
4159
new_pack = None