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.graph import Graph
26
from bzrlib.lockable_files import LockableFiles, TransportLock
27
from bzrlib.lockdir import LockDir
28
from bzrlib.osutils import safe_unicode
29
from bzrlib.revision import NULL_REVISION
30
import bzrlib.errors as errors
31
import bzrlib.gpg as gpg
32
from bzrlib.store import copy_all
33
from bzrlib.store.weave import WeaveStore
34
from bzrlib.store.text import TextStore
35
from bzrlib.symbol_versioning import *
36
from bzrlib.trace import mutter
37
from bzrlib.tree import RevisionTree
38
from bzrlib.testament import Testament
39
from bzrlib.tree import EmptyTree
44
class Repository(object):
45
"""Repository holding history for one or more branches.
47
The repository holds and retrieves historical information including
48
revisions and file history. It's normally accessed only by the Branch,
49
which views a particular line of development through that history.
51
The Repository builds on top of Stores and a Transport, which respectively
52
describe the disk data format and the way of accessing the (possibly
57
def add_inventory(self, revid, inv, parents):
58
"""Add the inventory inv to the repository as revid.
60
:param parents: The revision ids of the parents that revid
61
is known to have and are in the repository already.
63
returns the sha1 of the serialized inventory.
65
inv_text = bzrlib.xml5.serializer_v5.write_inventory_to_string(inv)
66
inv_sha1 = bzrlib.osutils.sha_string(inv_text)
67
self.control_weaves.add_text('inventory', revid,
68
bzrlib.osutils.split_lines(inv_text), parents,
69
self.get_transaction())
73
def add_revision(self, rev_id, rev, inv=None, config=None):
74
"""Add rev to the revision store as rev_id.
76
:param rev_id: the revision id to use.
77
:param rev: The revision object.
78
:param inv: The inventory for the revision. if None, it will be looked
79
up in the inventory storer
80
:param config: If None no digital signature will be created.
81
If supplied its signature_needed method will be used
82
to determine if a signature should be made.
84
if config is not None and config.signature_needed():
86
inv = self.get_inventory(rev_id)
87
plaintext = Testament(rev, inv).as_short_text()
88
self.store_revision_signature(
89
gpg.GPGStrategy(config), plaintext, rev_id)
90
if not rev_id in self.get_inventory_weave():
92
raise errors.WeaveRevisionNotPresent(rev_id,
93
self.get_inventory_weave())
95
# yes, this is not suitable for adding with ghosts.
96
self.add_inventory(rev_id, inv, rev.parent_ids)
99
bzrlib.xml5.serializer_v5.write_revision(rev, rev_tmp)
101
self.revision_store.add(rev_tmp, rev_id)
102
mutter('added revision_id {%s}', rev_id)
105
def _all_possible_ids(self):
106
"""Return all the possible revisions that we could find."""
107
return self.get_inventory_weave().names()
110
def all_revision_ids(self):
111
"""Returns a list of all the revision ids in the repository.
113
These are in as much topological order as the underlying store can
114
present: for weaves ghosts may lead to a lack of correctness until
115
the reweave updates the parents list.
117
result = self._all_possible_ids()
118
return self._eliminate_revisions_not_present(result)
121
def _eliminate_revisions_not_present(self, revision_ids):
122
"""Check every revision id in revision_ids to see if we have it.
124
Returns a set of the present revisions.
127
for id in revision_ids:
128
if self.has_revision(id):
133
def create(a_bzrdir):
134
"""Construct the current default format repository in a_bzrdir."""
135
return RepositoryFormat.get_default_format().initialize(a_bzrdir)
137
def __init__(self, _format, a_bzrdir, control_files, revision_store):
138
"""instantiate a Repository.
140
:param _format: The format of the repository on disk.
141
:param a_bzrdir: The BzrDir of the repository.
143
In the future we will have a single api for all stores for
144
getting file texts, inventories and revisions, then
145
this construct will accept instances of those things.
147
object.__init__(self)
148
self._format = _format
149
# the following are part of the public API for Repository:
150
self.bzrdir = a_bzrdir
151
self.control_files = control_files
152
self.revision_store = revision_store
154
def lock_write(self):
155
self.control_files.lock_write()
158
self.control_files.lock_read()
161
return self.control_files.is_locked()
164
def missing_revision_ids(self, other, revision_id=None):
165
"""Return the revision ids that other has that this does not.
167
These are returned in topological order.
169
revision_id: only return revision ids included by revision_id.
171
return InterRepository.get(other, self).missing_revision_ids(revision_id)
175
"""Open the repository rooted at base.
177
For instance, if the repository is at URL/.bzr/repository,
178
Repository.open(URL) -> a Repository instance.
180
control = bzrlib.bzrdir.BzrDir.open(base)
181
return control.open_repository()
183
def copy_content_into(self, destination, revision_id=None, basis=None):
184
"""Make a complete copy of the content in self into destination.
186
This is a destructive operation! Do not use it on existing
189
return InterRepository.get(self, destination).copy_content(revision_id, basis)
191
def fetch(self, source, revision_id=None, pb=None):
192
"""Fetch the content required to construct revision_id from source.
194
If revision_id is None all content is copied.
196
return InterRepository.get(source, self).fetch(revision_id=revision_id,
200
self.control_files.unlock()
203
def clone(self, a_bzrdir, revision_id=None, basis=None):
204
"""Clone this repository into a_bzrdir using the current format.
206
Currently no check is made that the format of this repository and
207
the bzrdir format are compatible. FIXME RBC 20060201.
209
if not isinstance(a_bzrdir._format, self.bzrdir._format.__class__):
210
# use target default format.
211
result = a_bzrdir.create_repository()
212
# FIXME RBC 20060209 split out the repository type to avoid this check ?
213
elif isinstance(a_bzrdir._format,
214
(bzrlib.bzrdir.BzrDirFormat4,
215
bzrlib.bzrdir.BzrDirFormat5,
216
bzrlib.bzrdir.BzrDirFormat6)):
217
result = a_bzrdir.open_repository()
219
result = self._format.initialize(a_bzrdir, shared=self.is_shared())
220
self.copy_content_into(result, revision_id, basis)
223
def has_revision(self, revision_id):
224
"""True if this branch has a copy of the revision.
226
This does not necessarily imply the revision is merge
227
or on the mainline."""
228
return (revision_id is None
229
or self.revision_store.has_id(revision_id))
232
def get_revision_xml_file(self, revision_id):
233
"""Return XML file object for revision object."""
234
if not revision_id or not isinstance(revision_id, basestring):
235
raise InvalidRevisionId(revision_id=revision_id, branch=self)
237
return self.revision_store.get(revision_id)
238
except (IndexError, KeyError):
239
raise bzrlib.errors.NoSuchRevision(self, revision_id)
242
def get_revision_xml(self, revision_id):
243
return self.get_revision_xml_file(revision_id).read()
246
def get_revision_reconcile(self, revision_id):
247
"""'reconcile' helper routine that allows access to a revision always.
249
This variant of get_revision does not cross check the weave graph
250
against the revision one as get_revision does: but it should only
251
be used by reconcile, or reconcile-alike commands that are correcting
252
or testing the revision graph.
254
xml_file = self.get_revision_xml_file(revision_id)
257
r = bzrlib.xml5.serializer_v5.read_revision(xml_file)
258
except SyntaxError, e:
259
raise bzrlib.errors.BzrError('failed to unpack revision_xml',
263
assert r.revision_id == revision_id
267
def get_revision(self, revision_id):
268
"""Return the Revision object for a named revision"""
269
r = self.get_revision_reconcile(revision_id)
270
# weave corruption can lead to absent revision markers that should be
272
# the following test is reasonably cheap (it needs a single weave read)
273
# and the weave is cached in read transactions. In write transactions
274
# it is not cached but typically we only read a small number of
275
# revisions. For knits when they are introduced we will probably want
276
# to ensure that caching write transactions are in use.
277
inv = self.get_inventory_weave()
278
self._check_revision_parents(r, inv)
281
def _check_revision_parents(self, revision, inventory):
282
"""Private to Repository and Fetch.
284
This checks the parentage of revision in an inventory weave for
285
consistency and is only applicable to inventory-weave-for-ancestry
286
using repository formats & fetchers.
288
weave_parents = inventory.parent_names(revision.revision_id)
289
weave_names = inventory.names()
290
for parent_id in revision.parent_ids:
291
if parent_id in weave_names:
292
# this parent must not be a ghost.
293
if not parent_id in weave_parents:
295
raise errors.CorruptRepository(self)
298
def get_revision_sha1(self, revision_id):
299
"""Hash the stored value of a revision, and return it."""
300
# In the future, revision entries will be signed. At that
301
# point, it is probably best *not* to include the signature
302
# in the revision hash. Because that lets you re-sign
303
# the revision, (add signatures/remove signatures) and still
304
# have all hash pointers stay consistent.
305
# But for now, just hash the contents.
306
return bzrlib.osutils.sha_file(self.get_revision_xml_file(revision_id))
309
def store_revision_signature(self, gpg_strategy, plaintext, revision_id):
310
self.revision_store.add(StringIO(gpg_strategy.sign(plaintext)),
313
def fileid_involved_between_revs(self, from_revid, to_revid):
314
"""Find file_id(s) which are involved in the changes between revisions.
316
This determines the set of revisions which are involved, and then
317
finds all file ids affected by those revisions.
319
# TODO: jam 20060119 This code assumes that w.inclusions will
320
# always be correct. But because of the presence of ghosts
321
# it is possible to be wrong.
322
# One specific example from Robert Collins:
323
# Two branches, with revisions ABC, and AD
324
# C is a ghost merge of D.
325
# Inclusions doesn't recognize D as an ancestor.
326
# If D is ever merged in the future, the weave
327
# won't be fixed, because AD never saw revision C
328
# to cause a conflict which would force a reweave.
329
w = self.get_inventory_weave()
330
from_set = set(w.inclusions([w.lookup(from_revid)]))
331
to_set = set(w.inclusions([w.lookup(to_revid)]))
332
included = to_set.difference(from_set)
333
changed = map(w.idx_to_name, included)
334
return self._fileid_involved_by_set(changed)
336
def fileid_involved(self, last_revid=None):
337
"""Find all file_ids modified in the ancestry of last_revid.
339
:param last_revid: If None, last_revision() will be used.
341
w = self.get_inventory_weave()
343
changed = set(w._names)
345
included = w.inclusions([w.lookup(last_revid)])
346
changed = map(w.idx_to_name, included)
347
return self._fileid_involved_by_set(changed)
349
def fileid_involved_by_set(self, changes):
350
"""Find all file_ids modified by the set of revisions passed in.
352
:param changes: A set() of revision ids
354
# TODO: jam 20060119 This line does *nothing*, remove it.
355
# or better yet, change _fileid_involved_by_set so
356
# that it takes the inventory weave, rather than
357
# pulling it out by itself.
358
return self._fileid_involved_by_set(changes)
360
def _fileid_involved_by_set(self, changes):
361
"""Find the set of file-ids affected by the set of revisions.
363
:param changes: A set() of revision ids.
364
:return: A set() of file ids.
366
This peaks at the Weave, interpreting each line, looking to
367
see if it mentions one of the revisions. And if so, includes
368
the file id mentioned.
369
This expects both the Weave format, and the serialization
370
to have a single line per file/directory, and to have
371
fileid="" and revision="" on that line.
373
assert isinstance(self._format, (RepositoryFormat5,
376
RepositoryFormatKnit1)), \
377
"fileid_involved only supported for branches which store inventory as unnested xml"
379
w = self.get_inventory_weave()
381
for line in w._weave:
383
# it is ugly, but it is due to the weave structure
384
if not isinstance(line, basestring): continue
386
start = line.find('file_id="')+9
387
if start < 9: continue
388
end = line.find('"', start)
390
file_id = xml.sax.saxutils.unescape(line[start:end])
392
# check if file_id is already present
393
if file_id in file_ids: continue
395
start = line.find('revision="')+10
396
if start < 10: continue
397
end = line.find('"', start)
399
revision_id = xml.sax.saxutils.unescape(line[start:end])
401
if revision_id in changes:
402
file_ids.add(file_id)
406
def get_inventory_weave(self):
407
return self.control_weaves.get_weave('inventory',
408
self.get_transaction())
411
def get_inventory(self, revision_id):
412
"""Get Inventory object by hash."""
413
xml = self.get_inventory_xml(revision_id)
414
return bzrlib.xml5.serializer_v5.read_inventory_from_string(xml)
417
def get_inventory_xml(self, revision_id):
418
"""Get inventory XML as a file object."""
420
assert isinstance(revision_id, basestring), type(revision_id)
421
iw = self.get_inventory_weave()
422
return iw.get_text(iw.lookup(revision_id))
424
raise bzrlib.errors.HistoryMissing(self, 'inventory', revision_id)
427
def get_inventory_sha1(self, revision_id):
428
"""Return the sha1 hash of the inventory entry
430
return self.get_revision(revision_id).inventory_sha1
433
def get_revision_graph(self, revision_id=None):
434
"""Return a dictionary containing the revision graph.
436
:return: a dictionary of revision_id->revision_parents_list.
438
weave = self.get_inventory_weave()
439
all_revisions = self._eliminate_revisions_not_present(weave.names())
440
entire_graph = dict([(node, weave.parent_names(node)) for
441
node in all_revisions])
442
if revision_id is None:
444
elif revision_id not in entire_graph:
445
raise errors.NoSuchRevision(self, revision_id)
447
# add what can be reached from revision_id
449
pending = set([revision_id])
450
while len(pending) > 0:
452
result[node] = entire_graph[node]
453
for revision_id in result[node]:
454
if revision_id not in result:
455
pending.add(revision_id)
459
def get_revision_graph_with_ghosts(self, revision_ids=None):
460
"""Return a graph of the revisions with ghosts marked as applicable.
462
:param revision_ids: an iterable of revisions to graph or None for all.
463
:return: a Graph object with the graph reachable from revision_ids.
467
pending = set(self.all_revision_ids())
470
pending = set(revision_ids)
471
required = set(revision_ids)
474
revision_id = pending.pop()
476
rev = self.get_revision(revision_id)
477
except errors.NoSuchRevision:
478
if revision_id in required:
481
result.add_ghost(revision_id)
483
for parent_id in rev.parent_ids:
484
# is this queued or done ?
485
if (parent_id not in pending and
486
parent_id not in done):
488
pending.add(parent_id)
489
result.add_node(revision_id, rev.parent_ids)
494
def get_revision_inventory(self, revision_id):
495
"""Return inventory of a past revision."""
496
# TODO: Unify this with get_inventory()
497
# bzr 0.0.6 and later imposes the constraint that the inventory_id
498
# must be the same as its revision, so this is trivial.
499
if revision_id is None:
500
# This does not make sense: if there is no revision,
501
# then it is the current tree inventory surely ?!
502
# and thus get_root_id() is something that looks at the last
503
# commit on the branch, and the get_root_id is an inventory check.
504
raise NotImplementedError
505
# return Inventory(self.get_root_id())
507
return self.get_inventory(revision_id)
511
"""Return True if this repository is flagged as a shared repository."""
512
# FIXME format 4-6 cannot be shared, this is technically faulty.
513
return self.control_files._transport.has('shared-storage')
516
def revision_tree(self, revision_id):
517
"""Return Tree for a revision on this branch.
519
`revision_id` may be None for the null revision, in which case
520
an `EmptyTree` is returned."""
521
# TODO: refactor this to use an existing revision object
522
# so we don't need to read it in twice.
523
if revision_id is None or revision_id == NULL_REVISION:
526
inv = self.get_revision_inventory(revision_id)
527
return RevisionTree(self, inv, revision_id)
530
def get_ancestry(self, revision_id):
531
"""Return a list of revision-ids integrated by a revision.
533
This is topologically sorted.
535
if revision_id is None:
537
if not self.has_revision(revision_id):
538
raise errors.NoSuchRevision(self, revision_id)
539
w = self.get_inventory_weave()
540
return [None] + map(w.idx_to_name,
541
w.inclusions([w.lookup(revision_id)]))
544
def print_file(self, file, revision_id):
545
"""Print `file` to stdout.
547
FIXME RBC 20060125 as John Meinel points out this is a bad api
548
- it writes to stdout, it assumes that that is valid etc. Fix
549
by creating a new more flexible convenience function.
551
tree = self.revision_tree(revision_id)
552
# use inventory as it was in that revision
553
file_id = tree.inventory.path2id(file)
555
raise BzrError("%r is not present in revision %s" % (file, revno))
557
revno = self.revision_id_to_revno(revision_id)
558
except errors.NoSuchRevision:
559
# TODO: This should not be BzrError,
560
# but NoSuchFile doesn't fit either
561
raise BzrError('%r is not present in revision %s'
562
% (file, revision_id))
564
raise BzrError('%r is not present in revision %s'
566
tree.print_file(file_id)
568
def get_transaction(self):
569
return self.control_files.get_transaction()
571
def revision_parents(self, revid):
572
return self.get_inventory_weave().parent_names(revid)
575
def set_make_working_trees(self, new_value):
576
"""Set the policy flag for making working trees when creating branches.
578
This only applies to branches that use this repository.
580
The default is 'True'.
581
:param new_value: True to restore the default, False to disable making
584
# FIXME: split out into a new class/strategy ?
585
if isinstance(self._format, (RepositoryFormat4,
588
raise NotImplementedError(self.set_make_working_trees)
591
self.control_files._transport.delete('no-working-trees')
592
except errors.NoSuchFile:
595
self.control_files.put_utf8('no-working-trees', '')
597
def make_working_trees(self):
598
"""Returns the policy for making working trees on new branches."""
599
# FIXME: split out into a new class/strategy ?
600
if isinstance(self._format, (RepositoryFormat4,
604
return not self.control_files._transport.has('no-working-trees')
607
def sign_revision(self, revision_id, gpg_strategy):
608
plaintext = Testament.from_revision(self, revision_id).as_short_text()
609
self.store_revision_signature(gpg_strategy, plaintext, revision_id)
612
class AllInOneRepository(Repository):
613
"""Legacy support - the repository behaviour for all-in-one branches."""
615
def __init__(self, _format, a_bzrdir, revision_store):
616
# we reuse one control files instance.
617
dir_mode = a_bzrdir._control_files._dir_mode
618
file_mode = a_bzrdir._control_files._file_mode
620
def get_weave(name, prefixed=False):
622
name = safe_unicode(name)
625
relpath = a_bzrdir._control_files._escape(name)
626
weave_transport = a_bzrdir._control_files._transport.clone(relpath)
627
ws = WeaveStore(weave_transport, prefixed=prefixed,
630
if a_bzrdir._control_files._transport.should_cache():
631
ws.enable_cache = True
634
def get_store(name, compressed=True, prefixed=False):
635
# FIXME: This approach of assuming stores are all entirely compressed
636
# or entirely uncompressed is tidy, but breaks upgrade from
637
# some existing branches where there's a mixture; we probably
638
# still want the option to look for both.
639
relpath = a_bzrdir._control_files._escape(name)
640
store = TextStore(a_bzrdir._control_files._transport.clone(relpath),
641
prefixed=prefixed, compressed=compressed,
644
#if self._transport.should_cache():
645
# cache_path = os.path.join(self.cache_root, name)
646
# os.mkdir(cache_path)
647
# store = bzrlib.store.CachedStore(store, cache_path)
650
# not broken out yet because the controlweaves|inventory_store
651
# and text_store | weave_store bits are still different.
652
if isinstance(_format, RepositoryFormat4):
653
self.inventory_store = get_store('inventory-store')
654
self.text_store = get_store('text-store')
655
elif isinstance(_format, RepositoryFormat5):
656
self.control_weaves = get_weave('')
657
self.weave_store = get_weave('weaves')
658
elif isinstance(_format, RepositoryFormat6):
659
self.control_weaves = get_weave('')
660
self.weave_store = get_weave('weaves', prefixed=True)
662
raise errors.BzrError('unreachable code: unexpected repository'
664
revision_store.register_suffix('sig')
665
super(AllInOneRepository, self).__init__(_format, a_bzrdir, a_bzrdir._control_files, revision_store)
668
class MetaDirRepository(Repository):
669
"""Repositories in the new meta-dir layout."""
671
def __init__(self, _format, a_bzrdir, control_files, revision_store):
672
super(MetaDirRepository, self).__init__(_format,
677
dir_mode = self.control_files._dir_mode
678
file_mode = self.control_files._file_mode
680
def get_weave(name, prefixed=False):
682
name = safe_unicode(name)
685
relpath = self.control_files._escape(name)
686
weave_transport = self.control_files._transport.clone(relpath)
687
ws = WeaveStore(weave_transport, prefixed=prefixed,
690
if self.control_files._transport.should_cache():
691
ws.enable_cache = True
694
if isinstance(self._format, RepositoryFormat7):
695
self.control_weaves = get_weave('')
696
self.weave_store = get_weave('weaves', prefixed=True)
697
elif isinstance(self._format, RepositoryFormatKnit1):
698
self.control_weaves = get_weave('')
699
self.weave_store = get_weave('knits', prefixed=True)
701
raise errors.BzrError('unreachable code: unexpected repository'
705
class RepositoryFormat(object):
706
"""A repository format.
708
Formats provide three things:
709
* An initialization routine to construct repository data on disk.
710
* a format string which is used when the BzrDir supports versioned
712
* an open routine which returns a Repository instance.
714
Formats are placed in an dict by their format string for reference
715
during opening. These should be subclasses of RepositoryFormat
718
Once a format is deprecated, just deprecate the initialize and open
719
methods on the format class. Do not deprecate the object, as the
720
object will be created every system load.
722
Common instance attributes:
723
_matchingbzrdir - the bzrdir format that the repository format was
724
originally written to work with. This can be used if manually
725
constructing a bzrdir and repository, or more commonly for test suite
729
_default_format = None
730
"""The default format used for new repositories."""
733
"""The known formats."""
736
def find_format(klass, a_bzrdir):
737
"""Return the format for the repository object in a_bzrdir."""
739
transport = a_bzrdir.get_repository_transport(None)
740
format_string = transport.get("format").read()
741
return klass._formats[format_string]
742
except errors.NoSuchFile:
743
raise errors.NoRepositoryPresent(a_bzrdir)
745
raise errors.UnknownFormatError(format_string)
748
def get_default_format(klass):
749
"""Return the current default format."""
750
return klass._default_format
752
def get_format_string(self):
753
"""Return the ASCII format string that identifies this format.
755
Note that in pre format ?? repositories the format string is
756
not permitted nor written to disk.
758
raise NotImplementedError(self.get_format_string)
760
def _get_revision_store(self, repo_transport, control_files):
761
"""Return the revision store object for this a_bzrdir."""
762
raise NotImplementedError(self._get_revision_store)
764
def _get_rev_store(self,
770
"""Common logic for getting a revision store for a repository.
772
see self._get_revision_store for the method to
773
get the store for a repository.
776
name = safe_unicode(name)
779
dir_mode = control_files._dir_mode
780
file_mode = control_files._file_mode
781
revision_store =TextStore(transport.clone(name),
783
compressed=compressed,
786
revision_store.register_suffix('sig')
787
return revision_store
789
def initialize(self, a_bzrdir, shared=False):
790
"""Initialize a repository of this format in a_bzrdir.
792
:param a_bzrdir: The bzrdir to put the new repository in it.
793
:param shared: The repository should be initialized as a sharable one.
795
This may raise UninitializableFormat if shared repository are not
796
compatible the a_bzrdir.
799
def is_supported(self):
800
"""Is this format supported?
802
Supported formats must be initializable and openable.
803
Unsupported formats may not support initialization or committing or
804
some other features depending on the reason for not being supported.
808
def open(self, a_bzrdir, _found=False):
809
"""Return an instance of this format for the bzrdir a_bzrdir.
811
_found is a private parameter, do not use it.
813
raise NotImplementedError(self.open)
816
def register_format(klass, format):
817
klass._formats[format.get_format_string()] = format
820
def set_default_format(klass, format):
821
klass._default_format = format
824
def unregister_format(klass, format):
825
assert klass._formats[format.get_format_string()] is format
826
del klass._formats[format.get_format_string()]
829
class PreSplitOutRepositoryFormat(RepositoryFormat):
830
"""Base class for the pre split out repository formats."""
832
def initialize(self, a_bzrdir, shared=False, _internal=False):
833
"""Create a weave repository.
835
TODO: when creating split out bzr branch formats, move this to a common
836
base for Format5, Format6. or something like that.
838
from bzrlib.weavefile import write_weave_v5
839
from bzrlib.weave import Weave
842
raise errors.IncompatibleFormat(self, a_bzrdir._format)
845
# always initialized when the bzrdir is.
846
return self.open(a_bzrdir, _found=True)
848
# Create an empty weave
850
bzrlib.weavefile.write_weave_v5(Weave(), sio)
851
empty_weave = sio.getvalue()
853
mutter('creating repository in %s.', a_bzrdir.transport.base)
854
dirs = ['revision-store', 'weaves']
855
files = [('inventory.weave', StringIO(empty_weave)),
858
# FIXME: RBC 20060125 dont peek under the covers
859
# NB: no need to escape relative paths that are url safe.
860
control_files = LockableFiles(a_bzrdir.transport, 'branch-lock',
862
control_files.create_lock()
863
control_files.lock_write()
864
control_files._transport.mkdir_multi(dirs,
865
mode=control_files._dir_mode)
867
for file, content in files:
868
control_files.put(file, content)
870
control_files.unlock()
871
return self.open(a_bzrdir, _found=True)
873
def open(self, a_bzrdir, _found=False):
874
"""See RepositoryFormat.open()."""
876
# we are being called directly and must probe.
877
raise NotImplementedError
879
repo_transport = a_bzrdir.get_repository_transport(None)
880
control_files = a_bzrdir._control_files
881
revision_store = self._get_revision_store(repo_transport, control_files)
882
return AllInOneRepository(_format=self,
884
revision_store=revision_store)
887
class RepositoryFormat4(PreSplitOutRepositoryFormat):
888
"""Bzr repository format 4.
890
This repository format has:
892
- TextStores for texts, inventories,revisions.
894
This format is deprecated: it indexes texts using a text id which is
895
removed in format 5; initializationa and write support for this format
900
super(RepositoryFormat4, self).__init__()
901
self._matchingbzrdir = bzrlib.bzrdir.BzrDirFormat4()
903
def initialize(self, url, shared=False, _internal=False):
904
"""Format 4 branches cannot be created."""
905
raise errors.UninitializableFormat(self)
907
def is_supported(self):
908
"""Format 4 is not supported.
910
It is not supported because the model changed from 4 to 5 and the
911
conversion logic is expensive - so doing it on the fly was not
916
def _get_revision_store(self, repo_transport, control_files):
917
"""See RepositoryFormat._get_revision_store()."""
918
return self._get_rev_store(repo_transport,
923
class RepositoryFormat5(PreSplitOutRepositoryFormat):
924
"""Bzr control format 5.
926
This repository format has:
927
- weaves for file texts and inventory
929
- TextStores for revisions and signatures.
933
super(RepositoryFormat5, self).__init__()
934
self._matchingbzrdir = bzrlib.bzrdir.BzrDirFormat5()
936
def _get_revision_store(self, repo_transport, control_files):
937
"""See RepositoryFormat._get_revision_store()."""
938
"""Return the revision store object for this a_bzrdir."""
939
return self._get_rev_store(repo_transport,
945
class RepositoryFormat6(PreSplitOutRepositoryFormat):
946
"""Bzr control format 6.
948
This repository format has:
949
- weaves for file texts and inventory
950
- hash subdirectory based stores.
951
- TextStores for revisions and signatures.
955
super(RepositoryFormat6, self).__init__()
956
self._matchingbzrdir = bzrlib.bzrdir.BzrDirFormat6()
958
def _get_revision_store(self, repo_transport, control_files):
959
"""See RepositoryFormat._get_revision_store()."""
960
return self._get_rev_store(repo_transport,
967
class MetaDirRepositoryFormat(RepositoryFormat):
968
"""Common base class for the new repositories using the metadir layour."""
971
super(MetaDirRepositoryFormat, self).__init__()
972
self._matchingbzrdir = bzrlib.bzrdir.BzrDirMetaFormat1()
974
def _create_control_files(self, a_bzrdir):
975
"""Create the required files and the initial control_files object."""
976
# FIXME: RBC 20060125 dont peek under the covers
977
# NB: no need to escape relative paths that are url safe.
978
repository_transport = a_bzrdir.get_repository_transport(self)
979
control_files = LockableFiles(repository_transport, 'lock', LockDir)
980
control_files.create_lock()
983
def _get_revision_store(self, repo_transport, control_files):
984
"""See RepositoryFormat._get_revision_store()."""
985
return self._get_rev_store(repo_transport,
992
def open(self, a_bzrdir, _found=False, _override_transport=None):
993
"""See RepositoryFormat.open().
995
:param _override_transport: INTERNAL USE ONLY. Allows opening the
996
repository at a slightly different url
997
than normal. I.e. during 'upgrade'.
1000
format = RepositoryFormat.find_format(a_bzrdir)
1001
assert format.__class__ == self.__class__
1002
if _override_transport is not None:
1003
repo_transport = _override_transport
1005
repo_transport = a_bzrdir.get_repository_transport(None)
1006
control_files = LockableFiles(repo_transport, 'lock', LockDir)
1007
revision_store = self._get_revision_store(repo_transport, control_files)
1008
return MetaDirRepository(_format=self,
1010
control_files=control_files,
1011
revision_store=revision_store)
1013
def _upload_blank_content(self, a_bzrdir, dirs, files, utf8_files, shared):
1014
"""Upload the initial blank content."""
1015
control_files = self._create_control_files(a_bzrdir)
1016
control_files.lock_write()
1018
control_files._transport.mkdir_multi(dirs,
1019
mode=control_files._dir_mode)
1020
for file, content in files:
1021
control_files.put(file, content)
1022
for file, content in utf8_files:
1023
control_files.put_utf8(file, content)
1025
control_files.put_utf8('shared-storage', '')
1027
control_files.unlock()
1030
class RepositoryFormat7(MetaDirRepositoryFormat):
1031
"""Bzr repository 7.
1033
This repository format has:
1034
- weaves for file texts and inventory
1035
- hash subdirectory based stores.
1036
- TextStores for revisions and signatures.
1037
- a format marker of its own
1038
- an optional 'shared-storage' flag
1039
- an optional 'no-working-trees' flag
1042
def get_format_string(self):
1043
"""See RepositoryFormat.get_format_string()."""
1044
return "Bazaar-NG Repository format 7"
1046
def initialize(self, a_bzrdir, shared=False):
1047
"""Create a weave repository.
1049
:param shared: If true the repository will be initialized as a shared
1052
from bzrlib.weavefile import write_weave_v5
1053
from bzrlib.weave import Weave
1055
# Create an empty weave
1057
bzrlib.weavefile.write_weave_v5(Weave(), sio)
1058
empty_weave = sio.getvalue()
1060
mutter('creating repository in %s.', a_bzrdir.transport.base)
1061
dirs = ['revision-store', 'weaves']
1062
files = [('inventory.weave', StringIO(empty_weave)),
1064
utf8_files = [('format', self.get_format_string())]
1066
self._upload_blank_content(a_bzrdir, dirs, files, utf8_files, shared)
1067
return self.open(a_bzrdir=a_bzrdir, _found=True)
1070
class RepositoryFormatKnit1(MetaDirRepositoryFormat):
1071
"""Bzr repository knit format 1.
1073
This repository format has:
1074
- knits for file texts and inventory
1075
- hash subdirectory based stores.
1076
- knits for revisions and signatures
1077
- TextStores for revisions and signatures.
1078
- a format marker of its own
1079
- an optional 'shared-storage' flag
1080
- an optional 'no-working-trees' flag
1084
def get_format_string(self):
1085
"""See RepositoryFormat.get_format_string()."""
1086
return "Bazaar-NG Knit Repository Format 1"
1088
def initialize(self, a_bzrdir, shared=False):
1089
"""Create a knit format 1 repository.
1091
:param shared: If true the repository will be initialized as a shared
1093
XXX NOTE that this current uses a Weave for testing and will become
1094
A Knit in due course.
1096
from bzrlib.weavefile import write_weave_v5
1097
from bzrlib.weave import Weave
1099
# Create an empty weave
1101
bzrlib.weavefile.write_weave_v5(Weave(), sio)
1102
empty_weave = sio.getvalue()
1104
mutter('creating repository in %s.', a_bzrdir.transport.base)
1105
dirs = ['revision-store', 'knits']
1106
files = [('inventory.weave', StringIO(empty_weave)),
1108
utf8_files = [('format', self.get_format_string())]
1110
self._upload_blank_content(a_bzrdir, dirs, files, utf8_files, shared)
1111
return self.open(a_bzrdir=a_bzrdir, _found=True)
1114
# formats which have no format string are not discoverable
1115
# and not independently creatable, so are not registered.
1116
_default_format = RepositoryFormat7()
1117
RepositoryFormat.register_format(_default_format)
1118
RepositoryFormat.register_format(RepositoryFormatKnit1())
1119
RepositoryFormat.set_default_format(_default_format)
1120
_legacy_formats = [RepositoryFormat4(),
1121
RepositoryFormat5(),
1122
RepositoryFormat6()]
1125
class InterRepository(object):
1126
"""This class represents operations taking place between two repositories.
1128
Its instances have methods like copy_content and fetch, and contain
1129
references to the source and target repositories these operations can be
1132
Often we will provide convenience methods on 'repository' which carry out
1133
operations with another repository - they will always forward to
1134
InterRepository.get(other).method_name(parameters).
1136
# XXX: FIXME: FUTURE: robertc
1137
# testing of these probably requires a factory in optimiser type, and
1138
# then a test adapter to test each type thoroughly.
1142
"""The available optimised InterRepository types."""
1144
def __init__(self, source, target):
1145
"""Construct a default InterRepository instance. Please use 'get'.
1147
Only subclasses of InterRepository should call
1148
InterRepository.__init__ - clients should call InterRepository.get
1149
instead which will create an optimised InterRepository if possible.
1151
self.source = source
1152
self.target = target
1155
def copy_content(self, revision_id=None, basis=None):
1156
"""Make a complete copy of the content in self into destination.
1158
This is a destructive operation! Do not use it on existing
1161
:param revision_id: Only copy the content needed to construct
1162
revision_id and its parents.
1163
:param basis: Copy the needed data preferentially from basis.
1166
self.target.set_make_working_trees(self.source.make_working_trees())
1167
except NotImplementedError:
1169
# grab the basis available data
1170
if basis is not None:
1171
self.target.fetch(basis, revision_id=revision_id)
1172
# but dont both fetching if we have the needed data now.
1173
if (revision_id not in (None, NULL_REVISION) and
1174
self.target.has_revision(revision_id)):
1176
self.target.fetch(self.source, revision_id=revision_id)
1178
def _double_lock(self, lock_source, lock_target):
1179
"""Take out too locks, rolling back the first if the second throws."""
1184
# we want to ensure that we don't leave source locked by mistake.
1185
# and any error on target should not confuse source.
1186
self.source.unlock()
1190
def fetch(self, revision_id=None, pb=None):
1191
"""Fetch the content required to construct revision_id.
1193
The content is copied from source to target.
1195
:param revision_id: if None all content is copied, if NULL_REVISION no
1197
:param pb: optional progress bar to use for progress reports. If not
1198
provided a default one will be created.
1200
Returns the copied revision count and the failed revisions in a tuple:
1203
from bzrlib.fetch import RepoFetcher
1204
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
1205
self.source, self.source._format, self.target, self.target._format)
1206
f = RepoFetcher(to_repository=self.target,
1207
from_repository=self.source,
1208
last_revision=revision_id,
1210
return f.count_copied, f.failed_revisions
1213
def get(klass, repository_source, repository_target):
1214
"""Retrieve a InterRepository worker object for these repositories.
1216
:param repository_source: the repository to be the 'source' member of
1217
the InterRepository instance.
1218
:param repository_target: the repository to be the 'target' member of
1219
the InterRepository instance.
1220
If an optimised InterRepository worker exists it will be used otherwise
1221
a default InterRepository instance will be created.
1223
for provider in klass._optimisers:
1224
if provider.is_compatible(repository_source, repository_target):
1225
return provider(repository_source, repository_target)
1226
return InterRepository(repository_source, repository_target)
1228
def lock_read(self):
1229
"""Take out a logical read lock.
1231
This will lock the source branch and the target branch. The source gets
1232
a read lock and the target a read lock.
1234
self._double_lock(self.source.lock_read, self.target.lock_read)
1236
def lock_write(self):
1237
"""Take out a logical write lock.
1239
This will lock the source branch and the target branch. The source gets
1240
a read lock and the target a write lock.
1242
self._double_lock(self.source.lock_read, self.target.lock_write)
1245
def missing_revision_ids(self, revision_id=None):
1246
"""Return the revision ids that source has that target does not.
1248
These are returned in topological order.
1250
:param revision_id: only return revision ids included by this
1253
# generic, possibly worst case, slow code path.
1254
target_ids = set(self.target.all_revision_ids())
1255
if revision_id is not None:
1256
source_ids = self.source.get_ancestry(revision_id)
1257
assert source_ids.pop(0) == None
1259
source_ids = self.source.all_revision_ids()
1260
result_set = set(source_ids).difference(target_ids)
1261
# this may look like a no-op: its not. It preserves the ordering
1262
# other_ids had while only returning the members from other_ids
1263
# that we've decided we need.
1264
return [rev_id for rev_id in source_ids if rev_id in result_set]
1267
def register_optimiser(klass, optimiser):
1268
"""Register an InterRepository optimiser."""
1269
klass._optimisers.add(optimiser)
1272
"""Release the locks on source and target."""
1274
self.target.unlock()
1276
self.source.unlock()
1279
def unregister_optimiser(klass, optimiser):
1280
"""Unregister an InterRepository optimiser."""
1281
klass._optimisers.remove(optimiser)
1284
class InterWeaveRepo(InterRepository):
1285
"""Optimised code paths between Weave based repositories."""
1287
_matching_repo_format = _default_format
1288
"""Repository format for testing with."""
1291
def is_compatible(source, target):
1292
"""Be compatible with known Weave formats.
1294
We dont test for the stores being of specific types becase that
1295
could lead to confusing results, and there is no need to be
1299
return (isinstance(source._format, (RepositoryFormat5,
1301
RepositoryFormat7)) and
1302
isinstance(target._format, (RepositoryFormat5,
1304
RepositoryFormat7)))
1305
except AttributeError:
1309
def copy_content(self, revision_id=None, basis=None):
1310
"""See InterRepository.copy_content()."""
1311
# weave specific optimised path:
1312
if basis is not None:
1313
# copy the basis in, then fetch remaining data.
1314
basis.copy_content_into(self.target, revision_id)
1315
# the basis copy_content_into could misset this.
1317
self.target.set_make_working_trees(self.source.make_working_trees())
1318
except NotImplementedError:
1320
self.target.fetch(self.source, revision_id=revision_id)
1323
self.target.set_make_working_trees(self.source.make_working_trees())
1324
except NotImplementedError:
1326
# FIXME do not peek!
1327
if self.source.control_files._transport.listable():
1328
pb = bzrlib.ui.ui_factory.nested_progress_bar()
1330
copy_all(self.source.weave_store,
1331
self.target.weave_store, pb=pb)
1332
pb.update('copying inventory', 0, 1)
1333
self.target.control_weaves.copy_multi(
1334
self.source.control_weaves, ['inventory'])
1335
copy_all(self.source.revision_store,
1336
self.target.revision_store, pb=pb)
1340
self.target.fetch(self.source, revision_id=revision_id)
1343
def fetch(self, revision_id=None, pb=None):
1344
"""See InterRepository.fetch()."""
1345
from bzrlib.fetch import RepoFetcher
1346
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
1347
self.source, self.source._format, self.target, self.target._format)
1348
f = RepoFetcher(to_repository=self.target,
1349
from_repository=self.source,
1350
last_revision=revision_id,
1352
return f.count_copied, f.failed_revisions
1355
def missing_revision_ids(self, revision_id=None):
1356
"""See InterRepository.missing_revision_ids()."""
1357
# we want all revisions to satisfy revision_id in source.
1358
# but we dont want to stat every file here and there.
1359
# we want then, all revisions other needs to satisfy revision_id
1360
# checked, but not those that we have locally.
1361
# so the first thing is to get a subset of the revisions to
1362
# satisfy revision_id in source, and then eliminate those that
1363
# we do already have.
1364
# this is slow on high latency connection to self, but as as this
1365
# disk format scales terribly for push anyway due to rewriting
1366
# inventory.weave, this is considered acceptable.
1368
if revision_id is not None:
1369
source_ids = self.source.get_ancestry(revision_id)
1370
assert source_ids.pop(0) == None
1372
source_ids = self.source._all_possible_ids()
1373
source_ids_set = set(source_ids)
1374
# source_ids is the worst possible case we may need to pull.
1375
# now we want to filter source_ids against what we actually
1376
# have in target, but dont try to check for existence where we know
1377
# we do not have a revision as that would be pointless.
1378
target_ids = set(self.target._all_possible_ids())
1379
possibly_present_revisions = target_ids.intersection(source_ids_set)
1380
actually_present_revisions = set(self.target._eliminate_revisions_not_present(possibly_present_revisions))
1381
required_revisions = source_ids_set.difference(actually_present_revisions)
1382
required_topo_revisions = [rev_id for rev_id in source_ids if rev_id in required_revisions]
1383
if revision_id is not None:
1384
# we used get_ancestry to determine source_ids then we are assured all
1385
# revisions referenced are present as they are installed in topological order.
1386
# and the tip revision was validated by get_ancestry.
1387
return required_topo_revisions
1389
# if we just grabbed the possibly available ids, then
1390
# we only have an estimate of whats available and need to validate
1391
# that against the revision records.
1392
return self.source._eliminate_revisions_not_present(required_topo_revisions)
1395
InterRepository.register_optimiser(InterWeaveRepo)
1398
class RepositoryTestProviderAdapter(object):
1399
"""A tool to generate a suite testing multiple repository formats at once.
1401
This is done by copying the test once for each transport and injecting
1402
the transport_server, transport_readonly_server, and bzrdir_format and
1403
repository_format classes into each copy. Each copy is also given a new id()
1404
to make it easy to identify.
1407
def __init__(self, transport_server, transport_readonly_server, formats):
1408
self._transport_server = transport_server
1409
self._transport_readonly_server = transport_readonly_server
1410
self._formats = formats
1412
def adapt(self, test):
1413
result = TestSuite()
1414
for repository_format, bzrdir_format in self._formats:
1415
new_test = deepcopy(test)
1416
new_test.transport_server = self._transport_server
1417
new_test.transport_readonly_server = self._transport_readonly_server
1418
new_test.bzrdir_format = bzrdir_format
1419
new_test.repository_format = repository_format
1420
def make_new_test_id():
1421
new_id = "%s(%s)" % (new_test.id(), repository_format.__class__.__name__)
1422
return lambda: new_id
1423
new_test.id = make_new_test_id()
1424
result.addTest(new_test)
1428
class InterRepositoryTestProviderAdapter(object):
1429
"""A tool to generate a suite testing multiple inter repository formats.
1431
This is done by copying the test once for each interrepo provider and injecting
1432
the transport_server, transport_readonly_server, repository_format and
1433
repository_to_format classes into each copy.
1434
Each copy is also given a new id() to make it easy to identify.
1437
def __init__(self, transport_server, transport_readonly_server, formats):
1438
self._transport_server = transport_server
1439
self._transport_readonly_server = transport_readonly_server
1440
self._formats = formats
1442
def adapt(self, test):
1443
result = TestSuite()
1444
for interrepo_class, repository_format, repository_format_to in self._formats:
1445
new_test = deepcopy(test)
1446
new_test.transport_server = self._transport_server
1447
new_test.transport_readonly_server = self._transport_readonly_server
1448
new_test.interrepo_class = interrepo_class
1449
new_test.repository_format = repository_format
1450
new_test.repository_format_to = repository_format_to
1451
def make_new_test_id():
1452
new_id = "%s(%s)" % (new_test.id(), interrepo_class.__name__)
1453
return lambda: new_id
1454
new_test.id = make_new_test_id()
1455
result.addTest(new_test)
1459
def default_test_list():
1460
"""Generate the default list of interrepo permutations to test."""
1462
# test the default InterRepository between format 6 and the current
1464
# XXX: robertc 20060220 reinstate this when there are two supported
1465
# formats which do not have an optimal code path between them.
1466
result.append((InterRepository,
1467
RepositoryFormat6(),
1468
RepositoryFormatKnit1()))
1469
for optimiser in InterRepository._optimisers:
1470
result.append((optimiser,
1471
optimiser._matching_repo_format,
1472
optimiser._matching_repo_format
1474
# if there are specific combinations we want to use, we can add them
1479
class CopyConverter(object):
1480
"""A repository conversion tool which just performs a copy of the content.
1482
This is slow but quite reliable.
1485
def __init__(self, target_format):
1486
"""Create a CopyConverter.
1488
:param target_format: The format the resulting repository should be.
1490
self.target_format = target_format
1492
def convert(self, repo, pb):
1493
"""Perform the conversion of to_convert, giving feedback via pb.
1495
:param to_convert: The disk object to convert.
1496
:param pb: a progress bar to use for progress information.
1501
# this is only useful with metadir layouts - separated repo content.
1502
# trigger an assertion if not such
1503
repo._format.get_format_string()
1504
self.repo_dir = repo.bzrdir
1505
self.step('Moving repository to repository.backup')
1506
self.repo_dir.transport.move('repository', 'repository.backup')
1507
backup_transport = self.repo_dir.transport.clone('repository.backup')
1508
self.source_repo = repo._format.open(self.repo_dir,
1510
_override_transport=backup_transport)
1511
self.step('Creating new repository')
1512
converted = self.target_format.initialize(self.repo_dir,
1513
self.source_repo.is_shared())
1514
converted.lock_write()
1516
self.step('Copying content into repository.')
1517
self.source_repo.copy_content_into(converted)
1520
self.step('Deleting old repository content.')
1521
self.repo_dir.transport.delete_tree('repository.backup')
1522
self.pb.note('repository converted')
1524
def step(self, message):
1525
"""Update the pb by a step."""
1527
self.pb.update(message, self.count, self.total)