187
202
self.random_revid = False
189
def _heads(self, file_id, revision_ids):
190
"""Calculate the graph heads for revision_ids in the graph of file_id.
192
This can use either a per-file graph or a global revision graph as we
193
have an identity relationship between the two graphs.
195
return self.__heads(revision_ids)
197
def _check_root(self, ie, parent_invs, tree):
198
"""Helper for record_entry_contents.
200
:param ie: An entry being added.
201
:param parent_invs: The inventories of the parent revisions of the
203
:param tree: The tree that is being committed.
205
# In this revision format, root entries have no knit or weave When
206
# serializing out to disk and back in root.revision is always
208
ie.revision = self._new_revision_id
210
def _get_delta(self, ie, basis_inv, path):
211
"""Get a delta against the basis inventory for ie."""
212
if ie.file_id not in basis_inv:
214
return (None, path, ie.file_id, ie)
215
elif ie != basis_inv[ie.file_id]:
217
# TODO: avoid tis id2path call.
218
return (basis_inv.id2path(ie.file_id), path, ie.file_id, ie)
223
def record_entry_contents(self, ie, parent_invs, path, tree,
225
"""Record the content of ie from tree into the commit if needed.
227
Side effect: sets ie.revision when unchanged
229
:param ie: An inventory entry present in the commit.
230
:param parent_invs: The inventories of the parent revisions of the
232
:param path: The path the entry is at in the tree.
233
:param tree: The tree which contains this entry and should be used to
235
:param content_summary: Summary data from the tree about the paths
236
content - stat, length, exec, sha/link target. This is only
237
accessed when the entry has a revision of None - that is when it is
238
a candidate to commit.
239
:return: A tuple (change_delta, version_recorded, fs_hash).
240
change_delta is an inventory_delta change for this entry against
241
the basis tree of the commit, or None if no change occured against
243
version_recorded is True if a new version of the entry has been
244
recorded. For instance, committing a merge where a file was only
245
changed on the other side will return (delta, False).
246
fs_hash is either None, or the hash details for the path (currently
247
a tuple of the contents sha1 and the statvalue returned by
248
tree.get_file_with_stat()).
250
if self.new_inventory.root is None:
251
if ie.parent_id is not None:
252
raise errors.RootMissing()
253
self._check_root(ie, parent_invs, tree)
254
if ie.revision is None:
255
kind = content_summary[0]
257
# ie is carried over from a prior commit
259
# XXX: repository specific check for nested tree support goes here - if
260
# the repo doesn't want nested trees we skip it ?
261
if (kind == 'tree-reference' and
262
not self.repository._format.supports_tree_reference):
263
# mismatch between commit builder logic and repository:
264
# this needs the entry creation pushed down into the builder.
265
raise NotImplementedError('Missing repository subtree support.')
266
self.new_inventory.add(ie)
268
# TODO: slow, take it out of the inner loop.
270
basis_inv = parent_invs[0]
272
basis_inv = Inventory(root_id=None)
274
# ie.revision is always None if the InventoryEntry is considered
275
# for committing. We may record the previous parents revision if the
276
# content is actually unchanged against a sole head.
277
if ie.revision is not None:
278
if not self._versioned_root and path == '':
279
# repositories that do not version the root set the root's
280
# revision to the new commit even when no change occurs, and
281
# this masks when a change may have occurred against the basis,
282
# so calculate if one happened.
283
if ie.file_id in basis_inv:
284
delta = (basis_inv.id2path(ie.file_id), path,
288
delta = (None, path, ie.file_id, ie)
289
return delta, False, None
291
# we don't need to commit this, because the caller already
292
# determined that an existing revision of this file is
293
# appropriate. If its not being considered for committing then
294
# it and all its parents to the root must be unaltered so
295
# no-change against the basis.
296
if ie.revision == self._new_revision_id:
297
raise AssertionError("Impossible situation, a skipped "
298
"inventory entry (%r) claims to be modified in this "
299
"commit (%r).", (ie, self._new_revision_id))
300
return None, False, None
301
# XXX: Friction: parent_candidates should return a list not a dict
302
# so that we don't have to walk the inventories again.
303
parent_candiate_entries = ie.parent_candidates(parent_invs)
304
head_set = self._heads(ie.file_id, parent_candiate_entries.keys())
306
for inv in parent_invs:
307
if ie.file_id in inv:
308
old_rev = inv[ie.file_id].revision
309
if old_rev in head_set:
310
heads.append(inv[ie.file_id].revision)
311
head_set.remove(inv[ie.file_id].revision)
314
# now we check to see if we need to write a new record to the
316
# We write a new entry unless there is one head to the ancestors, and
317
# the kind-derived content is unchanged.
319
# Cheapest check first: no ancestors, or more the one head in the
320
# ancestors, we write a new node.
324
# There is a single head, look it up for comparison
325
parent_entry = parent_candiate_entries[heads[0]]
326
# if the non-content specific data has changed, we'll be writing a
328
if (parent_entry.parent_id != ie.parent_id or
329
parent_entry.name != ie.name):
331
# now we need to do content specific checks:
333
# if the kind changed the content obviously has
334
if kind != parent_entry.kind:
336
# Stat cache fingerprint feedback for the caller - None as we usually
337
# don't generate one.
340
if content_summary[2] is None:
341
raise ValueError("Files must not have executable = None")
343
if (# if the file length changed we have to store:
344
parent_entry.text_size != content_summary[1] or
345
# if the exec bit has changed we have to store:
346
parent_entry.executable != content_summary[2]):
348
elif parent_entry.text_sha1 == content_summary[3]:
349
# all meta and content is unchanged (using a hash cache
350
# hit to check the sha)
351
ie.revision = parent_entry.revision
352
ie.text_size = parent_entry.text_size
353
ie.text_sha1 = parent_entry.text_sha1
354
ie.executable = parent_entry.executable
355
return self._get_delta(ie, basis_inv, path), False, None
357
# Either there is only a hash change(no hash cache entry,
358
# or same size content change), or there is no change on
360
# Provide the parent's hash to the store layer, so that the
361
# content is unchanged we will not store a new node.
362
nostore_sha = parent_entry.text_sha1
364
# We want to record a new node regardless of the presence or
365
# absence of a content change in the file.
367
ie.executable = content_summary[2]
368
file_obj, stat_value = tree.get_file_with_stat(ie.file_id, path)
370
lines = file_obj.readlines()
374
ie.text_sha1, ie.text_size = self._add_text_to_weave(
375
ie.file_id, lines, heads, nostore_sha)
376
# Let the caller know we generated a stat fingerprint.
377
fingerprint = (ie.text_sha1, stat_value)
378
except errors.ExistingContent:
379
# Turns out that the file content was unchanged, and we were
380
# only going to store a new node if it was changed. Carry over
382
ie.revision = parent_entry.revision
383
ie.text_size = parent_entry.text_size
384
ie.text_sha1 = parent_entry.text_sha1
385
ie.executable = parent_entry.executable
386
return self._get_delta(ie, basis_inv, path), False, None
387
elif kind == 'directory':
389
# all data is meta here, nothing specific to directory, so
391
ie.revision = parent_entry.revision
392
return self._get_delta(ie, basis_inv, path), False, None
394
self._add_text_to_weave(ie.file_id, lines, heads, None)
395
elif kind == 'symlink':
396
current_link_target = content_summary[3]
398
# symlink target is not generic metadata, check if it has
400
if current_link_target != parent_entry.symlink_target:
403
# unchanged, carry over.
404
ie.revision = parent_entry.revision
405
ie.symlink_target = parent_entry.symlink_target
406
return self._get_delta(ie, basis_inv, path), False, None
407
ie.symlink_target = current_link_target
409
self._add_text_to_weave(ie.file_id, lines, heads, None)
410
elif kind == 'tree-reference':
412
if content_summary[3] != parent_entry.reference_revision:
415
# unchanged, carry over.
416
ie.reference_revision = parent_entry.reference_revision
417
ie.revision = parent_entry.revision
418
return self._get_delta(ie, basis_inv, path), False, None
419
ie.reference_revision = content_summary[3]
421
self._add_text_to_weave(ie.file_id, lines, heads, None)
423
raise NotImplementedError('unknown kind')
424
ie.revision = self._new_revision_id
425
return self._get_delta(ie, basis_inv, path), True, fingerprint
427
def _add_text_to_weave(self, file_id, new_lines, parents, nostore_sha):
428
# Note: as we read the content directly from the tree, we know its not
429
# been turned into unicode or badly split - but a broken tree
430
# implementation could give us bad output from readlines() so this is
431
# not a guarantee of safety. What would be better is always checking
432
# the content during test suite execution. RBC 20070912
433
parent_keys = tuple((file_id, parent) for parent in parents)
434
return self.repository.texts.add_lines(
435
(file_id, self._new_revision_id), parent_keys, new_lines,
436
nostore_sha=nostore_sha, random_id=self.random_revid,
437
check_content=False)[0:2]
440
class RootCommitBuilder(CommitBuilder):
441
"""This commitbuilder actually records the root id"""
443
# the root entry gets versioned properly by this builder.
444
_versioned_root = True
446
def _check_root(self, ie, parent_invs, tree):
447
"""Helper for record_entry_contents.
449
:param ie: An entry being added.
450
:param parent_invs: The inventories of the parent revisions of the
452
:param tree: The tree that is being committed.
204
def will_record_deletes(self):
205
"""Tell the commit builder that deletes are being notified.
207
This enables the accumulation of an inventory delta; for the resulting
208
commit to be valid, deletes against the basis MUST be recorded via
209
builder.record_delete().
211
raise NotImplementedError(self.will_record_deletes)
213
def record_iter_changes(self, tree, basis_revision_id, iter_changes):
214
"""Record a new tree via iter_changes.
216
:param tree: The tree to obtain text contents from for changed objects.
217
:param basis_revision_id: The revision id of the tree the iter_changes
218
has been generated against. Currently assumed to be the same
219
as self.parents[0] - if it is not, errors may occur.
220
:param iter_changes: An iter_changes iterator with the changes to apply
221
to basis_revision_id. The iterator must not include any items with
222
a current kind of None - missing items must be either filtered out
223
or errored-on beefore record_iter_changes sees the item.
224
:return: A generator of (file_id, relpath, fs_hash) tuples for use with
227
raise NotImplementedError(self.record_iter_changes)
230
class RepositoryWriteLockResult(LogicalLockResult):
231
"""The result of write locking a repository.
233
:ivar repository_token: The token obtained from the underlying lock, or
235
:ivar unlock: A callable which will unlock the lock.
238
def __init__(self, unlock, repository_token):
239
LogicalLockResult.__init__(self, unlock)
240
self.repository_token = repository_token
243
return "RepositoryWriteLockResult(%s, %s)" % (self.repository_token,
456
247
######################################################################
459
class Repository(object):
251
class Repository(_RelockDebugMixin, controldir.ControlComponent):
460
252
"""Repository holding history for one or more branches.
462
254
The repository holds and retrieves historical information including
463
255
revisions and file history. It's normally accessed only by the Branch,
464
256
which views a particular line of development through that history.
466
The Repository builds on top of some byte storage facilies (the revisions,
467
signatures, inventories and texts attributes) and a Transport, which
468
respectively provide byte storage and a means to access the (possibly
471
The byte storage facilities are addressed via tuples, which we refer to
472
as 'keys' throughout the code base. Revision_keys, inventory_keys and
473
signature_keys are all 1-tuples: (revision_id,). text_keys are two-tuples:
474
(file_id, revision_id). We use this interface because it allows low
475
friction with the underlying code that implements disk indices, network
476
encoding and other parts of bzrlib.
478
:ivar revisions: A bzrlib.versionedfile.VersionedFiles instance containing
479
the serialised revisions for the repository. This can be used to obtain
480
revision graph information or to access raw serialised revisions.
481
The result of trying to insert data into the repository via this store
482
is undefined: it should be considered read-only except for implementors
484
:ivar signatures: A bzrlib.versionedfile.VersionedFiles instance containing
485
the serialised signatures for the repository. This can be used to
486
obtain access to raw serialised signatures. The result of trying to
487
insert data into the repository via this store is undefined: it should
488
be considered read-only except for implementors of repositories.
489
:ivar inventories: A bzrlib.versionedfile.VersionedFiles instance containing
490
the serialised inventories for the repository. This can be used to
491
obtain unserialised inventories. The result of trying to insert data
492
into the repository via this store is undefined: it should be
493
considered read-only except for implementors of repositories.
494
:ivar texts: A bzrlib.versionedfile.VersionedFiles instance containing the
495
texts of files and directories for the repository. This can be used to
496
obtain file texts or file graphs. Note that Repository.iter_file_bytes
497
is usually a better interface for accessing file texts.
498
The result of trying to insert data into the repository via this store
499
is undefined: it should be considered read-only except for implementors
501
:ivar _transport: Transport for file access to repository, typically
502
pointing to .bzr/repository.
258
See VersionedFileRepository in bzrlib.vf_repository for the
259
base class for most Bazaar repositories.
505
# What class to use for a CommitBuilder. Often its simpler to change this
506
# in a Repository class subclass rather than to override
507
# get_commit_builder.
508
_commit_builder_class = CommitBuilder
509
# The search regex used by xml based repositories to determine what things
510
# where changed in a single commit.
511
_file_ids_altered_regex = lazy_regex.lazy_compile(
512
r'file_id="(?P<file_id>[^"]+)"'
513
r'.* revision="(?P<revision_id>[^"]+)"'
516
def abort_write_group(self):
262
def abort_write_group(self, suppress_errors=False):
517
263
"""Commit the contents accrued within the current write group.
265
:param suppress_errors: if true, abort_write_group will catch and log
266
unexpected errors that happen during the abort, rather than
267
allowing them to propagate. Defaults to False.
519
269
:seealso: start_write_group.
521
271
if self._write_group is not self.get_transaction():
522
272
# has an unlock or relock occured ?
523
raise errors.BzrError('mismatched lock context and write group.')
524
self._abort_write_group()
275
'(suppressed) mismatched lock context and write group. %r, %r',
276
self._write_group, self.get_transaction())
278
raise errors.BzrError(
279
'mismatched lock context and write group. %r, %r' %
280
(self._write_group, self.get_transaction()))
282
self._abort_write_group()
283
except Exception, exc:
284
self._write_group = None
285
if not suppress_errors:
287
mutter('abort_write_group failed')
288
log_exception_quietly()
289
note(gettext('bzr: ERROR (ignored): %s'), exc)
525
290
self._write_group = None
527
292
def _abort_write_group(self):
528
293
"""Template method for per-repository write group cleanup.
530
This is called during abort before the write group is considered to be
295
This is called during abort before the write group is considered to be
531
296
finished and should cleanup any internal state accrued during the write
532
297
group. There is no requirement that data handed to the repository be
533
298
*not* made available - this is not a rollback - but neither should any
947
622
"""Commit the contents accrued within the current write group.
949
624
:seealso: start_write_group.
626
:return: it may return an opaque hint that can be passed to 'pack'.
951
628
if self._write_group is not self.get_transaction():
952
629
# has an unlock or relock occured ?
953
630
raise errors.BzrError('mismatched lock context %r and '
954
631
'write group %r.' %
955
632
(self.get_transaction(), self._write_group))
956
self._commit_write_group()
633
result = self._commit_write_group()
957
634
self._write_group = None
959
637
def _commit_write_group(self):
960
638
"""Template method for per-repository write group cleanup.
962
This is called before the write group is considered to be
640
This is called before the write group is considered to be
963
641
finished and should ensure that all data handed to the repository
964
for writing during the write group is safely committed (to the
642
for writing during the write group is safely committed (to the
965
643
extent possible considering file system caching etc).
968
def fetch(self, source, revision_id=None, pb=None, find_ghosts=False):
646
def suspend_write_group(self):
647
"""Suspend a write group.
649
:raise UnsuspendableWriteGroup: If the write group can not be
651
:return: List of tokens
653
raise errors.UnsuspendableWriteGroup(self)
655
def refresh_data(self):
656
"""Re-read any data needed to synchronise with disk.
658
This method is intended to be called after another repository instance
659
(such as one used by a smart server) has inserted data into the
660
repository. On all repositories this will work outside of write groups.
661
Some repository formats (pack and newer for bzrlib native formats)
662
support refresh_data inside write groups. If called inside a write
663
group on a repository that does not support refreshing in a write group
664
IsInWriteGroupError will be raised.
668
def resume_write_group(self, tokens):
669
if not self.is_write_locked():
670
raise errors.NotWriteLocked(self)
671
if self._write_group:
672
raise errors.BzrError('already in a write group')
673
self._resume_write_group(tokens)
674
# so we can detect unlock/relock - the write group is now entered.
675
self._write_group = self.get_transaction()
677
def _resume_write_group(self, tokens):
678
raise errors.UnsuspendableWriteGroup(self)
680
def fetch(self, source, revision_id=None, find_ghosts=False):
969
681
"""Fetch the content required to construct revision_id from source.
971
If revision_id is None all content is copied.
683
If revision_id is None, then all content is copied.
685
fetch() may not be used when the repository is in a write group -
686
either finish the current write group before using fetch, or use
687
fetch before starting the write group.
972
689
:param find_ghosts: Find and copy revisions in the source that are
973
690
ghosts in the target (and not reachable directly by walking out to
974
691
the first-present revision in target from revision_id).
692
:param revision_id: If specified, all the content needed for this
693
revision ID will be copied to the target. Fetch will determine for
694
itself which content needs to be copied.
696
if self.is_in_write_group():
697
raise errors.InternalBzrError(
698
"May not fetch while in a write group.")
976
699
# fast path same-url fetch operations
977
if self.has_same_location(source):
700
# TODO: lift out to somewhere common with RemoteRepository
701
# <https://bugs.launchpad.net/bzr/+bug/401646>
702
if (self.has_same_location(source)
703
and self._has_same_fallbacks(source)):
978
704
# check that last_revision is in 'from' and then return a
980
706
if (revision_id is not None and
981
707
not _mod_revision.is_null(revision_id)):
982
708
self.get_revision(revision_id)
984
# if there is no specific appropriate InterRepository, this will get
985
# the InterRepository base class, which raises an
986
# IncompatibleRepositories when asked to fetch.
987
710
inter = InterRepository.get(source, self)
988
return inter.fetch(revision_id=revision_id, pb=pb,
989
find_ghosts=find_ghosts)
711
return inter.fetch(revision_id=revision_id, find_ghosts=find_ghosts)
991
713
def create_bundle(self, target, base, fileobj, format=None):
992
714
return serializer.write_bundle(self, target, base, fileobj, format)
994
def get_commit_builder(self, branch, parents, config, timestamp=None,
716
def get_commit_builder(self, branch, parents, config_stack, timestamp=None,
995
717
timezone=None, committer=None, revprops=None,
718
revision_id=None, lossy=False):
997
719
"""Obtain a CommitBuilder for this repository.
999
721
:param branch: Branch to commit to.
1000
722
:param parents: Revision ids of the parents of the new revision.
1001
:param config: Configuration to use.
723
:param config_stack: Configuration stack to use.
1002
724
:param timestamp: Optional timestamp recorded for commit.
1003
725
:param timezone: Optional timezone for timestamp.
1004
726
:param committer: Optional committer to set for commit.
1005
727
:param revprops: Optional dictionary of revision properties.
1006
728
:param revision_id: Optional revision id.
729
:param lossy: Whether to discard data that can not be natively
730
represented, when pushing to a foreign VCS
1008
result = self._commit_builder_class(self, parents, config,
1009
timestamp, timezone, committer, revprops, revision_id)
1010
self.start_write_group()
732
raise NotImplementedError(self.get_commit_builder)
734
@only_raises(errors.LockNotHeld, errors.LockBroken)
1013
735
def unlock(self):
1014
736
if (self.control_files._lock_count == 1 and
1015
737
self.control_files._lock_mode == 'w'):
1180
890
yield trees[revision.revision_id].changes_from(old_tree)
1182
892
@needs_read_lock
1183
def get_revision_delta(self, revision_id):
893
def get_revision_delta(self, revision_id, specific_fileids=None):
1184
894
"""Return the delta for one revision.
1186
896
The delta is relative to the left-hand predecessor of the
899
:param specific_fileids: if not None, the result is filtered
900
so that only those file-ids, their parents and their
901
children are included.
1189
903
r = self.get_revision(revision_id)
1190
return list(self.get_deltas_for_revisions([r]))[0]
904
return list(self.get_deltas_for_revisions([r],
905
specific_fileids=specific_fileids))[0]
1192
907
@needs_write_lock
1193
908
def store_revision_signature(self, gpg_strategy, plaintext, revision_id):
1194
909
signature = gpg_strategy.sign(plaintext)
1195
910
self.add_signature_text(revision_id, signature)
1198
912
def add_signature_text(self, revision_id, signature):
1199
self.signatures.add_lines((revision_id,), (),
1200
osutils.split_lines(signature))
1202
def find_text_key_references(self):
1203
"""Find the text key references within the repository.
1205
:return: a dictionary mapping (file_id, revision_id) tuples to altered file-ids to an iterable of
1206
revision_ids. Each altered file-ids has the exact revision_ids that
1207
altered it listed explicitly.
1208
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1209
to whether they were referred to by the inventory of the
1210
revision_id that they contain. The inventory texts from all present
1211
revision ids are assessed to generate this report.
1213
revision_keys = self.revisions.keys()
1214
w = self.inventories
1215
pb = ui.ui_factory.nested_progress_bar()
1217
return self._find_text_key_references_from_xml_inventory_lines(
1218
w.iter_lines_added_or_present_in_keys(revision_keys, pb=pb))
1222
def _find_text_key_references_from_xml_inventory_lines(self,
1224
"""Core routine for extracting references to texts from inventories.
1226
This performs the translation of xml lines to revision ids.
1228
:param line_iterator: An iterator of lines, origin_version_id
1229
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1230
to whether they were referred to by the inventory of the
1231
revision_id that they contain. Note that if that revision_id was
1232
not part of the line_iterator's output then False will be given -
1233
even though it may actually refer to that key.
1235
if not self._serializer.support_altered_by_hack:
1236
raise AssertionError(
1237
"_find_text_key_references_from_xml_inventory_lines only "
1238
"supported for branches which store inventory as unnested xml"
1239
", not on %r" % self)
1242
# this code needs to read every new line in every inventory for the
1243
# inventories [revision_ids]. Seeing a line twice is ok. Seeing a line
1244
# not present in one of those inventories is unnecessary but not
1245
# harmful because we are filtering by the revision id marker in the
1246
# inventory lines : we only select file ids altered in one of those
1247
# revisions. We don't need to see all lines in the inventory because
1248
# only those added in an inventory in rev X can contain a revision=X
1250
unescape_revid_cache = {}
1251
unescape_fileid_cache = {}
1253
# jam 20061218 In a big fetch, this handles hundreds of thousands
1254
# of lines, so it has had a lot of inlining and optimizing done.
1255
# Sorry that it is a little bit messy.
1256
# Move several functions to be local variables, since this is a long
1258
search = self._file_ids_altered_regex.search
1259
unescape = _unescape_xml
1260
setdefault = result.setdefault
1261
for line, line_key in line_iterator:
1262
match = search(line)
1265
# One call to match.group() returning multiple items is quite a
1266
# bit faster than 2 calls to match.group() each returning 1
1267
file_id, revision_id = match.group('file_id', 'revision_id')
1269
# Inlining the cache lookups helps a lot when you make 170,000
1270
# lines and 350k ids, versus 8.4 unique ids.
1271
# Using a cache helps in 2 ways:
1272
# 1) Avoids unnecessary decoding calls
1273
# 2) Re-uses cached strings, which helps in future set and
1275
# (2) is enough that removing encoding entirely along with
1276
# the cache (so we are using plain strings) results in no
1277
# performance improvement.
1279
revision_id = unescape_revid_cache[revision_id]
1281
unescaped = unescape(revision_id)
1282
unescape_revid_cache[revision_id] = unescaped
1283
revision_id = unescaped
1285
# Note that unconditionally unescaping means that we deserialise
1286
# every fileid, which for general 'pull' is not great, but we don't
1287
# really want to have some many fulltexts that this matters anyway.
1290
file_id = unescape_fileid_cache[file_id]
1292
unescaped = unescape(file_id)
1293
unescape_fileid_cache[file_id] = unescaped
1296
key = (file_id, revision_id)
1297
setdefault(key, False)
1298
if revision_id == line_key[-1]:
1302
def _find_file_ids_from_xml_inventory_lines(self, line_iterator,
1304
"""Helper routine for fileids_altered_by_revision_ids.
1306
This performs the translation of xml lines to revision ids.
1308
:param line_iterator: An iterator of lines, origin_version_id
1309
:param revision_ids: The revision ids to filter for. This should be a
1310
set or other type which supports efficient __contains__ lookups, as
1311
the revision id from each parsed line will be looked up in the
1312
revision_ids filter.
1313
:return: a dictionary mapping altered file-ids to an iterable of
1314
revision_ids. Each altered file-ids has the exact revision_ids that
1315
altered it listed explicitly.
1318
setdefault = result.setdefault
1320
self._find_text_key_references_from_xml_inventory_lines(
1321
line_iterator).iterkeys():
1322
# once data is all ensured-consistent; then this is
1323
# if revision_id == version_id
1324
if key[-1:] in revision_ids:
1325
setdefault(key[0], set()).add(key[-1])
1328
def fileids_altered_by_revision_ids(self, revision_ids, _inv_weave=None):
1329
"""Find the file ids and versions affected by revisions.
1331
:param revisions: an iterable containing revision ids.
1332
:param _inv_weave: The inventory weave from this repository or None.
1333
If None, the inventory weave will be opened automatically.
1334
:return: a dictionary mapping altered file-ids to an iterable of
1335
revision_ids. Each altered file-ids has the exact revision_ids that
1336
altered it listed explicitly.
1338
selected_keys = set((revid,) for revid in revision_ids)
1339
w = _inv_weave or self.inventories
1340
pb = ui.ui_factory.nested_progress_bar()
1342
return self._find_file_ids_from_xml_inventory_lines(
1343
w.iter_lines_added_or_present_in_keys(
1344
selected_keys, pb=pb),
913
"""Store a signature text for a revision.
915
:param revision_id: Revision id of the revision
916
:param signature: Signature text.
918
raise NotImplementedError(self.add_signature_text)
920
def _find_parent_ids_of_revisions(self, revision_ids):
921
"""Find all parent ids that are mentioned in the revision graph.
923
:return: set of revisions that are parents of revision_ids which are
924
not part of revision_ids themselves
926
parent_map = self.get_parent_map(revision_ids)
928
map(parent_ids.update, parent_map.itervalues())
929
parent_ids.difference_update(revision_ids)
930
parent_ids.discard(_mod_revision.NULL_REVISION)
1349
933
def iter_files_bytes(self, desired_files):
1350
934
"""Iterate through file versions.
1357
941
uniquely identify the file version in the caller's context. (Examples:
1358
942
an index number or a TreeTransform trans_id.)
1360
bytes_iterator is an iterable of bytestrings for the file. The
1361
kind of iterable and length of the bytestrings are unspecified, but for
1362
this implementation, it is a list of bytes produced by
1363
VersionedFile.get_record_stream().
1365
944
:param desired_files: a list of (file_id, revision_id, identifier)
1368
transaction = self.get_transaction()
1370
for file_id, revision_id, callable_data in desired_files:
1371
text_keys[(file_id, revision_id)] = callable_data
1372
for record in self.texts.get_record_stream(text_keys, 'unordered', True):
1373
if record.storage_kind == 'absent':
1374
raise errors.RevisionNotPresent(record.key, self)
1375
yield text_keys[record.key], record.get_bytes_as('fulltext')
1377
def _generate_text_key_index(self, text_key_references=None,
1379
"""Generate a new text key index for the repository.
1381
This is an expensive function that will take considerable time to run.
1383
:return: A dict mapping text keys ((file_id, revision_id) tuples) to a
1384
list of parents, also text keys. When a given key has no parents,
1385
the parents list will be [NULL_REVISION].
1387
# All revisions, to find inventory parents.
1388
if ancestors is None:
1389
graph = self.get_graph()
1390
ancestors = graph.get_parent_map(self.all_revision_ids())
1391
if text_key_references is None:
1392
text_key_references = self.find_text_key_references()
1393
pb = ui.ui_factory.nested_progress_bar()
1395
return self._do_generate_text_key_index(ancestors,
1396
text_key_references, pb)
1400
def _do_generate_text_key_index(self, ancestors, text_key_references, pb):
1401
"""Helper for _generate_text_key_index to avoid deep nesting."""
1402
revision_order = tsort.topo_sort(ancestors)
1403
invalid_keys = set()
1405
for revision_id in revision_order:
1406
revision_keys[revision_id] = set()
1407
text_count = len(text_key_references)
1408
# a cache of the text keys to allow reuse; costs a dict of all the
1409
# keys, but saves a 2-tuple for every child of a given key.
1411
for text_key, valid in text_key_references.iteritems():
1413
invalid_keys.add(text_key)
1415
revision_keys[text_key[1]].add(text_key)
1416
text_key_cache[text_key] = text_key
1417
del text_key_references
1419
text_graph = graph.Graph(graph.DictParentsProvider(text_index))
1420
NULL_REVISION = _mod_revision.NULL_REVISION
1421
# Set a cache with a size of 10 - this suffices for bzr.dev but may be
1422
# too small for large or very branchy trees. However, for 55K path
1423
# trees, it would be easy to use too much memory trivially. Ideally we
1424
# could gauge this by looking at available real memory etc, but this is
1425
# always a tricky proposition.
1426
inventory_cache = lru_cache.LRUCache(10)
1427
batch_size = 10 # should be ~150MB on a 55K path tree
1428
batch_count = len(revision_order) / batch_size + 1
1430
pb.update("Calculating text parents.", processed_texts, text_count)
1431
for offset in xrange(batch_count):
1432
to_query = revision_order[offset * batch_size:(offset + 1) *
1436
for rev_tree in self.revision_trees(to_query):
1437
revision_id = rev_tree.get_revision_id()
1438
parent_ids = ancestors[revision_id]
1439
for text_key in revision_keys[revision_id]:
1440
pb.update("Calculating text parents.", processed_texts)
1441
processed_texts += 1
1442
candidate_parents = []
1443
for parent_id in parent_ids:
1444
parent_text_key = (text_key[0], parent_id)
1446
check_parent = parent_text_key not in \
1447
revision_keys[parent_id]
1449
# the parent parent_id is a ghost:
1450
check_parent = False
1451
# truncate the derived graph against this ghost.
1452
parent_text_key = None
1454
# look at the parent commit details inventories to
1455
# determine possible candidates in the per file graph.
1458
inv = inventory_cache[parent_id]
1460
inv = self.revision_tree(parent_id).inventory
1461
inventory_cache[parent_id] = inv
1462
parent_entry = inv._byid.get(text_key[0], None)
1463
if parent_entry is not None:
1465
text_key[0], parent_entry.revision)
1467
parent_text_key = None
1468
if parent_text_key is not None:
1469
candidate_parents.append(
1470
text_key_cache[parent_text_key])
1471
parent_heads = text_graph.heads(candidate_parents)
1472
new_parents = list(parent_heads)
1473
new_parents.sort(key=lambda x:candidate_parents.index(x))
1474
if new_parents == []:
1475
new_parents = [NULL_REVISION]
1476
text_index[text_key] = new_parents
1478
for text_key in invalid_keys:
1479
text_index[text_key] = [NULL_REVISION]
1482
def item_keys_introduced_by(self, revision_ids, _files_pb=None):
1483
"""Get an iterable listing the keys of all the data introduced by a set
1486
The keys will be ordered so that the corresponding items can be safely
1487
fetched and inserted in that order.
1489
:returns: An iterable producing tuples of (knit-kind, file-id,
1490
versions). knit-kind is one of 'file', 'inventory', 'signatures',
1491
'revisions'. file-id is None unless knit-kind is 'file'.
1493
# XXX: it's a bit weird to control the inventory weave caching in this
1494
# generator. Ideally the caching would be done in fetch.py I think. Or
1495
# maybe this generator should explicitly have the contract that it
1496
# should not be iterated until the previously yielded item has been
1498
inv_w = self.inventories
1500
# file ids that changed
1501
file_ids = self.fileids_altered_by_revision_ids(revision_ids, inv_w)
1503
num_file_ids = len(file_ids)
1504
for file_id, altered_versions in file_ids.iteritems():
1505
if _files_pb is not None:
1506
_files_pb.update("fetch texts", count, num_file_ids)
1508
yield ("file", file_id, altered_versions)
1509
# We're done with the files_pb. Note that it finished by the caller,
1510
# just as it was created by the caller.
1514
yield ("inventory", None, revision_ids)
1517
revisions_with_signatures = set()
1518
for rev_id in revision_ids:
1520
self.get_signature_text(rev_id)
1521
except errors.NoSuchRevision:
1525
revisions_with_signatures.add(rev_id)
1526
yield ("signatures", None, revisions_with_signatures)
1529
yield ("revisions", None, revision_ids)
1532
def get_inventory(self, revision_id):
1533
"""Get Inventory object by revision id."""
1534
return self.iter_inventories([revision_id]).next()
1536
def iter_inventories(self, revision_ids):
1537
"""Get many inventories by revision_ids.
1539
This will buffer some or all of the texts used in constructing the
1540
inventories in memory, but will only parse a single inventory at a
1543
:return: An iterator of inventories.
1545
if ((None in revision_ids)
1546
or (_mod_revision.NULL_REVISION in revision_ids)):
1547
raise ValueError('cannot get null revision inventory')
1548
return self._iter_inventories(revision_ids)
1550
def _iter_inventories(self, revision_ids):
1551
"""single-document based inventory iteration."""
1552
for text, revision_id in self._iter_inventory_xmls(revision_ids):
1553
yield self.deserialise_inventory(revision_id, text)
1555
def _iter_inventory_xmls(self, revision_ids):
1556
keys = [(revision_id,) for revision_id in revision_ids]
1557
stream = self.inventories.get_record_stream(keys, 'unordered', True)
1559
for record in stream:
1560
if record.storage_kind != 'absent':
1561
texts[record.key] = record.get_bytes_as('fulltext')
1563
raise errors.NoSuchRevision(self, record.key)
1565
yield texts[key], key[-1]
1567
def deserialise_inventory(self, revision_id, xml):
1568
"""Transform the xml into an inventory object.
1570
:param revision_id: The expected revision id of the inventory.
1571
:param xml: A serialised inventory.
1573
result = self._serializer.read_inventory_from_string(xml, revision_id)
1574
if result.revision_id != revision_id:
1575
raise AssertionError('revision id mismatch %s != %s' % (
1576
result.revision_id, revision_id))
1579
def serialise_inventory(self, inv):
1580
return self._serializer.write_inventory_to_string(inv)
1582
def _serialise_inventory_to_lines(self, inv):
1583
return self._serializer.write_inventory_to_lines(inv)
1585
def get_serializer_format(self):
1586
return self._serializer.format_num
1589
def get_inventory_xml(self, revision_id):
1590
"""Get inventory XML as a file object."""
1591
texts = self._iter_inventory_xmls([revision_id])
1593
text, revision_id = texts.next()
1594
except StopIteration:
1595
raise errors.HistoryMissing(self, 'inventory', revision_id)
1599
def get_inventory_sha1(self, revision_id):
1600
"""Return the sha1 hash of the inventory entry
1602
return self.get_revision(revision_id).inventory_sha1
1604
def iter_reverse_revision_history(self, revision_id):
1605
"""Iterate backwards through revision ids in the lefthand history
1607
:param revision_id: The revision id to start with. All its lefthand
1608
ancestors will be traversed.
1610
graph = self.get_graph()
1611
next_id = revision_id
1613
if next_id in (None, _mod_revision.NULL_REVISION):
1616
# Note: The following line may raise KeyError in the event of
1617
# truncated history. We decided not to have a try:except:raise
1618
# RevisionNotPresent here until we see a use for it, because of the
1619
# cost in an inner loop that is by its very nature O(history).
1620
# Robert Collins 20080326
1621
parents = graph.get_parent_map([next_id])[next_id]
1622
if len(parents) == 0:
1625
next_id = parents[0]
1628
def get_revision_inventory(self, revision_id):
1629
"""Return inventory of a past revision."""
1630
# TODO: Unify this with get_inventory()
1631
# bzr 0.0.6 and later imposes the constraint that the inventory_id
1632
# must be the same as its revision, so this is trivial.
1633
if revision_id is None:
1634
# This does not make sense: if there is no revision,
1635
# then it is the current tree inventory surely ?!
1636
# and thus get_root_id() is something that looks at the last
1637
# commit on the branch, and the get_root_id is an inventory check.
1638
raise NotImplementedError
1639
# return Inventory(self.get_root_id())
1641
return self.get_inventory(revision_id)
947
raise NotImplementedError(self.iter_files_bytes)
949
def get_rev_id_for_revno(self, revno, known_pair):
950
"""Return the revision id of a revno, given a later (revno, revid)
951
pair in the same history.
953
:return: if found (True, revid). If the available history ran out
954
before reaching the revno, then this returns
955
(False, (closest_revno, closest_revid)).
957
known_revno, known_revid = known_pair
958
partial_history = [known_revid]
959
distance_from_known = known_revno - revno
960
if distance_from_known < 0:
962
'requested revno (%d) is later than given known revno (%d)'
963
% (revno, known_revno))
966
self, partial_history, stop_index=distance_from_known)
967
except errors.RevisionNotPresent, err:
968
if err.revision_id == known_revid:
969
# The start revision (known_revid) wasn't found.
971
# This is a stacked repository with no fallbacks, or a there's a
972
# left-hand ghost. Either way, even though the revision named in
973
# the error isn't in this repo, we know it's the next step in this
975
partial_history.append(err.revision_id)
976
if len(partial_history) <= distance_from_known:
977
# Didn't find enough history to get a revid for the revno.
978
earliest_revno = known_revno - len(partial_history) + 1
979
return (False, (earliest_revno, partial_history[-1]))
980
if len(partial_history) - 1 > distance_from_known:
981
raise AssertionError('_iter_for_revno returned too much history')
982
return (True, partial_history[-1])
1643
984
def is_shared(self):
1644
985
"""Return True if this repository is flagged as a shared repository."""
2353
1629
InterRepository.get(other).method_name(parameters).
2356
_walk_to_common_revisions_batch_size = 1
2357
1632
_optimisers = []
2358
1633
"""The available optimised InterRepository types."""
2360
def __init__(self, source, target):
2361
InterObject.__init__(self, source, target)
2362
# These two attributes may be overridden by e.g. InterOtherToRemote to
2363
# provide a faster implementation.
2364
self.target_get_graph = self.target.get_graph
2365
self.target_get_parent_map = self.target.get_parent_map
2367
1636
def copy_content(self, revision_id=None):
2368
raise NotImplementedError(self.copy_content)
2370
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
1637
"""Make a complete copy of the content in self into destination.
1639
This is a destructive operation! Do not use it on existing
1642
:param revision_id: Only copy the content needed to construct
1643
revision_id and its parents.
1646
self.target.set_make_working_trees(self.source.make_working_trees())
1647
except NotImplementedError:
1649
self.target.fetch(self.source, revision_id=revision_id)
1652
def fetch(self, revision_id=None, find_ghosts=False):
2371
1653
"""Fetch the content required to construct revision_id.
2373
1655
The content is copied from self.source to self.target.
2375
1657
:param revision_id: if None all content is copied, if NULL_REVISION no
2376
1658
content is copied.
2377
:param pb: optional progress bar to use for progress reports. If not
2378
provided a default one will be created.
2380
:returns: (copied_revision_count, failures).
2382
# Normally we should find a specific InterRepository subclass to do
2383
# the fetch; if nothing else then at least InterSameDataRepository.
2384
# If none of them is suitable it looks like fetching is not possible;
2385
# we try to give a good message why. _assert_same_model will probably
2386
# give a helpful message; otherwise a generic one.
2387
self._assert_same_model(self.source, self.target)
2388
raise errors.IncompatibleRepositories(self.source, self.target,
2389
"no suitableInterRepository found")
2391
def _walk_to_common_revisions(self, revision_ids):
2392
"""Walk out from revision_ids in source to revisions target has.
2394
:param revision_ids: The start point for the search.
2395
:return: A set of revision ids.
2397
target_graph = self.target_get_graph()
2398
revision_ids = frozenset(revision_ids)
2399
# Fast path for the case where all the revisions are already in the
2401
# (Although this does incur an extra round trip for the
2402
# fairly common case where the target doesn't already have the revision
2404
if set(target_graph.get_parent_map(revision_ids)) == revision_ids:
2405
return graph.SearchResult(revision_ids, set(), 0, set())
2406
missing_revs = set()
2407
source_graph = self.source.get_graph()
2408
# ensure we don't pay silly lookup costs.
2409
searcher = source_graph._make_breadth_first_searcher(revision_ids)
2410
null_set = frozenset([_mod_revision.NULL_REVISION])
2411
searcher_exhausted = False
2415
# Iterate the searcher until we have enough next_revs
2416
while len(next_revs) < self._walk_to_common_revisions_batch_size:
2418
next_revs_part, ghosts_part = searcher.next_with_ghosts()
2419
next_revs.update(next_revs_part)
2420
ghosts.update(ghosts_part)
2421
except StopIteration:
2422
searcher_exhausted = True
2424
# If there are ghosts in the source graph, and the caller asked for
2425
# them, make sure that they are present in the target.
2426
# We don't care about other ghosts as we can't fetch them and
2427
# haven't been asked to.
2428
ghosts_to_check = set(revision_ids.intersection(ghosts))
2429
revs_to_get = set(next_revs).union(ghosts_to_check)
2431
have_revs = set(target_graph.get_parent_map(revs_to_get))
2432
# we always have NULL_REVISION present.
2433
have_revs = have_revs.union(null_set)
2434
# Check if the target is missing any ghosts we need.
2435
ghosts_to_check.difference_update(have_revs)
2437
# One of the caller's revision_ids is a ghost in both the
2438
# source and the target.
2439
raise errors.NoSuchRevision(
2440
self.source, ghosts_to_check.pop())
2441
missing_revs.update(next_revs - have_revs)
2442
# Because we may have walked past the original stop point, make
2443
# sure everything is stopped
2444
stop_revs = searcher.find_seen_ancestors(have_revs)
2445
searcher.stop_searching_any(stop_revs)
2446
if searcher_exhausted:
2448
return searcher.get_result()
2450
@deprecated_method(one_two)
2452
def missing_revision_ids(self, revision_id=None, find_ghosts=True):
2453
"""Return the revision ids that source has that target does not.
2455
These are returned in topological order.
2457
:param revision_id: only return revision ids included by this
2459
:param find_ghosts: If True find missing revisions in deep history
2460
rather than just finding the surface difference.
2462
return list(self.search_missing_revision_ids(
2463
revision_id, find_ghosts).get_keys())
2466
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2467
"""Return the revision ids that source has that target does not.
2469
:param revision_id: only return revision ids included by this
2471
:param find_ghosts: If True find missing revisions in deep history
2472
rather than just finding the surface difference.
1661
raise NotImplementedError(self.fetch)
1664
def search_missing_revision_ids(self,
1665
revision_id=symbol_versioning.DEPRECATED_PARAMETER,
1666
find_ghosts=True, revision_ids=None, if_present_ids=None,
1668
"""Return the revision ids that source has that target does not.
1670
:param revision_id: only return revision ids included by this
1672
:param revision_ids: return revision ids included by these
1673
revision_ids. NoSuchRevision will be raised if any of these
1674
revisions are not present.
1675
:param if_present_ids: like revision_ids, but will not cause
1676
NoSuchRevision if any of these are absent, instead they will simply
1677
not be in the result. This is useful for e.g. finding revisions
1678
to fetch for tags, which may reference absent revisions.
1679
:param find_ghosts: If True find missing revisions in deep history
1680
rather than just finding the surface difference.
1681
:param limit: Maximum number of revisions to return, topologically
2473
1683
:return: A bzrlib.graph.SearchResult.
2475
# stop searching at found target revisions.
2476
if not find_ghosts and revision_id is not None:
2477
return self._walk_to_common_revisions([revision_id])
2478
# generic, possibly worst case, slow code path.
2479
target_ids = set(self.target.all_revision_ids())
2480
if revision_id is not None:
2481
source_ids = self.source.get_ancestry(revision_id)
2482
if source_ids[0] is not None:
2483
raise AssertionError()
2486
source_ids = self.source.all_revision_ids()
2487
result_set = set(source_ids).difference(target_ids)
2488
return self.source.revision_ids_to_search_result(result_set)
1685
raise NotImplementedError(self.search_missing_revision_ids)
2491
1688
def _same_model(source, target):
2492
1689
"""True if source and target have the same data representation.
2494
1691
Note: this is always called on the base class; overriding it in a
2495
1692
subclass will have no effect.
2512
1709
"different serializers")
2515
class InterSameDataRepository(InterRepository):
2516
"""Code for converting between repositories that represent the same data.
2518
Data format and model must match for this to work.
2522
def _get_repo_format_to_test(self):
2523
"""Repository format for testing with.
2525
InterSameData can pull from subtree to subtree and from non-subtree to
2526
non-subtree, so we test this with the richest repository format.
2528
from bzrlib.repofmt import knitrepo
2529
return knitrepo.RepositoryFormatKnit3()
2532
def is_compatible(source, target):
2533
return InterRepository._same_model(source, target)
2536
def copy_content(self, revision_id=None):
2537
"""Make a complete copy of the content in self into destination.
2539
This copies both the repository's revision data, and configuration information
2540
such as the make_working_trees setting.
2542
This is a destructive operation! Do not use it on existing
2545
:param revision_id: Only copy the content needed to construct
2546
revision_id and its parents.
2549
self.target.set_make_working_trees(self.source.make_working_trees())
2550
except NotImplementedError:
2552
# but don't bother fetching if we have the needed data now.
2553
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
2554
self.target.has_revision(revision_id)):
2556
self.target.fetch(self.source, revision_id=revision_id)
2559
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2560
"""See InterRepository.fetch()."""
2561
from bzrlib.fetch import RepoFetcher
2562
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2563
self.source, self.source._format, self.target,
2564
self.target._format)
2565
f = RepoFetcher(to_repository=self.target,
2566
from_repository=self.source,
2567
last_revision=revision_id,
2568
pb=pb, find_ghosts=find_ghosts)
2569
return f.count_copied, f.failed_revisions
2572
class InterWeaveRepo(InterSameDataRepository):
2573
"""Optimised code paths between Weave based repositories.
2575
This should be in bzrlib/repofmt/weaverepo.py but we have not yet
2576
implemented lazy inter-object optimisation.
2580
def _get_repo_format_to_test(self):
2581
from bzrlib.repofmt import weaverepo
2582
return weaverepo.RepositoryFormat7()
2585
def is_compatible(source, target):
2586
"""Be compatible with known Weave formats.
2588
We don't test for the stores being of specific types because that
2589
could lead to confusing results, and there is no need to be
2592
from bzrlib.repofmt.weaverepo import (
2598
return (isinstance(source._format, (RepositoryFormat5,
2600
RepositoryFormat7)) and
2601
isinstance(target._format, (RepositoryFormat5,
2603
RepositoryFormat7)))
2604
except AttributeError:
2608
def copy_content(self, revision_id=None):
2609
"""See InterRepository.copy_content()."""
2610
# weave specific optimised path:
2612
self.target.set_make_working_trees(self.source.make_working_trees())
2613
except (errors.RepositoryUpgradeRequired, NotImplemented):
2615
# FIXME do not peek!
2616
if self.source._transport.listable():
2617
pb = ui.ui_factory.nested_progress_bar()
2619
self.target.texts.insert_record_stream(
2620
self.source.texts.get_record_stream(
2621
self.source.texts.keys(), 'topological', False))
2622
pb.update('copying inventory', 0, 1)
2623
self.target.inventories.insert_record_stream(
2624
self.source.inventories.get_record_stream(
2625
self.source.inventories.keys(), 'topological', False))
2626
self.target.signatures.insert_record_stream(
2627
self.source.signatures.get_record_stream(
2628
self.source.signatures.keys(),
2630
self.target.revisions.insert_record_stream(
2631
self.source.revisions.get_record_stream(
2632
self.source.revisions.keys(),
2633
'topological', True))
2637
self.target.fetch(self.source, revision_id=revision_id)
2640
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2641
"""See InterRepository.fetch()."""
2642
from bzrlib.fetch import RepoFetcher
2643
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2644
self.source, self.source._format, self.target, self.target._format)
2645
f = RepoFetcher(to_repository=self.target,
2646
from_repository=self.source,
2647
last_revision=revision_id,
2648
pb=pb, find_ghosts=find_ghosts)
2649
return f.count_copied, f.failed_revisions
2652
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2653
"""See InterRepository.missing_revision_ids()."""
2654
# we want all revisions to satisfy revision_id in source.
2655
# but we don't want to stat every file here and there.
2656
# we want then, all revisions other needs to satisfy revision_id
2657
# checked, but not those that we have locally.
2658
# so the first thing is to get a subset of the revisions to
2659
# satisfy revision_id in source, and then eliminate those that
2660
# we do already have.
2661
# this is slow on high latency connection to self, but as as this
2662
# disk format scales terribly for push anyway due to rewriting
2663
# inventory.weave, this is considered acceptable.
2665
if revision_id is not None:
2666
source_ids = self.source.get_ancestry(revision_id)
2667
if source_ids[0] is not None:
2668
raise AssertionError()
2671
source_ids = self.source._all_possible_ids()
2672
source_ids_set = set(source_ids)
2673
# source_ids is the worst possible case we may need to pull.
2674
# now we want to filter source_ids against what we actually
2675
# have in target, but don't try to check for existence where we know
2676
# we do not have a revision as that would be pointless.
2677
target_ids = set(self.target._all_possible_ids())
2678
possibly_present_revisions = target_ids.intersection(source_ids_set)
2679
actually_present_revisions = set(
2680
self.target._eliminate_revisions_not_present(possibly_present_revisions))
2681
required_revisions = source_ids_set.difference(actually_present_revisions)
2682
if revision_id is not None:
2683
# we used get_ancestry to determine source_ids then we are assured all
2684
# revisions referenced are present as they are installed in topological order.
2685
# and the tip revision was validated by get_ancestry.
2686
result_set = required_revisions
2688
# if we just grabbed the possibly available ids, then
2689
# we only have an estimate of whats available and need to validate
2690
# that against the revision records.
2692
self.source._eliminate_revisions_not_present(required_revisions))
2693
return self.source.revision_ids_to_search_result(result_set)
2696
class InterKnitRepo(InterSameDataRepository):
2697
"""Optimised code paths between Knit based repositories."""
2700
def _get_repo_format_to_test(self):
2701
from bzrlib.repofmt import knitrepo
2702
return knitrepo.RepositoryFormatKnit1()
2705
def is_compatible(source, target):
2706
"""Be compatible with known Knit formats.
2708
We don't test for the stores being of specific types because that
2709
could lead to confusing results, and there is no need to be
2712
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit
2714
are_knits = (isinstance(source._format, RepositoryFormatKnit) and
2715
isinstance(target._format, RepositoryFormatKnit))
2716
except AttributeError:
2718
return are_knits and InterRepository._same_model(source, target)
2721
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2722
"""See InterRepository.fetch()."""
2723
from bzrlib.fetch import RepoFetcher
2724
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2725
self.source, self.source._format, self.target, self.target._format)
2726
f = RepoFetcher(to_repository=self.target,
2727
from_repository=self.source,
2728
last_revision=revision_id,
2729
pb=pb, find_ghosts=find_ghosts)
2730
return f.count_copied, f.failed_revisions
2733
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2734
"""See InterRepository.missing_revision_ids()."""
2735
if revision_id is not None:
2736
source_ids = self.source.get_ancestry(revision_id)
2737
if source_ids[0] is not None:
2738
raise AssertionError()
2741
source_ids = self.source.all_revision_ids()
2742
source_ids_set = set(source_ids)
2743
# source_ids is the worst possible case we may need to pull.
2744
# now we want to filter source_ids against what we actually
2745
# have in target, but don't try to check for existence where we know
2746
# we do not have a revision as that would be pointless.
2747
target_ids = set(self.target.all_revision_ids())
2748
possibly_present_revisions = target_ids.intersection(source_ids_set)
2749
actually_present_revisions = set(
2750
self.target._eliminate_revisions_not_present(possibly_present_revisions))
2751
required_revisions = source_ids_set.difference(actually_present_revisions)
2752
if revision_id is not None:
2753
# we used get_ancestry to determine source_ids then we are assured all
2754
# revisions referenced are present as they are installed in topological order.
2755
# and the tip revision was validated by get_ancestry.
2756
result_set = required_revisions
2758
# if we just grabbed the possibly available ids, then
2759
# we only have an estimate of whats available and need to validate
2760
# that against the revision records.
2762
self.source._eliminate_revisions_not_present(required_revisions))
2763
return self.source.revision_ids_to_search_result(result_set)
2766
class InterPackRepo(InterSameDataRepository):
2767
"""Optimised code paths between Pack based repositories."""
2770
def _get_repo_format_to_test(self):
2771
from bzrlib.repofmt import pack_repo
2772
return pack_repo.RepositoryFormatKnitPack1()
2775
def is_compatible(source, target):
2776
"""Be compatible with known Pack formats.
2778
We don't test for the stores being of specific types because that
2779
could lead to confusing results, and there is no need to be
2782
from bzrlib.repofmt.pack_repo import RepositoryFormatPack
2784
are_packs = (isinstance(source._format, RepositoryFormatPack) and
2785
isinstance(target._format, RepositoryFormatPack))
2786
except AttributeError:
2788
return are_packs and InterRepository._same_model(source, target)
2791
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2792
"""See InterRepository.fetch()."""
2793
if (len(self.source._fallback_repositories) > 0 or
2794
len(self.target._fallback_repositories) > 0):
2795
# The pack layer is not aware of fallback repositories, so when
2796
# fetching from a stacked repository or into a stacked repository
2797
# we use the generic fetch logic which uses the VersionedFiles
2798
# attributes on repository.
2799
from bzrlib.fetch import RepoFetcher
2800
fetcher = RepoFetcher(self.target, self.source, revision_id,
2802
return fetcher.count_copied, fetcher.failed_revisions
2803
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2804
self.source, self.source._format, self.target, self.target._format)
2805
self.count_copied = 0
2806
if revision_id is None:
2808
# everything to do - use pack logic
2809
# to fetch from all packs to one without
2810
# inventory parsing etc, IFF nothing to be copied is in the target.
2812
source_revision_ids = frozenset(self.source.all_revision_ids())
2813
revision_ids = source_revision_ids - \
2814
frozenset(self.target_get_parent_map(source_revision_ids))
2815
revision_keys = [(revid,) for revid in revision_ids]
2816
target_pack_collection = self._get_target_pack_collection()
2817
index = target_pack_collection.revision_index.combined_index
2818
present_revision_ids = set(item[1][0] for item in
2819
index.iter_entries(revision_keys))
2820
revision_ids = set(revision_ids) - present_revision_ids
2821
# implementing the TODO will involve:
2822
# - detecting when all of a pack is selected
2823
# - avoiding as much as possible pre-selection, so the
2824
# more-core routines such as create_pack_from_packs can filter in
2825
# a just-in-time fashion. (though having a HEADS list on a
2826
# repository might make this a lot easier, because we could
2827
# sensibly detect 'new revisions' without doing a full index scan.
2828
elif _mod_revision.is_null(revision_id):
2833
revision_ids = self.search_missing_revision_ids(revision_id,
2834
find_ghosts=find_ghosts).get_keys()
2835
except errors.NoSuchRevision:
2836
raise errors.InstallFailed([revision_id])
2837
if len(revision_ids) == 0:
2839
return self._pack(self.source, self.target, revision_ids)
2841
def _pack(self, source, target, revision_ids):
2842
from bzrlib.repofmt.pack_repo import Packer
2843
target_pack_collection = self._get_target_pack_collection()
2844
packs = source._pack_collection.all_packs()
2845
pack = Packer(target_pack_collection, packs, '.fetch',
2846
revision_ids).pack()
2847
if pack is not None:
2848
target_pack_collection._save_pack_names()
2849
copied_revs = pack.get_revision_count()
2850
# Trigger an autopack. This may duplicate effort as we've just done
2851
# a pack creation, but for now it is simpler to think about as
2852
# 'upload data, then repack if needed'.
2854
return (copied_revs, [])
2858
def _autopack(self):
2859
self.target._pack_collection.autopack()
2861
def _get_target_pack_collection(self):
2862
return self.target._pack_collection
2865
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2866
"""See InterRepository.missing_revision_ids().
2868
:param find_ghosts: Find ghosts throughout the ancestry of
2871
if not find_ghosts and revision_id is not None:
2872
return self._walk_to_common_revisions([revision_id])
2873
elif revision_id is not None:
2874
# Find ghosts: search for revisions pointing from one repository to
2875
# the other, and vice versa, anywhere in the history of revision_id.
2876
graph = self.target_get_graph(other_repository=self.source)
2877
searcher = graph._make_breadth_first_searcher([revision_id])
2881
next_revs, ghosts = searcher.next_with_ghosts()
2882
except StopIteration:
2884
if revision_id in ghosts:
2885
raise errors.NoSuchRevision(self.source, revision_id)
2886
found_ids.update(next_revs)
2887
found_ids.update(ghosts)
2888
found_ids = frozenset(found_ids)
2889
# Double query here: should be able to avoid this by changing the
2890
# graph api further.
2891
result_set = found_ids - frozenset(
2892
self.target_get_parent_map(found_ids))
2894
source_ids = self.source.all_revision_ids()
2895
# source_ids is the worst possible case we may need to pull.
2896
# now we want to filter source_ids against what we actually
2897
# have in target, but don't try to check for existence where we know
2898
# we do not have a revision as that would be pointless.
2899
target_ids = set(self.target.all_revision_ids())
2900
result_set = set(source_ids).difference(target_ids)
2901
return self.source.revision_ids_to_search_result(result_set)
2904
class InterModel1and2(InterRepository):
2907
def _get_repo_format_to_test(self):
2911
def is_compatible(source, target):
2912
if not source.supports_rich_root() and target.supports_rich_root():
2918
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2919
"""See InterRepository.fetch()."""
2920
from bzrlib.fetch import Model1toKnit2Fetcher
2921
f = Model1toKnit2Fetcher(to_repository=self.target,
2922
from_repository=self.source,
2923
last_revision=revision_id,
2924
pb=pb, find_ghosts=find_ghosts)
2925
return f.count_copied, f.failed_revisions
2928
def copy_content(self, revision_id=None):
2929
"""Make a complete copy of the content in self into destination.
2931
This is a destructive operation! Do not use it on existing
2934
:param revision_id: Only copy the content needed to construct
2935
revision_id and its parents.
2938
self.target.set_make_working_trees(self.source.make_working_trees())
2939
except NotImplementedError:
2941
# but don't bother fetching if we have the needed data now.
2942
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
2943
self.target.has_revision(revision_id)):
2945
self.target.fetch(self.source, revision_id=revision_id)
2948
class InterKnit1and2(InterKnitRepo):
2951
def _get_repo_format_to_test(self):
2955
def is_compatible(source, target):
2956
"""Be compatible with Knit1 source and Knit3 target"""
2958
from bzrlib.repofmt.knitrepo import (
2959
RepositoryFormatKnit1,
2960
RepositoryFormatKnit3,
2962
from bzrlib.repofmt.pack_repo import (
2963
RepositoryFormatKnitPack1,
2964
RepositoryFormatKnitPack3,
2965
RepositoryFormatKnitPack4,
2966
RepositoryFormatKnitPack5,
2967
RepositoryFormatKnitPack5RichRoot,
2968
RepositoryFormatPackDevelopment2,
2969
RepositoryFormatPackDevelopment2Subtree,
2972
RepositoryFormatKnit1, # no rr, no subtree
2973
RepositoryFormatKnitPack1, # no rr, no subtree
2974
RepositoryFormatPackDevelopment2, # no rr, no subtree
2975
RepositoryFormatKnitPack5, # no rr, no subtree
2978
RepositoryFormatKnit3, # rr, subtree
2979
RepositoryFormatKnitPack3, # rr, subtree
2980
RepositoryFormatKnitPack4, # rr, no subtree
2981
RepositoryFormatKnitPack5RichRoot,# rr, no subtree
2982
RepositoryFormatPackDevelopment2Subtree, # rr, subtree
2984
for format in norichroot:
2985
if format.rich_root_data:
2986
raise AssertionError('Format %s is a rich-root format'
2987
' but is included in the non-rich-root list'
2989
for format in richroot:
2990
if not format.rich_root_data:
2991
raise AssertionError('Format %s is not a rich-root format'
2992
' but is included in the rich-root list'
2994
# TODO: One alternative is to just check format.rich_root_data,
2995
# instead of keeping membership lists. However, the formats
2996
# *also* have to use the same 'Knit' style of storage
2997
# (line-deltas, fulltexts, etc.)
2998
return (isinstance(source._format, norichroot) and
2999
isinstance(target._format, richroot))
3000
except AttributeError:
3004
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3005
"""See InterRepository.fetch()."""
3006
from bzrlib.fetch import Knit1to2Fetcher
3007
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
3008
self.source, self.source._format, self.target,
3009
self.target._format)
3010
f = Knit1to2Fetcher(to_repository=self.target,
3011
from_repository=self.source,
3012
last_revision=revision_id,
3013
pb=pb, find_ghosts=find_ghosts)
3014
return f.count_copied, f.failed_revisions
3017
class InterDifferingSerializer(InterKnitRepo):
3020
def _get_repo_format_to_test(self):
3024
def is_compatible(source, target):
3025
"""Be compatible with Knit2 source and Knit3 target"""
3026
if source.supports_rich_root() != target.supports_rich_root():
3028
# Ideally, we'd support fetching if the source had no tree references
3029
# even if it supported them...
3030
if (getattr(source, '_format.supports_tree_reference', False) and
3031
not getattr(target, '_format.supports_tree_reference', False)):
3036
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3037
"""See InterRepository.fetch()."""
3038
revision_ids = self.target.search_missing_revision_ids(self.source,
3039
revision_id, find_ghosts=find_ghosts).get_keys()
3040
revision_ids = tsort.topo_sort(
3041
self.source.get_graph().get_parent_map(revision_ids))
3042
def revisions_iterator():
3043
for current_revision_id in revision_ids:
3044
revision = self.source.get_revision(current_revision_id)
3045
tree = self.source.revision_tree(current_revision_id)
3047
signature = self.source.get_signature_text(
3048
current_revision_id)
3049
except errors.NoSuchRevision:
3051
yield revision, tree, signature
3053
my_pb = ui.ui_factory.nested_progress_bar()
3058
install_revisions(self.target, revisions_iterator(),
3059
len(revision_ids), pb)
3061
if my_pb is not None:
3063
return len(revision_ids), 0
3066
class InterOtherToRemote(InterRepository):
3067
"""An InterRepository that simply delegates to the 'real' InterRepository
3068
calculated for (source, target._real_repository).
3071
_walk_to_common_revisions_batch_size = 50
3073
def __init__(self, source, target):
3074
InterRepository.__init__(self, source, target)
3075
self._real_inter = None
3078
def is_compatible(source, target):
3079
if isinstance(target, remote.RemoteRepository):
3083
def _ensure_real_inter(self):
3084
if self._real_inter is None:
3085
self.target._ensure_real()
3086
real_target = self.target._real_repository
3087
self._real_inter = InterRepository.get(self.source, real_target)
3088
# Make _real_inter use the RemoteRepository for get_parent_map
3089
self._real_inter.target_get_graph = self.target.get_graph
3090
self._real_inter.target_get_parent_map = self.target.get_parent_map
3092
def copy_content(self, revision_id=None):
3093
self._ensure_real_inter()
3094
self._real_inter.copy_content(revision_id=revision_id)
3096
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3097
self._ensure_real_inter()
3098
return self._real_inter.fetch(revision_id=revision_id, pb=pb,
3099
find_ghosts=find_ghosts)
3102
def _get_repo_format_to_test(self):
3106
class InterRemoteToOther(InterRepository):
3108
def __init__(self, source, target):
3109
InterRepository.__init__(self, source, target)
3110
self._real_inter = None
3113
def is_compatible(source, target):
3114
if not isinstance(source, remote.RemoteRepository):
3116
# Is source's model compatible with target's model?
3117
source._ensure_real()
3118
real_source = source._real_repository
3119
if isinstance(real_source, remote.RemoteRepository):
3120
raise NotImplementedError(
3121
"We don't support remote repos backed by remote repos yet.")
3122
return InterRepository._same_model(real_source, target)
3124
def _ensure_real_inter(self):
3125
if self._real_inter is None:
3126
self.source._ensure_real()
3127
real_source = self.source._real_repository
3128
self._real_inter = InterRepository.get(real_source, self.target)
3130
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3131
self._ensure_real_inter()
3132
return self._real_inter.fetch(revision_id=revision_id, pb=pb,
3133
find_ghosts=find_ghosts)
3135
def copy_content(self, revision_id=None):
3136
self._ensure_real_inter()
3137
self._real_inter.copy_content(revision_id=revision_id)
3140
def _get_repo_format_to_test(self):
3145
class InterPackToRemotePack(InterPackRepo):
3146
"""A specialisation of InterPackRepo for a target that is a
3149
This will use the get_parent_map RPC rather than plain readvs, and also
3150
uses an RPC for autopacking.
3153
_walk_to_common_revisions_batch_size = 50
3156
def is_compatible(source, target):
3157
from bzrlib.repofmt.pack_repo import RepositoryFormatPack
3158
if isinstance(source._format, RepositoryFormatPack):
3159
if isinstance(target, remote.RemoteRepository):
3160
target._ensure_real()
3161
if isinstance(target._real_repository._format,
3162
RepositoryFormatPack):
3163
if InterRepository._same_model(source, target):
3167
def _autopack(self):
3168
self.target.autopack()
3170
def _get_target_pack_collection(self):
3171
return self.target._real_repository._pack_collection
3174
def _get_repo_format_to_test(self):
3178
InterRepository.register_optimiser(InterDifferingSerializer)
3179
InterRepository.register_optimiser(InterSameDataRepository)
3180
InterRepository.register_optimiser(InterWeaveRepo)
3181
InterRepository.register_optimiser(InterKnitRepo)
3182
InterRepository.register_optimiser(InterModel1and2)
3183
InterRepository.register_optimiser(InterKnit1and2)
3184
InterRepository.register_optimiser(InterPackRepo)
3185
InterRepository.register_optimiser(InterOtherToRemote)
3186
InterRepository.register_optimiser(InterRemoteToOther)
3187
InterRepository.register_optimiser(InterPackToRemotePack)
3190
1712
class CopyConverter(object):
3191
1713
"""A repository conversion tool which just performs a copy of the content.
3193
1715
This is slow but quite reliable.