202
191
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,
193
def _heads(self, file_id, revision_ids):
194
"""Calculate the graph heads for revision_ids in the graph of file_id.
196
This can use either a per-file graph or a global revision graph as we
197
have an identity relationship between the two graphs.
199
return self.__heads(revision_ids)
201
def _check_root(self, ie, parent_invs, tree):
202
"""Helper for record_entry_contents.
204
:param ie: An entry being added.
205
:param parent_invs: The inventories of the parent revisions of the
207
:param tree: The tree that is being committed.
209
# In this revision format, root entries have no knit or weave When
210
# serializing out to disk and back in root.revision is always
212
ie.revision = self._new_revision_id
214
def _get_delta(self, ie, basis_inv, path):
215
"""Get a delta against the basis inventory for ie."""
216
if ie.file_id not in basis_inv:
218
return (None, path, ie.file_id, ie)
219
elif ie != basis_inv[ie.file_id]:
221
# TODO: avoid tis id2path call.
222
return (basis_inv.id2path(ie.file_id), path, ie.file_id, ie)
227
def record_entry_contents(self, ie, parent_invs, path, tree,
229
"""Record the content of ie from tree into the commit if needed.
231
Side effect: sets ie.revision when unchanged
233
:param ie: An inventory entry present in the commit.
234
:param parent_invs: The inventories of the parent revisions of the
236
:param path: The path the entry is at in the tree.
237
:param tree: The tree which contains this entry and should be used to
239
:param content_summary: Summary data from the tree about the paths
240
content - stat, length, exec, sha/link target. This is only
241
accessed when the entry has a revision of None - that is when it is
242
a candidate to commit.
243
:return: A tuple (change_delta, version_recorded). change_delta is
244
an inventory_delta change for this entry against the basis tree of
245
the commit, or None if no change occured against the basis tree.
246
version_recorded is True if a new version of the entry has been
247
recorded. For instance, committing a merge where a file was only
248
changed on the other side will return (delta, False).
250
if self.new_inventory.root is None:
251
if ie.parent_id is not None:
252
raise errors.RootMissing()
253
self._check_root(ie, parent_invs, tree)
254
if ie.revision is None:
255
kind = content_summary[0]
257
# ie is carried over from a prior commit
259
# XXX: repository specific check for nested tree support goes here - if
260
# the repo doesn't want nested trees we skip it ?
261
if (kind == 'tree-reference' and
262
not self.repository._format.supports_tree_reference):
263
# mismatch between commit builder logic and repository:
264
# this needs the entry creation pushed down into the builder.
265
raise NotImplementedError('Missing repository subtree support.')
266
self.new_inventory.add(ie)
268
# TODO: slow, take it out of the inner loop.
270
basis_inv = parent_invs[0]
272
basis_inv = Inventory(root_id=None)
274
# ie.revision is always None if the InventoryEntry is considered
275
# for committing. We may record the previous parents revision if the
276
# content is actually unchanged against a sole head.
277
if ie.revision is not None:
278
if not self._versioned_root and path == '':
279
# repositories that do not version the root set the root's
280
# revision to the new commit even when no change occurs, and
281
# this masks when a change may have occurred against the basis,
282
# so calculate if one happened.
283
if ie.file_id in basis_inv:
284
delta = (basis_inv.id2path(ie.file_id), path,
288
delta = (None, path, ie.file_id, ie)
291
# we don't need to commit this, because the caller already
292
# determined that an existing revision of this file is
294
return None, (ie.revision == self._new_revision_id)
295
# XXX: Friction: parent_candidates should return a list not a dict
296
# so that we don't have to walk the inventories again.
297
parent_candiate_entries = ie.parent_candidates(parent_invs)
298
head_set = self._heads(ie.file_id, parent_candiate_entries.keys())
300
for inv in parent_invs:
301
if ie.file_id in inv:
302
old_rev = inv[ie.file_id].revision
303
if old_rev in head_set:
304
heads.append(inv[ie.file_id].revision)
305
head_set.remove(inv[ie.file_id].revision)
308
# now we check to see if we need to write a new record to the
310
# We write a new entry unless there is one head to the ancestors, and
311
# the kind-derived content is unchanged.
313
# Cheapest check first: no ancestors, or more the one head in the
314
# ancestors, we write a new node.
318
# There is a single head, look it up for comparison
319
parent_entry = parent_candiate_entries[heads[0]]
320
# if the non-content specific data has changed, we'll be writing a
322
if (parent_entry.parent_id != ie.parent_id or
323
parent_entry.name != ie.name):
325
# now we need to do content specific checks:
327
# if the kind changed the content obviously has
328
if kind != parent_entry.kind:
331
assert content_summary[2] is not None, \
332
"Files must not have executable = None"
334
if (# if the file length changed we have to store:
335
parent_entry.text_size != content_summary[1] or
336
# if the exec bit has changed we have to store:
337
parent_entry.executable != content_summary[2]):
339
elif parent_entry.text_sha1 == content_summary[3]:
340
# all meta and content is unchanged (using a hash cache
341
# hit to check the sha)
342
ie.revision = parent_entry.revision
343
ie.text_size = parent_entry.text_size
344
ie.text_sha1 = parent_entry.text_sha1
345
ie.executable = parent_entry.executable
346
return self._get_delta(ie, basis_inv, path), False
348
# Either there is only a hash change(no hash cache entry,
349
# or same size content change), or there is no change on
351
# Provide the parent's hash to the store layer, so that the
352
# content is unchanged we will not store a new node.
353
nostore_sha = parent_entry.text_sha1
355
# We want to record a new node regardless of the presence or
356
# absence of a content change in the file.
358
ie.executable = content_summary[2]
359
lines = tree.get_file(ie.file_id, path).readlines()
361
ie.text_sha1, ie.text_size = self._add_text_to_weave(
362
ie.file_id, lines, heads, nostore_sha)
363
except errors.ExistingContent:
364
# Turns out that the file content was unchanged, and we were
365
# only going to store a new node if it was changed. Carry over
367
ie.revision = parent_entry.revision
368
ie.text_size = parent_entry.text_size
369
ie.text_sha1 = parent_entry.text_sha1
370
ie.executable = parent_entry.executable
371
return self._get_delta(ie, basis_inv, path), False
372
elif kind == 'directory':
374
# all data is meta here, nothing specific to directory, so
376
ie.revision = parent_entry.revision
377
return self._get_delta(ie, basis_inv, path), False
379
self._add_text_to_weave(ie.file_id, lines, heads, None)
380
elif kind == 'symlink':
381
current_link_target = content_summary[3]
383
# symlink target is not generic metadata, check if it has
385
if current_link_target != parent_entry.symlink_target:
388
# unchanged, carry over.
389
ie.revision = parent_entry.revision
390
ie.symlink_target = parent_entry.symlink_target
391
return self._get_delta(ie, basis_inv, path), False
392
ie.symlink_target = current_link_target
394
self._add_text_to_weave(ie.file_id, lines, heads, None)
395
elif kind == 'tree-reference':
397
if content_summary[3] != parent_entry.reference_revision:
400
# unchanged, carry over.
401
ie.reference_revision = parent_entry.reference_revision
402
ie.revision = parent_entry.revision
403
return self._get_delta(ie, basis_inv, path), False
404
ie.reference_revision = content_summary[3]
406
self._add_text_to_weave(ie.file_id, lines, heads, None)
408
raise NotImplementedError('unknown kind')
409
ie.revision = self._new_revision_id
410
return self._get_delta(ie, basis_inv, path), True
412
def _add_text_to_weave(self, file_id, new_lines, parents, nostore_sha):
413
versionedfile = self.repository.weave_store.get_weave_or_empty(
414
file_id, self.repository.get_transaction())
415
# Don't change this to add_lines - add_lines_with_ghosts is cheaper
416
# than add_lines, and allows committing when a parent is ghosted for
418
# Note: as we read the content directly from the tree, we know its not
419
# been turned into unicode or badly split - but a broken tree
420
# implementation could give us bad output from readlines() so this is
421
# not a guarantee of safety. What would be better is always checking
422
# the content during test suite execution. RBC 20070912
423
return versionedfile.add_lines_with_ghosts(
424
self._new_revision_id, parents, new_lines,
425
nostore_sha=nostore_sha, random_id=self.random_revid,
426
check_content=False)[0:2]
429
class RootCommitBuilder(CommitBuilder):
430
"""This commitbuilder actually records the root id"""
432
# the root entry gets versioned properly by this builder.
433
_versioned_root = True
435
def _check_root(self, ie, parent_invs, tree):
436
"""Helper for record_entry_contents.
438
:param ie: An entry being added.
439
:param parent_invs: The inventories of the parent revisions of the
441
:param tree: The tree that is being committed.
247
445
######################################################################
251
class Repository(_RelockDebugMixin, controldir.ControlComponent):
448
class Repository(object):
252
449
"""Repository holding history for one or more branches.
254
451
The repository holds and retrieves historical information including
255
452
revisions and file history. It's normally accessed only by the Branch,
256
453
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.
455
The Repository builds on top of Stores and a Transport, which respectively
456
describe the disk data format and the way of accessing the (possibly
262
def abort_write_group(self, suppress_errors=False):
460
# What class to use for a CommitBuilder. Often its simpler to change this
461
# in a Repository class subclass rather than to override
462
# get_commit_builder.
463
_commit_builder_class = CommitBuilder
464
# The search regex used by xml based repositories to determine what things
465
# where changed in a single commit.
466
_file_ids_altered_regex = lazy_regex.lazy_compile(
467
r'file_id="(?P<file_id>[^"]+)"'
468
r'.* revision="(?P<revision_id>[^"]+)"'
471
def abort_write_group(self):
263
472
"""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
474
:seealso: start_write_group.
271
476
if self._write_group is not self.get_transaction():
272
477
# 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)
478
raise errors.BzrError('mismatched lock context and write group.')
479
self._abort_write_group()
290
480
self._write_group = None
292
482
def _abort_write_group(self):
293
483
"""Template method for per-repository write group cleanup.
295
This is called during abort before the write group is considered to be
485
This is called during abort before the write group is considered to be
296
486
finished and should cleanup any internal state accrued during the write
297
487
group. There is no requirement that data handed to the repository be
298
488
*not* made available - this is not a rollback - but neither should any
828
1064
raise NotImplementedError(self.has_revisions)
1066
return self._revision_store.has_revision_id(revision_id,
1067
self.get_transaction())
830
1069
@needs_read_lock
831
1070
def get_revision(self, revision_id):
832
1071
"""Return the Revision object for a named revision."""
833
1072
return self.get_revisions([revision_id])[0]
835
1075
def get_revision_reconcile(self, revision_id):
836
1076
"""'reconcile' helper routine that allows access to a revision always.
838
1078
This variant of get_revision does not cross check the weave graph
839
1079
against the revision one as get_revision does: but it should only
840
1080
be used by reconcile, or reconcile-alike commands that are correcting
841
1081
or testing the revision graph.
843
raise NotImplementedError(self.get_revision_reconcile)
1083
return self._get_revisions([revision_id])[0]
845
1086
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):
1087
"""Get many revisions at once."""
1088
return self._get_revisions(revision_ids)
1091
def _get_revisions(self, revision_ids):
1092
"""Core work logic to get many revisions without sanity checks."""
1093
for rev_id in revision_ids:
1094
if not rev_id or not isinstance(rev_id, basestring):
1095
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1096
revs = self._revision_store.get_revisions(revision_ids,
1097
self.get_transaction())
1099
assert not isinstance(rev.revision_id, unicode)
1100
for parent_id in rev.parent_ids:
1101
assert not isinstance(parent_id, unicode)
1105
def get_revision_xml(self, revision_id):
1106
# TODO: jam 20070210 This shouldn't be necessary since get_revision
1107
# would have already do it.
1108
# TODO: jam 20070210 Just use _serializer.write_revision_to_string()
1109
rev = self.get_revision(revision_id)
1110
rev_tmp = StringIO()
1111
# the current serializer..
1112
self._revision_store._serializer.write_revision(rev, rev_tmp)
1114
return rev_tmp.getvalue()
1116
def get_deltas_for_revisions(self, revisions):
854
1117
"""Produce a generator of revision deltas.
856
1119
Note that the input is a sequence of REVISIONS, not revision_ids.
857
1120
Trees will be held in memory until the generator exits.
858
1121
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
1123
required_trees = set()
866
1124
for revision in revisions:
867
1125
required_trees.add(revision.revision_id)
868
1126
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
1127
trees = dict((t.get_revision_id(), t) for
1128
t in self.revision_trees(required_trees))
885
1129
for revision in revisions:
886
1130
if not revision.parent_ids:
887
old_tree = self.revision_tree(_mod_revision.NULL_REVISION)
1131
old_tree = self.revision_tree(None)
889
1133
old_tree = trees[revision.parent_ids[0]]
890
1134
yield trees[revision.revision_id].changes_from(old_tree)
892
1136
@needs_read_lock
893
def get_revision_delta(self, revision_id, specific_fileids=None):
1137
def get_revision_delta(self, revision_id):
894
1138
"""Return the delta for one revision.
896
1140
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
1143
r = self.get_revision(revision_id)
904
return list(self.get_deltas_for_revisions([r],
905
specific_fileids=specific_fileids))[0]
1144
return list(self.get_deltas_for_revisions([r]))[0]
907
1146
@needs_write_lock
908
1147
def store_revision_signature(self, gpg_strategy, plaintext, revision_id):
909
1148
signature = gpg_strategy.sign(plaintext)
910
1149
self.add_signature_text(revision_id, signature)
912
1152
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)
1153
self._revision_store.add_revision_signature_text(revision_id,
1155
self.get_transaction())
1157
def find_text_key_references(self):
1158
"""Find the text key references within the repository.
1160
:return: a dictionary mapping (file_id, revision_id) tuples to altered file-ids to an iterable of
1161
revision_ids. Each altered file-ids has the exact revision_ids that
1162
altered it listed explicitly.
1163
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1164
to whether they were referred to by the inventory of the
1165
revision_id that they contain. The inventory texts from all present
1166
revision ids are assessed to generate this report.
1168
revision_ids = self.all_revision_ids()
1169
w = self.get_inventory_weave()
1170
pb = ui.ui_factory.nested_progress_bar()
1172
return self._find_text_key_references_from_xml_inventory_lines(
1173
w.iter_lines_added_or_present_in_versions(revision_ids, pb=pb))
1177
def _find_text_key_references_from_xml_inventory_lines(self,
1179
"""Core routine for extracting references to texts from inventories.
1181
This performs the translation of xml lines to revision ids.
1183
:param line_iterator: An iterator of lines, origin_version_id
1184
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1185
to whether they were referred to by the inventory of the
1186
revision_id that they contain. Note that if that revision_id was
1187
not part of the line_iterator's output then False will be given -
1188
even though it may actually refer to that key.
1190
if not self._serializer.support_altered_by_hack:
1191
raise AssertionError(
1192
"_find_text_key_references_from_xml_inventory_lines only "
1193
"supported for branches which store inventory as unnested xml"
1194
", not on %r" % self)
1197
# this code needs to read every new line in every inventory for the
1198
# inventories [revision_ids]. Seeing a line twice is ok. Seeing a line
1199
# not present in one of those inventories is unnecessary but not
1200
# harmful because we are filtering by the revision id marker in the
1201
# inventory lines : we only select file ids altered in one of those
1202
# revisions. We don't need to see all lines in the inventory because
1203
# only those added in an inventory in rev X can contain a revision=X
1205
unescape_revid_cache = {}
1206
unescape_fileid_cache = {}
1208
# jam 20061218 In a big fetch, this handles hundreds of thousands
1209
# of lines, so it has had a lot of inlining and optimizing done.
1210
# Sorry that it is a little bit messy.
1211
# Move several functions to be local variables, since this is a long
1213
search = self._file_ids_altered_regex.search
1214
unescape = _unescape_xml
1215
setdefault = result.setdefault
1216
for line, version_id in line_iterator:
1217
match = search(line)
1220
# One call to match.group() returning multiple items is quite a
1221
# bit faster than 2 calls to match.group() each returning 1
1222
file_id, revision_id = match.group('file_id', 'revision_id')
1224
# Inlining the cache lookups helps a lot when you make 170,000
1225
# lines and 350k ids, versus 8.4 unique ids.
1226
# Using a cache helps in 2 ways:
1227
# 1) Avoids unnecessary decoding calls
1228
# 2) Re-uses cached strings, which helps in future set and
1230
# (2) is enough that removing encoding entirely along with
1231
# the cache (so we are using plain strings) results in no
1232
# performance improvement.
1234
revision_id = unescape_revid_cache[revision_id]
1236
unescaped = unescape(revision_id)
1237
unescape_revid_cache[revision_id] = unescaped
1238
revision_id = unescaped
1240
# Note that unconditionally unescaping means that we deserialise
1241
# every fileid, which for general 'pull' is not great, but we don't
1242
# really want to have some many fulltexts that this matters anyway.
1245
file_id = unescape_fileid_cache[file_id]
1247
unescaped = unescape(file_id)
1248
unescape_fileid_cache[file_id] = unescaped
1251
key = (file_id, revision_id)
1252
setdefault(key, False)
1253
if revision_id == version_id:
1257
def _find_file_ids_from_xml_inventory_lines(self, line_iterator,
1259
"""Helper routine for fileids_altered_by_revision_ids.
1261
This performs the translation of xml lines to revision ids.
1263
:param line_iterator: An iterator of lines, origin_version_id
1264
:param revision_ids: The revision ids to filter for. This should be a
1265
set or other type which supports efficient __contains__ lookups, as
1266
the revision id from each parsed line will be looked up in the
1267
revision_ids filter.
1268
:return: a dictionary mapping altered file-ids to an iterable of
1269
revision_ids. Each altered file-ids has the exact revision_ids that
1270
altered it listed explicitly.
1273
setdefault = result.setdefault
1274
for file_id, revision_id in \
1275
self._find_text_key_references_from_xml_inventory_lines(
1276
line_iterator).iterkeys():
1277
# once data is all ensured-consistent; then this is
1278
# if revision_id == version_id
1279
if revision_id in revision_ids:
1280
setdefault(file_id, set()).add(revision_id)
1283
def fileids_altered_by_revision_ids(self, revision_ids, _inv_weave=None):
1284
"""Find the file ids and versions affected by revisions.
1286
:param revisions: an iterable containing revision ids.
1287
:param _inv_weave: The inventory weave from this repository or None.
1288
If None, the inventory weave will be opened automatically.
1289
:return: a dictionary mapping altered file-ids to an iterable of
1290
revision_ids. Each altered file-ids has the exact revision_ids that
1291
altered it listed explicitly.
1293
selected_revision_ids = set(revision_ids)
1294
w = _inv_weave or self.get_inventory_weave()
1295
pb = ui.ui_factory.nested_progress_bar()
1297
return self._find_file_ids_from_xml_inventory_lines(
1298
w.iter_lines_added_or_present_in_versions(
1299
selected_revision_ids, pb=pb),
1300
selected_revision_ids)
933
1304
def iter_files_bytes(self, desired_files):
934
1305
"""Iterate through file versions.
941
1312
uniquely identify the file version in the caller's context. (Examples:
942
1313
an index number or a TreeTransform trans_id.)
1315
bytes_iterator is an iterable of bytestrings for the file. The
1316
kind of iterable and length of the bytestrings are unspecified, but for
1317
this implementation, it is a list of lines produced by
1318
VersionedFile.get_lines().
944
1320
: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])
1323
transaction = self.get_transaction()
1324
for file_id, revision_id, callable_data in desired_files:
1326
weave = self.weave_store.get_weave(file_id, transaction)
1327
except errors.NoSuchFile:
1328
raise errors.NoSuchIdInRepository(self, file_id)
1329
yield callable_data, weave.get_lines(revision_id)
1331
def _generate_text_key_index(self, text_key_references=None,
1333
"""Generate a new text key index for the repository.
1335
This is an expensive function that will take considerable time to run.
1337
:return: A dict mapping text keys ((file_id, revision_id) tuples) to a
1338
list of parents, also text keys. When a given key has no parents,
1339
the parents list will be [NULL_REVISION].
1341
# All revisions, to find inventory parents.
1342
if ancestors is None:
1343
graph = self.get_graph()
1344
ancestors = graph.get_parent_map(self.all_revision_ids())
1345
if text_key_references is None:
1346
text_key_references = self.find_text_key_references()
1347
pb = ui.ui_factory.nested_progress_bar()
1349
return self._do_generate_text_key_index(ancestors,
1350
text_key_references, pb)
1354
def _do_generate_text_key_index(self, ancestors, text_key_references, pb):
1355
"""Helper for _generate_text_key_index to avoid deep nesting."""
1356
revision_order = tsort.topo_sort(ancestors)
1357
invalid_keys = set()
1359
for revision_id in revision_order:
1360
revision_keys[revision_id] = set()
1361
text_count = len(text_key_references)
1362
# a cache of the text keys to allow reuse; costs a dict of all the
1363
# keys, but saves a 2-tuple for every child of a given key.
1365
for text_key, valid in text_key_references.iteritems():
1367
invalid_keys.add(text_key)
1369
revision_keys[text_key[1]].add(text_key)
1370
text_key_cache[text_key] = text_key
1371
del text_key_references
1373
text_graph = graph.Graph(graph.DictParentsProvider(text_index))
1374
NULL_REVISION = _mod_revision.NULL_REVISION
1375
# Set a cache with a size of 10 - this suffices for bzr.dev but may be
1376
# too small for large or very branchy trees. However, for 55K path
1377
# trees, it would be easy to use too much memory trivially. Ideally we
1378
# could gauge this by looking at available real memory etc, but this is
1379
# always a tricky proposition.
1380
inventory_cache = lru_cache.LRUCache(10)
1381
batch_size = 10 # should be ~150MB on a 55K path tree
1382
batch_count = len(revision_order) / batch_size + 1
1384
pb.update("Calculating text parents.", processed_texts, text_count)
1385
for offset in xrange(batch_count):
1386
to_query = revision_order[offset * batch_size:(offset + 1) *
1390
for rev_tree in self.revision_trees(to_query):
1391
revision_id = rev_tree.get_revision_id()
1392
parent_ids = ancestors[revision_id]
1393
for text_key in revision_keys[revision_id]:
1394
pb.update("Calculating text parents.", processed_texts)
1395
processed_texts += 1
1396
candidate_parents = []
1397
for parent_id in parent_ids:
1398
parent_text_key = (text_key[0], parent_id)
1400
check_parent = parent_text_key not in \
1401
revision_keys[parent_id]
1403
# the parent parent_id is a ghost:
1404
check_parent = False
1405
# truncate the derived graph against this ghost.
1406
parent_text_key = None
1408
# look at the parent commit details inventories to
1409
# determine possible candidates in the per file graph.
1412
inv = inventory_cache[parent_id]
1414
inv = self.revision_tree(parent_id).inventory
1415
inventory_cache[parent_id] = inv
1416
parent_entry = inv._byid.get(text_key[0], None)
1417
if parent_entry is not None:
1419
text_key[0], parent_entry.revision)
1421
parent_text_key = None
1422
if parent_text_key is not None:
1423
candidate_parents.append(
1424
text_key_cache[parent_text_key])
1425
parent_heads = text_graph.heads(candidate_parents)
1426
new_parents = list(parent_heads)
1427
new_parents.sort(key=lambda x:candidate_parents.index(x))
1428
if new_parents == []:
1429
new_parents = [NULL_REVISION]
1430
text_index[text_key] = new_parents
1432
for text_key in invalid_keys:
1433
text_index[text_key] = [NULL_REVISION]
1436
def item_keys_introduced_by(self, revision_ids, _files_pb=None):
1437
"""Get an iterable listing the keys of all the data introduced by a set
1440
The keys will be ordered so that the corresponding items can be safely
1441
fetched and inserted in that order.
1443
:returns: An iterable producing tuples of (knit-kind, file-id,
1444
versions). knit-kind is one of 'file', 'inventory', 'signatures',
1445
'revisions'. file-id is None unless knit-kind is 'file'.
1447
# XXX: it's a bit weird to control the inventory weave caching in this
1448
# generator. Ideally the caching would be done in fetch.py I think. Or
1449
# maybe this generator should explicitly have the contract that it
1450
# should not be iterated until the previously yielded item has been
1452
inv_w = self.get_inventory_weave()
1454
# file ids that changed
1455
file_ids = self.fileids_altered_by_revision_ids(revision_ids, inv_w)
1457
num_file_ids = len(file_ids)
1458
for file_id, altered_versions in file_ids.iteritems():
1459
if _files_pb is not None:
1460
_files_pb.update("fetch texts", count, num_file_ids)
1462
yield ("file", file_id, altered_versions)
1463
# We're done with the files_pb. Note that it finished by the caller,
1464
# just as it was created by the caller.
1468
yield ("inventory", None, revision_ids)
1471
revisions_with_signatures = set()
1472
for rev_id in revision_ids:
1474
self.get_signature_text(rev_id)
1475
except errors.NoSuchRevision:
1479
revisions_with_signatures.add(rev_id)
1480
yield ("signatures", None, revisions_with_signatures)
1483
yield ("revisions", None, revision_ids)
1486
def get_inventory_weave(self):
1487
return self.control_weaves.get_weave('inventory',
1488
self.get_transaction())
1491
def get_inventory(self, revision_id):
1492
"""Get Inventory object by revision id."""
1493
return self.iter_inventories([revision_id]).next()
1495
def iter_inventories(self, revision_ids):
1496
"""Get many inventories by revision_ids.
1498
This will buffer some or all of the texts used in constructing the
1499
inventories in memory, but will only parse a single inventory at a
1502
:return: An iterator of inventories.
1504
assert None not in revision_ids
1505
assert _mod_revision.NULL_REVISION not in revision_ids
1506
return self._iter_inventories(revision_ids)
1508
def _iter_inventories(self, revision_ids):
1509
"""single-document based inventory iteration."""
1510
texts = self.get_inventory_weave().get_texts(revision_ids)
1511
for text, revision_id in zip(texts, revision_ids):
1512
yield self.deserialise_inventory(revision_id, text)
1514
def deserialise_inventory(self, revision_id, xml):
1515
"""Transform the xml into an inventory object.
1517
:param revision_id: The expected revision id of the inventory.
1518
:param xml: A serialised inventory.
1520
result = self._serializer.read_inventory_from_string(xml, revision_id)
1521
if result.revision_id != revision_id:
1522
raise AssertionError('revision id mismatch %s != %s' % (
1523
result.revision_id, revision_id))
1526
def serialise_inventory(self, inv):
1527
return self._serializer.write_inventory_to_string(inv)
1529
def _serialise_inventory_to_lines(self, inv):
1530
return self._serializer.write_inventory_to_lines(inv)
1532
def get_serializer_format(self):
1533
return self._serializer.format_num
1536
def get_inventory_xml(self, revision_id):
1537
"""Get inventory XML as a file object."""
1539
assert isinstance(revision_id, str), type(revision_id)
1540
iw = self.get_inventory_weave()
1541
return iw.get_text(revision_id)
1543
raise errors.HistoryMissing(self, 'inventory', revision_id)
1546
def get_inventory_sha1(self, revision_id):
1547
"""Return the sha1 hash of the inventory entry
1549
return self.get_revision(revision_id).inventory_sha1
1552
@deprecated_method(symbol_versioning.one_four)
1553
def get_revision_graph(self, revision_id=None):
1554
"""Return a dictionary containing the revision graph.
1556
NB: This method should not be used as it accesses the entire graph all
1557
at once, which is much more data than most operations should require.
1559
:param revision_id: The revision_id to get a graph from. If None, then
1560
the entire revision graph is returned. This is a deprecated mode of
1561
operation and will be removed in the future.
1562
:return: a dictionary of revision_id->revision_parents_list.
1564
raise NotImplementedError(self.get_revision_graph)
1567
@deprecated_method(symbol_versioning.one_three)
1568
def get_revision_graph_with_ghosts(self, revision_ids=None):
1569
"""Return a graph of the revisions with ghosts marked as applicable.
1571
:param revision_ids: an iterable of revisions to graph or None for all.
1572
:return: a Graph object with the graph reachable from revision_ids.
1574
if 'evil' in debug.debug_flags:
1576
"get_revision_graph_with_ghosts scales with size of history.")
1577
result = deprecated_graph.Graph()
1578
if not revision_ids:
1579
pending = set(self.all_revision_ids())
1582
pending = set(revision_ids)
1583
# special case NULL_REVISION
1584
if _mod_revision.NULL_REVISION in pending:
1585
pending.remove(_mod_revision.NULL_REVISION)
1586
required = set(pending)
1589
revision_id = pending.pop()
1591
rev = self.get_revision(revision_id)
1592
except errors.NoSuchRevision:
1593
if revision_id in required:
1596
result.add_ghost(revision_id)
1598
for parent_id in rev.parent_ids:
1599
# is this queued or done ?
1600
if (parent_id not in pending and
1601
parent_id not in done):
1603
pending.add(parent_id)
1604
result.add_node(revision_id, rev.parent_ids)
1605
done.add(revision_id)
1608
def iter_reverse_revision_history(self, revision_id):
1609
"""Iterate backwards through revision ids in the lefthand history
1611
:param revision_id: The revision id to start with. All its lefthand
1612
ancestors will be traversed.
1614
graph = self.get_graph()
1615
next_id = revision_id
1617
if next_id in (None, _mod_revision.NULL_REVISION):
1620
# Note: The following line may raise KeyError in the event of
1621
# truncated history. We decided not to have a try:except:raise
1622
# RevisionNotPresent here until we see a use for it, because of the
1623
# cost in an inner loop that is by its very nature O(history).
1624
# Robert Collins 20080326
1625
parents = graph.get_parent_map([next_id])[next_id]
1626
if len(parents) == 0:
1629
next_id = parents[0]
1632
def get_revision_inventory(self, revision_id):
1633
"""Return inventory of a past revision."""
1634
# TODO: Unify this with get_inventory()
1635
# bzr 0.0.6 and later imposes the constraint that the inventory_id
1636
# must be the same as its revision, so this is trivial.
1637
if revision_id is None:
1638
# This does not make sense: if there is no revision,
1639
# then it is the current tree inventory surely ?!
1640
# and thus get_root_id() is something that looks at the last
1641
# commit on the branch, and the get_root_id is an inventory check.
1642
raise NotImplementedError
1643
# return Inventory(self.get_root_id())
1645
return self.get_inventory(revision_id)
984
1648
def is_shared(self):
985
1649
"""Return True if this repository is flagged as a shared repository."""
986
1650
raise NotImplementedError(self.is_shared)
1000
1664
for repositories to maintain loaded indices across multiple locks
1001
1665
by checking inside their implementation of this method to see
1002
1666
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
1667
the disk format being validatable in this manner.
1010
1670
@needs_read_lock
1011
1671
def revision_tree(self, revision_id):
1012
1672
"""Return Tree for a revision on this branch.
1014
`revision_id` may be NULL_REVISION for the empty tree revision.
1674
`revision_id` may be None for the empty tree revision.
1016
raise NotImplementedError(self.revision_tree)
1676
# TODO: refactor this to use an existing revision object
1677
# so we don't need to read it in twice.
1678
if revision_id is None or revision_id == _mod_revision.NULL_REVISION:
1679
return RevisionTree(self, Inventory(root_id=None),
1680
_mod_revision.NULL_REVISION)
1682
inv = self.get_revision_inventory(revision_id)
1683
return RevisionTree(self, inv, revision_id)
1018
1685
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:'
1686
"""Return Tree for a revision on this branch.
1688
`revision_id` may not be None or 'null:'"""
1689
inventories = self.iter_inventories(revision_ids)
1690
for inv in inventories:
1691
yield RevisionTree(self, inv, inv.revision_id)
1694
def get_ancestry(self, revision_id, topo_sorted=True):
1695
"""Return a list of revision-ids integrated by a revision.
1697
The first element of the list is always None, indicating the origin
1698
revision. This might change when we have history horizons, or
1699
perhaps we should have a new API.
1701
This is topologically sorted.
1024
raise NotImplementedError(self.revision_trees)
1703
if _mod_revision.is_null(revision_id):
1705
if not self.has_revision(revision_id):
1706
raise errors.NoSuchRevision(self, revision_id)
1707
w = self.get_inventory_weave()
1708
candidates = w.get_ancestry(revision_id, topo_sorted)
1709
return [None] + candidates # self._eliminate_revisions_not_present(candidates)
1026
def pack(self, hint=None, clean_obsolete_packs=False):
1027
1712
"""Compress the data within the repository.
1029
1714
This operation only makes sense for some repository types. For other
1030
1715
types it should be a no-op that just returns.
1032
1717
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
1718
@needs_write_lock as this is a long running call its reasonable to
1034
1719
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
1723
def print_file(self, file, revision_id):
1724
"""Print `file` to stdout.
1726
FIXME RBC 20060125 as John Meinel points out this is a bad api
1727
- it writes to stdout, it assumes that that is valid etc. Fix
1728
by creating a new more flexible convenience function.
1730
tree = self.revision_tree(revision_id)
1731
# use inventory as it was in that revision
1732
file_id = tree.inventory.path2id(file)
1734
# TODO: jam 20060427 Write a test for this code path
1735
# it had a bug in it, and was raising the wrong
1737
raise errors.BzrError("%r is not present in revision %s" % (file, revision_id))
1738
tree.print_file(file_id)
1047
1740
def get_transaction(self):
1048
1741
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:
1743
@deprecated_method(symbol_versioning.one_five)
1744
def revision_parents(self, revision_id):
1745
return self.get_inventory_weave().parent_names(revision_id)
1747
@deprecated_method(symbol_versioning.one_one)
1748
def get_parents(self, revision_ids):
1749
"""See StackedParentsProvider.get_parents"""
1750
parent_map = self.get_parent_map(revision_ids)
1751
return [parent_map.get(r, None) for r in revision_ids]
1753
def get_parent_map(self, keys):
1754
"""See graph._StackedParentsProvider.get_parent_map"""
1756
for revision_id in keys:
1061
1757
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,)
1758
parent_map[revision_id] = ()
1761
parent_id_list = self.get_revision(revision_id).parent_ids
1762
except errors.NoSuchRevision:
1765
if len(parent_id_list) == 0:
1766
parent_ids = (_mod_revision.NULL_REVISION,)
1768
parent_ids = tuple(parent_id_list)
1769
parent_map[revision_id] = parent_ids
1077
1772
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
1775
def get_graph(self, other_repository=None):
1099
1776
"""Return the graph walker for this repository format"""
1100
1777
parents_provider = self._make_parents_provider()
1101
1778
if (other_repository is not None and
1102
1779
not self.has_same_location(other_repository)):
1103
parents_provider = graph.StackedParentsProvider(
1780
parents_provider = graph._StackedParentsProvider(
1104
1781
[parents_provider, other_repository._make_parents_provider()])
1105
1782
return graph.Graph(parents_provider)
1784
def _get_versioned_file_checker(self):
1785
"""Return an object suitable for checking versioned files."""
1786
return _VersionedFileChecker(self)
1788
def revision_ids_to_search_result(self, result_set):
1789
"""Convert a set of revision ids to a graph SearchResult."""
1790
result_parents = set()
1791
for parents in self.get_graph().get_parent_map(
1792
result_set).itervalues():
1793
result_parents.update(parents)
1794
included_keys = result_set.intersection(result_parents)
1795
start_keys = result_set.difference(included_keys)
1796
exclude_keys = result_parents.difference(result_set)
1797
result = graph.SearchResult(start_keys, exclude_keys,
1798
len(result_set), result_set)
1107
1801
@needs_write_lock
1108
1802
def set_make_working_trees(self, new_value):
1109
1803
"""Set the policy flag for making working trees when creating branches.
1220
1873
revision_id.decode('ascii')
1221
1874
except UnicodeDecodeError:
1222
1875
raise errors.NonAsciiRevisionId(method, self)
1877
def revision_graph_can_have_wrong_parents(self):
1878
"""Is it possible for this repository to have a revision graph with
1881
If True, then this repository must also implement
1882
_find_inconsistent_revision_parents so that check and reconcile can
1883
check for inconsistencies before proceeding with other checks that may
1884
depend on the revision index being consistent.
1886
raise NotImplementedError(self.revision_graph_can_have_wrong_parents)
1889
# remove these delegates a while after bzr 0.15
1890
def __make_delegated(name, from_module):
1891
def _deprecated_repository_forwarder():
1892
symbol_versioning.warn('%s moved to %s in bzr 0.15'
1893
% (name, from_module),
1896
m = __import__(from_module, globals(), locals(), [name])
1898
return getattr(m, name)
1899
except AttributeError:
1900
raise AttributeError('module %s has no name %s'
1902
globals()[name] = _deprecated_repository_forwarder
1905
'AllInOneRepository',
1906
'WeaveMetaDirRepository',
1907
'PreSplitOutRepositoryFormat',
1908
'RepositoryFormat4',
1909
'RepositoryFormat5',
1910
'RepositoryFormat6',
1911
'RepositoryFormat7',
1913
__make_delegated(_name, 'bzrlib.repofmt.weaverepo')
1917
'RepositoryFormatKnit',
1918
'RepositoryFormatKnit1',
1920
__make_delegated(_name, 'bzrlib.repofmt.knitrepo')
1923
def install_revision(repository, rev, revision_tree):
1924
"""Install all revision data into a repository."""
1925
install_revisions(repository, [(rev, revision_tree, None)])
1928
def install_revisions(repository, iterable, num_revisions=None, pb=None):
1929
"""Install all revision data into a repository.
1931
Accepts an iterable of revision, tree, signature tuples. The signature
1934
repository.start_write_group()
1936
for n, (revision, revision_tree, signature) in enumerate(iterable):
1937
_install_revision(repository, revision, revision_tree, signature)
1939
pb.update('Transferring revisions', n + 1, num_revisions)
1941
repository.abort_write_group()
1944
repository.commit_write_group()
1947
def _install_revision(repository, rev, revision_tree, signature):
1948
"""Install all revision data into a repository."""
1949
present_parents = []
1951
for p_id in rev.parent_ids:
1952
if repository.has_revision(p_id):
1953
present_parents.append(p_id)
1954
parent_trees[p_id] = repository.revision_tree(p_id)
1956
parent_trees[p_id] = repository.revision_tree(None)
1958
inv = revision_tree.inventory
1959
entries = inv.iter_entries()
1960
# backwards compatibility hack: skip the root id.
1961
if not repository.supports_rich_root():
1962
path, root = entries.next()
1963
if root.revision != rev.revision_id:
1964
raise errors.IncompatibleRevision(repr(repository))
1965
# Add the texts that are not already present
1966
for path, ie in entries:
1967
w = repository.weave_store.get_weave_or_empty(ie.file_id,
1968
repository.get_transaction())
1969
if ie.revision not in w:
1971
# FIXME: TODO: The following loop *may* be overlapping/duplicate
1972
# with InventoryEntry.find_previous_heads(). if it is, then there
1973
# is a latent bug here where the parents may have ancestors of each
1975
for revision, tree in parent_trees.iteritems():
1976
if ie.file_id not in tree:
1978
parent_id = tree.inventory[ie.file_id].revision
1979
if parent_id in text_parents:
1981
text_parents.append(parent_id)
1983
vfile = repository.weave_store.get_weave_or_empty(ie.file_id,
1984
repository.get_transaction())
1985
lines = revision_tree.get_file(ie.file_id).readlines()
1986
vfile.add_lines(rev.revision_id, text_parents, lines)
1988
# install the inventory
1989
repository.add_inventory(rev.revision_id, inv, present_parents)
1990
except errors.RevisionAlreadyPresent:
1992
if signature is not None:
1993
repository.add_signature_text(rev.revision_id, signature)
1994
repository.add_revision(rev.revision_id, rev, inv)
1225
1997
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
1998
"""Repositories in the new meta-dir layout."""
2000
def __init__(self, _format, a_bzrdir, control_files, _revision_store, control_store, text_store):
2001
super(MetaDirRepository, self).__init__(_format,
2007
dir_mode = self.control_files._dir_mode
2008
file_mode = self.control_files._file_mode
1236
2011
def is_shared(self):
1237
2012
"""Return True if this repository is flagged as a shared repository."""
1238
return self._transport.has('shared-storage')
2013
return self.control_files._transport.has('shared-storage')
1240
2015
@needs_write_lock
1241
2016
def set_make_working_trees(self, new_value):
1632
2360
_optimisers = []
1633
2361
"""The available optimised InterRepository types."""
1636
2363
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):
2364
raise NotImplementedError(self.copy_content)
2366
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
1653
2367
"""Fetch the content required to construct revision_id.
1655
2369
The content is copied from self.source to self.target.
1657
2371
:param revision_id: if None all content is copied, if NULL_REVISION no
1658
2372
content is copied.
2373
:param pb: optional progress bar to use for progress reports. If not
2374
provided a default one will be created.
2376
Returns the copied revision count and the failed revisions in a tuple:
1661
2379
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
2381
def _walk_to_common_revisions(self, revision_ids):
2382
"""Walk out from revision_ids in source to revisions target has.
2384
:param revision_ids: The start point for the search.
2385
:return: A set of revision ids.
2387
target_graph = self.target.get_graph()
2388
revision_ids = frozenset(revision_ids)
2389
if set(target_graph.get_parent_map(revision_ids)) == revision_ids:
2390
return graph.SearchResult(revision_ids, set(), 0, set())
2391
missing_revs = set()
2392
source_graph = self.source.get_graph()
2393
# ensure we don't pay silly lookup costs.
2394
searcher = source_graph._make_breadth_first_searcher(revision_ids)
2395
null_set = frozenset([_mod_revision.NULL_REVISION])
2398
next_revs, ghosts = searcher.next_with_ghosts()
2399
except StopIteration:
2401
if revision_ids.intersection(ghosts):
2402
absent_ids = set(revision_ids.intersection(ghosts))
2403
# If all absent_ids are present in target, no error is needed.
2404
absent_ids.difference_update(
2405
set(target_graph.get_parent_map(absent_ids)))
2407
raise errors.NoSuchRevision(self.source, absent_ids.pop())
2408
# we don't care about other ghosts as we can't fetch them and
2409
# haven't been asked to.
2410
next_revs = set(next_revs)
2411
# we always have NULL_REVISION present.
2412
have_revs = set(target_graph.get_parent_map(next_revs)).union(null_set)
2413
missing_revs.update(next_revs - have_revs)
2414
searcher.stop_searching_any(have_revs)
2415
return searcher.get_result()
2417
@deprecated_method(symbol_versioning.one_two)
2419
def missing_revision_ids(self, revision_id=None, find_ghosts=True):
2420
"""Return the revision ids that source has that target does not.
2422
These are returned in topological order.
2424
:param revision_id: only return revision ids included by this
2426
:param find_ghosts: If True find missing revisions in deep history
2427
rather than just finding the surface difference.
2429
return list(self.search_missing_revision_ids(
2430
revision_id, find_ghosts).get_keys())
2433
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2434
"""Return the revision ids that source has that target does not.
2436
:param revision_id: only return revision ids included by this
2438
:param find_ghosts: If True find missing revisions in deep history
2439
rather than just finding the surface difference.
1683
2440
:return: A bzrlib.graph.SearchResult.
1685
raise NotImplementedError(self.search_missing_revision_ids)
2442
# stop searching at found target revisions.
2443
if not find_ghosts and revision_id is not None:
2444
return self._walk_to_common_revisions([revision_id])
2445
# generic, possibly worst case, slow code path.
2446
target_ids = set(self.target.all_revision_ids())
2447
if revision_id is not None:
2448
source_ids = self.source.get_ancestry(revision_id)
2449
assert source_ids[0] is None
2452
source_ids = self.source.all_revision_ids()
2453
result_set = set(source_ids).difference(target_ids)
2454
return self.source.revision_ids_to_search_result(result_set)
1688
2457
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:
2458
"""True if source and target have the same data representation."""
2459
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
2461
if source._serializer != target._serializer:
1708
raise errors.IncompatibleRepositories(source, target,
1709
"different serializers")
2466
class InterSameDataRepository(InterRepository):
2467
"""Code for converting between repositories that represent the same data.
2469
Data format and model must match for this to work.
2473
def _get_repo_format_to_test(self):
2474
"""Repository format for testing with.
2476
InterSameData can pull from subtree to subtree and from non-subtree to
2477
non-subtree, so we test this with the richest repository format.
2479
from bzrlib.repofmt import knitrepo
2480
return knitrepo.RepositoryFormatKnit3()
2483
def is_compatible(source, target):
2484
return InterRepository._same_model(source, target)
2487
def copy_content(self, revision_id=None):
2488
"""Make a complete copy of the content in self into destination.
2490
This copies both the repository's revision data, and configuration information
2491
such as the make_working_trees setting.
2493
This is a destructive operation! Do not use it on existing
2496
:param revision_id: Only copy the content needed to construct
2497
revision_id and its parents.
2500
self.target.set_make_working_trees(self.source.make_working_trees())
2501
except NotImplementedError:
2503
# but don't bother fetching if we have the needed data now.
2504
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
2505
self.target.has_revision(revision_id)):
2507
self.target.fetch(self.source, revision_id=revision_id)
2510
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2511
"""See InterRepository.fetch()."""
2512
from bzrlib.fetch import GenericRepoFetcher
2513
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2514
self.source, self.source._format, self.target,
2515
self.target._format)
2516
f = GenericRepoFetcher(to_repository=self.target,
2517
from_repository=self.source,
2518
last_revision=revision_id,
2519
pb=pb, find_ghosts=find_ghosts)
2520
return f.count_copied, f.failed_revisions
2523
class InterWeaveRepo(InterSameDataRepository):
2524
"""Optimised code paths between Weave based repositories.
2526
This should be in bzrlib/repofmt/weaverepo.py but we have not yet
2527
implemented lazy inter-object optimisation.
2531
def _get_repo_format_to_test(self):
2532
from bzrlib.repofmt import weaverepo
2533
return weaverepo.RepositoryFormat7()
2536
def is_compatible(source, target):
2537
"""Be compatible with known Weave formats.
2539
We don't test for the stores being of specific types because that
2540
could lead to confusing results, and there is no need to be
2543
from bzrlib.repofmt.weaverepo import (
2549
return (isinstance(source._format, (RepositoryFormat5,
2551
RepositoryFormat7)) and
2552
isinstance(target._format, (RepositoryFormat5,
2554
RepositoryFormat7)))
2555
except AttributeError:
2559
def copy_content(self, revision_id=None):
2560
"""See InterRepository.copy_content()."""
2561
# weave specific optimised path:
2563
self.target.set_make_working_trees(self.source.make_working_trees())
2564
except (errors.RepositoryUpgradeRequired, NotImplemented):
2566
# FIXME do not peek!
2567
if self.source.control_files._transport.listable():
2568
pb = ui.ui_factory.nested_progress_bar()
2570
self.target.weave_store.copy_all_ids(
2571
self.source.weave_store,
2573
from_transaction=self.source.get_transaction(),
2574
to_transaction=self.target.get_transaction())
2575
pb.update('copying inventory', 0, 1)
2576
self.target.control_weaves.copy_multi(
2577
self.source.control_weaves, ['inventory'],
2578
from_transaction=self.source.get_transaction(),
2579
to_transaction=self.target.get_transaction())
2580
self.target._revision_store.text_store.copy_all_ids(
2581
self.source._revision_store.text_store,
2586
self.target.fetch(self.source, revision_id=revision_id)
2589
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2590
"""See InterRepository.fetch()."""
2591
from bzrlib.fetch import GenericRepoFetcher
2592
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2593
self.source, self.source._format, self.target, self.target._format)
2594
f = GenericRepoFetcher(to_repository=self.target,
2595
from_repository=self.source,
2596
last_revision=revision_id,
2597
pb=pb, find_ghosts=find_ghosts)
2598
return f.count_copied, f.failed_revisions
2601
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2602
"""See InterRepository.missing_revision_ids()."""
2603
# we want all revisions to satisfy revision_id in source.
2604
# but we don't want to stat every file here and there.
2605
# we want then, all revisions other needs to satisfy revision_id
2606
# checked, but not those that we have locally.
2607
# so the first thing is to get a subset of the revisions to
2608
# satisfy revision_id in source, and then eliminate those that
2609
# we do already have.
2610
# this is slow on high latency connection to self, but as as this
2611
# disk format scales terribly for push anyway due to rewriting
2612
# inventory.weave, this is considered acceptable.
2614
if revision_id is not None:
2615
source_ids = self.source.get_ancestry(revision_id)
2616
assert source_ids[0] is None
2619
source_ids = self.source._all_possible_ids()
2620
source_ids_set = set(source_ids)
2621
# source_ids is the worst possible case we may need to pull.
2622
# now we want to filter source_ids against what we actually
2623
# have in target, but don't try to check for existence where we know
2624
# we do not have a revision as that would be pointless.
2625
target_ids = set(self.target._all_possible_ids())
2626
possibly_present_revisions = target_ids.intersection(source_ids_set)
2627
actually_present_revisions = set(
2628
self.target._eliminate_revisions_not_present(possibly_present_revisions))
2629
required_revisions = source_ids_set.difference(actually_present_revisions)
2630
if revision_id is not None:
2631
# we used get_ancestry to determine source_ids then we are assured all
2632
# revisions referenced are present as they are installed in topological order.
2633
# and the tip revision was validated by get_ancestry.
2634
result_set = required_revisions
2636
# if we just grabbed the possibly available ids, then
2637
# we only have an estimate of whats available and need to validate
2638
# that against the revision records.
2640
self.source._eliminate_revisions_not_present(required_revisions))
2641
return self.source.revision_ids_to_search_result(result_set)
2644
class InterKnitRepo(InterSameDataRepository):
2645
"""Optimised code paths between Knit based repositories."""
2648
def _get_repo_format_to_test(self):
2649
from bzrlib.repofmt import knitrepo
2650
return knitrepo.RepositoryFormatKnit1()
2653
def is_compatible(source, target):
2654
"""Be compatible with known Knit formats.
2656
We don't test for the stores being of specific types because that
2657
could lead to confusing results, and there is no need to be
2660
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit
2662
are_knits = (isinstance(source._format, RepositoryFormatKnit) and
2663
isinstance(target._format, RepositoryFormatKnit))
2664
except AttributeError:
2666
return are_knits and InterRepository._same_model(source, target)
2669
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2670
"""See InterRepository.fetch()."""
2671
from bzrlib.fetch import KnitRepoFetcher
2672
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2673
self.source, self.source._format, self.target, self.target._format)
2674
f = KnitRepoFetcher(to_repository=self.target,
2675
from_repository=self.source,
2676
last_revision=revision_id,
2677
pb=pb, find_ghosts=find_ghosts)
2678
return f.count_copied, f.failed_revisions
2681
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2682
"""See InterRepository.missing_revision_ids()."""
2683
if revision_id is not None:
2684
source_ids = self.source.get_ancestry(revision_id)
2685
assert source_ids[0] is None
2688
source_ids = self.source.all_revision_ids()
2689
source_ids_set = set(source_ids)
2690
# source_ids is the worst possible case we may need to pull.
2691
# now we want to filter source_ids against what we actually
2692
# have in target, but don't try to check for existence where we know
2693
# we do not have a revision as that would be pointless.
2694
target_ids = set(self.target.all_revision_ids())
2695
possibly_present_revisions = target_ids.intersection(source_ids_set)
2696
actually_present_revisions = set(
2697
self.target._eliminate_revisions_not_present(possibly_present_revisions))
2698
required_revisions = source_ids_set.difference(actually_present_revisions)
2699
if revision_id is not None:
2700
# we used get_ancestry to determine source_ids then we are assured all
2701
# revisions referenced are present as they are installed in topological order.
2702
# and the tip revision was validated by get_ancestry.
2703
result_set = required_revisions
2705
# if we just grabbed the possibly available ids, then
2706
# we only have an estimate of whats available and need to validate
2707
# that against the revision records.
2709
self.source._eliminate_revisions_not_present(required_revisions))
2710
return self.source.revision_ids_to_search_result(result_set)
2713
class InterPackRepo(InterSameDataRepository):
2714
"""Optimised code paths between Pack based repositories."""
2717
def _get_repo_format_to_test(self):
2718
from bzrlib.repofmt import pack_repo
2719
return pack_repo.RepositoryFormatKnitPack1()
2722
def is_compatible(source, target):
2723
"""Be compatible with known Pack formats.
2725
We don't test for the stores being of specific types because that
2726
could lead to confusing results, and there is no need to be
2729
from bzrlib.repofmt.pack_repo import RepositoryFormatPack
2731
are_packs = (isinstance(source._format, RepositoryFormatPack) and
2732
isinstance(target._format, RepositoryFormatPack))
2733
except AttributeError:
2735
return are_packs and InterRepository._same_model(source, target)
2738
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2739
"""See InterRepository.fetch()."""
2740
from bzrlib.repofmt.pack_repo import Packer
2741
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2742
self.source, self.source._format, self.target, self.target._format)
2743
self.count_copied = 0
2744
if revision_id is None:
2746
# everything to do - use pack logic
2747
# to fetch from all packs to one without
2748
# inventory parsing etc, IFF nothing to be copied is in the target.
2750
revision_ids = self.source.all_revision_ids()
2751
revision_keys = [(revid,) for revid in revision_ids]
2752
index = self.target._pack_collection.revision_index.combined_index
2753
present_revision_ids = set(item[1][0] for item in
2754
index.iter_entries(revision_keys))
2755
revision_ids = set(revision_ids) - present_revision_ids
2756
# implementing the TODO will involve:
2757
# - detecting when all of a pack is selected
2758
# - avoiding as much as possible pre-selection, so the
2759
# more-core routines such as create_pack_from_packs can filter in
2760
# a just-in-time fashion. (though having a HEADS list on a
2761
# repository might make this a lot easier, because we could
2762
# sensibly detect 'new revisions' without doing a full index scan.
2763
elif _mod_revision.is_null(revision_id):
2768
revision_ids = self.search_missing_revision_ids(revision_id,
2769
find_ghosts=find_ghosts).get_keys()
2770
except errors.NoSuchRevision:
2771
raise errors.InstallFailed([revision_id])
2772
if len(revision_ids) == 0:
2774
packs = self.source._pack_collection.all_packs()
2775
pack = Packer(self.target._pack_collection, packs, '.fetch',
2776
revision_ids).pack()
2777
if pack is not None:
2778
self.target._pack_collection._save_pack_names()
2779
# Trigger an autopack. This may duplicate effort as we've just done
2780
# a pack creation, but for now it is simpler to think about as
2781
# 'upload data, then repack if needed'.
2782
self.target._pack_collection.autopack()
2783
return (pack.get_revision_count(), [])
2788
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2789
"""See InterRepository.missing_revision_ids().
2791
:param find_ghosts: Find ghosts throughout the ancestry of
2794
if not find_ghosts and revision_id is not None:
2795
return self._walk_to_common_revisions([revision_id])
2796
elif revision_id is not None:
2797
source_ids = self.source.get_ancestry(revision_id)
2798
assert source_ids[0] is None
2801
source_ids = self.source.all_revision_ids()
2802
# source_ids is the worst possible case we may need to pull.
2803
# now we want to filter source_ids against what we actually
2804
# have in target, but don't try to check for existence where we know
2805
# we do not have a revision as that would be pointless.
2806
target_ids = set(self.target.all_revision_ids())
2807
result_set = set(source_ids).difference(target_ids)
2808
return self.source.revision_ids_to_search_result(result_set)
2811
class InterModel1and2(InterRepository):
2814
def _get_repo_format_to_test(self):
2818
def is_compatible(source, target):
2819
if not source.supports_rich_root() and target.supports_rich_root():
2825
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2826
"""See InterRepository.fetch()."""
2827
from bzrlib.fetch import Model1toKnit2Fetcher
2828
f = Model1toKnit2Fetcher(to_repository=self.target,
2829
from_repository=self.source,
2830
last_revision=revision_id,
2831
pb=pb, find_ghosts=find_ghosts)
2832
return f.count_copied, f.failed_revisions
2835
def copy_content(self, revision_id=None):
2836
"""Make a complete copy of the content in self into destination.
2838
This is a destructive operation! Do not use it on existing
2841
:param revision_id: Only copy the content needed to construct
2842
revision_id and its parents.
2845
self.target.set_make_working_trees(self.source.make_working_trees())
2846
except NotImplementedError:
2848
# but don't bother fetching if we have the needed data now.
2849
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
2850
self.target.has_revision(revision_id)):
2852
self.target.fetch(self.source, revision_id=revision_id)
2855
class InterKnit1and2(InterKnitRepo):
2858
def _get_repo_format_to_test(self):
2862
def is_compatible(source, target):
2863
"""Be compatible with Knit1 source and Knit3 target"""
2864
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit3
2866
from bzrlib.repofmt.knitrepo import (RepositoryFormatKnit1,
2867
RepositoryFormatKnit3)
2868
from bzrlib.repofmt.pack_repo import (
2869
RepositoryFormatKnitPack1,
2870
RepositoryFormatKnitPack3,
2871
RepositoryFormatPackDevelopment0,
2872
RepositoryFormatPackDevelopment0Subtree,
2875
RepositoryFormatKnit1,
2876
RepositoryFormatKnitPack1,
2877
RepositoryFormatPackDevelopment0,
2880
RepositoryFormatKnit3,
2881
RepositoryFormatKnitPack3,
2882
RepositoryFormatPackDevelopment0Subtree,
2884
return (isinstance(source._format, nosubtrees) and
2885
isinstance(target._format, subtrees))
2886
except AttributeError:
2890
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2891
"""See InterRepository.fetch()."""
2892
from bzrlib.fetch import Knit1to2Fetcher
2893
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2894
self.source, self.source._format, self.target,
2895
self.target._format)
2896
f = Knit1to2Fetcher(to_repository=self.target,
2897
from_repository=self.source,
2898
last_revision=revision_id,
2899
pb=pb, find_ghosts=find_ghosts)
2900
return f.count_copied, f.failed_revisions
2903
class InterDifferingSerializer(InterKnitRepo):
2906
def _get_repo_format_to_test(self):
2910
def is_compatible(source, target):
2911
"""Be compatible with Knit2 source and Knit3 target"""
2912
if source.supports_rich_root() != target.supports_rich_root():
2914
# Ideally, we'd support fetching if the source had no tree references
2915
# even if it supported them...
2916
if (getattr(source, '_format.supports_tree_reference', False) and
2917
not getattr(target, '_format.supports_tree_reference', False)):
2922
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2923
"""See InterRepository.fetch()."""
2924
revision_ids = self.target.search_missing_revision_ids(self.source,
2925
revision_id, find_ghosts=find_ghosts).get_keys()
2926
revision_ids = tsort.topo_sort(
2927
self.source.get_graph().get_parent_map(revision_ids))
2928
def revisions_iterator():
2929
for current_revision_id in revision_ids:
2930
revision = self.source.get_revision(current_revision_id)
2931
tree = self.source.revision_tree(current_revision_id)
2933
signature = self.source.get_signature_text(
2934
current_revision_id)
2935
except errors.NoSuchRevision:
2937
yield revision, tree, signature
2939
my_pb = ui.ui_factory.nested_progress_bar()
2944
install_revisions(self.target, revisions_iterator(),
2945
len(revision_ids), pb)
2947
if my_pb is not None:
2949
return len(revision_ids), 0
2952
class InterRemoteToOther(InterRepository):
2954
def __init__(self, source, target):
2955
InterRepository.__init__(self, source, target)
2956
self._real_inter = None
2959
def is_compatible(source, target):
2960
if not isinstance(source, remote.RemoteRepository):
2962
# Is source's model compatible with target's model?
2963
source._ensure_real()
2964
real_source = source._real_repository
2965
assert not isinstance(real_source, remote.RemoteRepository), (
2966
"We don't support remote repos backed by remote repos yet.")
2967
return InterRepository._same_model(real_source, target)
2970
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2971
"""See InterRepository.fetch()."""
2972
from bzrlib.fetch import RemoteToOtherFetcher
2973
mutter("Using fetch logic to copy between %s(remote) and %s(%s)",
2974
self.source, self.target, self.target._format)
2975
# TODO: jam 20070210 This should be an assert, not a translate
2976
revision_id = osutils.safe_revision_id(revision_id)
2977
f = RemoteToOtherFetcher(to_repository=self.target,
2978
from_repository=self.source,
2979
last_revision=revision_id,
2980
pb=pb, find_ghosts=find_ghosts)
2981
return f.count_copied, f.failed_revisions
2984
def _get_repo_format_to_test(self):
2988
class InterOtherToRemote(InterRepository):
2990
def __init__(self, source, target):
2991
InterRepository.__init__(self, source, target)
2992
self._real_inter = None
2995
def is_compatible(source, target):
2996
if isinstance(target, remote.RemoteRepository):
3000
def _ensure_real_inter(self):
3001
if self._real_inter is None:
3002
self.target._ensure_real()
3003
real_target = self.target._real_repository
3004
self._real_inter = InterRepository.get(self.source, real_target)
3006
def copy_content(self, revision_id=None):
3007
self._ensure_real_inter()
3008
self._real_inter.copy_content(revision_id=revision_id)
3010
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3011
self._ensure_real_inter()
3012
self._real_inter.fetch(revision_id=revision_id, pb=pb,
3013
find_ghosts=find_ghosts)
3016
def _get_repo_format_to_test(self):
3020
InterRepository.register_optimiser(InterDifferingSerializer)
3021
InterRepository.register_optimiser(InterSameDataRepository)
3022
InterRepository.register_optimiser(InterWeaveRepo)
3023
InterRepository.register_optimiser(InterKnitRepo)
3024
InterRepository.register_optimiser(InterModel1and2)
3025
InterRepository.register_optimiser(InterKnit1and2)
3026
InterRepository.register_optimiser(InterPackRepo)
3027
InterRepository.register_optimiser(InterRemoteToOther)
3028
InterRepository.register_optimiser(InterOtherToRemote)
1712
3031
class CopyConverter(object):
1713
3032
"""A repository conversion tool which just performs a copy of the content.
1715
3034
This is slow but quite reliable.