1
# Copyright (C) 2005, 2006, 2007 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 cStringIO import StringIO
19
from bzrlib.lazy_import import lazy_import
20
lazy_import(globals(), """
40
revision as _mod_revision,
46
from bzrlib.bundle import serializer
47
from bzrlib.revisiontree import RevisionTree
48
from bzrlib.store.versioned import VersionedFileStore
49
from bzrlib.store.text import TextStore
50
from bzrlib.testament import Testament
51
from bzrlib.util import bencode
54
from bzrlib.decorators import needs_read_lock, needs_write_lock
55
from bzrlib.inter import InterObject
56
from bzrlib.inventory import Inventory, InventoryDirectory, ROOT_ID
57
from bzrlib.symbol_versioning import (
60
from bzrlib.trace import mutter, mutter_callsite, note, warning
63
# Old formats display a warning, but only once
64
_deprecation_warning_done = False
67
class CommitBuilder(object):
68
"""Provides an interface to build up a commit.
70
This allows describing a tree to be committed without needing to
71
know the internals of the format of the repository.
74
# all clients should supply tree roots.
75
record_root_entry = True
76
# the default CommitBuilder does not manage trees whose root is versioned.
77
_versioned_root = False
79
def __init__(self, repository, parents, config, timestamp=None,
80
timezone=None, committer=None, revprops=None,
82
"""Initiate a CommitBuilder.
84
:param repository: Repository to commit to.
85
:param parents: Revision ids of the parents of the new revision.
86
:param config: Configuration to use.
87
:param timestamp: Optional timestamp recorded for commit.
88
:param timezone: Optional timezone for timestamp.
89
:param committer: Optional committer to set for commit.
90
:param revprops: Optional dictionary of revision properties.
91
:param revision_id: Optional revision id.
96
self._committer = self._config.username()
98
assert isinstance(committer, basestring), type(committer)
99
self._committer = committer
101
self.new_inventory = Inventory(None)
102
self._new_revision_id = revision_id
103
self.parents = parents
104
self.repository = repository
107
if revprops is not None:
108
self._revprops.update(revprops)
110
if timestamp is None:
111
timestamp = time.time()
112
# Restrict resolution to 1ms
113
self._timestamp = round(timestamp, 3)
116
self._timezone = osutils.local_time_offset()
118
self._timezone = int(timezone)
120
self._generate_revision_if_needed()
121
self._heads = graph.HeadsCache(repository.get_graph()).heads
123
def commit(self, message):
124
"""Make the actual commit.
126
:return: The revision id of the recorded revision.
128
rev = _mod_revision.Revision(
129
timestamp=self._timestamp,
130
timezone=self._timezone,
131
committer=self._committer,
133
inventory_sha1=self.inv_sha1,
134
revision_id=self._new_revision_id,
135
properties=self._revprops)
136
rev.parent_ids = self.parents
137
self.repository.add_revision(self._new_revision_id, rev,
138
self.new_inventory, self._config)
139
self.repository.commit_write_group()
140
return self._new_revision_id
143
"""Abort the commit that is being built.
145
self.repository.abort_write_group()
147
def revision_tree(self):
148
"""Return the tree that was just committed.
150
After calling commit() this can be called to get a RevisionTree
151
representing the newly committed tree. This is preferred to
152
calling Repository.revision_tree() because that may require
153
deserializing the inventory, while we already have a copy in
156
return RevisionTree(self.repository, self.new_inventory,
157
self._new_revision_id)
159
def finish_inventory(self):
160
"""Tell the builder that the inventory is finished."""
161
if self.new_inventory.root is None:
162
raise AssertionError('Root entry should be supplied to'
163
' record_entry_contents, as of bzr 0.10.',
164
DeprecationWarning, stacklevel=2)
165
self.new_inventory.add(InventoryDirectory(ROOT_ID, '', None))
166
self.new_inventory.revision_id = self._new_revision_id
167
self.inv_sha1 = self.repository.add_inventory(
168
self._new_revision_id,
173
def _gen_revision_id(self):
174
"""Return new revision-id."""
175
return generate_ids.gen_revision_id(self._config.username(),
178
def _generate_revision_if_needed(self):
179
"""Create a revision id if None was supplied.
181
If the repository can not support user-specified revision ids
182
they should override this function and raise CannotSetRevisionId
183
if _new_revision_id is not None.
185
:raises: CannotSetRevisionId
187
if self._new_revision_id is None:
188
self._new_revision_id = self._gen_revision_id()
189
self.random_revid = True
191
self.random_revid = False
193
def _check_root(self, ie, parent_invs, tree):
194
"""Helper for record_entry_contents.
196
:param ie: An entry being added.
197
:param parent_invs: The inventories of the parent revisions of the
199
:param tree: The tree that is being committed.
201
# In this revision format, root entries have no knit or weave When
202
# serializing out to disk and back in root.revision is always
204
ie.revision = self._new_revision_id
206
def _get_delta(self, ie, basis_inv, path):
207
"""Get a delta against the basis inventory for ie."""
208
if ie.file_id not in basis_inv:
210
return (None, path, ie.file_id, ie)
211
elif ie != basis_inv[ie.file_id]:
213
# TODO: avoid tis id2path call.
214
return (basis_inv.id2path(ie.file_id), path, ie.file_id, ie)
219
def record_entry_contents(self, ie, parent_invs, path, tree,
221
"""Record the content of ie from tree into the commit if needed.
223
Side effect: sets ie.revision when unchanged
225
:param ie: An inventory entry present in the commit.
226
:param parent_invs: The inventories of the parent revisions of the
228
:param path: The path the entry is at in the tree.
229
:param tree: The tree which contains this entry and should be used to
231
:param content_summary: Summary data from the tree about the paths
232
content - stat, length, exec, sha/link target. This is only
233
accessed when the entry has a revision of None - that is when it is
234
a candidate to commit.
235
:return: A tuple (change_delta, version_recorded). change_delta is
236
an inventory_delta change for this entry against the basis tree of
237
the commit, or None if no change occured against the basis tree.
238
version_recorded is True if a new version of the entry has been
239
recorded. For instance, committing a merge where a file was only
240
changed on the other side will return (delta, False).
242
if self.new_inventory.root is None:
243
if ie.parent_id is not None:
244
raise errors.RootMissing()
245
self._check_root(ie, parent_invs, tree)
246
if ie.revision is None:
247
kind = content_summary[0]
249
# ie is carried over from a prior commit
251
# XXX: repository specific check for nested tree support goes here - if
252
# the repo doesn't want nested trees we skip it ?
253
if (kind == 'tree-reference' and
254
not self.repository._format.supports_tree_reference):
255
# mismatch between commit builder logic and repository:
256
# this needs the entry creation pushed down into the builder.
257
raise NotImplementedError('Missing repository subtree support.')
258
self.new_inventory.add(ie)
260
# TODO: slow, take it out of the inner loop.
262
basis_inv = parent_invs[0]
264
basis_inv = Inventory(root_id=None)
266
# ie.revision is always None if the InventoryEntry is considered
267
# for committing. We may record the previous parents revision if the
268
# content is actually unchanged against a sole head.
269
if ie.revision is not None:
270
if not self._versioned_root and path == '':
271
# repositories that do not version the root set the root's
272
# revision to the new commit even when no change occurs, and
273
# this masks when a change may have occurred against the basis,
274
# so calculate if one happened.
275
if ie.file_id in basis_inv:
276
delta = (basis_inv.id2path(ie.file_id), path,
280
delta = (None, path, ie.file_id, ie)
283
# we don't need to commit this, because the caller already
284
# determined that an existing revision of this file is
286
return None, (ie.revision == self._new_revision_id)
287
# XXX: Friction: parent_candidates should return a list not a dict
288
# so that we don't have to walk the inventories again.
289
parent_candiate_entries = ie.parent_candidates(parent_invs)
290
head_set = self._heads(parent_candiate_entries.keys())
292
for inv in parent_invs:
293
if ie.file_id in inv:
294
old_rev = inv[ie.file_id].revision
295
if old_rev in head_set:
296
heads.append(inv[ie.file_id].revision)
297
head_set.remove(inv[ie.file_id].revision)
300
# now we check to see if we need to write a new record to the
302
# We write a new entry unless there is one head to the ancestors, and
303
# the kind-derived content is unchanged.
305
# Cheapest check first: no ancestors, or more the one head in the
306
# ancestors, we write a new node.
310
# There is a single head, look it up for comparison
311
parent_entry = parent_candiate_entries[heads[0]]
312
# if the non-content specific data has changed, we'll be writing a
314
if (parent_entry.parent_id != ie.parent_id or
315
parent_entry.name != ie.name):
317
# now we need to do content specific checks:
319
# if the kind changed the content obviously has
320
if kind != parent_entry.kind:
323
assert content_summary[2] is not None, \
324
"Files must not have executable = None"
326
if (# if the file length changed we have to store:
327
parent_entry.text_size != content_summary[1] or
328
# if the exec bit has changed we have to store:
329
parent_entry.executable != content_summary[2]):
331
elif parent_entry.text_sha1 == content_summary[3]:
332
# all meta and content is unchanged (using a hash cache
333
# hit to check the sha)
334
ie.revision = parent_entry.revision
335
ie.text_size = parent_entry.text_size
336
ie.text_sha1 = parent_entry.text_sha1
337
ie.executable = parent_entry.executable
338
return self._get_delta(ie, basis_inv, path), False
340
# Either there is only a hash change(no hash cache entry,
341
# or same size content change), or there is no change on
343
# Provide the parent's hash to the store layer, so that the
344
# content is unchanged we will not store a new node.
345
nostore_sha = parent_entry.text_sha1
347
# We want to record a new node regardless of the presence or
348
# absence of a content change in the file.
350
ie.executable = content_summary[2]
351
lines = tree.get_file(ie.file_id, path).readlines()
353
ie.text_sha1, ie.text_size = self._add_text_to_weave(
354
ie.file_id, lines, heads, nostore_sha)
355
except errors.ExistingContent:
356
# Turns out that the file content was unchanged, and we were
357
# only going to store a new node if it was changed. Carry over
359
ie.revision = parent_entry.revision
360
ie.text_size = parent_entry.text_size
361
ie.text_sha1 = parent_entry.text_sha1
362
ie.executable = parent_entry.executable
363
return self._get_delta(ie, basis_inv, path), False
364
elif kind == 'directory':
366
# all data is meta here, nothing specific to directory, so
368
ie.revision = parent_entry.revision
369
return self._get_delta(ie, basis_inv, path), False
371
self._add_text_to_weave(ie.file_id, lines, heads, None)
372
elif kind == 'symlink':
373
current_link_target = content_summary[3]
375
# symlink target is not generic metadata, check if it has
377
if current_link_target != parent_entry.symlink_target:
380
# unchanged, carry over.
381
ie.revision = parent_entry.revision
382
ie.symlink_target = parent_entry.symlink_target
383
return self._get_delta(ie, basis_inv, path), False
384
ie.symlink_target = current_link_target
386
self._add_text_to_weave(ie.file_id, lines, heads, None)
387
elif kind == 'tree-reference':
389
if content_summary[3] != parent_entry.reference_revision:
392
# unchanged, carry over.
393
ie.reference_revision = parent_entry.reference_revision
394
ie.revision = parent_entry.revision
395
return self._get_delta(ie, basis_inv, path), False
396
ie.reference_revision = content_summary[3]
398
self._add_text_to_weave(ie.file_id, lines, heads, None)
400
raise NotImplementedError('unknown kind')
401
ie.revision = self._new_revision_id
402
return self._get_delta(ie, basis_inv, path), True
404
def _add_text_to_weave(self, file_id, new_lines, parents, nostore_sha):
405
versionedfile = self.repository.weave_store.get_weave_or_empty(
406
file_id, self.repository.get_transaction())
407
# Don't change this to add_lines - add_lines_with_ghosts is cheaper
408
# than add_lines, and allows committing when a parent is ghosted for
410
# Note: as we read the content directly from the tree, we know its not
411
# been turned into unicode or badly split - but a broken tree
412
# implementation could give us bad output from readlines() so this is
413
# not a guarantee of safety. What would be better is always checking
414
# the content during test suite execution. RBC 20070912
416
return versionedfile.add_lines_with_ghosts(
417
self._new_revision_id, parents, new_lines,
418
nostore_sha=nostore_sha, random_id=self.random_revid,
419
check_content=False)[0:2]
421
versionedfile.clear_cache()
424
class RootCommitBuilder(CommitBuilder):
425
"""This commitbuilder actually records the root id"""
427
# the root entry gets versioned properly by this builder.
428
_versioned_root = True
430
def _check_root(self, ie, parent_invs, tree):
431
"""Helper for record_entry_contents.
433
:param ie: An entry being added.
434
:param parent_invs: The inventories of the parent revisions of the
436
:param tree: The tree that is being committed.
440
######################################################################
443
class Repository(object):
444
"""Repository holding history for one or more branches.
446
The repository holds and retrieves historical information including
447
revisions and file history. It's normally accessed only by the Branch,
448
which views a particular line of development through that history.
450
The Repository builds on top of Stores and a Transport, which respectively
451
describe the disk data format and the way of accessing the (possibly
455
# What class to use for a CommitBuilder. Often its simpler to change this
456
# in a Repository class subclass rather than to override
457
# get_commit_builder.
458
_commit_builder_class = CommitBuilder
459
# The search regex used by xml based repositories to determine what things
460
# where changed in a single commit.
461
_file_ids_altered_regex = lazy_regex.lazy_compile(
462
r'file_id="(?P<file_id>[^"]+)"'
463
r'.* revision="(?P<revision_id>[^"]+)"'
466
def abort_write_group(self):
467
"""Commit the contents accrued within the current write group.
469
:seealso: start_write_group.
471
if self._write_group is not self.get_transaction():
472
# has an unlock or relock occured ?
473
raise errors.BzrError('mismatched lock context and write group.')
474
self._abort_write_group()
475
self._write_group = None
477
def _abort_write_group(self):
478
"""Template method for per-repository write group cleanup.
480
This is called during abort before the write group is considered to be
481
finished and should cleanup any internal state accrued during the write
482
group. There is no requirement that data handed to the repository be
483
*not* made available - this is not a rollback - but neither should any
484
attempt be made to ensure that data added is fully commited. Abort is
485
invoked when an error has occured so futher disk or network operations
486
may not be possible or may error and if possible should not be
491
def add_inventory(self, revision_id, inv, parents):
492
"""Add the inventory inv to the repository as revision_id.
494
:param parents: The revision ids of the parents that revision_id
495
is known to have and are in the repository already.
497
returns the sha1 of the serialized inventory.
499
assert self.is_in_write_group()
500
_mod_revision.check_not_reserved_id(revision_id)
501
assert inv.revision_id is None or inv.revision_id == revision_id, \
502
"Mismatch between inventory revision" \
503
" id and insertion revid (%r, %r)" % (inv.revision_id, revision_id)
504
assert inv.root is not None
505
inv_lines = self._serialise_inventory_to_lines(inv)
506
inv_vf = self.get_inventory_weave()
507
return self._inventory_add_lines(inv_vf, revision_id, parents,
508
inv_lines, check_content=False)
510
def _inventory_add_lines(self, inv_vf, revision_id, parents, lines,
512
"""Store lines in inv_vf and return the sha1 of the inventory."""
514
for parent in parents:
516
final_parents.append(parent)
517
return inv_vf.add_lines(revision_id, final_parents, lines,
518
check_content=check_content)[0]
521
def add_revision(self, revision_id, rev, inv=None, config=None):
522
"""Add rev to the revision store as revision_id.
524
:param revision_id: the revision id to use.
525
:param rev: The revision object.
526
:param inv: The inventory for the revision. if None, it will be looked
527
up in the inventory storer
528
:param config: If None no digital signature will be created.
529
If supplied its signature_needed method will be used
530
to determine if a signature should be made.
532
# TODO: jam 20070210 Shouldn't we check rev.revision_id and
534
_mod_revision.check_not_reserved_id(revision_id)
535
if config is not None and config.signature_needed():
537
inv = self.get_inventory(revision_id)
538
plaintext = Testament(rev, inv).as_short_text()
539
self.store_revision_signature(
540
gpg.GPGStrategy(config), plaintext, revision_id)
541
if not revision_id in self.get_inventory_weave():
543
raise errors.WeaveRevisionNotPresent(revision_id,
544
self.get_inventory_weave())
546
# yes, this is not suitable for adding with ghosts.
547
self.add_inventory(revision_id, inv, rev.parent_ids)
548
self._revision_store.add_revision(rev, self.get_transaction())
550
def _add_revision_text(self, revision_id, text):
551
revision = self._revision_store._serializer.read_revision_from_string(
553
self._revision_store._add_revision(revision, StringIO(text),
554
self.get_transaction())
556
def all_revision_ids(self):
557
"""Returns a list of all the revision ids in the repository.
559
This is deprecated because code should generally work on the graph
560
reachable from a particular revision, and ignore any other revisions
561
that might be present. There is no direct replacement method.
563
if 'evil' in debug.debug_flags:
564
mutter_callsite(2, "all_revision_ids is linear with history.")
565
return self._all_revision_ids()
567
def _all_revision_ids(self):
568
"""Returns a list of all the revision ids in the repository.
570
These are in as much topological order as the underlying store can
573
raise NotImplementedError(self._all_revision_ids)
575
def break_lock(self):
576
"""Break a lock if one is present from another instance.
578
Uses the ui factory to ask for confirmation if the lock may be from
581
self.control_files.break_lock()
584
def _eliminate_revisions_not_present(self, revision_ids):
585
"""Check every revision id in revision_ids to see if we have it.
587
Returns a set of the present revisions.
590
for id in revision_ids:
591
if self.has_revision(id):
596
def create(a_bzrdir):
597
"""Construct the current default format repository in a_bzrdir."""
598
return RepositoryFormat.get_default_format().initialize(a_bzrdir)
600
def __init__(self, _format, a_bzrdir, control_files, _revision_store, control_store, text_store):
601
"""instantiate a Repository.
603
:param _format: The format of the repository on disk.
604
:param a_bzrdir: The BzrDir of the repository.
606
In the future we will have a single api for all stores for
607
getting file texts, inventories and revisions, then
608
this construct will accept instances of those things.
610
super(Repository, self).__init__()
611
self._format = _format
612
# the following are part of the public API for Repository:
613
self.bzrdir = a_bzrdir
614
self.control_files = control_files
615
self._revision_store = _revision_store
616
# backwards compatibility
617
self.weave_store = text_store
619
self._reconcile_does_inventory_gc = True
620
self._reconcile_fixes_text_parents = False
621
self._reconcile_backsup_inventory = True
622
# not right yet - should be more semantically clear ?
624
self.control_store = control_store
625
self.control_weaves = control_store
626
# TODO: make sure to construct the right store classes, etc, depending
627
# on whether escaping is required.
628
self._warn_if_deprecated()
629
self._write_group = None
630
self.base = control_files._transport.base
633
return '%s(%r)' % (self.__class__.__name__,
636
def has_same_location(self, other):
637
"""Returns a boolean indicating if this repository is at the same
638
location as another repository.
640
This might return False even when two repository objects are accessing
641
the same physical repository via different URLs.
643
if self.__class__ is not other.__class__:
645
return (self.control_files._transport.base ==
646
other.control_files._transport.base)
648
def is_in_write_group(self):
649
"""Return True if there is an open write group.
651
:seealso: start_write_group.
653
return self._write_group is not None
656
return self.control_files.is_locked()
658
def is_write_locked(self):
659
"""Return True if this object is write locked."""
660
return self.is_locked() and self.control_files._lock_mode == 'w'
662
def lock_write(self, token=None):
663
"""Lock this repository for writing.
665
This causes caching within the repository obejct to start accumlating
666
data during reads, and allows a 'write_group' to be obtained. Write
667
groups must be used for actual data insertion.
669
:param token: if this is already locked, then lock_write will fail
670
unless the token matches the existing lock.
671
:returns: a token if this instance supports tokens, otherwise None.
672
:raises TokenLockingNotSupported: when a token is given but this
673
instance doesn't support using token locks.
674
:raises MismatchedToken: if the specified token doesn't match the token
675
of the existing lock.
676
:seealso: start_write_group.
678
A token should be passed in if you know that you have locked the object
679
some other way, and need to synchronise this object's state with that
682
XXX: this docstring is duplicated in many places, e.g. lockable_files.py
684
result = self.control_files.lock_write(token=token)
689
self.control_files.lock_read()
692
def get_physical_lock_status(self):
693
return self.control_files.get_physical_lock_status()
695
def leave_lock_in_place(self):
696
"""Tell this repository not to release the physical lock when this
699
If lock_write doesn't return a token, then this method is not supported.
701
self.control_files.leave_in_place()
703
def dont_leave_lock_in_place(self):
704
"""Tell this repository to release the physical lock when this
705
object is unlocked, even if it didn't originally acquire it.
707
If lock_write doesn't return a token, then this method is not supported.
709
self.control_files.dont_leave_in_place()
712
def gather_stats(self, revid=None, committers=None):
713
"""Gather statistics from a revision id.
715
:param revid: The revision id to gather statistics from, if None, then
716
no revision specific statistics are gathered.
717
:param committers: Optional parameter controlling whether to grab
718
a count of committers from the revision specific statistics.
719
:return: A dictionary of statistics. Currently this contains:
720
committers: The number of committers if requested.
721
firstrev: A tuple with timestamp, timezone for the penultimate left
722
most ancestor of revid, if revid is not the NULL_REVISION.
723
latestrev: A tuple with timestamp, timezone for revid, if revid is
724
not the NULL_REVISION.
725
revisions: The total revision count in the repository.
726
size: An estimate disk size of the repository in bytes.
729
if revid and committers:
730
result['committers'] = 0
731
if revid and revid != _mod_revision.NULL_REVISION:
733
all_committers = set()
734
revisions = self.get_ancestry(revid)
735
# pop the leading None
737
first_revision = None
739
# ignore the revisions in the middle - just grab first and last
740
revisions = revisions[0], revisions[-1]
741
for revision in self.get_revisions(revisions):
742
if not first_revision:
743
first_revision = revision
745
all_committers.add(revision.committer)
746
last_revision = revision
748
result['committers'] = len(all_committers)
749
result['firstrev'] = (first_revision.timestamp,
750
first_revision.timezone)
751
result['latestrev'] = (last_revision.timestamp,
752
last_revision.timezone)
754
# now gather global repository information
755
if self.bzrdir.root_transport.listable():
756
c, t = self._revision_store.total_size(self.get_transaction())
757
result['revisions'] = c
761
def get_data_stream(self, revision_ids):
762
raise NotImplementedError(self.get_data_stream)
764
def insert_data_stream(self, stream):
765
"""XXX What does this really do?
767
Is it a substitute for fetch?
768
Should it manage its own write group ?
770
for item_key, bytes in stream:
771
if item_key[0] == 'file':
772
(file_id,) = item_key[1:]
773
knit = self.weave_store.get_weave_or_empty(
774
file_id, self.get_transaction())
775
elif item_key == ('inventory',):
776
knit = self.get_inventory_weave()
777
elif item_key == ('revisions',):
778
knit = self._revision_store.get_revision_file(
779
self.get_transaction())
780
elif item_key == ('signatures',):
781
knit = self._revision_store.get_signature_file(
782
self.get_transaction())
784
raise RepositoryDataStreamError(
785
"Unrecognised data stream key '%s'" % (item_key,))
786
decoded_list = bencode.bdecode(bytes)
787
format = decoded_list.pop(0)
790
for version, options, parents, some_bytes in decoded_list:
791
data_list.append((version, options, len(some_bytes), parents))
792
knit_bytes += some_bytes
793
knit.insert_data_stream(
794
(format, data_list, StringIO(knit_bytes).read))
797
def missing_revision_ids(self, other, revision_id=None, find_ghosts=True):
798
"""Return the revision ids that other has that this does not.
800
These are returned in topological order.
802
revision_id: only return revision ids included by revision_id.
804
return InterRepository.get(other, self).missing_revision_ids(
805
revision_id, find_ghosts)
809
"""Open the repository rooted at base.
811
For instance, if the repository is at URL/.bzr/repository,
812
Repository.open(URL) -> a Repository instance.
814
control = bzrdir.BzrDir.open(base)
815
return control.open_repository()
817
def copy_content_into(self, destination, revision_id=None):
818
"""Make a complete copy of the content in self into destination.
820
This is a destructive operation! Do not use it on existing
823
return InterRepository.get(self, destination).copy_content(revision_id)
825
def commit_write_group(self):
826
"""Commit the contents accrued within the current write group.
828
:seealso: start_write_group.
830
if self._write_group is not self.get_transaction():
831
# has an unlock or relock occured ?
832
raise errors.BzrError('mismatched lock context %r and '
834
(self.get_transaction(), self._write_group))
835
self._commit_write_group()
836
self._write_group = None
838
def _commit_write_group(self):
839
"""Template method for per-repository write group cleanup.
841
This is called before the write group is considered to be
842
finished and should ensure that all data handed to the repository
843
for writing during the write group is safely committed (to the
844
extent possible considering file system caching etc).
847
def fetch(self, source, revision_id=None, pb=None, find_ghosts=False):
848
"""Fetch the content required to construct revision_id from source.
850
If revision_id is None all content is copied.
851
:param find_ghosts: Find and copy revisions in the source that are
852
ghosts in the target (and not reachable directly by walking out to
853
the first-present revision in target from revision_id).
855
# fast path same-url fetch operations
856
if self.has_same_location(source):
857
# check that last_revision is in 'from' and then return a
859
if (revision_id is not None and
860
not _mod_revision.is_null(revision_id)):
861
self.get_revision(revision_id)
863
inter = InterRepository.get(source, self)
865
return inter.fetch(revision_id=revision_id, pb=pb, find_ghosts=find_ghosts)
866
except NotImplementedError:
867
raise errors.IncompatibleRepositories(source, self)
869
def create_bundle(self, target, base, fileobj, format=None):
870
return serializer.write_bundle(self, target, base, fileobj, format)
872
def get_commit_builder(self, branch, parents, config, timestamp=None,
873
timezone=None, committer=None, revprops=None,
875
"""Obtain a CommitBuilder for this repository.
877
:param branch: Branch to commit to.
878
:param parents: Revision ids of the parents of the new revision.
879
:param config: Configuration to use.
880
:param timestamp: Optional timestamp recorded for commit.
881
:param timezone: Optional timezone for timestamp.
882
:param committer: Optional committer to set for commit.
883
:param revprops: Optional dictionary of revision properties.
884
:param revision_id: Optional revision id.
886
result = self._commit_builder_class(self, parents, config,
887
timestamp, timezone, committer, revprops, revision_id)
888
self.start_write_group()
892
if (self.control_files._lock_count == 1 and
893
self.control_files._lock_mode == 'w'):
894
if self._write_group is not None:
895
self.abort_write_group()
896
self.control_files.unlock()
897
raise errors.BzrError(
898
'Must end write groups before releasing write locks.')
899
self.control_files.unlock()
902
def clone(self, a_bzrdir, revision_id=None):
903
"""Clone this repository into a_bzrdir using the current format.
905
Currently no check is made that the format of this repository and
906
the bzrdir format are compatible. FIXME RBC 20060201.
908
:return: The newly created destination repository.
910
# TODO: deprecate after 0.16; cloning this with all its settings is
911
# probably not very useful -- mbp 20070423
912
dest_repo = self._create_sprouting_repo(a_bzrdir, shared=self.is_shared())
913
self.copy_content_into(dest_repo, revision_id)
916
def start_write_group(self):
917
"""Start a write group in the repository.
919
Write groups are used by repositories which do not have a 1:1 mapping
920
between file ids and backend store to manage the insertion of data from
921
both fetch and commit operations.
923
A write lock is required around the start_write_group/commit_write_group
924
for the support of lock-requiring repository formats.
926
One can only insert data into a repository inside a write group.
930
if not self.is_write_locked():
931
raise errors.NotWriteLocked(self)
932
if self._write_group:
933
raise errors.BzrError('already in a write group')
934
self._start_write_group()
935
# so we can detect unlock/relock - the write group is now entered.
936
self._write_group = self.get_transaction()
938
def _start_write_group(self):
939
"""Template method for per-repository write group startup.
941
This is called before the write group is considered to be
946
def sprout(self, to_bzrdir, revision_id=None):
947
"""Create a descendent repository for new development.
949
Unlike clone, this does not copy the settings of the repository.
951
dest_repo = self._create_sprouting_repo(to_bzrdir, shared=False)
952
dest_repo.fetch(self, revision_id=revision_id)
955
def _create_sprouting_repo(self, a_bzrdir, shared):
956
if not isinstance(a_bzrdir._format, self.bzrdir._format.__class__):
957
# use target default format.
958
dest_repo = a_bzrdir.create_repository()
960
# Most control formats need the repository to be specifically
961
# created, but on some old all-in-one formats it's not needed
963
dest_repo = self._format.initialize(a_bzrdir, shared=shared)
964
except errors.UninitializableFormat:
965
dest_repo = a_bzrdir.open_repository()
969
def has_revision(self, revision_id):
970
"""True if this repository has a copy of the revision."""
971
if 'evil' in debug.debug_flags:
972
mutter_callsite(3, "has_revision is a LBYL symptom.")
973
return self._revision_store.has_revision_id(revision_id,
974
self.get_transaction())
977
def get_revision(self, revision_id):
978
"""Return the Revision object for a named revision."""
979
return self.get_revisions([revision_id])[0]
982
def get_revision_reconcile(self, revision_id):
983
"""'reconcile' helper routine that allows access to a revision always.
985
This variant of get_revision does not cross check the weave graph
986
against the revision one as get_revision does: but it should only
987
be used by reconcile, or reconcile-alike commands that are correcting
988
or testing the revision graph.
990
return self._get_revisions([revision_id])[0]
993
def get_revisions(self, revision_ids):
994
"""Get many revisions at once."""
995
return self._get_revisions(revision_ids)
998
def _get_revisions(self, revision_ids):
999
"""Core work logic to get many revisions without sanity checks."""
1000
for rev_id in revision_ids:
1001
if not rev_id or not isinstance(rev_id, basestring):
1002
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1003
revs = self._revision_store.get_revisions(revision_ids,
1004
self.get_transaction())
1006
assert not isinstance(rev.revision_id, unicode)
1007
for parent_id in rev.parent_ids:
1008
assert not isinstance(parent_id, unicode)
1012
def get_revision_xml(self, revision_id):
1013
# TODO: jam 20070210 This shouldn't be necessary since get_revision
1014
# would have already do it.
1015
# TODO: jam 20070210 Just use _serializer.write_revision_to_string()
1016
rev = self.get_revision(revision_id)
1017
rev_tmp = StringIO()
1018
# the current serializer..
1019
self._revision_store._serializer.write_revision(rev, rev_tmp)
1021
return rev_tmp.getvalue()
1024
def get_deltas_for_revisions(self, revisions):
1025
"""Produce a generator of revision deltas.
1027
Note that the input is a sequence of REVISIONS, not revision_ids.
1028
Trees will be held in memory until the generator exits.
1029
Each delta is relative to the revision's lefthand predecessor.
1031
required_trees = set()
1032
for revision in revisions:
1033
required_trees.add(revision.revision_id)
1034
required_trees.update(revision.parent_ids[:1])
1035
trees = dict((t.get_revision_id(), t) for
1036
t in self.revision_trees(required_trees))
1037
for revision in revisions:
1038
if not revision.parent_ids:
1039
old_tree = self.revision_tree(None)
1041
old_tree = trees[revision.parent_ids[0]]
1042
yield trees[revision.revision_id].changes_from(old_tree)
1045
def get_revision_delta(self, revision_id):
1046
"""Return the delta for one revision.
1048
The delta is relative to the left-hand predecessor of the
1051
r = self.get_revision(revision_id)
1052
return list(self.get_deltas_for_revisions([r]))[0]
1055
def store_revision_signature(self, gpg_strategy, plaintext, revision_id):
1056
signature = gpg_strategy.sign(plaintext)
1057
self.add_signature_text(revision_id, signature)
1060
def add_signature_text(self, revision_id, signature):
1061
self._revision_store.add_revision_signature_text(revision_id,
1063
self.get_transaction())
1065
def find_text_key_references(self):
1066
"""Find the text key references within the repository.
1068
:return: a dictionary mapping (file_id, revision_id) tuples to altered file-ids to an iterable of
1069
revision_ids. Each altered file-ids has the exact revision_ids that
1070
altered it listed explicitly.
1071
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1072
to whether they were referred to by the inventory of the
1073
revision_id that they contain. The inventory texts from all present
1074
revision ids are assessed to generate this report.
1076
revision_ids = self.all_revision_ids()
1077
w = self.get_inventory_weave()
1078
pb = ui.ui_factory.nested_progress_bar()
1080
return self._find_text_key_references_from_xml_inventory_lines(
1081
w.iter_lines_added_or_present_in_versions(revision_ids, pb=pb))
1085
def _find_text_key_references_from_xml_inventory_lines(self,
1087
"""Core routine for extracting references to texts from inventories.
1089
This performs the translation of xml lines to revision ids.
1091
:param line_iterator: An iterator of lines, origin_version_id
1092
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1093
to whether they were referred to by the inventory of the
1094
revision_id that they contain. Note that if that revision_id was
1095
not part of the line_iterator's output then False will be given -
1096
even though it may actually refer to that key.
1098
if not self._serializer.support_altered_by_hack:
1099
raise AssertionError(
1100
"_find_text_key_references_from_xml_inventory_lines only "
1101
"supported for branches which store inventory as unnested xml"
1102
", not on %r" % self)
1105
# this code needs to read every new line in every inventory for the
1106
# inventories [revision_ids]. Seeing a line twice is ok. Seeing a line
1107
# not present in one of those inventories is unnecessary but not
1108
# harmful because we are filtering by the revision id marker in the
1109
# inventory lines : we only select file ids altered in one of those
1110
# revisions. We don't need to see all lines in the inventory because
1111
# only those added in an inventory in rev X can contain a revision=X
1113
unescape_revid_cache = {}
1114
unescape_fileid_cache = {}
1116
# jam 20061218 In a big fetch, this handles hundreds of thousands
1117
# of lines, so it has had a lot of inlining and optimizing done.
1118
# Sorry that it is a little bit messy.
1119
# Move several functions to be local variables, since this is a long
1121
search = self._file_ids_altered_regex.search
1122
unescape = _unescape_xml
1123
setdefault = result.setdefault
1124
for line, version_id in line_iterator:
1125
match = search(line)
1128
# One call to match.group() returning multiple items is quite a
1129
# bit faster than 2 calls to match.group() each returning 1
1130
file_id, revision_id = match.group('file_id', 'revision_id')
1132
# Inlining the cache lookups helps a lot when you make 170,000
1133
# lines and 350k ids, versus 8.4 unique ids.
1134
# Using a cache helps in 2 ways:
1135
# 1) Avoids unnecessary decoding calls
1136
# 2) Re-uses cached strings, which helps in future set and
1138
# (2) is enough that removing encoding entirely along with
1139
# the cache (so we are using plain strings) results in no
1140
# performance improvement.
1142
revision_id = unescape_revid_cache[revision_id]
1144
unescaped = unescape(revision_id)
1145
unescape_revid_cache[revision_id] = unescaped
1146
revision_id = unescaped
1148
# Note that unconditionally unescaping means that we deserialise
1149
# every fileid, which for general 'pull' is not great, but we don't
1150
# really want to have some many fulltexts that this matters anyway.
1153
file_id = unescape_fileid_cache[file_id]
1155
unescaped = unescape(file_id)
1156
unescape_fileid_cache[file_id] = unescaped
1159
key = (file_id, revision_id)
1160
setdefault(key, False)
1161
if revision_id == version_id:
1165
def _find_file_ids_from_xml_inventory_lines(self, line_iterator,
1167
"""Helper routine for fileids_altered_by_revision_ids.
1169
This performs the translation of xml lines to revision ids.
1171
:param line_iterator: An iterator of lines, origin_version_id
1172
:param revision_ids: The revision ids to filter for. This should be a
1173
set or other type which supports efficient __contains__ lookups, as
1174
the revision id from each parsed line will be looked up in the
1175
revision_ids filter.
1176
:return: a dictionary mapping altered file-ids to an iterable of
1177
revision_ids. Each altered file-ids has the exact revision_ids that
1178
altered it listed explicitly.
1181
setdefault = result.setdefault
1182
for file_id, revision_id in \
1183
self._find_text_key_references_from_xml_inventory_lines(
1184
line_iterator).iterkeys():
1185
# once data is all ensured-consistent; then this is
1186
# if revision_id == version_id
1187
if revision_id in revision_ids:
1188
setdefault(file_id, set()).add(revision_id)
1191
def fileids_altered_by_revision_ids(self, revision_ids):
1192
"""Find the file ids and versions affected by revisions.
1194
:param revisions: an iterable containing revision ids.
1195
:return: a dictionary mapping altered file-ids to an iterable of
1196
revision_ids. Each altered file-ids has the exact revision_ids that
1197
altered it listed explicitly.
1199
selected_revision_ids = set(revision_ids)
1200
w = self.get_inventory_weave()
1201
pb = ui.ui_factory.nested_progress_bar()
1203
return self._find_file_ids_from_xml_inventory_lines(
1204
w.iter_lines_added_or_present_in_versions(
1205
selected_revision_ids, pb=pb),
1206
selected_revision_ids)
1210
def iter_files_bytes(self, desired_files):
1211
"""Iterate through file versions.
1213
Files will not necessarily be returned in the order they occur in
1214
desired_files. No specific order is guaranteed.
1216
Yields pairs of identifier, bytes_iterator. identifier is an opaque
1217
value supplied by the caller as part of desired_files. It should
1218
uniquely identify the file version in the caller's context. (Examples:
1219
an index number or a TreeTransform trans_id.)
1221
bytes_iterator is an iterable of bytestrings for the file. The
1222
kind of iterable and length of the bytestrings are unspecified, but for
1223
this implementation, it is a list of lines produced by
1224
VersionedFile.get_lines().
1226
:param desired_files: a list of (file_id, revision_id, identifier)
1229
transaction = self.get_transaction()
1230
for file_id, revision_id, callable_data in desired_files:
1232
weave = self.weave_store.get_weave(file_id, transaction)
1233
except errors.NoSuchFile:
1234
raise errors.NoSuchIdInRepository(self, file_id)
1235
yield callable_data, weave.get_lines(revision_id)
1237
def _generate_text_key_index(self):
1238
"""Generate a new text key index for the repository.
1240
This is an expensive function that will take considerable time to run.
1242
:return: A dict mapping text keys ((file_id, revision_id) tuples) to a
1243
list of parents, also text keys. When a given key has no parents,
1244
the parents list will be [NULL_REVISION].
1246
# All revisions, to find inventory parents.
1247
revision_graph = self.get_revision_graph_with_ghosts()
1248
ancestors = revision_graph.get_ancestors()
1249
text_key_references = self.find_text_key_references()
1250
pb = ui.ui_factory.nested_progress_bar()
1252
return self._do_generate_text_key_index(ancestors,
1253
text_key_references, pb)
1257
def _do_generate_text_key_index(self, ancestors, text_key_references, pb):
1258
"""Helper for _generate_text_key_index to avoid deep nesting."""
1259
revision_order = tsort.topo_sort(ancestors)
1260
invalid_keys = set()
1262
for revision_id in revision_order:
1263
revision_keys[revision_id] = set()
1264
text_count = len(text_key_references)
1265
# a cache of the text keys to allow reuse; costs a dict of all the
1266
# keys, but saves a 2-tuple for every child of a given key.
1268
for text_key, valid in text_key_references.iteritems():
1270
invalid_keys.add(text_key)
1272
revision_keys[text_key[1]].add(text_key)
1273
text_key_cache[text_key] = text_key
1274
del text_key_references
1276
text_graph = graph.Graph(graph.DictParentsProvider(text_index))
1277
NULL_REVISION = _mod_revision.NULL_REVISION
1278
# Set a cache with a size of 10 - this suffices for bzr.dev but may be
1279
# too small for large or very branchy trees. However, for 55K path
1280
# trees, it would be easy to use too much memory trivially. Ideally we
1281
# could gauge this by looking at available real memory etc, but this is
1282
# always a tricky proposition.
1283
inventory_cache = lru_cache.LRUCache(10)
1284
batch_size = 10 # should be ~150MB on a 55K path tree
1285
batch_count = len(revision_order) / batch_size + 1
1287
pb.update("Calculating text parents.", processed_texts, text_count)
1288
for offset in xrange(batch_count):
1289
to_query = revision_order[offset * batch_size:(offset + 1) *
1293
for rev_tree in self.revision_trees(to_query):
1294
revision_id = rev_tree.get_revision_id()
1295
parent_ids = ancestors[revision_id]
1296
for text_key in revision_keys[revision_id]:
1297
pb.update("Calculating text parents.", processed_texts)
1298
processed_texts += 1
1299
candidate_parents = []
1300
for parent_id in parent_ids:
1301
parent_text_key = (text_key[0], parent_id)
1303
check_parent = parent_text_key not in \
1304
revision_keys[parent_id]
1306
# the parent parent_id is a ghost:
1307
check_parent = False
1308
# truncate the derived graph against this ghost.
1309
parent_text_key = None
1311
# look at the parent commit details inventories to
1312
# determine possible candidates in the per file graph.
1315
inv = inventory_cache[parent_id]
1317
inv = self.revision_tree(parent_id).inventory
1318
inventory_cache[parent_id] = inv
1319
parent_entry = inv._byid.get(text_key[0], None)
1320
if parent_entry is not None:
1322
text_key[0], parent_entry.revision)
1324
parent_text_key = None
1325
if parent_text_key is not None:
1326
candidate_parents.append(
1327
text_key_cache[parent_text_key])
1328
parent_heads = text_graph.heads(candidate_parents)
1329
new_parents = list(parent_heads)
1330
new_parents.sort(key=lambda x:candidate_parents.index(x))
1331
if new_parents == []:
1332
new_parents = [NULL_REVISION]
1333
text_index[text_key] = new_parents
1335
for text_key in invalid_keys:
1336
text_index[text_key] = [NULL_REVISION]
1339
def item_keys_introduced_by(self, revision_ids, _files_pb=None):
1340
"""Get an iterable listing the keys of all the data introduced by a set
1343
The keys will be ordered so that the corresponding items can be safely
1344
fetched and inserted in that order.
1346
:returns: An iterable producing tuples of (knit-kind, file-id,
1347
versions). knit-kind is one of 'file', 'inventory', 'signatures',
1348
'revisions'. file-id is None unless knit-kind is 'file'.
1350
# XXX: it's a bit weird to control the inventory weave caching in this
1351
# generator. Ideally the caching would be done in fetch.py I think. Or
1352
# maybe this generator should explicitly have the contract that it
1353
# should not be iterated until the previously yielded item has been
1356
inv_w = self.get_inventory_weave()
1357
inv_w.enable_cache()
1359
# file ids that changed
1360
file_ids = self.fileids_altered_by_revision_ids(revision_ids)
1362
num_file_ids = len(file_ids)
1363
for file_id, altered_versions in file_ids.iteritems():
1364
if _files_pb is not None:
1365
_files_pb.update("fetch texts", count, num_file_ids)
1367
yield ("file", file_id, altered_versions)
1368
# We're done with the files_pb. Note that it finished by the caller,
1369
# just as it was created by the caller.
1373
yield ("inventory", None, revision_ids)
1377
revisions_with_signatures = set()
1378
for rev_id in revision_ids:
1380
self.get_signature_text(rev_id)
1381
except errors.NoSuchRevision:
1385
revisions_with_signatures.add(rev_id)
1387
yield ("signatures", None, revisions_with_signatures)
1390
yield ("revisions", None, revision_ids)
1393
def get_inventory_weave(self):
1394
return self.control_weaves.get_weave('inventory',
1395
self.get_transaction())
1398
def get_inventory(self, revision_id):
1399
"""Get Inventory object by hash."""
1400
return self.deserialise_inventory(
1401
revision_id, self.get_inventory_xml(revision_id))
1403
def deserialise_inventory(self, revision_id, xml):
1404
"""Transform the xml into an inventory object.
1406
:param revision_id: The expected revision id of the inventory.
1407
:param xml: A serialised inventory.
1409
return self._serializer.read_inventory_from_string(xml, revision_id)
1411
def serialise_inventory(self, inv):
1412
return self._serializer.write_inventory_to_string(inv)
1414
def _serialise_inventory_to_lines(self, inv):
1415
return self._serializer.write_inventory_to_lines(inv)
1417
def get_serializer_format(self):
1418
return self._serializer.format_num
1421
def get_inventory_xml(self, revision_id):
1422
"""Get inventory XML as a file object."""
1424
assert isinstance(revision_id, str), type(revision_id)
1425
iw = self.get_inventory_weave()
1426
return iw.get_text(revision_id)
1428
raise errors.HistoryMissing(self, 'inventory', revision_id)
1431
def get_inventory_sha1(self, revision_id):
1432
"""Return the sha1 hash of the inventory entry
1434
return self.get_revision(revision_id).inventory_sha1
1437
def get_revision_graph(self, revision_id=None):
1438
"""Return a dictionary containing the revision graph.
1440
NB: This method should not be used as it accesses the entire graph all
1441
at once, which is much more data than most operations should require.
1443
:param revision_id: The revision_id to get a graph from. If None, then
1444
the entire revision graph is returned. This is a deprecated mode of
1445
operation and will be removed in the future.
1446
:return: a dictionary of revision_id->revision_parents_list.
1448
raise NotImplementedError(self.get_revision_graph)
1451
def get_revision_graph_with_ghosts(self, revision_ids=None):
1452
"""Return a graph of the revisions with ghosts marked as applicable.
1454
:param revision_ids: an iterable of revisions to graph or None for all.
1455
:return: a Graph object with the graph reachable from revision_ids.
1457
if 'evil' in debug.debug_flags:
1459
"get_revision_graph_with_ghosts scales with size of history.")
1460
result = deprecated_graph.Graph()
1461
if not revision_ids:
1462
pending = set(self.all_revision_ids())
1465
pending = set(revision_ids)
1466
# special case NULL_REVISION
1467
if _mod_revision.NULL_REVISION in pending:
1468
pending.remove(_mod_revision.NULL_REVISION)
1469
required = set(pending)
1472
revision_id = pending.pop()
1474
rev = self.get_revision(revision_id)
1475
except errors.NoSuchRevision:
1476
if revision_id in required:
1479
result.add_ghost(revision_id)
1481
for parent_id in rev.parent_ids:
1482
# is this queued or done ?
1483
if (parent_id not in pending and
1484
parent_id not in done):
1486
pending.add(parent_id)
1487
result.add_node(revision_id, rev.parent_ids)
1488
done.add(revision_id)
1491
def _get_history_vf(self):
1492
"""Get a versionedfile whose history graph reflects all revisions.
1494
For weave repositories, this is the inventory weave.
1496
return self.get_inventory_weave()
1498
def iter_reverse_revision_history(self, revision_id):
1499
"""Iterate backwards through revision ids in the lefthand history
1501
:param revision_id: The revision id to start with. All its lefthand
1502
ancestors will be traversed.
1504
if revision_id in (None, _mod_revision.NULL_REVISION):
1506
next_id = revision_id
1507
versionedfile = self._get_history_vf()
1510
parents = versionedfile.get_parents(next_id)
1511
if len(parents) == 0:
1514
next_id = parents[0]
1517
def get_revision_inventory(self, revision_id):
1518
"""Return inventory of a past revision."""
1519
# TODO: Unify this with get_inventory()
1520
# bzr 0.0.6 and later imposes the constraint that the inventory_id
1521
# must be the same as its revision, so this is trivial.
1522
if revision_id is None:
1523
# This does not make sense: if there is no revision,
1524
# then it is the current tree inventory surely ?!
1525
# and thus get_root_id() is something that looks at the last
1526
# commit on the branch, and the get_root_id is an inventory check.
1527
raise NotImplementedError
1528
# return Inventory(self.get_root_id())
1530
return self.get_inventory(revision_id)
1533
def is_shared(self):
1534
"""Return True if this repository is flagged as a shared repository."""
1535
raise NotImplementedError(self.is_shared)
1538
def reconcile(self, other=None, thorough=False):
1539
"""Reconcile this repository."""
1540
from bzrlib.reconcile import RepoReconciler
1541
reconciler = RepoReconciler(self, thorough=thorough)
1542
reconciler.reconcile()
1545
def _refresh_data(self):
1546
"""Helper called from lock_* to ensure coherency with disk.
1548
The default implementation does nothing; it is however possible
1549
for repositories to maintain loaded indices across multiple locks
1550
by checking inside their implementation of this method to see
1551
whether their indices are still valid. This depends of course on
1552
the disk format being validatable in this manner.
1556
def revision_tree(self, revision_id):
1557
"""Return Tree for a revision on this branch.
1559
`revision_id` may be None for the empty tree revision.
1561
# TODO: refactor this to use an existing revision object
1562
# so we don't need to read it in twice.
1563
if revision_id is None or revision_id == _mod_revision.NULL_REVISION:
1564
return RevisionTree(self, Inventory(root_id=None),
1565
_mod_revision.NULL_REVISION)
1567
inv = self.get_revision_inventory(revision_id)
1568
return RevisionTree(self, inv, revision_id)
1571
def revision_trees(self, revision_ids):
1572
"""Return Tree for a revision on this branch.
1574
`revision_id` may not be None or 'null:'"""
1575
assert None not in revision_ids
1576
assert _mod_revision.NULL_REVISION not in revision_ids
1577
texts = self.get_inventory_weave().get_texts(revision_ids)
1578
for text, revision_id in zip(texts, revision_ids):
1579
inv = self.deserialise_inventory(revision_id, text)
1580
yield RevisionTree(self, inv, revision_id)
1583
def get_ancestry(self, revision_id, topo_sorted=True):
1584
"""Return a list of revision-ids integrated by a revision.
1586
The first element of the list is always None, indicating the origin
1587
revision. This might change when we have history horizons, or
1588
perhaps we should have a new API.
1590
This is topologically sorted.
1592
if _mod_revision.is_null(revision_id):
1594
if not self.has_revision(revision_id):
1595
raise errors.NoSuchRevision(self, revision_id)
1596
w = self.get_inventory_weave()
1597
candidates = w.get_ancestry(revision_id, topo_sorted)
1598
return [None] + candidates # self._eliminate_revisions_not_present(candidates)
1601
"""Compress the data within the repository.
1603
This operation only makes sense for some repository types. For other
1604
types it should be a no-op that just returns.
1606
This stub method does not require a lock, but subclasses should use
1607
@needs_write_lock as this is a long running call its reasonable to
1608
implicitly lock for the user.
1612
def print_file(self, file, revision_id):
1613
"""Print `file` to stdout.
1615
FIXME RBC 20060125 as John Meinel points out this is a bad api
1616
- it writes to stdout, it assumes that that is valid etc. Fix
1617
by creating a new more flexible convenience function.
1619
tree = self.revision_tree(revision_id)
1620
# use inventory as it was in that revision
1621
file_id = tree.inventory.path2id(file)
1623
# TODO: jam 20060427 Write a test for this code path
1624
# it had a bug in it, and was raising the wrong
1626
raise errors.BzrError("%r is not present in revision %s" % (file, revision_id))
1627
tree.print_file(file_id)
1629
def get_transaction(self):
1630
return self.control_files.get_transaction()
1632
def revision_parents(self, revision_id):
1633
return self.get_inventory_weave().parent_names(revision_id)
1635
def get_parents(self, revision_ids):
1636
"""See StackedParentsProvider.get_parents"""
1638
for revision_id in revision_ids:
1639
if revision_id == _mod_revision.NULL_REVISION:
1643
parents = self.get_revision(revision_id).parent_ids
1644
except errors.NoSuchRevision:
1647
if len(parents) == 0:
1648
parents = [_mod_revision.NULL_REVISION]
1649
parents_list.append(parents)
1652
def _make_parents_provider(self):
1655
def get_graph(self, other_repository=None):
1656
"""Return the graph walker for this repository format"""
1657
parents_provider = self._make_parents_provider()
1658
if (other_repository is not None and
1659
other_repository.bzrdir.transport.base !=
1660
self.bzrdir.transport.base):
1661
parents_provider = graph._StackedParentsProvider(
1662
[parents_provider, other_repository._make_parents_provider()])
1663
return graph.Graph(parents_provider)
1665
def get_versioned_file_checker(self):
1666
"""Return an object suitable for checking versioned files."""
1667
return VersionedFileChecker(self)
1670
def set_make_working_trees(self, new_value):
1671
"""Set the policy flag for making working trees when creating branches.
1673
This only applies to branches that use this repository.
1675
The default is 'True'.
1676
:param new_value: True to restore the default, False to disable making
1679
raise NotImplementedError(self.set_make_working_trees)
1681
def make_working_trees(self):
1682
"""Returns the policy for making working trees on new branches."""
1683
raise NotImplementedError(self.make_working_trees)
1686
def sign_revision(self, revision_id, gpg_strategy):
1687
plaintext = Testament.from_revision(self, revision_id).as_short_text()
1688
self.store_revision_signature(gpg_strategy, plaintext, revision_id)
1691
def has_signature_for_revision_id(self, revision_id):
1692
"""Query for a revision signature for revision_id in the repository."""
1693
return self._revision_store.has_signature(revision_id,
1694
self.get_transaction())
1697
def get_signature_text(self, revision_id):
1698
"""Return the text for a signature."""
1699
return self._revision_store.get_signature_text(revision_id,
1700
self.get_transaction())
1703
def check(self, revision_ids=None):
1704
"""Check consistency of all history of given revision_ids.
1706
Different repository implementations should override _check().
1708
:param revision_ids: A non-empty list of revision_ids whose ancestry
1709
will be checked. Typically the last revision_id of a branch.
1711
return self._check(revision_ids)
1713
def _check(self, revision_ids):
1714
result = check.Check(self)
1718
def _warn_if_deprecated(self):
1719
global _deprecation_warning_done
1720
if _deprecation_warning_done:
1722
_deprecation_warning_done = True
1723
warning("Format %s for %s is deprecated - please use 'bzr upgrade' to get better performance"
1724
% (self._format, self.bzrdir.transport.base))
1726
def supports_rich_root(self):
1727
return self._format.rich_root_data
1729
def _check_ascii_revisionid(self, revision_id, method):
1730
"""Private helper for ascii-only repositories."""
1731
# weave repositories refuse to store revisionids that are non-ascii.
1732
if revision_id is not None:
1733
# weaves require ascii revision ids.
1734
if isinstance(revision_id, unicode):
1736
revision_id.encode('ascii')
1737
except UnicodeEncodeError:
1738
raise errors.NonAsciiRevisionId(method, self)
1741
revision_id.decode('ascii')
1742
except UnicodeDecodeError:
1743
raise errors.NonAsciiRevisionId(method, self)
1745
def revision_graph_can_have_wrong_parents(self):
1746
"""Is it possible for this repository to have a revision graph with
1749
If True, then this repository must also implement
1750
_find_inconsistent_revision_parents so that check and reconcile can
1751
check for inconsistencies before proceeding with other checks that may
1752
depend on the revision index being consistent.
1754
raise NotImplementedError(self.revision_graph_can_have_wrong_parents)
1756
# remove these delegates a while after bzr 0.15
1757
def __make_delegated(name, from_module):
1758
def _deprecated_repository_forwarder():
1759
symbol_versioning.warn('%s moved to %s in bzr 0.15'
1760
% (name, from_module),
1763
m = __import__(from_module, globals(), locals(), [name])
1765
return getattr(m, name)
1766
except AttributeError:
1767
raise AttributeError('module %s has no name %s'
1769
globals()[name] = _deprecated_repository_forwarder
1772
'AllInOneRepository',
1773
'WeaveMetaDirRepository',
1774
'PreSplitOutRepositoryFormat',
1775
'RepositoryFormat4',
1776
'RepositoryFormat5',
1777
'RepositoryFormat6',
1778
'RepositoryFormat7',
1780
__make_delegated(_name, 'bzrlib.repofmt.weaverepo')
1784
'RepositoryFormatKnit',
1785
'RepositoryFormatKnit1',
1787
__make_delegated(_name, 'bzrlib.repofmt.knitrepo')
1790
def install_revision(repository, rev, revision_tree):
1791
"""Install all revision data into a repository."""
1792
install_revisions(repository, [(rev, revision_tree, None)])
1795
def install_revisions(repository, iterable):
1796
"""Install all revision data into a repository.
1798
Accepts an iterable of revision, tree, signature tuples. The signature
1801
repository.start_write_group()
1803
for revision, revision_tree, signature in iterable:
1804
_install_revision(repository, revision, revision_tree, signature)
1806
repository.abort_write_group()
1809
repository.commit_write_group()
1812
def _install_revision(repository, rev, revision_tree, signature):
1813
"""Install all revision data into a repository."""
1814
present_parents = []
1816
for p_id in rev.parent_ids:
1817
if repository.has_revision(p_id):
1818
present_parents.append(p_id)
1819
parent_trees[p_id] = repository.revision_tree(p_id)
1821
parent_trees[p_id] = repository.revision_tree(None)
1823
inv = revision_tree.inventory
1824
entries = inv.iter_entries()
1825
# backwards compatibility hack: skip the root id.
1826
if not repository.supports_rich_root():
1827
path, root = entries.next()
1828
if root.revision != rev.revision_id:
1829
raise errors.IncompatibleRevision(repr(repository))
1830
# Add the texts that are not already present
1831
for path, ie in entries:
1832
w = repository.weave_store.get_weave_or_empty(ie.file_id,
1833
repository.get_transaction())
1834
if ie.revision not in w:
1836
# FIXME: TODO: The following loop *may* be overlapping/duplicate
1837
# with InventoryEntry.find_previous_heads(). if it is, then there
1838
# is a latent bug here where the parents may have ancestors of each
1840
for revision, tree in parent_trees.iteritems():
1841
if ie.file_id not in tree:
1843
parent_id = tree.inventory[ie.file_id].revision
1844
if parent_id in text_parents:
1846
text_parents.append(parent_id)
1848
vfile = repository.weave_store.get_weave_or_empty(ie.file_id,
1849
repository.get_transaction())
1850
lines = revision_tree.get_file(ie.file_id).readlines()
1851
vfile.add_lines(rev.revision_id, text_parents, lines)
1853
# install the inventory
1854
repository.add_inventory(rev.revision_id, inv, present_parents)
1855
except errors.RevisionAlreadyPresent:
1857
if signature is not None:
1858
repository.add_signature_text(rev.revision_id, signature)
1859
repository.add_revision(rev.revision_id, rev, inv)
1862
class MetaDirRepository(Repository):
1863
"""Repositories in the new meta-dir layout."""
1865
def __init__(self, _format, a_bzrdir, control_files, _revision_store, control_store, text_store):
1866
super(MetaDirRepository, self).__init__(_format,
1872
dir_mode = self.control_files._dir_mode
1873
file_mode = self.control_files._file_mode
1876
def is_shared(self):
1877
"""Return True if this repository is flagged as a shared repository."""
1878
return self.control_files._transport.has('shared-storage')
1881
def set_make_working_trees(self, new_value):
1882
"""Set the policy flag for making working trees when creating branches.
1884
This only applies to branches that use this repository.
1886
The default is 'True'.
1887
:param new_value: True to restore the default, False to disable making
1892
self.control_files._transport.delete('no-working-trees')
1893
except errors.NoSuchFile:
1896
self.control_files.put_utf8('no-working-trees', '')
1898
def make_working_trees(self):
1899
"""Returns the policy for making working trees on new branches."""
1900
return not self.control_files._transport.has('no-working-trees')
1903
class RepositoryFormatRegistry(registry.Registry):
1904
"""Registry of RepositoryFormats."""
1906
def get(self, format_string):
1907
r = registry.Registry.get(self, format_string)
1913
format_registry = RepositoryFormatRegistry()
1914
"""Registry of formats, indexed by their identifying format string.
1916
This can contain either format instances themselves, or classes/factories that
1917
can be called to obtain one.
1921
#####################################################################
1922
# Repository Formats
1924
class RepositoryFormat(object):
1925
"""A repository format.
1927
Formats provide three things:
1928
* An initialization routine to construct repository data on disk.
1929
* a format string which is used when the BzrDir supports versioned
1931
* an open routine which returns a Repository instance.
1933
There is one and only one Format subclass for each on-disk format. But
1934
there can be one Repository subclass that is used for several different
1935
formats. The _format attribute on a Repository instance can be used to
1936
determine the disk format.
1938
Formats are placed in an dict by their format string for reference
1939
during opening. These should be subclasses of RepositoryFormat
1942
Once a format is deprecated, just deprecate the initialize and open
1943
methods on the format class. Do not deprecate the object, as the
1944
object will be created every system load.
1946
Common instance attributes:
1947
_matchingbzrdir - the bzrdir format that the repository format was
1948
originally written to work with. This can be used if manually
1949
constructing a bzrdir and repository, or more commonly for test suite
1953
# Set to True or False in derived classes. True indicates that the format
1954
# supports ghosts gracefully.
1955
supports_ghosts = None
1958
return "<%s>" % self.__class__.__name__
1960
def __eq__(self, other):
1961
# format objects are generally stateless
1962
return isinstance(other, self.__class__)
1964
def __ne__(self, other):
1965
return not self == other
1968
def find_format(klass, a_bzrdir):
1969
"""Return the format for the repository object in a_bzrdir.
1971
This is used by bzr native formats that have a "format" file in
1972
the repository. Other methods may be used by different types of
1976
transport = a_bzrdir.get_repository_transport(None)
1977
format_string = transport.get("format").read()
1978
return format_registry.get(format_string)
1979
except errors.NoSuchFile:
1980
raise errors.NoRepositoryPresent(a_bzrdir)
1982
raise errors.UnknownFormatError(format=format_string)
1985
def register_format(klass, format):
1986
format_registry.register(format.get_format_string(), format)
1989
def unregister_format(klass, format):
1990
format_registry.remove(format.get_format_string())
1993
def get_default_format(klass):
1994
"""Return the current default format."""
1995
from bzrlib import bzrdir
1996
return bzrdir.format_registry.make_bzrdir('default').repository_format
1998
def _get_control_store(self, repo_transport, control_files):
1999
"""Return the control store for this repository."""
2000
raise NotImplementedError(self._get_control_store)
2002
def get_format_string(self):
2003
"""Return the ASCII format string that identifies this format.
2005
Note that in pre format ?? repositories the format string is
2006
not permitted nor written to disk.
2008
raise NotImplementedError(self.get_format_string)
2010
def get_format_description(self):
2011
"""Return the short description for this format."""
2012
raise NotImplementedError(self.get_format_description)
2014
def _get_revision_store(self, repo_transport, control_files):
2015
"""Return the revision store object for this a_bzrdir."""
2016
raise NotImplementedError(self._get_revision_store)
2018
def _get_text_rev_store(self,
2025
"""Common logic for getting a revision store for a repository.
2027
see self._get_revision_store for the subclass-overridable method to
2028
get the store for a repository.
2030
from bzrlib.store.revision.text import TextRevisionStore
2031
dir_mode = control_files._dir_mode
2032
file_mode = control_files._file_mode
2033
text_store = TextStore(transport.clone(name),
2035
compressed=compressed,
2037
file_mode=file_mode)
2038
_revision_store = TextRevisionStore(text_store, serializer)
2039
return _revision_store
2041
# TODO: this shouldn't be in the base class, it's specific to things that
2042
# use weaves or knits -- mbp 20070207
2043
def _get_versioned_file_store(self,
2048
versionedfile_class=None,
2049
versionedfile_kwargs={},
2051
if versionedfile_class is None:
2052
versionedfile_class = self._versionedfile_class
2053
weave_transport = control_files._transport.clone(name)
2054
dir_mode = control_files._dir_mode
2055
file_mode = control_files._file_mode
2056
return VersionedFileStore(weave_transport, prefixed=prefixed,
2058
file_mode=file_mode,
2059
versionedfile_class=versionedfile_class,
2060
versionedfile_kwargs=versionedfile_kwargs,
2063
def initialize(self, a_bzrdir, shared=False):
2064
"""Initialize a repository of this format in a_bzrdir.
2066
:param a_bzrdir: The bzrdir to put the new repository in it.
2067
:param shared: The repository should be initialized as a sharable one.
2068
:returns: The new repository object.
2070
This may raise UninitializableFormat if shared repository are not
2071
compatible the a_bzrdir.
2073
raise NotImplementedError(self.initialize)
2075
def is_supported(self):
2076
"""Is this format supported?
2078
Supported formats must be initializable and openable.
2079
Unsupported formats may not support initialization or committing or
2080
some other features depending on the reason for not being supported.
2084
def check_conversion_target(self, target_format):
2085
raise NotImplementedError(self.check_conversion_target)
2087
def open(self, a_bzrdir, _found=False):
2088
"""Return an instance of this format for the bzrdir a_bzrdir.
2090
_found is a private parameter, do not use it.
2092
raise NotImplementedError(self.open)
2095
class MetaDirRepositoryFormat(RepositoryFormat):
2096
"""Common base class for the new repositories using the metadir layout."""
2098
rich_root_data = False
2099
supports_tree_reference = False
2100
_matchingbzrdir = bzrdir.BzrDirMetaFormat1()
2103
super(MetaDirRepositoryFormat, self).__init__()
2105
def _create_control_files(self, a_bzrdir):
2106
"""Create the required files and the initial control_files object."""
2107
# FIXME: RBC 20060125 don't peek under the covers
2108
# NB: no need to escape relative paths that are url safe.
2109
repository_transport = a_bzrdir.get_repository_transport(self)
2110
control_files = lockable_files.LockableFiles(repository_transport,
2111
'lock', lockdir.LockDir)
2112
control_files.create_lock()
2113
return control_files
2115
def _upload_blank_content(self, a_bzrdir, dirs, files, utf8_files, shared):
2116
"""Upload the initial blank content."""
2117
control_files = self._create_control_files(a_bzrdir)
2118
control_files.lock_write()
2120
control_files._transport.mkdir_multi(dirs,
2121
mode=control_files._dir_mode)
2122
for file, content in files:
2123
control_files.put(file, content)
2124
for file, content in utf8_files:
2125
control_files.put_utf8(file, content)
2127
control_files.put_utf8('shared-storage', '')
2129
control_files.unlock()
2132
# formats which have no format string are not discoverable
2133
# and not independently creatable, so are not registered. They're
2134
# all in bzrlib.repofmt.weaverepo now. When an instance of one of these is
2135
# needed, it's constructed directly by the BzrDir. Non-native formats where
2136
# the repository is not separately opened are similar.
2138
format_registry.register_lazy(
2139
'Bazaar-NG Repository format 7',
2140
'bzrlib.repofmt.weaverepo',
2144
# KEEP in sync with bzrdir.format_registry default, which controls the overall
2145
# default control directory format
2146
format_registry.register_lazy(
2147
'Bazaar-NG Knit Repository Format 1',
2148
'bzrlib.repofmt.knitrepo',
2149
'RepositoryFormatKnit1',
2151
format_registry.default_key = 'Bazaar-NG Knit Repository Format 1'
2153
format_registry.register_lazy(
2154
'Bazaar Knit Repository Format 3 (bzr 0.15)\n',
2155
'bzrlib.repofmt.knitrepo',
2156
'RepositoryFormatKnit3',
2159
format_registry.register_lazy(
2160
'Bazaar Knit Repository Format 4 (bzr 1.0)\n',
2161
'bzrlib.repofmt.knitrepo',
2162
'RepositoryFormatKnit4',
2165
# Pack-based formats. There is one format for pre-subtrees, and one for
2166
# post-subtrees to allow ease of testing.
2167
# NOTE: These are experimental in 0.92.
2168
format_registry.register_lazy(
2169
'Bazaar pack repository format 1 (needs bzr 0.92)\n',
2170
'bzrlib.repofmt.pack_repo',
2171
'RepositoryFormatKnitPack1',
2173
format_registry.register_lazy(
2174
'Bazaar pack repository format 1 with subtree support (needs bzr 0.92)\n',
2175
'bzrlib.repofmt.pack_repo',
2176
'RepositoryFormatKnitPack3',
2180
class InterRepository(InterObject):
2181
"""This class represents operations taking place between two repositories.
2183
Its instances have methods like copy_content and fetch, and contain
2184
references to the source and target repositories these operations can be
2187
Often we will provide convenience methods on 'repository' which carry out
2188
operations with another repository - they will always forward to
2189
InterRepository.get(other).method_name(parameters).
2193
"""The available optimised InterRepository types."""
2195
def copy_content(self, revision_id=None):
2196
raise NotImplementedError(self.copy_content)
2198
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2199
"""Fetch the content required to construct revision_id.
2201
The content is copied from self.source to self.target.
2203
:param revision_id: if None all content is copied, if NULL_REVISION no
2205
:param pb: optional progress bar to use for progress reports. If not
2206
provided a default one will be created.
2208
Returns the copied revision count and the failed revisions in a tuple:
2211
raise NotImplementedError(self.fetch)
2214
def missing_revision_ids(self, revision_id=None, find_ghosts=True):
2215
"""Return the revision ids that source has that target does not.
2217
These are returned in topological order.
2219
:param revision_id: only return revision ids included by this
2222
# generic, possibly worst case, slow code path.
2223
target_ids = set(self.target.all_revision_ids())
2224
if revision_id is not None:
2225
source_ids = self.source.get_ancestry(revision_id)
2226
assert source_ids[0] is None
2229
source_ids = self.source.all_revision_ids()
2230
result_set = set(source_ids).difference(target_ids)
2231
# this may look like a no-op: its not. It preserves the ordering
2232
# other_ids had while only returning the members from other_ids
2233
# that we've decided we need.
2234
return [rev_id for rev_id in source_ids if rev_id in result_set]
2237
def _same_model(source, target):
2238
"""True if source and target have the same data representation."""
2239
if source.supports_rich_root() != target.supports_rich_root():
2241
if source._serializer != target._serializer:
2246
class InterSameDataRepository(InterRepository):
2247
"""Code for converting between repositories that represent the same data.
2249
Data format and model must match for this to work.
2253
def _get_repo_format_to_test(self):
2254
"""Repository format for testing with.
2256
InterSameData can pull from subtree to subtree and from non-subtree to
2257
non-subtree, so we test this with the richest repository format.
2259
from bzrlib.repofmt import knitrepo
2260
return knitrepo.RepositoryFormatKnit3()
2263
def is_compatible(source, target):
2264
return InterRepository._same_model(source, target)
2267
def copy_content(self, revision_id=None):
2268
"""Make a complete copy of the content in self into destination.
2270
This copies both the repository's revision data, and configuration information
2271
such as the make_working_trees setting.
2273
This is a destructive operation! Do not use it on existing
2276
:param revision_id: Only copy the content needed to construct
2277
revision_id and its parents.
2280
self.target.set_make_working_trees(self.source.make_working_trees())
2281
except NotImplementedError:
2283
# but don't bother fetching if we have the needed data now.
2284
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
2285
self.target.has_revision(revision_id)):
2287
self.target.fetch(self.source, revision_id=revision_id)
2290
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2291
"""See InterRepository.fetch()."""
2292
from bzrlib.fetch import GenericRepoFetcher
2293
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2294
self.source, self.source._format, self.target,
2295
self.target._format)
2296
f = GenericRepoFetcher(to_repository=self.target,
2297
from_repository=self.source,
2298
last_revision=revision_id,
2300
return f.count_copied, f.failed_revisions
2303
class InterWeaveRepo(InterSameDataRepository):
2304
"""Optimised code paths between Weave based repositories.
2306
This should be in bzrlib/repofmt/weaverepo.py but we have not yet
2307
implemented lazy inter-object optimisation.
2311
def _get_repo_format_to_test(self):
2312
from bzrlib.repofmt import weaverepo
2313
return weaverepo.RepositoryFormat7()
2316
def is_compatible(source, target):
2317
"""Be compatible with known Weave formats.
2319
We don't test for the stores being of specific types because that
2320
could lead to confusing results, and there is no need to be
2323
from bzrlib.repofmt.weaverepo import (
2329
return (isinstance(source._format, (RepositoryFormat5,
2331
RepositoryFormat7)) and
2332
isinstance(target._format, (RepositoryFormat5,
2334
RepositoryFormat7)))
2335
except AttributeError:
2339
def copy_content(self, revision_id=None):
2340
"""See InterRepository.copy_content()."""
2341
# weave specific optimised path:
2343
self.target.set_make_working_trees(self.source.make_working_trees())
2344
except NotImplementedError:
2346
# FIXME do not peek!
2347
if self.source.control_files._transport.listable():
2348
pb = ui.ui_factory.nested_progress_bar()
2350
self.target.weave_store.copy_all_ids(
2351
self.source.weave_store,
2353
from_transaction=self.source.get_transaction(),
2354
to_transaction=self.target.get_transaction())
2355
pb.update('copying inventory', 0, 1)
2356
self.target.control_weaves.copy_multi(
2357
self.source.control_weaves, ['inventory'],
2358
from_transaction=self.source.get_transaction(),
2359
to_transaction=self.target.get_transaction())
2360
self.target._revision_store.text_store.copy_all_ids(
2361
self.source._revision_store.text_store,
2366
self.target.fetch(self.source, revision_id=revision_id)
2369
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2370
"""See InterRepository.fetch()."""
2371
from bzrlib.fetch import GenericRepoFetcher
2372
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2373
self.source, self.source._format, self.target, self.target._format)
2374
f = GenericRepoFetcher(to_repository=self.target,
2375
from_repository=self.source,
2376
last_revision=revision_id,
2378
return f.count_copied, f.failed_revisions
2381
def missing_revision_ids(self, revision_id=None, find_ghosts=True):
2382
"""See InterRepository.missing_revision_ids()."""
2383
# we want all revisions to satisfy revision_id in source.
2384
# but we don't want to stat every file here and there.
2385
# we want then, all revisions other needs to satisfy revision_id
2386
# checked, but not those that we have locally.
2387
# so the first thing is to get a subset of the revisions to
2388
# satisfy revision_id in source, and then eliminate those that
2389
# we do already have.
2390
# this is slow on high latency connection to self, but as as this
2391
# disk format scales terribly for push anyway due to rewriting
2392
# inventory.weave, this is considered acceptable.
2394
if revision_id is not None:
2395
source_ids = self.source.get_ancestry(revision_id)
2396
assert source_ids[0] is None
2399
source_ids = self.source._all_possible_ids()
2400
source_ids_set = set(source_ids)
2401
# source_ids is the worst possible case we may need to pull.
2402
# now we want to filter source_ids against what we actually
2403
# have in target, but don't try to check for existence where we know
2404
# we do not have a revision as that would be pointless.
2405
target_ids = set(self.target._all_possible_ids())
2406
possibly_present_revisions = target_ids.intersection(source_ids_set)
2407
actually_present_revisions = set(self.target._eliminate_revisions_not_present(possibly_present_revisions))
2408
required_revisions = source_ids_set.difference(actually_present_revisions)
2409
required_topo_revisions = [rev_id for rev_id in source_ids if rev_id in required_revisions]
2410
if revision_id is not None:
2411
# we used get_ancestry to determine source_ids then we are assured all
2412
# revisions referenced are present as they are installed in topological order.
2413
# and the tip revision was validated by get_ancestry.
2414
return required_topo_revisions
2416
# if we just grabbed the possibly available ids, then
2417
# we only have an estimate of whats available and need to validate
2418
# that against the revision records.
2419
return self.source._eliminate_revisions_not_present(required_topo_revisions)
2422
class InterKnitRepo(InterSameDataRepository):
2423
"""Optimised code paths between Knit based repositories."""
2426
def _get_repo_format_to_test(self):
2427
from bzrlib.repofmt import knitrepo
2428
return knitrepo.RepositoryFormatKnit1()
2431
def is_compatible(source, target):
2432
"""Be compatible with known Knit formats.
2434
We don't test for the stores being of specific types because that
2435
could lead to confusing results, and there is no need to be
2438
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit
2440
are_knits = (isinstance(source._format, RepositoryFormatKnit) and
2441
isinstance(target._format, RepositoryFormatKnit))
2442
except AttributeError:
2444
return are_knits and InterRepository._same_model(source, target)
2447
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2448
"""See InterRepository.fetch()."""
2449
from bzrlib.fetch import KnitRepoFetcher
2450
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2451
self.source, self.source._format, self.target, self.target._format)
2452
f = KnitRepoFetcher(to_repository=self.target,
2453
from_repository=self.source,
2454
last_revision=revision_id,
2456
return f.count_copied, f.failed_revisions
2459
def missing_revision_ids(self, revision_id=None, find_ghosts=True):
2460
"""See InterRepository.missing_revision_ids()."""
2461
if revision_id is not None:
2462
source_ids = self.source.get_ancestry(revision_id)
2463
assert source_ids[0] is None
2466
source_ids = self.source.all_revision_ids()
2467
source_ids_set = set(source_ids)
2468
# source_ids is the worst possible case we may need to pull.
2469
# now we want to filter source_ids against what we actually
2470
# have in target, but don't try to check for existence where we know
2471
# we do not have a revision as that would be pointless.
2472
target_ids = set(self.target.all_revision_ids())
2473
possibly_present_revisions = target_ids.intersection(source_ids_set)
2474
actually_present_revisions = set(self.target._eliminate_revisions_not_present(possibly_present_revisions))
2475
required_revisions = source_ids_set.difference(actually_present_revisions)
2476
required_topo_revisions = [rev_id for rev_id in source_ids if rev_id in required_revisions]
2477
if revision_id is not None:
2478
# we used get_ancestry to determine source_ids then we are assured all
2479
# revisions referenced are present as they are installed in topological order.
2480
# and the tip revision was validated by get_ancestry.
2481
return required_topo_revisions
2483
# if we just grabbed the possibly available ids, then
2484
# we only have an estimate of whats available and need to validate
2485
# that against the revision records.
2486
return self.source._eliminate_revisions_not_present(required_topo_revisions)
2489
class InterPackRepo(InterSameDataRepository):
2490
"""Optimised code paths between Pack based repositories."""
2493
def _get_repo_format_to_test(self):
2494
from bzrlib.repofmt import pack_repo
2495
return pack_repo.RepositoryFormatKnitPack1()
2498
def is_compatible(source, target):
2499
"""Be compatible with known Pack formats.
2501
We don't test for the stores being of specific types because that
2502
could lead to confusing results, and there is no need to be
2505
from bzrlib.repofmt.pack_repo import RepositoryFormatPack
2507
are_packs = (isinstance(source._format, RepositoryFormatPack) and
2508
isinstance(target._format, RepositoryFormatPack))
2509
except AttributeError:
2511
return are_packs and InterRepository._same_model(source, target)
2514
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2515
"""See InterRepository.fetch()."""
2516
from bzrlib.repofmt.pack_repo import Packer
2517
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2518
self.source, self.source._format, self.target, self.target._format)
2519
self.count_copied = 0
2520
if revision_id is None:
2522
# everything to do - use pack logic
2523
# to fetch from all packs to one without
2524
# inventory parsing etc, IFF nothing to be copied is in the target.
2526
revision_ids = self.source.all_revision_ids()
2527
# implementing the TODO will involve:
2528
# - detecting when all of a pack is selected
2529
# - avoiding as much as possible pre-selection, so the
2530
# more-core routines such as create_pack_from_packs can filter in
2531
# a just-in-time fashion. (though having a HEADS list on a
2532
# repository might make this a lot easier, because we could
2533
# sensibly detect 'new revisions' without doing a full index scan.
2534
elif _mod_revision.is_null(revision_id):
2539
revision_ids = self.missing_revision_ids(revision_id,
2540
find_ghosts=find_ghosts)
2541
except errors.NoSuchRevision:
2542
raise errors.InstallFailed([revision_id])
2543
packs = self.source._pack_collection.all_packs()
2544
pack = Packer(self.target._pack_collection, packs, '.fetch',
2545
revision_ids).pack()
2546
if pack is not None:
2547
self.target._pack_collection._save_pack_names()
2548
# Trigger an autopack. This may duplicate effort as we've just done
2549
# a pack creation, but for now it is simpler to think about as
2550
# 'upload data, then repack if needed'.
2551
self.target._pack_collection.autopack()
2552
return (pack.get_revision_count(), [])
2557
def missing_revision_ids(self, revision_id=None, find_ghosts=True):
2558
"""See InterRepository.missing_revision_ids().
2560
:param find_ghosts: Find ghosts throughough the ancestry of
2563
if not find_ghosts and revision_id is not None:
2564
graph = self.source.get_graph()
2565
missing_revs = set()
2566
searcher = graph._make_breadth_first_searcher([revision_id])
2568
self.target._pack_collection.revision_index.combined_index
2569
null_set = frozenset([_mod_revision.NULL_REVISION])
2572
next_revs = set(searcher.next())
2573
except StopIteration:
2575
next_revs.difference_update(null_set)
2576
target_keys = [(key,) for key in next_revs]
2577
have_revs = frozenset(node[1][0] for node in
2578
target_index.iter_entries(target_keys))
2579
missing_revs.update(next_revs - have_revs)
2580
searcher.stop_searching_any(have_revs)
2581
if next_revs - have_revs == set([revision_id]):
2582
# we saw the start rev itself, but no parents from it (or
2583
# next_revs would have been updated to e.g. set(). We remove
2584
# have_revs because if we found revision_id locally we
2585
# stop_searching at the first time around.
2586
raise errors.NoSuchRevision(self.source, revision_id)
2588
elif revision_id is not None:
2589
source_ids = self.source.get_ancestry(revision_id)
2590
assert source_ids[0] is None
2593
source_ids = self.source.all_revision_ids()
2594
# source_ids is the worst possible case we may need to pull.
2595
# now we want to filter source_ids against what we actually
2596
# have in target, but don't try to check for existence where we know
2597
# we do not have a revision as that would be pointless.
2598
target_ids = set(self.target.all_revision_ids())
2599
return [r for r in source_ids if (r not in target_ids)]
2602
class InterModel1and2(InterRepository):
2605
def _get_repo_format_to_test(self):
2609
def is_compatible(source, target):
2610
if not source.supports_rich_root() and target.supports_rich_root():
2616
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2617
"""See InterRepository.fetch()."""
2618
from bzrlib.fetch import Model1toKnit2Fetcher
2619
f = Model1toKnit2Fetcher(to_repository=self.target,
2620
from_repository=self.source,
2621
last_revision=revision_id,
2623
return f.count_copied, f.failed_revisions
2626
def copy_content(self, revision_id=None):
2627
"""Make a complete copy of the content in self into destination.
2629
This is a destructive operation! Do not use it on existing
2632
:param revision_id: Only copy the content needed to construct
2633
revision_id and its parents.
2636
self.target.set_make_working_trees(self.source.make_working_trees())
2637
except NotImplementedError:
2639
# but don't bother fetching if we have the needed data now.
2640
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
2641
self.target.has_revision(revision_id)):
2643
self.target.fetch(self.source, revision_id=revision_id)
2646
class InterKnit1and2(InterKnitRepo):
2649
def _get_repo_format_to_test(self):
2653
def is_compatible(source, target):
2654
"""Be compatible with Knit1 source and Knit3 target"""
2655
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit3
2657
from bzrlib.repofmt.knitrepo import (RepositoryFormatKnit1,
2658
RepositoryFormatKnit3)
2659
from bzrlib.repofmt.pack_repo import (RepositoryFormatKnitPack1,
2660
RepositoryFormatKnitPack3)
2661
return (isinstance(source._format,
2662
(RepositoryFormatKnit1, RepositoryFormatKnitPack1)) and
2663
isinstance(target._format,
2664
(RepositoryFormatKnit3, RepositoryFormatKnitPack3))
2666
except AttributeError:
2670
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2671
"""See InterRepository.fetch()."""
2672
from bzrlib.fetch import Knit1to2Fetcher
2673
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2674
self.source, self.source._format, self.target,
2675
self.target._format)
2676
f = Knit1to2Fetcher(to_repository=self.target,
2677
from_repository=self.source,
2678
last_revision=revision_id,
2680
return f.count_copied, f.failed_revisions
2683
class InterDifferingSerializer(InterKnitRepo):
2686
def _get_repo_format_to_test(self):
2690
def is_compatible(source, target):
2691
"""Be compatible with Knit2 source and Knit3 target"""
2692
if source.supports_rich_root() != target.supports_rich_root():
2694
# Ideally, we'd support fetching if the source had no tree references
2695
# even if it supported them...
2696
if (getattr(source, '_format.supports_tree_reference', False) and
2697
not getattr(target, '_format.supports_tree_reference', False)):
2702
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2703
"""See InterRepository.fetch()."""
2704
revision_ids = self.target.missing_revision_ids(self.source,
2706
def revisions_iterator():
2707
for current_revision_id in revision_ids:
2708
revision = self.source.get_revision(current_revision_id)
2709
tree = self.source.revision_tree(current_revision_id)
2711
signature = self.source.get_signature_text(
2712
current_revision_id)
2713
except errors.NoSuchRevision:
2715
yield revision, tree, signature
2716
install_revisions(self.target, revisions_iterator())
2717
return len(revision_ids), 0
2720
class InterRemoteToOther(InterRepository):
2722
def __init__(self, source, target):
2723
InterRepository.__init__(self, source, target)
2724
self._real_inter = None
2727
def is_compatible(source, target):
2728
if not isinstance(source, remote.RemoteRepository):
2730
source._ensure_real()
2731
real_source = source._real_repository
2732
# Is source's model compatible with target's model, and are they the
2733
# same format? Currently we can only optimise fetching from an
2734
# identical model & format repo.
2735
assert not isinstance(real_source, remote.RemoteRepository), (
2736
"We don't support remote repos backed by remote repos yet.")
2737
return real_source._format == target._format
2740
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2741
"""See InterRepository.fetch()."""
2742
from bzrlib.fetch import RemoteToOtherFetcher
2743
mutter("Using fetch logic to copy between %s(remote) and %s(%s)",
2744
self.source, self.target, self.target._format)
2745
# TODO: jam 20070210 This should be an assert, not a translate
2746
revision_id = osutils.safe_revision_id(revision_id)
2747
f = RemoteToOtherFetcher(to_repository=self.target,
2748
from_repository=self.source,
2749
last_revision=revision_id,
2751
return f.count_copied, f.failed_revisions
2754
def _get_repo_format_to_test(self):
2758
class InterOtherToRemote(InterRepository):
2760
def __init__(self, source, target):
2761
InterRepository.__init__(self, source, target)
2762
self._real_inter = None
2765
def is_compatible(source, target):
2766
if isinstance(target, remote.RemoteRepository):
2770
def _ensure_real_inter(self):
2771
if self._real_inter is None:
2772
self.target._ensure_real()
2773
real_target = self.target._real_repository
2774
self._real_inter = InterRepository.get(self.source, real_target)
2776
def copy_content(self, revision_id=None):
2777
self._ensure_real_inter()
2778
self._real_inter.copy_content(revision_id=revision_id)
2780
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2781
self._ensure_real_inter()
2782
self._real_inter.fetch(revision_id=revision_id, pb=pb)
2785
def _get_repo_format_to_test(self):
2789
InterRepository.register_optimiser(InterDifferingSerializer)
2790
InterRepository.register_optimiser(InterSameDataRepository)
2791
InterRepository.register_optimiser(InterWeaveRepo)
2792
InterRepository.register_optimiser(InterKnitRepo)
2793
InterRepository.register_optimiser(InterModel1and2)
2794
InterRepository.register_optimiser(InterKnit1and2)
2795
InterRepository.register_optimiser(InterPackRepo)
2796
InterRepository.register_optimiser(InterRemoteToOther)
2797
InterRepository.register_optimiser(InterOtherToRemote)
2800
class CopyConverter(object):
2801
"""A repository conversion tool which just performs a copy of the content.
2803
This is slow but quite reliable.
2806
def __init__(self, target_format):
2807
"""Create a CopyConverter.
2809
:param target_format: The format the resulting repository should be.
2811
self.target_format = target_format
2813
def convert(self, repo, pb):
2814
"""Perform the conversion of to_convert, giving feedback via pb.
2816
:param to_convert: The disk object to convert.
2817
:param pb: a progress bar to use for progress information.
2822
# this is only useful with metadir layouts - separated repo content.
2823
# trigger an assertion if not such
2824
repo._format.get_format_string()
2825
self.repo_dir = repo.bzrdir
2826
self.step('Moving repository to repository.backup')
2827
self.repo_dir.transport.move('repository', 'repository.backup')
2828
backup_transport = self.repo_dir.transport.clone('repository.backup')
2829
repo._format.check_conversion_target(self.target_format)
2830
self.source_repo = repo._format.open(self.repo_dir,
2832
_override_transport=backup_transport)
2833
self.step('Creating new repository')
2834
converted = self.target_format.initialize(self.repo_dir,
2835
self.source_repo.is_shared())
2836
converted.lock_write()
2838
self.step('Copying content into repository.')
2839
self.source_repo.copy_content_into(converted)
2842
self.step('Deleting old repository content.')
2843
self.repo_dir.transport.delete_tree('repository.backup')
2844
self.pb.note('repository converted')
2846
def step(self, message):
2847
"""Update the pb by a step."""
2849
self.pb.update(message, self.count, self.total)
2861
def _unescaper(match, _map=_unescape_map):
2862
code = match.group(1)
2866
if not code.startswith('#'):
2868
return unichr(int(code[1:])).encode('utf8')
2874
def _unescape_xml(data):
2875
"""Unescape predefined XML entities in a string of data."""
2877
if _unescape_re is None:
2878
_unescape_re = re.compile('\&([^;]*);')
2879
return _unescape_re.sub(_unescaper, data)
2882
class VersionedFileChecker(object):
2884
def __init__(self, repository):
2885
self.repository = repository
2886
self.text_index = self.repository._generate_text_key_index()
2888
def calculate_file_version_parents(self, revision_id, file_id):
2889
"""Calculate the correct parents for a file version according to
2892
parent_keys = self.text_index[(file_id, revision_id)]
2893
if parent_keys == [_mod_revision.NULL_REVISION]:
2895
# strip the file_id, for the weave api
2896
return tuple([revision_id for file_id, revision_id in parent_keys])
2898
def check_file_version_parents(self, weave, file_id, planned_revisions):
2899
"""Check the parents stored in a versioned file are correct.
2901
It also detects file versions that are not referenced by their
2902
corresponding revision's inventory.
2904
:returns: A tuple of (wrong_parents, dangling_file_versions).
2905
wrong_parents is a dict mapping {revision_id: (stored_parents,
2906
correct_parents)} for each revision_id where the stored parents
2907
are not correct. dangling_file_versions is a set of (file_id,
2908
revision_id) tuples for versions that are present in this versioned
2909
file, but not used by the corresponding inventory.
2912
unused_versions = set()
2913
for num, revision_id in enumerate(planned_revisions):
2915
correct_parents = self.calculate_file_version_parents(
2916
revision_id, file_id)
2918
# we were asked to investigate a non-existant version.
2919
unused_versions.add(revision_id)
2922
knit_parents = tuple(weave.get_parents(revision_id))
2923
except errors.RevisionNotPresent:
2925
if correct_parents != knit_parents:
2926
wrong_parents[revision_id] = (knit_parents, correct_parents)
2927
return wrong_parents, unused_versions
added
removed
bzrlib/repository.py
