202
193
self.random_revid = False
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,
195
def _heads(self, file_id, revision_ids):
196
"""Calculate the graph heads for revision_ids in the graph of file_id.
198
This can use either a per-file graph or a global revision graph as we
199
have an identity relationship between the two graphs.
201
return self.__heads(revision_ids)
203
def _check_root(self, ie, parent_invs, tree):
204
"""Helper for record_entry_contents.
206
:param ie: An entry being added.
207
:param parent_invs: The inventories of the parent revisions of the
209
:param tree: The tree that is being committed.
211
# In this revision format, root entries have no knit or weave When
212
# serializing out to disk and back in root.revision is always
214
ie.revision = self._new_revision_id
216
def _get_delta(self, ie, basis_inv, path):
217
"""Get a delta against the basis inventory for ie."""
218
if ie.file_id not in basis_inv:
220
return (None, path, ie.file_id, ie)
221
elif ie != basis_inv[ie.file_id]:
223
# TODO: avoid tis id2path call.
224
return (basis_inv.id2path(ie.file_id), path, ie.file_id, ie)
229
def record_entry_contents(self, ie, parent_invs, path, tree,
231
"""Record the content of ie from tree into the commit if needed.
233
Side effect: sets ie.revision when unchanged
235
:param ie: An inventory entry present in the commit.
236
:param parent_invs: The inventories of the parent revisions of the
238
:param path: The path the entry is at in the tree.
239
:param tree: The tree which contains this entry and should be used to
241
:param content_summary: Summary data from the tree about the paths
242
content - stat, length, exec, sha/link target. This is only
243
accessed when the entry has a revision of None - that is when it is
244
a candidate to commit.
245
:return: A tuple (change_delta, version_recorded). change_delta is
246
an inventory_delta change for this entry against the basis tree of
247
the commit, or None if no change occured against the basis tree.
248
version_recorded is True if a new version of the entry has been
249
recorded. For instance, committing a merge where a file was only
250
changed on the other side will return (delta, False).
252
if self.new_inventory.root is None:
253
if ie.parent_id is not None:
254
raise errors.RootMissing()
255
self._check_root(ie, parent_invs, tree)
256
if ie.revision is None:
257
kind = content_summary[0]
259
# ie is carried over from a prior commit
261
# XXX: repository specific check for nested tree support goes here - if
262
# the repo doesn't want nested trees we skip it ?
263
if (kind == 'tree-reference' and
264
not self.repository._format.supports_tree_reference):
265
# mismatch between commit builder logic and repository:
266
# this needs the entry creation pushed down into the builder.
267
raise NotImplementedError('Missing repository subtree support.')
268
self.new_inventory.add(ie)
270
# TODO: slow, take it out of the inner loop.
272
basis_inv = parent_invs[0]
274
basis_inv = Inventory(root_id=None)
276
# ie.revision is always None if the InventoryEntry is considered
277
# for committing. We may record the previous parents revision if the
278
# content is actually unchanged against a sole head.
279
if ie.revision is not None:
280
if not self._versioned_root and path == '':
281
# repositories that do not version the root set the root's
282
# revision to the new commit even when no change occurs, and
283
# this masks when a change may have occurred against the basis,
284
# so calculate if one happened.
285
if ie.file_id in basis_inv:
286
delta = (basis_inv.id2path(ie.file_id), path,
290
delta = (None, path, ie.file_id, ie)
293
# we don't need to commit this, because the caller already
294
# determined that an existing revision of this file is
296
return None, (ie.revision == self._new_revision_id)
297
# XXX: Friction: parent_candidates should return a list not a dict
298
# so that we don't have to walk the inventories again.
299
parent_candiate_entries = ie.parent_candidates(parent_invs)
300
head_set = self._heads(ie.file_id, parent_candiate_entries.keys())
302
for inv in parent_invs:
303
if ie.file_id in inv:
304
old_rev = inv[ie.file_id].revision
305
if old_rev in head_set:
306
heads.append(inv[ie.file_id].revision)
307
head_set.remove(inv[ie.file_id].revision)
310
# now we check to see if we need to write a new record to the
312
# We write a new entry unless there is one head to the ancestors, and
313
# the kind-derived content is unchanged.
315
# Cheapest check first: no ancestors, or more the one head in the
316
# ancestors, we write a new node.
320
# There is a single head, look it up for comparison
321
parent_entry = parent_candiate_entries[heads[0]]
322
# if the non-content specific data has changed, we'll be writing a
324
if (parent_entry.parent_id != ie.parent_id or
325
parent_entry.name != ie.name):
327
# now we need to do content specific checks:
329
# if the kind changed the content obviously has
330
if kind != parent_entry.kind:
333
if content_summary[2] is None:
334
raise ValueError("Files must not have executable = None")
336
if (# if the file length changed we have to store:
337
parent_entry.text_size != content_summary[1] or
338
# if the exec bit has changed we have to store:
339
parent_entry.executable != content_summary[2]):
341
elif parent_entry.text_sha1 == content_summary[3]:
342
# all meta and content is unchanged (using a hash cache
343
# hit to check the sha)
344
ie.revision = parent_entry.revision
345
ie.text_size = parent_entry.text_size
346
ie.text_sha1 = parent_entry.text_sha1
347
ie.executable = parent_entry.executable
348
return self._get_delta(ie, basis_inv, path), False
350
# Either there is only a hash change(no hash cache entry,
351
# or same size content change), or there is no change on
353
# Provide the parent's hash to the store layer, so that the
354
# content is unchanged we will not store a new node.
355
nostore_sha = parent_entry.text_sha1
357
# We want to record a new node regardless of the presence or
358
# absence of a content change in the file.
360
ie.executable = content_summary[2]
361
lines = tree.get_file(ie.file_id, path).readlines()
363
ie.text_sha1, ie.text_size = self._add_text_to_weave(
364
ie.file_id, lines, heads, nostore_sha)
365
except errors.ExistingContent:
366
# Turns out that the file content was unchanged, and we were
367
# only going to store a new node if it was changed. Carry over
369
ie.revision = parent_entry.revision
370
ie.text_size = parent_entry.text_size
371
ie.text_sha1 = parent_entry.text_sha1
372
ie.executable = parent_entry.executable
373
return self._get_delta(ie, basis_inv, path), False
374
elif kind == 'directory':
376
# all data is meta here, nothing specific to directory, so
378
ie.revision = parent_entry.revision
379
return self._get_delta(ie, basis_inv, path), False
381
self._add_text_to_weave(ie.file_id, lines, heads, None)
382
elif kind == 'symlink':
383
current_link_target = content_summary[3]
385
# symlink target is not generic metadata, check if it has
387
if current_link_target != parent_entry.symlink_target:
390
# unchanged, carry over.
391
ie.revision = parent_entry.revision
392
ie.symlink_target = parent_entry.symlink_target
393
return self._get_delta(ie, basis_inv, path), False
394
ie.symlink_target = current_link_target
396
self._add_text_to_weave(ie.file_id, lines, heads, None)
397
elif kind == 'tree-reference':
399
if content_summary[3] != parent_entry.reference_revision:
402
# unchanged, carry over.
403
ie.reference_revision = parent_entry.reference_revision
404
ie.revision = parent_entry.revision
405
return self._get_delta(ie, basis_inv, path), False
406
ie.reference_revision = content_summary[3]
408
self._add_text_to_weave(ie.file_id, lines, heads, None)
410
raise NotImplementedError('unknown kind')
411
ie.revision = self._new_revision_id
412
return self._get_delta(ie, basis_inv, path), True
414
def _add_text_to_weave(self, file_id, new_lines, parents, nostore_sha):
415
# Note: as we read the content directly from the tree, we know its not
416
# been turned into unicode or badly split - but a broken tree
417
# implementation could give us bad output from readlines() so this is
418
# not a guarantee of safety. What would be better is always checking
419
# the content during test suite execution. RBC 20070912
420
parent_keys = tuple((file_id, parent) for parent in parents)
421
return self.repository.texts.add_lines(
422
(file_id, self._new_revision_id), parent_keys, new_lines,
423
nostore_sha=nostore_sha, random_id=self.random_revid,
424
check_content=False)[0:2]
427
class RootCommitBuilder(CommitBuilder):
428
"""This commitbuilder actually records the root id"""
430
# the root entry gets versioned properly by this builder.
431
_versioned_root = True
433
def _check_root(self, ie, parent_invs, tree):
434
"""Helper for record_entry_contents.
436
:param ie: An entry being added.
437
:param parent_invs: The inventories of the parent revisions of the
439
:param tree: The tree that is being committed.
247
443
######################################################################
251
class Repository(_RelockDebugMixin, controldir.ControlComponent):
446
class Repository(object):
252
447
"""Repository holding history for one or more branches.
254
449
The repository holds and retrieves historical information including
255
450
revisions and file history. It's normally accessed only by the Branch,
256
451
which views a particular line of development through that history.
258
See VersionedFileRepository in bzrlib.vf_repository for the
259
base class for most Bazaar repositories.
453
The Repository builds on top of some byte storage facilies (the revisions,
454
signatures, inventories and texts attributes) and a Transport, which
455
respectively provide byte storage and a means to access the (possibly
458
The byte storage facilities are addressed via tuples, which we refer to
459
as 'keys' throughout the code base. Revision_keys, inventory_keys and
460
signature_keys are all 1-tuples: (revision_id,). text_keys are two-tuples:
461
(file_id, revision_id). We use this interface because it allows low
462
friction with the underlying code that implements disk indices, network
463
encoding and other parts of bzrlib.
465
:ivar revisions: A bzrlib.versionedfile.VersionedFiles instance containing
466
the serialised revisions for the repository. This can be used to obtain
467
revision graph information or to access raw serialised revisions.
468
The result of trying to insert data into the repository via this store
469
is undefined: it should be considered read-only except for implementors
471
:ivar signatures: A bzrlib.versionedfile.VersionedFiles instance containing
472
the serialised signatures for the repository. This can be used to
473
obtain access to raw serialised signatures. The result of trying to
474
insert data into the repository via this store is undefined: it should
475
be considered read-only except for implementors of repositories.
476
:ivar inventories: A bzrlib.versionedfile.VersionedFiles instance containing
477
the serialised inventories for the repository. This can be used to
478
obtain unserialised inventories. The result of trying to insert data
479
into the repository via this store is undefined: it should be
480
considered read-only except for implementors of repositories.
481
:ivar texts: A bzrlib.versionedfile.VersionedFiles instance containing the
482
texts of files and directories for the repository. This can be used to
483
obtain file texts or file graphs. Note that Repository.iter_file_bytes
484
is usually a better interface for accessing file texts.
485
The result of trying to insert data into the repository via this store
486
is undefined: it should be considered read-only except for implementors
488
:ivar _transport: Transport for file access to repository, typically
489
pointing to .bzr/repository.
262
def abort_write_group(self, suppress_errors=False):
492
# What class to use for a CommitBuilder. Often its simpler to change this
493
# in a Repository class subclass rather than to override
494
# get_commit_builder.
495
_commit_builder_class = CommitBuilder
496
# The search regex used by xml based repositories to determine what things
497
# where changed in a single commit.
498
_file_ids_altered_regex = lazy_regex.lazy_compile(
499
r'file_id="(?P<file_id>[^"]+)"'
500
r'.* revision="(?P<revision_id>[^"]+)"'
503
def abort_write_group(self):
263
504
"""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.
269
506
:seealso: start_write_group.
271
508
if self._write_group is not self.get_transaction():
272
509
# has an unlock or relock occured ?
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)
510
raise errors.BzrError('mismatched lock context and write group.')
511
self._abort_write_group()
290
512
self._write_group = None
292
514
def _abort_write_group(self):
293
515
"""Template method for per-repository write group cleanup.
295
This is called during abort before the write group is considered to be
517
This is called during abort before the write group is considered to be
296
518
finished and should cleanup any internal state accrued during the write
297
519
group. There is no requirement that data handed to the repository be
298
520
*not* made available - this is not a rollback - but neither should any
825
1041
:param revision_ids: An iterable of revision_ids.
826
1042
:return: A set of the revision_ids that were present.
828
raise NotImplementedError(self.has_revisions)
1044
parent_map = self.revisions.get_parent_map(
1045
[(rev_id,) for rev_id in revision_ids])
1047
if _mod_revision.NULL_REVISION in revision_ids:
1048
result.add(_mod_revision.NULL_REVISION)
1049
result.update([key[0] for key in parent_map])
830
1052
@needs_read_lock
831
1053
def get_revision(self, revision_id):
832
1054
"""Return the Revision object for a named revision."""
833
1055
return self.get_revisions([revision_id])[0]
835
1058
def get_revision_reconcile(self, revision_id):
836
1059
"""'reconcile' helper routine that allows access to a revision always.
838
1061
This variant of get_revision does not cross check the weave graph
839
1062
against the revision one as get_revision does: but it should only
840
1063
be used by reconcile, or reconcile-alike commands that are correcting
841
1064
or testing the revision graph.
843
raise NotImplementedError(self.get_revision_reconcile)
1066
return self._get_revisions([revision_id])[0]
845
1069
def get_revisions(self, revision_ids):
846
"""Get many revisions at once.
848
Repositories that need to check data on every revision read should
849
subclass this method.
851
raise NotImplementedError(self.get_revisions)
853
def get_deltas_for_revisions(self, revisions, specific_fileids=None):
1070
"""Get many revisions at once."""
1071
return self._get_revisions(revision_ids)
1074
def _get_revisions(self, revision_ids):
1075
"""Core work logic to get many revisions without sanity checks."""
1076
for rev_id in revision_ids:
1077
if not rev_id or not isinstance(rev_id, basestring):
1078
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1079
keys = [(key,) for key in revision_ids]
1080
stream = self.revisions.get_record_stream(keys, 'unordered', True)
1082
for record in stream:
1083
if record.storage_kind == 'absent':
1084
raise errors.NoSuchRevision(self, record.key[0])
1085
text = record.get_bytes_as('fulltext')
1086
rev = self._serializer.read_revision_from_string(text)
1087
revs[record.key[0]] = rev
1088
return [revs[revid] for revid in revision_ids]
1091
def get_revision_xml(self, revision_id):
1092
# TODO: jam 20070210 This shouldn't be necessary since get_revision
1093
# would have already do it.
1094
# TODO: jam 20070210 Just use _serializer.write_revision_to_string()
1095
rev = self.get_revision(revision_id)
1096
rev_tmp = StringIO()
1097
# the current serializer..
1098
self._serializer.write_revision(rev, rev_tmp)
1100
return rev_tmp.getvalue()
1102
def get_deltas_for_revisions(self, revisions):
854
1103
"""Produce a generator of revision deltas.
856
1105
Note that the input is a sequence of REVISIONS, not revision_ids.
857
1106
Trees will be held in memory until the generator exits.
858
1107
Each delta is relative to the revision's lefthand predecessor.
860
:param specific_fileids: if not None, the result is filtered
861
so that only those file-ids, their parents and their
862
children are included.
864
# Get the revision-ids of interest
865
1109
required_trees = set()
866
1110
for revision in revisions:
867
1111
required_trees.add(revision.revision_id)
868
1112
required_trees.update(revision.parent_ids[:1])
870
# Get the matching filtered trees. Note that it's more
871
# efficient to pass filtered trees to changes_from() rather
872
# than doing the filtering afterwards. changes_from() could
873
# arguably do the filtering itself but it's path-based, not
874
# file-id based, so filtering before or afterwards is
876
if specific_fileids is None:
877
trees = dict((t.get_revision_id(), t) for
878
t in self.revision_trees(required_trees))
880
trees = dict((t.get_revision_id(), t) for
881
t in self._filtered_revision_trees(required_trees,
884
# Calculate the deltas
1113
trees = dict((t.get_revision_id(), t) for
1114
t in self.revision_trees(required_trees))
885
1115
for revision in revisions:
886
1116
if not revision.parent_ids:
887
old_tree = self.revision_tree(_mod_revision.NULL_REVISION)
1117
old_tree = self.revision_tree(None)
889
1119
old_tree = trees[revision.parent_ids[0]]
890
1120
yield trees[revision.revision_id].changes_from(old_tree)
892
1122
@needs_read_lock
893
def get_revision_delta(self, revision_id, specific_fileids=None):
1123
def get_revision_delta(self, revision_id):
894
1124
"""Return the delta for one revision.
896
1126
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.
903
1129
r = self.get_revision(revision_id)
904
return list(self.get_deltas_for_revisions([r],
905
specific_fileids=specific_fileids))[0]
1130
return list(self.get_deltas_for_revisions([r]))[0]
907
1132
@needs_write_lock
908
1133
def store_revision_signature(self, gpg_strategy, plaintext, revision_id):
909
1134
signature = gpg_strategy.sign(plaintext)
910
1135
self.add_signature_text(revision_id, signature)
912
1138
def add_signature_text(self, revision_id, signature):
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)
1139
self.signatures.add_lines((revision_id,), (),
1140
osutils.split_lines(signature))
1142
def find_text_key_references(self):
1143
"""Find the text key references within the repository.
1145
:return: a dictionary mapping (file_id, revision_id) tuples to altered file-ids to an iterable of
1146
revision_ids. Each altered file-ids has the exact revision_ids that
1147
altered it listed explicitly.
1148
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1149
to whether they were referred to by the inventory of the
1150
revision_id that they contain. The inventory texts from all present
1151
revision ids are assessed to generate this report.
1153
revision_keys = self.revisions.keys()
1154
w = self.inventories
1155
pb = ui.ui_factory.nested_progress_bar()
1157
return self._find_text_key_references_from_xml_inventory_lines(
1158
w.iter_lines_added_or_present_in_keys(revision_keys, pb=pb))
1162
def _find_text_key_references_from_xml_inventory_lines(self,
1164
"""Core routine for extracting references to texts from inventories.
1166
This performs the translation of xml lines to revision ids.
1168
:param line_iterator: An iterator of lines, origin_version_id
1169
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1170
to whether they were referred to by the inventory of the
1171
revision_id that they contain. Note that if that revision_id was
1172
not part of the line_iterator's output then False will be given -
1173
even though it may actually refer to that key.
1175
if not self._serializer.support_altered_by_hack:
1176
raise AssertionError(
1177
"_find_text_key_references_from_xml_inventory_lines only "
1178
"supported for branches which store inventory as unnested xml"
1179
", not on %r" % self)
1182
# this code needs to read every new line in every inventory for the
1183
# inventories [revision_ids]. Seeing a line twice is ok. Seeing a line
1184
# not present in one of those inventories is unnecessary but not
1185
# harmful because we are filtering by the revision id marker in the
1186
# inventory lines : we only select file ids altered in one of those
1187
# revisions. We don't need to see all lines in the inventory because
1188
# only those added in an inventory in rev X can contain a revision=X
1190
unescape_revid_cache = {}
1191
unescape_fileid_cache = {}
1193
# jam 20061218 In a big fetch, this handles hundreds of thousands
1194
# of lines, so it has had a lot of inlining and optimizing done.
1195
# Sorry that it is a little bit messy.
1196
# Move several functions to be local variables, since this is a long
1198
search = self._file_ids_altered_regex.search
1199
unescape = _unescape_xml
1200
setdefault = result.setdefault
1201
for line, line_key in line_iterator:
1202
match = search(line)
1205
# One call to match.group() returning multiple items is quite a
1206
# bit faster than 2 calls to match.group() each returning 1
1207
file_id, revision_id = match.group('file_id', 'revision_id')
1209
# Inlining the cache lookups helps a lot when you make 170,000
1210
# lines and 350k ids, versus 8.4 unique ids.
1211
# Using a cache helps in 2 ways:
1212
# 1) Avoids unnecessary decoding calls
1213
# 2) Re-uses cached strings, which helps in future set and
1215
# (2) is enough that removing encoding entirely along with
1216
# the cache (so we are using plain strings) results in no
1217
# performance improvement.
1219
revision_id = unescape_revid_cache[revision_id]
1221
unescaped = unescape(revision_id)
1222
unescape_revid_cache[revision_id] = unescaped
1223
revision_id = unescaped
1225
# Note that unconditionally unescaping means that we deserialise
1226
# every fileid, which for general 'pull' is not great, but we don't
1227
# really want to have some many fulltexts that this matters anyway.
1230
file_id = unescape_fileid_cache[file_id]
1232
unescaped = unescape(file_id)
1233
unescape_fileid_cache[file_id] = unescaped
1236
key = (file_id, revision_id)
1237
setdefault(key, False)
1238
if revision_id == line_key[-1]:
1242
def _find_file_ids_from_xml_inventory_lines(self, line_iterator,
1244
"""Helper routine for fileids_altered_by_revision_ids.
1246
This performs the translation of xml lines to revision ids.
1248
:param line_iterator: An iterator of lines, origin_version_id
1249
:param revision_ids: The revision ids to filter for. This should be a
1250
set or other type which supports efficient __contains__ lookups, as
1251
the revision id from each parsed line will be looked up in the
1252
revision_ids filter.
1253
:return: a dictionary mapping altered file-ids to an iterable of
1254
revision_ids. Each altered file-ids has the exact revision_ids that
1255
altered it listed explicitly.
1258
setdefault = result.setdefault
1260
self._find_text_key_references_from_xml_inventory_lines(
1261
line_iterator).iterkeys():
1262
# once data is all ensured-consistent; then this is
1263
# if revision_id == version_id
1264
if key[-1:] in revision_ids:
1265
setdefault(key[0], set()).add(key[-1])
1268
def fileids_altered_by_revision_ids(self, revision_ids, _inv_weave=None):
1269
"""Find the file ids and versions affected by revisions.
1271
:param revisions: an iterable containing revision ids.
1272
:param _inv_weave: The inventory weave from this repository or None.
1273
If None, the inventory weave will be opened automatically.
1274
:return: a dictionary mapping altered file-ids to an iterable of
1275
revision_ids. Each altered file-ids has the exact revision_ids that
1276
altered it listed explicitly.
1278
selected_keys = set((revid,) for revid in revision_ids)
1279
w = _inv_weave or self.inventories
1280
pb = ui.ui_factory.nested_progress_bar()
1282
return self._find_file_ids_from_xml_inventory_lines(
1283
w.iter_lines_added_or_present_in_keys(
1284
selected_keys, pb=pb),
933
1289
def iter_files_bytes(self, desired_files):
934
1290
"""Iterate through file versions.
941
1297
uniquely identify the file version in the caller's context. (Examples:
942
1298
an index number or a TreeTransform trans_id.)
1300
bytes_iterator is an iterable of bytestrings for the file. The
1301
kind of iterable and length of the bytestrings are unspecified, but for
1302
this implementation, it is a list of bytes produced by
1303
VersionedFile.get_record_stream().
944
1305
:param desired_files: a list of (file_id, revision_id, identifier)
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])
1308
transaction = self.get_transaction()
1310
for file_id, revision_id, callable_data in desired_files:
1311
text_keys[(file_id, revision_id)] = callable_data
1312
for record in self.texts.get_record_stream(text_keys, 'unordered', True):
1313
if record.storage_kind == 'absent':
1314
raise errors.RevisionNotPresent(record.key, self)
1315
yield text_keys[record.key], record.get_bytes_as('fulltext')
1317
def _generate_text_key_index(self, text_key_references=None,
1319
"""Generate a new text key index for the repository.
1321
This is an expensive function that will take considerable time to run.
1323
:return: A dict mapping text keys ((file_id, revision_id) tuples) to a
1324
list of parents, also text keys. When a given key has no parents,
1325
the parents list will be [NULL_REVISION].
1327
# All revisions, to find inventory parents.
1328
if ancestors is None:
1329
graph = self.get_graph()
1330
ancestors = graph.get_parent_map(self.all_revision_ids())
1331
if text_key_references is None:
1332
text_key_references = self.find_text_key_references()
1333
pb = ui.ui_factory.nested_progress_bar()
1335
return self._do_generate_text_key_index(ancestors,
1336
text_key_references, pb)
1340
def _do_generate_text_key_index(self, ancestors, text_key_references, pb):
1341
"""Helper for _generate_text_key_index to avoid deep nesting."""
1342
revision_order = tsort.topo_sort(ancestors)
1343
invalid_keys = set()
1345
for revision_id in revision_order:
1346
revision_keys[revision_id] = set()
1347
text_count = len(text_key_references)
1348
# a cache of the text keys to allow reuse; costs a dict of all the
1349
# keys, but saves a 2-tuple for every child of a given key.
1351
for text_key, valid in text_key_references.iteritems():
1353
invalid_keys.add(text_key)
1355
revision_keys[text_key[1]].add(text_key)
1356
text_key_cache[text_key] = text_key
1357
del text_key_references
1359
text_graph = graph.Graph(graph.DictParentsProvider(text_index))
1360
NULL_REVISION = _mod_revision.NULL_REVISION
1361
# Set a cache with a size of 10 - this suffices for bzr.dev but may be
1362
# too small for large or very branchy trees. However, for 55K path
1363
# trees, it would be easy to use too much memory trivially. Ideally we
1364
# could gauge this by looking at available real memory etc, but this is
1365
# always a tricky proposition.
1366
inventory_cache = lru_cache.LRUCache(10)
1367
batch_size = 10 # should be ~150MB on a 55K path tree
1368
batch_count = len(revision_order) / batch_size + 1
1370
pb.update("Calculating text parents.", processed_texts, text_count)
1371
for offset in xrange(batch_count):
1372
to_query = revision_order[offset * batch_size:(offset + 1) *
1376
for rev_tree in self.revision_trees(to_query):
1377
revision_id = rev_tree.get_revision_id()
1378
parent_ids = ancestors[revision_id]
1379
for text_key in revision_keys[revision_id]:
1380
pb.update("Calculating text parents.", processed_texts)
1381
processed_texts += 1
1382
candidate_parents = []
1383
for parent_id in parent_ids:
1384
parent_text_key = (text_key[0], parent_id)
1386
check_parent = parent_text_key not in \
1387
revision_keys[parent_id]
1389
# the parent parent_id is a ghost:
1390
check_parent = False
1391
# truncate the derived graph against this ghost.
1392
parent_text_key = None
1394
# look at the parent commit details inventories to
1395
# determine possible candidates in the per file graph.
1398
inv = inventory_cache[parent_id]
1400
inv = self.revision_tree(parent_id).inventory
1401
inventory_cache[parent_id] = inv
1402
parent_entry = inv._byid.get(text_key[0], None)
1403
if parent_entry is not None:
1405
text_key[0], parent_entry.revision)
1407
parent_text_key = None
1408
if parent_text_key is not None:
1409
candidate_parents.append(
1410
text_key_cache[parent_text_key])
1411
parent_heads = text_graph.heads(candidate_parents)
1412
new_parents = list(parent_heads)
1413
new_parents.sort(key=lambda x:candidate_parents.index(x))
1414
if new_parents == []:
1415
new_parents = [NULL_REVISION]
1416
text_index[text_key] = new_parents
1418
for text_key in invalid_keys:
1419
text_index[text_key] = [NULL_REVISION]
1422
def item_keys_introduced_by(self, revision_ids, _files_pb=None):
1423
"""Get an iterable listing the keys of all the data introduced by a set
1426
The keys will be ordered so that the corresponding items can be safely
1427
fetched and inserted in that order.
1429
:returns: An iterable producing tuples of (knit-kind, file-id,
1430
versions). knit-kind is one of 'file', 'inventory', 'signatures',
1431
'revisions'. file-id is None unless knit-kind is 'file'.
1433
# XXX: it's a bit weird to control the inventory weave caching in this
1434
# generator. Ideally the caching would be done in fetch.py I think. Or
1435
# maybe this generator should explicitly have the contract that it
1436
# should not be iterated until the previously yielded item has been
1438
inv_w = self.inventories
1440
# file ids that changed
1441
file_ids = self.fileids_altered_by_revision_ids(revision_ids, inv_w)
1443
num_file_ids = len(file_ids)
1444
for file_id, altered_versions in file_ids.iteritems():
1445
if _files_pb is not None:
1446
_files_pb.update("fetch texts", count, num_file_ids)
1448
yield ("file", file_id, altered_versions)
1449
# We're done with the files_pb. Note that it finished by the caller,
1450
# just as it was created by the caller.
1454
yield ("inventory", None, revision_ids)
1457
revisions_with_signatures = set()
1458
for rev_id in revision_ids:
1460
self.get_signature_text(rev_id)
1461
except errors.NoSuchRevision:
1465
revisions_with_signatures.add(rev_id)
1466
yield ("signatures", None, revisions_with_signatures)
1469
yield ("revisions", None, revision_ids)
1472
def get_inventory(self, revision_id):
1473
"""Get Inventory object by revision id."""
1474
return self.iter_inventories([revision_id]).next()
1476
def iter_inventories(self, revision_ids):
1477
"""Get many inventories by revision_ids.
1479
This will buffer some or all of the texts used in constructing the
1480
inventories in memory, but will only parse a single inventory at a
1483
:return: An iterator of inventories.
1485
if ((None in revision_ids)
1486
or (_mod_revision.NULL_REVISION in revision_ids)):
1487
raise ValueError('cannot get null revision inventory')
1488
return self._iter_inventories(revision_ids)
1490
def _iter_inventories(self, revision_ids):
1491
"""single-document based inventory iteration."""
1492
for text, revision_id in self._iter_inventory_xmls(revision_ids):
1493
yield self.deserialise_inventory(revision_id, text)
1495
def _iter_inventory_xmls(self, revision_ids):
1496
keys = [(revision_id,) for revision_id in revision_ids]
1497
stream = self.inventories.get_record_stream(keys, 'unordered', True)
1499
for record in stream:
1500
if record.storage_kind != 'absent':
1501
texts[record.key] = record.get_bytes_as('fulltext')
1503
raise errors.NoSuchRevision(self, record.key)
1505
yield texts[key], key[-1]
1507
def deserialise_inventory(self, revision_id, xml):
1508
"""Transform the xml into an inventory object.
1510
:param revision_id: The expected revision id of the inventory.
1511
:param xml: A serialised inventory.
1513
result = self._serializer.read_inventory_from_string(xml, revision_id)
1514
if result.revision_id != revision_id:
1515
raise AssertionError('revision id mismatch %s != %s' % (
1516
result.revision_id, revision_id))
1519
def serialise_inventory(self, inv):
1520
return self._serializer.write_inventory_to_string(inv)
1522
def _serialise_inventory_to_lines(self, inv):
1523
return self._serializer.write_inventory_to_lines(inv)
1525
def get_serializer_format(self):
1526
return self._serializer.format_num
1529
def get_inventory_xml(self, revision_id):
1530
"""Get inventory XML as a file object."""
1531
texts = self._iter_inventory_xmls([revision_id])
1533
text, revision_id = texts.next()
1534
except StopIteration:
1535
raise errors.HistoryMissing(self, 'inventory', revision_id)
1539
def get_inventory_sha1(self, revision_id):
1540
"""Return the sha1 hash of the inventory entry
1542
return self.get_revision(revision_id).inventory_sha1
1544
def iter_reverse_revision_history(self, revision_id):
1545
"""Iterate backwards through revision ids in the lefthand history
1547
:param revision_id: The revision id to start with. All its lefthand
1548
ancestors will be traversed.
1550
graph = self.get_graph()
1551
next_id = revision_id
1553
if next_id in (None, _mod_revision.NULL_REVISION):
1556
# Note: The following line may raise KeyError in the event of
1557
# truncated history. We decided not to have a try:except:raise
1558
# RevisionNotPresent here until we see a use for it, because of the
1559
# cost in an inner loop that is by its very nature O(history).
1560
# Robert Collins 20080326
1561
parents = graph.get_parent_map([next_id])[next_id]
1562
if len(parents) == 0:
1565
next_id = parents[0]
1568
def get_revision_inventory(self, revision_id):
1569
"""Return inventory of a past revision."""
1570
# TODO: Unify this with get_inventory()
1571
# bzr 0.0.6 and later imposes the constraint that the inventory_id
1572
# must be the same as its revision, so this is trivial.
1573
if revision_id is None:
1574
# This does not make sense: if there is no revision,
1575
# then it is the current tree inventory surely ?!
1576
# and thus get_root_id() is something that looks at the last
1577
# commit on the branch, and the get_root_id is an inventory check.
1578
raise NotImplementedError
1579
# return Inventory(self.get_root_id())
1581
return self.get_inventory(revision_id)
984
1584
def is_shared(self):
985
1585
"""Return True if this repository is flagged as a shared repository."""
986
1586
raise NotImplementedError(self.is_shared)
1000
1600
for repositories to maintain loaded indices across multiple locks
1001
1601
by checking inside their implementation of this method to see
1002
1602
whether their indices are still valid. This depends of course on
1003
the disk format being validatable in this manner. This method is
1004
also called by the refresh_data() public interface to cause a refresh
1005
to occur while in a write lock so that data inserted by a smart server
1006
push operation is visible on the client's instance of the physical
1603
the disk format being validatable in this manner.
1010
1606
@needs_read_lock
1011
1607
def revision_tree(self, revision_id):
1012
1608
"""Return Tree for a revision on this branch.
1014
`revision_id` may be NULL_REVISION for the empty tree revision.
1610
`revision_id` may be None for the empty tree revision.
1016
raise NotImplementedError(self.revision_tree)
1612
# TODO: refactor this to use an existing revision object
1613
# so we don't need to read it in twice.
1614
if revision_id is None or revision_id == _mod_revision.NULL_REVISION:
1615
return RevisionTree(self, Inventory(root_id=None),
1616
_mod_revision.NULL_REVISION)
1618
inv = self.get_revision_inventory(revision_id)
1619
return RevisionTree(self, inv, revision_id)
1018
1621
def revision_trees(self, revision_ids):
1019
"""Return Trees for revisions in this repository.
1021
:param revision_ids: a sequence of revision-ids;
1022
a revision-id may not be None or 'null:'
1622
"""Return Tree for a revision on this branch.
1624
`revision_id` may not be None or 'null:'"""
1625
inventories = self.iter_inventories(revision_ids)
1626
for inv in inventories:
1627
yield RevisionTree(self, inv, inv.revision_id)
1630
def get_ancestry(self, revision_id, topo_sorted=True):
1631
"""Return a list of revision-ids integrated by a revision.
1633
The first element of the list is always None, indicating the origin
1634
revision. This might change when we have history horizons, or
1635
perhaps we should have a new API.
1637
This is topologically sorted.
1024
raise NotImplementedError(self.revision_trees)
1639
if _mod_revision.is_null(revision_id):
1641
if not self.has_revision(revision_id):
1642
raise errors.NoSuchRevision(self, revision_id)
1643
graph = self.get_graph()
1645
search = graph._make_breadth_first_searcher([revision_id])
1648
found, ghosts = search.next_with_ghosts()
1649
except StopIteration:
1652
if _mod_revision.NULL_REVISION in keys:
1653
keys.remove(_mod_revision.NULL_REVISION)
1655
parent_map = graph.get_parent_map(keys)
1656
keys = tsort.topo_sort(parent_map)
1657
return [None] + list(keys)
1026
def pack(self, hint=None, clean_obsolete_packs=False):
1027
1660
"""Compress the data within the repository.
1029
1662
This operation only makes sense for some repository types. For other
1030
1663
types it should be a no-op that just returns.
1032
1665
This stub method does not require a lock, but subclasses should use
1033
@needs_write_lock as this is a long running call it's reasonable to
1666
@needs_write_lock as this is a long running call its reasonable to
1034
1667
implicitly lock for the user.
1036
:param hint: If not supplied, the whole repository is packed.
1037
If supplied, the repository may use the hint parameter as a
1038
hint for the parts of the repository to pack. A hint can be
1039
obtained from the result of commit_write_group(). Out of
1040
date hints are simply ignored, because concurrent operations
1041
can obsolete them rapidly.
1043
:param clean_obsolete_packs: Clean obsolete packs immediately after
1671
@deprecated_method(one_six)
1672
def print_file(self, file, revision_id):
1673
"""Print `file` to stdout.
1675
FIXME RBC 20060125 as John Meinel points out this is a bad api
1676
- it writes to stdout, it assumes that that is valid etc. Fix
1677
by creating a new more flexible convenience function.
1679
tree = self.revision_tree(revision_id)
1680
# use inventory as it was in that revision
1681
file_id = tree.inventory.path2id(file)
1683
# TODO: jam 20060427 Write a test for this code path
1684
# it had a bug in it, and was raising the wrong
1686
raise errors.BzrError("%r is not present in revision %s" % (file, revision_id))
1687
tree.print_file(file_id)
1047
1689
def get_transaction(self):
1048
1690
return self.control_files.get_transaction()
1050
def get_parent_map(self, revision_ids):
1051
"""See graph.StackedParentsProvider.get_parent_map"""
1052
raise NotImplementedError(self.get_parent_map)
1692
@deprecated_method(one_one)
1693
def get_parents(self, revision_ids):
1694
"""See StackedParentsProvider.get_parents"""
1695
parent_map = self.get_parent_map(revision_ids)
1696
return [parent_map.get(r, None) for r in revision_ids]
1054
def _get_parent_map_no_fallbacks(self, revision_ids):
1055
"""Same as Repository.get_parent_map except doesn't query fallbacks."""
1056
# revisions index works in keys; this just works in revisions
1057
# therefore wrap and unwrap
1060
for revision_id in revision_ids:
1698
def get_parent_map(self, keys):
1699
"""See graph._StackedParentsProvider.get_parent_map"""
1701
for revision_id in keys:
1702
if revision_id is None:
1703
raise ValueError('get_parent_map(None) is not valid')
1061
1704
if revision_id == _mod_revision.NULL_REVISION:
1062
result[revision_id] = ()
1063
elif revision_id is None:
1064
raise ValueError('get_parent_map(None) is not valid')
1066
query_keys.append((revision_id ,))
1067
vf = self.revisions.without_fallbacks()
1068
for ((revision_id,), parent_keys) in \
1069
vf.get_parent_map(query_keys).iteritems():
1071
result[revision_id] = tuple([parent_revid
1072
for (parent_revid,) in parent_keys])
1074
result[revision_id] = (_mod_revision.NULL_REVISION,)
1705
parent_map[revision_id] = ()
1708
parent_id_list = self.get_revision(revision_id).parent_ids
1709
except errors.NoSuchRevision:
1712
if len(parent_id_list) == 0:
1713
parent_ids = (_mod_revision.NULL_REVISION,)
1715
parent_ids = tuple(parent_id_list)
1716
parent_map[revision_id] = parent_ids
1077
1719
def _make_parents_provider(self):
1078
if not self._format.supports_external_lookups:
1080
return graph.StackedParentsProvider(_LazyListJoin(
1081
[self._make_parents_provider_unstacked()],
1082
self._fallback_repositories))
1084
def _make_parents_provider_unstacked(self):
1085
return graph.CallableToParentsProviderAdapter(
1086
self._get_parent_map_no_fallbacks)
1089
def get_known_graph_ancestry(self, revision_ids):
1090
"""Return the known graph for a set of revision ids and their ancestors.
1092
raise NotImplementedError(self.get_known_graph_ancestry)
1094
def get_file_graph(self):
1095
"""Return the graph walker for files."""
1096
raise NotImplementedError(self.get_file_graph)
1098
1722
def get_graph(self, other_repository=None):
1099
1723
"""Return the graph walker for this repository format"""
1100
1724
parents_provider = self._make_parents_provider()
1101
1725
if (other_repository is not None and
1102
1726
not self.has_same_location(other_repository)):
1103
parents_provider = graph.StackedParentsProvider(
1727
parents_provider = graph._StackedParentsProvider(
1104
1728
[parents_provider, other_repository._make_parents_provider()])
1105
1729
return graph.Graph(parents_provider)
1731
def _get_versioned_file_checker(self):
1732
"""Return an object suitable for checking versioned files."""
1733
return _VersionedFileChecker(self)
1735
def revision_ids_to_search_result(self, result_set):
1736
"""Convert a set of revision ids to a graph SearchResult."""
1737
result_parents = set()
1738
for parents in self.get_graph().get_parent_map(
1739
result_set).itervalues():
1740
result_parents.update(parents)
1741
included_keys = result_set.intersection(result_parents)
1742
start_keys = result_set.difference(included_keys)
1743
exclude_keys = result_parents.difference(result_set)
1744
result = graph.SearchResult(start_keys, exclude_keys,
1745
len(result_set), result_set)
1107
1748
@needs_write_lock
1108
1749
def set_make_working_trees(self, new_value):
1109
1750
"""Set the policy flag for making working trees when creating branches.
1632
2281
_optimisers = []
1633
2282
"""The available optimised InterRepository types."""
1636
2284
def copy_content(self, revision_id=None):
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):
2285
raise NotImplementedError(self.copy_content)
2287
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
1653
2288
"""Fetch the content required to construct revision_id.
1655
2290
The content is copied from self.source to self.target.
1657
2292
:param revision_id: if None all content is copied, if NULL_REVISION no
1658
2293
content is copied.
2294
:param pb: optional progress bar to use for progress reports. If not
2295
provided a default one will be created.
2297
Returns the copied revision count and the failed revisions in a tuple:
1661
2300
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
2302
def _walk_to_common_revisions(self, revision_ids):
2303
"""Walk out from revision_ids in source to revisions target has.
2305
:param revision_ids: The start point for the search.
2306
:return: A set of revision ids.
2308
target_graph = self.target.get_graph()
2309
revision_ids = frozenset(revision_ids)
2310
if set(target_graph.get_parent_map(revision_ids)) == revision_ids:
2311
return graph.SearchResult(revision_ids, set(), 0, set())
2312
missing_revs = set()
2313
source_graph = self.source.get_graph()
2314
# ensure we don't pay silly lookup costs.
2315
searcher = source_graph._make_breadth_first_searcher(revision_ids)
2316
null_set = frozenset([_mod_revision.NULL_REVISION])
2319
next_revs, ghosts = searcher.next_with_ghosts()
2320
except StopIteration:
2322
if revision_ids.intersection(ghosts):
2323
absent_ids = set(revision_ids.intersection(ghosts))
2324
# If all absent_ids are present in target, no error is needed.
2325
absent_ids.difference_update(
2326
set(target_graph.get_parent_map(absent_ids)))
2328
raise errors.NoSuchRevision(self.source, absent_ids.pop())
2329
# we don't care about other ghosts as we can't fetch them and
2330
# haven't been asked to.
2331
next_revs = set(next_revs)
2332
# we always have NULL_REVISION present.
2333
have_revs = set(target_graph.get_parent_map(next_revs)).union(null_set)
2334
missing_revs.update(next_revs - have_revs)
2335
searcher.stop_searching_any(have_revs)
2336
return searcher.get_result()
2338
@deprecated_method(one_two)
2340
def missing_revision_ids(self, revision_id=None, find_ghosts=True):
2341
"""Return the revision ids that source has that target does not.
2343
These are returned in topological order.
2345
:param revision_id: only return revision ids included by this
2347
:param find_ghosts: If True find missing revisions in deep history
2348
rather than just finding the surface difference.
2350
return list(self.search_missing_revision_ids(
2351
revision_id, find_ghosts).get_keys())
2354
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2355
"""Return the revision ids that source has that target does not.
2357
:param revision_id: only return revision ids included by this
2359
:param find_ghosts: If True find missing revisions in deep history
2360
rather than just finding the surface difference.
1683
2361
:return: A bzrlib.graph.SearchResult.
1685
raise NotImplementedError(self.search_missing_revision_ids)
2363
# stop searching at found target revisions.
2364
if not find_ghosts and revision_id is not None:
2365
return self._walk_to_common_revisions([revision_id])
2366
# generic, possibly worst case, slow code path.
2367
target_ids = set(self.target.all_revision_ids())
2368
if revision_id is not None:
2369
source_ids = self.source.get_ancestry(revision_id)
2370
if source_ids[0] is not None:
2371
raise AssertionError()
2374
source_ids = self.source.all_revision_ids()
2375
result_set = set(source_ids).difference(target_ids)
2376
return self.source.revision_ids_to_search_result(result_set)
1688
2379
def _same_model(source, target):
1689
"""True if source and target have the same data representation.
1691
Note: this is always called on the base class; overriding it in a
1692
subclass will have no effect.
1695
InterRepository._assert_same_model(source, target)
1697
except errors.IncompatibleRepositories, e:
2380
"""True if source and target have the same data representation."""
2381
if source.supports_rich_root() != target.supports_rich_root():
1701
def _assert_same_model(source, target):
1702
"""Raise an exception if two repositories do not use the same model.
1704
if source.supports_rich_root() != target.supports_rich_root():
1705
raise errors.IncompatibleRepositories(source, target,
1706
"different rich-root support")
1707
2383
if source._serializer != target._serializer:
1708
raise errors.IncompatibleRepositories(source, target,
1709
"different serializers")
2388
class InterSameDataRepository(InterRepository):
2389
"""Code for converting between repositories that represent the same data.
2391
Data format and model must match for this to work.
2395
def _get_repo_format_to_test(self):
2396
"""Repository format for testing with.
2398
InterSameData can pull from subtree to subtree and from non-subtree to
2399
non-subtree, so we test this with the richest repository format.
2401
from bzrlib.repofmt import knitrepo
2402
return knitrepo.RepositoryFormatKnit3()
2405
def is_compatible(source, target):
2406
return InterRepository._same_model(source, target)
2409
def copy_content(self, revision_id=None):
2410
"""Make a complete copy of the content in self into destination.
2412
This copies both the repository's revision data, and configuration information
2413
such as the make_working_trees setting.
2415
This is a destructive operation! Do not use it on existing
2418
:param revision_id: Only copy the content needed to construct
2419
revision_id and its parents.
2422
self.target.set_make_working_trees(self.source.make_working_trees())
2423
except NotImplementedError:
2425
# but don't bother fetching if we have the needed data now.
2426
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
2427
self.target.has_revision(revision_id)):
2429
self.target.fetch(self.source, revision_id=revision_id)
2432
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2433
"""See InterRepository.fetch()."""
2434
from bzrlib.fetch import GenericRepoFetcher
2435
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2436
self.source, self.source._format, self.target,
2437
self.target._format)
2438
f = GenericRepoFetcher(to_repository=self.target,
2439
from_repository=self.source,
2440
last_revision=revision_id,
2441
pb=pb, find_ghosts=find_ghosts)
2442
return f.count_copied, f.failed_revisions
2445
class InterWeaveRepo(InterSameDataRepository):
2446
"""Optimised code paths between Weave based repositories.
2448
This should be in bzrlib/repofmt/weaverepo.py but we have not yet
2449
implemented lazy inter-object optimisation.
2453
def _get_repo_format_to_test(self):
2454
from bzrlib.repofmt import weaverepo
2455
return weaverepo.RepositoryFormat7()
2458
def is_compatible(source, target):
2459
"""Be compatible with known Weave formats.
2461
We don't test for the stores being of specific types because that
2462
could lead to confusing results, and there is no need to be
2465
from bzrlib.repofmt.weaverepo import (
2471
return (isinstance(source._format, (RepositoryFormat5,
2473
RepositoryFormat7)) and
2474
isinstance(target._format, (RepositoryFormat5,
2476
RepositoryFormat7)))
2477
except AttributeError:
2481
def copy_content(self, revision_id=None):
2482
"""See InterRepository.copy_content()."""
2483
# weave specific optimised path:
2485
self.target.set_make_working_trees(self.source.make_working_trees())
2486
except (errors.RepositoryUpgradeRequired, NotImplemented):
2488
# FIXME do not peek!
2489
if self.source._transport.listable():
2490
pb = ui.ui_factory.nested_progress_bar()
2492
self.target.texts.insert_record_stream(
2493
self.source.texts.get_record_stream(
2494
self.source.texts.keys(), 'topological', False))
2495
pb.update('copying inventory', 0, 1)
2496
self.target.inventories.insert_record_stream(
2497
self.source.inventories.get_record_stream(
2498
self.source.inventories.keys(), 'topological', False))
2499
self.target.signatures.insert_record_stream(
2500
self.source.signatures.get_record_stream(
2501
self.source.signatures.keys(),
2503
self.target.revisions.insert_record_stream(
2504
self.source.revisions.get_record_stream(
2505
self.source.revisions.keys(),
2506
'topological', True))
2510
self.target.fetch(self.source, revision_id=revision_id)
2513
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2514
"""See InterRepository.fetch()."""
2515
from bzrlib.fetch import GenericRepoFetcher
2516
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2517
self.source, self.source._format, self.target, self.target._format)
2518
f = GenericRepoFetcher(to_repository=self.target,
2519
from_repository=self.source,
2520
last_revision=revision_id,
2521
pb=pb, find_ghosts=find_ghosts)
2522
return f.count_copied, f.failed_revisions
2525
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2526
"""See InterRepository.missing_revision_ids()."""
2527
# we want all revisions to satisfy revision_id in source.
2528
# but we don't want to stat every file here and there.
2529
# we want then, all revisions other needs to satisfy revision_id
2530
# checked, but not those that we have locally.
2531
# so the first thing is to get a subset of the revisions to
2532
# satisfy revision_id in source, and then eliminate those that
2533
# we do already have.
2534
# this is slow on high latency connection to self, but as as this
2535
# disk format scales terribly for push anyway due to rewriting
2536
# inventory.weave, this is considered acceptable.
2538
if revision_id is not None:
2539
source_ids = self.source.get_ancestry(revision_id)
2540
if source_ids[0] is not None:
2541
raise AssertionError()
2544
source_ids = self.source._all_possible_ids()
2545
source_ids_set = set(source_ids)
2546
# source_ids is the worst possible case we may need to pull.
2547
# now we want to filter source_ids against what we actually
2548
# have in target, but don't try to check for existence where we know
2549
# we do not have a revision as that would be pointless.
2550
target_ids = set(self.target._all_possible_ids())
2551
possibly_present_revisions = target_ids.intersection(source_ids_set)
2552
actually_present_revisions = set(
2553
self.target._eliminate_revisions_not_present(possibly_present_revisions))
2554
required_revisions = source_ids_set.difference(actually_present_revisions)
2555
if revision_id is not None:
2556
# we used get_ancestry to determine source_ids then we are assured all
2557
# revisions referenced are present as they are installed in topological order.
2558
# and the tip revision was validated by get_ancestry.
2559
result_set = required_revisions
2561
# if we just grabbed the possibly available ids, then
2562
# we only have an estimate of whats available and need to validate
2563
# that against the revision records.
2565
self.source._eliminate_revisions_not_present(required_revisions))
2566
return self.source.revision_ids_to_search_result(result_set)
2569
class InterKnitRepo(InterSameDataRepository):
2570
"""Optimised code paths between Knit based repositories."""
2573
def _get_repo_format_to_test(self):
2574
from bzrlib.repofmt import knitrepo
2575
return knitrepo.RepositoryFormatKnit1()
2578
def is_compatible(source, target):
2579
"""Be compatible with known Knit formats.
2581
We don't test for the stores being of specific types because that
2582
could lead to confusing results, and there is no need to be
2585
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit
2587
are_knits = (isinstance(source._format, RepositoryFormatKnit) and
2588
isinstance(target._format, RepositoryFormatKnit))
2589
except AttributeError:
2591
return are_knits and InterRepository._same_model(source, target)
2594
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2595
"""See InterRepository.fetch()."""
2596
from bzrlib.fetch import KnitRepoFetcher
2597
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2598
self.source, self.source._format, self.target, self.target._format)
2599
f = KnitRepoFetcher(to_repository=self.target,
2600
from_repository=self.source,
2601
last_revision=revision_id,
2602
pb=pb, find_ghosts=find_ghosts)
2603
return f.count_copied, f.failed_revisions
2606
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2607
"""See InterRepository.missing_revision_ids()."""
2608
if revision_id is not None:
2609
source_ids = self.source.get_ancestry(revision_id)
2610
if source_ids[0] is not None:
2611
raise AssertionError()
2614
source_ids = self.source.all_revision_ids()
2615
source_ids_set = set(source_ids)
2616
# source_ids is the worst possible case we may need to pull.
2617
# now we want to filter source_ids against what we actually
2618
# have in target, but don't try to check for existence where we know
2619
# we do not have a revision as that would be pointless.
2620
target_ids = set(self.target.all_revision_ids())
2621
possibly_present_revisions = target_ids.intersection(source_ids_set)
2622
actually_present_revisions = set(
2623
self.target._eliminate_revisions_not_present(possibly_present_revisions))
2624
required_revisions = source_ids_set.difference(actually_present_revisions)
2625
if revision_id is not None:
2626
# we used get_ancestry to determine source_ids then we are assured all
2627
# revisions referenced are present as they are installed in topological order.
2628
# and the tip revision was validated by get_ancestry.
2629
result_set = required_revisions
2631
# if we just grabbed the possibly available ids, then
2632
# we only have an estimate of whats available and need to validate
2633
# that against the revision records.
2635
self.source._eliminate_revisions_not_present(required_revisions))
2636
return self.source.revision_ids_to_search_result(result_set)
2639
class InterPackRepo(InterSameDataRepository):
2640
"""Optimised code paths between Pack based repositories."""
2643
def _get_repo_format_to_test(self):
2644
from bzrlib.repofmt import pack_repo
2645
return pack_repo.RepositoryFormatKnitPack1()
2648
def is_compatible(source, target):
2649
"""Be compatible with known Pack formats.
2651
We don't test for the stores being of specific types because that
2652
could lead to confusing results, and there is no need to be
2655
from bzrlib.repofmt.pack_repo import RepositoryFormatPack
2657
are_packs = (isinstance(source._format, RepositoryFormatPack) and
2658
isinstance(target._format, RepositoryFormatPack))
2659
except AttributeError:
2661
return are_packs and InterRepository._same_model(source, target)
2664
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2665
"""See InterRepository.fetch()."""
2666
from bzrlib.repofmt.pack_repo import Packer
2667
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2668
self.source, self.source._format, self.target, self.target._format)
2669
self.count_copied = 0
2670
if revision_id is None:
2672
# everything to do - use pack logic
2673
# to fetch from all packs to one without
2674
# inventory parsing etc, IFF nothing to be copied is in the target.
2676
revision_ids = self.source.all_revision_ids()
2677
revision_keys = [(revid,) for revid in revision_ids]
2678
index = self.target._pack_collection.revision_index.combined_index
2679
present_revision_ids = set(item[1][0] for item in
2680
index.iter_entries(revision_keys))
2681
revision_ids = set(revision_ids) - present_revision_ids
2682
# implementing the TODO will involve:
2683
# - detecting when all of a pack is selected
2684
# - avoiding as much as possible pre-selection, so the
2685
# more-core routines such as create_pack_from_packs can filter in
2686
# a just-in-time fashion. (though having a HEADS list on a
2687
# repository might make this a lot easier, because we could
2688
# sensibly detect 'new revisions' without doing a full index scan.
2689
elif _mod_revision.is_null(revision_id):
2694
revision_ids = self.search_missing_revision_ids(revision_id,
2695
find_ghosts=find_ghosts).get_keys()
2696
except errors.NoSuchRevision:
2697
raise errors.InstallFailed([revision_id])
2698
if len(revision_ids) == 0:
2700
packs = self.source._pack_collection.all_packs()
2701
pack = Packer(self.target._pack_collection, packs, '.fetch',
2702
revision_ids).pack()
2703
if pack is not None:
2704
self.target._pack_collection._save_pack_names()
2705
# Trigger an autopack. This may duplicate effort as we've just done
2706
# a pack creation, but for now it is simpler to think about as
2707
# 'upload data, then repack if needed'.
2708
self.target._pack_collection.autopack()
2709
return (pack.get_revision_count(), [])
2714
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2715
"""See InterRepository.missing_revision_ids().
2717
:param find_ghosts: Find ghosts throughout the ancestry of
2720
if not find_ghosts and revision_id is not None:
2721
return self._walk_to_common_revisions([revision_id])
2722
elif revision_id is not None:
2723
source_ids = self.source.get_ancestry(revision_id)
2724
if source_ids[0] is not None:
2725
raise AssertionError()
2728
source_ids = self.source.all_revision_ids()
2729
# source_ids is the worst possible case we may need to pull.
2730
# now we want to filter source_ids against what we actually
2731
# have in target, but don't try to check for existence where we know
2732
# we do not have a revision as that would be pointless.
2733
target_ids = set(self.target.all_revision_ids())
2734
result_set = set(source_ids).difference(target_ids)
2735
return self.source.revision_ids_to_search_result(result_set)
2738
class InterModel1and2(InterRepository):
2741
def _get_repo_format_to_test(self):
2745
def is_compatible(source, target):
2746
if not source.supports_rich_root() and target.supports_rich_root():
2752
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2753
"""See InterRepository.fetch()."""
2754
from bzrlib.fetch import Model1toKnit2Fetcher
2755
f = Model1toKnit2Fetcher(to_repository=self.target,
2756
from_repository=self.source,
2757
last_revision=revision_id,
2758
pb=pb, find_ghosts=find_ghosts)
2759
return f.count_copied, f.failed_revisions
2762
def copy_content(self, revision_id=None):
2763
"""Make a complete copy of the content in self into destination.
2765
This is a destructive operation! Do not use it on existing
2768
:param revision_id: Only copy the content needed to construct
2769
revision_id and its parents.
2772
self.target.set_make_working_trees(self.source.make_working_trees())
2773
except NotImplementedError:
2775
# but don't bother fetching if we have the needed data now.
2776
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
2777
self.target.has_revision(revision_id)):
2779
self.target.fetch(self.source, revision_id=revision_id)
2782
class InterKnit1and2(InterKnitRepo):
2785
def _get_repo_format_to_test(self):
2789
def is_compatible(source, target):
2790
"""Be compatible with Knit1 source and Knit3 target"""
2791
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit3
2793
from bzrlib.repofmt.knitrepo import (RepositoryFormatKnit1,
2794
RepositoryFormatKnit3)
2795
from bzrlib.repofmt.pack_repo import (
2796
RepositoryFormatKnitPack1,
2797
RepositoryFormatKnitPack3,
2798
RepositoryFormatPackDevelopment0,
2799
RepositoryFormatPackDevelopment0Subtree,
2802
RepositoryFormatKnit1,
2803
RepositoryFormatKnitPack1,
2804
RepositoryFormatPackDevelopment0,
2807
RepositoryFormatKnit3,
2808
RepositoryFormatKnitPack3,
2809
RepositoryFormatPackDevelopment0Subtree,
2811
return (isinstance(source._format, nosubtrees) and
2812
isinstance(target._format, subtrees))
2813
except AttributeError:
2817
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2818
"""See InterRepository.fetch()."""
2819
from bzrlib.fetch import Knit1to2Fetcher
2820
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2821
self.source, self.source._format, self.target,
2822
self.target._format)
2823
f = Knit1to2Fetcher(to_repository=self.target,
2824
from_repository=self.source,
2825
last_revision=revision_id,
2826
pb=pb, find_ghosts=find_ghosts)
2827
return f.count_copied, f.failed_revisions
2830
class InterDifferingSerializer(InterKnitRepo):
2833
def _get_repo_format_to_test(self):
2837
def is_compatible(source, target):
2838
"""Be compatible with Knit2 source and Knit3 target"""
2839
if source.supports_rich_root() != target.supports_rich_root():
2841
# Ideally, we'd support fetching if the source had no tree references
2842
# even if it supported them...
2843
if (getattr(source, '_format.supports_tree_reference', False) and
2844
not getattr(target, '_format.supports_tree_reference', False)):
2849
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2850
"""See InterRepository.fetch()."""
2851
revision_ids = self.target.search_missing_revision_ids(self.source,
2852
revision_id, find_ghosts=find_ghosts).get_keys()
2853
revision_ids = tsort.topo_sort(
2854
self.source.get_graph().get_parent_map(revision_ids))
2855
def revisions_iterator():
2856
for current_revision_id in revision_ids:
2857
revision = self.source.get_revision(current_revision_id)
2858
tree = self.source.revision_tree(current_revision_id)
2860
signature = self.source.get_signature_text(
2861
current_revision_id)
2862
except errors.NoSuchRevision:
2864
yield revision, tree, signature
2866
my_pb = ui.ui_factory.nested_progress_bar()
2871
install_revisions(self.target, revisions_iterator(),
2872
len(revision_ids), pb)
2874
if my_pb is not None:
2876
return len(revision_ids), 0
2879
class InterOtherToRemote(InterRepository):
2881
def __init__(self, source, target):
2882
InterRepository.__init__(self, source, target)
2883
self._real_inter = None
2886
def is_compatible(source, target):
2887
if isinstance(target, remote.RemoteRepository):
2891
def _ensure_real_inter(self):
2892
if self._real_inter is None:
2893
self.target._ensure_real()
2894
real_target = self.target._real_repository
2895
self._real_inter = InterRepository.get(self.source, real_target)
2897
def copy_content(self, revision_id=None):
2898
self._ensure_real_inter()
2899
self._real_inter.copy_content(revision_id=revision_id)
2901
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2902
self._ensure_real_inter()
2903
return self._real_inter.fetch(revision_id=revision_id, pb=pb,
2904
find_ghosts=find_ghosts)
2907
def _get_repo_format_to_test(self):
2911
InterRepository.register_optimiser(InterDifferingSerializer)
2912
InterRepository.register_optimiser(InterSameDataRepository)
2913
InterRepository.register_optimiser(InterWeaveRepo)
2914
InterRepository.register_optimiser(InterKnitRepo)
2915
InterRepository.register_optimiser(InterModel1and2)
2916
InterRepository.register_optimiser(InterKnit1and2)
2917
InterRepository.register_optimiser(InterPackRepo)
2918
InterRepository.register_optimiser(InterOtherToRemote)
1712
2921
class CopyConverter(object):
1713
2922
"""A repository conversion tool which just performs a copy of the content.
1715
2924
This is slow but quite reliable.
added
removed
bzrlib/repository.py
