202
189
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,
191
def _heads(self, file_id, revision_ids):
192
"""Calculate the graph heads for revision_ids in the graph of file_id.
194
This can use either a per-file graph or a global revision graph as we
195
have an identity relationship between the two graphs.
197
return self.__heads(revision_ids)
199
def _check_root(self, ie, parent_invs, tree):
200
"""Helper for record_entry_contents.
202
:param ie: An entry being added.
203
:param parent_invs: The inventories of the parent revisions of the
205
:param tree: The tree that is being committed.
207
# In this revision format, root entries have no knit or weave When
208
# serializing out to disk and back in root.revision is always
210
ie.revision = self._new_revision_id
212
def _get_delta(self, ie, basis_inv, path):
213
"""Get a delta against the basis inventory for ie."""
214
if ie.file_id not in basis_inv:
216
return (None, path, ie.file_id, ie)
217
elif ie != basis_inv[ie.file_id]:
219
# TODO: avoid tis id2path call.
220
return (basis_inv.id2path(ie.file_id), path, ie.file_id, ie)
225
def record_entry_contents(self, ie, parent_invs, path, tree,
227
"""Record the content of ie from tree into the commit if needed.
229
Side effect: sets ie.revision when unchanged
231
:param ie: An inventory entry present in the commit.
232
:param parent_invs: The inventories of the parent revisions of the
234
:param path: The path the entry is at in the tree.
235
:param tree: The tree which contains this entry and should be used to
237
:param content_summary: Summary data from the tree about the paths
238
content - stat, length, exec, sha/link target. This is only
239
accessed when the entry has a revision of None - that is when it is
240
a candidate to commit.
241
:return: A tuple (change_delta, version_recorded). change_delta is
242
an inventory_delta change for this entry against the basis tree of
243
the commit, or None if no change occured against the basis tree.
244
version_recorded is True if a new version of the entry has been
245
recorded. For instance, committing a merge where a file was only
246
changed on the other side will return (delta, False).
248
if self.new_inventory.root is None:
249
if ie.parent_id is not None:
250
raise errors.RootMissing()
251
self._check_root(ie, parent_invs, tree)
252
if ie.revision is None:
253
kind = content_summary[0]
255
# ie is carried over from a prior commit
257
# XXX: repository specific check for nested tree support goes here - if
258
# the repo doesn't want nested trees we skip it ?
259
if (kind == 'tree-reference' and
260
not self.repository._format.supports_tree_reference):
261
# mismatch between commit builder logic and repository:
262
# this needs the entry creation pushed down into the builder.
263
raise NotImplementedError('Missing repository subtree support.')
264
self.new_inventory.add(ie)
266
# TODO: slow, take it out of the inner loop.
268
basis_inv = parent_invs[0]
270
basis_inv = Inventory(root_id=None)
272
# ie.revision is always None if the InventoryEntry is considered
273
# for committing. We may record the previous parents revision if the
274
# content is actually unchanged against a sole head.
275
if ie.revision is not None:
276
if not self._versioned_root and path == '':
277
# repositories that do not version the root set the root's
278
# revision to the new commit even when no change occurs, and
279
# this masks when a change may have occurred against the basis,
280
# so calculate if one happened.
281
if ie.file_id in basis_inv:
282
delta = (basis_inv.id2path(ie.file_id), path,
286
delta = (None, path, ie.file_id, ie)
289
# we don't need to commit this, because the caller already
290
# determined that an existing revision of this file is
292
return None, (ie.revision == self._new_revision_id)
293
# XXX: Friction: parent_candidates should return a list not a dict
294
# so that we don't have to walk the inventories again.
295
parent_candiate_entries = ie.parent_candidates(parent_invs)
296
head_set = self._heads(ie.file_id, parent_candiate_entries.keys())
298
for inv in parent_invs:
299
if ie.file_id in inv:
300
old_rev = inv[ie.file_id].revision
301
if old_rev in head_set:
302
heads.append(inv[ie.file_id].revision)
303
head_set.remove(inv[ie.file_id].revision)
306
# now we check to see if we need to write a new record to the
308
# We write a new entry unless there is one head to the ancestors, and
309
# the kind-derived content is unchanged.
311
# Cheapest check first: no ancestors, or more the one head in the
312
# ancestors, we write a new node.
316
# There is a single head, look it up for comparison
317
parent_entry = parent_candiate_entries[heads[0]]
318
# if the non-content specific data has changed, we'll be writing a
320
if (parent_entry.parent_id != ie.parent_id or
321
parent_entry.name != ie.name):
323
# now we need to do content specific checks:
325
# if the kind changed the content obviously has
326
if kind != parent_entry.kind:
329
if content_summary[2] is None:
330
raise ValueError("Files must not have executable = None")
332
if (# if the file length changed we have to store:
333
parent_entry.text_size != content_summary[1] or
334
# if the exec bit has changed we have to store:
335
parent_entry.executable != content_summary[2]):
337
elif parent_entry.text_sha1 == content_summary[3]:
338
# all meta and content is unchanged (using a hash cache
339
# hit to check the sha)
340
ie.revision = parent_entry.revision
341
ie.text_size = parent_entry.text_size
342
ie.text_sha1 = parent_entry.text_sha1
343
ie.executable = parent_entry.executable
344
return self._get_delta(ie, basis_inv, path), False
346
# Either there is only a hash change(no hash cache entry,
347
# or same size content change), or there is no change on
349
# Provide the parent's hash to the store layer, so that the
350
# content is unchanged we will not store a new node.
351
nostore_sha = parent_entry.text_sha1
353
# We want to record a new node regardless of the presence or
354
# absence of a content change in the file.
356
ie.executable = content_summary[2]
357
lines = tree.get_file(ie.file_id, path).readlines()
359
ie.text_sha1, ie.text_size = self._add_text_to_weave(
360
ie.file_id, lines, heads, nostore_sha)
361
except errors.ExistingContent:
362
# Turns out that the file content was unchanged, and we were
363
# only going to store a new node if it was changed. Carry over
365
ie.revision = parent_entry.revision
366
ie.text_size = parent_entry.text_size
367
ie.text_sha1 = parent_entry.text_sha1
368
ie.executable = parent_entry.executable
369
return self._get_delta(ie, basis_inv, path), False
370
elif kind == 'directory':
372
# all data is meta here, nothing specific to directory, so
374
ie.revision = parent_entry.revision
375
return self._get_delta(ie, basis_inv, path), False
377
self._add_text_to_weave(ie.file_id, lines, heads, None)
378
elif kind == 'symlink':
379
current_link_target = content_summary[3]
381
# symlink target is not generic metadata, check if it has
383
if current_link_target != parent_entry.symlink_target:
386
# unchanged, carry over.
387
ie.revision = parent_entry.revision
388
ie.symlink_target = parent_entry.symlink_target
389
return self._get_delta(ie, basis_inv, path), False
390
ie.symlink_target = current_link_target
392
self._add_text_to_weave(ie.file_id, lines, heads, None)
393
elif kind == 'tree-reference':
395
if content_summary[3] != parent_entry.reference_revision:
398
# unchanged, carry over.
399
ie.reference_revision = parent_entry.reference_revision
400
ie.revision = parent_entry.revision
401
return self._get_delta(ie, basis_inv, path), False
402
ie.reference_revision = content_summary[3]
404
self._add_text_to_weave(ie.file_id, lines, heads, None)
406
raise NotImplementedError('unknown kind')
407
ie.revision = self._new_revision_id
408
return self._get_delta(ie, basis_inv, path), True
410
def _add_text_to_weave(self, file_id, new_lines, parents, nostore_sha):
411
versionedfile = self.repository.weave_store.get_weave_or_empty(
412
file_id, self.repository.get_transaction())
413
# Don't change this to add_lines - add_lines_with_ghosts is cheaper
414
# than add_lines, and allows committing when a parent is ghosted for
416
# Note: as we read the content directly from the tree, we know its not
417
# been turned into unicode or badly split - but a broken tree
418
# implementation could give us bad output from readlines() so this is
419
# not a guarantee of safety. What would be better is always checking
420
# the content during test suite execution. RBC 20070912
421
return versionedfile.add_lines_with_ghosts(
422
self._new_revision_id, parents, 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 Stores and a Transport, which respectively
454
describe the disk data format and the way of accessing the (possibly
457
:ivar _transport: Transport for file access to repository, typically
458
pointing to .bzr/repository.
262
def abort_write_group(self, suppress_errors=False):
461
# What class to use for a CommitBuilder. Often its simpler to change this
462
# in a Repository class subclass rather than to override
463
# get_commit_builder.
464
_commit_builder_class = CommitBuilder
465
# The search regex used by xml based repositories to determine what things
466
# where changed in a single commit.
467
_file_ids_altered_regex = lazy_regex.lazy_compile(
468
r'file_id="(?P<file_id>[^"]+)"'
469
r'.* revision="(?P<revision_id>[^"]+)"'
472
def abort_write_group(self):
263
473
"""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
475
:seealso: start_write_group.
271
477
if self._write_group is not self.get_transaction():
272
478
# 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)
479
raise errors.BzrError('mismatched lock context and write group.')
480
self._abort_write_group()
290
481
self._write_group = None
292
483
def _abort_write_group(self):
293
484
"""Template method for per-repository write group cleanup.
295
This is called during abort before the write group is considered to be
486
This is called during abort before the write group is considered to be
296
487
finished and should cleanup any internal state accrued during the write
297
488
group. There is no requirement that data handed to the repository be
298
489
*not* made available - this is not a rollback - but neither should any
828
1070
raise NotImplementedError(self.has_revisions)
1072
return self._revision_store.has_revision_id(revision_id,
1073
self.get_transaction())
830
1075
@needs_read_lock
831
1076
def get_revision(self, revision_id):
832
1077
"""Return the Revision object for a named revision."""
833
1078
return self.get_revisions([revision_id])[0]
835
1081
def get_revision_reconcile(self, revision_id):
836
1082
"""'reconcile' helper routine that allows access to a revision always.
838
1084
This variant of get_revision does not cross check the weave graph
839
1085
against the revision one as get_revision does: but it should only
840
1086
be used by reconcile, or reconcile-alike commands that are correcting
841
1087
or testing the revision graph.
843
raise NotImplementedError(self.get_revision_reconcile)
1089
return self._get_revisions([revision_id])[0]
845
1092
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):
1093
"""Get many revisions at once."""
1094
return self._get_revisions(revision_ids)
1097
def _get_revisions(self, revision_ids):
1098
"""Core work logic to get many revisions without sanity checks."""
1099
for rev_id in revision_ids:
1100
if not rev_id or not isinstance(rev_id, basestring):
1101
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1102
revs = self._revision_store.get_revisions(revision_ids,
1103
self.get_transaction())
1107
def get_revision_xml(self, revision_id):
1108
# TODO: jam 20070210 This shouldn't be necessary since get_revision
1109
# would have already do it.
1110
# TODO: jam 20070210 Just use _serializer.write_revision_to_string()
1111
rev = self.get_revision(revision_id)
1112
rev_tmp = StringIO()
1113
# the current serializer..
1114
self._revision_store._serializer.write_revision(rev, rev_tmp)
1116
return rev_tmp.getvalue()
1118
def get_deltas_for_revisions(self, revisions):
854
1119
"""Produce a generator of revision deltas.
856
1121
Note that the input is a sequence of REVISIONS, not revision_ids.
857
1122
Trees will be held in memory until the generator exits.
858
1123
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
1125
required_trees = set()
866
1126
for revision in revisions:
867
1127
required_trees.add(revision.revision_id)
868
1128
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
1129
trees = dict((t.get_revision_id(), t) for
1130
t in self.revision_trees(required_trees))
885
1131
for revision in revisions:
886
1132
if not revision.parent_ids:
887
old_tree = self.revision_tree(_mod_revision.NULL_REVISION)
1133
old_tree = self.revision_tree(None)
889
1135
old_tree = trees[revision.parent_ids[0]]
890
1136
yield trees[revision.revision_id].changes_from(old_tree)
892
1138
@needs_read_lock
893
def get_revision_delta(self, revision_id, specific_fileids=None):
1139
def get_revision_delta(self, revision_id):
894
1140
"""Return the delta for one revision.
896
1142
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
1145
r = self.get_revision(revision_id)
904
return list(self.get_deltas_for_revisions([r],
905
specific_fileids=specific_fileids))[0]
1146
return list(self.get_deltas_for_revisions([r]))[0]
907
1148
@needs_write_lock
908
1149
def store_revision_signature(self, gpg_strategy, plaintext, revision_id):
909
1150
signature = gpg_strategy.sign(plaintext)
910
1151
self.add_signature_text(revision_id, signature)
912
1154
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)
1155
self._revision_store.add_revision_signature_text(revision_id,
1157
self.get_transaction())
1159
def find_text_key_references(self):
1160
"""Find the text key references within the repository.
1162
:return: a dictionary mapping (file_id, revision_id) tuples to altered file-ids to an iterable of
1163
revision_ids. Each altered file-ids has the exact revision_ids that
1164
altered it listed explicitly.
1165
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1166
to whether they were referred to by the inventory of the
1167
revision_id that they contain. The inventory texts from all present
1168
revision ids are assessed to generate this report.
1170
revision_ids = self.all_revision_ids()
1171
w = self.get_inventory_weave()
1172
pb = ui.ui_factory.nested_progress_bar()
1174
return self._find_text_key_references_from_xml_inventory_lines(
1175
w.iter_lines_added_or_present_in_versions(revision_ids, pb=pb))
1179
def _find_text_key_references_from_xml_inventory_lines(self,
1181
"""Core routine for extracting references to texts from inventories.
1183
This performs the translation of xml lines to revision ids.
1185
:param line_iterator: An iterator of lines, origin_version_id
1186
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1187
to whether they were referred to by the inventory of the
1188
revision_id that they contain. Note that if that revision_id was
1189
not part of the line_iterator's output then False will be given -
1190
even though it may actually refer to that key.
1192
if not self._serializer.support_altered_by_hack:
1193
raise AssertionError(
1194
"_find_text_key_references_from_xml_inventory_lines only "
1195
"supported for branches which store inventory as unnested xml"
1196
", not on %r" % self)
1199
# this code needs to read every new line in every inventory for the
1200
# inventories [revision_ids]. Seeing a line twice is ok. Seeing a line
1201
# not present in one of those inventories is unnecessary but not
1202
# harmful because we are filtering by the revision id marker in the
1203
# inventory lines : we only select file ids altered in one of those
1204
# revisions. We don't need to see all lines in the inventory because
1205
# only those added in an inventory in rev X can contain a revision=X
1207
unescape_revid_cache = {}
1208
unescape_fileid_cache = {}
1210
# jam 20061218 In a big fetch, this handles hundreds of thousands
1211
# of lines, so it has had a lot of inlining and optimizing done.
1212
# Sorry that it is a little bit messy.
1213
# Move several functions to be local variables, since this is a long
1215
search = self._file_ids_altered_regex.search
1216
unescape = _unescape_xml
1217
setdefault = result.setdefault
1218
for line, version_id in line_iterator:
1219
match = search(line)
1222
# One call to match.group() returning multiple items is quite a
1223
# bit faster than 2 calls to match.group() each returning 1
1224
file_id, revision_id = match.group('file_id', 'revision_id')
1226
# Inlining the cache lookups helps a lot when you make 170,000
1227
# lines and 350k ids, versus 8.4 unique ids.
1228
# Using a cache helps in 2 ways:
1229
# 1) Avoids unnecessary decoding calls
1230
# 2) Re-uses cached strings, which helps in future set and
1232
# (2) is enough that removing encoding entirely along with
1233
# the cache (so we are using plain strings) results in no
1234
# performance improvement.
1236
revision_id = unescape_revid_cache[revision_id]
1238
unescaped = unescape(revision_id)
1239
unescape_revid_cache[revision_id] = unescaped
1240
revision_id = unescaped
1242
# Note that unconditionally unescaping means that we deserialise
1243
# every fileid, which for general 'pull' is not great, but we don't
1244
# really want to have some many fulltexts that this matters anyway.
1247
file_id = unescape_fileid_cache[file_id]
1249
unescaped = unescape(file_id)
1250
unescape_fileid_cache[file_id] = unescaped
1253
key = (file_id, revision_id)
1254
setdefault(key, False)
1255
if revision_id == version_id:
1259
def _find_file_ids_from_xml_inventory_lines(self, line_iterator,
1261
"""Helper routine for fileids_altered_by_revision_ids.
1263
This performs the translation of xml lines to revision ids.
1265
:param line_iterator: An iterator of lines, origin_version_id
1266
:param revision_ids: The revision ids to filter for. This should be a
1267
set or other type which supports efficient __contains__ lookups, as
1268
the revision id from each parsed line will be looked up in the
1269
revision_ids filter.
1270
:return: a dictionary mapping altered file-ids to an iterable of
1271
revision_ids. Each altered file-ids has the exact revision_ids that
1272
altered it listed explicitly.
1275
setdefault = result.setdefault
1276
for file_id, revision_id in \
1277
self._find_text_key_references_from_xml_inventory_lines(
1278
line_iterator).iterkeys():
1279
# once data is all ensured-consistent; then this is
1280
# if revision_id == version_id
1281
if revision_id in revision_ids:
1282
setdefault(file_id, set()).add(revision_id)
1285
def fileids_altered_by_revision_ids(self, revision_ids, _inv_weave=None):
1286
"""Find the file ids and versions affected by revisions.
1288
:param revisions: an iterable containing revision ids.
1289
:param _inv_weave: The inventory weave from this repository or None.
1290
If None, the inventory weave will be opened automatically.
1291
:return: a dictionary mapping altered file-ids to an iterable of
1292
revision_ids. Each altered file-ids has the exact revision_ids that
1293
altered it listed explicitly.
1295
selected_revision_ids = set(revision_ids)
1296
w = _inv_weave or self.get_inventory_weave()
1297
pb = ui.ui_factory.nested_progress_bar()
1299
return self._find_file_ids_from_xml_inventory_lines(
1300
w.iter_lines_added_or_present_in_versions(
1301
selected_revision_ids, pb=pb),
1302
selected_revision_ids)
933
1306
def iter_files_bytes(self, desired_files):
934
1307
"""Iterate through file versions.
941
1314
uniquely identify the file version in the caller's context. (Examples:
942
1315
an index number or a TreeTransform trans_id.)
1317
bytes_iterator is an iterable of bytestrings for the file. The
1318
kind of iterable and length of the bytestrings are unspecified, but for
1319
this implementation, it is a list of lines produced by
1320
VersionedFile.get_lines().
944
1322
: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])
1325
transaction = self.get_transaction()
1326
for file_id, revision_id, callable_data in desired_files:
1328
weave = self.weave_store.get_weave(file_id, transaction)
1329
except errors.NoSuchFile:
1330
raise errors.NoSuchIdInRepository(self, file_id)
1331
yield callable_data, weave.get_lines(revision_id)
1333
def _generate_text_key_index(self, text_key_references=None,
1335
"""Generate a new text key index for the repository.
1337
This is an expensive function that will take considerable time to run.
1339
:return: A dict mapping text keys ((file_id, revision_id) tuples) to a
1340
list of parents, also text keys. When a given key has no parents,
1341
the parents list will be [NULL_REVISION].
1343
# All revisions, to find inventory parents.
1344
if ancestors is None:
1345
graph = self.get_graph()
1346
ancestors = graph.get_parent_map(self.all_revision_ids())
1347
if text_key_references is None:
1348
text_key_references = self.find_text_key_references()
1349
pb = ui.ui_factory.nested_progress_bar()
1351
return self._do_generate_text_key_index(ancestors,
1352
text_key_references, pb)
1356
def _do_generate_text_key_index(self, ancestors, text_key_references, pb):
1357
"""Helper for _generate_text_key_index to avoid deep nesting."""
1358
revision_order = tsort.topo_sort(ancestors)
1359
invalid_keys = set()
1361
for revision_id in revision_order:
1362
revision_keys[revision_id] = set()
1363
text_count = len(text_key_references)
1364
# a cache of the text keys to allow reuse; costs a dict of all the
1365
# keys, but saves a 2-tuple for every child of a given key.
1367
for text_key, valid in text_key_references.iteritems():
1369
invalid_keys.add(text_key)
1371
revision_keys[text_key[1]].add(text_key)
1372
text_key_cache[text_key] = text_key
1373
del text_key_references
1375
text_graph = graph.Graph(graph.DictParentsProvider(text_index))
1376
NULL_REVISION = _mod_revision.NULL_REVISION
1377
# Set a cache with a size of 10 - this suffices for bzr.dev but may be
1378
# too small for large or very branchy trees. However, for 55K path
1379
# trees, it would be easy to use too much memory trivially. Ideally we
1380
# could gauge this by looking at available real memory etc, but this is
1381
# always a tricky proposition.
1382
inventory_cache = lru_cache.LRUCache(10)
1383
batch_size = 10 # should be ~150MB on a 55K path tree
1384
batch_count = len(revision_order) / batch_size + 1
1386
pb.update("Calculating text parents.", processed_texts, text_count)
1387
for offset in xrange(batch_count):
1388
to_query = revision_order[offset * batch_size:(offset + 1) *
1392
for rev_tree in self.revision_trees(to_query):
1393
revision_id = rev_tree.get_revision_id()
1394
parent_ids = ancestors[revision_id]
1395
for text_key in revision_keys[revision_id]:
1396
pb.update("Calculating text parents.", processed_texts)
1397
processed_texts += 1
1398
candidate_parents = []
1399
for parent_id in parent_ids:
1400
parent_text_key = (text_key[0], parent_id)
1402
check_parent = parent_text_key not in \
1403
revision_keys[parent_id]
1405
# the parent parent_id is a ghost:
1406
check_parent = False
1407
# truncate the derived graph against this ghost.
1408
parent_text_key = None
1410
# look at the parent commit details inventories to
1411
# determine possible candidates in the per file graph.
1414
inv = inventory_cache[parent_id]
1416
inv = self.revision_tree(parent_id).inventory
1417
inventory_cache[parent_id] = inv
1418
parent_entry = inv._byid.get(text_key[0], None)
1419
if parent_entry is not None:
1421
text_key[0], parent_entry.revision)
1423
parent_text_key = None
1424
if parent_text_key is not None:
1425
candidate_parents.append(
1426
text_key_cache[parent_text_key])
1427
parent_heads = text_graph.heads(candidate_parents)
1428
new_parents = list(parent_heads)
1429
new_parents.sort(key=lambda x:candidate_parents.index(x))
1430
if new_parents == []:
1431
new_parents = [NULL_REVISION]
1432
text_index[text_key] = new_parents
1434
for text_key in invalid_keys:
1435
text_index[text_key] = [NULL_REVISION]
1438
def item_keys_introduced_by(self, revision_ids, _files_pb=None):
1439
"""Get an iterable listing the keys of all the data introduced by a set
1442
The keys will be ordered so that the corresponding items can be safely
1443
fetched and inserted in that order.
1445
:returns: An iterable producing tuples of (knit-kind, file-id,
1446
versions). knit-kind is one of 'file', 'inventory', 'signatures',
1447
'revisions'. file-id is None unless knit-kind is 'file'.
1449
# XXX: it's a bit weird to control the inventory weave caching in this
1450
# generator. Ideally the caching would be done in fetch.py I think. Or
1451
# maybe this generator should explicitly have the contract that it
1452
# should not be iterated until the previously yielded item has been
1454
inv_w = self.get_inventory_weave()
1456
# file ids that changed
1457
file_ids = self.fileids_altered_by_revision_ids(revision_ids, inv_w)
1459
num_file_ids = len(file_ids)
1460
for file_id, altered_versions in file_ids.iteritems():
1461
if _files_pb is not None:
1462
_files_pb.update("fetch texts", count, num_file_ids)
1464
yield ("file", file_id, altered_versions)
1465
# We're done with the files_pb. Note that it finished by the caller,
1466
# just as it was created by the caller.
1470
yield ("inventory", None, revision_ids)
1473
revisions_with_signatures = set()
1474
for rev_id in revision_ids:
1476
self.get_signature_text(rev_id)
1477
except errors.NoSuchRevision:
1481
revisions_with_signatures.add(rev_id)
1482
yield ("signatures", None, revisions_with_signatures)
1485
yield ("revisions", None, revision_ids)
1488
def get_inventory_weave(self):
1489
return self.control_weaves.get_weave('inventory',
1490
self.get_transaction())
1493
def get_inventory(self, revision_id):
1494
"""Get Inventory object by revision id."""
1495
return self.iter_inventories([revision_id]).next()
1497
def iter_inventories(self, revision_ids):
1498
"""Get many inventories by revision_ids.
1500
This will buffer some or all of the texts used in constructing the
1501
inventories in memory, but will only parse a single inventory at a
1504
:return: An iterator of inventories.
1506
if ((None in revision_ids)
1507
or (_mod_revision.NULL_REVISION in revision_ids)):
1508
raise ValueError('cannot get null revision inventory')
1509
return self._iter_inventories(revision_ids)
1511
def _iter_inventories(self, revision_ids):
1512
"""single-document based inventory iteration."""
1513
texts = self.get_inventory_weave().get_texts(revision_ids)
1514
for text, revision_id in zip(texts, revision_ids):
1515
yield self.deserialise_inventory(revision_id, text)
1517
def deserialise_inventory(self, revision_id, xml):
1518
"""Transform the xml into an inventory object.
1520
:param revision_id: The expected revision id of the inventory.
1521
:param xml: A serialised inventory.
1523
result = self._serializer.read_inventory_from_string(xml, revision_id)
1524
if result.revision_id != revision_id:
1525
raise AssertionError('revision id mismatch %s != %s' % (
1526
result.revision_id, revision_id))
1529
def serialise_inventory(self, inv):
1530
return self._serializer.write_inventory_to_string(inv)
1532
def _serialise_inventory_to_lines(self, inv):
1533
return self._serializer.write_inventory_to_lines(inv)
1535
def get_serializer_format(self):
1536
return self._serializer.format_num
1539
def get_inventory_xml(self, revision_id):
1540
"""Get inventory XML as a file object."""
1542
iw = self.get_inventory_weave()
1543
return iw.get_text(revision_id)
1545
raise errors.HistoryMissing(self, 'inventory', revision_id)
1548
def get_inventory_sha1(self, revision_id):
1549
"""Return the sha1 hash of the inventory entry
1551
return self.get_revision(revision_id).inventory_sha1
1553
def iter_reverse_revision_history(self, revision_id):
1554
"""Iterate backwards through revision ids in the lefthand history
1556
:param revision_id: The revision id to start with. All its lefthand
1557
ancestors will be traversed.
1559
graph = self.get_graph()
1560
next_id = revision_id
1562
if next_id in (None, _mod_revision.NULL_REVISION):
1565
# Note: The following line may raise KeyError in the event of
1566
# truncated history. We decided not to have a try:except:raise
1567
# RevisionNotPresent here until we see a use for it, because of the
1568
# cost in an inner loop that is by its very nature O(history).
1569
# Robert Collins 20080326
1570
parents = graph.get_parent_map([next_id])[next_id]
1571
if len(parents) == 0:
1574
next_id = parents[0]
1577
def get_revision_inventory(self, revision_id):
1578
"""Return inventory of a past revision."""
1579
# TODO: Unify this with get_inventory()
1580
# bzr 0.0.6 and later imposes the constraint that the inventory_id
1581
# must be the same as its revision, so this is trivial.
1582
if revision_id is None:
1583
# This does not make sense: if there is no revision,
1584
# then it is the current tree inventory surely ?!
1585
# and thus get_root_id() is something that looks at the last
1586
# commit on the branch, and the get_root_id is an inventory check.
1587
raise NotImplementedError
1588
# return Inventory(self.get_root_id())
1590
return self.get_inventory(revision_id)
984
1593
def is_shared(self):
985
1594
"""Return True if this repository is flagged as a shared repository."""
986
1595
raise NotImplementedError(self.is_shared)
1000
1609
for repositories to maintain loaded indices across multiple locks
1001
1610
by checking inside their implementation of this method to see
1002
1611
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
1612
the disk format being validatable in this manner.
1010
1615
@needs_read_lock
1011
1616
def revision_tree(self, revision_id):
1012
1617
"""Return Tree for a revision on this branch.
1014
`revision_id` may be NULL_REVISION for the empty tree revision.
1619
`revision_id` may be None for the empty tree revision.
1016
raise NotImplementedError(self.revision_tree)
1621
# TODO: refactor this to use an existing revision object
1622
# so we don't need to read it in twice.
1623
if revision_id is None or revision_id == _mod_revision.NULL_REVISION:
1624
return RevisionTree(self, Inventory(root_id=None),
1625
_mod_revision.NULL_REVISION)
1627
inv = self.get_revision_inventory(revision_id)
1628
return RevisionTree(self, inv, revision_id)
1018
1630
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:'
1631
"""Return Tree for a revision on this branch.
1633
`revision_id` may not be None or 'null:'"""
1634
inventories = self.iter_inventories(revision_ids)
1635
for inv in inventories:
1636
yield RevisionTree(self, inv, inv.revision_id)
1639
def get_ancestry(self, revision_id, topo_sorted=True):
1640
"""Return a list of revision-ids integrated by a revision.
1642
The first element of the list is always None, indicating the origin
1643
revision. This might change when we have history horizons, or
1644
perhaps we should have a new API.
1646
This is topologically sorted.
1024
raise NotImplementedError(self.revision_trees)
1648
if _mod_revision.is_null(revision_id):
1650
if not self.has_revision(revision_id):
1651
raise errors.NoSuchRevision(self, revision_id)
1652
w = self.get_inventory_weave()
1653
candidates = w.get_ancestry(revision_id, topo_sorted)
1654
return [None] + candidates # self._eliminate_revisions_not_present(candidates)
1026
def pack(self, hint=None, clean_obsolete_packs=False):
1027
1657
"""Compress the data within the repository.
1029
1659
This operation only makes sense for some repository types. For other
1030
1660
types it should be a no-op that just returns.
1032
1662
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
1663
@needs_write_lock as this is a long running call its reasonable to
1034
1664
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
1668
def print_file(self, file, revision_id):
1669
"""Print `file` to stdout.
1671
FIXME RBC 20060125 as John Meinel points out this is a bad api
1672
- it writes to stdout, it assumes that that is valid etc. Fix
1673
by creating a new more flexible convenience function.
1675
tree = self.revision_tree(revision_id)
1676
# use inventory as it was in that revision
1677
file_id = tree.inventory.path2id(file)
1679
# TODO: jam 20060427 Write a test for this code path
1680
# it had a bug in it, and was raising the wrong
1682
raise errors.BzrError("%r is not present in revision %s" % (file, revision_id))
1683
tree.print_file(file_id)
1047
1685
def get_transaction(self):
1048
1686
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)
1688
@deprecated_method(symbol_versioning.one_one)
1689
def get_parents(self, revision_ids):
1690
"""See StackedParentsProvider.get_parents"""
1691
parent_map = self.get_parent_map(revision_ids)
1692
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:
1694
def get_parent_map(self, keys):
1695
"""See graph._StackedParentsProvider.get_parent_map"""
1697
for revision_id in keys:
1698
if revision_id is None:
1699
raise ValueError('get_parent_map(None) is not valid')
1061
1700
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,)
1701
parent_map[revision_id] = ()
1704
parent_id_list = self.get_revision(revision_id).parent_ids
1705
except errors.NoSuchRevision:
1708
if len(parent_id_list) == 0:
1709
parent_ids = (_mod_revision.NULL_REVISION,)
1711
parent_ids = tuple(parent_id_list)
1712
parent_map[revision_id] = parent_ids
1077
1715
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
1718
def get_graph(self, other_repository=None):
1099
1719
"""Return the graph walker for this repository format"""
1100
1720
parents_provider = self._make_parents_provider()
1101
1721
if (other_repository is not None and
1102
1722
not self.has_same_location(other_repository)):
1103
parents_provider = graph.StackedParentsProvider(
1723
parents_provider = graph._StackedParentsProvider(
1104
1724
[parents_provider, other_repository._make_parents_provider()])
1105
1725
return graph.Graph(parents_provider)
1727
def _get_versioned_file_checker(self):
1728
"""Return an object suitable for checking versioned files."""
1729
return _VersionedFileChecker(self)
1731
def revision_ids_to_search_result(self, result_set):
1732
"""Convert a set of revision ids to a graph SearchResult."""
1733
result_parents = set()
1734
for parents in self.get_graph().get_parent_map(
1735
result_set).itervalues():
1736
result_parents.update(parents)
1737
included_keys = result_set.intersection(result_parents)
1738
start_keys = result_set.difference(included_keys)
1739
exclude_keys = result_parents.difference(result_set)
1740
result = graph.SearchResult(start_keys, exclude_keys,
1741
len(result_set), result_set)
1107
1744
@needs_write_lock
1108
1745
def set_make_working_trees(self, new_value):
1109
1746
"""Set the policy flag for making working trees when creating branches.
1632
2309
_optimisers = []
1633
2310
"""The available optimised InterRepository types."""
1636
2312
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):
2313
raise NotImplementedError(self.copy_content)
2315
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
1653
2316
"""Fetch the content required to construct revision_id.
1655
2318
The content is copied from self.source to self.target.
1657
2320
:param revision_id: if None all content is copied, if NULL_REVISION no
1658
2321
content is copied.
2322
:param pb: optional progress bar to use for progress reports. If not
2323
provided a default one will be created.
2325
Returns the copied revision count and the failed revisions in a tuple:
1661
2328
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
2330
def _walk_to_common_revisions(self, revision_ids):
2331
"""Walk out from revision_ids in source to revisions target has.
2333
:param revision_ids: The start point for the search.
2334
:return: A set of revision ids.
2336
target_graph = self.target.get_graph()
2337
revision_ids = frozenset(revision_ids)
2338
if set(target_graph.get_parent_map(revision_ids)) == revision_ids:
2339
return graph.SearchResult(revision_ids, set(), 0, set())
2340
missing_revs = set()
2341
source_graph = self.source.get_graph()
2342
# ensure we don't pay silly lookup costs.
2343
searcher = source_graph._make_breadth_first_searcher(revision_ids)
2344
null_set = frozenset([_mod_revision.NULL_REVISION])
2347
next_revs, ghosts = searcher.next_with_ghosts()
2348
except StopIteration:
2350
if revision_ids.intersection(ghosts):
2351
absent_ids = set(revision_ids.intersection(ghosts))
2352
# If all absent_ids are present in target, no error is needed.
2353
absent_ids.difference_update(
2354
set(target_graph.get_parent_map(absent_ids)))
2356
raise errors.NoSuchRevision(self.source, absent_ids.pop())
2357
# we don't care about other ghosts as we can't fetch them and
2358
# haven't been asked to.
2359
next_revs = set(next_revs)
2360
# we always have NULL_REVISION present.
2361
have_revs = set(target_graph.get_parent_map(next_revs)).union(null_set)
2362
missing_revs.update(next_revs - have_revs)
2363
searcher.stop_searching_any(have_revs)
2364
return searcher.get_result()
2366
@deprecated_method(symbol_versioning.one_two)
2368
def missing_revision_ids(self, revision_id=None, find_ghosts=True):
2369
"""Return the revision ids that source has that target does not.
2371
These are returned in topological order.
2373
:param revision_id: only return revision ids included by this
2375
:param find_ghosts: If True find missing revisions in deep history
2376
rather than just finding the surface difference.
2378
return list(self.search_missing_revision_ids(
2379
revision_id, find_ghosts).get_keys())
2382
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2383
"""Return the revision ids that source has that target does not.
2385
:param revision_id: only return revision ids included by this
2387
:param find_ghosts: If True find missing revisions in deep history
2388
rather than just finding the surface difference.
1683
2389
:return: A bzrlib.graph.SearchResult.
1685
raise NotImplementedError(self.search_missing_revision_ids)
2391
# stop searching at found target revisions.
2392
if not find_ghosts and revision_id is not None:
2393
return self._walk_to_common_revisions([revision_id])
2394
# generic, possibly worst case, slow code path.
2395
target_ids = set(self.target.all_revision_ids())
2396
if revision_id is not None:
2397
source_ids = self.source.get_ancestry(revision_id)
2398
if source_ids[0] is not None:
2399
raise AssertionError()
2402
source_ids = self.source.all_revision_ids()
2403
result_set = set(source_ids).difference(target_ids)
2404
return self.source.revision_ids_to_search_result(result_set)
1688
2407
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:
2408
"""True if source and target have the same data representation."""
2409
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
2411
if source._serializer != target._serializer:
1708
raise errors.IncompatibleRepositories(source, target,
1709
"different serializers")
2416
class InterSameDataRepository(InterRepository):
2417
"""Code for converting between repositories that represent the same data.
2419
Data format and model must match for this to work.
2423
def _get_repo_format_to_test(self):
2424
"""Repository format for testing with.
2426
InterSameData can pull from subtree to subtree and from non-subtree to
2427
non-subtree, so we test this with the richest repository format.
2429
from bzrlib.repofmt import knitrepo
2430
return knitrepo.RepositoryFormatKnit3()
2433
def is_compatible(source, target):
2434
return InterRepository._same_model(source, target)
2437
def copy_content(self, revision_id=None):
2438
"""Make a complete copy of the content in self into destination.
2440
This copies both the repository's revision data, and configuration information
2441
such as the make_working_trees setting.
2443
This is a destructive operation! Do not use it on existing
2446
:param revision_id: Only copy the content needed to construct
2447
revision_id and its parents.
2450
self.target.set_make_working_trees(self.source.make_working_trees())
2451
except NotImplementedError:
2453
# but don't bother fetching if we have the needed data now.
2454
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
2455
self.target.has_revision(revision_id)):
2457
self.target.fetch(self.source, revision_id=revision_id)
2460
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2461
"""See InterRepository.fetch()."""
2462
from bzrlib.fetch import GenericRepoFetcher
2463
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2464
self.source, self.source._format, self.target,
2465
self.target._format)
2466
f = GenericRepoFetcher(to_repository=self.target,
2467
from_repository=self.source,
2468
last_revision=revision_id,
2469
pb=pb, find_ghosts=find_ghosts)
2470
return f.count_copied, f.failed_revisions
2473
class InterWeaveRepo(InterSameDataRepository):
2474
"""Optimised code paths between Weave based repositories.
2476
This should be in bzrlib/repofmt/weaverepo.py but we have not yet
2477
implemented lazy inter-object optimisation.
2481
def _get_repo_format_to_test(self):
2482
from bzrlib.repofmt import weaverepo
2483
return weaverepo.RepositoryFormat7()
2486
def is_compatible(source, target):
2487
"""Be compatible with known Weave formats.
2489
We don't test for the stores being of specific types because that
2490
could lead to confusing results, and there is no need to be
2493
from bzrlib.repofmt.weaverepo import (
2499
return (isinstance(source._format, (RepositoryFormat5,
2501
RepositoryFormat7)) and
2502
isinstance(target._format, (RepositoryFormat5,
2504
RepositoryFormat7)))
2505
except AttributeError:
2509
def copy_content(self, revision_id=None):
2510
"""See InterRepository.copy_content()."""
2511
# weave specific optimised path:
2513
self.target.set_make_working_trees(self.source.make_working_trees())
2514
except (errors.RepositoryUpgradeRequired, NotImplemented):
2516
# FIXME do not peek!
2517
if self.source._transport.listable():
2518
pb = ui.ui_factory.nested_progress_bar()
2520
self.target.weave_store.copy_all_ids(
2521
self.source.weave_store,
2523
from_transaction=self.source.get_transaction(),
2524
to_transaction=self.target.get_transaction())
2525
pb.update('copying inventory', 0, 1)
2526
self.target.control_weaves.copy_multi(
2527
self.source.control_weaves, ['inventory'],
2528
from_transaction=self.source.get_transaction(),
2529
to_transaction=self.target.get_transaction())
2530
self.target._revision_store.text_store.copy_all_ids(
2531
self.source._revision_store.text_store,
2536
self.target.fetch(self.source, revision_id=revision_id)
2539
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2540
"""See InterRepository.fetch()."""
2541
from bzrlib.fetch import GenericRepoFetcher
2542
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2543
self.source, self.source._format, self.target, self.target._format)
2544
f = GenericRepoFetcher(to_repository=self.target,
2545
from_repository=self.source,
2546
last_revision=revision_id,
2547
pb=pb, find_ghosts=find_ghosts)
2548
return f.count_copied, f.failed_revisions
2551
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2552
"""See InterRepository.missing_revision_ids()."""
2553
# we want all revisions to satisfy revision_id in source.
2554
# but we don't want to stat every file here and there.
2555
# we want then, all revisions other needs to satisfy revision_id
2556
# checked, but not those that we have locally.
2557
# so the first thing is to get a subset of the revisions to
2558
# satisfy revision_id in source, and then eliminate those that
2559
# we do already have.
2560
# this is slow on high latency connection to self, but as as this
2561
# disk format scales terribly for push anyway due to rewriting
2562
# inventory.weave, this is considered acceptable.
2564
if revision_id is not None:
2565
source_ids = self.source.get_ancestry(revision_id)
2566
if source_ids[0] is not None:
2567
raise AssertionError()
2570
source_ids = self.source._all_possible_ids()
2571
source_ids_set = set(source_ids)
2572
# source_ids is the worst possible case we may need to pull.
2573
# now we want to filter source_ids against what we actually
2574
# have in target, but don't try to check for existence where we know
2575
# we do not have a revision as that would be pointless.
2576
target_ids = set(self.target._all_possible_ids())
2577
possibly_present_revisions = target_ids.intersection(source_ids_set)
2578
actually_present_revisions = set(
2579
self.target._eliminate_revisions_not_present(possibly_present_revisions))
2580
required_revisions = source_ids_set.difference(actually_present_revisions)
2581
if revision_id is not None:
2582
# we used get_ancestry to determine source_ids then we are assured all
2583
# revisions referenced are present as they are installed in topological order.
2584
# and the tip revision was validated by get_ancestry.
2585
result_set = required_revisions
2587
# if we just grabbed the possibly available ids, then
2588
# we only have an estimate of whats available and need to validate
2589
# that against the revision records.
2591
self.source._eliminate_revisions_not_present(required_revisions))
2592
return self.source.revision_ids_to_search_result(result_set)
2595
class InterKnitRepo(InterSameDataRepository):
2596
"""Optimised code paths between Knit based repositories."""
2599
def _get_repo_format_to_test(self):
2600
from bzrlib.repofmt import knitrepo
2601
return knitrepo.RepositoryFormatKnit1()
2604
def is_compatible(source, target):
2605
"""Be compatible with known Knit formats.
2607
We don't test for the stores being of specific types because that
2608
could lead to confusing results, and there is no need to be
2611
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit
2613
are_knits = (isinstance(source._format, RepositoryFormatKnit) and
2614
isinstance(target._format, RepositoryFormatKnit))
2615
except AttributeError:
2617
return are_knits and InterRepository._same_model(source, target)
2620
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2621
"""See InterRepository.fetch()."""
2622
from bzrlib.fetch import KnitRepoFetcher
2623
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2624
self.source, self.source._format, self.target, self.target._format)
2625
f = KnitRepoFetcher(to_repository=self.target,
2626
from_repository=self.source,
2627
last_revision=revision_id,
2628
pb=pb, find_ghosts=find_ghosts)
2629
return f.count_copied, f.failed_revisions
2632
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2633
"""See InterRepository.missing_revision_ids()."""
2634
if revision_id is not None:
2635
source_ids = self.source.get_ancestry(revision_id)
2636
if source_ids[0] is not None:
2637
raise AssertionError()
2640
source_ids = self.source.all_revision_ids()
2641
source_ids_set = set(source_ids)
2642
# source_ids is the worst possible case we may need to pull.
2643
# now we want to filter source_ids against what we actually
2644
# have in target, but don't try to check for existence where we know
2645
# we do not have a revision as that would be pointless.
2646
target_ids = set(self.target.all_revision_ids())
2647
possibly_present_revisions = target_ids.intersection(source_ids_set)
2648
actually_present_revisions = set(
2649
self.target._eliminate_revisions_not_present(possibly_present_revisions))
2650
required_revisions = source_ids_set.difference(actually_present_revisions)
2651
if revision_id is not None:
2652
# we used get_ancestry to determine source_ids then we are assured all
2653
# revisions referenced are present as they are installed in topological order.
2654
# and the tip revision was validated by get_ancestry.
2655
result_set = required_revisions
2657
# if we just grabbed the possibly available ids, then
2658
# we only have an estimate of whats available and need to validate
2659
# that against the revision records.
2661
self.source._eliminate_revisions_not_present(required_revisions))
2662
return self.source.revision_ids_to_search_result(result_set)
2665
class InterPackRepo(InterSameDataRepository):
2666
"""Optimised code paths between Pack based repositories."""
2669
def _get_repo_format_to_test(self):
2670
from bzrlib.repofmt import pack_repo
2671
return pack_repo.RepositoryFormatKnitPack1()
2674
def is_compatible(source, target):
2675
"""Be compatible with known Pack formats.
2677
We don't test for the stores being of specific types because that
2678
could lead to confusing results, and there is no need to be
2681
from bzrlib.repofmt.pack_repo import RepositoryFormatPack
2683
are_packs = (isinstance(source._format, RepositoryFormatPack) and
2684
isinstance(target._format, RepositoryFormatPack))
2685
except AttributeError:
2687
return are_packs and InterRepository._same_model(source, target)
2690
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2691
"""See InterRepository.fetch()."""
2692
from bzrlib.repofmt.pack_repo import Packer
2693
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2694
self.source, self.source._format, self.target, self.target._format)
2695
self.count_copied = 0
2696
if revision_id is None:
2698
# everything to do - use pack logic
2699
# to fetch from all packs to one without
2700
# inventory parsing etc, IFF nothing to be copied is in the target.
2702
revision_ids = self.source.all_revision_ids()
2703
revision_keys = [(revid,) for revid in revision_ids]
2704
index = self.target._pack_collection.revision_index.combined_index
2705
present_revision_ids = set(item[1][0] for item in
2706
index.iter_entries(revision_keys))
2707
revision_ids = set(revision_ids) - present_revision_ids
2708
# implementing the TODO will involve:
2709
# - detecting when all of a pack is selected
2710
# - avoiding as much as possible pre-selection, so the
2711
# more-core routines such as create_pack_from_packs can filter in
2712
# a just-in-time fashion. (though having a HEADS list on a
2713
# repository might make this a lot easier, because we could
2714
# sensibly detect 'new revisions' without doing a full index scan.
2715
elif _mod_revision.is_null(revision_id):
2720
revision_ids = self.search_missing_revision_ids(revision_id,
2721
find_ghosts=find_ghosts).get_keys()
2722
except errors.NoSuchRevision:
2723
raise errors.InstallFailed([revision_id])
2724
if len(revision_ids) == 0:
2726
packs = self.source._pack_collection.all_packs()
2727
pack = Packer(self.target._pack_collection, packs, '.fetch',
2728
revision_ids).pack()
2729
if pack is not None:
2730
self.target._pack_collection._save_pack_names()
2731
# Trigger an autopack. This may duplicate effort as we've just done
2732
# a pack creation, but for now it is simpler to think about as
2733
# 'upload data, then repack if needed'.
2734
self.target._pack_collection.autopack()
2735
return (pack.get_revision_count(), [])
2740
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2741
"""See InterRepository.missing_revision_ids().
2743
:param find_ghosts: Find ghosts throughout the ancestry of
2746
if not find_ghosts and revision_id is not None:
2747
return self._walk_to_common_revisions([revision_id])
2748
elif revision_id is not None:
2749
source_ids = self.source.get_ancestry(revision_id)
2750
if source_ids[0] is not None:
2751
raise AssertionError()
2754
source_ids = self.source.all_revision_ids()
2755
# source_ids is the worst possible case we may need to pull.
2756
# now we want to filter source_ids against what we actually
2757
# have in target, but don't try to check for existence where we know
2758
# we do not have a revision as that would be pointless.
2759
target_ids = set(self.target.all_revision_ids())
2760
result_set = set(source_ids).difference(target_ids)
2761
return self.source.revision_ids_to_search_result(result_set)
2764
class InterModel1and2(InterRepository):
2767
def _get_repo_format_to_test(self):
2771
def is_compatible(source, target):
2772
if not source.supports_rich_root() and target.supports_rich_root():
2778
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2779
"""See InterRepository.fetch()."""
2780
from bzrlib.fetch import Model1toKnit2Fetcher
2781
f = Model1toKnit2Fetcher(to_repository=self.target,
2782
from_repository=self.source,
2783
last_revision=revision_id,
2784
pb=pb, find_ghosts=find_ghosts)
2785
return f.count_copied, f.failed_revisions
2788
def copy_content(self, revision_id=None):
2789
"""Make a complete copy of the content in self into destination.
2791
This is a destructive operation! Do not use it on existing
2794
:param revision_id: Only copy the content needed to construct
2795
revision_id and its parents.
2798
self.target.set_make_working_trees(self.source.make_working_trees())
2799
except NotImplementedError:
2801
# but don't bother fetching if we have the needed data now.
2802
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
2803
self.target.has_revision(revision_id)):
2805
self.target.fetch(self.source, revision_id=revision_id)
2808
class InterKnit1and2(InterKnitRepo):
2811
def _get_repo_format_to_test(self):
2815
def is_compatible(source, target):
2816
"""Be compatible with Knit1 source and Knit3 target"""
2817
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit3
2819
from bzrlib.repofmt.knitrepo import (RepositoryFormatKnit1,
2820
RepositoryFormatKnit3)
2821
from bzrlib.repofmt.pack_repo import (
2822
RepositoryFormatKnitPack1,
2823
RepositoryFormatKnitPack3,
2824
RepositoryFormatPackDevelopment0,
2825
RepositoryFormatPackDevelopment0Subtree,
2828
RepositoryFormatKnit1,
2829
RepositoryFormatKnitPack1,
2830
RepositoryFormatPackDevelopment0,
2833
RepositoryFormatKnit3,
2834
RepositoryFormatKnitPack3,
2835
RepositoryFormatPackDevelopment0Subtree,
2837
return (isinstance(source._format, nosubtrees) and
2838
isinstance(target._format, subtrees))
2839
except AttributeError:
2843
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2844
"""See InterRepository.fetch()."""
2845
from bzrlib.fetch import Knit1to2Fetcher
2846
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2847
self.source, self.source._format, self.target,
2848
self.target._format)
2849
f = Knit1to2Fetcher(to_repository=self.target,
2850
from_repository=self.source,
2851
last_revision=revision_id,
2852
pb=pb, find_ghosts=find_ghosts)
2853
return f.count_copied, f.failed_revisions
2856
class InterDifferingSerializer(InterKnitRepo):
2859
def _get_repo_format_to_test(self):
2863
def is_compatible(source, target):
2864
"""Be compatible with Knit2 source and Knit3 target"""
2865
if source.supports_rich_root() != target.supports_rich_root():
2867
# Ideally, we'd support fetching if the source had no tree references
2868
# even if it supported them...
2869
if (getattr(source, '_format.supports_tree_reference', False) and
2870
not getattr(target, '_format.supports_tree_reference', False)):
2875
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2876
"""See InterRepository.fetch()."""
2877
revision_ids = self.target.search_missing_revision_ids(self.source,
2878
revision_id, find_ghosts=find_ghosts).get_keys()
2879
revision_ids = tsort.topo_sort(
2880
self.source.get_graph().get_parent_map(revision_ids))
2881
def revisions_iterator():
2882
for current_revision_id in revision_ids:
2883
revision = self.source.get_revision(current_revision_id)
2884
tree = self.source.revision_tree(current_revision_id)
2886
signature = self.source.get_signature_text(
2887
current_revision_id)
2888
except errors.NoSuchRevision:
2890
yield revision, tree, signature
2892
my_pb = ui.ui_factory.nested_progress_bar()
2897
install_revisions(self.target, revisions_iterator(),
2898
len(revision_ids), pb)
2900
if my_pb is not None:
2902
return len(revision_ids), 0
2905
class InterRemoteToOther(InterRepository):
2907
def __init__(self, source, target):
2908
InterRepository.__init__(self, source, target)
2909
self._real_inter = None
2912
def is_compatible(source, target):
2913
if not isinstance(source, remote.RemoteRepository):
2915
# Is source's model compatible with target's model?
2916
source._ensure_real()
2917
real_source = source._real_repository
2918
if isinstance(real_source, remote.RemoteRepository):
2919
raise NotImplementedError(
2920
"We don't support remote repos backed by remote repos yet.")
2921
return InterRepository._same_model(real_source, target)
2924
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2925
"""See InterRepository.fetch()."""
2926
from bzrlib.fetch import RemoteToOtherFetcher
2927
mutter("Using fetch logic to copy between %s(remote) and %s(%s)",
2928
self.source, self.target, self.target._format)
2929
# TODO: jam 20070210 This should be an assert, not a translate
2930
revision_id = osutils.safe_revision_id(revision_id)
2931
f = RemoteToOtherFetcher(to_repository=self.target,
2932
from_repository=self.source,
2933
last_revision=revision_id,
2934
pb=pb, find_ghosts=find_ghosts)
2935
return f.count_copied, f.failed_revisions
2938
def _get_repo_format_to_test(self):
2942
class InterOtherToRemote(InterRepository):
2944
def __init__(self, source, target):
2945
InterRepository.__init__(self, source, target)
2946
self._real_inter = None
2949
def is_compatible(source, target):
2950
if isinstance(target, remote.RemoteRepository):
2954
def _ensure_real_inter(self):
2955
if self._real_inter is None:
2956
self.target._ensure_real()
2957
real_target = self.target._real_repository
2958
self._real_inter = InterRepository.get(self.source, real_target)
2960
def copy_content(self, revision_id=None):
2961
self._ensure_real_inter()
2962
self._real_inter.copy_content(revision_id=revision_id)
2964
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2965
self._ensure_real_inter()
2966
self._real_inter.fetch(revision_id=revision_id, pb=pb,
2967
find_ghosts=find_ghosts)
2970
def _get_repo_format_to_test(self):
2974
InterRepository.register_optimiser(InterDifferingSerializer)
2975
InterRepository.register_optimiser(InterSameDataRepository)
2976
InterRepository.register_optimiser(InterWeaveRepo)
2977
InterRepository.register_optimiser(InterKnitRepo)
2978
InterRepository.register_optimiser(InterModel1and2)
2979
InterRepository.register_optimiser(InterKnit1and2)
2980
InterRepository.register_optimiser(InterPackRepo)
2981
InterRepository.register_optimiser(InterRemoteToOther)
2982
InterRepository.register_optimiser(InterOtherToRemote)
1712
2985
class CopyConverter(object):
1713
2986
"""A repository conversion tool which just performs a copy of the content.
1715
2988
This is slow but quite reliable.