196
187
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,
189
def _heads(self, file_id, revision_ids):
190
"""Calculate the graph heads for revision_ids in the graph of file_id.
192
This can use either a per-file graph or a global revision graph as we
193
have an identity relationship between the two graphs.
195
return self.__heads(revision_ids)
197
def _check_root(self, ie, parent_invs, tree):
198
"""Helper for record_entry_contents.
200
:param ie: An entry being added.
201
:param parent_invs: The inventories of the parent revisions of the
203
:param tree: The tree that is being committed.
205
# In this revision format, root entries have no knit or weave When
206
# serializing out to disk and back in root.revision is always
208
ie.revision = self._new_revision_id
210
def _get_delta(self, ie, basis_inv, path):
211
"""Get a delta against the basis inventory for ie."""
212
if ie.file_id not in basis_inv:
214
return (None, path, ie.file_id, ie)
215
elif ie != basis_inv[ie.file_id]:
217
# TODO: avoid tis id2path call.
218
return (basis_inv.id2path(ie.file_id), path, ie.file_id, ie)
223
def record_entry_contents(self, ie, parent_invs, path, tree,
225
"""Record the content of ie from tree into the commit if needed.
227
Side effect: sets ie.revision when unchanged
229
:param ie: An inventory entry present in the commit.
230
:param parent_invs: The inventories of the parent revisions of the
232
:param path: The path the entry is at in the tree.
233
:param tree: The tree which contains this entry and should be used to
235
:param content_summary: Summary data from the tree about the paths
236
content - stat, length, exec, sha/link target. This is only
237
accessed when the entry has a revision of None - that is when it is
238
a candidate to commit.
239
:return: A tuple (change_delta, version_recorded, fs_hash).
240
change_delta is an inventory_delta change for this entry against
241
the basis tree of the commit, or None if no change occured against
243
version_recorded is True if a new version of the entry has been
244
recorded. For instance, committing a merge where a file was only
245
changed on the other side will return (delta, False).
246
fs_hash is either None, or the hash details for the path (currently
247
a tuple of the contents sha1 and the statvalue returned by
248
tree.get_file_with_stat()).
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)
289
return delta, False, None
291
# we don't need to commit this, because the caller already
292
# determined that an existing revision of this file is
293
# appropriate. If its not being considered for committing then
294
# it and all its parents to the root must be unaltered so
295
# no-change against the basis.
296
if ie.revision == self._new_revision_id:
297
raise AssertionError("Impossible situation, a skipped "
298
"inventory entry (%r) claims to be modified in this "
299
"commit (%r).", (ie, self._new_revision_id))
300
return None, False, None
301
# XXX: Friction: parent_candidates should return a list not a dict
302
# so that we don't have to walk the inventories again.
303
parent_candiate_entries = ie.parent_candidates(parent_invs)
304
head_set = self._heads(ie.file_id, parent_candiate_entries.keys())
306
for inv in parent_invs:
307
if ie.file_id in inv:
308
old_rev = inv[ie.file_id].revision
309
if old_rev in head_set:
310
heads.append(inv[ie.file_id].revision)
311
head_set.remove(inv[ie.file_id].revision)
314
# now we check to see if we need to write a new record to the
316
# We write a new entry unless there is one head to the ancestors, and
317
# the kind-derived content is unchanged.
319
# Cheapest check first: no ancestors, or more the one head in the
320
# ancestors, we write a new node.
324
# There is a single head, look it up for comparison
325
parent_entry = parent_candiate_entries[heads[0]]
326
# if the non-content specific data has changed, we'll be writing a
328
if (parent_entry.parent_id != ie.parent_id or
329
parent_entry.name != ie.name):
331
# now we need to do content specific checks:
333
# if the kind changed the content obviously has
334
if kind != parent_entry.kind:
336
# Stat cache fingerprint feedback for the caller - None as we usually
337
# don't generate one.
340
if content_summary[2] is None:
341
raise ValueError("Files must not have executable = None")
343
if (# if the file length changed we have to store:
344
parent_entry.text_size != content_summary[1] or
345
# if the exec bit has changed we have to store:
346
parent_entry.executable != content_summary[2]):
348
elif parent_entry.text_sha1 == content_summary[3]:
349
# all meta and content is unchanged (using a hash cache
350
# hit to check the sha)
351
ie.revision = parent_entry.revision
352
ie.text_size = parent_entry.text_size
353
ie.text_sha1 = parent_entry.text_sha1
354
ie.executable = parent_entry.executable
355
return self._get_delta(ie, basis_inv, path), False, None
357
# Either there is only a hash change(no hash cache entry,
358
# or same size content change), or there is no change on
360
# Provide the parent's hash to the store layer, so that the
361
# content is unchanged we will not store a new node.
362
nostore_sha = parent_entry.text_sha1
364
# We want to record a new node regardless of the presence or
365
# absence of a content change in the file.
367
ie.executable = content_summary[2]
368
file_obj, stat_value = tree.get_file_with_stat(ie.file_id, path)
370
lines = file_obj.readlines()
374
ie.text_sha1, ie.text_size = self._add_text_to_weave(
375
ie.file_id, lines, heads, nostore_sha)
376
# Let the caller know we generated a stat fingerprint.
377
fingerprint = (ie.text_sha1, stat_value)
378
except errors.ExistingContent:
379
# Turns out that the file content was unchanged, and we were
380
# only going to store a new node if it was changed. Carry over
382
ie.revision = parent_entry.revision
383
ie.text_size = parent_entry.text_size
384
ie.text_sha1 = parent_entry.text_sha1
385
ie.executable = parent_entry.executable
386
return self._get_delta(ie, basis_inv, path), False, None
387
elif kind == 'directory':
389
# all data is meta here, nothing specific to directory, so
391
ie.revision = parent_entry.revision
392
return self._get_delta(ie, basis_inv, path), False, None
394
self._add_text_to_weave(ie.file_id, lines, heads, None)
395
elif kind == 'symlink':
396
current_link_target = content_summary[3]
398
# symlink target is not generic metadata, check if it has
400
if current_link_target != parent_entry.symlink_target:
403
# unchanged, carry over.
404
ie.revision = parent_entry.revision
405
ie.symlink_target = parent_entry.symlink_target
406
return self._get_delta(ie, basis_inv, path), False, None
407
ie.symlink_target = current_link_target
409
self._add_text_to_weave(ie.file_id, lines, heads, None)
410
elif kind == 'tree-reference':
412
if content_summary[3] != parent_entry.reference_revision:
415
# unchanged, carry over.
416
ie.reference_revision = parent_entry.reference_revision
417
ie.revision = parent_entry.revision
418
return self._get_delta(ie, basis_inv, path), False, None
419
ie.reference_revision = content_summary[3]
421
self._add_text_to_weave(ie.file_id, lines, heads, None)
423
raise NotImplementedError('unknown kind')
424
ie.revision = self._new_revision_id
425
return self._get_delta(ie, basis_inv, path), True, fingerprint
427
def _add_text_to_weave(self, file_id, new_lines, parents, nostore_sha):
428
# Note: as we read the content directly from the tree, we know its not
429
# been turned into unicode or badly split - but a broken tree
430
# implementation could give us bad output from readlines() so this is
431
# not a guarantee of safety. What would be better is always checking
432
# the content during test suite execution. RBC 20070912
433
parent_keys = tuple((file_id, parent) for parent in parents)
434
return self.repository.texts.add_lines(
435
(file_id, self._new_revision_id), parent_keys, new_lines,
436
nostore_sha=nostore_sha, random_id=self.random_revid,
437
check_content=False)[0:2]
440
class RootCommitBuilder(CommitBuilder):
441
"""This commitbuilder actually records the root id"""
443
# the root entry gets versioned properly by this builder.
444
_versioned_root = True
446
def _check_root(self, ie, parent_invs, tree):
447
"""Helper for record_entry_contents.
449
:param ie: An entry being added.
450
:param parent_invs: The inventories of the parent revisions of the
452
:param tree: The tree that is being committed.
241
456
######################################################################
245
class Repository(_RelockDebugMixin, controldir.ControlComponent):
459
class Repository(object):
246
460
"""Repository holding history for one or more branches.
248
462
The repository holds and retrieves historical information including
249
463
revisions and file history. It's normally accessed only by the Branch,
250
464
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.
466
The Repository builds on top of some byte storage facilies (the revisions,
467
signatures, inventories and texts attributes) and a Transport, which
468
respectively provide byte storage and a means to access the (possibly
471
The byte storage facilities are addressed via tuples, which we refer to
472
as 'keys' throughout the code base. Revision_keys, inventory_keys and
473
signature_keys are all 1-tuples: (revision_id,). text_keys are two-tuples:
474
(file_id, revision_id). We use this interface because it allows low
475
friction with the underlying code that implements disk indices, network
476
encoding and other parts of bzrlib.
478
:ivar revisions: A bzrlib.versionedfile.VersionedFiles instance containing
479
the serialised revisions for the repository. This can be used to obtain
480
revision graph information or to access raw serialised revisions.
481
The result of trying to insert data into the repository via this store
482
is undefined: it should be considered read-only except for implementors
484
:ivar signatures: A bzrlib.versionedfile.VersionedFiles instance containing
485
the serialised signatures for the repository. This can be used to
486
obtain access to raw serialised signatures. The result of trying to
487
insert data into the repository via this store is undefined: it should
488
be considered read-only except for implementors of repositories.
489
:ivar inventories: A bzrlib.versionedfile.VersionedFiles instance containing
490
the serialised inventories for the repository. This can be used to
491
obtain unserialised inventories. The result of trying to insert data
492
into the repository via this store is undefined: it should be
493
considered read-only except for implementors of repositories.
494
:ivar texts: A bzrlib.versionedfile.VersionedFiles instance containing the
495
texts of files and directories for the repository. This can be used to
496
obtain file texts or file graphs. Note that Repository.iter_file_bytes
497
is usually a better interface for accessing file texts.
498
The result of trying to insert data into the repository via this store
499
is undefined: it should be considered read-only except for implementors
501
:ivar _transport: Transport for file access to repository, typically
502
pointing to .bzr/repository.
256
def abort_write_group(self, suppress_errors=False):
505
# What class to use for a CommitBuilder. Often its simpler to change this
506
# in a Repository class subclass rather than to override
507
# get_commit_builder.
508
_commit_builder_class = CommitBuilder
509
# The search regex used by xml based repositories to determine what things
510
# where changed in a single commit.
511
_file_ids_altered_regex = lazy_regex.lazy_compile(
512
r'file_id="(?P<file_id>[^"]+)"'
513
r'.* revision="(?P<revision_id>[^"]+)"'
516
def abort_write_group(self):
257
517
"""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
519
:seealso: start_write_group.
265
521
if self._write_group is not self.get_transaction():
266
522
# 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)
523
raise errors.BzrError('mismatched lock context and write group.')
524
self._abort_write_group()
284
525
self._write_group = None
286
527
def _abort_write_group(self):
287
528
"""Template method for per-repository write group cleanup.
289
This is called during abort before the write group is considered to be
530
This is called during abort before the write group is considered to be
290
531
finished and should cleanup any internal state accrued during the write
291
532
group. There is no requirement that data handed to the repository be
292
533
*not* made available - this is not a rollback - but neither should any
618
947
"""Commit the contents accrued within the current write group.
620
949
:seealso: start_write_group.
622
:return: it may return an opaque hint that can be passed to 'pack'.
624
951
if self._write_group is not self.get_transaction():
625
952
# has an unlock or relock occured ?
626
953
raise errors.BzrError('mismatched lock context %r and '
627
954
'write group %r.' %
628
955
(self.get_transaction(), self._write_group))
629
result = self._commit_write_group()
956
self._commit_write_group()
630
957
self._write_group = None
633
959
def _commit_write_group(self):
634
960
"""Template method for per-repository write group cleanup.
636
This is called before the write group is considered to be
962
This is called before the write group is considered to be
637
963
finished and should ensure that all data handed to the repository
638
for writing during the write group is safely committed (to the
964
for writing during the write group is safely committed (to the
639
965
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,
968
def fetch(self, source, revision_id=None, pb=None, find_ghosts=False):
672
969
"""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.
971
If revision_id is None all content is copied.
681
972
:param find_ghosts: Find and copy revisions in the source that are
682
973
ghosts in the target (and not reachable directly by walking out to
683
974
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
976
# 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)):
977
if self.has_same_location(source):
704
978
# check that last_revision is in 'from' and then return a
706
980
if (revision_id is not None and
707
981
not _mod_revision.is_null(revision_id)):
708
982
self.get_revision(revision_id)
984
# if there is no specific appropriate InterRepository, this will get
985
# the InterRepository base class, which raises an
986
# IncompatibleRepositories when asked to fetch.
710
987
inter = InterRepository.get(source, self)
711
return inter.fetch(revision_id=revision_id,
712
find_ghosts=find_ghosts, fetch_spec=fetch_spec)
988
return inter.fetch(revision_id=revision_id, pb=pb,
989
find_ghosts=find_ghosts)
714
991
def create_bundle(self, target, base, fileobj, format=None):
715
992
return serializer.write_bundle(self, target, base, fileobj, format)
717
994
def get_commit_builder(self, branch, parents, config, timestamp=None,
718
995
timezone=None, committer=None, revprops=None,
719
revision_id=None, lossy=False):
720
997
"""Obtain a CommitBuilder for this repository.
722
999
:param branch: Branch to commit to.
723
1000
:param parents: Revision ids of the parents of the new revision.
724
1001
:param config: Configuration to use.
890
1180
yield trees[revision.revision_id].changes_from(old_tree)
892
1182
@needs_read_lock
893
def get_revision_delta(self, revision_id, specific_fileids=None):
1183
def get_revision_delta(self, revision_id):
894
1184
"""Return the delta for one revision.
896
1186
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
1189
r = self.get_revision(revision_id)
904
return list(self.get_deltas_for_revisions([r],
905
specific_fileids=specific_fileids))[0]
1190
return list(self.get_deltas_for_revisions([r]))[0]
907
1192
@needs_write_lock
908
1193
def store_revision_signature(self, gpg_strategy, plaintext, revision_id):
909
1194
signature = gpg_strategy.sign(plaintext)
910
1195
self.add_signature_text(revision_id, signature)
912
1198
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):
1199
self.signatures.add_lines((revision_id,), (),
1200
osutils.split_lines(signature))
1202
def find_text_key_references(self):
1203
"""Find the text key references within the repository.
1205
:return: a dictionary mapping (file_id, revision_id) tuples to altered file-ids to an iterable of
1206
revision_ids. Each altered file-ids has the exact revision_ids that
1207
altered it listed explicitly.
1208
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1209
to whether they were referred to by the inventory of the
1210
revision_id that they contain. The inventory texts from all present
1211
revision ids are assessed to generate this report.
1213
revision_keys = self.revisions.keys()
1214
w = self.inventories
1215
pb = ui.ui_factory.nested_progress_bar()
1217
return self._find_text_key_references_from_xml_inventory_lines(
1218
w.iter_lines_added_or_present_in_keys(revision_keys, pb=pb))
1222
def _find_text_key_references_from_xml_inventory_lines(self,
1224
"""Core routine for extracting references to texts from inventories.
1226
This performs the translation of xml lines to revision ids.
1228
:param line_iterator: An iterator of lines, origin_version_id
1229
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1230
to whether they were referred to by the inventory of the
1231
revision_id that they contain. Note that if that revision_id was
1232
not part of the line_iterator's output then False will be given -
1233
even though it may actually refer to that key.
1235
if not self._serializer.support_altered_by_hack:
1236
raise AssertionError(
1237
"_find_text_key_references_from_xml_inventory_lines only "
1238
"supported for branches which store inventory as unnested xml"
1239
", not on %r" % self)
1242
# this code needs to read every new line in every inventory for the
1243
# inventories [revision_ids]. Seeing a line twice is ok. Seeing a line
1244
# not present in one of those inventories is unnecessary but not
1245
# harmful because we are filtering by the revision id marker in the
1246
# inventory lines : we only select file ids altered in one of those
1247
# revisions. We don't need to see all lines in the inventory because
1248
# only those added in an inventory in rev X can contain a revision=X
1250
unescape_revid_cache = {}
1251
unescape_fileid_cache = {}
1253
# jam 20061218 In a big fetch, this handles hundreds of thousands
1254
# of lines, so it has had a lot of inlining and optimizing done.
1255
# Sorry that it is a little bit messy.
1256
# Move several functions to be local variables, since this is a long
1258
search = self._file_ids_altered_regex.search
1259
unescape = _unescape_xml
1260
setdefault = result.setdefault
1261
for line, line_key in line_iterator:
1262
match = search(line)
1265
# One call to match.group() returning multiple items is quite a
1266
# bit faster than 2 calls to match.group() each returning 1
1267
file_id, revision_id = match.group('file_id', 'revision_id')
1269
# Inlining the cache lookups helps a lot when you make 170,000
1270
# lines and 350k ids, versus 8.4 unique ids.
1271
# Using a cache helps in 2 ways:
1272
# 1) Avoids unnecessary decoding calls
1273
# 2) Re-uses cached strings, which helps in future set and
1275
# (2) is enough that removing encoding entirely along with
1276
# the cache (so we are using plain strings) results in no
1277
# performance improvement.
1279
revision_id = unescape_revid_cache[revision_id]
1281
unescaped = unescape(revision_id)
1282
unescape_revid_cache[revision_id] = unescaped
1283
revision_id = unescaped
1285
# Note that unconditionally unescaping means that we deserialise
1286
# every fileid, which for general 'pull' is not great, but we don't
1287
# really want to have some many fulltexts that this matters anyway.
1290
file_id = unescape_fileid_cache[file_id]
1292
unescaped = unescape(file_id)
1293
unescape_fileid_cache[file_id] = unescaped
1296
key = (file_id, revision_id)
1297
setdefault(key, False)
1298
if revision_id == line_key[-1]:
1302
def _find_file_ids_from_xml_inventory_lines(self, line_iterator,
1304
"""Helper routine for fileids_altered_by_revision_ids.
1306
This performs the translation of xml lines to revision ids.
1308
:param line_iterator: An iterator of lines, origin_version_id
1309
:param revision_ids: The revision ids to filter for. This should be a
1310
set or other type which supports efficient __contains__ lookups, as
1311
the revision id from each parsed line will be looked up in the
1312
revision_ids filter.
1313
:return: a dictionary mapping altered file-ids to an iterable of
1314
revision_ids. Each altered file-ids has the exact revision_ids that
1315
altered it listed explicitly.
1318
setdefault = result.setdefault
1320
self._find_text_key_references_from_xml_inventory_lines(
1321
line_iterator).iterkeys():
1322
# once data is all ensured-consistent; then this is
1323
# if revision_id == version_id
1324
if key[-1:] in revision_ids:
1325
setdefault(key[0], set()).add(key[-1])
1328
def fileids_altered_by_revision_ids(self, revision_ids, _inv_weave=None):
934
1329
"""Find the file ids and versions affected by revisions.
936
1331
:param revisions: an iterable containing revision ids.
1332
:param _inv_weave: The inventory weave from this repository or None.
1333
If None, the inventory weave will be opened automatically.
937
1334
: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.
1335
revision_ids. Each altered file-ids has the exact revision_ids that
1336
altered it listed explicitly.
941
raise NotImplementedError(self.fileids_altered_by_revision_ids)
1338
selected_keys = set((revid,) for revid in revision_ids)
1339
w = _inv_weave or self.inventories
1340
pb = ui.ui_factory.nested_progress_bar()
1342
return self._find_file_ids_from_xml_inventory_lines(
1343
w.iter_lines_added_or_present_in_keys(
1344
selected_keys, pb=pb),
943
1349
def iter_files_bytes(self, desired_files):
944
1350
"""Iterate through file versions.
951
1357
uniquely identify the file version in the caller's context. (Examples:
952
1358
an index number or a TreeTransform trans_id.)
1360
bytes_iterator is an iterable of bytestrings for the file. The
1361
kind of iterable and length of the bytestrings are unspecified, but for
1362
this implementation, it is a list of bytes produced by
1363
VersionedFile.get_record_stream().
954
1365
: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)))
1368
transaction = self.get_transaction()
1370
for file_id, revision_id, callable_data in desired_files:
1371
text_keys[(file_id, revision_id)] = callable_data
1372
for record in self.texts.get_record_stream(text_keys, 'unordered', True):
1373
if record.storage_kind == 'absent':
1374
raise errors.RevisionNotPresent(record.key, self)
1375
yield text_keys[record.key], record.get_bytes_as('fulltext')
1377
def _generate_text_key_index(self, text_key_references=None,
1379
"""Generate a new text key index for the repository.
1381
This is an expensive function that will take considerable time to run.
1383
:return: A dict mapping text keys ((file_id, revision_id) tuples) to a
1384
list of parents, also text keys. When a given key has no parents,
1385
the parents list will be [NULL_REVISION].
1387
# All revisions, to find inventory parents.
1388
if ancestors is None:
1389
graph = self.get_graph()
1390
ancestors = graph.get_parent_map(self.all_revision_ids())
1391
if text_key_references is None:
1392
text_key_references = self.find_text_key_references()
1393
pb = ui.ui_factory.nested_progress_bar()
1395
return self._do_generate_text_key_index(ancestors,
1396
text_key_references, pb)
1400
def _do_generate_text_key_index(self, ancestors, text_key_references, pb):
1401
"""Helper for _generate_text_key_index to avoid deep nesting."""
1402
revision_order = tsort.topo_sort(ancestors)
1403
invalid_keys = set()
1405
for revision_id in revision_order:
1406
revision_keys[revision_id] = set()
1407
text_count = len(text_key_references)
1408
# a cache of the text keys to allow reuse; costs a dict of all the
1409
# keys, but saves a 2-tuple for every child of a given key.
1411
for text_key, valid in text_key_references.iteritems():
1413
invalid_keys.add(text_key)
1415
revision_keys[text_key[1]].add(text_key)
1416
text_key_cache[text_key] = text_key
1417
del text_key_references
1419
text_graph = graph.Graph(graph.DictParentsProvider(text_index))
1420
NULL_REVISION = _mod_revision.NULL_REVISION
1421
# Set a cache with a size of 10 - this suffices for bzr.dev but may be
1422
# too small for large or very branchy trees. However, for 55K path
1423
# trees, it would be easy to use too much memory trivially. Ideally we
1424
# could gauge this by looking at available real memory etc, but this is
1425
# always a tricky proposition.
1426
inventory_cache = lru_cache.LRUCache(10)
1427
batch_size = 10 # should be ~150MB on a 55K path tree
1428
batch_count = len(revision_order) / batch_size + 1
1430
pb.update("Calculating text parents.", processed_texts, text_count)
1431
for offset in xrange(batch_count):
1432
to_query = revision_order[offset * batch_size:(offset + 1) *
1436
for rev_tree in self.revision_trees(to_query):
1437
revision_id = rev_tree.get_revision_id()
1438
parent_ids = ancestors[revision_id]
1439
for text_key in revision_keys[revision_id]:
1440
pb.update("Calculating text parents.", processed_texts)
1441
processed_texts += 1
1442
candidate_parents = []
1443
for parent_id in parent_ids:
1444
parent_text_key = (text_key[0], parent_id)
1446
check_parent = parent_text_key not in \
1447
revision_keys[parent_id]
1449
# the parent parent_id is a ghost:
1450
check_parent = False
1451
# truncate the derived graph against this ghost.
1452
parent_text_key = None
1454
# look at the parent commit details inventories to
1455
# determine possible candidates in the per file graph.
1458
inv = inventory_cache[parent_id]
1460
inv = self.revision_tree(parent_id).inventory
1461
inventory_cache[parent_id] = inv
1462
parent_entry = inv._byid.get(text_key[0], None)
1463
if parent_entry is not None:
1465
text_key[0], parent_entry.revision)
1467
parent_text_key = None
1468
if parent_text_key is not None:
1469
candidate_parents.append(
1470
text_key_cache[parent_text_key])
1471
parent_heads = text_graph.heads(candidate_parents)
1472
new_parents = list(parent_heads)
1473
new_parents.sort(key=lambda x:candidate_parents.index(x))
1474
if new_parents == []:
1475
new_parents = [NULL_REVISION]
1476
text_index[text_key] = new_parents
1478
for text_key in invalid_keys:
1479
text_index[text_key] = [NULL_REVISION]
1482
def item_keys_introduced_by(self, revision_ids, _files_pb=None):
1483
"""Get an iterable listing the keys of all the data introduced by a set
1486
The keys will be ordered so that the corresponding items can be safely
1487
fetched and inserted in that order.
1489
:returns: An iterable producing tuples of (knit-kind, file-id,
1490
versions). knit-kind is one of 'file', 'inventory', 'signatures',
1491
'revisions'. file-id is None unless knit-kind is 'file'.
1493
# XXX: it's a bit weird to control the inventory weave caching in this
1494
# generator. Ideally the caching would be done in fetch.py I think. Or
1495
# maybe this generator should explicitly have the contract that it
1496
# should not be iterated until the previously yielded item has been
1498
inv_w = self.inventories
1500
# file ids that changed
1501
file_ids = self.fileids_altered_by_revision_ids(revision_ids, inv_w)
1503
num_file_ids = len(file_ids)
1504
for file_id, altered_versions in file_ids.iteritems():
1505
if _files_pb is not None:
1506
_files_pb.update("fetch texts", count, num_file_ids)
1508
yield ("file", file_id, altered_versions)
1509
# We're done with the files_pb. Note that it finished by the caller,
1510
# just as it was created by the caller.
1514
yield ("inventory", None, revision_ids)
1517
revisions_with_signatures = set()
1518
for rev_id in revision_ids:
1520
self.get_signature_text(rev_id)
1521
except errors.NoSuchRevision:
1525
revisions_with_signatures.add(rev_id)
1526
yield ("signatures", None, revisions_with_signatures)
1529
yield ("revisions", None, revision_ids)
1532
def get_inventory(self, revision_id):
1533
"""Get Inventory object by revision id."""
1534
return self.iter_inventories([revision_id]).next()
1536
def iter_inventories(self, revision_ids):
1537
"""Get many inventories by revision_ids.
1539
This will buffer some or all of the texts used in constructing the
1540
inventories in memory, but will only parse a single inventory at a
1543
:return: An iterator of inventories.
1545
if ((None in revision_ids)
1546
or (_mod_revision.NULL_REVISION in revision_ids)):
1547
raise ValueError('cannot get null revision inventory')
1548
return self._iter_inventories(revision_ids)
1550
def _iter_inventories(self, revision_ids):
1551
"""single-document based inventory iteration."""
1552
for text, revision_id in self._iter_inventory_xmls(revision_ids):
1553
yield self.deserialise_inventory(revision_id, text)
1555
def _iter_inventory_xmls(self, revision_ids):
1556
keys = [(revision_id,) for revision_id in revision_ids]
1557
stream = self.inventories.get_record_stream(keys, 'unordered', True)
1559
for record in stream:
1560
if record.storage_kind != 'absent':
1561
texts[record.key] = record.get_bytes_as('fulltext')
1563
raise errors.NoSuchRevision(self, record.key)
1565
yield texts[key], key[-1]
1567
def deserialise_inventory(self, revision_id, xml):
1568
"""Transform the xml into an inventory object.
1570
:param revision_id: The expected revision id of the inventory.
1571
:param xml: A serialised inventory.
1573
result = self._serializer.read_inventory_from_string(xml, revision_id)
1574
if result.revision_id != revision_id:
1575
raise AssertionError('revision id mismatch %s != %s' % (
1576
result.revision_id, revision_id))
1579
def serialise_inventory(self, inv):
1580
return self._serializer.write_inventory_to_string(inv)
1582
def _serialise_inventory_to_lines(self, inv):
1583
return self._serializer.write_inventory_to_lines(inv)
1585
def get_serializer_format(self):
1586
return self._serializer.format_num
1589
def get_inventory_xml(self, revision_id):
1590
"""Get inventory XML as a file object."""
1591
texts = self._iter_inventory_xmls([revision_id])
1593
text, revision_id = texts.next()
1594
except StopIteration:
1595
raise errors.HistoryMissing(self, 'inventory', revision_id)
1599
def get_inventory_sha1(self, revision_id):
1600
"""Return the sha1 hash of the inventory entry
1602
return self.get_revision(revision_id).inventory_sha1
995
1604
def iter_reverse_revision_history(self, revision_id):
996
1605
"""Iterate backwards through revision ids in the lefthand history
1695
2356
_optimisers = []
1696
2357
"""The available optimised InterRepository types."""
1699
2359
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,
2360
raise NotImplementedError(self.copy_content)
2362
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
1717
2363
"""Fetch the content required to construct revision_id.
1719
2365
The content is copied from self.source to self.target.
1721
2367
:param revision_id: if None all content is copied, if NULL_REVISION no
1722
2368
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
2369
:param pb: optional progress bar to use for progress reports. If not
2370
provided a default one will be created.
2372
:returns: (copied_revision_count, failures).
2374
# Normally we should find a specific InterRepository subclass to do
2375
# the fetch; if nothing else then at least InterSameDataRepository.
2376
# If none of them is suitable it looks like fetching is not possible;
2377
# we try to give a good message why. _assert_same_model will probably
2378
# give a helpful message; otherwise a generic one.
2379
self._assert_same_model(self.source, self.target)
2380
raise errors.IncompatibleRepositories(self.source, self.target,
2381
"no suitableInterRepository found")
2383
def _walk_to_common_revisions(self, revision_ids):
2384
"""Walk out from revision_ids in source to revisions target has.
2386
:param revision_ids: The start point for the search.
2387
:return: A set of revision ids.
2389
target_graph = self.target.get_graph()
2390
revision_ids = frozenset(revision_ids)
2391
if set(target_graph.get_parent_map(revision_ids)) == revision_ids:
2392
return graph.SearchResult(revision_ids, set(), 0, set())
2393
missing_revs = set()
2394
source_graph = self.source.get_graph()
2395
# ensure we don't pay silly lookup costs.
2396
searcher = source_graph._make_breadth_first_searcher(revision_ids)
2397
null_set = frozenset([_mod_revision.NULL_REVISION])
2400
next_revs, ghosts = searcher.next_with_ghosts()
2401
except StopIteration:
2403
if revision_ids.intersection(ghosts):
2404
absent_ids = set(revision_ids.intersection(ghosts))
2405
# If all absent_ids are present in target, no error is needed.
2406
absent_ids.difference_update(
2407
set(target_graph.get_parent_map(absent_ids)))
2409
raise errors.NoSuchRevision(self.source, absent_ids.pop())
2410
# we don't care about other ghosts as we can't fetch them and
2411
# haven't been asked to.
2412
next_revs = set(next_revs)
2413
# we always have NULL_REVISION present.
2414
have_revs = set(target_graph.get_parent_map(next_revs)).union(null_set)
2415
missing_revs.update(next_revs - have_revs)
2416
searcher.stop_searching_any(have_revs)
2417
return searcher.get_result()
2419
@deprecated_method(one_two)
2421
def missing_revision_ids(self, revision_id=None, find_ghosts=True):
2422
"""Return the revision ids that source has that target does not.
2424
These are returned in topological order.
2426
:param revision_id: only return revision ids included by this
2428
:param find_ghosts: If True find missing revisions in deep history
2429
rather than just finding the surface difference.
2431
return list(self.search_missing_revision_ids(
2432
revision_id, find_ghosts).get_keys())
2435
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2436
"""Return the revision ids that source has that target does not.
2438
:param revision_id: only return revision ids included by this
2440
:param find_ghosts: If True find missing revisions in deep history
2441
rather than just finding the surface difference.
1747
2442
:return: A bzrlib.graph.SearchResult.
1749
raise NotImplementedError(self.search_missing_revision_ids)
2444
# stop searching at found target revisions.
2445
if not find_ghosts and revision_id is not None:
2446
return self._walk_to_common_revisions([revision_id])
2447
# generic, possibly worst case, slow code path.
2448
target_ids = set(self.target.all_revision_ids())
2449
if revision_id is not None:
2450
source_ids = self.source.get_ancestry(revision_id)
2451
if source_ids[0] is not None:
2452
raise AssertionError()
2455
source_ids = self.source.all_revision_ids()
2456
result_set = set(source_ids).difference(target_ids)
2457
return self.source.revision_ids_to_search_result(result_set)
1752
2460
def _same_model(source, target):
1753
2461
"""True if source and target have the same data representation.
1755
2463
Note: this is always called on the base class; overriding it in a
1756
2464
subclass will have no effect.
1773
2481
"different serializers")
2484
class InterSameDataRepository(InterRepository):
2485
"""Code for converting between repositories that represent the same data.
2487
Data format and model must match for this to work.
2491
def _get_repo_format_to_test(self):
2492
"""Repository format for testing with.
2494
InterSameData can pull from subtree to subtree and from non-subtree to
2495
non-subtree, so we test this with the richest repository format.
2497
from bzrlib.repofmt import knitrepo
2498
return knitrepo.RepositoryFormatKnit3()
2501
def is_compatible(source, target):
2502
return InterRepository._same_model(source, target)
2505
def copy_content(self, revision_id=None):
2506
"""Make a complete copy of the content in self into destination.
2508
This copies both the repository's revision data, and configuration information
2509
such as the make_working_trees setting.
2511
This is a destructive operation! Do not use it on existing
2514
:param revision_id: Only copy the content needed to construct
2515
revision_id and its parents.
2518
self.target.set_make_working_trees(self.source.make_working_trees())
2519
except NotImplementedError:
2521
# but don't bother fetching if we have the needed data now.
2522
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
2523
self.target.has_revision(revision_id)):
2525
self.target.fetch(self.source, revision_id=revision_id)
2528
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2529
"""See InterRepository.fetch()."""
2530
from bzrlib.fetch import RepoFetcher
2531
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2532
self.source, self.source._format, self.target,
2533
self.target._format)
2534
f = RepoFetcher(to_repository=self.target,
2535
from_repository=self.source,
2536
last_revision=revision_id,
2537
pb=pb, find_ghosts=find_ghosts)
2538
return f.count_copied, f.failed_revisions
2541
class InterWeaveRepo(InterSameDataRepository):
2542
"""Optimised code paths between Weave based repositories.
2544
This should be in bzrlib/repofmt/weaverepo.py but we have not yet
2545
implemented lazy inter-object optimisation.
2549
def _get_repo_format_to_test(self):
2550
from bzrlib.repofmt import weaverepo
2551
return weaverepo.RepositoryFormat7()
2554
def is_compatible(source, target):
2555
"""Be compatible with known Weave formats.
2557
We don't test for the stores being of specific types because that
2558
could lead to confusing results, and there is no need to be
2561
from bzrlib.repofmt.weaverepo import (
2567
return (isinstance(source._format, (RepositoryFormat5,
2569
RepositoryFormat7)) and
2570
isinstance(target._format, (RepositoryFormat5,
2572
RepositoryFormat7)))
2573
except AttributeError:
2577
def copy_content(self, revision_id=None):
2578
"""See InterRepository.copy_content()."""
2579
# weave specific optimised path:
2581
self.target.set_make_working_trees(self.source.make_working_trees())
2582
except (errors.RepositoryUpgradeRequired, NotImplemented):
2584
# FIXME do not peek!
2585
if self.source._transport.listable():
2586
pb = ui.ui_factory.nested_progress_bar()
2588
self.target.texts.insert_record_stream(
2589
self.source.texts.get_record_stream(
2590
self.source.texts.keys(), 'topological', False))
2591
pb.update('copying inventory', 0, 1)
2592
self.target.inventories.insert_record_stream(
2593
self.source.inventories.get_record_stream(
2594
self.source.inventories.keys(), 'topological', False))
2595
self.target.signatures.insert_record_stream(
2596
self.source.signatures.get_record_stream(
2597
self.source.signatures.keys(),
2599
self.target.revisions.insert_record_stream(
2600
self.source.revisions.get_record_stream(
2601
self.source.revisions.keys(),
2602
'topological', True))
2606
self.target.fetch(self.source, revision_id=revision_id)
2609
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2610
"""See InterRepository.fetch()."""
2611
from bzrlib.fetch import RepoFetcher
2612
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2613
self.source, self.source._format, self.target, self.target._format)
2614
f = RepoFetcher(to_repository=self.target,
2615
from_repository=self.source,
2616
last_revision=revision_id,
2617
pb=pb, find_ghosts=find_ghosts)
2618
return f.count_copied, f.failed_revisions
2621
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2622
"""See InterRepository.missing_revision_ids()."""
2623
# we want all revisions to satisfy revision_id in source.
2624
# but we don't want to stat every file here and there.
2625
# we want then, all revisions other needs to satisfy revision_id
2626
# checked, but not those that we have locally.
2627
# so the first thing is to get a subset of the revisions to
2628
# satisfy revision_id in source, and then eliminate those that
2629
# we do already have.
2630
# this is slow on high latency connection to self, but as as this
2631
# disk format scales terribly for push anyway due to rewriting
2632
# inventory.weave, this is considered acceptable.
2634
if revision_id is not None:
2635
source_ids = self.source.get_ancestry(revision_id)
2636
if source_ids[0] is not None:
2637
raise AssertionError()
2640
source_ids = self.source._all_possible_ids()
2641
source_ids_set = set(source_ids)
2642
# source_ids is the worst possible case we may need to pull.
2643
# now we want to filter source_ids against what we actually
2644
# have in target, but don't try to check for existence where we know
2645
# we do not have a revision as that would be pointless.
2646
target_ids = set(self.target._all_possible_ids())
2647
possibly_present_revisions = target_ids.intersection(source_ids_set)
2648
actually_present_revisions = set(
2649
self.target._eliminate_revisions_not_present(possibly_present_revisions))
2650
required_revisions = source_ids_set.difference(actually_present_revisions)
2651
if revision_id is not None:
2652
# we used get_ancestry to determine source_ids then we are assured all
2653
# revisions referenced are present as they are installed in topological order.
2654
# and the tip revision was validated by get_ancestry.
2655
result_set = required_revisions
2657
# if we just grabbed the possibly available ids, then
2658
# we only have an estimate of whats available and need to validate
2659
# that against the revision records.
2661
self.source._eliminate_revisions_not_present(required_revisions))
2662
return self.source.revision_ids_to_search_result(result_set)
2665
class InterKnitRepo(InterSameDataRepository):
2666
"""Optimised code paths between Knit based repositories."""
2669
def _get_repo_format_to_test(self):
2670
from bzrlib.repofmt import knitrepo
2671
return knitrepo.RepositoryFormatKnit1()
2674
def is_compatible(source, target):
2675
"""Be compatible with known Knit formats.
2677
We don't test for the stores being of specific types because that
2678
could lead to confusing results, and there is no need to be
2681
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit
2683
are_knits = (isinstance(source._format, RepositoryFormatKnit) and
2684
isinstance(target._format, RepositoryFormatKnit))
2685
except AttributeError:
2687
return are_knits and InterRepository._same_model(source, target)
2690
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2691
"""See InterRepository.fetch()."""
2692
from bzrlib.fetch import RepoFetcher
2693
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2694
self.source, self.source._format, self.target, self.target._format)
2695
f = RepoFetcher(to_repository=self.target,
2696
from_repository=self.source,
2697
last_revision=revision_id,
2698
pb=pb, find_ghosts=find_ghosts)
2699
return f.count_copied, f.failed_revisions
2702
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2703
"""See InterRepository.missing_revision_ids()."""
2704
if revision_id is not None:
2705
source_ids = self.source.get_ancestry(revision_id)
2706
if source_ids[0] is not None:
2707
raise AssertionError()
2710
source_ids = self.source.all_revision_ids()
2711
source_ids_set = set(source_ids)
2712
# source_ids is the worst possible case we may need to pull.
2713
# now we want to filter source_ids against what we actually
2714
# have in target, but don't try to check for existence where we know
2715
# we do not have a revision as that would be pointless.
2716
target_ids = set(self.target.all_revision_ids())
2717
possibly_present_revisions = target_ids.intersection(source_ids_set)
2718
actually_present_revisions = set(
2719
self.target._eliminate_revisions_not_present(possibly_present_revisions))
2720
required_revisions = source_ids_set.difference(actually_present_revisions)
2721
if revision_id is not None:
2722
# we used get_ancestry to determine source_ids then we are assured all
2723
# revisions referenced are present as they are installed in topological order.
2724
# and the tip revision was validated by get_ancestry.
2725
result_set = required_revisions
2727
# if we just grabbed the possibly available ids, then
2728
# we only have an estimate of whats available and need to validate
2729
# that against the revision records.
2731
self.source._eliminate_revisions_not_present(required_revisions))
2732
return self.source.revision_ids_to_search_result(result_set)
2735
class InterPackRepo(InterSameDataRepository):
2736
"""Optimised code paths between Pack based repositories."""
2739
def _get_repo_format_to_test(self):
2740
from bzrlib.repofmt import pack_repo
2741
return pack_repo.RepositoryFormatKnitPack1()
2744
def is_compatible(source, target):
2745
"""Be compatible with known Pack formats.
2747
We don't test for the stores being of specific types because that
2748
could lead to confusing results, and there is no need to be
2751
from bzrlib.repofmt.pack_repo import RepositoryFormatPack
2753
are_packs = (isinstance(source._format, RepositoryFormatPack) and
2754
isinstance(target._format, RepositoryFormatPack))
2755
except AttributeError:
2757
return are_packs and InterRepository._same_model(source, target)
2760
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2761
"""See InterRepository.fetch()."""
2762
if (len(self.source._fallback_repositories) > 0 or
2763
len(self.target._fallback_repositories) > 0):
2764
# The pack layer is not aware of fallback repositories, so when
2765
# fetching from a stacked repository or into a stacked repository
2766
# we use the generic fetch logic which uses the VersionedFiles
2767
# attributes on repository.
2768
from bzrlib.fetch import RepoFetcher
2769
fetcher = RepoFetcher(self.target, self.source, revision_id,
2771
return fetcher.count_copied, fetcher.failed_revisions
2772
from bzrlib.repofmt.pack_repo import Packer
2773
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2774
self.source, self.source._format, self.target, self.target._format)
2775
self.count_copied = 0
2776
if revision_id is None:
2778
# everything to do - use pack logic
2779
# to fetch from all packs to one without
2780
# inventory parsing etc, IFF nothing to be copied is in the target.
2782
source_revision_ids = frozenset(self.source.all_revision_ids())
2783
revision_ids = source_revision_ids - \
2784
frozenset(self.target.get_parent_map(source_revision_ids))
2785
revision_keys = [(revid,) for revid in revision_ids]
2786
index = self.target._pack_collection.revision_index.combined_index
2787
present_revision_ids = set(item[1][0] for item in
2788
index.iter_entries(revision_keys))
2789
revision_ids = set(revision_ids) - present_revision_ids
2790
# implementing the TODO will involve:
2791
# - detecting when all of a pack is selected
2792
# - avoiding as much as possible pre-selection, so the
2793
# more-core routines such as create_pack_from_packs can filter in
2794
# a just-in-time fashion. (though having a HEADS list on a
2795
# repository might make this a lot easier, because we could
2796
# sensibly detect 'new revisions' without doing a full index scan.
2797
elif _mod_revision.is_null(revision_id):
2802
revision_ids = self.search_missing_revision_ids(revision_id,
2803
find_ghosts=find_ghosts).get_keys()
2804
except errors.NoSuchRevision:
2805
raise errors.InstallFailed([revision_id])
2806
if len(revision_ids) == 0:
2808
packs = self.source._pack_collection.all_packs()
2809
pack = Packer(self.target._pack_collection, packs, '.fetch',
2810
revision_ids).pack()
2811
if pack is not None:
2812
self.target._pack_collection._save_pack_names()
2813
# Trigger an autopack. This may duplicate effort as we've just done
2814
# a pack creation, but for now it is simpler to think about as
2815
# 'upload data, then repack if needed'.
2816
self.target._pack_collection.autopack()
2817
return (pack.get_revision_count(), [])
2822
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2823
"""See InterRepository.missing_revision_ids().
2825
:param find_ghosts: Find ghosts throughout the ancestry of
2828
if not find_ghosts and revision_id is not None:
2829
return self._walk_to_common_revisions([revision_id])
2830
elif revision_id is not None:
2831
# Find ghosts: search for revisions pointing from one repository to
2832
# the other, and vice versa, anywhere in the history of revision_id.
2833
graph = self.target.get_graph(other_repository=self.source)
2834
searcher = graph._make_breadth_first_searcher([revision_id])
2838
next_revs, ghosts = searcher.next_with_ghosts()
2839
except StopIteration:
2841
if revision_id in ghosts:
2842
raise errors.NoSuchRevision(self.source, revision_id)
2843
found_ids.update(next_revs)
2844
found_ids.update(ghosts)
2845
found_ids = frozenset(found_ids)
2846
# Double query here: should be able to avoid this by changing the
2847
# graph api further.
2848
result_set = found_ids - frozenset(
2849
self.target.get_parent_map(found_ids))
2851
source_ids = self.source.all_revision_ids()
2852
# source_ids is the worst possible case we may need to pull.
2853
# now we want to filter source_ids against what we actually
2854
# have in target, but don't try to check for existence where we know
2855
# we do not have a revision as that would be pointless.
2856
target_ids = set(self.target.all_revision_ids())
2857
result_set = set(source_ids).difference(target_ids)
2858
return self.source.revision_ids_to_search_result(result_set)
2861
class InterModel1and2(InterRepository):
2864
def _get_repo_format_to_test(self):
2868
def is_compatible(source, target):
2869
if not source.supports_rich_root() and target.supports_rich_root():
2875
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2876
"""See InterRepository.fetch()."""
2877
from bzrlib.fetch import Model1toKnit2Fetcher
2878
f = Model1toKnit2Fetcher(to_repository=self.target,
2879
from_repository=self.source,
2880
last_revision=revision_id,
2881
pb=pb, find_ghosts=find_ghosts)
2882
return f.count_copied, f.failed_revisions
2885
def copy_content(self, revision_id=None):
2886
"""Make a complete copy of the content in self into destination.
2888
This is a destructive operation! Do not use it on existing
2891
:param revision_id: Only copy the content needed to construct
2892
revision_id and its parents.
2895
self.target.set_make_working_trees(self.source.make_working_trees())
2896
except NotImplementedError:
2898
# but don't bother fetching if we have the needed data now.
2899
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
2900
self.target.has_revision(revision_id)):
2902
self.target.fetch(self.source, revision_id=revision_id)
2905
class InterKnit1and2(InterKnitRepo):
2908
def _get_repo_format_to_test(self):
2912
def is_compatible(source, target):
2913
"""Be compatible with Knit1 source and Knit3 target"""
2915
from bzrlib.repofmt.knitrepo import (
2916
RepositoryFormatKnit1,
2917
RepositoryFormatKnit3,
2919
from bzrlib.repofmt.pack_repo import (
2920
RepositoryFormatKnitPack1,
2921
RepositoryFormatKnitPack3,
2922
RepositoryFormatKnitPack4,
2923
RepositoryFormatKnitPack5,
2924
RepositoryFormatKnitPack5RichRoot,
2925
RepositoryFormatPackDevelopment2,
2926
RepositoryFormatPackDevelopment2Subtree,
2929
RepositoryFormatKnit1, # no rr, no subtree
2930
RepositoryFormatKnitPack1, # no rr, no subtree
2931
RepositoryFormatPackDevelopment2, # no rr, no subtree
2932
RepositoryFormatKnitPack5, # no rr, no subtree
2935
RepositoryFormatKnit3, # rr, subtree
2936
RepositoryFormatKnitPack3, # rr, subtree
2937
RepositoryFormatKnitPack4, # rr, no subtree
2938
RepositoryFormatKnitPack5RichRoot,# rr, no subtree
2939
RepositoryFormatPackDevelopment2Subtree, # rr, subtree
2941
for format in norichroot:
2942
if format.rich_root_data:
2943
raise AssertionError('Format %s is a rich-root format'
2944
' but is included in the non-rich-root list'
2946
for format in richroot:
2947
if not format.rich_root_data:
2948
raise AssertionError('Format %s is not a rich-root format'
2949
' but is included in the rich-root list'
2951
# TODO: One alternative is to just check format.rich_root_data,
2952
# instead of keeping membership lists. However, the formats
2953
# *also* have to use the same 'Knit' style of storage
2954
# (line-deltas, fulltexts, etc.)
2955
return (isinstance(source._format, norichroot) and
2956
isinstance(target._format, richroot))
2957
except AttributeError:
2961
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2962
"""See InterRepository.fetch()."""
2963
from bzrlib.fetch import Knit1to2Fetcher
2964
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2965
self.source, self.source._format, self.target,
2966
self.target._format)
2967
f = Knit1to2Fetcher(to_repository=self.target,
2968
from_repository=self.source,
2969
last_revision=revision_id,
2970
pb=pb, find_ghosts=find_ghosts)
2971
return f.count_copied, f.failed_revisions
2974
class InterDifferingSerializer(InterKnitRepo):
2977
def _get_repo_format_to_test(self):
2981
def is_compatible(source, target):
2982
"""Be compatible with Knit2 source and Knit3 target"""
2983
if source.supports_rich_root() != target.supports_rich_root():
2985
# Ideally, we'd support fetching if the source had no tree references
2986
# even if it supported them...
2987
if (getattr(source, '_format.supports_tree_reference', False) and
2988
not getattr(target, '_format.supports_tree_reference', False)):
2993
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2994
"""See InterRepository.fetch()."""
2995
revision_ids = self.target.search_missing_revision_ids(self.source,
2996
revision_id, find_ghosts=find_ghosts).get_keys()
2997
revision_ids = tsort.topo_sort(
2998
self.source.get_graph().get_parent_map(revision_ids))
2999
def revisions_iterator():
3000
for current_revision_id in revision_ids:
3001
revision = self.source.get_revision(current_revision_id)
3002
tree = self.source.revision_tree(current_revision_id)
3004
signature = self.source.get_signature_text(
3005
current_revision_id)
3006
except errors.NoSuchRevision:
3008
yield revision, tree, signature
3010
my_pb = ui.ui_factory.nested_progress_bar()
3015
install_revisions(self.target, revisions_iterator(),
3016
len(revision_ids), pb)
3018
if my_pb is not None:
3020
return len(revision_ids), 0
3023
class InterOtherToRemote(InterRepository):
3025
def __init__(self, source, target):
3026
InterRepository.__init__(self, source, target)
3027
self._real_inter = None
3030
def is_compatible(source, target):
3031
if isinstance(target, remote.RemoteRepository):
3035
def _ensure_real_inter(self):
3036
if self._real_inter is None:
3037
self.target._ensure_real()
3038
real_target = self.target._real_repository
3039
self._real_inter = InterRepository.get(self.source, real_target)
3041
def copy_content(self, revision_id=None):
3042
self._ensure_real_inter()
3043
self._real_inter.copy_content(revision_id=revision_id)
3045
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3046
self._ensure_real_inter()
3047
return self._real_inter.fetch(revision_id=revision_id, pb=pb,
3048
find_ghosts=find_ghosts)
3051
def _get_repo_format_to_test(self):
3055
class InterRemoteToOther(InterRepository):
3057
def __init__(self, source, target):
3058
InterRepository.__init__(self, source, target)
3059
self._real_inter = None
3062
def is_compatible(source, target):
3063
if not isinstance(source, remote.RemoteRepository):
3065
# Is source's model compatible with target's model?
3066
source._ensure_real()
3067
real_source = source._real_repository
3068
if isinstance(real_source, remote.RemoteRepository):
3069
raise NotImplementedError(
3070
"We don't support remote repos backed by remote repos yet.")
3071
return InterRepository._same_model(real_source, target)
3073
def _ensure_real_inter(self):
3074
if self._real_inter is None:
3075
self.source._ensure_real()
3076
real_source = self.source._real_repository
3077
self._real_inter = InterRepository.get(real_source, self.target)
3079
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3080
self._ensure_real_inter()
3081
return self._real_inter.fetch(revision_id=revision_id, pb=pb,
3082
find_ghosts=find_ghosts)
3084
def copy_content(self, revision_id=None):
3085
self._ensure_real_inter()
3086
self._real_inter.copy_content(revision_id=revision_id)
3089
def _get_repo_format_to_test(self):
3094
InterRepository.register_optimiser(InterDifferingSerializer)
3095
InterRepository.register_optimiser(InterSameDataRepository)
3096
InterRepository.register_optimiser(InterWeaveRepo)
3097
InterRepository.register_optimiser(InterKnitRepo)
3098
InterRepository.register_optimiser(InterModel1and2)
3099
InterRepository.register_optimiser(InterKnit1and2)
3100
InterRepository.register_optimiser(InterPackRepo)
3101
InterRepository.register_optimiser(InterOtherToRemote)
3102
InterRepository.register_optimiser(InterRemoteToOther)
1776
3105
class CopyConverter(object):
1777
3106
"""A repository conversion tool which just performs a copy of the content.
1779
3108
This is slow but quite reliable.
1785
3114
:param target_format: The format the resulting repository should be.
1787
3116
self.target_format = target_format
1789
3118
def convert(self, repo, pb):
1790
3119
"""Perform the conversion of to_convert, giving feedback via pb.
1792
3121
:param to_convert: The disk object to convert.
1793
3122
:param pb: a progress bar to use for progress information.
1795
pb = ui.ui_factory.nested_progress_bar()
1798
3127
# this is only useful with metadir layouts - separated repo content.
1799
3128
# trigger an assertion if not such
1800
3129
repo._format.get_format_string()
1801
3130
self.repo_dir = repo.bzrdir
1802
pb.update('Moving repository to repository.backup')
3131
self.step('Moving repository to repository.backup')
1803
3132
self.repo_dir.transport.move('repository', 'repository.backup')
1804
3133
backup_transport = self.repo_dir.transport.clone('repository.backup')
1805
3134
repo._format.check_conversion_target(self.target_format)
1806
3135
self.source_repo = repo._format.open(self.repo_dir,
1808
3137
_override_transport=backup_transport)
1809
pb.update('Creating new repository')
3138
self.step('Creating new repository')
1810
3139
converted = self.target_format.initialize(self.repo_dir,
1811
3140
self.source_repo.is_shared())
1812
3141
converted.lock_write()
1814
pb.update('Copying content')
3143
self.step('Copying content into repository.')
1815
3144
self.source_repo.copy_content_into(converted)
1817
3146
converted.unlock()
1818
pb.update('Deleting old repository content')
3147
self.step('Deleting old repository content.')
1819
3148
self.repo_dir.transport.delete_tree('repository.backup')
1820
ui.ui_factory.note('repository converted')
3149
self.pb.note('repository converted')
3151
def step(self, message):
3152
"""Update the pb by a step."""
3154
self.pb.update(message, self.count, self.total)
3166
def _unescaper(match, _map=_unescape_map):
3167
code = match.group(1)
3171
if not code.startswith('#'):
3173
return unichr(int(code[1:])).encode('utf8')
3179
def _unescape_xml(data):
3180
"""Unescape predefined XML entities in a string of data."""
3182
if _unescape_re is None:
3183
_unescape_re = re.compile('\&([^;]*);')
3184
return _unescape_re.sub(_unescaper, data)
3187
class _VersionedFileChecker(object):
3189
def __init__(self, repository):
3190
self.repository = repository
3191
self.text_index = self.repository._generate_text_key_index()
3193
def calculate_file_version_parents(self, text_key):
3194
"""Calculate the correct parents for a file version according to
3197
parent_keys = self.text_index[text_key]
3198
if parent_keys == [_mod_revision.NULL_REVISION]:
3200
return tuple(parent_keys)
3202
def check_file_version_parents(self, texts, progress_bar=None):
3203
"""Check the parents stored in a versioned file are correct.
3205
It also detects file versions that are not referenced by their
3206
corresponding revision's inventory.
3208
:returns: A tuple of (wrong_parents, dangling_file_versions).
3209
wrong_parents is a dict mapping {revision_id: (stored_parents,
3210
correct_parents)} for each revision_id where the stored parents
3211
are not correct. dangling_file_versions is a set of (file_id,
3212
revision_id) tuples for versions that are present in this versioned
3213
file, but not used by the corresponding inventory.
3216
self.file_ids = set([file_id for file_id, _ in
3217
self.text_index.iterkeys()])
3218
# text keys is now grouped by file_id
3219
n_weaves = len(self.file_ids)
3220
files_in_revisions = {}
3221
revisions_of_files = {}
3222
n_versions = len(self.text_index)
3223
progress_bar.update('loading text store', 0, n_versions)
3224
parent_map = self.repository.texts.get_parent_map(self.text_index)
3225
# On unlistable transports this could well be empty/error...
3226
text_keys = self.repository.texts.keys()
3227
unused_keys = frozenset(text_keys) - set(self.text_index)
3228
for num, key in enumerate(self.text_index.iterkeys()):
3229
if progress_bar is not None:
3230
progress_bar.update('checking text graph', num, n_versions)
3231
correct_parents = self.calculate_file_version_parents(key)
3233
knit_parents = parent_map[key]
3234
except errors.RevisionNotPresent:
3237
if correct_parents != knit_parents:
3238
wrong_parents[key] = (knit_parents, correct_parents)
3239
return wrong_parents, unused_keys
3242
def _old_get_graph(repository, revision_id):
3243
"""DO NOT USE. That is all. I'm serious."""
3244
graph = repository.get_graph()
3245
revision_graph = dict(((key, value) for key, value in
3246
graph.iter_ancestry([revision_id]) if value is not None))
3247
return _strip_NULL_ghosts(revision_graph)
1824
3250
def _strip_NULL_ghosts(revision_graph):