196
192
self.random_revid = False
198
def will_record_deletes(self):
199
"""Tell the commit builder that deletes are being notified.
201
This enables the accumulation of an inventory delta; for the resulting
202
commit to be valid, deletes against the basis MUST be recorded via
203
builder.record_delete().
205
raise NotImplementedError(self.will_record_deletes)
207
def record_iter_changes(self, tree, basis_revision_id, iter_changes):
208
"""Record a new tree via iter_changes.
210
:param tree: The tree to obtain text contents from for changed objects.
211
:param basis_revision_id: The revision id of the tree the iter_changes
212
has been generated against. Currently assumed to be the same
213
as self.parents[0] - if it is not, errors may occur.
214
:param iter_changes: An iter_changes iterator with the changes to apply
215
to basis_revision_id. The iterator must not include any items with
216
a current kind of None - missing items must be either filtered out
217
or errored-on beefore record_iter_changes sees the item.
218
:return: A generator of (file_id, relpath, fs_hash) tuples for use with
221
raise NotImplementedError(self.record_iter_changes)
224
class RepositoryWriteLockResult(LogicalLockResult):
225
"""The result of write locking a repository.
227
:ivar repository_token: The token obtained from the underlying lock, or
229
:ivar unlock: A callable which will unlock the lock.
232
def __init__(self, unlock, repository_token):
233
LogicalLockResult.__init__(self, unlock)
234
self.repository_token = repository_token
237
return "RepositoryWriteLockResult(%s, %s)" % (self.repository_token,
194
def _heads(self, file_id, revision_ids):
195
"""Calculate the graph heads for revision_ids in the graph of file_id.
197
This can use either a per-file graph or a global revision graph as we
198
have an identity relationship between the two graphs.
200
return self.__heads(revision_ids)
202
def _check_root(self, ie, parent_invs, tree):
203
"""Helper for record_entry_contents.
205
:param ie: An entry being added.
206
:param parent_invs: The inventories of the parent revisions of the
208
:param tree: The tree that is being committed.
210
# In this revision format, root entries have no knit or weave When
211
# serializing out to disk and back in root.revision is always
213
ie.revision = self._new_revision_id
215
def _get_delta(self, ie, basis_inv, path):
216
"""Get a delta against the basis inventory for ie."""
217
if ie.file_id not in basis_inv:
219
return (None, path, ie.file_id, ie)
220
elif ie != basis_inv[ie.file_id]:
222
# TODO: avoid tis id2path call.
223
return (basis_inv.id2path(ie.file_id), path, ie.file_id, ie)
228
def record_entry_contents(self, ie, parent_invs, path, tree,
230
"""Record the content of ie from tree into the commit if needed.
232
Side effect: sets ie.revision when unchanged
234
:param ie: An inventory entry present in the commit.
235
:param parent_invs: The inventories of the parent revisions of the
237
:param path: The path the entry is at in the tree.
238
:param tree: The tree which contains this entry and should be used to
240
:param content_summary: Summary data from the tree about the paths
241
content - stat, length, exec, sha/link target. This is only
242
accessed when the entry has a revision of None - that is when it is
243
a candidate to commit.
244
:return: A tuple (change_delta, version_recorded). change_delta is
245
an inventory_delta change for this entry against the basis tree of
246
the commit, or None if no change occured against the basis tree.
247
version_recorded is True if a new version of the entry has been
248
recorded. For instance, committing a merge where a file was only
249
changed on the other side will return (delta, False).
251
if self.new_inventory.root is None:
252
if ie.parent_id is not None:
253
raise errors.RootMissing()
254
self._check_root(ie, parent_invs, tree)
255
if ie.revision is None:
256
kind = content_summary[0]
258
# ie is carried over from a prior commit
260
# XXX: repository specific check for nested tree support goes here - if
261
# the repo doesn't want nested trees we skip it ?
262
if (kind == 'tree-reference' and
263
not self.repository._format.supports_tree_reference):
264
# mismatch between commit builder logic and repository:
265
# this needs the entry creation pushed down into the builder.
266
raise NotImplementedError('Missing repository subtree support.')
267
self.new_inventory.add(ie)
269
# TODO: slow, take it out of the inner loop.
271
basis_inv = parent_invs[0]
273
basis_inv = Inventory(root_id=None)
275
# ie.revision is always None if the InventoryEntry is considered
276
# for committing. We may record the previous parents revision if the
277
# content is actually unchanged against a sole head.
278
if ie.revision is not None:
279
if not self._versioned_root and path == '':
280
# repositories that do not version the root set the root's
281
# revision to the new commit even when no change occurs, and
282
# this masks when a change may have occurred against the basis,
283
# so calculate if one happened.
284
if ie.file_id in basis_inv:
285
delta = (basis_inv.id2path(ie.file_id), path,
289
delta = (None, path, ie.file_id, ie)
292
# we don't need to commit this, because the caller already
293
# determined that an existing revision of this file is
294
# appropriate. If its not being considered for committing then
295
# it and all its parents to the root must be unaltered so
296
# no-change against the basis.
297
if ie.revision == self._new_revision_id:
298
raise AssertionError("Impossible situation, a skipped "
299
"inventory entry (%r) claims to be modified in this "
300
"commit (%r).", (ie, self._new_revision_id))
302
# XXX: Friction: parent_candidates should return a list not a dict
303
# so that we don't have to walk the inventories again.
304
parent_candiate_entries = ie.parent_candidates(parent_invs)
305
head_set = self._heads(ie.file_id, parent_candiate_entries.keys())
307
for inv in parent_invs:
308
if ie.file_id in inv:
309
old_rev = inv[ie.file_id].revision
310
if old_rev in head_set:
311
heads.append(inv[ie.file_id].revision)
312
head_set.remove(inv[ie.file_id].revision)
315
# now we check to see if we need to write a new record to the
317
# We write a new entry unless there is one head to the ancestors, and
318
# the kind-derived content is unchanged.
320
# Cheapest check first: no ancestors, or more the one head in the
321
# ancestors, we write a new node.
325
# There is a single head, look it up for comparison
326
parent_entry = parent_candiate_entries[heads[0]]
327
# if the non-content specific data has changed, we'll be writing a
329
if (parent_entry.parent_id != ie.parent_id or
330
parent_entry.name != ie.name):
332
# now we need to do content specific checks:
334
# if the kind changed the content obviously has
335
if kind != parent_entry.kind:
338
if content_summary[2] is None:
339
raise ValueError("Files must not have executable = None")
341
if (# if the file length changed we have to store:
342
parent_entry.text_size != content_summary[1] or
343
# if the exec bit has changed we have to store:
344
parent_entry.executable != content_summary[2]):
346
elif parent_entry.text_sha1 == content_summary[3]:
347
# all meta and content is unchanged (using a hash cache
348
# hit to check the sha)
349
ie.revision = parent_entry.revision
350
ie.text_size = parent_entry.text_size
351
ie.text_sha1 = parent_entry.text_sha1
352
ie.executable = parent_entry.executable
353
return self._get_delta(ie, basis_inv, path), False
355
# Either there is only a hash change(no hash cache entry,
356
# or same size content change), or there is no change on
358
# Provide the parent's hash to the store layer, so that the
359
# content is unchanged we will not store a new node.
360
nostore_sha = parent_entry.text_sha1
362
# We want to record a new node regardless of the presence or
363
# absence of a content change in the file.
365
ie.executable = content_summary[2]
366
lines = tree.get_file(ie.file_id, path).readlines()
368
ie.text_sha1, ie.text_size = self._add_text_to_weave(
369
ie.file_id, lines, heads, nostore_sha)
370
except errors.ExistingContent:
371
# Turns out that the file content was unchanged, and we were
372
# only going to store a new node if it was changed. Carry over
374
ie.revision = parent_entry.revision
375
ie.text_size = parent_entry.text_size
376
ie.text_sha1 = parent_entry.text_sha1
377
ie.executable = parent_entry.executable
378
return self._get_delta(ie, basis_inv, path), False
379
elif kind == 'directory':
381
# all data is meta here, nothing specific to directory, so
383
ie.revision = parent_entry.revision
384
return self._get_delta(ie, basis_inv, path), False
386
self._add_text_to_weave(ie.file_id, lines, heads, None)
387
elif kind == 'symlink':
388
current_link_target = content_summary[3]
390
# symlink target is not generic metadata, check if it has
392
if current_link_target != parent_entry.symlink_target:
395
# unchanged, carry over.
396
ie.revision = parent_entry.revision
397
ie.symlink_target = parent_entry.symlink_target
398
return self._get_delta(ie, basis_inv, path), False
399
ie.symlink_target = current_link_target
401
self._add_text_to_weave(ie.file_id, lines, heads, None)
402
elif kind == 'tree-reference':
404
if content_summary[3] != parent_entry.reference_revision:
407
# unchanged, carry over.
408
ie.reference_revision = parent_entry.reference_revision
409
ie.revision = parent_entry.revision
410
return self._get_delta(ie, basis_inv, path), False
411
ie.reference_revision = content_summary[3]
413
self._add_text_to_weave(ie.file_id, lines, heads, None)
415
raise NotImplementedError('unknown kind')
416
ie.revision = self._new_revision_id
417
return self._get_delta(ie, basis_inv, path), True
419
def _add_text_to_weave(self, file_id, new_lines, parents, nostore_sha):
420
# Note: as we read the content directly from the tree, we know its not
421
# been turned into unicode or badly split - but a broken tree
422
# implementation could give us bad output from readlines() so this is
423
# not a guarantee of safety. What would be better is always checking
424
# the content during test suite execution. RBC 20070912
425
parent_keys = tuple((file_id, parent) for parent in parents)
426
return self.repository.texts.add_lines(
427
(file_id, self._new_revision_id), parent_keys, new_lines,
428
nostore_sha=nostore_sha, random_id=self.random_revid,
429
check_content=False)[0:2]
432
class RootCommitBuilder(CommitBuilder):
433
"""This commitbuilder actually records the root id"""
435
# the root entry gets versioned properly by this builder.
436
_versioned_root = True
438
def _check_root(self, ie, parent_invs, tree):
439
"""Helper for record_entry_contents.
441
:param ie: An entry being added.
442
:param parent_invs: The inventories of the parent revisions of the
444
:param tree: The tree that is being committed.
241
448
######################################################################
245
class Repository(_RelockDebugMixin, controldir.ControlComponent):
451
class Repository(object):
246
452
"""Repository holding history for one or more branches.
248
454
The repository holds and retrieves historical information including
249
455
revisions and file history. It's normally accessed only by the Branch,
250
456
which views a particular line of development through that history.
252
See VersionedFileRepository in bzrlib.vf_repository for the
253
base class for most Bazaar repositories.
458
The Repository builds on top of some byte storage facilies (the revisions,
459
signatures, inventories and texts attributes) and a Transport, which
460
respectively provide byte storage and a means to access the (possibly
463
The byte storage facilities are addressed via tuples, which we refer to
464
as 'keys' throughout the code base. Revision_keys, inventory_keys and
465
signature_keys are all 1-tuples: (revision_id,). text_keys are two-tuples:
466
(file_id, revision_id). We use this interface because it allows low
467
friction with the underlying code that implements disk indices, network
468
encoding and other parts of bzrlib.
470
:ivar revisions: A bzrlib.versionedfile.VersionedFiles instance containing
471
the serialised revisions for the repository. This can be used to obtain
472
revision graph information or to access raw serialised revisions.
473
The result of trying to insert data into the repository via this store
474
is undefined: it should be considered read-only except for implementors
476
:ivar signatures: A bzrlib.versionedfile.VersionedFiles instance containing
477
the serialised signatures for the repository. This can be used to
478
obtain access to raw serialised signatures. The result of trying to
479
insert data into the repository via this store is undefined: it should
480
be considered read-only except for implementors of repositories.
481
:ivar inventories: A bzrlib.versionedfile.VersionedFiles instance containing
482
the serialised inventories for the repository. This can be used to
483
obtain unserialised inventories. The result of trying to insert data
484
into the repository via this store is undefined: it should be
485
considered read-only except for implementors of repositories.
486
:ivar texts: A bzrlib.versionedfile.VersionedFiles instance containing the
487
texts of files and directories for the repository. This can be used to
488
obtain file texts or file graphs. Note that Repository.iter_file_bytes
489
is usually a better interface for accessing file texts.
490
The result of trying to insert data into the repository via this store
491
is undefined: it should be considered read-only except for implementors
493
:ivar _transport: Transport for file access to repository, typically
494
pointing to .bzr/repository.
256
def abort_write_group(self, suppress_errors=False):
497
# What class to use for a CommitBuilder. Often its simpler to change this
498
# in a Repository class subclass rather than to override
499
# get_commit_builder.
500
_commit_builder_class = CommitBuilder
501
# The search regex used by xml based repositories to determine what things
502
# where changed in a single commit.
503
_file_ids_altered_regex = lazy_regex.lazy_compile(
504
r'file_id="(?P<file_id>[^"]+)"'
505
r'.* revision="(?P<revision_id>[^"]+)"'
508
def abort_write_group(self):
257
509
"""Commit the contents accrued within the current write group.
259
:param suppress_errors: if true, abort_write_group will catch and log
260
unexpected errors that happen during the abort, rather than
261
allowing them to propagate. Defaults to False.
263
511
:seealso: start_write_group.
265
513
if self._write_group is not self.get_transaction():
266
514
# has an unlock or relock occured ?
269
'(suppressed) mismatched lock context and write group. %r, %r',
270
self._write_group, self.get_transaction())
272
raise errors.BzrError(
273
'mismatched lock context and write group. %r, %r' %
274
(self._write_group, self.get_transaction()))
276
self._abort_write_group()
277
except Exception, exc:
278
self._write_group = None
279
if not suppress_errors:
281
mutter('abort_write_group failed')
282
log_exception_quietly()
283
note('bzr: ERROR (ignored): %s', exc)
515
raise errors.BzrError('mismatched lock context and write group.')
516
self._abort_write_group()
284
517
self._write_group = None
286
519
def _abort_write_group(self):
287
520
"""Template method for per-repository write group cleanup.
289
This is called during abort before the write group is considered to be
522
This is called during abort before the write group is considered to be
290
523
finished and should cleanup any internal state accrued during the write
291
524
group. There is no requirement that data handed to the repository be
292
525
*not* made available - this is not a rollback - but neither should any
618
939
"""Commit the contents accrued within the current write group.
620
941
:seealso: start_write_group.
622
:return: it may return an opaque hint that can be passed to 'pack'.
624
943
if self._write_group is not self.get_transaction():
625
944
# has an unlock or relock occured ?
626
945
raise errors.BzrError('mismatched lock context %r and '
627
946
'write group %r.' %
628
947
(self.get_transaction(), self._write_group))
629
result = self._commit_write_group()
948
self._commit_write_group()
630
949
self._write_group = None
633
951
def _commit_write_group(self):
634
952
"""Template method for per-repository write group cleanup.
636
This is called before the write group is considered to be
954
This is called before the write group is considered to be
637
955
finished and should ensure that all data handed to the repository
638
for writing during the write group is safely committed (to the
956
for writing during the write group is safely committed (to the
639
957
extent possible considering file system caching etc).
642
def suspend_write_group(self):
643
raise errors.UnsuspendableWriteGroup(self)
645
def refresh_data(self):
646
"""Re-read any data needed to synchronise with disk.
648
This method is intended to be called after another repository instance
649
(such as one used by a smart server) has inserted data into the
650
repository. On all repositories this will work outside of write groups.
651
Some repository formats (pack and newer for bzrlib native formats)
652
support refresh_data inside write groups. If called inside a write
653
group on a repository that does not support refreshing in a write group
654
IsInWriteGroupError will be raised.
658
def resume_write_group(self, tokens):
659
if not self.is_write_locked():
660
raise errors.NotWriteLocked(self)
661
if self._write_group:
662
raise errors.BzrError('already in a write group')
663
self._resume_write_group(tokens)
664
# so we can detect unlock/relock - the write group is now entered.
665
self._write_group = self.get_transaction()
667
def _resume_write_group(self, tokens):
668
raise errors.UnsuspendableWriteGroup(self)
670
def fetch(self, source, revision_id=None, find_ghosts=False,
960
def fetch(self, source, revision_id=None, pb=None, find_ghosts=False):
672
961
"""Fetch the content required to construct revision_id from source.
674
If revision_id is None and fetch_spec is None, then all content is
677
fetch() may not be used when the repository is in a write group -
678
either finish the current write group before using fetch, or use
679
fetch before starting the write group.
963
If revision_id is None all content is copied.
681
964
:param find_ghosts: Find and copy revisions in the source that are
682
965
ghosts in the target (and not reachable directly by walking out to
683
966
the first-present revision in target from revision_id).
684
:param revision_id: If specified, all the content needed for this
685
revision ID will be copied to the target. Fetch will determine for
686
itself which content needs to be copied.
687
:param fetch_spec: If specified, a SearchResult or
688
PendingAncestryResult that describes which revisions to copy. This
689
allows copying multiple heads at once. Mutually exclusive with
692
if fetch_spec is not None and revision_id is not None:
693
raise AssertionError(
694
"fetch_spec and revision_id are mutually exclusive.")
695
if self.is_in_write_group():
696
raise errors.InternalBzrError(
697
"May not fetch while in a write group.")
698
968
# fast path same-url fetch operations
699
# TODO: lift out to somewhere common with RemoteRepository
700
# <https://bugs.launchpad.net/bzr/+bug/401646>
701
if (self.has_same_location(source)
702
and fetch_spec is None
703
and self._has_same_fallbacks(source)):
969
if self.has_same_location(source):
704
970
# check that last_revision is in 'from' and then return a
706
972
if (revision_id is not None and
707
973
not _mod_revision.is_null(revision_id)):
708
974
self.get_revision(revision_id)
976
# if there is no specific appropriate InterRepository, this will get
977
# the InterRepository base class, which raises an
978
# IncompatibleRepositories when asked to fetch.
710
979
inter = InterRepository.get(source, self)
711
return inter.fetch(revision_id=revision_id,
712
find_ghosts=find_ghosts, fetch_spec=fetch_spec)
980
return inter.fetch(revision_id=revision_id, pb=pb,
981
find_ghosts=find_ghosts)
714
983
def create_bundle(self, target, base, fileobj, format=None):
715
984
return serializer.write_bundle(self, target, base, fileobj, format)
717
986
def get_commit_builder(self, branch, parents, config, timestamp=None,
718
987
timezone=None, committer=None, revprops=None,
719
revision_id=None, lossy=False):
720
989
"""Obtain a CommitBuilder for this repository.
722
991
:param branch: Branch to commit to.
723
992
:param parents: Revision ids of the parents of the new revision.
724
993
:param config: Configuration to use.
890
1172
yield trees[revision.revision_id].changes_from(old_tree)
892
1174
@needs_read_lock
893
def get_revision_delta(self, revision_id, specific_fileids=None):
1175
def get_revision_delta(self, revision_id):
894
1176
"""Return the delta for one revision.
896
1178
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
1181
r = self.get_revision(revision_id)
904
return list(self.get_deltas_for_revisions([r],
905
specific_fileids=specific_fileids))[0]
1182
return list(self.get_deltas_for_revisions([r]))[0]
907
1184
@needs_write_lock
908
1185
def store_revision_signature(self, gpg_strategy, plaintext, revision_id):
909
1186
signature = gpg_strategy.sign(plaintext)
910
1187
self.add_signature_text(revision_id, signature)
912
1190
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)
933
def fileids_altered_by_revision_ids(self, revision_ids):
1191
self.signatures.add_lines((revision_id,), (),
1192
osutils.split_lines(signature))
1194
def find_text_key_references(self):
1195
"""Find the text key references within the repository.
1197
:return: a dictionary mapping (file_id, revision_id) tuples to altered file-ids to an iterable of
1198
revision_ids. Each altered file-ids has the exact revision_ids that
1199
altered it listed explicitly.
1200
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1201
to whether they were referred to by the inventory of the
1202
revision_id that they contain. The inventory texts from all present
1203
revision ids are assessed to generate this report.
1205
revision_keys = self.revisions.keys()
1206
w = self.inventories
1207
pb = ui.ui_factory.nested_progress_bar()
1209
return self._find_text_key_references_from_xml_inventory_lines(
1210
w.iter_lines_added_or_present_in_keys(revision_keys, pb=pb))
1214
def _find_text_key_references_from_xml_inventory_lines(self,
1216
"""Core routine for extracting references to texts from inventories.
1218
This performs the translation of xml lines to revision ids.
1220
:param line_iterator: An iterator of lines, origin_version_id
1221
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1222
to whether they were referred to by the inventory of the
1223
revision_id that they contain. Note that if that revision_id was
1224
not part of the line_iterator's output then False will be given -
1225
even though it may actually refer to that key.
1227
if not self._serializer.support_altered_by_hack:
1228
raise AssertionError(
1229
"_find_text_key_references_from_xml_inventory_lines only "
1230
"supported for branches which store inventory as unnested xml"
1231
", not on %r" % self)
1234
# this code needs to read every new line in every inventory for the
1235
# inventories [revision_ids]. Seeing a line twice is ok. Seeing a line
1236
# not present in one of those inventories is unnecessary but not
1237
# harmful because we are filtering by the revision id marker in the
1238
# inventory lines : we only select file ids altered in one of those
1239
# revisions. We don't need to see all lines in the inventory because
1240
# only those added in an inventory in rev X can contain a revision=X
1242
unescape_revid_cache = {}
1243
unescape_fileid_cache = {}
1245
# jam 20061218 In a big fetch, this handles hundreds of thousands
1246
# of lines, so it has had a lot of inlining and optimizing done.
1247
# Sorry that it is a little bit messy.
1248
# Move several functions to be local variables, since this is a long
1250
search = self._file_ids_altered_regex.search
1251
unescape = _unescape_xml
1252
setdefault = result.setdefault
1253
for line, line_key in line_iterator:
1254
match = search(line)
1257
# One call to match.group() returning multiple items is quite a
1258
# bit faster than 2 calls to match.group() each returning 1
1259
file_id, revision_id = match.group('file_id', 'revision_id')
1261
# Inlining the cache lookups helps a lot when you make 170,000
1262
# lines and 350k ids, versus 8.4 unique ids.
1263
# Using a cache helps in 2 ways:
1264
# 1) Avoids unnecessary decoding calls
1265
# 2) Re-uses cached strings, which helps in future set and
1267
# (2) is enough that removing encoding entirely along with
1268
# the cache (so we are using plain strings) results in no
1269
# performance improvement.
1271
revision_id = unescape_revid_cache[revision_id]
1273
unescaped = unescape(revision_id)
1274
unescape_revid_cache[revision_id] = unescaped
1275
revision_id = unescaped
1277
# Note that unconditionally unescaping means that we deserialise
1278
# every fileid, which for general 'pull' is not great, but we don't
1279
# really want to have some many fulltexts that this matters anyway.
1282
file_id = unescape_fileid_cache[file_id]
1284
unescaped = unescape(file_id)
1285
unescape_fileid_cache[file_id] = unescaped
1288
key = (file_id, revision_id)
1289
setdefault(key, False)
1290
if revision_id == line_key[-1]:
1294
def _find_file_ids_from_xml_inventory_lines(self, line_iterator,
1296
"""Helper routine for fileids_altered_by_revision_ids.
1298
This performs the translation of xml lines to revision ids.
1300
:param line_iterator: An iterator of lines, origin_version_id
1301
:param revision_ids: The revision ids to filter for. This should be a
1302
set or other type which supports efficient __contains__ lookups, as
1303
the revision id from each parsed line will be looked up in the
1304
revision_ids filter.
1305
:return: a dictionary mapping altered file-ids to an iterable of
1306
revision_ids. Each altered file-ids has the exact revision_ids that
1307
altered it listed explicitly.
1310
setdefault = result.setdefault
1312
self._find_text_key_references_from_xml_inventory_lines(
1313
line_iterator).iterkeys():
1314
# once data is all ensured-consistent; then this is
1315
# if revision_id == version_id
1316
if key[-1:] in revision_ids:
1317
setdefault(key[0], set()).add(key[-1])
1320
def fileids_altered_by_revision_ids(self, revision_ids, _inv_weave=None):
934
1321
"""Find the file ids and versions affected by revisions.
936
1323
:param revisions: an iterable containing revision ids.
1324
:param _inv_weave: The inventory weave from this repository or None.
1325
If None, the inventory weave will be opened automatically.
937
1326
:return: a dictionary mapping altered file-ids to an iterable of
938
revision_ids. Each altered file-ids has the exact revision_ids
939
that altered it listed explicitly.
1327
revision_ids. Each altered file-ids has the exact revision_ids that
1328
altered it listed explicitly.
941
raise NotImplementedError(self.fileids_altered_by_revision_ids)
1330
selected_keys = set((revid,) for revid in revision_ids)
1331
w = _inv_weave or self.inventories
1332
pb = ui.ui_factory.nested_progress_bar()
1334
return self._find_file_ids_from_xml_inventory_lines(
1335
w.iter_lines_added_or_present_in_keys(
1336
selected_keys, pb=pb),
943
1341
def iter_files_bytes(self, desired_files):
944
1342
"""Iterate through file versions.
951
1349
uniquely identify the file version in the caller's context. (Examples:
952
1350
an index number or a TreeTransform trans_id.)
1352
bytes_iterator is an iterable of bytestrings for the file. The
1353
kind of iterable and length of the bytestrings are unspecified, but for
1354
this implementation, it is a list of bytes produced by
1355
VersionedFile.get_record_stream().
954
1357
:param desired_files: a list of (file_id, revision_id, identifier)
957
raise NotImplementedError(self.iter_files_bytes)
959
def get_rev_id_for_revno(self, revno, known_pair):
960
"""Return the revision id of a revno, given a later (revno, revid)
961
pair in the same history.
963
:return: if found (True, revid). If the available history ran out
964
before reaching the revno, then this returns
965
(False, (closest_revno, closest_revid)).
967
known_revno, known_revid = known_pair
968
partial_history = [known_revid]
969
distance_from_known = known_revno - revno
970
if distance_from_known < 0:
972
'requested revno (%d) is later than given known revno (%d)'
973
% (revno, known_revno))
976
self, partial_history, stop_index=distance_from_known)
977
except errors.RevisionNotPresent, err:
978
if err.revision_id == known_revid:
979
# The start revision (known_revid) wasn't found.
981
# This is a stacked repository with no fallbacks, or a there's a
982
# left-hand ghost. Either way, even though the revision named in
983
# the error isn't in this repo, we know it's the next step in this
985
partial_history.append(err.revision_id)
986
if len(partial_history) <= distance_from_known:
987
# Didn't find enough history to get a revid for the revno.
988
earliest_revno = known_revno - len(partial_history) + 1
989
return (False, (earliest_revno, partial_history[-1]))
990
if len(partial_history) - 1 > distance_from_known:
991
raise AssertionError('_iter_for_revno returned too much history')
992
return (True, partial_history[-1])
994
@symbol_versioning.deprecated_method(symbol_versioning.deprecated_in((2, 4, 0)))
1360
transaction = self.get_transaction()
1362
for file_id, revision_id, callable_data in desired_files:
1363
text_keys[(file_id, revision_id)] = callable_data
1364
for record in self.texts.get_record_stream(text_keys, 'unordered', True):
1365
if record.storage_kind == 'absent':
1366
raise errors.RevisionNotPresent(record.key, self)
1367
yield text_keys[record.key], record.get_bytes_as('fulltext')
1369
def _generate_text_key_index(self, text_key_references=None,
1371
"""Generate a new text key index for the repository.
1373
This is an expensive function that will take considerable time to run.
1375
:return: A dict mapping text keys ((file_id, revision_id) tuples) to a
1376
list of parents, also text keys. When a given key has no parents,
1377
the parents list will be [NULL_REVISION].
1379
# All revisions, to find inventory parents.
1380
if ancestors is None:
1381
graph = self.get_graph()
1382
ancestors = graph.get_parent_map(self.all_revision_ids())
1383
if text_key_references is None:
1384
text_key_references = self.find_text_key_references()
1385
pb = ui.ui_factory.nested_progress_bar()
1387
return self._do_generate_text_key_index(ancestors,
1388
text_key_references, pb)
1392
def _do_generate_text_key_index(self, ancestors, text_key_references, pb):
1393
"""Helper for _generate_text_key_index to avoid deep nesting."""
1394
revision_order = tsort.topo_sort(ancestors)
1395
invalid_keys = set()
1397
for revision_id in revision_order:
1398
revision_keys[revision_id] = set()
1399
text_count = len(text_key_references)
1400
# a cache of the text keys to allow reuse; costs a dict of all the
1401
# keys, but saves a 2-tuple for every child of a given key.
1403
for text_key, valid in text_key_references.iteritems():
1405
invalid_keys.add(text_key)
1407
revision_keys[text_key[1]].add(text_key)
1408
text_key_cache[text_key] = text_key
1409
del text_key_references
1411
text_graph = graph.Graph(graph.DictParentsProvider(text_index))
1412
NULL_REVISION = _mod_revision.NULL_REVISION
1413
# Set a cache with a size of 10 - this suffices for bzr.dev but may be
1414
# too small for large or very branchy trees. However, for 55K path
1415
# trees, it would be easy to use too much memory trivially. Ideally we
1416
# could gauge this by looking at available real memory etc, but this is
1417
# always a tricky proposition.
1418
inventory_cache = lru_cache.LRUCache(10)
1419
batch_size = 10 # should be ~150MB on a 55K path tree
1420
batch_count = len(revision_order) / batch_size + 1
1422
pb.update("Calculating text parents.", processed_texts, text_count)
1423
for offset in xrange(batch_count):
1424
to_query = revision_order[offset * batch_size:(offset + 1) *
1428
for rev_tree in self.revision_trees(to_query):
1429
revision_id = rev_tree.get_revision_id()
1430
parent_ids = ancestors[revision_id]
1431
for text_key in revision_keys[revision_id]:
1432
pb.update("Calculating text parents.", processed_texts)
1433
processed_texts += 1
1434
candidate_parents = []
1435
for parent_id in parent_ids:
1436
parent_text_key = (text_key[0], parent_id)
1438
check_parent = parent_text_key not in \
1439
revision_keys[parent_id]
1441
# the parent parent_id is a ghost:
1442
check_parent = False
1443
# truncate the derived graph against this ghost.
1444
parent_text_key = None
1446
# look at the parent commit details inventories to
1447
# determine possible candidates in the per file graph.
1450
inv = inventory_cache[parent_id]
1452
inv = self.revision_tree(parent_id).inventory
1453
inventory_cache[parent_id] = inv
1454
parent_entry = inv._byid.get(text_key[0], None)
1455
if parent_entry is not None:
1457
text_key[0], parent_entry.revision)
1459
parent_text_key = None
1460
if parent_text_key is not None:
1461
candidate_parents.append(
1462
text_key_cache[parent_text_key])
1463
parent_heads = text_graph.heads(candidate_parents)
1464
new_parents = list(parent_heads)
1465
new_parents.sort(key=lambda x:candidate_parents.index(x))
1466
if new_parents == []:
1467
new_parents = [NULL_REVISION]
1468
text_index[text_key] = new_parents
1470
for text_key in invalid_keys:
1471
text_index[text_key] = [NULL_REVISION]
1474
def item_keys_introduced_by(self, revision_ids, _files_pb=None):
1475
"""Get an iterable listing the keys of all the data introduced by a set
1478
The keys will be ordered so that the corresponding items can be safely
1479
fetched and inserted in that order.
1481
:returns: An iterable producing tuples of (knit-kind, file-id,
1482
versions). knit-kind is one of 'file', 'inventory', 'signatures',
1483
'revisions'. file-id is None unless knit-kind is 'file'.
1485
# XXX: it's a bit weird to control the inventory weave caching in this
1486
# generator. Ideally the caching would be done in fetch.py I think. Or
1487
# maybe this generator should explicitly have the contract that it
1488
# should not be iterated until the previously yielded item has been
1490
inv_w = self.inventories
1492
# file ids that changed
1493
file_ids = self.fileids_altered_by_revision_ids(revision_ids, inv_w)
1495
num_file_ids = len(file_ids)
1496
for file_id, altered_versions in file_ids.iteritems():
1497
if _files_pb is not None:
1498
_files_pb.update("fetch texts", count, num_file_ids)
1500
yield ("file", file_id, altered_versions)
1501
# We're done with the files_pb. Note that it finished by the caller,
1502
# just as it was created by the caller.
1506
yield ("inventory", None, revision_ids)
1509
revisions_with_signatures = set()
1510
for rev_id in revision_ids:
1512
self.get_signature_text(rev_id)
1513
except errors.NoSuchRevision:
1517
revisions_with_signatures.add(rev_id)
1518
yield ("signatures", None, revisions_with_signatures)
1521
yield ("revisions", None, revision_ids)
1524
def get_inventory(self, revision_id):
1525
"""Get Inventory object by revision id."""
1526
return self.iter_inventories([revision_id]).next()
1528
def iter_inventories(self, revision_ids):
1529
"""Get many inventories by revision_ids.
1531
This will buffer some or all of the texts used in constructing the
1532
inventories in memory, but will only parse a single inventory at a
1535
:return: An iterator of inventories.
1537
if ((None in revision_ids)
1538
or (_mod_revision.NULL_REVISION in revision_ids)):
1539
raise ValueError('cannot get null revision inventory')
1540
return self._iter_inventories(revision_ids)
1542
def _iter_inventories(self, revision_ids):
1543
"""single-document based inventory iteration."""
1544
for text, revision_id in self._iter_inventory_xmls(revision_ids):
1545
yield self.deserialise_inventory(revision_id, text)
1547
def _iter_inventory_xmls(self, revision_ids):
1548
keys = [(revision_id,) for revision_id in revision_ids]
1549
stream = self.inventories.get_record_stream(keys, 'unordered', True)
1551
for record in stream:
1552
if record.storage_kind != 'absent':
1553
texts[record.key] = record.get_bytes_as('fulltext')
1555
raise errors.NoSuchRevision(self, record.key)
1557
yield texts[key], key[-1]
1559
def deserialise_inventory(self, revision_id, xml):
1560
"""Transform the xml into an inventory object.
1562
:param revision_id: The expected revision id of the inventory.
1563
:param xml: A serialised inventory.
1565
result = self._serializer.read_inventory_from_string(xml, revision_id)
1566
if result.revision_id != revision_id:
1567
raise AssertionError('revision id mismatch %s != %s' % (
1568
result.revision_id, revision_id))
1571
def serialise_inventory(self, inv):
1572
return self._serializer.write_inventory_to_string(inv)
1574
def _serialise_inventory_to_lines(self, inv):
1575
return self._serializer.write_inventory_to_lines(inv)
1577
def get_serializer_format(self):
1578
return self._serializer.format_num
1581
def get_inventory_xml(self, revision_id):
1582
"""Get inventory XML as a file object."""
1583
texts = self._iter_inventory_xmls([revision_id])
1585
text, revision_id = texts.next()
1586
except StopIteration:
1587
raise errors.HistoryMissing(self, 'inventory', revision_id)
1591
def get_inventory_sha1(self, revision_id):
1592
"""Return the sha1 hash of the inventory entry
1594
return self.get_revision(revision_id).inventory_sha1
995
1596
def iter_reverse_revision_history(self, revision_id):
996
1597
"""Iterate backwards through revision ids in the lefthand history
1695
2349
_optimisers = []
1696
2350
"""The available optimised InterRepository types."""
1699
2352
def copy_content(self, revision_id=None):
1700
"""Make a complete copy of the content in self into destination.
1702
This is a destructive operation! Do not use it on existing
1705
:param revision_id: Only copy the content needed to construct
1706
revision_id and its parents.
1709
self.target.set_make_working_trees(self.source.make_working_trees())
1710
except NotImplementedError:
1712
self.target.fetch(self.source, revision_id=revision_id)
1715
def fetch(self, revision_id=None, find_ghosts=False,
2353
raise NotImplementedError(self.copy_content)
2355
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
1717
2356
"""Fetch the content required to construct revision_id.
1719
2358
The content is copied from self.source to self.target.
1721
2360
:param revision_id: if None all content is copied, if NULL_REVISION no
1722
2361
content is copied.
1725
raise NotImplementedError(self.fetch)
1728
def search_missing_revision_ids(self,
1729
revision_id=symbol_versioning.DEPRECATED_PARAMETER,
1730
find_ghosts=True, revision_ids=None, if_present_ids=None,
1732
"""Return the revision ids that source has that target does not.
1734
:param revision_id: only return revision ids included by this
1736
:param revision_ids: return revision ids included by these
1737
revision_ids. NoSuchRevision will be raised if any of these
1738
revisions are not present.
1739
:param if_present_ids: like revision_ids, but will not cause
1740
NoSuchRevision if any of these are absent, instead they will simply
1741
not be in the result. This is useful for e.g. finding revisions
1742
to fetch for tags, which may reference absent revisions.
1743
:param find_ghosts: If True find missing revisions in deep history
1744
rather than just finding the surface difference.
1745
:param limit: Maximum number of revisions to return, topologically
2362
:param pb: optional progress bar to use for progress reports. If not
2363
provided a default one will be created.
2365
:returns: (copied_revision_count, failures).
2367
# Normally we should find a specific InterRepository subclass to do
2368
# the fetch; if nothing else then at least InterSameDataRepository.
2369
# If none of them is suitable it looks like fetching is not possible;
2370
# we try to give a good message why. _assert_same_model will probably
2371
# give a helpful message; otherwise a generic one.
2372
self._assert_same_model(self.source, self.target)
2373
raise errors.IncompatibleRepositories(self.source, self.target,
2374
"no suitableInterRepository found")
2376
def _walk_to_common_revisions(self, revision_ids):
2377
"""Walk out from revision_ids in source to revisions target has.
2379
:param revision_ids: The start point for the search.
2380
:return: A set of revision ids.
2382
target_graph = self.target.get_graph()
2383
revision_ids = frozenset(revision_ids)
2384
if set(target_graph.get_parent_map(revision_ids)) == revision_ids:
2385
return graph.SearchResult(revision_ids, set(), 0, set())
2386
missing_revs = set()
2387
source_graph = self.source.get_graph()
2388
# ensure we don't pay silly lookup costs.
2389
searcher = source_graph._make_breadth_first_searcher(revision_ids)
2390
null_set = frozenset([_mod_revision.NULL_REVISION])
2393
next_revs, ghosts = searcher.next_with_ghosts()
2394
except StopIteration:
2396
if revision_ids.intersection(ghosts):
2397
absent_ids = set(revision_ids.intersection(ghosts))
2398
# If all absent_ids are present in target, no error is needed.
2399
absent_ids.difference_update(
2400
set(target_graph.get_parent_map(absent_ids)))
2402
raise errors.NoSuchRevision(self.source, absent_ids.pop())
2403
# we don't care about other ghosts as we can't fetch them and
2404
# haven't been asked to.
2405
next_revs = set(next_revs)
2406
# we always have NULL_REVISION present.
2407
have_revs = set(target_graph.get_parent_map(next_revs)).union(null_set)
2408
missing_revs.update(next_revs - have_revs)
2409
searcher.stop_searching_any(have_revs)
2410
return searcher.get_result()
2412
@deprecated_method(one_two)
2414
def missing_revision_ids(self, revision_id=None, find_ghosts=True):
2415
"""Return the revision ids that source has that target does not.
2417
These are returned in topological order.
2419
:param revision_id: only return revision ids included by this
2421
:param find_ghosts: If True find missing revisions in deep history
2422
rather than just finding the surface difference.
2424
return list(self.search_missing_revision_ids(
2425
revision_id, find_ghosts).get_keys())
2428
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2429
"""Return the revision ids that source has that target does not.
2431
:param revision_id: only return revision ids included by this
2433
:param find_ghosts: If True find missing revisions in deep history
2434
rather than just finding the surface difference.
1747
2435
:return: A bzrlib.graph.SearchResult.
1749
raise NotImplementedError(self.search_missing_revision_ids)
2437
# stop searching at found target revisions.
2438
if not find_ghosts and revision_id is not None:
2439
return self._walk_to_common_revisions([revision_id])
2440
# generic, possibly worst case, slow code path.
2441
target_ids = set(self.target.all_revision_ids())
2442
if revision_id is not None:
2443
source_ids = self.source.get_ancestry(revision_id)
2444
if source_ids[0] is not None:
2445
raise AssertionError()
2448
source_ids = self.source.all_revision_ids()
2449
result_set = set(source_ids).difference(target_ids)
2450
return self.source.revision_ids_to_search_result(result_set)
1752
2453
def _same_model(source, target):
1753
2454
"""True if source and target have the same data representation.
1755
2456
Note: this is always called on the base class; overriding it in a
1756
2457
subclass will have no effect.
1773
2474
"different serializers")
2477
class InterSameDataRepository(InterRepository):
2478
"""Code for converting between repositories that represent the same data.
2480
Data format and model must match for this to work.
2484
def _get_repo_format_to_test(self):
2485
"""Repository format for testing with.
2487
InterSameData can pull from subtree to subtree and from non-subtree to
2488
non-subtree, so we test this with the richest repository format.
2490
from bzrlib.repofmt import knitrepo
2491
return knitrepo.RepositoryFormatKnit3()
2494
def is_compatible(source, target):
2495
return InterRepository._same_model(source, target)
2498
def copy_content(self, revision_id=None):
2499
"""Make a complete copy of the content in self into destination.
2501
This copies both the repository's revision data, and configuration information
2502
such as the make_working_trees setting.
2504
This is a destructive operation! Do not use it on existing
2507
:param revision_id: Only copy the content needed to construct
2508
revision_id and its parents.
2511
self.target.set_make_working_trees(self.source.make_working_trees())
2512
except NotImplementedError:
2514
# but don't bother fetching if we have the needed data now.
2515
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
2516
self.target.has_revision(revision_id)):
2518
self.target.fetch(self.source, revision_id=revision_id)
2521
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2522
"""See InterRepository.fetch()."""
2523
from bzrlib.fetch import RepoFetcher
2524
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2525
self.source, self.source._format, self.target,
2526
self.target._format)
2527
f = RepoFetcher(to_repository=self.target,
2528
from_repository=self.source,
2529
last_revision=revision_id,
2530
pb=pb, find_ghosts=find_ghosts)
2531
return f.count_copied, f.failed_revisions
2534
class InterWeaveRepo(InterSameDataRepository):
2535
"""Optimised code paths between Weave based repositories.
2537
This should be in bzrlib/repofmt/weaverepo.py but we have not yet
2538
implemented lazy inter-object optimisation.
2542
def _get_repo_format_to_test(self):
2543
from bzrlib.repofmt import weaverepo
2544
return weaverepo.RepositoryFormat7()
2547
def is_compatible(source, target):
2548
"""Be compatible with known Weave formats.
2550
We don't test for the stores being of specific types because that
2551
could lead to confusing results, and there is no need to be
2554
from bzrlib.repofmt.weaverepo import (
2560
return (isinstance(source._format, (RepositoryFormat5,
2562
RepositoryFormat7)) and
2563
isinstance(target._format, (RepositoryFormat5,
2565
RepositoryFormat7)))
2566
except AttributeError:
2570
def copy_content(self, revision_id=None):
2571
"""See InterRepository.copy_content()."""
2572
# weave specific optimised path:
2574
self.target.set_make_working_trees(self.source.make_working_trees())
2575
except (errors.RepositoryUpgradeRequired, NotImplemented):
2577
# FIXME do not peek!
2578
if self.source._transport.listable():
2579
pb = ui.ui_factory.nested_progress_bar()
2581
self.target.texts.insert_record_stream(
2582
self.source.texts.get_record_stream(
2583
self.source.texts.keys(), 'topological', False))
2584
pb.update('copying inventory', 0, 1)
2585
self.target.inventories.insert_record_stream(
2586
self.source.inventories.get_record_stream(
2587
self.source.inventories.keys(), 'topological', False))
2588
self.target.signatures.insert_record_stream(
2589
self.source.signatures.get_record_stream(
2590
self.source.signatures.keys(),
2592
self.target.revisions.insert_record_stream(
2593
self.source.revisions.get_record_stream(
2594
self.source.revisions.keys(),
2595
'topological', True))
2599
self.target.fetch(self.source, revision_id=revision_id)
2602
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2603
"""See InterRepository.fetch()."""
2604
from bzrlib.fetch import RepoFetcher
2605
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2606
self.source, self.source._format, self.target, self.target._format)
2607
f = RepoFetcher(to_repository=self.target,
2608
from_repository=self.source,
2609
last_revision=revision_id,
2610
pb=pb, find_ghosts=find_ghosts)
2611
return f.count_copied, f.failed_revisions
2614
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2615
"""See InterRepository.missing_revision_ids()."""
2616
# we want all revisions to satisfy revision_id in source.
2617
# but we don't want to stat every file here and there.
2618
# we want then, all revisions other needs to satisfy revision_id
2619
# checked, but not those that we have locally.
2620
# so the first thing is to get a subset of the revisions to
2621
# satisfy revision_id in source, and then eliminate those that
2622
# we do already have.
2623
# this is slow on high latency connection to self, but as as this
2624
# disk format scales terribly for push anyway due to rewriting
2625
# inventory.weave, this is considered acceptable.
2627
if revision_id is not None:
2628
source_ids = self.source.get_ancestry(revision_id)
2629
if source_ids[0] is not None:
2630
raise AssertionError()
2633
source_ids = self.source._all_possible_ids()
2634
source_ids_set = set(source_ids)
2635
# source_ids is the worst possible case we may need to pull.
2636
# now we want to filter source_ids against what we actually
2637
# have in target, but don't try to check for existence where we know
2638
# we do not have a revision as that would be pointless.
2639
target_ids = set(self.target._all_possible_ids())
2640
possibly_present_revisions = target_ids.intersection(source_ids_set)
2641
actually_present_revisions = set(
2642
self.target._eliminate_revisions_not_present(possibly_present_revisions))
2643
required_revisions = source_ids_set.difference(actually_present_revisions)
2644
if revision_id is not None:
2645
# we used get_ancestry to determine source_ids then we are assured all
2646
# revisions referenced are present as they are installed in topological order.
2647
# and the tip revision was validated by get_ancestry.
2648
result_set = required_revisions
2650
# if we just grabbed the possibly available ids, then
2651
# we only have an estimate of whats available and need to validate
2652
# that against the revision records.
2654
self.source._eliminate_revisions_not_present(required_revisions))
2655
return self.source.revision_ids_to_search_result(result_set)
2658
class InterKnitRepo(InterSameDataRepository):
2659
"""Optimised code paths between Knit based repositories."""
2662
def _get_repo_format_to_test(self):
2663
from bzrlib.repofmt import knitrepo
2664
return knitrepo.RepositoryFormatKnit1()
2667
def is_compatible(source, target):
2668
"""Be compatible with known Knit formats.
2670
We don't test for the stores being of specific types because that
2671
could lead to confusing results, and there is no need to be
2674
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit
2676
are_knits = (isinstance(source._format, RepositoryFormatKnit) and
2677
isinstance(target._format, RepositoryFormatKnit))
2678
except AttributeError:
2680
return are_knits and InterRepository._same_model(source, target)
2683
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2684
"""See InterRepository.fetch()."""
2685
from bzrlib.fetch import RepoFetcher
2686
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2687
self.source, self.source._format, self.target, self.target._format)
2688
f = RepoFetcher(to_repository=self.target,
2689
from_repository=self.source,
2690
last_revision=revision_id,
2691
pb=pb, find_ghosts=find_ghosts)
2692
return f.count_copied, f.failed_revisions
2695
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2696
"""See InterRepository.missing_revision_ids()."""
2697
if revision_id is not None:
2698
source_ids = self.source.get_ancestry(revision_id)
2699
if source_ids[0] is not None:
2700
raise AssertionError()
2703
source_ids = self.source.all_revision_ids()
2704
source_ids_set = set(source_ids)
2705
# source_ids is the worst possible case we may need to pull.
2706
# now we want to filter source_ids against what we actually
2707
# have in target, but don't try to check for existence where we know
2708
# we do not have a revision as that would be pointless.
2709
target_ids = set(self.target.all_revision_ids())
2710
possibly_present_revisions = target_ids.intersection(source_ids_set)
2711
actually_present_revisions = set(
2712
self.target._eliminate_revisions_not_present(possibly_present_revisions))
2713
required_revisions = source_ids_set.difference(actually_present_revisions)
2714
if revision_id is not None:
2715
# we used get_ancestry to determine source_ids then we are assured all
2716
# revisions referenced are present as they are installed in topological order.
2717
# and the tip revision was validated by get_ancestry.
2718
result_set = required_revisions
2720
# if we just grabbed the possibly available ids, then
2721
# we only have an estimate of whats available and need to validate
2722
# that against the revision records.
2724
self.source._eliminate_revisions_not_present(required_revisions))
2725
return self.source.revision_ids_to_search_result(result_set)
2728
class InterPackRepo(InterSameDataRepository):
2729
"""Optimised code paths between Pack based repositories."""
2732
def _get_repo_format_to_test(self):
2733
from bzrlib.repofmt import pack_repo
2734
return pack_repo.RepositoryFormatKnitPack1()
2737
def is_compatible(source, target):
2738
"""Be compatible with known Pack formats.
2740
We don't test for the stores being of specific types because that
2741
could lead to confusing results, and there is no need to be
2744
from bzrlib.repofmt.pack_repo import RepositoryFormatPack
2746
are_packs = (isinstance(source._format, RepositoryFormatPack) and
2747
isinstance(target._format, RepositoryFormatPack))
2748
except AttributeError:
2750
return are_packs and InterRepository._same_model(source, target)
2753
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2754
"""See InterRepository.fetch()."""
2755
if (len(self.source._fallback_repositories) > 0 or
2756
len(self.target._fallback_repositories) > 0):
2757
# The pack layer is not aware of fallback repositories, so when
2758
# fetching from a stacked repository or into a stacked repository
2759
# we use the generic fetch logic which uses the VersionedFiles
2760
# attributes on repository.
2761
from bzrlib.fetch import RepoFetcher
2762
fetcher = RepoFetcher(self.target, self.source, revision_id,
2764
return fetcher.count_copied, fetcher.failed_revisions
2765
from bzrlib.repofmt.pack_repo import Packer
2766
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2767
self.source, self.source._format, self.target, self.target._format)
2768
self.count_copied = 0
2769
if revision_id is None:
2771
# everything to do - use pack logic
2772
# to fetch from all packs to one without
2773
# inventory parsing etc, IFF nothing to be copied is in the target.
2775
source_revision_ids = frozenset(self.source.all_revision_ids())
2776
revision_ids = source_revision_ids - \
2777
frozenset(self.target.get_parent_map(source_revision_ids))
2778
revision_keys = [(revid,) for revid in revision_ids]
2779
index = self.target._pack_collection.revision_index.combined_index
2780
present_revision_ids = set(item[1][0] for item in
2781
index.iter_entries(revision_keys))
2782
revision_ids = set(revision_ids) - present_revision_ids
2783
# implementing the TODO will involve:
2784
# - detecting when all of a pack is selected
2785
# - avoiding as much as possible pre-selection, so the
2786
# more-core routines such as create_pack_from_packs can filter in
2787
# a just-in-time fashion. (though having a HEADS list on a
2788
# repository might make this a lot easier, because we could
2789
# sensibly detect 'new revisions' without doing a full index scan.
2790
elif _mod_revision.is_null(revision_id):
2795
revision_ids = self.search_missing_revision_ids(revision_id,
2796
find_ghosts=find_ghosts).get_keys()
2797
except errors.NoSuchRevision:
2798
raise errors.InstallFailed([revision_id])
2799
if len(revision_ids) == 0:
2801
packs = self.source._pack_collection.all_packs()
2802
pack = Packer(self.target._pack_collection, packs, '.fetch',
2803
revision_ids).pack()
2804
if pack is not None:
2805
self.target._pack_collection._save_pack_names()
2806
# Trigger an autopack. This may duplicate effort as we've just done
2807
# a pack creation, but for now it is simpler to think about as
2808
# 'upload data, then repack if needed'.
2809
self.target._pack_collection.autopack()
2810
return (pack.get_revision_count(), [])
2815
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2816
"""See InterRepository.missing_revision_ids().
2818
:param find_ghosts: Find ghosts throughout the ancestry of
2821
if not find_ghosts and revision_id is not None:
2822
return self._walk_to_common_revisions([revision_id])
2823
elif revision_id is not None:
2824
# Find ghosts: search for revisions pointing from one repository to
2825
# the other, and vice versa, anywhere in the history of revision_id.
2826
graph = self.target.get_graph(other_repository=self.source)
2827
searcher = graph._make_breadth_first_searcher([revision_id])
2831
next_revs, ghosts = searcher.next_with_ghosts()
2832
except StopIteration:
2834
if revision_id in ghosts:
2835
raise errors.NoSuchRevision(self.source, revision_id)
2836
found_ids.update(next_revs)
2837
found_ids.update(ghosts)
2838
found_ids = frozenset(found_ids)
2839
# Double query here: should be able to avoid this by changing the
2840
# graph api further.
2841
result_set = found_ids - frozenset(
2842
self.target.get_parent_map(found_ids))
2844
source_ids = self.source.all_revision_ids()
2845
# source_ids is the worst possible case we may need to pull.
2846
# now we want to filter source_ids against what we actually
2847
# have in target, but don't try to check for existence where we know
2848
# we do not have a revision as that would be pointless.
2849
target_ids = set(self.target.all_revision_ids())
2850
result_set = set(source_ids).difference(target_ids)
2851
return self.source.revision_ids_to_search_result(result_set)
2854
class InterModel1and2(InterRepository):
2857
def _get_repo_format_to_test(self):
2861
def is_compatible(source, target):
2862
if not source.supports_rich_root() and target.supports_rich_root():
2868
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2869
"""See InterRepository.fetch()."""
2870
from bzrlib.fetch import Model1toKnit2Fetcher
2871
f = Model1toKnit2Fetcher(to_repository=self.target,
2872
from_repository=self.source,
2873
last_revision=revision_id,
2874
pb=pb, find_ghosts=find_ghosts)
2875
return f.count_copied, f.failed_revisions
2878
def copy_content(self, revision_id=None):
2879
"""Make a complete copy of the content in self into destination.
2881
This is a destructive operation! Do not use it on existing
2884
:param revision_id: Only copy the content needed to construct
2885
revision_id and its parents.
2888
self.target.set_make_working_trees(self.source.make_working_trees())
2889
except NotImplementedError:
2891
# but don't bother fetching if we have the needed data now.
2892
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
2893
self.target.has_revision(revision_id)):
2895
self.target.fetch(self.source, revision_id=revision_id)
2898
class InterKnit1and2(InterKnitRepo):
2901
def _get_repo_format_to_test(self):
2905
def is_compatible(source, target):
2906
"""Be compatible with Knit1 source and Knit3 target"""
2908
from bzrlib.repofmt.knitrepo import (
2909
RepositoryFormatKnit1,
2910
RepositoryFormatKnit3,
2912
from bzrlib.repofmt.pack_repo import (
2913
RepositoryFormatKnitPack1,
2914
RepositoryFormatKnitPack3,
2915
RepositoryFormatKnitPack4,
2916
RepositoryFormatKnitPack5,
2917
RepositoryFormatKnitPack5RichRoot,
2918
RepositoryFormatPackDevelopment1,
2919
RepositoryFormatPackDevelopment1Subtree,
2922
RepositoryFormatKnit1, # no rr, no subtree
2923
RepositoryFormatKnitPack1, # no rr, no subtree
2924
RepositoryFormatPackDevelopment1, # no rr, no subtree
2925
RepositoryFormatKnitPack5, # no rr, no subtree
2928
RepositoryFormatKnit3, # rr, subtree
2929
RepositoryFormatKnitPack3, # rr, subtree
2930
RepositoryFormatKnitPack4, # rr, no subtree
2931
RepositoryFormatKnitPack5RichRoot,# rr, no subtree
2932
RepositoryFormatPackDevelopment1Subtree, # rr, subtree
2934
for format in norichroot:
2935
if format.rich_root_data:
2936
raise AssertionError('Format %s is a rich-root format'
2937
' but is included in the non-rich-root list'
2939
for format in richroot:
2940
if not format.rich_root_data:
2941
raise AssertionError('Format %s is not a rich-root format'
2942
' but is included in the rich-root list'
2944
# TODO: One alternative is to just check format.rich_root_data,
2945
# instead of keeping membership lists. However, the formats
2946
# *also* have to use the same 'Knit' style of storage
2947
# (line-deltas, fulltexts, etc.)
2948
return (isinstance(source._format, norichroot) and
2949
isinstance(target._format, richroot))
2950
except AttributeError:
2954
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2955
"""See InterRepository.fetch()."""
2956
from bzrlib.fetch import Knit1to2Fetcher
2957
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2958
self.source, self.source._format, self.target,
2959
self.target._format)
2960
f = Knit1to2Fetcher(to_repository=self.target,
2961
from_repository=self.source,
2962
last_revision=revision_id,
2963
pb=pb, find_ghosts=find_ghosts)
2964
return f.count_copied, f.failed_revisions
2967
class InterDifferingSerializer(InterKnitRepo):
2970
def _get_repo_format_to_test(self):
2974
def is_compatible(source, target):
2975
"""Be compatible with Knit2 source and Knit3 target"""
2976
if source.supports_rich_root() != target.supports_rich_root():
2978
# Ideally, we'd support fetching if the source had no tree references
2979
# even if it supported them...
2980
if (getattr(source, '_format.supports_tree_reference', False) and
2981
not getattr(target, '_format.supports_tree_reference', False)):
2986
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2987
"""See InterRepository.fetch()."""
2988
revision_ids = self.target.search_missing_revision_ids(self.source,
2989
revision_id, find_ghosts=find_ghosts).get_keys()
2990
revision_ids = tsort.topo_sort(
2991
self.source.get_graph().get_parent_map(revision_ids))
2992
def revisions_iterator():
2993
for current_revision_id in revision_ids:
2994
revision = self.source.get_revision(current_revision_id)
2995
tree = self.source.revision_tree(current_revision_id)
2997
signature = self.source.get_signature_text(
2998
current_revision_id)
2999
except errors.NoSuchRevision:
3001
yield revision, tree, signature
3003
my_pb = ui.ui_factory.nested_progress_bar()
3008
install_revisions(self.target, revisions_iterator(),
3009
len(revision_ids), pb)
3011
if my_pb is not None:
3013
return len(revision_ids), 0
3016
class InterOtherToRemote(InterRepository):
3018
def __init__(self, source, target):
3019
InterRepository.__init__(self, source, target)
3020
self._real_inter = None
3023
def is_compatible(source, target):
3024
if isinstance(target, remote.RemoteRepository):
3028
def _ensure_real_inter(self):
3029
if self._real_inter is None:
3030
self.target._ensure_real()
3031
real_target = self.target._real_repository
3032
self._real_inter = InterRepository.get(self.source, real_target)
3034
def copy_content(self, revision_id=None):
3035
self._ensure_real_inter()
3036
self._real_inter.copy_content(revision_id=revision_id)
3038
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3039
self._ensure_real_inter()
3040
return self._real_inter.fetch(revision_id=revision_id, pb=pb,
3041
find_ghosts=find_ghosts)
3044
def _get_repo_format_to_test(self):
3048
class InterRemoteToOther(InterRepository):
3050
def __init__(self, source, target):
3051
InterRepository.__init__(self, source, target)
3052
self._real_inter = None
3055
def is_compatible(source, target):
3056
if not isinstance(source, remote.RemoteRepository):
3058
# Is source's model compatible with target's model?
3059
source._ensure_real()
3060
real_source = source._real_repository
3061
if isinstance(real_source, remote.RemoteRepository):
3062
raise NotImplementedError(
3063
"We don't support remote repos backed by remote repos yet.")
3064
return InterRepository._same_model(real_source, target)
3066
def _ensure_real_inter(self):
3067
if self._real_inter is None:
3068
self.source._ensure_real()
3069
real_source = self.source._real_repository
3070
self._real_inter = InterRepository.get(real_source, self.target)
3072
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3073
self._ensure_real_inter()
3074
return self._real_inter.fetch(revision_id=revision_id, pb=pb,
3075
find_ghosts=find_ghosts)
3077
def copy_content(self, revision_id=None):
3078
self._ensure_real_inter()
3079
self._real_inter.copy_content(revision_id=revision_id)
3082
def _get_repo_format_to_test(self):
3087
InterRepository.register_optimiser(InterDifferingSerializer)
3088
InterRepository.register_optimiser(InterSameDataRepository)
3089
InterRepository.register_optimiser(InterWeaveRepo)
3090
InterRepository.register_optimiser(InterKnitRepo)
3091
InterRepository.register_optimiser(InterModel1and2)
3092
InterRepository.register_optimiser(InterKnit1and2)
3093
InterRepository.register_optimiser(InterPackRepo)
3094
InterRepository.register_optimiser(InterOtherToRemote)
3095
InterRepository.register_optimiser(InterRemoteToOther)
1776
3098
class CopyConverter(object):
1777
3099
"""A repository conversion tool which just performs a copy of the content.
1779
3101
This is slow but quite reliable.
1785
3107
:param target_format: The format the resulting repository should be.
1787
3109
self.target_format = target_format
1789
3111
def convert(self, repo, pb):
1790
3112
"""Perform the conversion of to_convert, giving feedback via pb.
1792
3114
:param to_convert: The disk object to convert.
1793
3115
:param pb: a progress bar to use for progress information.
1795
pb = ui.ui_factory.nested_progress_bar()
1798
3120
# this is only useful with metadir layouts - separated repo content.
1799
3121
# trigger an assertion if not such
1800
3122
repo._format.get_format_string()
1801
3123
self.repo_dir = repo.bzrdir
1802
pb.update('Moving repository to repository.backup')
3124
self.step('Moving repository to repository.backup')
1803
3125
self.repo_dir.transport.move('repository', 'repository.backup')
1804
3126
backup_transport = self.repo_dir.transport.clone('repository.backup')
1805
3127
repo._format.check_conversion_target(self.target_format)
1806
3128
self.source_repo = repo._format.open(self.repo_dir,
1808
3130
_override_transport=backup_transport)
1809
pb.update('Creating new repository')
3131
self.step('Creating new repository')
1810
3132
converted = self.target_format.initialize(self.repo_dir,
1811
3133
self.source_repo.is_shared())
1812
3134
converted.lock_write()
1814
pb.update('Copying content')
3136
self.step('Copying content into repository.')
1815
3137
self.source_repo.copy_content_into(converted)
1817
3139
converted.unlock()
1818
pb.update('Deleting old repository content')
3140
self.step('Deleting old repository content.')
1819
3141
self.repo_dir.transport.delete_tree('repository.backup')
1820
ui.ui_factory.note('repository converted')
3142
self.pb.note('repository converted')
3144
def step(self, message):
3145
"""Update the pb by a step."""
3147
self.pb.update(message, self.count, self.total)
3159
def _unescaper(match, _map=_unescape_map):
3160
code = match.group(1)
3164
if not code.startswith('#'):
3166
return unichr(int(code[1:])).encode('utf8')
3172
def _unescape_xml(data):
3173
"""Unescape predefined XML entities in a string of data."""
3175
if _unescape_re is None:
3176
_unescape_re = re.compile('\&([^;]*);')
3177
return _unescape_re.sub(_unescaper, data)
3180
class _VersionedFileChecker(object):
3182
def __init__(self, repository):
3183
self.repository = repository
3184
self.text_index = self.repository._generate_text_key_index()
3186
def calculate_file_version_parents(self, text_key):
3187
"""Calculate the correct parents for a file version according to
3190
parent_keys = self.text_index[text_key]
3191
if parent_keys == [_mod_revision.NULL_REVISION]:
3193
return tuple(parent_keys)
3195
def check_file_version_parents(self, texts, progress_bar=None):
3196
"""Check the parents stored in a versioned file are correct.
3198
It also detects file versions that are not referenced by their
3199
corresponding revision's inventory.
3201
:returns: A tuple of (wrong_parents, dangling_file_versions).
3202
wrong_parents is a dict mapping {revision_id: (stored_parents,
3203
correct_parents)} for each revision_id where the stored parents
3204
are not correct. dangling_file_versions is a set of (file_id,
3205
revision_id) tuples for versions that are present in this versioned
3206
file, but not used by the corresponding inventory.
3209
self.file_ids = set([file_id for file_id, _ in
3210
self.text_index.iterkeys()])
3211
# text keys is now grouped by file_id
3212
n_weaves = len(self.file_ids)
3213
files_in_revisions = {}
3214
revisions_of_files = {}
3215
n_versions = len(self.text_index)
3216
progress_bar.update('loading text store', 0, n_versions)
3217
parent_map = self.repository.texts.get_parent_map(self.text_index)
3218
# On unlistable transports this could well be empty/error...
3219
text_keys = self.repository.texts.keys()
3220
unused_keys = frozenset(text_keys) - set(self.text_index)
3221
for num, key in enumerate(self.text_index.iterkeys()):
3222
if progress_bar is not None:
3223
progress_bar.update('checking text graph', num, n_versions)
3224
correct_parents = self.calculate_file_version_parents(key)
3226
knit_parents = parent_map[key]
3227
except errors.RevisionNotPresent:
3230
if correct_parents != knit_parents:
3231
wrong_parents[key] = (knit_parents, correct_parents)
3232
return wrong_parents, unused_keys
3235
def _old_get_graph(repository, revision_id):
3236
"""DO NOT USE. That is all. I'm serious."""
3237
graph = repository.get_graph()
3238
revision_graph = dict(((key, value) for key, value in
3239
graph.iter_ancestry([revision_id]) if value is not None))
3240
return _strip_NULL_ghosts(revision_graph)
1824
3243
def _strip_NULL_ghosts(revision_graph):