390
545
# unchanged, carry over.
391
546
ie.reference_revision = parent_entry.reference_revision
392
547
ie.revision = parent_entry.revision
393
return self._get_delta(ie, basis_inv, path), False
548
return self._get_delta(ie, basis_inv, path), False, None
394
549
ie.reference_revision = content_summary[3]
396
self._add_text_to_weave(ie.file_id, lines, heads, None)
550
if ie.reference_revision is None:
551
raise AssertionError("invalid content_summary for nested tree: %r"
552
% (content_summary,))
553
self._add_text_to_weave(ie.file_id, '', heads, None)
398
555
raise NotImplementedError('unknown kind')
399
556
ie.revision = self._new_revision_id
400
return self._get_delta(ie, basis_inv, path), True
402
def _add_text_to_weave(self, file_id, new_lines, parents, nostore_sha):
403
versionedfile = self.repository.weave_store.get_weave_or_empty(
404
file_id, self.repository.get_transaction())
405
# Don't change this to add_lines - add_lines_with_ghosts is cheaper
406
# than add_lines, and allows committing when a parent is ghosted for
408
# Note: as we read the content directly from the tree, we know its not
409
# been turned into unicode or badly split - but a broken tree
410
# implementation could give us bad output from readlines() so this is
411
# not a guarantee of safety. What would be better is always checking
412
# the content during test suite execution. RBC 20070912
557
self._any_changes = True
558
return self._get_delta(ie, basis_inv, path), True, fingerprint
560
def record_iter_changes(self, tree, basis_revision_id, iter_changes,
561
_entry_factory=entry_factory):
562
"""Record a new tree via iter_changes.
564
:param tree: The tree to obtain text contents from for changed objects.
565
:param basis_revision_id: The revision id of the tree the iter_changes
566
has been generated against. Currently assumed to be the same
567
as self.parents[0] - if it is not, errors may occur.
568
:param iter_changes: An iter_changes iterator with the changes to apply
569
to basis_revision_id. The iterator must not include any items with
570
a current kind of None - missing items must be either filtered out
571
or errored-on beefore record_iter_changes sees the item.
572
:param _entry_factory: Private method to bind entry_factory locally for
574
:return: A generator of (file_id, relpath, fs_hash) tuples for use with
577
# Create an inventory delta based on deltas between all the parents and
578
# deltas between all the parent inventories. We use inventory delta's
579
# between the inventory objects because iter_changes masks
580
# last-changed-field only changes.
582
# file_id -> change map, change is fileid, paths, changed, versioneds,
583
# parents, names, kinds, executables
585
# {file_id -> revision_id -> inventory entry, for entries in parent
586
# trees that are not parents[0]
414
return versionedfile.add_lines_with_ghosts(
415
self._new_revision_id, parents, new_lines,
416
nostore_sha=nostore_sha, random_id=self.random_revid,
417
check_content=False)[0:2]
419
versionedfile.clear_cache()
590
revtrees = list(self.repository.revision_trees(self.parents))
591
except errors.NoSuchRevision:
592
# one or more ghosts, slow path.
594
for revision_id in self.parents:
596
revtrees.append(self.repository.revision_tree(revision_id))
597
except errors.NoSuchRevision:
599
basis_revision_id = _mod_revision.NULL_REVISION
601
revtrees.append(self.repository.revision_tree(
602
_mod_revision.NULL_REVISION))
603
# The basis inventory from a repository
605
basis_inv = revtrees[0].inventory
607
basis_inv = self.repository.revision_tree(
608
_mod_revision.NULL_REVISION).inventory
609
if len(self.parents) > 0:
610
if basis_revision_id != self.parents[0] and not ghost_basis:
612
"arbitrary basis parents not yet supported with merges")
613
for revtree in revtrees[1:]:
614
for change in revtree.inventory._make_delta(basis_inv):
615
if change[1] is None:
616
# Not present in this parent.
618
if change[2] not in merged_ids:
619
if change[0] is not None:
620
basis_entry = basis_inv[change[2]]
621
merged_ids[change[2]] = [
623
basis_entry.revision,
626
parent_entries[change[2]] = {
628
basis_entry.revision:basis_entry,
630
change[3].revision:change[3],
633
merged_ids[change[2]] = [change[3].revision]
634
parent_entries[change[2]] = {change[3].revision:change[3]}
636
merged_ids[change[2]].append(change[3].revision)
637
parent_entries[change[2]][change[3].revision] = change[3]
640
# Setup the changes from the tree:
641
# changes maps file_id -> (change, [parent revision_ids])
643
for change in iter_changes:
644
# This probably looks up in basis_inv way to much.
645
if change[1][0] is not None:
646
head_candidate = [basis_inv[change[0]].revision]
649
changes[change[0]] = change, merged_ids.get(change[0],
651
unchanged_merged = set(merged_ids) - set(changes)
652
# Extend the changes dict with synthetic changes to record merges of
654
for file_id in unchanged_merged:
655
# Record a merged version of these items that did not change vs the
656
# basis. This can be either identical parallel changes, or a revert
657
# of a specific file after a merge. The recorded content will be
658
# that of the current tree (which is the same as the basis), but
659
# the per-file graph will reflect a merge.
660
# NB:XXX: We are reconstructing path information we had, this
661
# should be preserved instead.
662
# inv delta change: (file_id, (path_in_source, path_in_target),
663
# changed_content, versioned, parent, name, kind,
666
basis_entry = basis_inv[file_id]
667
except errors.NoSuchId:
668
# a change from basis->some_parents but file_id isn't in basis
669
# so was new in the merge, which means it must have changed
670
# from basis -> current, and as it hasn't the add was reverted
671
# by the user. So we discard this change.
675
(basis_inv.id2path(file_id), tree.id2path(file_id)),
677
(basis_entry.parent_id, basis_entry.parent_id),
678
(basis_entry.name, basis_entry.name),
679
(basis_entry.kind, basis_entry.kind),
680
(basis_entry.executable, basis_entry.executable))
681
changes[file_id] = (change, merged_ids[file_id])
682
# changes contains tuples with the change and a set of inventory
683
# candidates for the file.
685
# old_path, new_path, file_id, new_inventory_entry
686
seen_root = False # Is the root in the basis delta?
687
inv_delta = self._basis_delta
688
modified_rev = self._new_revision_id
689
for change, head_candidates in changes.values():
690
if change[3][1]: # versioned in target.
691
# Several things may be happening here:
692
# We may have a fork in the per-file graph
693
# - record a change with the content from tree
694
# We may have a change against < all trees
695
# - carry over the tree that hasn't changed
696
# We may have a change against all trees
697
# - record the change with the content from tree
700
entry = _entry_factory[kind](file_id, change[5][1],
702
head_set = self._heads(change[0], set(head_candidates))
705
for head_candidate in head_candidates:
706
if head_candidate in head_set:
707
heads.append(head_candidate)
708
head_set.remove(head_candidate)
711
# Could be a carry-over situation:
712
parent_entry_revs = parent_entries.get(file_id, None)
713
if parent_entry_revs:
714
parent_entry = parent_entry_revs.get(heads[0], None)
717
if parent_entry is None:
718
# The parent iter_changes was called against is the one
719
# that is the per-file head, so any change is relevant
720
# iter_changes is valid.
721
carry_over_possible = False
723
# could be a carry over situation
724
# A change against the basis may just indicate a merge,
725
# we need to check the content against the source of the
726
# merge to determine if it was changed after the merge
728
if (parent_entry.kind != entry.kind or
729
parent_entry.parent_id != entry.parent_id or
730
parent_entry.name != entry.name):
731
# Metadata common to all entries has changed
732
# against per-file parent
733
carry_over_possible = False
735
carry_over_possible = True
736
# per-type checks for changes against the parent_entry
739
# Cannot be a carry-over situation
740
carry_over_possible = False
741
# Populate the entry in the delta
743
# XXX: There is still a small race here: If someone reverts the content of a file
744
# after iter_changes examines and decides it has changed,
745
# we will unconditionally record a new version even if some
746
# other process reverts it while commit is running (with
747
# the revert happening after iter_changes did it's
750
entry.executable = True
752
entry.executable = False
753
if (carry_over_possible and
754
parent_entry.executable == entry.executable):
755
# Check the file length, content hash after reading
757
nostore_sha = parent_entry.text_sha1
760
file_obj, stat_value = tree.get_file_with_stat(file_id, change[1][1])
762
text = file_obj.read()
766
entry.text_sha1, entry.text_size = self._add_text_to_weave(
767
file_id, text, heads, nostore_sha)
768
yield file_id, change[1][1], (entry.text_sha1, stat_value)
769
except errors.ExistingContent:
770
# No content change against a carry_over parent
771
# Perhaps this should also yield a fs hash update?
773
entry.text_size = parent_entry.text_size
774
entry.text_sha1 = parent_entry.text_sha1
775
elif kind == 'symlink':
777
entry.symlink_target = tree.get_symlink_target(file_id)
778
if (carry_over_possible and
779
parent_entry.symlink_target == entry.symlink_target):
782
self._add_text_to_weave(change[0], '', heads, None)
783
elif kind == 'directory':
784
if carry_over_possible:
787
# Nothing to set on the entry.
788
# XXX: split into the Root and nonRoot versions.
789
if change[1][1] != '' or self.repository.supports_rich_root():
790
self._add_text_to_weave(change[0], '', heads, None)
791
elif kind == 'tree-reference':
792
if not self.repository._format.supports_tree_reference:
793
# This isn't quite sane as an error, but we shouldn't
794
# ever see this code path in practice: tree's don't
795
# permit references when the repo doesn't support tree
797
raise errors.UnsupportedOperation(tree.add_reference,
799
reference_revision = tree.get_reference_revision(change[0])
800
entry.reference_revision = reference_revision
801
if (carry_over_possible and
802
parent_entry.reference_revision == reference_revision):
805
self._add_text_to_weave(change[0], '', heads, None)
807
raise AssertionError('unknown kind %r' % kind)
809
entry.revision = modified_rev
811
entry.revision = parent_entry.revision
814
new_path = change[1][1]
815
inv_delta.append((change[1][0], new_path, change[0], entry))
818
self.new_inventory = None
820
# This should perhaps be guarded by a check that the basis we
821
# commit against is the basis for the commit and if not do a delta
823
self._any_changes = True
825
# housekeeping root entry changes do not affect no-change commits.
826
self._require_root_change(tree)
827
self.basis_delta_revision = basis_revision_id
829
def _add_text_to_weave(self, file_id, new_text, parents, nostore_sha):
830
parent_keys = tuple([(file_id, parent) for parent in parents])
831
return self.repository.texts._add_text(
832
(file_id, self._new_revision_id), parent_keys, new_text,
833
nostore_sha=nostore_sha, random_id=self.random_revid)[0:2]
422
836
class RootCommitBuilder(CommitBuilder):
423
837
"""This commitbuilder actually records the root id"""
425
839
# the root entry gets versioned properly by this builder.
426
840
_versioned_root = True
973
def add_fallback_repository(self, repository):
974
"""Add a repository to use for looking up data not held locally.
976
:param repository: A repository.
978
if not self._format.supports_external_lookups:
979
raise errors.UnstackableRepositoryFormat(self._format, self.base)
981
# This repository will call fallback.unlock() when we transition to
982
# the unlocked state, so we make sure to increment the lock count
983
repository.lock_read()
984
self._check_fallback_repository(repository)
985
self._fallback_repositories.append(repository)
986
self.texts.add_fallback_versioned_files(repository.texts)
987
self.inventories.add_fallback_versioned_files(repository.inventories)
988
self.revisions.add_fallback_versioned_files(repository.revisions)
989
self.signatures.add_fallback_versioned_files(repository.signatures)
990
if self.chk_bytes is not None:
991
self.chk_bytes.add_fallback_versioned_files(repository.chk_bytes)
993
def _check_fallback_repository(self, repository):
994
"""Check that this repository can fallback to repository safely.
996
Raise an error if not.
998
:param repository: A repository to fallback to.
1000
return InterRepository._assert_same_model(self, repository)
489
1002
def add_inventory(self, revision_id, inv, parents):
490
1003
"""Add the inventory inv to the repository as revision_id.
492
1005
:param parents: The revision ids of the parents that revision_id
493
1006
is known to have and are in the repository already.
495
returns the sha1 of the serialized inventory.
1008
:returns: The validator(which is a sha1 digest, though what is sha'd is
1009
repository format specific) of the serialized inventory.
497
assert self.is_in_write_group()
1011
if not self.is_in_write_group():
1012
raise AssertionError("%r not in write group" % (self,))
498
1013
_mod_revision.check_not_reserved_id(revision_id)
499
assert inv.revision_id is None or inv.revision_id == revision_id, \
500
"Mismatch between inventory revision" \
501
" id and insertion revid (%r, %r)" % (inv.revision_id, revision_id)
502
assert inv.root is not None
503
inv_lines = self._serialise_inventory_to_lines(inv)
504
inv_vf = self.get_inventory_weave()
505
return self._inventory_add_lines(inv_vf, revision_id, parents,
1014
if not (inv.revision_id is None or inv.revision_id == revision_id):
1015
raise AssertionError(
1016
"Mismatch between inventory revision"
1017
" id and insertion revid (%r, %r)"
1018
% (inv.revision_id, revision_id))
1019
if inv.root is None:
1020
raise AssertionError()
1021
return self._add_inventory_checked(revision_id, inv, parents)
1023
def _add_inventory_checked(self, revision_id, inv, parents):
1024
"""Add inv to the repository after checking the inputs.
1026
This function can be overridden to allow different inventory styles.
1028
:seealso: add_inventory, for the contract.
1030
inv_lines = self._serializer.write_inventory_to_lines(inv)
1031
return self._inventory_add_lines(revision_id, parents,
506
1032
inv_lines, check_content=False)
508
def _inventory_add_lines(self, inv_vf, revision_id, parents, lines,
1034
def add_inventory_by_delta(self, basis_revision_id, delta, new_revision_id,
1035
parents, basis_inv=None, propagate_caches=False):
1036
"""Add a new inventory expressed as a delta against another revision.
1038
See the inventory developers documentation for the theory behind
1041
:param basis_revision_id: The inventory id the delta was created
1042
against. (This does not have to be a direct parent.)
1043
:param delta: The inventory delta (see Inventory.apply_delta for
1045
:param new_revision_id: The revision id that the inventory is being
1047
:param parents: The revision ids of the parents that revision_id is
1048
known to have and are in the repository already. These are supplied
1049
for repositories that depend on the inventory graph for revision
1050
graph access, as well as for those that pun ancestry with delta
1052
:param basis_inv: The basis inventory if it is already known,
1054
:param propagate_caches: If True, the caches for this inventory are
1055
copied to and updated for the result if possible.
1057
:returns: (validator, new_inv)
1058
The validator(which is a sha1 digest, though what is sha'd is
1059
repository format specific) of the serialized inventory, and the
1060
resulting inventory.
1062
if not self.is_in_write_group():
1063
raise AssertionError("%r not in write group" % (self,))
1064
_mod_revision.check_not_reserved_id(new_revision_id)
1065
basis_tree = self.revision_tree(basis_revision_id)
1066
basis_tree.lock_read()
1068
# Note that this mutates the inventory of basis_tree, which not all
1069
# inventory implementations may support: A better idiom would be to
1070
# return a new inventory, but as there is no revision tree cache in
1071
# repository this is safe for now - RBC 20081013
1072
if basis_inv is None:
1073
basis_inv = basis_tree.inventory
1074
basis_inv.apply_delta(delta)
1075
basis_inv.revision_id = new_revision_id
1076
return (self.add_inventory(new_revision_id, basis_inv, parents),
1081
def _inventory_add_lines(self, revision_id, parents, lines,
509
1082
check_content=True):
510
1083
"""Store lines in inv_vf and return the sha1 of the inventory."""
512
for parent in parents:
514
final_parents.append(parent)
515
return inv_vf.add_lines(revision_id, final_parents, lines,
1084
parents = [(parent,) for parent in parents]
1085
result = self.inventories.add_lines((revision_id,), parents, lines,
516
1086
check_content=check_content)[0]
1087
self.inventories._access.flush()
519
1090
def add_revision(self, revision_id, rev, inv=None, config=None):
520
1091
"""Add rev to the revision store as revision_id.
585
1162
Returns a set of the present revisions.
588
for id in revision_ids:
589
if self.has_revision(id):
1165
graph = self.get_graph()
1166
parent_map = graph.get_parent_map(revision_ids)
1167
# The old API returned a list, should this actually be a set?
1168
return parent_map.keys()
1170
def _check_inventories(self, checker):
1171
"""Check the inventories found from the revision scan.
1173
This is responsible for verifying the sha1 of inventories and
1174
creating a pending_keys set that covers data referenced by inventories.
1176
bar = ui.ui_factory.nested_progress_bar()
1178
self._do_check_inventories(checker, bar)
1182
def _do_check_inventories(self, checker, bar):
1183
"""Helper for _check_inventories."""
1185
keys = {'chk_bytes':set(), 'inventories':set(), 'texts':set()}
1186
kinds = ['chk_bytes', 'texts']
1187
count = len(checker.pending_keys)
1188
bar.update("inventories", 0, 2)
1189
current_keys = checker.pending_keys
1190
checker.pending_keys = {}
1191
# Accumulate current checks.
1192
for key in current_keys:
1193
if key[0] != 'inventories' and key[0] not in kinds:
1194
checker._report_items.append('unknown key type %r' % (key,))
1195
keys[key[0]].add(key[1:])
1196
if keys['inventories']:
1197
# NB: output order *should* be roughly sorted - topo or
1198
# inverse topo depending on repository - either way decent
1199
# to just delta against. However, pre-CHK formats didn't
1200
# try to optimise inventory layout on disk. As such the
1201
# pre-CHK code path does not use inventory deltas.
1203
for record in self.inventories.check(keys=keys['inventories']):
1204
if record.storage_kind == 'absent':
1205
checker._report_items.append(
1206
'Missing inventory {%s}' % (record.key,))
1208
last_object = self._check_record('inventories', record,
1209
checker, last_object,
1210
current_keys[('inventories',) + record.key])
1211
del keys['inventories']
1214
bar.update("texts", 1)
1215
while (checker.pending_keys or keys['chk_bytes']
1217
# Something to check.
1218
current_keys = checker.pending_keys
1219
checker.pending_keys = {}
1220
# Accumulate current checks.
1221
for key in current_keys:
1222
if key[0] not in kinds:
1223
checker._report_items.append('unknown key type %r' % (key,))
1224
keys[key[0]].add(key[1:])
1225
# Check the outermost kind only - inventories || chk_bytes || texts
1229
for record in getattr(self, kind).check(keys=keys[kind]):
1230
if record.storage_kind == 'absent':
1231
checker._report_items.append(
1232
'Missing %s {%s}' % (kind, record.key,))
1234
last_object = self._check_record(kind, record,
1235
checker, last_object, current_keys[(kind,) + record.key])
1239
def _check_record(self, kind, record, checker, last_object, item_data):
1240
"""Check a single text from this repository."""
1241
if kind == 'inventories':
1242
rev_id = record.key[0]
1243
inv = self._deserialise_inventory(rev_id,
1244
record.get_bytes_as('fulltext'))
1245
if last_object is not None:
1246
delta = inv._make_delta(last_object)
1247
for old_path, path, file_id, ie in delta:
1250
ie.check(checker, rev_id, inv)
1252
for path, ie in inv.iter_entries():
1253
ie.check(checker, rev_id, inv)
1254
if self._format.fast_deltas:
1256
elif kind == 'chk_bytes':
1257
# No code written to check chk_bytes for this repo format.
1258
checker._report_items.append(
1259
'unsupported key type chk_bytes for %s' % (record.key,))
1260
elif kind == 'texts':
1261
self._check_text(record, checker, item_data)
1263
checker._report_items.append(
1264
'unknown key type %s for %s' % (kind, record.key))
1266
def _check_text(self, record, checker, item_data):
1267
"""Check a single text."""
1268
# Check it is extractable.
1269
# TODO: check length.
1270
if record.storage_kind == 'chunked':
1271
chunks = record.get_bytes_as(record.storage_kind)
1272
sha1 = osutils.sha_strings(chunks)
1273
length = sum(map(len, chunks))
1275
content = record.get_bytes_as('fulltext')
1276
sha1 = osutils.sha_string(content)
1277
length = len(content)
1278
if item_data and sha1 != item_data[1]:
1279
checker._report_items.append(
1280
'sha1 mismatch: %s has sha1 %s expected %s referenced by %s' %
1281
(record.key, sha1, item_data[1], item_data[2]))
594
1284
def create(a_bzrdir):
595
1285
"""Construct the current default format repository in a_bzrdir."""
596
1286
return RepositoryFormat.get_default_format().initialize(a_bzrdir)
598
def __init__(self, _format, a_bzrdir, control_files, _revision_store, control_store, text_store):
1288
def __init__(self, _format, a_bzrdir, control_files):
599
1289
"""instantiate a Repository.
601
1291
:param _format: The format of the repository on disk.
822
1546
"""Commit the contents accrued within the current write group.
824
1548
:seealso: start_write_group.
1550
:return: it may return an opaque hint that can be passed to 'pack'.
826
1552
if self._write_group is not self.get_transaction():
827
1553
# has an unlock or relock occured ?
828
1554
raise errors.BzrError('mismatched lock context %r and '
829
1555
'write group %r.' %
830
1556
(self.get_transaction(), self._write_group))
831
self._commit_write_group()
1557
result = self._commit_write_group()
832
1558
self._write_group = None
834
1561
def _commit_write_group(self):
835
1562
"""Template method for per-repository write group cleanup.
837
This is called before the write group is considered to be
1564
This is called before the write group is considered to be
838
1565
finished and should ensure that all data handed to the repository
839
for writing during the write group is safely committed (to the
1566
for writing during the write group is safely committed (to the
840
1567
extent possible considering file system caching etc).
843
def fetch(self, source, revision_id=None, pb=None, find_ghosts=False):
1570
def suspend_write_group(self):
1571
raise errors.UnsuspendableWriteGroup(self)
1573
def get_missing_parent_inventories(self, check_for_missing_texts=True):
1574
"""Return the keys of missing inventory parents for revisions added in
1577
A revision is not complete if the inventory delta for that revision
1578
cannot be calculated. Therefore if the parent inventories of a
1579
revision are not present, the revision is incomplete, and e.g. cannot
1580
be streamed by a smart server. This method finds missing inventory
1581
parents for revisions added in this write group.
1583
if not self._format.supports_external_lookups:
1584
# This is only an issue for stacked repositories
1586
if not self.is_in_write_group():
1587
raise AssertionError('not in a write group')
1589
# XXX: We assume that every added revision already has its
1590
# corresponding inventory, so we only check for parent inventories that
1591
# might be missing, rather than all inventories.
1592
parents = set(self.revisions._index.get_missing_parents())
1593
parents.discard(_mod_revision.NULL_REVISION)
1594
unstacked_inventories = self.inventories._index
1595
present_inventories = unstacked_inventories.get_parent_map(
1596
key[-1:] for key in parents)
1597
parents.difference_update(present_inventories)
1598
if len(parents) == 0:
1599
# No missing parent inventories.
1601
if not check_for_missing_texts:
1602
return set(('inventories', rev_id) for (rev_id,) in parents)
1603
# Ok, now we have a list of missing inventories. But these only matter
1604
# if the inventories that reference them are missing some texts they
1605
# appear to introduce.
1606
# XXX: Texts referenced by all added inventories need to be present,
1607
# but at the moment we're only checking for texts referenced by
1608
# inventories at the graph's edge.
1609
key_deps = self.revisions._index._key_dependencies
1610
key_deps.satisfy_refs_for_keys(present_inventories)
1611
referrers = frozenset(r[0] for r in key_deps.get_referrers())
1612
file_ids = self.fileids_altered_by_revision_ids(referrers)
1613
missing_texts = set()
1614
for file_id, version_ids in file_ids.iteritems():
1615
missing_texts.update(
1616
(file_id, version_id) for version_id in version_ids)
1617
present_texts = self.texts.get_parent_map(missing_texts)
1618
missing_texts.difference_update(present_texts)
1619
if not missing_texts:
1620
# No texts are missing, so all revisions and their deltas are
1623
# Alternatively the text versions could be returned as the missing
1624
# keys, but this is likely to be less data.
1625
missing_keys = set(('inventories', rev_id) for (rev_id,) in parents)
1628
def refresh_data(self):
1629
"""Re-read any data needed to to synchronise with disk.
1631
This method is intended to be called after another repository instance
1632
(such as one used by a smart server) has inserted data into the
1633
repository. It may not be called during a write group, but may be
1634
called at any other time.
1636
if self.is_in_write_group():
1637
raise errors.InternalBzrError(
1638
"May not refresh_data while in a write group.")
1639
self._refresh_data()
1641
def resume_write_group(self, tokens):
1642
if not self.is_write_locked():
1643
raise errors.NotWriteLocked(self)
1644
if self._write_group:
1645
raise errors.BzrError('already in a write group')
1646
self._resume_write_group(tokens)
1647
# so we can detect unlock/relock - the write group is now entered.
1648
self._write_group = self.get_transaction()
1650
def _resume_write_group(self, tokens):
1651
raise errors.UnsuspendableWriteGroup(self)
1653
def fetch(self, source, revision_id=None, pb=None, find_ghosts=False,
844
1655
"""Fetch the content required to construct revision_id from source.
846
If revision_id is None all content is copied.
1657
If revision_id is None and fetch_spec is None, then all content is
1660
fetch() may not be used when the repository is in a write group -
1661
either finish the current write group before using fetch, or use
1662
fetch before starting the write group.
847
1664
:param find_ghosts: Find and copy revisions in the source that are
848
1665
ghosts in the target (and not reachable directly by walking out to
849
1666
the first-present revision in target from revision_id).
1667
:param revision_id: If specified, all the content needed for this
1668
revision ID will be copied to the target. Fetch will determine for
1669
itself which content needs to be copied.
1670
:param fetch_spec: If specified, a SearchResult or
1671
PendingAncestryResult that describes which revisions to copy. This
1672
allows copying multiple heads at once. Mutually exclusive with
1675
if fetch_spec is not None and revision_id is not None:
1676
raise AssertionError(
1677
"fetch_spec and revision_id are mutually exclusive.")
1678
if self.is_in_write_group():
1679
raise errors.InternalBzrError(
1680
"May not fetch while in a write group.")
851
1681
# fast path same-url fetch operations
852
if self.has_same_location(source):
1682
# TODO: lift out to somewhere common with RemoteRepository
1683
# <https://bugs.edge.launchpad.net/bzr/+bug/401646>
1684
if (self.has_same_location(source)
1685
and fetch_spec is None
1686
and self._has_same_fallbacks(source)):
853
1687
# check that last_revision is in 'from' and then return a
855
1689
if (revision_id is not None and
856
1690
not _mod_revision.is_null(revision_id)):
857
1691
self.get_revision(revision_id)
1693
# if there is no specific appropriate InterRepository, this will get
1694
# the InterRepository base class, which raises an
1695
# IncompatibleRepositories when asked to fetch.
859
1696
inter = InterRepository.get(source, self)
861
return inter.fetch(revision_id=revision_id, pb=pb, find_ghosts=find_ghosts)
862
except NotImplementedError:
863
raise errors.IncompatibleRepositories(source, self)
1697
return inter.fetch(revision_id=revision_id, pb=pb,
1698
find_ghosts=find_ghosts, fetch_spec=fetch_spec)
865
1700
def create_bundle(self, target, base, fileobj, format=None):
866
1701
return serializer.write_bundle(self, target, base, fileobj, format)
988
1852
@needs_read_lock
989
1853
def get_revisions(self, revision_ids):
990
"""Get many revisions at once."""
1854
"""Get many revisions at once.
1856
Repositories that need to check data on every revision read should
1857
subclass this method.
991
1859
return self._get_revisions(revision_ids)
993
1861
@needs_read_lock
994
1862
def _get_revisions(self, revision_ids):
995
1863
"""Core work logic to get many revisions without sanity checks."""
996
for rev_id in revision_ids:
997
if not rev_id or not isinstance(rev_id, basestring):
998
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
999
revs = self._revision_store.get_revisions(revision_ids,
1000
self.get_transaction())
1002
assert not isinstance(rev.revision_id, unicode)
1003
for parent_id in rev.parent_ids:
1004
assert not isinstance(parent_id, unicode)
1008
def get_revision_xml(self, revision_id):
1009
# TODO: jam 20070210 This shouldn't be necessary since get_revision
1010
# would have already do it.
1011
# TODO: jam 20070210 Just use _serializer.write_revision_to_string()
1012
rev = self.get_revision(revision_id)
1013
rev_tmp = StringIO()
1014
# the current serializer..
1015
self._revision_store._serializer.write_revision(rev, rev_tmp)
1017
return rev_tmp.getvalue()
1020
def get_deltas_for_revisions(self, revisions):
1865
for revid, rev in self._iter_revisions(revision_ids):
1867
raise errors.NoSuchRevision(self, revid)
1869
return [revs[revid] for revid in revision_ids]
1871
def _iter_revisions(self, revision_ids):
1872
"""Iterate over revision objects.
1874
:param revision_ids: An iterable of revisions to examine. None may be
1875
passed to request all revisions known to the repository. Note that
1876
not all repositories can find unreferenced revisions; for those
1877
repositories only referenced ones will be returned.
1878
:return: An iterator of (revid, revision) tuples. Absent revisions (
1879
those asked for but not available) are returned as (revid, None).
1881
if revision_ids is None:
1882
revision_ids = self.all_revision_ids()
1884
for rev_id in revision_ids:
1885
if not rev_id or not isinstance(rev_id, basestring):
1886
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1887
keys = [(key,) for key in revision_ids]
1888
stream = self.revisions.get_record_stream(keys, 'unordered', True)
1889
for record in stream:
1890
revid = record.key[0]
1891
if record.storage_kind == 'absent':
1894
text = record.get_bytes_as('fulltext')
1895
rev = self._serializer.read_revision_from_string(text)
1898
def get_deltas_for_revisions(self, revisions, specific_fileids=None):
1021
1899
"""Produce a generator of revision deltas.
1023
1901
Note that the input is a sequence of REVISIONS, not revision_ids.
1024
1902
Trees will be held in memory until the generator exits.
1025
1903
Each delta is relative to the revision's lefthand predecessor.
1905
:param specific_fileids: if not None, the result is filtered
1906
so that only those file-ids, their parents and their
1907
children are included.
1909
# Get the revision-ids of interest
1027
1910
required_trees = set()
1028
1911
for revision in revisions:
1029
1912
required_trees.add(revision.revision_id)
1030
1913
required_trees.update(revision.parent_ids[:1])
1031
trees = dict((t.get_revision_id(), t) for
1032
t in self.revision_trees(required_trees))
1915
# Get the matching filtered trees. Note that it's more
1916
# efficient to pass filtered trees to changes_from() rather
1917
# than doing the filtering afterwards. changes_from() could
1918
# arguably do the filtering itself but it's path-based, not
1919
# file-id based, so filtering before or afterwards is
1921
if specific_fileids is None:
1922
trees = dict((t.get_revision_id(), t) for
1923
t in self.revision_trees(required_trees))
1925
trees = dict((t.get_revision_id(), t) for
1926
t in self._filtered_revision_trees(required_trees,
1929
# Calculate the deltas
1033
1930
for revision in revisions:
1034
1931
if not revision.parent_ids:
1035
old_tree = self.revision_tree(None)
1932
old_tree = self.revision_tree(_mod_revision.NULL_REVISION)
1037
1934
old_tree = trees[revision.parent_ids[0]]
1038
1935
yield trees[revision.revision_id].changes_from(old_tree)
1040
1937
@needs_read_lock
1041
def get_revision_delta(self, revision_id):
1938
def get_revision_delta(self, revision_id, specific_fileids=None):
1042
1939
"""Return the delta for one revision.
1044
1941
The delta is relative to the left-hand predecessor of the
1944
:param specific_fileids: if not None, the result is filtered
1945
so that only those file-ids, their parents and their
1946
children are included.
1047
1948
r = self.get_revision(revision_id)
1048
return list(self.get_deltas_for_revisions([r]))[0]
1949
return list(self.get_deltas_for_revisions([r],
1950
specific_fileids=specific_fileids))[0]
1050
1952
@needs_write_lock
1051
1953
def store_revision_signature(self, gpg_strategy, plaintext, revision_id):
1052
1954
signature = gpg_strategy.sign(plaintext)
1053
self._revision_store.add_revision_signature_text(revision_id,
1055
self.get_transaction())
1057
def _find_file_ids_from_xml_inventory_lines(self, line_iterator,
1059
"""Helper routine for fileids_altered_by_revision_ids.
1955
self.add_signature_text(revision_id, signature)
1958
def add_signature_text(self, revision_id, signature):
1959
self.signatures.add_lines((revision_id,), (),
1960
osutils.split_lines(signature))
1962
def find_text_key_references(self):
1963
"""Find the text key references within the repository.
1965
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1966
to whether they were referred to by the inventory of the
1967
revision_id that they contain. The inventory texts from all present
1968
revision ids are assessed to generate this report.
1970
revision_keys = self.revisions.keys()
1971
w = self.inventories
1972
pb = ui.ui_factory.nested_progress_bar()
1974
return self._find_text_key_references_from_xml_inventory_lines(
1975
w.iter_lines_added_or_present_in_keys(revision_keys, pb=pb))
1979
def _find_text_key_references_from_xml_inventory_lines(self,
1981
"""Core routine for extracting references to texts from inventories.
1061
1983
This performs the translation of xml lines to revision ids.
1063
:param line_iterator: An iterator of lines
1064
:param revision_ids: The revision ids to filter for. This should be a
1065
set or other type which supports efficient __contains__ lookups, as
1066
the revision id from each parsed line will be looked up in the
1067
revision_ids filter.
1068
:return: a dictionary mapping altered file-ids to an iterable of
1069
revision_ids. Each altered file-ids has the exact revision_ids that
1070
altered it listed explicitly.
1985
:param line_iterator: An iterator of lines, origin_version_id
1986
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1987
to whether they were referred to by the inventory of the
1988
revision_id that they contain. Note that if that revision_id was
1989
not part of the line_iterator's output then False will be given -
1990
even though it may actually refer to that key.
1992
if not self._serializer.support_altered_by_hack:
1993
raise AssertionError(
1994
"_find_text_key_references_from_xml_inventory_lines only "
1995
"supported for branches which store inventory as unnested xml"
1996
", not on %r" % self)
1074
1999
# this code needs to read every new line in every inventory for the
1075
2000
# inventories [revision_ids]. Seeing a line twice is ok. Seeing a line
1076
# not present in one of those inventories is unnecessary but not
2001
# not present in one of those inventories is unnecessary but not
1077
2002
# harmful because we are filtering by the revision id marker in the
1078
# inventory lines : we only select file ids altered in one of those
2003
# inventory lines : we only select file ids altered in one of those
1079
2004
# revisions. We don't need to see all lines in the inventory because
1080
2005
# only those added in an inventory in rev X can contain a revision=X
1114
2039
unescape_revid_cache[revision_id] = unescaped
1115
2040
revision_id = unescaped
1117
if revision_id in revision_ids:
1119
file_id = unescape_fileid_cache[file_id]
1121
unescaped = unescape(file_id)
1122
unescape_fileid_cache[file_id] = unescaped
1124
setdefault(file_id, set()).add(revision_id)
1127
def fileids_altered_by_revision_ids(self, revision_ids):
2042
# Note that unconditionally unescaping means that we deserialise
2043
# every fileid, which for general 'pull' is not great, but we don't
2044
# really want to have some many fulltexts that this matters anyway.
2047
file_id = unescape_fileid_cache[file_id]
2049
unescaped = unescape(file_id)
2050
unescape_fileid_cache[file_id] = unescaped
2053
key = (file_id, revision_id)
2054
setdefault(key, False)
2055
if revision_id == line_key[-1]:
2059
def _inventory_xml_lines_for_keys(self, keys):
2060
"""Get a line iterator of the sort needed for findind references.
2062
Not relevant for non-xml inventory repositories.
2064
Ghosts in revision_keys are ignored.
2066
:param revision_keys: The revision keys for the inventories to inspect.
2067
:return: An iterator over (inventory line, revid) for the fulltexts of
2068
all of the xml inventories specified by revision_keys.
2070
stream = self.inventories.get_record_stream(keys, 'unordered', True)
2071
for record in stream:
2072
if record.storage_kind != 'absent':
2073
chunks = record.get_bytes_as('chunked')
2074
revid = record.key[-1]
2075
lines = osutils.chunks_to_lines(chunks)
2079
def _find_file_ids_from_xml_inventory_lines(self, line_iterator,
2081
"""Helper routine for fileids_altered_by_revision_ids.
2083
This performs the translation of xml lines to revision ids.
2085
:param line_iterator: An iterator of lines, origin_version_id
2086
:param revision_keys: The revision ids to filter for. This should be a
2087
set or other type which supports efficient __contains__ lookups, as
2088
the revision key from each parsed line will be looked up in the
2089
revision_keys filter.
2090
:return: a dictionary mapping altered file-ids to an iterable of
2091
revision_ids. Each altered file-ids has the exact revision_ids that
2092
altered it listed explicitly.
2094
seen = set(self._find_text_key_references_from_xml_inventory_lines(
2095
line_iterator).iterkeys())
2096
parent_keys = self._find_parent_keys_of_revisions(revision_keys)
2097
parent_seen = set(self._find_text_key_references_from_xml_inventory_lines(
2098
self._inventory_xml_lines_for_keys(parent_keys)))
2099
new_keys = seen - parent_seen
2101
setdefault = result.setdefault
2102
for key in new_keys:
2103
setdefault(key[0], set()).add(key[-1])
2106
def _find_parent_ids_of_revisions(self, revision_ids):
2107
"""Find all parent ids that are mentioned in the revision graph.
2109
:return: set of revisions that are parents of revision_ids which are
2110
not part of revision_ids themselves
2112
parent_map = self.get_parent_map(revision_ids)
2114
map(parent_ids.update, parent_map.itervalues())
2115
parent_ids.difference_update(revision_ids)
2116
parent_ids.discard(_mod_revision.NULL_REVISION)
2119
def _find_parent_keys_of_revisions(self, revision_keys):
2120
"""Similar to _find_parent_ids_of_revisions, but used with keys.
2122
:param revision_keys: An iterable of revision_keys.
2123
:return: The parents of all revision_keys that are not already in
2126
parent_map = self.revisions.get_parent_map(revision_keys)
2128
map(parent_keys.update, parent_map.itervalues())
2129
parent_keys.difference_update(revision_keys)
2130
parent_keys.discard(_mod_revision.NULL_REVISION)
2133
def fileids_altered_by_revision_ids(self, revision_ids, _inv_weave=None):
1128
2134
"""Find the file ids and versions affected by revisions.
1130
2136
:param revisions: an iterable containing revision ids.
2137
:param _inv_weave: The inventory weave from this repository or None.
2138
If None, the inventory weave will be opened automatically.
1131
2139
:return: a dictionary mapping altered file-ids to an iterable of
1132
2140
revision_ids. Each altered file-ids has the exact revision_ids that
1133
2141
altered it listed explicitly.
1135
assert self._serializer.support_altered_by_hack, \
1136
("fileids_altered_by_revision_ids only supported for branches "
1137
"which store inventory as unnested xml, not on %r" % self)
1138
selected_revision_ids = set(revision_ids)
1139
w = self.get_inventory_weave()
1140
pb = ui.ui_factory.nested_progress_bar()
1142
return self._find_file_ids_from_xml_inventory_lines(
1143
w.iter_lines_added_or_present_in_versions(
1144
selected_revision_ids, pb=pb),
1145
selected_revision_ids)
2143
selected_keys = set((revid,) for revid in revision_ids)
2144
w = _inv_weave or self.inventories
2145
return self._find_file_ids_from_xml_inventory_lines(
2146
w.iter_lines_added_or_present_in_keys(
2147
selected_keys, pb=None),
1149
2150
def iter_files_bytes(self, desired_files):
1150
2151
"""Iterate through file versions.
1160
2161
bytes_iterator is an iterable of bytestrings for the file. The
1161
2162
kind of iterable and length of the bytestrings are unspecified, but for
1162
this implementation, it is a list of lines produced by
1163
VersionedFile.get_lines().
2163
this implementation, it is a list of bytes produced by
2164
VersionedFile.get_record_stream().
1165
2166
:param desired_files: a list of (file_id, revision_id, identifier)
1168
transaction = self.get_transaction()
1169
2170
for file_id, revision_id, callable_data in desired_files:
1171
weave = self.weave_store.get_weave(file_id, transaction)
1172
except errors.NoSuchFile:
1173
raise errors.NoSuchIdInRepository(self, file_id)
1174
yield callable_data, weave.get_lines(revision_id)
2171
text_keys[(file_id, revision_id)] = callable_data
2172
for record in self.texts.get_record_stream(text_keys, 'unordered', True):
2173
if record.storage_kind == 'absent':
2174
raise errors.RevisionNotPresent(record.key, self)
2175
yield text_keys[record.key], record.get_bytes_as('chunked')
2177
def _generate_text_key_index(self, text_key_references=None,
2179
"""Generate a new text key index for the repository.
2181
This is an expensive function that will take considerable time to run.
2183
:return: A dict mapping text keys ((file_id, revision_id) tuples) to a
2184
list of parents, also text keys. When a given key has no parents,
2185
the parents list will be [NULL_REVISION].
2187
# All revisions, to find inventory parents.
2188
if ancestors is None:
2189
graph = self.get_graph()
2190
ancestors = graph.get_parent_map(self.all_revision_ids())
2191
if text_key_references is None:
2192
text_key_references = self.find_text_key_references()
2193
pb = ui.ui_factory.nested_progress_bar()
2195
return self._do_generate_text_key_index(ancestors,
2196
text_key_references, pb)
2200
def _do_generate_text_key_index(self, ancestors, text_key_references, pb):
2201
"""Helper for _generate_text_key_index to avoid deep nesting."""
2202
revision_order = tsort.topo_sort(ancestors)
2203
invalid_keys = set()
2205
for revision_id in revision_order:
2206
revision_keys[revision_id] = set()
2207
text_count = len(text_key_references)
2208
# a cache of the text keys to allow reuse; costs a dict of all the
2209
# keys, but saves a 2-tuple for every child of a given key.
2211
for text_key, valid in text_key_references.iteritems():
2213
invalid_keys.add(text_key)
2215
revision_keys[text_key[1]].add(text_key)
2216
text_key_cache[text_key] = text_key
2217
del text_key_references
2219
text_graph = graph.Graph(graph.DictParentsProvider(text_index))
2220
NULL_REVISION = _mod_revision.NULL_REVISION
2221
# Set a cache with a size of 10 - this suffices for bzr.dev but may be
2222
# too small for large or very branchy trees. However, for 55K path
2223
# trees, it would be easy to use too much memory trivially. Ideally we
2224
# could gauge this by looking at available real memory etc, but this is
2225
# always a tricky proposition.
2226
inventory_cache = lru_cache.LRUCache(10)
2227
batch_size = 10 # should be ~150MB on a 55K path tree
2228
batch_count = len(revision_order) / batch_size + 1
2230
pb.update("Calculating text parents", processed_texts, text_count)
2231
for offset in xrange(batch_count):
2232
to_query = revision_order[offset * batch_size:(offset + 1) *
2236
for revision_id in to_query:
2237
parent_ids = ancestors[revision_id]
2238
for text_key in revision_keys[revision_id]:
2239
pb.update("Calculating text parents", processed_texts)
2240
processed_texts += 1
2241
candidate_parents = []
2242
for parent_id in parent_ids:
2243
parent_text_key = (text_key[0], parent_id)
2245
check_parent = parent_text_key not in \
2246
revision_keys[parent_id]
2248
# the parent parent_id is a ghost:
2249
check_parent = False
2250
# truncate the derived graph against this ghost.
2251
parent_text_key = None
2253
# look at the parent commit details inventories to
2254
# determine possible candidates in the per file graph.
2257
inv = inventory_cache[parent_id]
2259
inv = self.revision_tree(parent_id).inventory
2260
inventory_cache[parent_id] = inv
2262
parent_entry = inv[text_key[0]]
2263
except (KeyError, errors.NoSuchId):
2265
if parent_entry is not None:
2267
text_key[0], parent_entry.revision)
2269
parent_text_key = None
2270
if parent_text_key is not None:
2271
candidate_parents.append(
2272
text_key_cache[parent_text_key])
2273
parent_heads = text_graph.heads(candidate_parents)
2274
new_parents = list(parent_heads)
2275
new_parents.sort(key=lambda x:candidate_parents.index(x))
2276
if new_parents == []:
2277
new_parents = [NULL_REVISION]
2278
text_index[text_key] = new_parents
2280
for text_key in invalid_keys:
2281
text_index[text_key] = [NULL_REVISION]
1176
2284
def item_keys_introduced_by(self, revision_ids, _files_pb=None):
1177
2285
"""Get an iterable listing the keys of all the data introduced by a set
1184
2292
versions). knit-kind is one of 'file', 'inventory', 'signatures',
1185
2293
'revisions'. file-id is None unless knit-kind is 'file'.
2295
for result in self._find_file_keys_to_fetch(revision_ids, _files_pb):
2298
for result in self._find_non_file_keys_to_fetch(revision_ids):
2301
def _find_file_keys_to_fetch(self, revision_ids, pb):
1187
2302
# XXX: it's a bit weird to control the inventory weave caching in this
1188
2303
# generator. Ideally the caching would be done in fetch.py I think. Or
1189
2304
# maybe this generator should explicitly have the contract that it
1190
2305
# should not be iterated until the previously yielded item has been
1193
inv_w = self.get_inventory_weave()
1194
inv_w.enable_cache()
2307
inv_w = self.inventories
1196
2309
# file ids that changed
1197
file_ids = self.fileids_altered_by_revision_ids(revision_ids)
2310
file_ids = self.fileids_altered_by_revision_ids(revision_ids, inv_w)
1199
2312
num_file_ids = len(file_ids)
1200
2313
for file_id, altered_versions in file_ids.iteritems():
1201
if _files_pb is not None:
1202
_files_pb.update("fetch texts", count, num_file_ids)
2315
pb.update("Fetch texts", count, num_file_ids)
1204
2317
yield ("file", file_id, altered_versions)
1205
# We're done with the files_pb. Note that it finished by the caller,
1206
# just as it was created by the caller.
2319
def _find_non_file_keys_to_fetch(self, revision_ids):
1210
2321
yield ("inventory", None, revision_ids)
1214
revisions_with_signatures = set()
1215
for rev_id in revision_ids:
1217
self.get_signature_text(rev_id)
1218
except errors.NoSuchRevision:
1222
revisions_with_signatures.add(rev_id)
2324
# XXX: Note ATM no callers actually pay attention to this return
2325
# instead they just use the list of revision ids and ignore
2326
# missing sigs. Consider removing this work entirely
2327
revisions_with_signatures = set(self.signatures.get_parent_map(
2328
[(r,) for r in revision_ids]))
2329
revisions_with_signatures = set(
2330
[r for (r,) in revisions_with_signatures])
2331
revisions_with_signatures.intersection_update(revision_ids)
1224
2332
yield ("signatures", None, revisions_with_signatures)
1227
2335
yield ("revisions", None, revision_ids)
1229
2337
@needs_read_lock
1230
def get_inventory_weave(self):
1231
return self.control_weaves.get_weave('inventory',
1232
self.get_transaction())
1235
2338
def get_inventory(self, revision_id):
1236
"""Get Inventory object by hash."""
1237
return self.deserialise_inventory(
1238
revision_id, self.get_inventory_xml(revision_id))
1240
def deserialise_inventory(self, revision_id, xml):
1241
"""Transform the xml into an inventory object.
2339
"""Get Inventory object by revision id."""
2340
return self.iter_inventories([revision_id]).next()
2342
def iter_inventories(self, revision_ids, ordering=None):
2343
"""Get many inventories by revision_ids.
2345
This will buffer some or all of the texts used in constructing the
2346
inventories in memory, but will only parse a single inventory at a
2349
:param revision_ids: The expected revision ids of the inventories.
2350
:param ordering: optional ordering, e.g. 'topological'. If not
2351
specified, the order of revision_ids will be preserved (by
2352
buffering if necessary).
2353
:return: An iterator of inventories.
2355
if ((None in revision_ids)
2356
or (_mod_revision.NULL_REVISION in revision_ids)):
2357
raise ValueError('cannot get null revision inventory')
2358
return self._iter_inventories(revision_ids, ordering)
2360
def _iter_inventories(self, revision_ids, ordering):
2361
"""single-document based inventory iteration."""
2362
inv_xmls = self._iter_inventory_xmls(revision_ids, ordering)
2363
for text, revision_id in inv_xmls:
2364
yield self._deserialise_inventory(revision_id, text)
2366
def _iter_inventory_xmls(self, revision_ids, ordering):
2367
if ordering is None:
2368
order_as_requested = True
2369
ordering = 'unordered'
2371
order_as_requested = False
2372
keys = [(revision_id,) for revision_id in revision_ids]
2375
if order_as_requested:
2376
key_iter = iter(keys)
2377
next_key = key_iter.next()
2378
stream = self.inventories.get_record_stream(keys, ordering, True)
2380
for record in stream:
2381
if record.storage_kind != 'absent':
2382
chunks = record.get_bytes_as('chunked')
2383
if order_as_requested:
2384
text_chunks[record.key] = chunks
2386
yield ''.join(chunks), record.key[-1]
2388
raise errors.NoSuchRevision(self, record.key)
2389
if order_as_requested:
2390
# Yield as many results as we can while preserving order.
2391
while next_key in text_chunks:
2392
chunks = text_chunks.pop(next_key)
2393
yield ''.join(chunks), next_key[-1]
2395
next_key = key_iter.next()
2396
except StopIteration:
2397
# We still want to fully consume the get_record_stream,
2398
# just in case it is not actually finished at this point
2402
def _deserialise_inventory(self, revision_id, xml):
2403
"""Transform the xml into an inventory object.
1243
2405
:param revision_id: The expected revision id of the inventory.
1244
2406
:param xml: A serialised inventory.
1246
return self._serializer.read_inventory_from_string(xml, revision_id)
1248
def serialise_inventory(self, inv):
1249
return self._serializer.write_inventory_to_string(inv)
1251
def _serialise_inventory_to_lines(self, inv):
1252
return self._serializer.write_inventory_to_lines(inv)
2408
result = self._serializer.read_inventory_from_string(xml, revision_id,
2409
entry_cache=self._inventory_entry_cache,
2410
return_from_cache=self._safe_to_return_from_cache)
2411
if result.revision_id != revision_id:
2412
raise AssertionError('revision id mismatch %s != %s' % (
2413
result.revision_id, revision_id))
1254
2416
def get_serializer_format(self):
1255
2417
return self._serializer.format_num
1257
2419
@needs_read_lock
1258
def get_inventory_xml(self, revision_id):
1259
"""Get inventory XML as a file object."""
2420
def _get_inventory_xml(self, revision_id):
2421
"""Get serialized inventory as a string."""
2422
texts = self._iter_inventory_xmls([revision_id], 'unordered')
1261
assert isinstance(revision_id, str), type(revision_id)
1262
iw = self.get_inventory_weave()
1263
return iw.get_text(revision_id)
2424
text, revision_id = texts.next()
2425
except StopIteration:
1265
2426
raise errors.HistoryMissing(self, 'inventory', revision_id)
1268
def get_inventory_sha1(self, revision_id):
1269
"""Return the sha1 hash of the inventory entry
1271
return self.get_revision(revision_id).inventory_sha1
1274
def get_revision_graph(self, revision_id=None):
1275
"""Return a dictionary containing the revision graph.
1277
NB: This method should not be used as it accesses the entire graph all
1278
at once, which is much more data than most operations should require.
1280
:param revision_id: The revision_id to get a graph from. If None, then
1281
the entire revision graph is returned. This is a deprecated mode of
1282
operation and will be removed in the future.
1283
:return: a dictionary of revision_id->revision_parents_list.
1285
raise NotImplementedError(self.get_revision_graph)
1288
def get_revision_graph_with_ghosts(self, revision_ids=None):
1289
"""Return a graph of the revisions with ghosts marked as applicable.
1291
:param revision_ids: an iterable of revisions to graph or None for all.
1292
:return: a Graph object with the graph reachable from revision_ids.
1294
if 'evil' in debug.debug_flags:
1296
"get_revision_graph_with_ghosts scales with size of history.")
1297
result = deprecated_graph.Graph()
1298
if not revision_ids:
1299
pending = set(self.all_revision_ids())
1302
pending = set(revision_ids)
1303
# special case NULL_REVISION
1304
if _mod_revision.NULL_REVISION in pending:
1305
pending.remove(_mod_revision.NULL_REVISION)
1306
required = set(pending)
1309
revision_id = pending.pop()
1311
rev = self.get_revision(revision_id)
1312
except errors.NoSuchRevision:
1313
if revision_id in required:
1316
result.add_ghost(revision_id)
1318
for parent_id in rev.parent_ids:
1319
# is this queued or done ?
1320
if (parent_id not in pending and
1321
parent_id not in done):
1323
pending.add(parent_id)
1324
result.add_node(revision_id, rev.parent_ids)
1325
done.add(revision_id)
1328
def _get_history_vf(self):
1329
"""Get a versionedfile whose history graph reflects all revisions.
1331
For weave repositories, this is the inventory weave.
1333
return self.get_inventory_weave()
2429
def get_rev_id_for_revno(self, revno, known_pair):
2430
"""Return the revision id of a revno, given a later (revno, revid)
2431
pair in the same history.
2433
:return: if found (True, revid). If the available history ran out
2434
before reaching the revno, then this returns
2435
(False, (closest_revno, closest_revid)).
2437
known_revno, known_revid = known_pair
2438
partial_history = [known_revid]
2439
distance_from_known = known_revno - revno
2440
if distance_from_known < 0:
2442
'requested revno (%d) is later than given known revno (%d)'
2443
% (revno, known_revno))
2446
self, partial_history, stop_index=distance_from_known)
2447
except errors.RevisionNotPresent, err:
2448
if err.revision_id == known_revid:
2449
# The start revision (known_revid) wasn't found.
2451
# This is a stacked repository with no fallbacks, or a there's a
2452
# left-hand ghost. Either way, even though the revision named in
2453
# the error isn't in this repo, we know it's the next step in this
2454
# left-hand history.
2455
partial_history.append(err.revision_id)
2456
if len(partial_history) <= distance_from_known:
2457
# Didn't find enough history to get a revid for the revno.
2458
earliest_revno = known_revno - len(partial_history) + 1
2459
return (False, (earliest_revno, partial_history[-1]))
2460
if len(partial_history) - 1 > distance_from_known:
2461
raise AssertionError('_iter_for_revno returned too much history')
2462
return (True, partial_history[-1])
1335
2464
def iter_reverse_revision_history(self, revision_id):
1336
2465
"""Iterate backwards through revision ids in the lefthand history
1386
2501
for repositories to maintain loaded indices across multiple locks
1387
2502
by checking inside their implementation of this method to see
1388
2503
whether their indices are still valid. This depends of course on
1389
the disk format being validatable in this manner.
2504
the disk format being validatable in this manner. This method is
2505
also called by the refresh_data() public interface to cause a refresh
2506
to occur while in a write lock so that data inserted by a smart server
2507
push operation is visible on the client's instance of the physical
1392
2511
@needs_read_lock
1393
2512
def revision_tree(self, revision_id):
1394
2513
"""Return Tree for a revision on this branch.
1396
`revision_id` may be None for the empty tree revision.
2515
`revision_id` may be NULL_REVISION for the empty tree revision.
2517
revision_id = _mod_revision.ensure_null(revision_id)
1398
2518
# TODO: refactor this to use an existing revision object
1399
2519
# so we don't need to read it in twice.
1400
if revision_id is None or revision_id == _mod_revision.NULL_REVISION:
1401
return RevisionTree(self, Inventory(root_id=None),
2520
if revision_id == _mod_revision.NULL_REVISION:
2521
return RevisionTree(self, Inventory(root_id=None),
1402
2522
_mod_revision.NULL_REVISION)
1404
inv = self.get_revision_inventory(revision_id)
2524
inv = self.get_inventory(revision_id)
1405
2525
return RevisionTree(self, inv, revision_id)
1408
2527
def revision_trees(self, revision_ids):
1409
"""Return Tree for a revision on this branch.
1411
`revision_id` may not be None or 'null:'"""
1412
assert None not in revision_ids
1413
assert _mod_revision.NULL_REVISION not in revision_ids
1414
texts = self.get_inventory_weave().get_texts(revision_ids)
1415
for text, revision_id in zip(texts, revision_ids):
1416
inv = self.deserialise_inventory(revision_id, text)
1417
yield RevisionTree(self, inv, revision_id)
2528
"""Return Trees for revisions in this repository.
2530
:param revision_ids: a sequence of revision-ids;
2531
a revision-id may not be None or 'null:'
2533
inventories = self.iter_inventories(revision_ids)
2534
for inv in inventories:
2535
yield RevisionTree(self, inv, inv.revision_id)
2537
def _filtered_revision_trees(self, revision_ids, file_ids):
2538
"""Return Tree for a revision on this branch with only some files.
2540
:param revision_ids: a sequence of revision-ids;
2541
a revision-id may not be None or 'null:'
2542
:param file_ids: if not None, the result is filtered
2543
so that only those file-ids, their parents and their
2544
children are included.
2546
inventories = self.iter_inventories(revision_ids)
2547
for inv in inventories:
2548
# Should we introduce a FilteredRevisionTree class rather
2549
# than pre-filter the inventory here?
2550
filtered_inv = inv.filter(file_ids)
2551
yield RevisionTree(self, filtered_inv, filtered_inv.revision_id)
1419
2553
@needs_read_lock
1420
2554
def get_ancestry(self, revision_id, topo_sorted=True):
1421
2555
"""Return a list of revision-ids integrated by a revision.
1423
The first element of the list is always None, indicating the origin
1424
revision. This might change when we have history horizons, or
2557
The first element of the list is always None, indicating the origin
2558
revision. This might change when we have history horizons, or
1425
2559
perhaps we should have a new API.
1427
2561
This is topologically sorted.
1429
2563
if _mod_revision.is_null(revision_id):
1431
2565
if not self.has_revision(revision_id):
1432
2566
raise errors.NoSuchRevision(self, revision_id)
1433
w = self.get_inventory_weave()
1434
candidates = w.get_ancestry(revision_id, topo_sorted)
1435
return [None] + candidates # self._eliminate_revisions_not_present(candidates)
2567
graph = self.get_graph()
2569
search = graph._make_breadth_first_searcher([revision_id])
2572
found, ghosts = search.next_with_ghosts()
2573
except StopIteration:
2576
if _mod_revision.NULL_REVISION in keys:
2577
keys.remove(_mod_revision.NULL_REVISION)
2579
parent_map = graph.get_parent_map(keys)
2580
keys = tsort.topo_sort(parent_map)
2581
return [None] + list(keys)
2583
def pack(self, hint=None):
1438
2584
"""Compress the data within the repository.
1440
2586
This operation only makes sense for some repository types. For other
1441
2587
types it should be a no-op that just returns.
1443
2589
This stub method does not require a lock, but subclasses should use
1444
@needs_write_lock as this is a long running call its reasonable to
2590
@needs_write_lock as this is a long running call its reasonable to
1445
2591
implicitly lock for the user.
1449
def print_file(self, file, revision_id):
1450
"""Print `file` to stdout.
1452
FIXME RBC 20060125 as John Meinel points out this is a bad api
1453
- it writes to stdout, it assumes that that is valid etc. Fix
1454
by creating a new more flexible convenience function.
2593
:param hint: If not supplied, the whole repository is packed.
2594
If supplied, the repository may use the hint parameter as a
2595
hint for the parts of the repository to pack. A hint can be
2596
obtained from the result of commit_write_group(). Out of
2597
date hints are simply ignored, because concurrent operations
2598
can obsolete them rapidly.
1456
tree = self.revision_tree(revision_id)
1457
# use inventory as it was in that revision
1458
file_id = tree.inventory.path2id(file)
1460
# TODO: jam 20060427 Write a test for this code path
1461
# it had a bug in it, and was raising the wrong
1463
raise errors.BzrError("%r is not present in revision %s" % (file, revision_id))
1464
tree.print_file(file_id)
1466
2601
def get_transaction(self):
1467
2602
return self.control_files.get_transaction()
1469
def revision_parents(self, revision_id):
1470
return self.get_inventory_weave().parent_names(revision_id)
1472
def get_parents(self, revision_ids):
1473
"""See StackedParentsProvider.get_parents"""
2604
def get_parent_map(self, revision_ids):
2605
"""See graph.StackedParentsProvider.get_parent_map"""
2606
# revisions index works in keys; this just works in revisions
2607
# therefore wrap and unwrap
1475
2610
for revision_id in revision_ids:
1476
2611
if revision_id == _mod_revision.NULL_REVISION:
1480
parents = self.get_revision(revision_id).parent_ids
1481
except errors.NoSuchRevision:
1484
if len(parents) == 0:
1485
parents = [_mod_revision.NULL_REVISION]
1486
parents_list.append(parents)
2612
result[revision_id] = ()
2613
elif revision_id is None:
2614
raise ValueError('get_parent_map(None) is not valid')
2616
query_keys.append((revision_id ,))
2617
for ((revision_id,), parent_keys) in \
2618
self.revisions.get_parent_map(query_keys).iteritems():
2620
result[revision_id] = tuple([parent_revid
2621
for (parent_revid,) in parent_keys])
2623
result[revision_id] = (_mod_revision.NULL_REVISION,)
1489
2626
def _make_parents_provider(self):
2007
3339
InterRepository.get(other).method_name(parameters).
3342
_walk_to_common_revisions_batch_size = 50
2010
3343
_optimisers = []
2011
3344
"""The available optimised InterRepository types."""
2013
3347
def copy_content(self, revision_id=None):
2014
raise NotImplementedError(self.copy_content)
2016
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3348
"""Make a complete copy of the content in self into destination.
3350
This is a destructive operation! Do not use it on existing
3353
:param revision_id: Only copy the content needed to construct
3354
revision_id and its parents.
3357
self.target.set_make_working_trees(self.source.make_working_trees())
3358
except NotImplementedError:
3360
self.target.fetch(self.source, revision_id=revision_id)
3363
def fetch(self, revision_id=None, pb=None, find_ghosts=False,
2017
3365
"""Fetch the content required to construct revision_id.
2019
3367
The content is copied from self.source to self.target.
2021
3369
:param revision_id: if None all content is copied, if NULL_REVISION no
2022
3370
content is copied.
2023
:param pb: optional progress bar to use for progress reports. If not
2024
provided a default one will be created.
2026
Returns the copied revision count and the failed revisions in a tuple:
2029
raise NotImplementedError(self.fetch)
3374
ui.ui_factory.warn_experimental_format_fetch(self)
3375
f = _mod_fetch.RepoFetcher(to_repository=self.target,
3376
from_repository=self.source,
3377
last_revision=revision_id,
3378
fetch_spec=fetch_spec,
3379
find_ghosts=find_ghosts)
3381
def _walk_to_common_revisions(self, revision_ids):
3382
"""Walk out from revision_ids in source to revisions target has.
3384
:param revision_ids: The start point for the search.
3385
:return: A set of revision ids.
3387
target_graph = self.target.get_graph()
3388
revision_ids = frozenset(revision_ids)
3389
missing_revs = set()
3390
source_graph = self.source.get_graph()
3391
# ensure we don't pay silly lookup costs.
3392
searcher = source_graph._make_breadth_first_searcher(revision_ids)
3393
null_set = frozenset([_mod_revision.NULL_REVISION])
3394
searcher_exhausted = False
3398
# Iterate the searcher until we have enough next_revs
3399
while len(next_revs) < self._walk_to_common_revisions_batch_size:
3401
next_revs_part, ghosts_part = searcher.next_with_ghosts()
3402
next_revs.update(next_revs_part)
3403
ghosts.update(ghosts_part)
3404
except StopIteration:
3405
searcher_exhausted = True
3407
# If there are ghosts in the source graph, and the caller asked for
3408
# them, make sure that they are present in the target.
3409
# We don't care about other ghosts as we can't fetch them and
3410
# haven't been asked to.
3411
ghosts_to_check = set(revision_ids.intersection(ghosts))
3412
revs_to_get = set(next_revs).union(ghosts_to_check)
3414
have_revs = set(target_graph.get_parent_map(revs_to_get))
3415
# we always have NULL_REVISION present.
3416
have_revs = have_revs.union(null_set)
3417
# Check if the target is missing any ghosts we need.
3418
ghosts_to_check.difference_update(have_revs)
3420
# One of the caller's revision_ids is a ghost in both the
3421
# source and the target.
3422
raise errors.NoSuchRevision(
3423
self.source, ghosts_to_check.pop())
3424
missing_revs.update(next_revs - have_revs)
3425
# Because we may have walked past the original stop point, make
3426
# sure everything is stopped
3427
stop_revs = searcher.find_seen_ancestors(have_revs)
3428
searcher.stop_searching_any(stop_revs)
3429
if searcher_exhausted:
3431
return searcher.get_result()
2031
3433
@needs_read_lock
2032
def missing_revision_ids(self, revision_id=None):
3434
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2033
3435
"""Return the revision ids that source has that target does not.
2035
These are returned in topological order.
2037
3437
:param revision_id: only return revision ids included by this
3439
:param find_ghosts: If True find missing revisions in deep history
3440
rather than just finding the surface difference.
3441
:return: A bzrlib.graph.SearchResult.
3443
# stop searching at found target revisions.
3444
if not find_ghosts and revision_id is not None:
3445
return self._walk_to_common_revisions([revision_id])
2040
3446
# generic, possibly worst case, slow code path.
2041
3447
target_ids = set(self.target.all_revision_ids())
2042
3448
if revision_id is not None:
2043
3449
source_ids = self.source.get_ancestry(revision_id)
2044
assert source_ids[0] is None
3450
if source_ids[0] is not None:
3451
raise AssertionError()
2045
3452
source_ids.pop(0)
2047
3454
source_ids = self.source.all_revision_ids()
2048
3455
result_set = set(source_ids).difference(target_ids)
2049
# this may look like a no-op: its not. It preserves the ordering
2050
# other_ids had while only returning the members from other_ids
2051
# that we've decided we need.
2052
return [rev_id for rev_id in source_ids if rev_id in result_set]
3456
return self.source.revision_ids_to_search_result(result_set)
2055
3459
def _same_model(source, target):
2056
"""True if source and target have the same data representation."""
3460
"""True if source and target have the same data representation.
3462
Note: this is always called on the base class; overriding it in a
3463
subclass will have no effect.
3466
InterRepository._assert_same_model(source, target)
3468
except errors.IncompatibleRepositories, e:
3472
def _assert_same_model(source, target):
3473
"""Raise an exception if two repositories do not use the same model.
2057
3475
if source.supports_rich_root() != target.supports_rich_root():
3476
raise errors.IncompatibleRepositories(source, target,
3477
"different rich-root support")
2059
3478
if source._serializer != target._serializer:
3479
raise errors.IncompatibleRepositories(source, target,
3480
"different serializers")
2064
3483
class InterSameDataRepository(InterRepository):
2065
3484
"""Code for converting between repositories that represent the same data.
2067
3486
Data format and model must match for this to work.
2071
3490
def _get_repo_format_to_test(self):
2072
3491
"""Repository format for testing with.
2074
3493
InterSameData can pull from subtree to subtree and from non-subtree to
2075
3494
non-subtree, so we test this with the richest repository format.
2289
3654
# we do not have a revision as that would be pointless.
2290
3655
target_ids = set(self.target.all_revision_ids())
2291
3656
possibly_present_revisions = target_ids.intersection(source_ids_set)
2292
actually_present_revisions = set(self.target._eliminate_revisions_not_present(possibly_present_revisions))
3657
actually_present_revisions = set(
3658
self.target._eliminate_revisions_not_present(possibly_present_revisions))
2293
3659
required_revisions = source_ids_set.difference(actually_present_revisions)
2294
required_topo_revisions = [rev_id for rev_id in source_ids if rev_id in required_revisions]
2295
3660
if revision_id is not None:
2296
3661
# we used get_ancestry to determine source_ids then we are assured all
2297
3662
# revisions referenced are present as they are installed in topological order.
2298
3663
# and the tip revision was validated by get_ancestry.
2299
return required_topo_revisions
3664
result_set = required_revisions
2301
# if we just grabbed the possibly available ids, then
3666
# if we just grabbed the possibly available ids, then
2302
3667
# we only have an estimate of whats available and need to validate
2303
3668
# that against the revision records.
2304
return self.source._eliminate_revisions_not_present(required_topo_revisions)
2307
class InterPackRepo(InterSameDataRepository):
2308
"""Optimised code paths between Pack based repositories."""
3670
self.source._eliminate_revisions_not_present(required_revisions))
3671
return self.source.revision_ids_to_search_result(result_set)
3674
class InterDifferingSerializer(InterRepository):
2311
3677
def _get_repo_format_to_test(self):
2312
from bzrlib.repofmt import pack_repo
2313
return pack_repo.RepositoryFormatKnitPack1()
2316
3681
def is_compatible(source, target):
2317
"""Be compatible with known Pack formats.
2319
We don't test for the stores being of specific types because that
2320
could lead to confusing results, and there is no need to be
2323
from bzrlib.repofmt.pack_repo import RepositoryFormatPack
2325
are_packs = (isinstance(source._format, RepositoryFormatPack) and
2326
isinstance(target._format, RepositoryFormatPack))
2327
except AttributeError:
2329
return are_packs and InterRepository._same_model(source, target)
2332
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2333
"""See InterRepository.fetch()."""
2334
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2335
self.source, self.source._format, self.target, self.target._format)
2336
self.count_copied = 0
2337
if revision_id is None:
2339
# everything to do - use pack logic
2340
# to fetch from all packs to one without
2341
# inventory parsing etc, IFF nothing to be copied is in the target.
2343
revision_ids = self.source.all_revision_ids()
2344
# implementing the TODO will involve:
2345
# - detecting when all of a pack is selected
2346
# - avoiding as much as possible pre-selection, so the
2347
# more-core routines such as create_pack_from_packs can filter in
2348
# a just-in-time fashion. (though having a HEADS list on a
2349
# repository might make this a lot easier, because we could
2350
# sensibly detect 'new revisions' without doing a full index scan.
2351
elif _mod_revision.is_null(revision_id):
3682
"""Be compatible with Knit2 source and Knit3 target"""
3683
# This is redundant with format.check_conversion_target(), however that
3684
# raises an exception, and we just want to say "False" as in we won't
3685
# support converting between these formats.
3686
if 'IDS_never' in debug.debug_flags:
3688
if source.supports_rich_root() and not target.supports_rich_root():
3690
if (source._format.supports_tree_reference
3691
and not target._format.supports_tree_reference):
3693
if target._fallback_repositories and target._format.supports_chks:
3694
# IDS doesn't know how to copy CHKs for the parent inventories it
3695
# adds to stacked repos.
3697
if 'IDS_always' in debug.debug_flags:
3699
# Only use this code path for local source and target. IDS does far
3700
# too much IO (both bandwidth and roundtrips) over a network.
3701
if not source.bzrdir.transport.base.startswith('file:///'):
3703
if not target.bzrdir.transport.base.startswith('file:///'):
3707
def _get_trees(self, revision_ids, cache):
3709
for rev_id in revision_ids:
3711
possible_trees.append((rev_id, cache[rev_id]))
3713
# Not cached, but inventory might be present anyway.
3715
tree = self.source.revision_tree(rev_id)
3716
except errors.NoSuchRevision:
3717
# Nope, parent is ghost.
3720
cache[rev_id] = tree
3721
possible_trees.append((rev_id, tree))
3722
return possible_trees
3724
def _get_delta_for_revision(self, tree, parent_ids, possible_trees):
3725
"""Get the best delta and base for this revision.
3727
:return: (basis_id, delta)
3730
# Generate deltas against each tree, to find the shortest.
3731
texts_possibly_new_in_tree = set()
3732
for basis_id, basis_tree in possible_trees:
3733
delta = tree.inventory._make_delta(basis_tree.inventory)
3734
for old_path, new_path, file_id, new_entry in delta:
3735
if new_path is None:
3736
# This file_id isn't present in the new rev, so we don't
3740
# Rich roots are handled elsewhere...
3742
kind = new_entry.kind
3743
if kind != 'directory' and kind != 'file':
3744
# No text record associated with this inventory entry.
3746
# This is a directory or file that has changed somehow.
3747
texts_possibly_new_in_tree.add((file_id, new_entry.revision))
3748
deltas.append((len(delta), basis_id, delta))
3750
return deltas[0][1:]
3752
def _fetch_parent_invs_for_stacking(self, parent_map, cache):
3753
"""Find all parent revisions that are absent, but for which the
3754
inventory is present, and copy those inventories.
3756
This is necessary to preserve correctness when the source is stacked
3757
without fallbacks configured. (Note that in cases like upgrade the
3758
source may be not have _fallback_repositories even though it is
3762
for parents in parent_map.values():
3763
parent_revs.update(parents)
3764
present_parents = self.source.get_parent_map(parent_revs)
3765
absent_parents = set(parent_revs).difference(present_parents)
3766
parent_invs_keys_for_stacking = self.source.inventories.get_parent_map(
3767
(rev_id,) for rev_id in absent_parents)
3768
parent_inv_ids = [key[-1] for key in parent_invs_keys_for_stacking]
3769
for parent_tree in self.source.revision_trees(parent_inv_ids):
3770
current_revision_id = parent_tree.get_revision_id()
3771
parents_parents_keys = parent_invs_keys_for_stacking[
3772
(current_revision_id,)]
3773
parents_parents = [key[-1] for key in parents_parents_keys]
3774
basis_id = _mod_revision.NULL_REVISION
3775
basis_tree = self.source.revision_tree(basis_id)
3776
delta = parent_tree.inventory._make_delta(basis_tree.inventory)
3777
self.target.add_inventory_by_delta(
3778
basis_id, delta, current_revision_id, parents_parents)
3779
cache[current_revision_id] = parent_tree
3781
def _fetch_batch(self, revision_ids, basis_id, cache, a_graph=None):
3782
"""Fetch across a few revisions.
3784
:param revision_ids: The revisions to copy
3785
:param basis_id: The revision_id of a tree that must be in cache, used
3786
as a basis for delta when no other base is available
3787
:param cache: A cache of RevisionTrees that we can use.
3788
:param a_graph: A Graph object to determine the heads() of the
3789
rich-root data stream.
3790
:return: The revision_id of the last converted tree. The RevisionTree
3791
for it will be in cache
3793
# Walk though all revisions; get inventory deltas, copy referenced
3794
# texts that delta references, insert the delta, revision and
3796
root_keys_to_create = set()
3799
pending_revisions = []
3800
parent_map = self.source.get_parent_map(revision_ids)
3801
self._fetch_parent_invs_for_stacking(parent_map, cache)
3802
self.source._safe_to_return_from_cache = True
3803
for tree in self.source.revision_trees(revision_ids):
3804
# Find a inventory delta for this revision.
3805
# Find text entries that need to be copied, too.
3806
current_revision_id = tree.get_revision_id()
3807
parent_ids = parent_map.get(current_revision_id, ())
3808
parent_trees = self._get_trees(parent_ids, cache)
3809
possible_trees = list(parent_trees)
3810
if len(possible_trees) == 0:
3811
# There either aren't any parents, or the parents are ghosts,
3812
# so just use the last converted tree.
3813
possible_trees.append((basis_id, cache[basis_id]))
3814
basis_id, delta = self._get_delta_for_revision(tree, parent_ids,
3816
revision = self.source.get_revision(current_revision_id)
3817
pending_deltas.append((basis_id, delta,
3818
current_revision_id, revision.parent_ids))
3819
if self._converting_to_rich_root:
3820
self._revision_id_to_root_id[current_revision_id] = \
3822
# Determine which texts are in present in this revision but not in
3823
# any of the available parents.
3824
texts_possibly_new_in_tree = set()
3825
for old_path, new_path, file_id, entry in delta:
3826
if new_path is None:
3827
# This file_id isn't present in the new rev
3831
if not self.target.supports_rich_root():
3832
# The target doesn't support rich root, so we don't
3835
if self._converting_to_rich_root:
3836
# This can't be copied normally, we have to insert
3838
root_keys_to_create.add((file_id, entry.revision))
3841
texts_possibly_new_in_tree.add((file_id, entry.revision))
3842
for basis_id, basis_tree in possible_trees:
3843
basis_inv = basis_tree.inventory
3844
for file_key in list(texts_possibly_new_in_tree):
3845
file_id, file_revision = file_key
3847
entry = basis_inv[file_id]
3848
except errors.NoSuchId:
3850
if entry.revision == file_revision:
3851
texts_possibly_new_in_tree.remove(file_key)
3852
text_keys.update(texts_possibly_new_in_tree)
3853
pending_revisions.append(revision)
3854
cache[current_revision_id] = tree
3855
basis_id = current_revision_id
3856
self.source._safe_to_return_from_cache = False
3858
from_texts = self.source.texts
3859
to_texts = self.target.texts
3860
if root_keys_to_create:
3861
root_stream = _mod_fetch._new_root_data_stream(
3862
root_keys_to_create, self._revision_id_to_root_id, parent_map,
3863
self.source, graph=a_graph)
3864
to_texts.insert_record_stream(root_stream)
3865
to_texts.insert_record_stream(from_texts.get_record_stream(
3866
text_keys, self.target._format._fetch_order,
3867
not self.target._format._fetch_uses_deltas))
3868
# insert inventory deltas
3869
for delta in pending_deltas:
3870
self.target.add_inventory_by_delta(*delta)
3871
if self.target._fallback_repositories:
3872
# Make sure this stacked repository has all the parent inventories
3873
# for the new revisions that we are about to insert. We do this
3874
# before adding the revisions so that no revision is added until
3875
# all the inventories it may depend on are added.
3876
# Note that this is overzealous, as we may have fetched these in an
3879
revision_ids = set()
3880
for revision in pending_revisions:
3881
revision_ids.add(revision.revision_id)
3882
parent_ids.update(revision.parent_ids)
3883
parent_ids.difference_update(revision_ids)
3884
parent_ids.discard(_mod_revision.NULL_REVISION)
3885
parent_map = self.source.get_parent_map(parent_ids)
3886
# we iterate over parent_map and not parent_ids because we don't
3887
# want to try copying any revision which is a ghost
3888
for parent_tree in self.source.revision_trees(parent_map):
3889
current_revision_id = parent_tree.get_revision_id()
3890
parents_parents = parent_map[current_revision_id]
3891
possible_trees = self._get_trees(parents_parents, cache)
3892
if len(possible_trees) == 0:
3893
# There either aren't any parents, or the parents are
3894
# ghosts, so just use the last converted tree.
3895
possible_trees.append((basis_id, cache[basis_id]))
3896
basis_id, delta = self._get_delta_for_revision(parent_tree,
3897
parents_parents, possible_trees)
3898
self.target.add_inventory_by_delta(
3899
basis_id, delta, current_revision_id, parents_parents)
3900
# insert signatures and revisions
3901
for revision in pending_revisions:
2356
revision_ids = self.missing_revision_ids(revision_id,
2357
find_ghosts=find_ghosts)
3903
signature = self.source.get_signature_text(
3904
revision.revision_id)
3905
self.target.add_signature_text(revision.revision_id,
2358
3907
except errors.NoSuchRevision:
2359
raise errors.InstallFailed([revision_id])
2360
packs = self.source._pack_collection.all_packs()
2361
pack = self.target._pack_collection.create_pack_from_packs(
2362
packs, '.fetch', revision_ids,
2364
if pack is not None:
2365
self.target._pack_collection._save_pack_names()
2366
# Trigger an autopack. This may duplicate effort as we've just done
2367
# a pack creation, but for now it is simpler to think about as
2368
# 'upload data, then repack if needed'.
2369
self.target._pack_collection.autopack()
2370
return pack.get_revision_count()
2375
def missing_revision_ids(self, revision_id=None, find_ghosts=True):
2376
"""See InterRepository.missing_revision_ids().
2378
:param find_ghosts: Find ghosts throughough the ancestry of
2381
if not find_ghosts and revision_id is not None:
2382
graph = self.source.get_graph()
2383
missing_revs = set()
2384
searcher = graph._make_breadth_first_searcher([revision_id])
2386
self.target._pack_collection.revision_index.combined_index
2387
null_set = frozenset([_mod_revision.NULL_REVISION])
2390
next_revs = set(searcher.next())
2391
except StopIteration:
2393
next_revs.difference_update(null_set)
2394
target_keys = [(key,) for key in next_revs]
2395
have_revs = frozenset(node[1][0] for node in
2396
target_index.iter_entries(target_keys))
2397
missing_revs.update(next_revs - have_revs)
2398
searcher.stop_searching_any(have_revs)
2400
elif revision_id is not None:
2401
source_ids = self.source.get_ancestry(revision_id)
2402
assert source_ids[0] is None
2405
source_ids = self.source.all_revision_ids()
2406
# source_ids is the worst possible case we may need to pull.
2407
# now we want to filter source_ids against what we actually
2408
# have in target, but don't try to check for existence where we know
2409
# we do not have a revision as that would be pointless.
2410
target_ids = set(self.target.all_revision_ids())
2411
return [r for r in source_ids if (r not in target_ids)]
2414
class InterModel1and2(InterRepository):
2417
def _get_repo_format_to_test(self):
2421
def is_compatible(source, target):
2422
if not source.supports_rich_root() and target.supports_rich_root():
2428
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2429
"""See InterRepository.fetch()."""
2430
from bzrlib.fetch import Model1toKnit2Fetcher
2431
f = Model1toKnit2Fetcher(to_repository=self.target,
2432
from_repository=self.source,
2433
last_revision=revision_id,
2435
return f.count_copied, f.failed_revisions
2438
def copy_content(self, revision_id=None):
2439
"""Make a complete copy of the content in self into destination.
2441
This is a destructive operation! Do not use it on existing
2444
:param revision_id: Only copy the content needed to construct
2445
revision_id and its parents.
2448
self.target.set_make_working_trees(self.source.make_working_trees())
2449
except NotImplementedError:
2451
# but don't bother fetching if we have the needed data now.
2452
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
2453
self.target.has_revision(revision_id)):
2455
self.target.fetch(self.source, revision_id=revision_id)
2458
class InterKnit1and2(InterKnitRepo):
2461
def _get_repo_format_to_test(self):
2465
def is_compatible(source, target):
2466
"""Be compatible with Knit1 source and Knit3 target"""
2467
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit3
2469
from bzrlib.repofmt.knitrepo import (RepositoryFormatKnit1,
2470
RepositoryFormatKnit3)
2471
from bzrlib.repofmt.pack_repo import (RepositoryFormatKnitPack1,
2472
RepositoryFormatKnitPack3)
2473
return (isinstance(source._format,
2474
(RepositoryFormatKnit1, RepositoryFormatKnitPack1)) and
2475
isinstance(target._format,
2476
(RepositoryFormatKnit3, RepositoryFormatKnitPack3))
2478
except AttributeError:
2482
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2483
"""See InterRepository.fetch()."""
2484
from bzrlib.fetch import Knit1to2Fetcher
2485
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2486
self.source, self.source._format, self.target,
2487
self.target._format)
2488
f = Knit1to2Fetcher(to_repository=self.target,
2489
from_repository=self.source,
2490
last_revision=revision_id,
2492
return f.count_copied, f.failed_revisions
2495
class InterRemoteToOther(InterRepository):
2497
def __init__(self, source, target):
2498
InterRepository.__init__(self, source, target)
2499
self._real_inter = None
2502
def is_compatible(source, target):
2503
if not isinstance(source, remote.RemoteRepository):
2505
source._ensure_real()
2506
real_source = source._real_repository
2507
# Is source's model compatible with target's model, and are they the
2508
# same format? Currently we can only optimise fetching from an
2509
# identical model & format repo.
2510
assert not isinstance(real_source, remote.RemoteRepository), (
2511
"We don't support remote repos backed by remote repos yet.")
2512
return real_source._format == target._format
2515
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2516
"""See InterRepository.fetch()."""
2517
from bzrlib.fetch import RemoteToOtherFetcher
2518
mutter("Using fetch logic to copy between %s(remote) and %s(%s)",
2519
self.source, self.target, self.target._format)
2520
# TODO: jam 20070210 This should be an assert, not a translate
2521
revision_id = osutils.safe_revision_id(revision_id)
2522
f = RemoteToOtherFetcher(to_repository=self.target,
2523
from_repository=self.source,
2524
last_revision=revision_id,
2526
return f.count_copied, f.failed_revisions
2529
def _get_repo_format_to_test(self):
2533
class InterOtherToRemote(InterRepository):
2535
def __init__(self, source, target):
2536
InterRepository.__init__(self, source, target)
2537
self._real_inter = None
2540
def is_compatible(source, target):
2541
if isinstance(target, remote.RemoteRepository):
2545
def _ensure_real_inter(self):
2546
if self._real_inter is None:
2547
self.target._ensure_real()
2548
real_target = self.target._real_repository
2549
self._real_inter = InterRepository.get(self.source, real_target)
2551
def copy_content(self, revision_id=None):
2552
self._ensure_real_inter()
2553
self._real_inter.copy_content(revision_id=revision_id)
2555
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2556
self._ensure_real_inter()
2557
self._real_inter.fetch(revision_id=revision_id, pb=pb)
2560
def _get_repo_format_to_test(self):
3909
self.target.add_revision(revision.revision_id, revision)
3912
def _fetch_all_revisions(self, revision_ids, pb):
3913
"""Fetch everything for the list of revisions.
3915
:param revision_ids: The list of revisions to fetch. Must be in
3917
:param pb: A ProgressTask
3920
basis_id, basis_tree = self._get_basis(revision_ids[0])
3922
cache = lru_cache.LRUCache(100)
3923
cache[basis_id] = basis_tree
3924
del basis_tree # We don't want to hang on to it here
3926
if self._converting_to_rich_root and len(revision_ids) > 100:
3927
a_graph = _mod_fetch._get_rich_root_heads_graph(self.source,
3932
for offset in range(0, len(revision_ids), batch_size):
3933
self.target.start_write_group()
3935
pb.update('Transferring revisions', offset,
3937
batch = revision_ids[offset:offset+batch_size]
3938
basis_id = self._fetch_batch(batch, basis_id, cache,
3941
self.source._safe_to_return_from_cache = False
3942
self.target.abort_write_group()
3945
hint = self.target.commit_write_group()
3948
if hints and self.target._format.pack_compresses:
3949
self.target.pack(hint=hints)
3950
pb.update('Transferring revisions', len(revision_ids),
3954
def fetch(self, revision_id=None, pb=None, find_ghosts=False,
3956
"""See InterRepository.fetch()."""
3957
if fetch_spec is not None:
3958
raise AssertionError("Not implemented yet...")
3959
# See <https://launchpad.net/bugs/456077> asking for a warning here
3961
# nb this is only active for local-local fetches; other things using
3963
ui.ui_factory.warn_cross_format_fetch(self.source._format,
3964
self.target._format)
3965
ui.ui_factory.warn_experimental_format_fetch(self)
3966
if (not self.source.supports_rich_root()
3967
and self.target.supports_rich_root()):
3968
self._converting_to_rich_root = True
3969
self._revision_id_to_root_id = {}
3971
self._converting_to_rich_root = False
3972
revision_ids = self.target.search_missing_revision_ids(self.source,
3973
revision_id, find_ghosts=find_ghosts).get_keys()
3974
if not revision_ids:
3976
revision_ids = tsort.topo_sort(
3977
self.source.get_graph().get_parent_map(revision_ids))
3978
if not revision_ids:
3980
# Walk though all revisions; get inventory deltas, copy referenced
3981
# texts that delta references, insert the delta, revision and
3984
my_pb = ui.ui_factory.nested_progress_bar()
3987
symbol_versioning.warn(
3988
symbol_versioning.deprecated_in((1, 14, 0))
3989
% "pb parameter to fetch()")
3992
self._fetch_all_revisions(revision_ids, pb)
3994
if my_pb is not None:
3996
return len(revision_ids), 0
3998
def _get_basis(self, first_revision_id):
3999
"""Get a revision and tree which exists in the target.
4001
This assumes that first_revision_id is selected for transmission
4002
because all other ancestors are already present. If we can't find an
4003
ancestor we fall back to NULL_REVISION since we know that is safe.
4005
:return: (basis_id, basis_tree)
4007
first_rev = self.source.get_revision(first_revision_id)
4009
basis_id = first_rev.parent_ids[0]
4010
# only valid as a basis if the target has it
4011
self.target.get_revision(basis_id)
4012
# Try to get a basis tree - if its a ghost it will hit the
4013
# NoSuchRevision case.
4014
basis_tree = self.source.revision_tree(basis_id)
4015
except (IndexError, errors.NoSuchRevision):
4016
basis_id = _mod_revision.NULL_REVISION
4017
basis_tree = self.source.revision_tree(basis_id)
4018
return basis_id, basis_tree
4021
InterRepository.register_optimiser(InterDifferingSerializer)
2564
4022
InterRepository.register_optimiser(InterSameDataRepository)
2565
4023
InterRepository.register_optimiser(InterWeaveRepo)
2566
4024
InterRepository.register_optimiser(InterKnitRepo)
2567
InterRepository.register_optimiser(InterModel1and2)
2568
InterRepository.register_optimiser(InterKnit1and2)
2569
InterRepository.register_optimiser(InterPackRepo)
2570
InterRepository.register_optimiser(InterRemoteToOther)
2571
InterRepository.register_optimiser(InterOtherToRemote)
2574
4027
class CopyConverter(object):
2575
4028
"""A repository conversion tool which just performs a copy of the content.
2577
4030
This is slow but quite reliable.
2653
4102
return _unescape_re.sub(_unescaper, data)
2656
class _RevisionTextVersionCache(object):
2657
"""A cache of the versionedfile versions for revision and file-id."""
2659
def __init__(self, repository):
2660
self.repository = repository
2661
self.revision_versions = {}
2662
self.revision_parents = {}
2663
self.repo_graph = self.repository.get_graph()
2664
# XXX: RBC: I haven't tracked down what uses this, but it would be
2665
# better to use the headscache directly I think.
2666
self.heads = graph.HeadsCache(self.repo_graph).heads
2668
def add_revision_text_versions(self, tree):
2669
"""Cache text version data from the supplied revision tree"""
2671
for path, entry in tree.iter_entries_by_dir():
2672
inv_revisions[entry.file_id] = entry.revision
2673
self.revision_versions[tree.get_revision_id()] = inv_revisions
2674
return inv_revisions
2676
def get_text_version(self, file_id, revision_id):
2677
"""Determine the text version for a given file-id and revision-id"""
2679
inv_revisions = self.revision_versions[revision_id]
2682
tree = self.repository.revision_tree(revision_id)
2683
except errors.RevisionNotPresent:
2684
self.revision_versions[revision_id] = inv_revisions = {}
2686
inv_revisions = self.add_revision_text_versions(tree)
2687
return inv_revisions.get(file_id)
2689
def prepopulate_revs(self, revision_ids):
2690
# Filter out versions that we don't have an inventory for, so that the
2691
# revision_trees() call won't fail.
2692
inv_weave = self.repository.get_inventory_weave()
2693
revs = [r for r in revision_ids if inv_weave.has_version(r)]
2694
# XXX: this loop is very similar to
2695
# bzrlib.fetch.Inter1and2Helper.iter_rev_trees.
2697
mutter('%d revisions left to prepopulate', len(revs))
2698
for tree in self.repository.revision_trees(revs[:100]):
2699
if tree.inventory.revision_id is None:
2700
tree.inventory.revision_id = tree.get_revision_id()
2701
self.add_revision_text_versions(tree)
2704
def get_parents(self, revision_id):
2706
return self.revision_parents[revision_id]
2708
parents = self.repository.get_parents([revision_id])[0]
2709
self.revision_parents[revision_id] = parents
2712
def used_file_versions(self):
2713
"""Return a set of (revision_id, file_id) pairs for each file version
2714
referenced by any inventory cached by this _RevisionTextVersionCache.
2716
If the entire repository has been cached, this can be used to find all
2717
file versions that are actually referenced by inventories. Thus any
2718
other file version is completely unused and can be removed safely.
2721
for inventory_summary in self.revision_versions.itervalues():
2722
result.update(inventory_summary.items())
2726
class VersionedFileChecker(object):
2728
def __init__(self, planned_revisions, revision_versions, repository):
2729
self.planned_revisions = planned_revisions
2730
self.revision_versions = revision_versions
2731
self.repository = repository
2733
def calculate_file_version_parents(self, revision_id, file_id):
4105
class _VersionedFileChecker(object):
4107
def __init__(self, repository, text_key_references=None, ancestors=None):
4108
self.repository = repository
4109
self.text_index = self.repository._generate_text_key_index(
4110
text_key_references=text_key_references, ancestors=ancestors)
4112
def calculate_file_version_parents(self, text_key):
2734
4113
"""Calculate the correct parents for a file version according to
2735
4114
the inventories.
2737
text_revision = self.revision_versions.get_text_version(
2738
file_id, revision_id)
2739
if text_revision is None:
2741
parents_of_text_revision = self.revision_versions.get_parents(
2743
parents_from_inventories = []
2744
for parent in parents_of_text_revision:
2745
if parent == _mod_revision.NULL_REVISION:
2747
introduced_in = self.revision_versions.get_text_version(file_id,
2749
if introduced_in is not None:
2750
parents_from_inventories.append(introduced_in)
2751
heads = set(self.revision_versions.heads(parents_from_inventories))
2753
for parent in parents_from_inventories:
2754
if parent in heads and parent not in new_parents:
2755
new_parents.append(parent)
2756
return tuple(new_parents)
4116
parent_keys = self.text_index[text_key]
4117
if parent_keys == [_mod_revision.NULL_REVISION]:
4119
return tuple(parent_keys)
2758
def check_file_version_parents(self, weave, file_id):
4121
def check_file_version_parents(self, texts, progress_bar=None):
2759
4122
"""Check the parents stored in a versioned file are correct.
2761
4124
It also detects file versions that are not referenced by their
2768
4131
revision_id) tuples for versions that are present in this versioned
2769
4132
file, but not used by the corresponding inventory.
4134
local_progress = None
4135
if progress_bar is None:
4136
local_progress = ui.ui_factory.nested_progress_bar()
4137
progress_bar = local_progress
4139
return self._check_file_version_parents(texts, progress_bar)
4142
local_progress.finished()
4144
def _check_file_version_parents(self, texts, progress_bar):
4145
"""See check_file_version_parents."""
2771
4146
wrong_parents = {}
2772
dangling_file_versions = set()
2773
for num, revision_id in enumerate(self.planned_revisions):
2774
correct_parents = self.calculate_file_version_parents(
2775
revision_id, file_id)
2776
if correct_parents is None:
2778
text_revision = self.revision_versions.get_text_version(
2779
file_id, revision_id)
4147
self.file_ids = set([file_id for file_id, _ in
4148
self.text_index.iterkeys()])
4149
# text keys is now grouped by file_id
4150
n_versions = len(self.text_index)
4151
progress_bar.update('loading text store', 0, n_versions)
4152
parent_map = self.repository.texts.get_parent_map(self.text_index)
4153
# On unlistable transports this could well be empty/error...
4154
text_keys = self.repository.texts.keys()
4155
unused_keys = frozenset(text_keys) - set(self.text_index)
4156
for num, key in enumerate(self.text_index.iterkeys()):
4157
progress_bar.update('checking text graph', num, n_versions)
4158
correct_parents = self.calculate_file_version_parents(key)
2781
knit_parents = tuple(weave.get_parents(revision_id))
4160
knit_parents = parent_map[key]
2782
4161
except errors.RevisionNotPresent:
2783
4163
knit_parents = None
2784
if text_revision != revision_id:
2785
# This file version is not referenced by its corresponding
2787
dangling_file_versions.add((file_id, revision_id))
2788
4164
if correct_parents != knit_parents:
2789
wrong_parents[revision_id] = (knit_parents, correct_parents)
2790
return wrong_parents, dangling_file_versions
4165
wrong_parents[key] = (knit_parents, correct_parents)
4166
return wrong_parents, unused_keys
4169
def _old_get_graph(repository, revision_id):
4170
"""DO NOT USE. That is all. I'm serious."""
4171
graph = repository.get_graph()
4172
revision_graph = dict(((key, value) for key, value in
4173
graph.iter_ancestry([revision_id]) if value is not None))
4174
return _strip_NULL_ghosts(revision_graph)
4177
def _strip_NULL_ghosts(revision_graph):
4178
"""Also don't use this. more compatibility code for unmigrated clients."""
4179
# Filter ghosts, and null:
4180
if _mod_revision.NULL_REVISION in revision_graph:
4181
del revision_graph[_mod_revision.NULL_REVISION]
4182
for key, parents in revision_graph.items():
4183
revision_graph[key] = tuple(parent for parent in parents if parent
4185
return revision_graph
4188
class StreamSink(object):
4189
"""An object that can insert a stream into a repository.
4191
This interface handles the complexity of reserialising inventories and
4192
revisions from different formats, and allows unidirectional insertion into
4193
stacked repositories without looking for the missing basis parents
4197
def __init__(self, target_repo):
4198
self.target_repo = target_repo
4200
def insert_stream(self, stream, src_format, resume_tokens):
4201
"""Insert a stream's content into the target repository.
4203
:param src_format: a bzr repository format.
4205
:return: a list of resume tokens and an iterable of keys additional
4206
items required before the insertion can be completed.
4208
self.target_repo.lock_write()
4211
self.target_repo.resume_write_group(resume_tokens)
4214
self.target_repo.start_write_group()
4217
# locked_insert_stream performs a commit|suspend.
4218
return self._locked_insert_stream(stream, src_format, is_resume)
4220
self.target_repo.abort_write_group(suppress_errors=True)
4223
self.target_repo.unlock()
4225
def _locked_insert_stream(self, stream, src_format, is_resume):
4226
to_serializer = self.target_repo._format._serializer
4227
src_serializer = src_format._serializer
4229
if to_serializer == src_serializer:
4230
# If serializers match and the target is a pack repository, set the
4231
# write cache size on the new pack. This avoids poor performance
4232
# on transports where append is unbuffered (such as
4233
# RemoteTransport). This is safe to do because nothing should read
4234
# back from the target repository while a stream with matching
4235
# serialization is being inserted.
4236
# The exception is that a delta record from the source that should
4237
# be a fulltext may need to be expanded by the target (see
4238
# test_fetch_revisions_with_deltas_into_pack); but we take care to
4239
# explicitly flush any buffered writes first in that rare case.
4241
new_pack = self.target_repo._pack_collection._new_pack
4242
except AttributeError:
4243
# Not a pack repository
4246
new_pack.set_write_cache_size(1024*1024)
4247
for substream_type, substream in stream:
4248
if 'stream' in debug.debug_flags:
4249
mutter('inserting substream: %s', substream_type)
4250
if substream_type == 'texts':
4251
self.target_repo.texts.insert_record_stream(substream)
4252
elif substream_type == 'inventories':
4253
if src_serializer == to_serializer:
4254
self.target_repo.inventories.insert_record_stream(
4257
self._extract_and_insert_inventories(
4258
substream, src_serializer)
4259
elif substream_type == 'inventory-deltas':
4260
ui.ui_factory.warn_cross_format_fetch(src_format,
4261
self.target_repo._format)
4262
self._extract_and_insert_inventory_deltas(
4263
substream, src_serializer)
4264
elif substream_type == 'chk_bytes':
4265
# XXX: This doesn't support conversions, as it assumes the
4266
# conversion was done in the fetch code.
4267
self.target_repo.chk_bytes.insert_record_stream(substream)
4268
elif substream_type == 'revisions':
4269
# This may fallback to extract-and-insert more often than
4270
# required if the serializers are different only in terms of
4272
if src_serializer == to_serializer:
4273
self.target_repo.revisions.insert_record_stream(
4276
self._extract_and_insert_revisions(substream,
4278
elif substream_type == 'signatures':
4279
self.target_repo.signatures.insert_record_stream(substream)
4281
raise AssertionError('kaboom! %s' % (substream_type,))
4282
# Done inserting data, and the missing_keys calculations will try to
4283
# read back from the inserted data, so flush the writes to the new pack
4284
# (if this is pack format).
4285
if new_pack is not None:
4286
new_pack._write_data('', flush=True)
4287
# Find all the new revisions (including ones from resume_tokens)
4288
missing_keys = self.target_repo.get_missing_parent_inventories(
4289
check_for_missing_texts=is_resume)
4291
for prefix, versioned_file in (
4292
('texts', self.target_repo.texts),
4293
('inventories', self.target_repo.inventories),
4294
('revisions', self.target_repo.revisions),
4295
('signatures', self.target_repo.signatures),
4296
('chk_bytes', self.target_repo.chk_bytes),
4298
if versioned_file is None:
4300
# TODO: key is often going to be a StaticTuple object
4301
# I don't believe we can define a method by which
4302
# (prefix,) + StaticTuple will work, though we could
4303
# define a StaticTuple.sq_concat that would allow you to
4304
# pass in either a tuple or a StaticTuple as the second
4305
# object, so instead we could have:
4306
# StaticTuple(prefix) + key here...
4307
missing_keys.update((prefix,) + key for key in
4308
versioned_file.get_missing_compression_parent_keys())
4309
except NotImplementedError:
4310
# cannot even attempt suspending, and missing would have failed
4311
# during stream insertion.
4312
missing_keys = set()
4315
# suspend the write group and tell the caller what we is
4316
# missing. We know we can suspend or else we would not have
4317
# entered this code path. (All repositories that can handle
4318
# missing keys can handle suspending a write group).
4319
write_group_tokens = self.target_repo.suspend_write_group()
4320
return write_group_tokens, missing_keys
4321
hint = self.target_repo.commit_write_group()
4322
if (to_serializer != src_serializer and
4323
self.target_repo._format.pack_compresses):
4324
self.target_repo.pack(hint=hint)
4327
def _extract_and_insert_inventory_deltas(self, substream, serializer):
4328
target_rich_root = self.target_repo._format.rich_root_data
4329
target_tree_refs = self.target_repo._format.supports_tree_reference
4330
for record in substream:
4331
# Insert the delta directly
4332
inventory_delta_bytes = record.get_bytes_as('fulltext')
4333
deserialiser = inventory_delta.InventoryDeltaDeserializer()
4335
parse_result = deserialiser.parse_text_bytes(
4336
inventory_delta_bytes)
4337
except inventory_delta.IncompatibleInventoryDelta, err:
4338
trace.mutter("Incompatible delta: %s", err.msg)
4339
raise errors.IncompatibleRevision(self.target_repo._format)
4340
basis_id, new_id, rich_root, tree_refs, inv_delta = parse_result
4341
revision_id = new_id
4342
parents = [key[0] for key in record.parents]
4343
self.target_repo.add_inventory_by_delta(
4344
basis_id, inv_delta, revision_id, parents)
4346
def _extract_and_insert_inventories(self, substream, serializer,
4348
"""Generate a new inventory versionedfile in target, converting data.
4350
The inventory is retrieved from the source, (deserializing it), and
4351
stored in the target (reserializing it in a different format).
4353
target_rich_root = self.target_repo._format.rich_root_data
4354
target_tree_refs = self.target_repo._format.supports_tree_reference
4355
for record in substream:
4356
# It's not a delta, so it must be a fulltext in the source
4357
# serializer's format.
4358
bytes = record.get_bytes_as('fulltext')
4359
revision_id = record.key[0]
4360
inv = serializer.read_inventory_from_string(bytes, revision_id)
4361
parents = [key[0] for key in record.parents]
4362
self.target_repo.add_inventory(revision_id, inv, parents)
4363
# No need to keep holding this full inv in memory when the rest of
4364
# the substream is likely to be all deltas.
4367
def _extract_and_insert_revisions(self, substream, serializer):
4368
for record in substream:
4369
bytes = record.get_bytes_as('fulltext')
4370
revision_id = record.key[0]
4371
rev = serializer.read_revision_from_string(bytes)
4372
if rev.revision_id != revision_id:
4373
raise AssertionError('wtf: %s != %s' % (rev, revision_id))
4374
self.target_repo.add_revision(revision_id, rev)
4377
if self.target_repo._format._fetch_reconcile:
4378
self.target_repo.reconcile()
4381
class StreamSource(object):
4382
"""A source of a stream for fetching between repositories."""
4384
def __init__(self, from_repository, to_format):
4385
"""Create a StreamSource streaming from from_repository."""
4386
self.from_repository = from_repository
4387
self.to_format = to_format
4389
def delta_on_metadata(self):
4390
"""Return True if delta's are permitted on metadata streams.
4392
That is on revisions and signatures.
4394
src_serializer = self.from_repository._format._serializer
4395
target_serializer = self.to_format._serializer
4396
return (self.to_format._fetch_uses_deltas and
4397
src_serializer == target_serializer)
4399
def _fetch_revision_texts(self, revs):
4400
# fetch signatures first and then the revision texts
4401
# may need to be a InterRevisionStore call here.
4402
from_sf = self.from_repository.signatures
4403
# A missing signature is just skipped.
4404
keys = [(rev_id,) for rev_id in revs]
4405
signatures = versionedfile.filter_absent(from_sf.get_record_stream(
4407
self.to_format._fetch_order,
4408
not self.to_format._fetch_uses_deltas))
4409
# If a revision has a delta, this is actually expanded inside the
4410
# insert_record_stream code now, which is an alternate fix for
4412
from_rf = self.from_repository.revisions
4413
revisions = from_rf.get_record_stream(
4415
self.to_format._fetch_order,
4416
not self.delta_on_metadata())
4417
return [('signatures', signatures), ('revisions', revisions)]
4419
def _generate_root_texts(self, revs):
4420
"""This will be called by get_stream between fetching weave texts and
4421
fetching the inventory weave.
4423
if self._rich_root_upgrade():
4424
return _mod_fetch.Inter1and2Helper(
4425
self.from_repository).generate_root_texts(revs)
4429
def get_stream(self, search):
4431
revs = search.get_keys()
4432
graph = self.from_repository.get_graph()
4433
revs = tsort.topo_sort(graph.get_parent_map(revs))
4434
data_to_fetch = self.from_repository.item_keys_introduced_by(revs)
4436
for knit_kind, file_id, revisions in data_to_fetch:
4437
if knit_kind != phase:
4439
# Make a new progress bar for this phase
4440
if knit_kind == "file":
4441
# Accumulate file texts
4442
text_keys.extend([(file_id, revision) for revision in
4444
elif knit_kind == "inventory":
4445
# Now copy the file texts.
4446
from_texts = self.from_repository.texts
4447
yield ('texts', from_texts.get_record_stream(
4448
text_keys, self.to_format._fetch_order,
4449
not self.to_format._fetch_uses_deltas))
4450
# Cause an error if a text occurs after we have done the
4453
# Before we process the inventory we generate the root
4454
# texts (if necessary) so that the inventories references
4456
for _ in self._generate_root_texts(revs):
4458
# we fetch only the referenced inventories because we do not
4459
# know for unselected inventories whether all their required
4460
# texts are present in the other repository - it could be
4462
for info in self._get_inventory_stream(revs):
4464
elif knit_kind == "signatures":
4465
# Nothing to do here; this will be taken care of when
4466
# _fetch_revision_texts happens.
4468
elif knit_kind == "revisions":
4469
for record in self._fetch_revision_texts(revs):
4472
raise AssertionError("Unknown knit kind %r" % knit_kind)
4474
def get_stream_for_missing_keys(self, missing_keys):
4475
# missing keys can only occur when we are byte copying and not
4476
# translating (because translation means we don't send
4477
# unreconstructable deltas ever).
4479
keys['texts'] = set()
4480
keys['revisions'] = set()
4481
keys['inventories'] = set()
4482
keys['chk_bytes'] = set()
4483
keys['signatures'] = set()
4484
for key in missing_keys:
4485
keys[key[0]].add(key[1:])
4486
if len(keys['revisions']):
4487
# If we allowed copying revisions at this point, we could end up
4488
# copying a revision without copying its required texts: a
4489
# violation of the requirements for repository integrity.
4490
raise AssertionError(
4491
'cannot copy revisions to fill in missing deltas %s' % (
4492
keys['revisions'],))
4493
for substream_kind, keys in keys.iteritems():
4494
vf = getattr(self.from_repository, substream_kind)
4495
if vf is None and keys:
4496
raise AssertionError(
4497
"cannot fill in keys for a versioned file we don't"
4498
" have: %s needs %s" % (substream_kind, keys))
4500
# No need to stream something we don't have
4502
if substream_kind == 'inventories':
4503
# Some missing keys are genuinely ghosts, filter those out.
4504
present = self.from_repository.inventories.get_parent_map(keys)
4505
revs = [key[0] for key in present]
4506
# Get the inventory stream more-or-less as we do for the
4507
# original stream; there's no reason to assume that records
4508
# direct from the source will be suitable for the sink. (Think
4509
# e.g. 2a -> 1.9-rich-root).
4510
for info in self._get_inventory_stream(revs, missing=True):
4514
# Ask for full texts always so that we don't need more round trips
4515
# after this stream.
4516
# Some of the missing keys are genuinely ghosts, so filter absent
4517
# records. The Sink is responsible for doing another check to
4518
# ensure that ghosts don't introduce missing data for future
4520
stream = versionedfile.filter_absent(vf.get_record_stream(keys,
4521
self.to_format._fetch_order, True))
4522
yield substream_kind, stream
4524
def inventory_fetch_order(self):
4525
if self._rich_root_upgrade():
4526
return 'topological'
4528
return self.to_format._fetch_order
4530
def _rich_root_upgrade(self):
4531
return (not self.from_repository._format.rich_root_data and
4532
self.to_format.rich_root_data)
4534
def _get_inventory_stream(self, revision_ids, missing=False):
4535
from_format = self.from_repository._format
4536
if (from_format.supports_chks and self.to_format.supports_chks and
4537
from_format.network_name() == self.to_format.network_name()):
4538
raise AssertionError(
4539
"this case should be handled by GroupCHKStreamSource")
4540
elif 'forceinvdeltas' in debug.debug_flags:
4541
return self._get_convertable_inventory_stream(revision_ids,
4542
delta_versus_null=missing)
4543
elif from_format.network_name() == self.to_format.network_name():
4545
return self._get_simple_inventory_stream(revision_ids,
4547
elif (not from_format.supports_chks and not self.to_format.supports_chks
4548
and from_format._serializer == self.to_format._serializer):
4549
# Essentially the same format.
4550
return self._get_simple_inventory_stream(revision_ids,
4553
# Any time we switch serializations, we want to use an
4554
# inventory-delta based approach.
4555
return self._get_convertable_inventory_stream(revision_ids,
4556
delta_versus_null=missing)
4558
def _get_simple_inventory_stream(self, revision_ids, missing=False):
4559
# NB: This currently reopens the inventory weave in source;
4560
# using a single stream interface instead would avoid this.
4561
from_weave = self.from_repository.inventories
4563
delta_closure = True
4565
delta_closure = not self.delta_on_metadata()
4566
yield ('inventories', from_weave.get_record_stream(
4567
[(rev_id,) for rev_id in revision_ids],
4568
self.inventory_fetch_order(), delta_closure))
4570
def _get_convertable_inventory_stream(self, revision_ids,
4571
delta_versus_null=False):
4572
# The two formats are sufficiently different that there is no fast
4573
# path, so we need to send just inventorydeltas, which any
4574
# sufficiently modern client can insert into any repository.
4575
# The StreamSink code expects to be able to
4576
# convert on the target, so we need to put bytes-on-the-wire that can
4577
# be converted. That means inventory deltas (if the remote is <1.19,
4578
# RemoteStreamSink will fallback to VFS to insert the deltas).
4579
yield ('inventory-deltas',
4580
self._stream_invs_as_deltas(revision_ids,
4581
delta_versus_null=delta_versus_null))
4583
def _stream_invs_as_deltas(self, revision_ids, delta_versus_null=False):
4584
"""Return a stream of inventory-deltas for the given rev ids.
4586
:param revision_ids: The list of inventories to transmit
4587
:param delta_versus_null: Don't try to find a minimal delta for this
4588
entry, instead compute the delta versus the NULL_REVISION. This
4589
effectively streams a complete inventory. Used for stuff like
4590
filling in missing parents, etc.
4592
from_repo = self.from_repository
4593
revision_keys = [(rev_id,) for rev_id in revision_ids]
4594
parent_map = from_repo.inventories.get_parent_map(revision_keys)
4595
# XXX: possibly repos could implement a more efficient iter_inv_deltas
4597
inventories = self.from_repository.iter_inventories(
4598
revision_ids, 'topological')
4599
format = from_repo._format
4600
invs_sent_so_far = set([_mod_revision.NULL_REVISION])
4601
inventory_cache = lru_cache.LRUCache(50)
4602
null_inventory = from_repo.revision_tree(
4603
_mod_revision.NULL_REVISION).inventory
4604
# XXX: ideally the rich-root/tree-refs flags would be per-revision, not
4605
# per-repo (e.g. streaming a non-rich-root revision out of a rich-root
4606
# repo back into a non-rich-root repo ought to be allowed)
4607
serializer = inventory_delta.InventoryDeltaSerializer(
4608
versioned_root=format.rich_root_data,
4609
tree_references=format.supports_tree_reference)
4610
for inv in inventories:
4611
key = (inv.revision_id,)
4612
parent_keys = parent_map.get(key, ())
4614
if not delta_versus_null and parent_keys:
4615
# The caller did not ask for complete inventories and we have
4616
# some parents that we can delta against. Make a delta against
4617
# each parent so that we can find the smallest.
4618
parent_ids = [parent_key[0] for parent_key in parent_keys]
4619
for parent_id in parent_ids:
4620
if parent_id not in invs_sent_so_far:
4621
# We don't know that the remote side has this basis, so
4624
if parent_id == _mod_revision.NULL_REVISION:
4625
parent_inv = null_inventory
4627
parent_inv = inventory_cache.get(parent_id, None)
4628
if parent_inv is None:
4629
parent_inv = from_repo.get_inventory(parent_id)
4630
candidate_delta = inv._make_delta(parent_inv)
4631
if (delta is None or
4632
len(delta) > len(candidate_delta)):
4633
delta = candidate_delta
4634
basis_id = parent_id
4636
# Either none of the parents ended up being suitable, or we
4637
# were asked to delta against NULL
4638
basis_id = _mod_revision.NULL_REVISION
4639
delta = inv._make_delta(null_inventory)
4640
invs_sent_so_far.add(inv.revision_id)
4641
inventory_cache[inv.revision_id] = inv
4642
delta_serialized = ''.join(
4643
serializer.delta_to_lines(basis_id, key[-1], delta))
4644
yield versionedfile.FulltextContentFactory(
4645
key, parent_keys, None, delta_serialized)
4648
def _iter_for_revno(repo, partial_history_cache, stop_index=None,
4649
stop_revision=None):
4650
"""Extend the partial history to include a given index
4652
If a stop_index is supplied, stop when that index has been reached.
4653
If a stop_revision is supplied, stop when that revision is
4654
encountered. Otherwise, stop when the beginning of history is
4657
:param stop_index: The index which should be present. When it is
4658
present, history extension will stop.
4659
:param stop_revision: The revision id which should be present. When
4660
it is encountered, history extension will stop.
4662
start_revision = partial_history_cache[-1]
4663
iterator = repo.iter_reverse_revision_history(start_revision)
4665
#skip the last revision in the list
4668
if (stop_index is not None and
4669
len(partial_history_cache) > stop_index):
4671
if partial_history_cache[-1] == stop_revision:
4673
revision_id = iterator.next()
4674
partial_history_cache.append(revision_id)
4675
except StopIteration:
added
removed
bzrlib/repository.py
