1
# Copyright (C) 2005 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
23
from bzrlib.decorators import needs_read_lock, needs_write_lock
24
import bzrlib.errors as errors
25
from bzrlib.errors import InvalidRevisionId
26
import bzrlib.gpg as gpg
27
from bzrlib.lockable_files import LockableFiles
28
from bzrlib.osutils import safe_unicode
29
from bzrlib.revision import NULL_REVISION
30
from bzrlib.store import copy_all
31
from bzrlib.store.weave import WeaveStore
32
from bzrlib.store.text import TextStore
33
from bzrlib.symbol_versioning import *
34
from bzrlib.trace import mutter
35
from bzrlib.tree import RevisionTree
36
from bzrlib.testament import Testament
37
from bzrlib.tree import EmptyTree
42
class Repository(object):
43
"""Repository holding history for one or more branches.
45
The repository holds and retrieves historical information including
46
revisions and file history. It's normally accessed only by the Branch,
47
which views a particular line of development through that history.
49
The Repository builds on top of Stores and a Transport, which respectively
50
describe the disk data format and the way of accessing the (possibly
55
def add_inventory(self, revid, inv, parents):
56
"""Add the inventory inv to the repository as revid.
58
:param parents: The revision ids of the parents that revid
59
is known to have and are in the repository already.
61
returns the sha1 of the serialized inventory.
63
inv_text = bzrlib.xml5.serializer_v5.write_inventory_to_string(inv)
64
inv_sha1 = bzrlib.osutils.sha_string(inv_text)
65
self.control_weaves.add_text('inventory', revid,
66
bzrlib.osutils.split_lines(inv_text), parents,
67
self.get_transaction())
71
def add_revision(self, rev_id, rev, inv=None, config=None):
72
"""Add rev to the revision store as rev_id.
74
:param rev_id: the revision id to use.
75
:param rev: The revision object.
76
:param inv: The inventory for the revision. if None, it will be looked
77
up in the inventory storer
78
:param config: If None no digital signature will be created.
79
If supplied its signature_needed method will be used
80
to determine if a signature should be made.
82
if config is not None and config.signature_needed():
84
inv = self.get_inventory(rev_id)
85
plaintext = Testament(rev, inv).as_short_text()
86
self.store_revision_signature(
87
gpg.GPGStrategy(config), plaintext, rev_id)
88
if not rev_id in self.get_inventory_weave():
90
raise errors.WeaveRevisionNotPresent(rev_id,
91
self.get_inventory_weave())
93
# yes, this is not suitable for adding with ghosts.
94
self.add_inventory(rev_id, inv, rev.parent_ids)
97
bzrlib.xml5.serializer_v5.write_revision(rev, rev_tmp)
99
self.revision_store.add(rev_tmp, rev_id)
100
mutter('added revision_id {%s}', rev_id)
103
def _all_possible_ids(self):
104
"""Return all the possible revisions that we could find."""
105
return self.get_inventory_weave().names()
108
def all_revision_ids(self):
109
"""Returns a list of all the revision ids in the repository.
111
These are in as much topological order as the underlying store can
112
present: for weaves ghosts may lead to a lack of correctness until
113
the reweave updates the parents list.
115
result = self._all_possible_ids()
116
return self._eliminate_revisions_not_present(result)
119
def _eliminate_revisions_not_present(self, revision_ids):
120
"""Check every revision id in revision_ids to see if we have it.
122
Returns a set of the present revisions.
125
for id in revision_ids:
126
if self.has_revision(id):
131
def create(a_bzrdir):
132
"""Construct the current default format repository in a_bzrdir."""
133
return RepositoryFormat.get_default_format().initialize(a_bzrdir)
135
def __init__(self, _format, a_bzrdir, control_files, revision_store):
136
"""instantiate a Repository.
138
:param _format: The format of the repository on disk.
139
:param a_bzrdir: The BzrDir of the repository.
141
In the future we will have a single api for all stores for
142
getting file texts, inventories and revisions, then
143
this construct will accept instances of those things.
145
object.__init__(self)
146
self._format = _format
147
# the following are part of the public API for Repository:
148
self.bzrdir = a_bzrdir
149
self.control_files = control_files
150
self.revision_store = revision_store
152
def lock_write(self):
153
self.control_files.lock_write()
156
self.control_files.lock_read()
159
def missing_revision_ids(self, other, revision_id=None):
160
"""Return the revision ids that other has that this does not.
162
These are returned in topological order.
164
revision_id: only return revision ids included by revision_id.
166
return InterRepository.get(other, self).missing_revision_ids(revision_id)
170
"""Open the repository rooted at base.
172
For instance, if the repository is at URL/.bzr/repository,
173
Repository.open(URL) -> a Repository instance.
175
control = bzrlib.bzrdir.BzrDir.open(base)
176
return control.open_repository()
178
def copy_content_into(self, destination, revision_id=None, basis=None):
179
"""Make a complete copy of the content in self into destination.
181
This is a destructive operation! Do not use it on existing
184
return InterRepository.get(self, destination).copy_content(revision_id, basis)
186
def fetch(self, source, revision_id=None, pb=None):
187
"""Fetch the content required to construct revision_id from source.
189
If revision_id is None all content is copied.
191
return InterRepository.get(source, self).fetch(revision_id=revision_id,
195
self.control_files.unlock()
198
def clone(self, a_bzrdir, revision_id=None, basis=None):
199
"""Clone this repository into a_bzrdir using the current format.
201
Currently no check is made that the format of this repository and
202
the bzrdir format are compatible. FIXME RBC 20060201.
204
if not isinstance(a_bzrdir._format, self.bzrdir._format.__class__):
205
# use target default format.
206
result = a_bzrdir.create_repository()
207
# FIXME RBC 20060209 split out the repository type to avoid this check ?
208
elif isinstance(a_bzrdir._format,
209
(bzrlib.bzrdir.BzrDirFormat4,
210
bzrlib.bzrdir.BzrDirFormat5,
211
bzrlib.bzrdir.BzrDirFormat6)):
212
result = a_bzrdir.open_repository()
214
result = self._format.initialize(a_bzrdir, shared=self.is_shared())
215
self.copy_content_into(result, revision_id, basis)
218
def has_revision(self, revision_id):
219
"""True if this branch has a copy of the revision.
221
This does not necessarily imply the revision is merge
222
or on the mainline."""
223
return (revision_id is None
224
or self.revision_store.has_id(revision_id))
227
def get_revision_xml_file(self, revision_id):
228
"""Return XML file object for revision object."""
229
if not revision_id or not isinstance(revision_id, basestring):
230
raise InvalidRevisionId(revision_id=revision_id, branch=self)
232
return self.revision_store.get(revision_id)
233
except (IndexError, KeyError):
234
raise bzrlib.errors.NoSuchRevision(self, revision_id)
237
def get_revision_xml(self, revision_id):
238
return self.get_revision_xml_file(revision_id).read()
241
def get_revision_reconcile(self, revision_id):
242
"""'reconcile' helper routine that allows access to a revision always.
244
This variant of get_revision does not cross check the weave graph
245
against the revision one as get_revision does: but it should only
246
be used by reconcile, or reconcile-alike commands that are correcting
247
or testing the revision graph.
249
xml_file = self.get_revision_xml_file(revision_id)
252
r = bzrlib.xml5.serializer_v5.read_revision(xml_file)
253
except SyntaxError, e:
254
raise bzrlib.errors.BzrError('failed to unpack revision_xml',
258
assert r.revision_id == revision_id
262
def get_revision(self, revision_id):
263
"""Return the Revision object for a named revision"""
264
r = self.get_revision_reconcile(revision_id)
265
# weave corruption can lead to absent revision markers that should be
267
# the following test is reasonably cheap (it needs a single weave read)
268
# and the weave is cached in read transactions. In write transactions
269
# it is not cached but typically we only read a small number of
270
# revisions. For knits when they are introduced we will probably want
271
# to ensure that caching write transactions are in use.
272
inv = self.get_inventory_weave()
273
self._check_revision_parents(r, inv)
276
def _check_revision_parents(self, revision, inventory):
277
"""Private to Repository and Fetch.
279
This checks the parentage of revision in an inventory weave for
280
consistency and is only applicable to inventory-weave-for-ancestry
281
using repository formats & fetchers.
283
weave_parents = inventory.parent_names(revision.revision_id)
284
weave_names = inventory.names()
285
for parent_id in revision.parent_ids:
286
if parent_id in weave_names:
287
# this parent must not be a ghost.
288
if not parent_id in weave_parents:
290
raise errors.CorruptRepository(self)
293
def get_revision_sha1(self, revision_id):
294
"""Hash the stored value of a revision, and return it."""
295
# In the future, revision entries will be signed. At that
296
# point, it is probably best *not* to include the signature
297
# in the revision hash. Because that lets you re-sign
298
# the revision, (add signatures/remove signatures) and still
299
# have all hash pointers stay consistent.
300
# But for now, just hash the contents.
301
return bzrlib.osutils.sha_file(self.get_revision_xml_file(revision_id))
304
def store_revision_signature(self, gpg_strategy, plaintext, revision_id):
305
self.revision_store.add(StringIO(gpg_strategy.sign(plaintext)),
308
def fileid_involved_between_revs(self, from_revid, to_revid):
309
"""Find file_id(s) which are involved in the changes between revisions.
311
This determines the set of revisions which are involved, and then
312
finds all file ids affected by those revisions.
314
# TODO: jam 20060119 This code assumes that w.inclusions will
315
# always be correct. But because of the presence of ghosts
316
# it is possible to be wrong.
317
# One specific example from Robert Collins:
318
# Two branches, with revisions ABC, and AD
319
# C is a ghost merge of D.
320
# Inclusions doesn't recognize D as an ancestor.
321
# If D is ever merged in the future, the weave
322
# won't be fixed, because AD never saw revision C
323
# to cause a conflict which would force a reweave.
324
w = self.get_inventory_weave()
325
from_set = set(w.inclusions([w.lookup(from_revid)]))
326
to_set = set(w.inclusions([w.lookup(to_revid)]))
327
included = to_set.difference(from_set)
328
changed = map(w.idx_to_name, included)
329
return self._fileid_involved_by_set(changed)
331
def fileid_involved(self, last_revid=None):
332
"""Find all file_ids modified in the ancestry of last_revid.
334
:param last_revid: If None, last_revision() will be used.
336
w = self.get_inventory_weave()
338
changed = set(w._names)
340
included = w.inclusions([w.lookup(last_revid)])
341
changed = map(w.idx_to_name, included)
342
return self._fileid_involved_by_set(changed)
344
def fileid_involved_by_set(self, changes):
345
"""Find all file_ids modified by the set of revisions passed in.
347
:param changes: A set() of revision ids
349
# TODO: jam 20060119 This line does *nothing*, remove it.
350
# or better yet, change _fileid_involved_by_set so
351
# that it takes the inventory weave, rather than
352
# pulling it out by itself.
353
return self._fileid_involved_by_set(changes)
355
def _fileid_involved_by_set(self, changes):
356
"""Find the set of file-ids affected by the set of revisions.
358
:param changes: A set() of revision ids.
359
:return: A set() of file ids.
361
This peaks at the Weave, interpreting each line, looking to
362
see if it mentions one of the revisions. And if so, includes
363
the file id mentioned.
364
This expects both the Weave format, and the serialization
365
to have a single line per file/directory, and to have
366
fileid="" and revision="" on that line.
368
assert isinstance(self._format, (RepositoryFormat5,
371
RepositoryFormatKnit1)), \
372
"fileid_involved only supported for branches which store inventory as unnested xml"
374
w = self.get_inventory_weave()
376
for line in w._weave:
378
# it is ugly, but it is due to the weave structure
379
if not isinstance(line, basestring): continue
381
start = line.find('file_id="')+9
382
if start < 9: continue
383
end = line.find('"', start)
385
file_id = xml.sax.saxutils.unescape(line[start:end])
387
# check if file_id is already present
388
if file_id in file_ids: continue
390
start = line.find('revision="')+10
391
if start < 10: continue
392
end = line.find('"', start)
394
revision_id = xml.sax.saxutils.unescape(line[start:end])
396
if revision_id in changes:
397
file_ids.add(file_id)
401
def get_inventory_weave(self):
402
return self.control_weaves.get_weave('inventory',
403
self.get_transaction())
406
def get_inventory(self, revision_id):
407
"""Get Inventory object by hash."""
408
xml = self.get_inventory_xml(revision_id)
409
return bzrlib.xml5.serializer_v5.read_inventory_from_string(xml)
412
def get_inventory_xml(self, revision_id):
413
"""Get inventory XML as a file object."""
415
assert isinstance(revision_id, basestring), type(revision_id)
416
iw = self.get_inventory_weave()
417
return iw.get_text(iw.lookup(revision_id))
419
raise bzrlib.errors.HistoryMissing(self, 'inventory', revision_id)
422
def get_inventory_sha1(self, revision_id):
423
"""Return the sha1 hash of the inventory entry
425
return self.get_revision(revision_id).inventory_sha1
428
def get_revision_graph(self, revision_id=None):
429
"""Return a dictionary containing the revision graph.
431
:return: a dictionary of revision_id->revision_parents_list.
433
weave = self.get_inventory_weave()
434
all_revisions = self._eliminate_revisions_not_present(weave.names())
435
entire_graph = dict([(node, weave.parent_names(node)) for
436
node in all_revisions])
437
if revision_id is None:
439
elif revision_id not in entire_graph:
440
raise errors.NoSuchRevision(self, revision_id)
442
# add what can be reached from revision_id
444
pending = set([revision_id])
445
while len(pending) > 0:
447
result[node] = entire_graph[node]
448
for revision_id in result[node]:
449
if revision_id not in result:
450
pending.add(revision_id)
454
def get_revision_inventory(self, revision_id):
455
"""Return inventory of a past revision."""
456
# TODO: Unify this with get_inventory()
457
# bzr 0.0.6 and later imposes the constraint that the inventory_id
458
# must be the same as its revision, so this is trivial.
459
if revision_id is None:
460
# This does not make sense: if there is no revision,
461
# then it is the current tree inventory surely ?!
462
# and thus get_root_id() is something that looks at the last
463
# commit on the branch, and the get_root_id is an inventory check.
464
raise NotImplementedError
465
# return Inventory(self.get_root_id())
467
return self.get_inventory(revision_id)
471
"""Return True if this repository is flagged as a shared repository."""
472
# FIXME format 4-6 cannot be shared, this is technically faulty.
473
return self.control_files._transport.has('shared-storage')
476
def revision_tree(self, revision_id):
477
"""Return Tree for a revision on this branch.
479
`revision_id` may be None for the null revision, in which case
480
an `EmptyTree` is returned."""
481
# TODO: refactor this to use an existing revision object
482
# so we don't need to read it in twice.
483
if revision_id is None or revision_id == NULL_REVISION:
486
inv = self.get_revision_inventory(revision_id)
487
return RevisionTree(self, inv, revision_id)
490
def get_ancestry(self, revision_id):
491
"""Return a list of revision-ids integrated by a revision.
493
This is topologically sorted.
495
if revision_id is None:
497
if not self.has_revision(revision_id):
498
raise errors.NoSuchRevision(self, revision_id)
499
w = self.get_inventory_weave()
500
return [None] + map(w.idx_to_name,
501
w.inclusions([w.lookup(revision_id)]))
504
def print_file(self, file, revision_id):
505
"""Print `file` to stdout.
507
FIXME RBC 20060125 as John Meinel points out this is a bad api
508
- it writes to stdout, it assumes that that is valid etc. Fix
509
by creating a new more flexible convenience function.
511
tree = self.revision_tree(revision_id)
512
# use inventory as it was in that revision
513
file_id = tree.inventory.path2id(file)
515
raise BzrError("%r is not present in revision %s" % (file, revno))
517
revno = self.revision_id_to_revno(revision_id)
518
except errors.NoSuchRevision:
519
# TODO: This should not be BzrError,
520
# but NoSuchFile doesn't fit either
521
raise BzrError('%r is not present in revision %s'
522
% (file, revision_id))
524
raise BzrError('%r is not present in revision %s'
526
tree.print_file(file_id)
528
def get_transaction(self):
529
return self.control_files.get_transaction()
531
def revision_parents(self, revid):
532
return self.get_inventory_weave().parent_names(revid)
535
def set_make_working_trees(self, new_value):
536
"""Set the policy flag for making working trees when creating branches.
538
This only applies to branches that use this repository.
540
The default is 'True'.
541
:param new_value: True to restore the default, False to disable making
544
# FIXME: split out into a new class/strategy ?
545
if isinstance(self._format, (RepositoryFormat4,
548
raise NotImplementedError(self.set_make_working_trees)
551
self.control_files._transport.delete('no-working-trees')
552
except errors.NoSuchFile:
555
self.control_files.put_utf8('no-working-trees', '')
557
def make_working_trees(self):
558
"""Returns the policy for making working trees on new branches."""
559
# FIXME: split out into a new class/strategy ?
560
if isinstance(self._format, (RepositoryFormat4,
564
return not self.control_files._transport.has('no-working-trees')
567
def sign_revision(self, revision_id, gpg_strategy):
568
plaintext = Testament.from_revision(self, revision_id).as_short_text()
569
self.store_revision_signature(gpg_strategy, plaintext, revision_id)
572
class AllInOneRepository(Repository):
573
"""Legacy support - the repository behaviour for all-in-one branches."""
575
def __init__(self, _format, a_bzrdir, revision_store):
576
# we reuse one control files instance.
577
dir_mode = a_bzrdir._control_files._dir_mode
578
file_mode = a_bzrdir._control_files._file_mode
580
def get_weave(name, prefixed=False):
582
name = safe_unicode(name)
585
relpath = a_bzrdir._control_files._escape(name)
586
weave_transport = a_bzrdir._control_files._transport.clone(relpath)
587
ws = WeaveStore(weave_transport, prefixed=prefixed,
590
if a_bzrdir._control_files._transport.should_cache():
591
ws.enable_cache = True
594
def get_store(name, compressed=True, prefixed=False):
595
# FIXME: This approach of assuming stores are all entirely compressed
596
# or entirely uncompressed is tidy, but breaks upgrade from
597
# some existing branches where there's a mixture; we probably
598
# still want the option to look for both.
599
relpath = a_bzrdir._control_files._escape(name)
600
store = TextStore(a_bzrdir._control_files._transport.clone(relpath),
601
prefixed=prefixed, compressed=compressed,
604
#if self._transport.should_cache():
605
# cache_path = os.path.join(self.cache_root, name)
606
# os.mkdir(cache_path)
607
# store = bzrlib.store.CachedStore(store, cache_path)
610
# not broken out yet because the controlweaves|inventory_store
611
# and text_store | weave_store bits are still different.
612
if isinstance(_format, RepositoryFormat4):
613
self.inventory_store = get_store('inventory-store')
614
self.text_store = get_store('text-store')
615
elif isinstance(_format, RepositoryFormat5):
616
self.control_weaves = get_weave('')
617
self.weave_store = get_weave('weaves')
618
elif isinstance(_format, RepositoryFormat6):
619
self.control_weaves = get_weave('')
620
self.weave_store = get_weave('weaves', prefixed=True)
622
raise errors.BzrError('unreachable code: unexpected repository'
624
revision_store.register_suffix('sig')
625
super(AllInOneRepository, self).__init__(_format, a_bzrdir, a_bzrdir._control_files, revision_store)
628
class MetaDirRepository(Repository):
629
"""Repositories in the new meta-dir layout."""
631
def __init__(self, _format, a_bzrdir, control_files, revision_store):
632
super(MetaDirRepository, self).__init__(_format,
637
dir_mode = self.control_files._dir_mode
638
file_mode = self.control_files._file_mode
640
def get_weave(name, prefixed=False):
642
name = safe_unicode(name)
645
relpath = self.control_files._escape(name)
646
weave_transport = self.control_files._transport.clone(relpath)
647
ws = WeaveStore(weave_transport, prefixed=prefixed,
650
if self.control_files._transport.should_cache():
651
ws.enable_cache = True
654
if isinstance(self._format, RepositoryFormat7):
655
self.control_weaves = get_weave('')
656
self.weave_store = get_weave('weaves', prefixed=True)
657
elif isinstance(self._format, RepositoryFormatKnit1):
658
self.control_weaves = get_weave('')
659
self.weave_store = get_weave('knits', prefixed=True)
661
raise errors.BzrError('unreachable code: unexpected repository'
665
class RepositoryFormat(object):
666
"""A repository format.
668
Formats provide three things:
669
* An initialization routine to construct repository data on disk.
670
* a format string which is used when the BzrDir supports versioned
672
* an open routine which returns a Repository instance.
674
Formats are placed in an dict by their format string for reference
675
during opening. These should be subclasses of RepositoryFormat
678
Once a format is deprecated, just deprecate the initialize and open
679
methods on the format class. Do not deprecate the object, as the
680
object will be created every system load.
682
Common instance attributes:
683
_matchingbzrdir - the bzrdir format that the repository format was
684
originally written to work with. This can be used if manually
685
constructing a bzrdir and repository, or more commonly for test suite
689
_default_format = None
690
"""The default format used for new repositories."""
693
"""The known formats."""
696
def find_format(klass, a_bzrdir):
697
"""Return the format for the repository object in a_bzrdir."""
699
transport = a_bzrdir.get_repository_transport(None)
700
format_string = transport.get("format").read()
701
return klass._formats[format_string]
702
except errors.NoSuchFile:
703
raise errors.NoRepositoryPresent(a_bzrdir)
705
raise errors.UnknownFormatError(format_string)
708
def get_default_format(klass):
709
"""Return the current default format."""
710
return klass._default_format
712
def get_format_string(self):
713
"""Return the ASCII format string that identifies this format.
715
Note that in pre format ?? repositories the format string is
716
not permitted nor written to disk.
718
raise NotImplementedError(self.get_format_string)
720
def _get_revision_store(self, repo_transport, control_files):
721
"""Return the revision store object for this a_bzrdir."""
722
raise NotImplementedError(self._get_revision_store)
724
def _get_rev_store(self,
730
"""Common logic for getting a revision store for a repository.
732
see self._get_revision_store for the method to
733
get the store for a repository.
736
name = safe_unicode(name)
739
dir_mode = control_files._dir_mode
740
file_mode = control_files._file_mode
741
revision_store =TextStore(transport.clone(name),
743
compressed=compressed,
746
revision_store.register_suffix('sig')
747
return revision_store
749
def initialize(self, a_bzrdir, shared=False):
750
"""Initialize a repository of this format in a_bzrdir.
752
:param a_bzrdir: The bzrdir to put the new repository in it.
753
:param shared: The repository should be initialized as a sharable one.
755
This may raise UninitializableFormat if shared repository are not
756
compatible the a_bzrdir.
759
def is_supported(self):
760
"""Is this format supported?
762
Supported formats must be initializable and openable.
763
Unsupported formats may not support initialization or committing or
764
some other features depending on the reason for not being supported.
768
def open(self, a_bzrdir, _found=False):
769
"""Return an instance of this format for the bzrdir a_bzrdir.
771
_found is a private parameter, do not use it.
773
raise NotImplementedError(self.open)
776
def register_format(klass, format):
777
klass._formats[format.get_format_string()] = format
780
def set_default_format(klass, format):
781
klass._default_format = format
784
def unregister_format(klass, format):
785
assert klass._formats[format.get_format_string()] is format
786
del klass._formats[format.get_format_string()]
789
class PreSplitOutRepositoryFormat(RepositoryFormat):
790
"""Base class for the pre split out repository formats."""
792
def initialize(self, a_bzrdir, shared=False, _internal=False):
793
"""Create a weave repository.
795
TODO: when creating split out bzr branch formats, move this to a common
796
base for Format5, Format6. or something like that.
798
from bzrlib.weavefile import write_weave_v5
799
from bzrlib.weave import Weave
802
raise errors.IncompatibleFormat(self, a_bzrdir._format)
805
# always initialized when the bzrdir is.
806
return self.open(a_bzrdir, _found=True)
808
# Create an empty weave
810
bzrlib.weavefile.write_weave_v5(Weave(), sio)
811
empty_weave = sio.getvalue()
813
mutter('creating repository in %s.', a_bzrdir.transport.base)
814
dirs = ['revision-store', 'weaves']
815
lock_file = 'branch-lock'
816
files = [('inventory.weave', StringIO(empty_weave)),
819
# FIXME: RBC 20060125 dont peek under the covers
820
# NB: no need to escape relative paths that are url safe.
821
control_files = LockableFiles(a_bzrdir.transport, 'branch-lock')
822
control_files.lock_write()
823
control_files._transport.mkdir_multi(dirs,
824
mode=control_files._dir_mode)
826
for file, content in files:
827
control_files.put(file, content)
829
control_files.unlock()
830
return self.open(a_bzrdir, _found=True)
832
def open(self, a_bzrdir, _found=False):
833
"""See RepositoryFormat.open()."""
835
# we are being called directly and must probe.
836
raise NotImplementedError
838
repo_transport = a_bzrdir.get_repository_transport(None)
839
control_files = a_bzrdir._control_files
840
revision_store = self._get_revision_store(repo_transport, control_files)
841
return AllInOneRepository(_format=self,
843
revision_store=revision_store)
846
class RepositoryFormat4(PreSplitOutRepositoryFormat):
847
"""Bzr repository format 4.
849
This repository format has:
851
- TextStores for texts, inventories,revisions.
853
This format is deprecated: it indexes texts using a text id which is
854
removed in format 5; initializationa and write support for this format
859
super(RepositoryFormat4, self).__init__()
860
self._matchingbzrdir = bzrlib.bzrdir.BzrDirFormat4()
862
def initialize(self, url, shared=False, _internal=False):
863
"""Format 4 branches cannot be created."""
864
raise errors.UninitializableFormat(self)
866
def is_supported(self):
867
"""Format 4 is not supported.
869
It is not supported because the model changed from 4 to 5 and the
870
conversion logic is expensive - so doing it on the fly was not
875
def _get_revision_store(self, repo_transport, control_files):
876
"""See RepositoryFormat._get_revision_store()."""
877
return self._get_rev_store(repo_transport,
882
class RepositoryFormat5(PreSplitOutRepositoryFormat):
883
"""Bzr control format 5.
885
This repository format has:
886
- weaves for file texts and inventory
888
- TextStores for revisions and signatures.
892
super(RepositoryFormat5, self).__init__()
893
self._matchingbzrdir = bzrlib.bzrdir.BzrDirFormat5()
895
def _get_revision_store(self, repo_transport, control_files):
896
"""See RepositoryFormat._get_revision_store()."""
897
"""Return the revision store object for this a_bzrdir."""
898
return self._get_rev_store(repo_transport,
904
class RepositoryFormat6(PreSplitOutRepositoryFormat):
905
"""Bzr control format 6.
907
This repository format has:
908
- weaves for file texts and inventory
909
- hash subdirectory based stores.
910
- TextStores for revisions and signatures.
914
super(RepositoryFormat6, self).__init__()
915
self._matchingbzrdir = bzrlib.bzrdir.BzrDirFormat6()
917
def _get_revision_store(self, repo_transport, control_files):
918
"""See RepositoryFormat._get_revision_store()."""
919
return self._get_rev_store(repo_transport,
926
class MetaDirRepositoryFormat(RepositoryFormat):
927
"""Common base class for the new repositories using the metadir layour."""
930
super(MetaDirRepositoryFormat, self).__init__()
931
self._matchingbzrdir = bzrlib.bzrdir.BzrDirMetaFormat1()
933
def _create_control_files(self, a_bzrdir):
934
"""Create the required files and the initial control_files object."""
935
# FIXME: RBC 20060125 dont peek under the covers
936
# NB: no need to escape relative paths that are url safe.
938
repository_transport = a_bzrdir.get_repository_transport(self)
939
repository_transport.put(lock_file, StringIO()) # TODO get the file mode from the bzrdir lock files., mode=file_mode)
940
control_files = LockableFiles(repository_transport, 'lock')
943
def _get_revision_store(self, repo_transport, control_files):
944
"""See RepositoryFormat._get_revision_store()."""
945
return self._get_rev_store(repo_transport,
952
def open(self, a_bzrdir, _found=False, _override_transport=None):
953
"""See RepositoryFormat.open().
955
:param _override_transport: INTERNAL USE ONLY. Allows opening the
956
repository at a slightly different url
957
than normal. I.e. during 'upgrade'.
960
format = RepositoryFormat.find_format(a_bzrdir)
961
assert format.__class__ == self.__class__
962
if _override_transport is not None:
963
repo_transport = _override_transport
965
repo_transport = a_bzrdir.get_repository_transport(None)
966
control_files = LockableFiles(repo_transport, 'lock')
967
revision_store = self._get_revision_store(repo_transport, control_files)
968
return MetaDirRepository(_format=self,
970
control_files=control_files,
971
revision_store=revision_store)
973
def _upload_blank_content(self, a_bzrdir, dirs, files, utf8_files, shared):
974
"""Upload the initial blank content."""
975
control_files = self._create_control_files(a_bzrdir)
976
control_files.lock_write()
977
control_files._transport.mkdir_multi(dirs,
978
mode=control_files._dir_mode)
980
for file, content in files:
981
control_files.put(file, content)
982
for file, content in utf8_files:
983
control_files.put_utf8(file, content)
985
control_files.put_utf8('shared-storage', '')
987
control_files.unlock()
990
class RepositoryFormat7(MetaDirRepositoryFormat):
993
This repository format has:
994
- weaves for file texts and inventory
995
- hash subdirectory based stores.
996
- TextStores for revisions and signatures.
997
- a format marker of its own
998
- an optional 'shared-storage' flag
999
- an optional 'no-working-trees' flag
1002
def get_format_string(self):
1003
"""See RepositoryFormat.get_format_string()."""
1004
return "Bazaar-NG Repository format 7"
1006
def initialize(self, a_bzrdir, shared=False):
1007
"""Create a weave repository.
1009
:param shared: If true the repository will be initialized as a shared
1012
from bzrlib.weavefile import write_weave_v5
1013
from bzrlib.weave import Weave
1015
# Create an empty weave
1017
bzrlib.weavefile.write_weave_v5(Weave(), sio)
1018
empty_weave = sio.getvalue()
1020
mutter('creating repository in %s.', a_bzrdir.transport.base)
1021
dirs = ['revision-store', 'weaves']
1022
files = [('inventory.weave', StringIO(empty_weave)),
1024
utf8_files = [('format', self.get_format_string())]
1026
self._upload_blank_content(a_bzrdir, dirs, files, utf8_files, shared)
1027
return self.open(a_bzrdir=a_bzrdir, _found=True)
1030
class RepositoryFormatKnit1(MetaDirRepositoryFormat):
1031
"""Bzr repository knit format 1.
1033
This repository format has:
1034
- knits for file texts and inventory
1035
- hash subdirectory based stores.
1036
- knits for revisions and signatures
1037
- TextStores for revisions and signatures.
1038
- a format marker of its own
1039
- an optional 'shared-storage' flag
1040
- an optional 'no-working-trees' flag
1043
def get_format_string(self):
1044
"""See RepositoryFormat.get_format_string()."""
1045
return "Bazaar-NG Knit Repository Format 1"
1047
def initialize(self, a_bzrdir, shared=False):
1048
"""Create a knit format 1 repository.
1050
:param shared: If true the repository will be initialized as a shared
1052
XXX NOTE that this current uses a Weave for testing and will become
1053
A Knit in due course.
1055
from bzrlib.weavefile import write_weave_v5
1056
from bzrlib.weave import Weave
1058
# Create an empty weave
1060
bzrlib.weavefile.write_weave_v5(Weave(), sio)
1061
empty_weave = sio.getvalue()
1063
mutter('creating repository in %s.', a_bzrdir.transport.base)
1064
dirs = ['revision-store', 'knits']
1065
files = [('inventory.weave', StringIO(empty_weave)),
1067
utf8_files = [('format', self.get_format_string())]
1069
self._upload_blank_content(a_bzrdir, dirs, files, utf8_files, shared)
1070
return self.open(a_bzrdir=a_bzrdir, _found=True)
1073
# formats which have no format string are not discoverable
1074
# and not independently creatable, so are not registered.
1075
_default_format = RepositoryFormat7()
1076
RepositoryFormat.register_format(_default_format)
1077
RepositoryFormat.register_format(RepositoryFormatKnit1())
1078
RepositoryFormat.set_default_format(_default_format)
1079
_legacy_formats = [RepositoryFormat4(),
1080
RepositoryFormat5(),
1081
RepositoryFormat6()]
1084
class InterRepository(object):
1085
"""This class represents operations taking place between two repositories.
1087
Its instances have methods like copy_content and fetch, and contain
1088
references to the source and target repositories these operations can be
1091
Often we will provide convenience methods on 'repository' which carry out
1092
operations with another repository - they will always forward to
1093
InterRepository.get(other).method_name(parameters).
1095
# XXX: FIXME: FUTURE: robertc
1096
# testing of these probably requires a factory in optimiser type, and
1097
# then a test adapter to test each type thoroughly.
1101
"""The available optimised InterRepository types."""
1103
def __init__(self, source, target):
1104
"""Construct a default InterRepository instance. Please use 'get'.
1106
Only subclasses of InterRepository should call
1107
InterRepository.__init__ - clients should call InterRepository.get
1108
instead which will create an optimised InterRepository if possible.
1110
self.source = source
1111
self.target = target
1114
def copy_content(self, revision_id=None, basis=None):
1115
"""Make a complete copy of the content in self into destination.
1117
This is a destructive operation! Do not use it on existing
1120
:param revision_id: Only copy the content needed to construct
1121
revision_id and its parents.
1122
:param basis: Copy the needed data preferentially from basis.
1125
self.target.set_make_working_trees(self.source.make_working_trees())
1126
except NotImplementedError:
1128
# grab the basis available data
1129
if basis is not None:
1130
self.target.fetch(basis, revision_id=revision_id)
1131
# but dont both fetching if we have the needed data now.
1132
if (revision_id not in (None, NULL_REVISION) and
1133
self.target.has_revision(revision_id)):
1135
self.target.fetch(self.source, revision_id=revision_id)
1137
def _double_lock(self, lock_source, lock_target):
1138
"""Take out too locks, rolling back the first if the second throws."""
1143
# we want to ensure that we don't leave source locked by mistake.
1144
# and any error on target should not confuse source.
1145
self.source.unlock()
1149
def fetch(self, revision_id=None, pb=None):
1150
"""Fetch the content required to construct revision_id.
1152
The content is copied from source to target.
1154
:param revision_id: if None all content is copied, if NULL_REVISION no
1156
:param pb: optional progress bar to use for progress reports. If not
1157
provided a default one will be created.
1159
Returns the copied revision count and the failed revisions in a tuple:
1162
from bzrlib.fetch import RepoFetcher
1163
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
1164
self.source, self.source._format, self.target, self.target._format)
1165
f = RepoFetcher(to_repository=self.target,
1166
from_repository=self.source,
1167
last_revision=revision_id,
1169
return f.count_copied, f.failed_revisions
1172
def get(klass, repository_source, repository_target):
1173
"""Retrieve a InterRepository worker object for these repositories.
1175
:param repository_source: the repository to be the 'source' member of
1176
the InterRepository instance.
1177
:param repository_target: the repository to be the 'target' member of
1178
the InterRepository instance.
1179
If an optimised InterRepository worker exists it will be used otherwise
1180
a default InterRepository instance will be created.
1182
for provider in klass._optimisers:
1183
if provider.is_compatible(repository_source, repository_target):
1184
return provider(repository_source, repository_target)
1185
return InterRepository(repository_source, repository_target)
1187
def lock_read(self):
1188
"""Take out a logical read lock.
1190
This will lock the source branch and the target branch. The source gets
1191
a read lock and the target a read lock.
1193
self._double_lock(self.source.lock_read, self.target.lock_read)
1195
def lock_write(self):
1196
"""Take out a logical write lock.
1198
This will lock the source branch and the target branch. The source gets
1199
a read lock and the target a write lock.
1201
self._double_lock(self.source.lock_read, self.target.lock_write)
1204
def missing_revision_ids(self, revision_id=None):
1205
"""Return the revision ids that source has that target does not.
1207
These are returned in topological order.
1209
:param revision_id: only return revision ids included by this
1212
# generic, possibly worst case, slow code path.
1213
target_ids = set(self.target.all_revision_ids())
1214
if revision_id is not None:
1215
source_ids = self.source.get_ancestry(revision_id)
1216
assert source_ids.pop(0) == None
1218
source_ids = self.source.all_revision_ids()
1219
result_set = set(source_ids).difference(target_ids)
1220
# this may look like a no-op: its not. It preserves the ordering
1221
# other_ids had while only returning the members from other_ids
1222
# that we've decided we need.
1223
return [rev_id for rev_id in source_ids if rev_id in result_set]
1226
def register_optimiser(klass, optimiser):
1227
"""Register an InterRepository optimiser."""
1228
klass._optimisers.add(optimiser)
1231
"""Release the locks on source and target."""
1233
self.target.unlock()
1235
self.source.unlock()
1238
def unregister_optimiser(klass, optimiser):
1239
"""Unregister an InterRepository optimiser."""
1240
klass._optimisers.remove(optimiser)
1243
class InterWeaveRepo(InterRepository):
1244
"""Optimised code paths between Weave based repositories."""
1246
_matching_repo_format = _default_format
1247
"""Repository format for testing with."""
1250
def is_compatible(source, target):
1251
"""Be compatible with known Weave formats.
1253
We dont test for the stores being of specific types becase that
1254
could lead to confusing results, and there is no need to be
1258
return (isinstance(source._format, (RepositoryFormat5,
1260
RepositoryFormat7)) and
1261
isinstance(target._format, (RepositoryFormat5,
1263
RepositoryFormat7)))
1264
except AttributeError:
1268
def copy_content(self, revision_id=None, basis=None):
1269
"""See InterRepository.copy_content()."""
1270
# weave specific optimised path:
1271
if basis is not None:
1272
# copy the basis in, then fetch remaining data.
1273
basis.copy_content_into(self.target, revision_id)
1274
# the basis copy_content_into could misset this.
1276
self.target.set_make_working_trees(self.source.make_working_trees())
1277
except NotImplementedError:
1279
self.target.fetch(self.source, revision_id=revision_id)
1282
self.target.set_make_working_trees(self.source.make_working_trees())
1283
except NotImplementedError:
1285
# FIXME do not peek!
1286
if self.source.control_files._transport.listable():
1287
pb = bzrlib.ui.ui_factory.progress_bar()
1288
copy_all(self.source.weave_store,
1289
self.target.weave_store, pb=pb)
1290
pb.update('copying inventory', 0, 1)
1291
self.target.control_weaves.copy_multi(
1292
self.source.control_weaves, ['inventory'])
1293
copy_all(self.source.revision_store,
1294
self.target.revision_store, pb=pb)
1296
self.target.fetch(self.source, revision_id=revision_id)
1299
def fetch(self, revision_id=None, pb=None):
1300
"""See InterRepository.fetch()."""
1301
from bzrlib.fetch import RepoFetcher
1302
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
1303
self.source, self.source._format, self.target, self.target._format)
1304
f = RepoFetcher(to_repository=self.target,
1305
from_repository=self.source,
1306
last_revision=revision_id,
1308
return f.count_copied, f.failed_revisions
1311
def missing_revision_ids(self, revision_id=None):
1312
"""See InterRepository.missing_revision_ids()."""
1313
# we want all revisions to satisfy revision_id in source.
1314
# but we dont want to stat every file here and there.
1315
# we want then, all revisions other needs to satisfy revision_id
1316
# checked, but not those that we have locally.
1317
# so the first thing is to get a subset of the revisions to
1318
# satisfy revision_id in source, and then eliminate those that
1319
# we do already have.
1320
# this is slow on high latency connection to self, but as as this
1321
# disk format scales terribly for push anyway due to rewriting
1322
# inventory.weave, this is considered acceptable.
1324
if revision_id is not None:
1325
source_ids = self.source.get_ancestry(revision_id)
1326
assert source_ids.pop(0) == None
1328
source_ids = self.source._all_possible_ids()
1329
source_ids_set = set(source_ids)
1330
# source_ids is the worst possible case we may need to pull.
1331
# now we want to filter source_ids against what we actually
1332
# have in target, but dont try to check for existence where we know
1333
# we do not have a revision as that would be pointless.
1334
target_ids = set(self.target._all_possible_ids())
1335
possibly_present_revisions = target_ids.intersection(source_ids_set)
1336
actually_present_revisions = set(self.target._eliminate_revisions_not_present(possibly_present_revisions))
1337
required_revisions = source_ids_set.difference(actually_present_revisions)
1338
required_topo_revisions = [rev_id for rev_id in source_ids if rev_id in required_revisions]
1339
if revision_id is not None:
1340
# we used get_ancestry to determine source_ids then we are assured all
1341
# revisions referenced are present as they are installed in topological order.
1342
# and the tip revision was validated by get_ancestry.
1343
return required_topo_revisions
1345
# if we just grabbed the possibly available ids, then
1346
# we only have an estimate of whats available and need to validate
1347
# that against the revision records.
1348
return self.source._eliminate_revisions_not_present(required_topo_revisions)
1351
InterRepository.register_optimiser(InterWeaveRepo)
1354
class RepositoryTestProviderAdapter(object):
1355
"""A tool to generate a suite testing multiple repository formats at once.
1357
This is done by copying the test once for each transport and injecting
1358
the transport_server, transport_readonly_server, and bzrdir_format and
1359
repository_format classes into each copy. Each copy is also given a new id()
1360
to make it easy to identify.
1363
def __init__(self, transport_server, transport_readonly_server, formats):
1364
self._transport_server = transport_server
1365
self._transport_readonly_server = transport_readonly_server
1366
self._formats = formats
1368
def adapt(self, test):
1369
result = TestSuite()
1370
for repository_format, bzrdir_format in self._formats:
1371
new_test = deepcopy(test)
1372
new_test.transport_server = self._transport_server
1373
new_test.transport_readonly_server = self._transport_readonly_server
1374
new_test.bzrdir_format = bzrdir_format
1375
new_test.repository_format = repository_format
1376
def make_new_test_id():
1377
new_id = "%s(%s)" % (new_test.id(), repository_format.__class__.__name__)
1378
return lambda: new_id
1379
new_test.id = make_new_test_id()
1380
result.addTest(new_test)
1384
class InterRepositoryTestProviderAdapter(object):
1385
"""A tool to generate a suite testing multiple inter repository formats.
1387
This is done by copying the test once for each interrepo provider and injecting
1388
the transport_server, transport_readonly_server, repository_format and
1389
repository_to_format classes into each copy.
1390
Each copy is also given a new id() to make it easy to identify.
1393
def __init__(self, transport_server, transport_readonly_server, formats):
1394
self._transport_server = transport_server
1395
self._transport_readonly_server = transport_readonly_server
1396
self._formats = formats
1398
def adapt(self, test):
1399
result = TestSuite()
1400
for interrepo_class, repository_format, repository_format_to in self._formats:
1401
new_test = deepcopy(test)
1402
new_test.transport_server = self._transport_server
1403
new_test.transport_readonly_server = self._transport_readonly_server
1404
new_test.interrepo_class = interrepo_class
1405
new_test.repository_format = repository_format
1406
new_test.repository_format_to = repository_format_to
1407
def make_new_test_id():
1408
new_id = "%s(%s)" % (new_test.id(), interrepo_class.__name__)
1409
return lambda: new_id
1410
new_test.id = make_new_test_id()
1411
result.addTest(new_test)
1415
def default_test_list():
1416
"""Generate the default list of interrepo permutations to test."""
1418
# test the default InterRepository between format 6 and the current
1420
# XXX: robertc 20060220 reinstate this when there are two supported
1421
# formats which do not have an optimal code path between them.
1422
result.append((InterRepository,
1423
RepositoryFormat6(),
1424
RepositoryFormatKnit1()))
1425
for optimiser in InterRepository._optimisers:
1426
result.append((optimiser,
1427
optimiser._matching_repo_format,
1428
optimiser._matching_repo_format
1430
# if there are specific combinations we want to use, we can add them
1435
class CopyConverter(object):
1436
"""A repository conversion tool which just performs a copy of the content.
1438
This is slow but quite reliable.
1441
def __init__(self, target_format):
1442
"""Create a CopyConverter.
1444
:param target_format: The format the resulting repository should be.
1446
self.target_format = target_format
1448
def convert(self, repo, pb):
1449
"""Perform the conversion of to_convert, giving feedback via pb.
1451
:param to_convert: The disk object to convert.
1452
:param pb: a progress bar to use for progress information.
1457
# this is only useful with metadir layouts - separated repo content.
1458
# trigger an assertion if not such
1459
repo._format.get_format_string()
1460
self.repo_dir = repo.bzrdir
1461
self.step('Moving repository to repository.backup')
1462
self.repo_dir.transport.move('repository', 'repository.backup')
1463
backup_transport = self.repo_dir.transport.clone('repository.backup')
1464
self.source_repo = repo._format.open(self.repo_dir,
1466
_override_transport=backup_transport)
1467
self.step('Creating new repository')
1468
converted = self.target_format.initialize(self.repo_dir,
1469
self.source_repo.is_shared())
1470
converted.lock_write()
1472
self.step('Copying content into repository.')
1473
self.source_repo.copy_content_into(converted)
1476
self.step('Deleting old repository content.')
1477
self.repo_dir.transport.delete_tree('repository.backup')
1478
self.pb.note('repository converted')
1480
def step(self, message):
1481
"""Update the pb by a step."""
1483
self.pb.update(message, self.count, self.total)