1
# Copyright (C) 2005, 2006 Canonical Ltd
3
# This program is free software; you can redistribute it and/or modify
4
# it under the terms of the GNU General Public License as published by
5
# the Free Software Foundation; either version 2 of the License, or
6
# (at your option) any later version.
8
# This program is distributed in the hope that it will be useful,
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11
# GNU General Public License for more details.
13
# You should have received a copy of the GNU General Public License
14
# along with this program; if not, write to the Free Software
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
17
from copy import deepcopy
18
from cStringIO import StringIO
19
from unittest import TestSuite
20
import xml.sax.saxutils
22
import bzrlib.bzrdir as bzrdir
23
from bzrlib.decorators import needs_read_lock, needs_write_lock
24
from bzrlib.errors import InvalidRevisionId
25
from bzrlib.lockable_files import LockableFiles, TransportLock
26
from bzrlib.lockdir import LockDir
27
from bzrlib.osutils import safe_unicode
28
from bzrlib.revision import NULL_REVISION
29
import bzrlib.errors as errors
30
import bzrlib.gpg as gpg
31
from bzrlib.store import copy_all
32
from bzrlib.store.weave import WeaveStore
33
from bzrlib.store.text import TextStore
34
from bzrlib.symbol_versioning import *
35
from bzrlib.trace import mutter
36
from bzrlib.tree import RevisionTree
37
from bzrlib.testament import Testament
38
from bzrlib.tree import EmptyTree
43
class Repository(object):
44
"""Repository holding history for one or more branches.
46
The repository holds and retrieves historical information including
47
revisions and file history. It's normally accessed only by the Branch,
48
which views a particular line of development through that history.
50
The Repository builds on top of Stores and a Transport, which respectively
51
describe the disk data format and the way of accessing the (possibly
56
def add_inventory(self, revid, inv, parents):
57
"""Add the inventory inv to the repository as revid.
59
:param parents: The revision ids of the parents that revid
60
is known to have and are in the repository already.
62
returns the sha1 of the serialized inventory.
64
inv_text = bzrlib.xml5.serializer_v5.write_inventory_to_string(inv)
65
inv_sha1 = bzrlib.osutils.sha_string(inv_text)
66
self.control_weaves.add_text('inventory', revid,
67
bzrlib.osutils.split_lines(inv_text), parents,
68
self.get_transaction())
72
def add_revision(self, rev_id, rev, inv=None, config=None):
73
"""Add rev to the revision store as rev_id.
75
:param rev_id: the revision id to use.
76
:param rev: The revision object.
77
:param inv: The inventory for the revision. if None, it will be looked
78
up in the inventory storer
79
:param config: If None no digital signature will be created.
80
If supplied its signature_needed method will be used
81
to determine if a signature should be made.
83
if config is not None and config.signature_needed():
85
inv = self.get_inventory(rev_id)
86
plaintext = Testament(rev, inv).as_short_text()
87
self.store_revision_signature(
88
gpg.GPGStrategy(config), plaintext, rev_id)
89
if not rev_id in self.get_inventory_weave():
91
raise errors.WeaveRevisionNotPresent(rev_id,
92
self.get_inventory_weave())
94
# yes, this is not suitable for adding with ghosts.
95
self.add_inventory(rev_id, inv, rev.parent_ids)
98
bzrlib.xml5.serializer_v5.write_revision(rev, rev_tmp)
100
self.revision_store.add(rev_tmp, rev_id)
101
mutter('added revision_id {%s}', rev_id)
104
def _all_possible_ids(self):
105
"""Return all the possible revisions that we could find."""
106
return self.get_inventory_weave().names()
109
def all_revision_ids(self):
110
"""Returns a list of all the revision ids in the repository.
112
These are in as much topological order as the underlying store can
113
present: for weaves ghosts may lead to a lack of correctness until
114
the reweave updates the parents list.
116
result = self._all_possible_ids()
117
return self._eliminate_revisions_not_present(result)
120
def _eliminate_revisions_not_present(self, revision_ids):
121
"""Check every revision id in revision_ids to see if we have it.
123
Returns a set of the present revisions.
126
for id in revision_ids:
127
if self.has_revision(id):
132
def create(a_bzrdir):
133
"""Construct the current default format repository in a_bzrdir."""
134
return RepositoryFormat.get_default_format().initialize(a_bzrdir)
136
def __init__(self, _format, a_bzrdir, control_files, revision_store):
137
"""instantiate a Repository.
139
:param _format: The format of the repository on disk.
140
:param a_bzrdir: The BzrDir of the repository.
142
In the future we will have a single api for all stores for
143
getting file texts, inventories and revisions, then
144
this construct will accept instances of those things.
146
object.__init__(self)
147
self._format = _format
148
# the following are part of the public API for Repository:
149
self.bzrdir = a_bzrdir
150
self.control_files = control_files
151
self.revision_store = revision_store
153
def lock_write(self):
154
self.control_files.lock_write()
157
self.control_files.lock_read()
160
return self.control_files.is_locked()
163
def missing_revision_ids(self, other, revision_id=None):
164
"""Return the revision ids that other has that this does not.
166
These are returned in topological order.
168
revision_id: only return revision ids included by revision_id.
170
return InterRepository.get(other, self).missing_revision_ids(revision_id)
174
"""Open the repository rooted at base.
176
For instance, if the repository is at URL/.bzr/repository,
177
Repository.open(URL) -> a Repository instance.
179
control = bzrlib.bzrdir.BzrDir.open(base)
180
return control.open_repository()
182
def copy_content_into(self, destination, revision_id=None, basis=None):
183
"""Make a complete copy of the content in self into destination.
185
This is a destructive operation! Do not use it on existing
188
return InterRepository.get(self, destination).copy_content(revision_id, basis)
190
def fetch(self, source, revision_id=None, pb=None):
191
"""Fetch the content required to construct revision_id from source.
193
If revision_id is None all content is copied.
195
return InterRepository.get(source, self).fetch(revision_id=revision_id,
199
self.control_files.unlock()
202
def clone(self, a_bzrdir, revision_id=None, basis=None):
203
"""Clone this repository into a_bzrdir using the current format.
205
Currently no check is made that the format of this repository and
206
the bzrdir format are compatible. FIXME RBC 20060201.
208
if not isinstance(a_bzrdir._format, self.bzrdir._format.__class__):
209
# use target default format.
210
result = a_bzrdir.create_repository()
211
# FIXME RBC 20060209 split out the repository type to avoid this check ?
212
elif isinstance(a_bzrdir._format,
213
(bzrlib.bzrdir.BzrDirFormat4,
214
bzrlib.bzrdir.BzrDirFormat5,
215
bzrlib.bzrdir.BzrDirFormat6)):
216
result = a_bzrdir.open_repository()
218
result = self._format.initialize(a_bzrdir, shared=self.is_shared())
219
self.copy_content_into(result, revision_id, basis)
222
def has_revision(self, revision_id):
223
"""True if this branch has a copy of the revision.
225
This does not necessarily imply the revision is merge
226
or on the mainline."""
227
return (revision_id is None
228
or self.revision_store.has_id(revision_id))
231
def get_revision_xml_file(self, revision_id):
232
"""Return XML file object for revision object."""
233
if not revision_id or not isinstance(revision_id, basestring):
234
raise InvalidRevisionId(revision_id=revision_id, branch=self)
236
return self.revision_store.get(revision_id)
237
except (IndexError, KeyError):
238
raise bzrlib.errors.NoSuchRevision(self, revision_id)
241
def get_revision_xml(self, revision_id):
242
return self.get_revision_xml_file(revision_id).read()
245
def get_revision_reconcile(self, revision_id):
246
"""'reconcile' helper routine that allows access to a revision always.
248
This variant of get_revision does not cross check the weave graph
249
against the revision one as get_revision does: but it should only
250
be used by reconcile, or reconcile-alike commands that are correcting
251
or testing the revision graph.
253
xml_file = self.get_revision_xml_file(revision_id)
256
r = bzrlib.xml5.serializer_v5.read_revision(xml_file)
257
except SyntaxError, e:
258
raise bzrlib.errors.BzrError('failed to unpack revision_xml',
262
assert r.revision_id == revision_id
266
def get_revision(self, revision_id):
267
"""Return the Revision object for a named revision"""
268
r = self.get_revision_reconcile(revision_id)
269
# weave corruption can lead to absent revision markers that should be
271
# the following test is reasonably cheap (it needs a single weave read)
272
# and the weave is cached in read transactions. In write transactions
273
# it is not cached but typically we only read a small number of
274
# revisions. For knits when they are introduced we will probably want
275
# to ensure that caching write transactions are in use.
276
inv = self.get_inventory_weave()
277
self._check_revision_parents(r, inv)
280
def _check_revision_parents(self, revision, inventory):
281
"""Private to Repository and Fetch.
283
This checks the parentage of revision in an inventory weave for
284
consistency and is only applicable to inventory-weave-for-ancestry
285
using repository formats & fetchers.
287
weave_parents = inventory.parent_names(revision.revision_id)
288
weave_names = inventory.names()
289
for parent_id in revision.parent_ids:
290
if parent_id in weave_names:
291
# this parent must not be a ghost.
292
if not parent_id in weave_parents:
294
raise errors.CorruptRepository(self)
297
def get_revision_sha1(self, revision_id):
298
"""Hash the stored value of a revision, and return it."""
299
# In the future, revision entries will be signed. At that
300
# point, it is probably best *not* to include the signature
301
# in the revision hash. Because that lets you re-sign
302
# the revision, (add signatures/remove signatures) and still
303
# have all hash pointers stay consistent.
304
# But for now, just hash the contents.
305
return bzrlib.osutils.sha_file(self.get_revision_xml_file(revision_id))
308
def store_revision_signature(self, gpg_strategy, plaintext, revision_id):
309
self.revision_store.add(StringIO(gpg_strategy.sign(plaintext)),
312
def fileid_involved_between_revs(self, from_revid, to_revid):
313
"""Find file_id(s) which are involved in the changes between revisions.
315
This determines the set of revisions which are involved, and then
316
finds all file ids affected by those revisions.
318
# TODO: jam 20060119 This code assumes that w.inclusions will
319
# always be correct. But because of the presence of ghosts
320
# it is possible to be wrong.
321
# One specific example from Robert Collins:
322
# Two branches, with revisions ABC, and AD
323
# C is a ghost merge of D.
324
# Inclusions doesn't recognize D as an ancestor.
325
# If D is ever merged in the future, the weave
326
# won't be fixed, because AD never saw revision C
327
# to cause a conflict which would force a reweave.
328
w = self.get_inventory_weave()
329
from_set = set(w.inclusions([w.lookup(from_revid)]))
330
to_set = set(w.inclusions([w.lookup(to_revid)]))
331
included = to_set.difference(from_set)
332
changed = map(w.idx_to_name, included)
333
return self._fileid_involved_by_set(changed)
335
def fileid_involved(self, last_revid=None):
336
"""Find all file_ids modified in the ancestry of last_revid.
338
:param last_revid: If None, last_revision() will be used.
340
w = self.get_inventory_weave()
342
changed = set(w._names)
344
included = w.inclusions([w.lookup(last_revid)])
345
changed = map(w.idx_to_name, included)
346
return self._fileid_involved_by_set(changed)
348
def fileid_involved_by_set(self, changes):
349
"""Find all file_ids modified by the set of revisions passed in.
351
:param changes: A set() of revision ids
353
# TODO: jam 20060119 This line does *nothing*, remove it.
354
# or better yet, change _fileid_involved_by_set so
355
# that it takes the inventory weave, rather than
356
# pulling it out by itself.
357
return self._fileid_involved_by_set(changes)
359
def _fileid_involved_by_set(self, changes):
360
"""Find the set of file-ids affected by the set of revisions.
362
:param changes: A set() of revision ids.
363
:return: A set() of file ids.
365
This peaks at the Weave, interpreting each line, looking to
366
see if it mentions one of the revisions. And if so, includes
367
the file id mentioned.
368
This expects both the Weave format, and the serialization
369
to have a single line per file/directory, and to have
370
fileid="" and revision="" on that line.
372
assert isinstance(self._format, (RepositoryFormat5,
375
RepositoryFormatKnit1)), \
376
"fileid_involved only supported for branches which store inventory as unnested xml"
378
w = self.get_inventory_weave()
380
for line in w._weave:
382
# it is ugly, but it is due to the weave structure
383
if not isinstance(line, basestring): continue
385
start = line.find('file_id="')+9
386
if start < 9: continue
387
end = line.find('"', start)
389
file_id = xml.sax.saxutils.unescape(line[start:end])
391
# check if file_id is already present
392
if file_id in file_ids: continue
394
start = line.find('revision="')+10
395
if start < 10: continue
396
end = line.find('"', start)
398
revision_id = xml.sax.saxutils.unescape(line[start:end])
400
if revision_id in changes:
401
file_ids.add(file_id)
405
def get_inventory_weave(self):
406
return self.control_weaves.get_weave('inventory',
407
self.get_transaction())
410
def get_inventory(self, revision_id):
411
"""Get Inventory object by hash."""
412
xml = self.get_inventory_xml(revision_id)
413
return bzrlib.xml5.serializer_v5.read_inventory_from_string(xml)
416
def get_inventory_xml(self, revision_id):
417
"""Get inventory XML as a file object."""
419
assert isinstance(revision_id, basestring), type(revision_id)
420
iw = self.get_inventory_weave()
421
return iw.get_text(iw.lookup(revision_id))
423
raise bzrlib.errors.HistoryMissing(self, 'inventory', revision_id)
426
def get_inventory_sha1(self, revision_id):
427
"""Return the sha1 hash of the inventory entry
429
return self.get_revision(revision_id).inventory_sha1
432
def get_revision_graph(self, revision_id=None):
433
"""Return a dictionary containing the revision graph.
435
:return: a dictionary of revision_id->revision_parents_list.
437
weave = self.get_inventory_weave()
438
all_revisions = self._eliminate_revisions_not_present(weave.names())
439
entire_graph = dict([(node, weave.parent_names(node)) for
440
node in all_revisions])
441
if revision_id is None:
443
elif revision_id not in entire_graph:
444
raise errors.NoSuchRevision(self, revision_id)
446
# add what can be reached from revision_id
448
pending = set([revision_id])
449
while len(pending) > 0:
451
result[node] = entire_graph[node]
452
for revision_id in result[node]:
453
if revision_id not in result:
454
pending.add(revision_id)
458
def get_revision_inventory(self, revision_id):
459
"""Return inventory of a past revision."""
460
# TODO: Unify this with get_inventory()
461
# bzr 0.0.6 and later imposes the constraint that the inventory_id
462
# must be the same as its revision, so this is trivial.
463
if revision_id is None:
464
# This does not make sense: if there is no revision,
465
# then it is the current tree inventory surely ?!
466
# and thus get_root_id() is something that looks at the last
467
# commit on the branch, and the get_root_id is an inventory check.
468
raise NotImplementedError
469
# return Inventory(self.get_root_id())
471
return self.get_inventory(revision_id)
475
"""Return True if this repository is flagged as a shared repository."""
476
# FIXME format 4-6 cannot be shared, this is technically faulty.
477
return self.control_files._transport.has('shared-storage')
480
def revision_tree(self, revision_id):
481
"""Return Tree for a revision on this branch.
483
`revision_id` may be None for the null revision, in which case
484
an `EmptyTree` is returned."""
485
# TODO: refactor this to use an existing revision object
486
# so we don't need to read it in twice.
487
if revision_id is None or revision_id == NULL_REVISION:
490
inv = self.get_revision_inventory(revision_id)
491
return RevisionTree(self, inv, revision_id)
494
def get_ancestry(self, revision_id):
495
"""Return a list of revision-ids integrated by a revision.
497
This is topologically sorted.
499
if revision_id is None:
501
if not self.has_revision(revision_id):
502
raise errors.NoSuchRevision(self, revision_id)
503
w = self.get_inventory_weave()
504
return [None] + map(w.idx_to_name,
505
w.inclusions([w.lookup(revision_id)]))
508
def print_file(self, file, revision_id):
509
"""Print `file` to stdout.
511
FIXME RBC 20060125 as John Meinel points out this is a bad api
512
- it writes to stdout, it assumes that that is valid etc. Fix
513
by creating a new more flexible convenience function.
515
tree = self.revision_tree(revision_id)
516
# use inventory as it was in that revision
517
file_id = tree.inventory.path2id(file)
519
raise BzrError("%r is not present in revision %s" % (file, revno))
521
revno = self.revision_id_to_revno(revision_id)
522
except errors.NoSuchRevision:
523
# TODO: This should not be BzrError,
524
# but NoSuchFile doesn't fit either
525
raise BzrError('%r is not present in revision %s'
526
% (file, revision_id))
528
raise BzrError('%r is not present in revision %s'
530
tree.print_file(file_id)
532
def get_transaction(self):
533
return self.control_files.get_transaction()
535
def revision_parents(self, revid):
536
return self.get_inventory_weave().parent_names(revid)
539
def set_make_working_trees(self, new_value):
540
"""Set the policy flag for making working trees when creating branches.
542
This only applies to branches that use this repository.
544
The default is 'True'.
545
:param new_value: True to restore the default, False to disable making
548
# FIXME: split out into a new class/strategy ?
549
if isinstance(self._format, (RepositoryFormat4,
552
raise NotImplementedError(self.set_make_working_trees)
555
self.control_files._transport.delete('no-working-trees')
556
except errors.NoSuchFile:
559
self.control_files.put_utf8('no-working-trees', '')
561
def make_working_trees(self):
562
"""Returns the policy for making working trees on new branches."""
563
# FIXME: split out into a new class/strategy ?
564
if isinstance(self._format, (RepositoryFormat4,
568
return not self.control_files._transport.has('no-working-trees')
571
def sign_revision(self, revision_id, gpg_strategy):
572
plaintext = Testament.from_revision(self, revision_id).as_short_text()
573
self.store_revision_signature(gpg_strategy, plaintext, revision_id)
576
class AllInOneRepository(Repository):
577
"""Legacy support - the repository behaviour for all-in-one branches."""
579
def __init__(self, _format, a_bzrdir, revision_store):
580
# we reuse one control files instance.
581
dir_mode = a_bzrdir._control_files._dir_mode
582
file_mode = a_bzrdir._control_files._file_mode
584
def get_weave(name, prefixed=False):
586
name = safe_unicode(name)
589
relpath = a_bzrdir._control_files._escape(name)
590
weave_transport = a_bzrdir._control_files._transport.clone(relpath)
591
ws = WeaveStore(weave_transport, prefixed=prefixed,
594
if a_bzrdir._control_files._transport.should_cache():
595
ws.enable_cache = True
598
def get_store(name, compressed=True, prefixed=False):
599
# FIXME: This approach of assuming stores are all entirely compressed
600
# or entirely uncompressed is tidy, but breaks upgrade from
601
# some existing branches where there's a mixture; we probably
602
# still want the option to look for both.
603
relpath = a_bzrdir._control_files._escape(name)
604
store = TextStore(a_bzrdir._control_files._transport.clone(relpath),
605
prefixed=prefixed, compressed=compressed,
608
#if self._transport.should_cache():
609
# cache_path = os.path.join(self.cache_root, name)
610
# os.mkdir(cache_path)
611
# store = bzrlib.store.CachedStore(store, cache_path)
614
# not broken out yet because the controlweaves|inventory_store
615
# and text_store | weave_store bits are still different.
616
if isinstance(_format, RepositoryFormat4):
617
self.inventory_store = get_store('inventory-store')
618
self.text_store = get_store('text-store')
619
elif isinstance(_format, RepositoryFormat5):
620
self.control_weaves = get_weave('')
621
self.weave_store = get_weave('weaves')
622
elif isinstance(_format, RepositoryFormat6):
623
self.control_weaves = get_weave('')
624
self.weave_store = get_weave('weaves', prefixed=True)
626
raise errors.BzrError('unreachable code: unexpected repository'
628
revision_store.register_suffix('sig')
629
super(AllInOneRepository, self).__init__(_format, a_bzrdir, a_bzrdir._control_files, revision_store)
632
class MetaDirRepository(Repository):
633
"""Repositories in the new meta-dir layout."""
635
def __init__(self, _format, a_bzrdir, control_files, revision_store):
636
super(MetaDirRepository, self).__init__(_format,
641
dir_mode = self.control_files._dir_mode
642
file_mode = self.control_files._file_mode
644
def get_weave(name, prefixed=False):
646
name = safe_unicode(name)
649
relpath = self.control_files._escape(name)
650
weave_transport = self.control_files._transport.clone(relpath)
651
ws = WeaveStore(weave_transport, prefixed=prefixed,
654
if self.control_files._transport.should_cache():
655
ws.enable_cache = True
658
if isinstance(self._format, RepositoryFormat7):
659
self.control_weaves = get_weave('')
660
self.weave_store = get_weave('weaves', prefixed=True)
661
elif isinstance(self._format, RepositoryFormatKnit1):
662
self.control_weaves = get_weave('')
663
self.weave_store = get_weave('knits', prefixed=True)
665
raise errors.BzrError('unreachable code: unexpected repository'
669
class RepositoryFormat(object):
670
"""A repository format.
672
Formats provide three things:
673
* An initialization routine to construct repository data on disk.
674
* a format string which is used when the BzrDir supports versioned
676
* an open routine which returns a Repository instance.
678
Formats are placed in an dict by their format string for reference
679
during opening. These should be subclasses of RepositoryFormat
682
Once a format is deprecated, just deprecate the initialize and open
683
methods on the format class. Do not deprecate the object, as the
684
object will be created every system load.
686
Common instance attributes:
687
_matchingbzrdir - the bzrdir format that the repository format was
688
originally written to work with. This can be used if manually
689
constructing a bzrdir and repository, or more commonly for test suite
693
_default_format = None
694
"""The default format used for new repositories."""
697
"""The known formats."""
700
def find_format(klass, a_bzrdir):
701
"""Return the format for the repository object in a_bzrdir."""
703
transport = a_bzrdir.get_repository_transport(None)
704
format_string = transport.get("format").read()
705
return klass._formats[format_string]
706
except errors.NoSuchFile:
707
raise errors.NoRepositoryPresent(a_bzrdir)
709
raise errors.UnknownFormatError(format_string)
712
def get_default_format(klass):
713
"""Return the current default format."""
714
return klass._default_format
716
def get_format_string(self):
717
"""Return the ASCII format string that identifies this format.
719
Note that in pre format ?? repositories the format string is
720
not permitted nor written to disk.
722
raise NotImplementedError(self.get_format_string)
724
def _get_revision_store(self, repo_transport, control_files):
725
"""Return the revision store object for this a_bzrdir."""
726
raise NotImplementedError(self._get_revision_store)
728
def _get_rev_store(self,
734
"""Common logic for getting a revision store for a repository.
736
see self._get_revision_store for the method to
737
get the store for a repository.
740
name = safe_unicode(name)
743
dir_mode = control_files._dir_mode
744
file_mode = control_files._file_mode
745
revision_store =TextStore(transport.clone(name),
747
compressed=compressed,
750
revision_store.register_suffix('sig')
751
return revision_store
753
def initialize(self, a_bzrdir, shared=False):
754
"""Initialize a repository of this format in a_bzrdir.
756
:param a_bzrdir: The bzrdir to put the new repository in it.
757
:param shared: The repository should be initialized as a sharable one.
759
This may raise UninitializableFormat if shared repository are not
760
compatible the a_bzrdir.
763
def is_supported(self):
764
"""Is this format supported?
766
Supported formats must be initializable and openable.
767
Unsupported formats may not support initialization or committing or
768
some other features depending on the reason for not being supported.
772
def open(self, a_bzrdir, _found=False):
773
"""Return an instance of this format for the bzrdir a_bzrdir.
775
_found is a private parameter, do not use it.
777
raise NotImplementedError(self.open)
780
def register_format(klass, format):
781
klass._formats[format.get_format_string()] = format
784
def set_default_format(klass, format):
785
klass._default_format = format
788
def unregister_format(klass, format):
789
assert klass._formats[format.get_format_string()] is format
790
del klass._formats[format.get_format_string()]
793
class PreSplitOutRepositoryFormat(RepositoryFormat):
794
"""Base class for the pre split out repository formats."""
796
def initialize(self, a_bzrdir, shared=False, _internal=False):
797
"""Create a weave repository.
799
TODO: when creating split out bzr branch formats, move this to a common
800
base for Format5, Format6. or something like that.
802
from bzrlib.weavefile import write_weave_v5
803
from bzrlib.weave import Weave
806
raise errors.IncompatibleFormat(self, a_bzrdir._format)
809
# always initialized when the bzrdir is.
810
return self.open(a_bzrdir, _found=True)
812
# Create an empty weave
814
bzrlib.weavefile.write_weave_v5(Weave(), sio)
815
empty_weave = sio.getvalue()
817
mutter('creating repository in %s.', a_bzrdir.transport.base)
818
dirs = ['revision-store', 'weaves']
819
files = [('inventory.weave', StringIO(empty_weave)),
822
# FIXME: RBC 20060125 dont peek under the covers
823
# NB: no need to escape relative paths that are url safe.
824
control_files = LockableFiles(a_bzrdir.transport, 'branch-lock',
826
control_files.create_lock()
827
control_files.lock_write()
828
control_files._transport.mkdir_multi(dirs,
829
mode=control_files._dir_mode)
831
for file, content in files:
832
control_files.put(file, content)
834
control_files.unlock()
835
return self.open(a_bzrdir, _found=True)
837
def open(self, a_bzrdir, _found=False):
838
"""See RepositoryFormat.open()."""
840
# we are being called directly and must probe.
841
raise NotImplementedError
843
repo_transport = a_bzrdir.get_repository_transport(None)
844
control_files = a_bzrdir._control_files
845
revision_store = self._get_revision_store(repo_transport, control_files)
846
return AllInOneRepository(_format=self,
848
revision_store=revision_store)
851
class RepositoryFormat4(PreSplitOutRepositoryFormat):
852
"""Bzr repository format 4.
854
This repository format has:
856
- TextStores for texts, inventories,revisions.
858
This format is deprecated: it indexes texts using a text id which is
859
removed in format 5; initializationa and write support for this format
864
super(RepositoryFormat4, self).__init__()
865
self._matchingbzrdir = bzrlib.bzrdir.BzrDirFormat4()
867
def initialize(self, url, shared=False, _internal=False):
868
"""Format 4 branches cannot be created."""
869
raise errors.UninitializableFormat(self)
871
def is_supported(self):
872
"""Format 4 is not supported.
874
It is not supported because the model changed from 4 to 5 and the
875
conversion logic is expensive - so doing it on the fly was not
880
def _get_revision_store(self, repo_transport, control_files):
881
"""See RepositoryFormat._get_revision_store()."""
882
return self._get_rev_store(repo_transport,
887
class RepositoryFormat5(PreSplitOutRepositoryFormat):
888
"""Bzr control format 5.
890
This repository format has:
891
- weaves for file texts and inventory
893
- TextStores for revisions and signatures.
897
super(RepositoryFormat5, self).__init__()
898
self._matchingbzrdir = bzrlib.bzrdir.BzrDirFormat5()
900
def _get_revision_store(self, repo_transport, control_files):
901
"""See RepositoryFormat._get_revision_store()."""
902
"""Return the revision store object for this a_bzrdir."""
903
return self._get_rev_store(repo_transport,
909
class RepositoryFormat6(PreSplitOutRepositoryFormat):
910
"""Bzr control format 6.
912
This repository format has:
913
- weaves for file texts and inventory
914
- hash subdirectory based stores.
915
- TextStores for revisions and signatures.
919
super(RepositoryFormat6, self).__init__()
920
self._matchingbzrdir = bzrlib.bzrdir.BzrDirFormat6()
922
def _get_revision_store(self, repo_transport, control_files):
923
"""See RepositoryFormat._get_revision_store()."""
924
return self._get_rev_store(repo_transport,
931
class MetaDirRepositoryFormat(RepositoryFormat):
932
"""Common base class for the new repositories using the metadir layour."""
935
super(MetaDirRepositoryFormat, self).__init__()
936
self._matchingbzrdir = bzrlib.bzrdir.BzrDirMetaFormat1()
938
def _create_control_files(self, a_bzrdir):
939
"""Create the required files and the initial control_files object."""
940
# FIXME: RBC 20060125 dont peek under the covers
941
# NB: no need to escape relative paths that are url safe.
942
repository_transport = a_bzrdir.get_repository_transport(self)
943
control_files = LockableFiles(repository_transport, 'lock', LockDir)
944
control_files.create_lock()
947
def _get_revision_store(self, repo_transport, control_files):
948
"""See RepositoryFormat._get_revision_store()."""
949
return self._get_rev_store(repo_transport,
956
def open(self, a_bzrdir, _found=False, _override_transport=None):
957
"""See RepositoryFormat.open().
959
:param _override_transport: INTERNAL USE ONLY. Allows opening the
960
repository at a slightly different url
961
than normal. I.e. during 'upgrade'.
964
format = RepositoryFormat.find_format(a_bzrdir)
965
assert format.__class__ == self.__class__
966
if _override_transport is not None:
967
repo_transport = _override_transport
969
repo_transport = a_bzrdir.get_repository_transport(None)
970
control_files = LockableFiles(repo_transport, 'lock', LockDir)
971
revision_store = self._get_revision_store(repo_transport, control_files)
972
return MetaDirRepository(_format=self,
974
control_files=control_files,
975
revision_store=revision_store)
977
def _upload_blank_content(self, a_bzrdir, dirs, files, utf8_files, shared):
978
"""Upload the initial blank content."""
979
control_files = self._create_control_files(a_bzrdir)
980
control_files.lock_write()
982
control_files._transport.mkdir_multi(dirs,
983
mode=control_files._dir_mode)
984
for file, content in files:
985
control_files.put(file, content)
986
for file, content in utf8_files:
987
control_files.put_utf8(file, content)
989
control_files.put_utf8('shared-storage', '')
991
control_files.unlock()
994
class RepositoryFormat7(MetaDirRepositoryFormat):
997
This repository format has:
998
- weaves for file texts and inventory
999
- hash subdirectory based stores.
1000
- TextStores for revisions and signatures.
1001
- a format marker of its own
1002
- an optional 'shared-storage' flag
1003
- an optional 'no-working-trees' flag
1006
def get_format_string(self):
1007
"""See RepositoryFormat.get_format_string()."""
1008
return "Bazaar-NG Repository format 7"
1010
def initialize(self, a_bzrdir, shared=False):
1011
"""Create a weave repository.
1013
:param shared: If true the repository will be initialized as a shared
1016
from bzrlib.weavefile import write_weave_v5
1017
from bzrlib.weave import Weave
1019
# Create an empty weave
1021
bzrlib.weavefile.write_weave_v5(Weave(), sio)
1022
empty_weave = sio.getvalue()
1024
mutter('creating repository in %s.', a_bzrdir.transport.base)
1025
dirs = ['revision-store', 'weaves']
1026
files = [('inventory.weave', StringIO(empty_weave)),
1028
utf8_files = [('format', self.get_format_string())]
1030
self._upload_blank_content(a_bzrdir, dirs, files, utf8_files, shared)
1031
return self.open(a_bzrdir=a_bzrdir, _found=True)
1034
class RepositoryFormatKnit1(MetaDirRepositoryFormat):
1035
"""Bzr repository knit format 1.
1037
This repository format has:
1038
- knits for file texts and inventory
1039
- hash subdirectory based stores.
1040
- knits for revisions and signatures
1041
- TextStores for revisions and signatures.
1042
- a format marker of its own
1043
- an optional 'shared-storage' flag
1044
- an optional 'no-working-trees' flag
1048
def get_format_string(self):
1049
"""See RepositoryFormat.get_format_string()."""
1050
return "Bazaar-NG Knit Repository Format 1"
1052
def initialize(self, a_bzrdir, shared=False):
1053
"""Create a knit format 1 repository.
1055
:param shared: If true the repository will be initialized as a shared
1057
XXX NOTE that this current uses a Weave for testing and will become
1058
A Knit in due course.
1060
from bzrlib.weavefile import write_weave_v5
1061
from bzrlib.weave import Weave
1063
# Create an empty weave
1065
bzrlib.weavefile.write_weave_v5(Weave(), sio)
1066
empty_weave = sio.getvalue()
1068
mutter('creating repository in %s.', a_bzrdir.transport.base)
1069
dirs = ['revision-store', 'knits']
1070
files = [('inventory.weave', StringIO(empty_weave)),
1072
utf8_files = [('format', self.get_format_string())]
1074
self._upload_blank_content(a_bzrdir, dirs, files, utf8_files, shared)
1075
return self.open(a_bzrdir=a_bzrdir, _found=True)
1078
# formats which have no format string are not discoverable
1079
# and not independently creatable, so are not registered.
1080
_default_format = RepositoryFormat7()
1081
RepositoryFormat.register_format(_default_format)
1082
RepositoryFormat.register_format(RepositoryFormatKnit1())
1083
RepositoryFormat.set_default_format(_default_format)
1084
_legacy_formats = [RepositoryFormat4(),
1085
RepositoryFormat5(),
1086
RepositoryFormat6()]
1089
class InterRepository(object):
1090
"""This class represents operations taking place between two repositories.
1092
Its instances have methods like copy_content and fetch, and contain
1093
references to the source and target repositories these operations can be
1096
Often we will provide convenience methods on 'repository' which carry out
1097
operations with another repository - they will always forward to
1098
InterRepository.get(other).method_name(parameters).
1100
# XXX: FIXME: FUTURE: robertc
1101
# testing of these probably requires a factory in optimiser type, and
1102
# then a test adapter to test each type thoroughly.
1106
"""The available optimised InterRepository types."""
1108
def __init__(self, source, target):
1109
"""Construct a default InterRepository instance. Please use 'get'.
1111
Only subclasses of InterRepository should call
1112
InterRepository.__init__ - clients should call InterRepository.get
1113
instead which will create an optimised InterRepository if possible.
1115
self.source = source
1116
self.target = target
1119
def copy_content(self, revision_id=None, basis=None):
1120
"""Make a complete copy of the content in self into destination.
1122
This is a destructive operation! Do not use it on existing
1125
:param revision_id: Only copy the content needed to construct
1126
revision_id and its parents.
1127
:param basis: Copy the needed data preferentially from basis.
1130
self.target.set_make_working_trees(self.source.make_working_trees())
1131
except NotImplementedError:
1133
# grab the basis available data
1134
if basis is not None:
1135
self.target.fetch(basis, revision_id=revision_id)
1136
# but dont both fetching if we have the needed data now.
1137
if (revision_id not in (None, NULL_REVISION) and
1138
self.target.has_revision(revision_id)):
1140
self.target.fetch(self.source, revision_id=revision_id)
1142
def _double_lock(self, lock_source, lock_target):
1143
"""Take out too locks, rolling back the first if the second throws."""
1148
# we want to ensure that we don't leave source locked by mistake.
1149
# and any error on target should not confuse source.
1150
self.source.unlock()
1154
def fetch(self, revision_id=None, pb=None):
1155
"""Fetch the content required to construct revision_id.
1157
The content is copied from source to target.
1159
:param revision_id: if None all content is copied, if NULL_REVISION no
1161
:param pb: optional progress bar to use for progress reports. If not
1162
provided a default one will be created.
1164
Returns the copied revision count and the failed revisions in a tuple:
1167
from bzrlib.fetch import RepoFetcher
1168
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
1169
self.source, self.source._format, self.target, self.target._format)
1170
f = RepoFetcher(to_repository=self.target,
1171
from_repository=self.source,
1172
last_revision=revision_id,
1174
return f.count_copied, f.failed_revisions
1177
def get(klass, repository_source, repository_target):
1178
"""Retrieve a InterRepository worker object for these repositories.
1180
:param repository_source: the repository to be the 'source' member of
1181
the InterRepository instance.
1182
:param repository_target: the repository to be the 'target' member of
1183
the InterRepository instance.
1184
If an optimised InterRepository worker exists it will be used otherwise
1185
a default InterRepository instance will be created.
1187
for provider in klass._optimisers:
1188
if provider.is_compatible(repository_source, repository_target):
1189
return provider(repository_source, repository_target)
1190
return InterRepository(repository_source, repository_target)
1192
def lock_read(self):
1193
"""Take out a logical read lock.
1195
This will lock the source branch and the target branch. The source gets
1196
a read lock and the target a read lock.
1198
self._double_lock(self.source.lock_read, self.target.lock_read)
1200
def lock_write(self):
1201
"""Take out a logical write lock.
1203
This will lock the source branch and the target branch. The source gets
1204
a read lock and the target a write lock.
1206
self._double_lock(self.source.lock_read, self.target.lock_write)
1209
def missing_revision_ids(self, revision_id=None):
1210
"""Return the revision ids that source has that target does not.
1212
These are returned in topological order.
1214
:param revision_id: only return revision ids included by this
1217
# generic, possibly worst case, slow code path.
1218
target_ids = set(self.target.all_revision_ids())
1219
if revision_id is not None:
1220
source_ids = self.source.get_ancestry(revision_id)
1221
assert source_ids.pop(0) == None
1223
source_ids = self.source.all_revision_ids()
1224
result_set = set(source_ids).difference(target_ids)
1225
# this may look like a no-op: its not. It preserves the ordering
1226
# other_ids had while only returning the members from other_ids
1227
# that we've decided we need.
1228
return [rev_id for rev_id in source_ids if rev_id in result_set]
1231
def register_optimiser(klass, optimiser):
1232
"""Register an InterRepository optimiser."""
1233
klass._optimisers.add(optimiser)
1236
"""Release the locks on source and target."""
1238
self.target.unlock()
1240
self.source.unlock()
1243
def unregister_optimiser(klass, optimiser):
1244
"""Unregister an InterRepository optimiser."""
1245
klass._optimisers.remove(optimiser)
1248
class InterWeaveRepo(InterRepository):
1249
"""Optimised code paths between Weave based repositories."""
1251
_matching_repo_format = _default_format
1252
"""Repository format for testing with."""
1255
def is_compatible(source, target):
1256
"""Be compatible with known Weave formats.
1258
We dont test for the stores being of specific types becase that
1259
could lead to confusing results, and there is no need to be
1263
return (isinstance(source._format, (RepositoryFormat5,
1265
RepositoryFormat7)) and
1266
isinstance(target._format, (RepositoryFormat5,
1268
RepositoryFormat7)))
1269
except AttributeError:
1273
def copy_content(self, revision_id=None, basis=None):
1274
"""See InterRepository.copy_content()."""
1275
# weave specific optimised path:
1276
if basis is not None:
1277
# copy the basis in, then fetch remaining data.
1278
basis.copy_content_into(self.target, revision_id)
1279
# the basis copy_content_into could misset this.
1281
self.target.set_make_working_trees(self.source.make_working_trees())
1282
except NotImplementedError:
1284
self.target.fetch(self.source, revision_id=revision_id)
1287
self.target.set_make_working_trees(self.source.make_working_trees())
1288
except NotImplementedError:
1290
# FIXME do not peek!
1291
if self.source.control_files._transport.listable():
1292
pb = bzrlib.ui.ui_factory.nested_progress_bar()
1294
copy_all(self.source.weave_store,
1295
self.target.weave_store, pb=pb)
1296
pb.update('copying inventory', 0, 1)
1297
self.target.control_weaves.copy_multi(
1298
self.source.control_weaves, ['inventory'])
1299
copy_all(self.source.revision_store,
1300
self.target.revision_store, pb=pb)
1304
self.target.fetch(self.source, revision_id=revision_id)
1307
def fetch(self, revision_id=None, pb=None):
1308
"""See InterRepository.fetch()."""
1309
from bzrlib.fetch import RepoFetcher
1310
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
1311
self.source, self.source._format, self.target, self.target._format)
1312
f = RepoFetcher(to_repository=self.target,
1313
from_repository=self.source,
1314
last_revision=revision_id,
1316
return f.count_copied, f.failed_revisions
1319
def missing_revision_ids(self, revision_id=None):
1320
"""See InterRepository.missing_revision_ids()."""
1321
# we want all revisions to satisfy revision_id in source.
1322
# but we dont want to stat every file here and there.
1323
# we want then, all revisions other needs to satisfy revision_id
1324
# checked, but not those that we have locally.
1325
# so the first thing is to get a subset of the revisions to
1326
# satisfy revision_id in source, and then eliminate those that
1327
# we do already have.
1328
# this is slow on high latency connection to self, but as as this
1329
# disk format scales terribly for push anyway due to rewriting
1330
# inventory.weave, this is considered acceptable.
1332
if revision_id is not None:
1333
source_ids = self.source.get_ancestry(revision_id)
1334
assert source_ids.pop(0) == None
1336
source_ids = self.source._all_possible_ids()
1337
source_ids_set = set(source_ids)
1338
# source_ids is the worst possible case we may need to pull.
1339
# now we want to filter source_ids against what we actually
1340
# have in target, but dont try to check for existence where we know
1341
# we do not have a revision as that would be pointless.
1342
target_ids = set(self.target._all_possible_ids())
1343
possibly_present_revisions = target_ids.intersection(source_ids_set)
1344
actually_present_revisions = set(self.target._eliminate_revisions_not_present(possibly_present_revisions))
1345
required_revisions = source_ids_set.difference(actually_present_revisions)
1346
required_topo_revisions = [rev_id for rev_id in source_ids if rev_id in required_revisions]
1347
if revision_id is not None:
1348
# we used get_ancestry to determine source_ids then we are assured all
1349
# revisions referenced are present as they are installed in topological order.
1350
# and the tip revision was validated by get_ancestry.
1351
return required_topo_revisions
1353
# if we just grabbed the possibly available ids, then
1354
# we only have an estimate of whats available and need to validate
1355
# that against the revision records.
1356
return self.source._eliminate_revisions_not_present(required_topo_revisions)
1359
InterRepository.register_optimiser(InterWeaveRepo)
1362
class RepositoryTestProviderAdapter(object):
1363
"""A tool to generate a suite testing multiple repository formats at once.
1365
This is done by copying the test once for each transport and injecting
1366
the transport_server, transport_readonly_server, and bzrdir_format and
1367
repository_format classes into each copy. Each copy is also given a new id()
1368
to make it easy to identify.
1371
def __init__(self, transport_server, transport_readonly_server, formats):
1372
self._transport_server = transport_server
1373
self._transport_readonly_server = transport_readonly_server
1374
self._formats = formats
1376
def adapt(self, test):
1377
result = TestSuite()
1378
for repository_format, bzrdir_format in self._formats:
1379
new_test = deepcopy(test)
1380
new_test.transport_server = self._transport_server
1381
new_test.transport_readonly_server = self._transport_readonly_server
1382
new_test.bzrdir_format = bzrdir_format
1383
new_test.repository_format = repository_format
1384
def make_new_test_id():
1385
new_id = "%s(%s)" % (new_test.id(), repository_format.__class__.__name__)
1386
return lambda: new_id
1387
new_test.id = make_new_test_id()
1388
result.addTest(new_test)
1392
class InterRepositoryTestProviderAdapter(object):
1393
"""A tool to generate a suite testing multiple inter repository formats.
1395
This is done by copying the test once for each interrepo provider and injecting
1396
the transport_server, transport_readonly_server, repository_format and
1397
repository_to_format classes into each copy.
1398
Each copy is also given a new id() to make it easy to identify.
1401
def __init__(self, transport_server, transport_readonly_server, formats):
1402
self._transport_server = transport_server
1403
self._transport_readonly_server = transport_readonly_server
1404
self._formats = formats
1406
def adapt(self, test):
1407
result = TestSuite()
1408
for interrepo_class, repository_format, repository_format_to in self._formats:
1409
new_test = deepcopy(test)
1410
new_test.transport_server = self._transport_server
1411
new_test.transport_readonly_server = self._transport_readonly_server
1412
new_test.interrepo_class = interrepo_class
1413
new_test.repository_format = repository_format
1414
new_test.repository_format_to = repository_format_to
1415
def make_new_test_id():
1416
new_id = "%s(%s)" % (new_test.id(), interrepo_class.__name__)
1417
return lambda: new_id
1418
new_test.id = make_new_test_id()
1419
result.addTest(new_test)
1423
def default_test_list():
1424
"""Generate the default list of interrepo permutations to test."""
1426
# test the default InterRepository between format 6 and the current
1428
# XXX: robertc 20060220 reinstate this when there are two supported
1429
# formats which do not have an optimal code path between them.
1430
result.append((InterRepository,
1431
RepositoryFormat6(),
1432
RepositoryFormatKnit1()))
1433
for optimiser in InterRepository._optimisers:
1434
result.append((optimiser,
1435
optimiser._matching_repo_format,
1436
optimiser._matching_repo_format
1438
# if there are specific combinations we want to use, we can add them
1443
class CopyConverter(object):
1444
"""A repository conversion tool which just performs a copy of the content.
1446
This is slow but quite reliable.
1449
def __init__(self, target_format):
1450
"""Create a CopyConverter.
1452
:param target_format: The format the resulting repository should be.
1454
self.target_format = target_format
1456
def convert(self, repo, pb):
1457
"""Perform the conversion of to_convert, giving feedback via pb.
1459
:param to_convert: The disk object to convert.
1460
:param pb: a progress bar to use for progress information.
1465
# this is only useful with metadir layouts - separated repo content.
1466
# trigger an assertion if not such
1467
repo._format.get_format_string()
1468
self.repo_dir = repo.bzrdir
1469
self.step('Moving repository to repository.backup')
1470
self.repo_dir.transport.move('repository', 'repository.backup')
1471
backup_transport = self.repo_dir.transport.clone('repository.backup')
1472
self.source_repo = repo._format.open(self.repo_dir,
1474
_override_transport=backup_transport)
1475
self.step('Creating new repository')
1476
converted = self.target_format.initialize(self.repo_dir,
1477
self.source_repo.is_shared())
1478
converted.lock_write()
1480
self.step('Copying content into repository.')
1481
self.source_repo.copy_content_into(converted)
1484
self.step('Deleting old repository content.')
1485
self.repo_dir.transport.delete_tree('repository.backup')
1486
self.pb.note('repository converted')
1488
def step(self, message):
1489
"""Update the pb by a step."""
1491
self.pb.update(message, self.count, self.total)
added
removed
bzrlib/repository.py
