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
262
def abort_write_group(self, suppress_errors=False):
458
# What class to use for a CommitBuilder. Often its simpler to change this
459
# in a Repository class subclass rather than to override
460
# get_commit_builder.
461
_commit_builder_class = CommitBuilder
462
# The search regex used by xml based repositories to determine what things
463
# where changed in a single commit.
464
_file_ids_altered_regex = lazy_regex.lazy_compile(
465
r'file_id="(?P<file_id>[^"]+)"'
466
r'.* revision="(?P<revision_id>[^"]+)"'
469
def abort_write_group(self):
263
470
"""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
472
:seealso: start_write_group.
271
474
if self._write_group is not self.get_transaction():
272
475
# 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)
476
raise errors.BzrError('mismatched lock context and write group.')
477
self._abort_write_group()
290
478
self._write_group = None
292
480
def _abort_write_group(self):
293
481
"""Template method for per-repository write group cleanup.
295
This is called during abort before the write group is considered to be
483
This is called during abort before the write group is considered to be
296
484
finished and should cleanup any internal state accrued during the write
297
485
group. There is no requirement that data handed to the repository be
298
486
*not* made available - this is not a rollback - but neither should any
828
1066
raise NotImplementedError(self.has_revisions)
1068
return self._revision_store.has_revision_id(revision_id,
1069
self.get_transaction())
830
1071
@needs_read_lock
831
1072
def get_revision(self, revision_id):
832
1073
"""Return the Revision object for a named revision."""
833
1074
return self.get_revisions([revision_id])[0]
835
1077
def get_revision_reconcile(self, revision_id):
836
1078
"""'reconcile' helper routine that allows access to a revision always.
838
1080
This variant of get_revision does not cross check the weave graph
839
1081
against the revision one as get_revision does: but it should only
840
1082
be used by reconcile, or reconcile-alike commands that are correcting
841
1083
or testing the revision graph.
843
raise NotImplementedError(self.get_revision_reconcile)
1085
return self._get_revisions([revision_id])[0]
845
1088
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):
1089
"""Get many revisions at once."""
1090
return self._get_revisions(revision_ids)
1093
def _get_revisions(self, revision_ids):
1094
"""Core work logic to get many revisions without sanity checks."""
1095
for rev_id in revision_ids:
1096
if not rev_id or not isinstance(rev_id, basestring):
1097
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1098
revs = self._revision_store.get_revisions(revision_ids,
1099
self.get_transaction())
1103
def get_revision_xml(self, revision_id):
1104
# TODO: jam 20070210 This shouldn't be necessary since get_revision
1105
# would have already do it.
1106
# TODO: jam 20070210 Just use _serializer.write_revision_to_string()
1107
rev = self.get_revision(revision_id)
1108
rev_tmp = StringIO()
1109
# the current serializer..
1110
self._revision_store._serializer.write_revision(rev, rev_tmp)
1112
return rev_tmp.getvalue()
1114
def get_deltas_for_revisions(self, revisions):
854
1115
"""Produce a generator of revision deltas.
856
1117
Note that the input is a sequence of REVISIONS, not revision_ids.
857
1118
Trees will be held in memory until the generator exits.
858
1119
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
1121
required_trees = set()
866
1122
for revision in revisions:
867
1123
required_trees.add(revision.revision_id)
868
1124
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
1125
trees = dict((t.get_revision_id(), t) for
1126
t in self.revision_trees(required_trees))
885
1127
for revision in revisions:
886
1128
if not revision.parent_ids:
887
old_tree = self.revision_tree(_mod_revision.NULL_REVISION)
1129
old_tree = self.revision_tree(None)
889
1131
old_tree = trees[revision.parent_ids[0]]
890
1132
yield trees[revision.revision_id].changes_from(old_tree)
892
1134
@needs_read_lock
893
def get_revision_delta(self, revision_id, specific_fileids=None):
1135
def get_revision_delta(self, revision_id):
894
1136
"""Return the delta for one revision.
896
1138
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
1141
r = self.get_revision(revision_id)
904
return list(self.get_deltas_for_revisions([r],
905
specific_fileids=specific_fileids))[0]
1142
return list(self.get_deltas_for_revisions([r]))[0]
907
1144
@needs_write_lock
908
1145
def store_revision_signature(self, gpg_strategy, plaintext, revision_id):
909
1146
signature = gpg_strategy.sign(plaintext)
910
1147
self.add_signature_text(revision_id, signature)
912
1150
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)
1151
self._revision_store.add_revision_signature_text(revision_id,
1153
self.get_transaction())
1155
def find_text_key_references(self):
1156
"""Find the text key references within the repository.
1158
:return: a dictionary mapping (file_id, revision_id) tuples to altered file-ids to an iterable of
1159
revision_ids. Each altered file-ids has the exact revision_ids that
1160
altered it listed explicitly.
1161
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1162
to whether they were referred to by the inventory of the
1163
revision_id that they contain. The inventory texts from all present
1164
revision ids are assessed to generate this report.
1166
revision_ids = self.all_revision_ids()
1167
w = self.get_inventory_weave()
1168
pb = ui.ui_factory.nested_progress_bar()
1170
return self._find_text_key_references_from_xml_inventory_lines(
1171
w.iter_lines_added_or_present_in_versions(revision_ids, pb=pb))
1175
def _find_text_key_references_from_xml_inventory_lines(self,
1177
"""Core routine for extracting references to texts from inventories.
1179
This performs the translation of xml lines to revision ids.
1181
:param line_iterator: An iterator of lines, origin_version_id
1182
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1183
to whether they were referred to by the inventory of the
1184
revision_id that they contain. Note that if that revision_id was
1185
not part of the line_iterator's output then False will be given -
1186
even though it may actually refer to that key.
1188
if not self._serializer.support_altered_by_hack:
1189
raise AssertionError(
1190
"_find_text_key_references_from_xml_inventory_lines only "
1191
"supported for branches which store inventory as unnested xml"
1192
", not on %r" % self)
1195
# this code needs to read every new line in every inventory for the
1196
# inventories [revision_ids]. Seeing a line twice is ok. Seeing a line
1197
# not present in one of those inventories is unnecessary but not
1198
# harmful because we are filtering by the revision id marker in the
1199
# inventory lines : we only select file ids altered in one of those
1200
# revisions. We don't need to see all lines in the inventory because
1201
# only those added in an inventory in rev X can contain a revision=X
1203
unescape_revid_cache = {}
1204
unescape_fileid_cache = {}
1206
# jam 20061218 In a big fetch, this handles hundreds of thousands
1207
# of lines, so it has had a lot of inlining and optimizing done.
1208
# Sorry that it is a little bit messy.
1209
# Move several functions to be local variables, since this is a long
1211
search = self._file_ids_altered_regex.search
1212
unescape = _unescape_xml
1213
setdefault = result.setdefault
1214
for line, version_id in line_iterator:
1215
match = search(line)
1218
# One call to match.group() returning multiple items is quite a
1219
# bit faster than 2 calls to match.group() each returning 1
1220
file_id, revision_id = match.group('file_id', 'revision_id')
1222
# Inlining the cache lookups helps a lot when you make 170,000
1223
# lines and 350k ids, versus 8.4 unique ids.
1224
# Using a cache helps in 2 ways:
1225
# 1) Avoids unnecessary decoding calls
1226
# 2) Re-uses cached strings, which helps in future set and
1228
# (2) is enough that removing encoding entirely along with
1229
# the cache (so we are using plain strings) results in no
1230
# performance improvement.
1232
revision_id = unescape_revid_cache[revision_id]
1234
unescaped = unescape(revision_id)
1235
unescape_revid_cache[revision_id] = unescaped
1236
revision_id = unescaped
1238
# Note that unconditionally unescaping means that we deserialise
1239
# every fileid, which for general 'pull' is not great, but we don't
1240
# really want to have some many fulltexts that this matters anyway.
1243
file_id = unescape_fileid_cache[file_id]
1245
unescaped = unescape(file_id)
1246
unescape_fileid_cache[file_id] = unescaped
1249
key = (file_id, revision_id)
1250
setdefault(key, False)
1251
if revision_id == version_id:
1255
def _find_file_ids_from_xml_inventory_lines(self, line_iterator,
1257
"""Helper routine for fileids_altered_by_revision_ids.
1259
This performs the translation of xml lines to revision ids.
1261
:param line_iterator: An iterator of lines, origin_version_id
1262
:param revision_ids: The revision ids to filter for. This should be a
1263
set or other type which supports efficient __contains__ lookups, as
1264
the revision id from each parsed line will be looked up in the
1265
revision_ids filter.
1266
:return: a dictionary mapping altered file-ids to an iterable of
1267
revision_ids. Each altered file-ids has the exact revision_ids that
1268
altered it listed explicitly.
1271
setdefault = result.setdefault
1272
for file_id, revision_id in \
1273
self._find_text_key_references_from_xml_inventory_lines(
1274
line_iterator).iterkeys():
1275
# once data is all ensured-consistent; then this is
1276
# if revision_id == version_id
1277
if revision_id in revision_ids:
1278
setdefault(file_id, set()).add(revision_id)
1281
def fileids_altered_by_revision_ids(self, revision_ids, _inv_weave=None):
1282
"""Find the file ids and versions affected by revisions.
1284
:param revisions: an iterable containing revision ids.
1285
:param _inv_weave: The inventory weave from this repository or None.
1286
If None, the inventory weave will be opened automatically.
1287
:return: a dictionary mapping altered file-ids to an iterable of
1288
revision_ids. Each altered file-ids has the exact revision_ids that
1289
altered it listed explicitly.
1291
selected_revision_ids = set(revision_ids)
1292
w = _inv_weave or self.get_inventory_weave()
1293
pb = ui.ui_factory.nested_progress_bar()
1295
return self._find_file_ids_from_xml_inventory_lines(
1296
w.iter_lines_added_or_present_in_versions(
1297
selected_revision_ids, pb=pb),
1298
selected_revision_ids)
933
1302
def iter_files_bytes(self, desired_files):
934
1303
"""Iterate through file versions.
941
1310
uniquely identify the file version in the caller's context. (Examples:
942
1311
an index number or a TreeTransform trans_id.)
1313
bytes_iterator is an iterable of bytestrings for the file. The
1314
kind of iterable and length of the bytestrings are unspecified, but for
1315
this implementation, it is a list of lines produced by
1316
VersionedFile.get_lines().
944
1318
: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])
1321
transaction = self.get_transaction()
1322
for file_id, revision_id, callable_data in desired_files:
1324
weave = self.weave_store.get_weave(file_id, transaction)
1325
except errors.NoSuchFile:
1326
raise errors.NoSuchIdInRepository(self, file_id)
1327
yield callable_data, weave.get_lines(revision_id)
1329
def _generate_text_key_index(self, text_key_references=None,
1331
"""Generate a new text key index for the repository.
1333
This is an expensive function that will take considerable time to run.
1335
:return: A dict mapping text keys ((file_id, revision_id) tuples) to a
1336
list of parents, also text keys. When a given key has no parents,
1337
the parents list will be [NULL_REVISION].
1339
# All revisions, to find inventory parents.
1340
if ancestors is None:
1341
graph = self.get_graph()
1342
ancestors = graph.get_parent_map(self.all_revision_ids())
1343
if text_key_references is None:
1344
text_key_references = self.find_text_key_references()
1345
pb = ui.ui_factory.nested_progress_bar()
1347
return self._do_generate_text_key_index(ancestors,
1348
text_key_references, pb)
1352
def _do_generate_text_key_index(self, ancestors, text_key_references, pb):
1353
"""Helper for _generate_text_key_index to avoid deep nesting."""
1354
revision_order = tsort.topo_sort(ancestors)
1355
invalid_keys = set()
1357
for revision_id in revision_order:
1358
revision_keys[revision_id] = set()
1359
text_count = len(text_key_references)
1360
# a cache of the text keys to allow reuse; costs a dict of all the
1361
# keys, but saves a 2-tuple for every child of a given key.
1363
for text_key, valid in text_key_references.iteritems():
1365
invalid_keys.add(text_key)
1367
revision_keys[text_key[1]].add(text_key)
1368
text_key_cache[text_key] = text_key
1369
del text_key_references
1371
text_graph = graph.Graph(graph.DictParentsProvider(text_index))
1372
NULL_REVISION = _mod_revision.NULL_REVISION
1373
# Set a cache with a size of 10 - this suffices for bzr.dev but may be
1374
# too small for large or very branchy trees. However, for 55K path
1375
# trees, it would be easy to use too much memory trivially. Ideally we
1376
# could gauge this by looking at available real memory etc, but this is
1377
# always a tricky proposition.
1378
inventory_cache = lru_cache.LRUCache(10)
1379
batch_size = 10 # should be ~150MB on a 55K path tree
1380
batch_count = len(revision_order) / batch_size + 1
1382
pb.update("Calculating text parents.", processed_texts, text_count)
1383
for offset in xrange(batch_count):
1384
to_query = revision_order[offset * batch_size:(offset + 1) *
1388
for rev_tree in self.revision_trees(to_query):
1389
revision_id = rev_tree.get_revision_id()
1390
parent_ids = ancestors[revision_id]
1391
for text_key in revision_keys[revision_id]:
1392
pb.update("Calculating text parents.", processed_texts)
1393
processed_texts += 1
1394
candidate_parents = []
1395
for parent_id in parent_ids:
1396
parent_text_key = (text_key[0], parent_id)
1398
check_parent = parent_text_key not in \
1399
revision_keys[parent_id]
1401
# the parent parent_id is a ghost:
1402
check_parent = False
1403
# truncate the derived graph against this ghost.
1404
parent_text_key = None
1406
# look at the parent commit details inventories to
1407
# determine possible candidates in the per file graph.
1410
inv = inventory_cache[parent_id]
1412
inv = self.revision_tree(parent_id).inventory
1413
inventory_cache[parent_id] = inv
1414
parent_entry = inv._byid.get(text_key[0], None)
1415
if parent_entry is not None:
1417
text_key[0], parent_entry.revision)
1419
parent_text_key = None
1420
if parent_text_key is not None:
1421
candidate_parents.append(
1422
text_key_cache[parent_text_key])
1423
parent_heads = text_graph.heads(candidate_parents)
1424
new_parents = list(parent_heads)
1425
new_parents.sort(key=lambda x:candidate_parents.index(x))
1426
if new_parents == []:
1427
new_parents = [NULL_REVISION]
1428
text_index[text_key] = new_parents
1430
for text_key in invalid_keys:
1431
text_index[text_key] = [NULL_REVISION]
1434
def item_keys_introduced_by(self, revision_ids, _files_pb=None):
1435
"""Get an iterable listing the keys of all the data introduced by a set
1438
The keys will be ordered so that the corresponding items can be safely
1439
fetched and inserted in that order.
1441
:returns: An iterable producing tuples of (knit-kind, file-id,
1442
versions). knit-kind is one of 'file', 'inventory', 'signatures',
1443
'revisions'. file-id is None unless knit-kind is 'file'.
1445
# XXX: it's a bit weird to control the inventory weave caching in this
1446
# generator. Ideally the caching would be done in fetch.py I think. Or
1447
# maybe this generator should explicitly have the contract that it
1448
# should not be iterated until the previously yielded item has been
1450
inv_w = self.get_inventory_weave()
1452
# file ids that changed
1453
file_ids = self.fileids_altered_by_revision_ids(revision_ids, inv_w)
1455
num_file_ids = len(file_ids)
1456
for file_id, altered_versions in file_ids.iteritems():
1457
if _files_pb is not None:
1458
_files_pb.update("fetch texts", count, num_file_ids)
1460
yield ("file", file_id, altered_versions)
1461
# We're done with the files_pb. Note that it finished by the caller,
1462
# just as it was created by the caller.
1466
yield ("inventory", None, revision_ids)
1469
revisions_with_signatures = set()
1470
for rev_id in revision_ids:
1472
self.get_signature_text(rev_id)
1473
except errors.NoSuchRevision:
1477
revisions_with_signatures.add(rev_id)
1478
yield ("signatures", None, revisions_with_signatures)
1481
yield ("revisions", None, revision_ids)
1484
def get_inventory_weave(self):
1485
return self.control_weaves.get_weave('inventory',
1486
self.get_transaction())
1489
def get_inventory(self, revision_id):
1490
"""Get Inventory object by revision id."""
1491
return self.iter_inventories([revision_id]).next()
1493
def iter_inventories(self, revision_ids):
1494
"""Get many inventories by revision_ids.
1496
This will buffer some or all of the texts used in constructing the
1497
inventories in memory, but will only parse a single inventory at a
1500
:return: An iterator of inventories.
1502
if ((None in revision_ids)
1503
or (_mod_revision.NULL_REVISION in revision_ids)):
1504
raise ValueError('cannot get null revision inventory')
1505
return self._iter_inventories(revision_ids)
1507
def _iter_inventories(self, revision_ids):
1508
"""single-document based inventory iteration."""
1509
texts = self.get_inventory_weave().get_texts(revision_ids)
1510
for text, revision_id in zip(texts, revision_ids):
1511
yield self.deserialise_inventory(revision_id, text)
1513
def deserialise_inventory(self, revision_id, xml):
1514
"""Transform the xml into an inventory object.
1516
:param revision_id: The expected revision id of the inventory.
1517
:param xml: A serialised inventory.
1519
result = self._serializer.read_inventory_from_string(xml, revision_id)
1520
if result.revision_id != revision_id:
1521
raise AssertionError('revision id mismatch %s != %s' % (
1522
result.revision_id, revision_id))
1525
def serialise_inventory(self, inv):
1526
return self._serializer.write_inventory_to_string(inv)
1528
def _serialise_inventory_to_lines(self, inv):
1529
return self._serializer.write_inventory_to_lines(inv)
1531
def get_serializer_format(self):
1532
return self._serializer.format_num
1535
def get_inventory_xml(self, revision_id):
1536
"""Get inventory XML as a file object."""
1538
iw = self.get_inventory_weave()
1539
return iw.get_text(revision_id)
1541
raise errors.HistoryMissing(self, 'inventory', revision_id)
1544
def get_inventory_sha1(self, revision_id):
1545
"""Return the sha1 hash of the inventory entry
1547
return self.get_revision(revision_id).inventory_sha1
1549
def iter_reverse_revision_history(self, revision_id):
1550
"""Iterate backwards through revision ids in the lefthand history
1552
:param revision_id: The revision id to start with. All its lefthand
1553
ancestors will be traversed.
1555
graph = self.get_graph()
1556
next_id = revision_id
1558
if next_id in (None, _mod_revision.NULL_REVISION):
1561
# Note: The following line may raise KeyError in the event of
1562
# truncated history. We decided not to have a try:except:raise
1563
# RevisionNotPresent here until we see a use for it, because of the
1564
# cost in an inner loop that is by its very nature O(history).
1565
# Robert Collins 20080326
1566
parents = graph.get_parent_map([next_id])[next_id]
1567
if len(parents) == 0:
1570
next_id = parents[0]
1573
def get_revision_inventory(self, revision_id):
1574
"""Return inventory of a past revision."""
1575
# TODO: Unify this with get_inventory()
1576
# bzr 0.0.6 and later imposes the constraint that the inventory_id
1577
# must be the same as its revision, so this is trivial.
1578
if revision_id is None:
1579
# This does not make sense: if there is no revision,
1580
# then it is the current tree inventory surely ?!
1581
# and thus get_root_id() is something that looks at the last
1582
# commit on the branch, and the get_root_id is an inventory check.
1583
raise NotImplementedError
1584
# return Inventory(self.get_root_id())
1586
return self.get_inventory(revision_id)
984
1589
def is_shared(self):
985
1590
"""Return True if this repository is flagged as a shared repository."""
986
1591
raise NotImplementedError(self.is_shared)
1000
1605
for repositories to maintain loaded indices across multiple locks
1001
1606
by checking inside their implementation of this method to see
1002
1607
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
1608
the disk format being validatable in this manner.
1010
1611
@needs_read_lock
1011
1612
def revision_tree(self, revision_id):
1012
1613
"""Return Tree for a revision on this branch.
1014
`revision_id` may be NULL_REVISION for the empty tree revision.
1615
`revision_id` may be None for the empty tree revision.
1016
raise NotImplementedError(self.revision_tree)
1617
# TODO: refactor this to use an existing revision object
1618
# so we don't need to read it in twice.
1619
if revision_id is None or revision_id == _mod_revision.NULL_REVISION:
1620
return RevisionTree(self, Inventory(root_id=None),
1621
_mod_revision.NULL_REVISION)
1623
inv = self.get_revision_inventory(revision_id)
1624
return RevisionTree(self, inv, revision_id)
1018
1626
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:'
1627
"""Return Tree for a revision on this branch.
1629
`revision_id` may not be None or 'null:'"""
1630
inventories = self.iter_inventories(revision_ids)
1631
for inv in inventories:
1632
yield RevisionTree(self, inv, inv.revision_id)
1635
def get_ancestry(self, revision_id, topo_sorted=True):
1636
"""Return a list of revision-ids integrated by a revision.
1638
The first element of the list is always None, indicating the origin
1639
revision. This might change when we have history horizons, or
1640
perhaps we should have a new API.
1642
This is topologically sorted.
1024
raise NotImplementedError(self.revision_trees)
1644
if _mod_revision.is_null(revision_id):
1646
if not self.has_revision(revision_id):
1647
raise errors.NoSuchRevision(self, revision_id)
1648
w = self.get_inventory_weave()
1649
candidates = w.get_ancestry(revision_id, topo_sorted)
1650
return [None] + candidates # self._eliminate_revisions_not_present(candidates)
1026
def pack(self, hint=None, clean_obsolete_packs=False):
1027
1653
"""Compress the data within the repository.
1029
1655
This operation only makes sense for some repository types. For other
1030
1656
types it should be a no-op that just returns.
1032
1658
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
1659
@needs_write_lock as this is a long running call its reasonable to
1034
1660
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
1664
def print_file(self, file, revision_id):
1665
"""Print `file` to stdout.
1667
FIXME RBC 20060125 as John Meinel points out this is a bad api
1668
- it writes to stdout, it assumes that that is valid etc. Fix
1669
by creating a new more flexible convenience function.
1671
tree = self.revision_tree(revision_id)
1672
# use inventory as it was in that revision
1673
file_id = tree.inventory.path2id(file)
1675
# TODO: jam 20060427 Write a test for this code path
1676
# it had a bug in it, and was raising the wrong
1678
raise errors.BzrError("%r is not present in revision %s" % (file, revision_id))
1679
tree.print_file(file_id)
1047
1681
def get_transaction(self):
1048
1682
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)
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:
1684
@deprecated_method(symbol_versioning.one_five)
1685
def revision_parents(self, revision_id):
1686
return self.get_inventory_weave().parent_names(revision_id)
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]
1694
def get_parent_map(self, keys):
1695
"""See graph._StackedParentsProvider.get_parent_map"""
1697
for revision_id in keys:
1061
1698
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,)
1699
parent_map[revision_id] = ()
1702
parent_id_list = self.get_revision(revision_id).parent_ids
1703
except errors.NoSuchRevision:
1706
if len(parent_id_list) == 0:
1707
parent_ids = (_mod_revision.NULL_REVISION,)
1709
parent_ids = tuple(parent_id_list)
1710
parent_map[revision_id] = parent_ids
1077
1713
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
1716
def get_graph(self, other_repository=None):
1099
1717
"""Return the graph walker for this repository format"""
1100
1718
parents_provider = self._make_parents_provider()
1101
1719
if (other_repository is not None and
1102
1720
not self.has_same_location(other_repository)):
1103
parents_provider = graph.StackedParentsProvider(
1721
parents_provider = graph._StackedParentsProvider(
1104
1722
[parents_provider, other_repository._make_parents_provider()])
1105
1723
return graph.Graph(parents_provider)
1725
def _get_versioned_file_checker(self):
1726
"""Return an object suitable for checking versioned files."""
1727
return _VersionedFileChecker(self)
1729
def revision_ids_to_search_result(self, result_set):
1730
"""Convert a set of revision ids to a graph SearchResult."""
1731
result_parents = set()
1732
for parents in self.get_graph().get_parent_map(
1733
result_set).itervalues():
1734
result_parents.update(parents)
1735
included_keys = result_set.intersection(result_parents)
1736
start_keys = result_set.difference(included_keys)
1737
exclude_keys = result_parents.difference(result_set)
1738
result = graph.SearchResult(start_keys, exclude_keys,
1739
len(result_set), result_set)
1107
1742
@needs_write_lock
1108
1743
def set_make_working_trees(self, new_value):
1109
1744
"""Set the policy flag for making working trees when creating branches.
1220
1814
revision_id.decode('ascii')
1221
1815
except UnicodeDecodeError:
1222
1816
raise errors.NonAsciiRevisionId(method, self)
1818
def revision_graph_can_have_wrong_parents(self):
1819
"""Is it possible for this repository to have a revision graph with
1822
If True, then this repository must also implement
1823
_find_inconsistent_revision_parents so that check and reconcile can
1824
check for inconsistencies before proceeding with other checks that may
1825
depend on the revision index being consistent.
1827
raise NotImplementedError(self.revision_graph_can_have_wrong_parents)
1830
# remove these delegates a while after bzr 0.15
1831
def __make_delegated(name, from_module):
1832
def _deprecated_repository_forwarder():
1833
symbol_versioning.warn('%s moved to %s in bzr 0.15'
1834
% (name, from_module),
1837
m = __import__(from_module, globals(), locals(), [name])
1839
return getattr(m, name)
1840
except AttributeError:
1841
raise AttributeError('module %s has no name %s'
1843
globals()[name] = _deprecated_repository_forwarder
1846
'AllInOneRepository',
1847
'WeaveMetaDirRepository',
1848
'PreSplitOutRepositoryFormat',
1849
'RepositoryFormat4',
1850
'RepositoryFormat5',
1851
'RepositoryFormat6',
1852
'RepositoryFormat7',
1854
__make_delegated(_name, 'bzrlib.repofmt.weaverepo')
1858
'RepositoryFormatKnit',
1859
'RepositoryFormatKnit1',
1861
__make_delegated(_name, 'bzrlib.repofmt.knitrepo')
1864
def install_revision(repository, rev, revision_tree):
1865
"""Install all revision data into a repository."""
1866
install_revisions(repository, [(rev, revision_tree, None)])
1869
def install_revisions(repository, iterable, num_revisions=None, pb=None):
1870
"""Install all revision data into a repository.
1872
Accepts an iterable of revision, tree, signature tuples. The signature
1875
repository.start_write_group()
1877
for n, (revision, revision_tree, signature) in enumerate(iterable):
1878
_install_revision(repository, revision, revision_tree, signature)
1880
pb.update('Transferring revisions', n + 1, num_revisions)
1882
repository.abort_write_group()
1885
repository.commit_write_group()
1888
def _install_revision(repository, rev, revision_tree, signature):
1889
"""Install all revision data into a repository."""
1890
present_parents = []
1892
for p_id in rev.parent_ids:
1893
if repository.has_revision(p_id):
1894
present_parents.append(p_id)
1895
parent_trees[p_id] = repository.revision_tree(p_id)
1897
parent_trees[p_id] = repository.revision_tree(None)
1899
inv = revision_tree.inventory
1900
entries = inv.iter_entries()
1901
# backwards compatibility hack: skip the root id.
1902
if not repository.supports_rich_root():
1903
path, root = entries.next()
1904
if root.revision != rev.revision_id:
1905
raise errors.IncompatibleRevision(repr(repository))
1906
# Add the texts that are not already present
1907
for path, ie in entries:
1908
w = repository.weave_store.get_weave_or_empty(ie.file_id,
1909
repository.get_transaction())
1910
if ie.revision not in w:
1912
# FIXME: TODO: The following loop *may* be overlapping/duplicate
1913
# with InventoryEntry.find_previous_heads(). if it is, then there
1914
# is a latent bug here where the parents may have ancestors of each
1916
for revision, tree in parent_trees.iteritems():
1917
if ie.file_id not in tree:
1919
parent_id = tree.inventory[ie.file_id].revision
1920
if parent_id in text_parents:
1922
text_parents.append(parent_id)
1924
vfile = repository.weave_store.get_weave_or_empty(ie.file_id,
1925
repository.get_transaction())
1926
lines = revision_tree.get_file(ie.file_id).readlines()
1927
vfile.add_lines(rev.revision_id, text_parents, lines)
1929
# install the inventory
1930
repository.add_inventory(rev.revision_id, inv, present_parents)
1931
except errors.RevisionAlreadyPresent:
1933
if signature is not None:
1934
repository.add_signature_text(rev.revision_id, signature)
1935
repository.add_revision(rev.revision_id, rev, inv)
1225
1938
class MetaDirRepository(Repository):
1226
"""Repositories in the new meta-dir layout.
1228
:ivar _transport: Transport for access to repository control files,
1229
typically pointing to .bzr/repository.
1232
def __init__(self, _format, a_bzrdir, control_files):
1233
super(MetaDirRepository, self).__init__(_format, a_bzrdir, control_files)
1234
self._transport = control_files._transport
1939
"""Repositories in the new meta-dir layout."""
1941
def __init__(self, _format, a_bzrdir, control_files, _revision_store, control_store, text_store):
1942
super(MetaDirRepository, self).__init__(_format,
1948
dir_mode = self.control_files._dir_mode
1949
file_mode = self.control_files._file_mode
1236
1952
def is_shared(self):
1237
1953
"""Return True if this repository is flagged as a shared repository."""
1238
return self._transport.has('shared-storage')
1954
return self.control_files._transport.has('shared-storage')
1240
1956
@needs_write_lock
1241
1957
def set_make_working_trees(self, new_value):
1632
2301
_optimisers = []
1633
2302
"""The available optimised InterRepository types."""
1636
2304
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):
2305
raise NotImplementedError(self.copy_content)
2307
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
1653
2308
"""Fetch the content required to construct revision_id.
1655
2310
The content is copied from self.source to self.target.
1657
2312
:param revision_id: if None all content is copied, if NULL_REVISION no
1658
2313
content is copied.
2314
:param pb: optional progress bar to use for progress reports. If not
2315
provided a default one will be created.
2317
Returns the copied revision count and the failed revisions in a tuple:
1661
2320
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
2322
def _walk_to_common_revisions(self, revision_ids):
2323
"""Walk out from revision_ids in source to revisions target has.
2325
:param revision_ids: The start point for the search.
2326
:return: A set of revision ids.
2328
target_graph = self.target.get_graph()
2329
revision_ids = frozenset(revision_ids)
2330
if set(target_graph.get_parent_map(revision_ids)) == revision_ids:
2331
return graph.SearchResult(revision_ids, set(), 0, set())
2332
missing_revs = set()
2333
source_graph = self.source.get_graph()
2334
# ensure we don't pay silly lookup costs.
2335
searcher = source_graph._make_breadth_first_searcher(revision_ids)
2336
null_set = frozenset([_mod_revision.NULL_REVISION])
2339
next_revs, ghosts = searcher.next_with_ghosts()
2340
except StopIteration:
2342
if revision_ids.intersection(ghosts):
2343
absent_ids = set(revision_ids.intersection(ghosts))
2344
# If all absent_ids are present in target, no error is needed.
2345
absent_ids.difference_update(
2346
set(target_graph.get_parent_map(absent_ids)))
2348
raise errors.NoSuchRevision(self.source, absent_ids.pop())
2349
# we don't care about other ghosts as we can't fetch them and
2350
# haven't been asked to.
2351
next_revs = set(next_revs)
2352
# we always have NULL_REVISION present.
2353
have_revs = set(target_graph.get_parent_map(next_revs)).union(null_set)
2354
missing_revs.update(next_revs - have_revs)
2355
searcher.stop_searching_any(have_revs)
2356
return searcher.get_result()
2358
@deprecated_method(symbol_versioning.one_two)
2360
def missing_revision_ids(self, revision_id=None, find_ghosts=True):
2361
"""Return the revision ids that source has that target does not.
2363
These are returned in topological order.
2365
:param revision_id: only return revision ids included by this
2367
:param find_ghosts: If True find missing revisions in deep history
2368
rather than just finding the surface difference.
2370
return list(self.search_missing_revision_ids(
2371
revision_id, find_ghosts).get_keys())
2374
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2375
"""Return the revision ids that source has that target does not.
2377
:param revision_id: only return revision ids included by this
2379
:param find_ghosts: If True find missing revisions in deep history
2380
rather than just finding the surface difference.
1683
2381
:return: A bzrlib.graph.SearchResult.
1685
raise NotImplementedError(self.search_missing_revision_ids)
2383
# stop searching at found target revisions.
2384
if not find_ghosts and revision_id is not None:
2385
return self._walk_to_common_revisions([revision_id])
2386
# generic, possibly worst case, slow code path.
2387
target_ids = set(self.target.all_revision_ids())
2388
if revision_id is not None:
2389
source_ids = self.source.get_ancestry(revision_id)
2390
if source_ids[0] is not None:
2391
raise AssertionError()
2394
source_ids = self.source.all_revision_ids()
2395
result_set = set(source_ids).difference(target_ids)
2396
return self.source.revision_ids_to_search_result(result_set)
1688
2399
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:
2400
"""True if source and target have the same data representation."""
2401
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
2403
if source._serializer != target._serializer:
1708
raise errors.IncompatibleRepositories(source, target,
1709
"different serializers")
2408
class InterSameDataRepository(InterRepository):
2409
"""Code for converting between repositories that represent the same data.
2411
Data format and model must match for this to work.
2415
def _get_repo_format_to_test(self):
2416
"""Repository format for testing with.
2418
InterSameData can pull from subtree to subtree and from non-subtree to
2419
non-subtree, so we test this with the richest repository format.
2421
from bzrlib.repofmt import knitrepo
2422
return knitrepo.RepositoryFormatKnit3()
2425
def is_compatible(source, target):
2426
return InterRepository._same_model(source, target)
2429
def copy_content(self, revision_id=None):
2430
"""Make a complete copy of the content in self into destination.
2432
This copies both the repository's revision data, and configuration information
2433
such as the make_working_trees setting.
2435
This is a destructive operation! Do not use it on existing
2438
:param revision_id: Only copy the content needed to construct
2439
revision_id and its parents.
2442
self.target.set_make_working_trees(self.source.make_working_trees())
2443
except NotImplementedError:
2445
# but don't bother fetching if we have the needed data now.
2446
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
2447
self.target.has_revision(revision_id)):
2449
self.target.fetch(self.source, revision_id=revision_id)
2452
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2453
"""See InterRepository.fetch()."""
2454
from bzrlib.fetch import GenericRepoFetcher
2455
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2456
self.source, self.source._format, self.target,
2457
self.target._format)
2458
f = GenericRepoFetcher(to_repository=self.target,
2459
from_repository=self.source,
2460
last_revision=revision_id,
2461
pb=pb, find_ghosts=find_ghosts)
2462
return f.count_copied, f.failed_revisions
2465
class InterWeaveRepo(InterSameDataRepository):
2466
"""Optimised code paths between Weave based repositories.
2468
This should be in bzrlib/repofmt/weaverepo.py but we have not yet
2469
implemented lazy inter-object optimisation.
2473
def _get_repo_format_to_test(self):
2474
from bzrlib.repofmt import weaverepo
2475
return weaverepo.RepositoryFormat7()
2478
def is_compatible(source, target):
2479
"""Be compatible with known Weave formats.
2481
We don't test for the stores being of specific types because that
2482
could lead to confusing results, and there is no need to be
2485
from bzrlib.repofmt.weaverepo import (
2491
return (isinstance(source._format, (RepositoryFormat5,
2493
RepositoryFormat7)) and
2494
isinstance(target._format, (RepositoryFormat5,
2496
RepositoryFormat7)))
2497
except AttributeError:
2501
def copy_content(self, revision_id=None):
2502
"""See InterRepository.copy_content()."""
2503
# weave specific optimised path:
2505
self.target.set_make_working_trees(self.source.make_working_trees())
2506
except (errors.RepositoryUpgradeRequired, NotImplemented):
2508
# FIXME do not peek!
2509
if self.source.control_files._transport.listable():
2510
pb = ui.ui_factory.nested_progress_bar()
2512
self.target.weave_store.copy_all_ids(
2513
self.source.weave_store,
2515
from_transaction=self.source.get_transaction(),
2516
to_transaction=self.target.get_transaction())
2517
pb.update('copying inventory', 0, 1)
2518
self.target.control_weaves.copy_multi(
2519
self.source.control_weaves, ['inventory'],
2520
from_transaction=self.source.get_transaction(),
2521
to_transaction=self.target.get_transaction())
2522
self.target._revision_store.text_store.copy_all_ids(
2523
self.source._revision_store.text_store,
2528
self.target.fetch(self.source, revision_id=revision_id)
2531
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2532
"""See InterRepository.fetch()."""
2533
from bzrlib.fetch import GenericRepoFetcher
2534
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2535
self.source, self.source._format, self.target, self.target._format)
2536
f = GenericRepoFetcher(to_repository=self.target,
2537
from_repository=self.source,
2538
last_revision=revision_id,
2539
pb=pb, find_ghosts=find_ghosts)
2540
return f.count_copied, f.failed_revisions
2543
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2544
"""See InterRepository.missing_revision_ids()."""
2545
# we want all revisions to satisfy revision_id in source.
2546
# but we don't want to stat every file here and there.
2547
# we want then, all revisions other needs to satisfy revision_id
2548
# checked, but not those that we have locally.
2549
# so the first thing is to get a subset of the revisions to
2550
# satisfy revision_id in source, and then eliminate those that
2551
# we do already have.
2552
# this is slow on high latency connection to self, but as as this
2553
# disk format scales terribly for push anyway due to rewriting
2554
# inventory.weave, this is considered acceptable.
2556
if revision_id is not None:
2557
source_ids = self.source.get_ancestry(revision_id)
2558
if source_ids[0] is not None:
2559
raise AssertionError()
2562
source_ids = self.source._all_possible_ids()
2563
source_ids_set = set(source_ids)
2564
# source_ids is the worst possible case we may need to pull.
2565
# now we want to filter source_ids against what we actually
2566
# have in target, but don't try to check for existence where we know
2567
# we do not have a revision as that would be pointless.
2568
target_ids = set(self.target._all_possible_ids())
2569
possibly_present_revisions = target_ids.intersection(source_ids_set)
2570
actually_present_revisions = set(
2571
self.target._eliminate_revisions_not_present(possibly_present_revisions))
2572
required_revisions = source_ids_set.difference(actually_present_revisions)
2573
if revision_id is not None:
2574
# we used get_ancestry to determine source_ids then we are assured all
2575
# revisions referenced are present as they are installed in topological order.
2576
# and the tip revision was validated by get_ancestry.
2577
result_set = required_revisions
2579
# if we just grabbed the possibly available ids, then
2580
# we only have an estimate of whats available and need to validate
2581
# that against the revision records.
2583
self.source._eliminate_revisions_not_present(required_revisions))
2584
return self.source.revision_ids_to_search_result(result_set)
2587
class InterKnitRepo(InterSameDataRepository):
2588
"""Optimised code paths between Knit based repositories."""
2591
def _get_repo_format_to_test(self):
2592
from bzrlib.repofmt import knitrepo
2593
return knitrepo.RepositoryFormatKnit1()
2596
def is_compatible(source, target):
2597
"""Be compatible with known Knit formats.
2599
We don't test for the stores being of specific types because that
2600
could lead to confusing results, and there is no need to be
2603
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit
2605
are_knits = (isinstance(source._format, RepositoryFormatKnit) and
2606
isinstance(target._format, RepositoryFormatKnit))
2607
except AttributeError:
2609
return are_knits and InterRepository._same_model(source, target)
2612
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2613
"""See InterRepository.fetch()."""
2614
from bzrlib.fetch import KnitRepoFetcher
2615
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2616
self.source, self.source._format, self.target, self.target._format)
2617
f = KnitRepoFetcher(to_repository=self.target,
2618
from_repository=self.source,
2619
last_revision=revision_id,
2620
pb=pb, find_ghosts=find_ghosts)
2621
return f.count_copied, f.failed_revisions
2624
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2625
"""See InterRepository.missing_revision_ids()."""
2626
if revision_id is not None:
2627
source_ids = self.source.get_ancestry(revision_id)
2628
if source_ids[0] is not None:
2629
raise AssertionError()
2632
source_ids = self.source.all_revision_ids()
2633
source_ids_set = set(source_ids)
2634
# source_ids is the worst possible case we may need to pull.
2635
# now we want to filter source_ids against what we actually
2636
# have in target, but don't try to check for existence where we know
2637
# we do not have a revision as that would be pointless.
2638
target_ids = set(self.target.all_revision_ids())
2639
possibly_present_revisions = target_ids.intersection(source_ids_set)
2640
actually_present_revisions = set(
2641
self.target._eliminate_revisions_not_present(possibly_present_revisions))
2642
required_revisions = source_ids_set.difference(actually_present_revisions)
2643
if revision_id is not None:
2644
# we used get_ancestry to determine source_ids then we are assured all
2645
# revisions referenced are present as they are installed in topological order.
2646
# and the tip revision was validated by get_ancestry.
2647
result_set = required_revisions
2649
# if we just grabbed the possibly available ids, then
2650
# we only have an estimate of whats available and need to validate
2651
# that against the revision records.
2653
self.source._eliminate_revisions_not_present(required_revisions))
2654
return self.source.revision_ids_to_search_result(result_set)
2657
class InterPackRepo(InterSameDataRepository):
2658
"""Optimised code paths between Pack based repositories."""
2661
def _get_repo_format_to_test(self):
2662
from bzrlib.repofmt import pack_repo
2663
return pack_repo.RepositoryFormatKnitPack1()
2666
def is_compatible(source, target):
2667
"""Be compatible with known Pack formats.
2669
We don't test for the stores being of specific types because that
2670
could lead to confusing results, and there is no need to be
2673
from bzrlib.repofmt.pack_repo import RepositoryFormatPack
2675
are_packs = (isinstance(source._format, RepositoryFormatPack) and
2676
isinstance(target._format, RepositoryFormatPack))
2677
except AttributeError:
2679
return are_packs and InterRepository._same_model(source, target)
2682
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2683
"""See InterRepository.fetch()."""
2684
from bzrlib.repofmt.pack_repo import Packer
2685
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2686
self.source, self.source._format, self.target, self.target._format)
2687
self.count_copied = 0
2688
if revision_id is None:
2690
# everything to do - use pack logic
2691
# to fetch from all packs to one without
2692
# inventory parsing etc, IFF nothing to be copied is in the target.
2694
revision_ids = self.source.all_revision_ids()
2695
revision_keys = [(revid,) for revid in revision_ids]
2696
index = self.target._pack_collection.revision_index.combined_index
2697
present_revision_ids = set(item[1][0] for item in
2698
index.iter_entries(revision_keys))
2699
revision_ids = set(revision_ids) - present_revision_ids
2700
# implementing the TODO will involve:
2701
# - detecting when all of a pack is selected
2702
# - avoiding as much as possible pre-selection, so the
2703
# more-core routines such as create_pack_from_packs can filter in
2704
# a just-in-time fashion. (though having a HEADS list on a
2705
# repository might make this a lot easier, because we could
2706
# sensibly detect 'new revisions' without doing a full index scan.
2707
elif _mod_revision.is_null(revision_id):
2712
revision_ids = self.search_missing_revision_ids(revision_id,
2713
find_ghosts=find_ghosts).get_keys()
2714
except errors.NoSuchRevision:
2715
raise errors.InstallFailed([revision_id])
2716
if len(revision_ids) == 0:
2718
packs = self.source._pack_collection.all_packs()
2719
pack = Packer(self.target._pack_collection, packs, '.fetch',
2720
revision_ids).pack()
2721
if pack is not None:
2722
self.target._pack_collection._save_pack_names()
2723
# Trigger an autopack. This may duplicate effort as we've just done
2724
# a pack creation, but for now it is simpler to think about as
2725
# 'upload data, then repack if needed'.
2726
self.target._pack_collection.autopack()
2727
return (pack.get_revision_count(), [])
2732
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2733
"""See InterRepository.missing_revision_ids().
2735
:param find_ghosts: Find ghosts throughout the ancestry of
2738
if not find_ghosts and revision_id is not None:
2739
return self._walk_to_common_revisions([revision_id])
2740
elif revision_id is not None:
2741
source_ids = self.source.get_ancestry(revision_id)
2742
if source_ids[0] is not None:
2743
raise AssertionError()
2746
source_ids = self.source.all_revision_ids()
2747
# source_ids is the worst possible case we may need to pull.
2748
# now we want to filter source_ids against what we actually
2749
# have in target, but don't try to check for existence where we know
2750
# we do not have a revision as that would be pointless.
2751
target_ids = set(self.target.all_revision_ids())
2752
result_set = set(source_ids).difference(target_ids)
2753
return self.source.revision_ids_to_search_result(result_set)
2756
class InterModel1and2(InterRepository):
2759
def _get_repo_format_to_test(self):
2763
def is_compatible(source, target):
2764
if not source.supports_rich_root() and target.supports_rich_root():
2770
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2771
"""See InterRepository.fetch()."""
2772
from bzrlib.fetch import Model1toKnit2Fetcher
2773
f = Model1toKnit2Fetcher(to_repository=self.target,
2774
from_repository=self.source,
2775
last_revision=revision_id,
2776
pb=pb, find_ghosts=find_ghosts)
2777
return f.count_copied, f.failed_revisions
2780
def copy_content(self, revision_id=None):
2781
"""Make a complete copy of the content in self into destination.
2783
This is a destructive operation! Do not use it on existing
2786
:param revision_id: Only copy the content needed to construct
2787
revision_id and its parents.
2790
self.target.set_make_working_trees(self.source.make_working_trees())
2791
except NotImplementedError:
2793
# but don't bother fetching if we have the needed data now.
2794
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
2795
self.target.has_revision(revision_id)):
2797
self.target.fetch(self.source, revision_id=revision_id)
2800
class InterKnit1and2(InterKnitRepo):
2803
def _get_repo_format_to_test(self):
2807
def is_compatible(source, target):
2808
"""Be compatible with Knit1 source and Knit3 target"""
2809
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit3
2811
from bzrlib.repofmt.knitrepo import (RepositoryFormatKnit1,
2812
RepositoryFormatKnit3)
2813
from bzrlib.repofmt.pack_repo import (
2814
RepositoryFormatKnitPack1,
2815
RepositoryFormatKnitPack3,
2816
RepositoryFormatPackDevelopment0,
2817
RepositoryFormatPackDevelopment0Subtree,
2820
RepositoryFormatKnit1,
2821
RepositoryFormatKnitPack1,
2822
RepositoryFormatPackDevelopment0,
2825
RepositoryFormatKnit3,
2826
RepositoryFormatKnitPack3,
2827
RepositoryFormatPackDevelopment0Subtree,
2829
return (isinstance(source._format, nosubtrees) and
2830
isinstance(target._format, subtrees))
2831
except AttributeError:
2835
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2836
"""See InterRepository.fetch()."""
2837
from bzrlib.fetch import Knit1to2Fetcher
2838
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2839
self.source, self.source._format, self.target,
2840
self.target._format)
2841
f = Knit1to2Fetcher(to_repository=self.target,
2842
from_repository=self.source,
2843
last_revision=revision_id,
2844
pb=pb, find_ghosts=find_ghosts)
2845
return f.count_copied, f.failed_revisions
2848
class InterDifferingSerializer(InterKnitRepo):
2851
def _get_repo_format_to_test(self):
2855
def is_compatible(source, target):
2856
"""Be compatible with Knit2 source and Knit3 target"""
2857
if source.supports_rich_root() != target.supports_rich_root():
2859
# Ideally, we'd support fetching if the source had no tree references
2860
# even if it supported them...
2861
if (getattr(source, '_format.supports_tree_reference', False) and
2862
not getattr(target, '_format.supports_tree_reference', False)):
2867
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2868
"""See InterRepository.fetch()."""
2869
revision_ids = self.target.search_missing_revision_ids(self.source,
2870
revision_id, find_ghosts=find_ghosts).get_keys()
2871
revision_ids = tsort.topo_sort(
2872
self.source.get_graph().get_parent_map(revision_ids))
2873
def revisions_iterator():
2874
for current_revision_id in revision_ids:
2875
revision = self.source.get_revision(current_revision_id)
2876
tree = self.source.revision_tree(current_revision_id)
2878
signature = self.source.get_signature_text(
2879
current_revision_id)
2880
except errors.NoSuchRevision:
2882
yield revision, tree, signature
2884
my_pb = ui.ui_factory.nested_progress_bar()
2889
install_revisions(self.target, revisions_iterator(),
2890
len(revision_ids), pb)
2892
if my_pb is not None:
2894
return len(revision_ids), 0
2897
class InterRemoteToOther(InterRepository):
2899
def __init__(self, source, target):
2900
InterRepository.__init__(self, source, target)
2901
self._real_inter = None
2904
def is_compatible(source, target):
2905
if not isinstance(source, remote.RemoteRepository):
2907
# Is source's model compatible with target's model?
2908
source._ensure_real()
2909
real_source = source._real_repository
2910
if isinstance(real_source, remote.RemoteRepository):
2911
raise NotImplementedError(
2912
"We don't support remote repos backed by remote repos yet.")
2913
return InterRepository._same_model(real_source, target)
2916
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2917
"""See InterRepository.fetch()."""
2918
from bzrlib.fetch import RemoteToOtherFetcher
2919
mutter("Using fetch logic to copy between %s(remote) and %s(%s)",
2920
self.source, self.target, self.target._format)
2921
# TODO: jam 20070210 This should be an assert, not a translate
2922
revision_id = osutils.safe_revision_id(revision_id)
2923
f = RemoteToOtherFetcher(to_repository=self.target,
2924
from_repository=self.source,
2925
last_revision=revision_id,
2926
pb=pb, find_ghosts=find_ghosts)
2927
return f.count_copied, f.failed_revisions
2930
def _get_repo_format_to_test(self):
2934
class InterOtherToRemote(InterRepository):
2936
def __init__(self, source, target):
2937
InterRepository.__init__(self, source, target)
2938
self._real_inter = None
2941
def is_compatible(source, target):
2942
if isinstance(target, remote.RemoteRepository):
2946
def _ensure_real_inter(self):
2947
if self._real_inter is None:
2948
self.target._ensure_real()
2949
real_target = self.target._real_repository
2950
self._real_inter = InterRepository.get(self.source, real_target)
2952
def copy_content(self, revision_id=None):
2953
self._ensure_real_inter()
2954
self._real_inter.copy_content(revision_id=revision_id)
2956
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2957
self._ensure_real_inter()
2958
self._real_inter.fetch(revision_id=revision_id, pb=pb,
2959
find_ghosts=find_ghosts)
2962
def _get_repo_format_to_test(self):
2966
InterRepository.register_optimiser(InterDifferingSerializer)
2967
InterRepository.register_optimiser(InterSameDataRepository)
2968
InterRepository.register_optimiser(InterWeaveRepo)
2969
InterRepository.register_optimiser(InterKnitRepo)
2970
InterRepository.register_optimiser(InterModel1and2)
2971
InterRepository.register_optimiser(InterKnit1and2)
2972
InterRepository.register_optimiser(InterPackRepo)
2973
InterRepository.register_optimiser(InterRemoteToOther)
2974
InterRepository.register_optimiser(InterOtherToRemote)
1712
2977
class CopyConverter(object):
1713
2978
"""A repository conversion tool which just performs a copy of the content.
1715
2980
This is slow but quite reliable.