548
492
ie.revision = parent_entry.revision
549
493
return self._get_delta(ie, basis_inv, path), False, None
550
494
ie.reference_revision = content_summary[3]
551
if ie.reference_revision is None:
552
raise AssertionError("invalid content_summary for nested tree: %r"
553
% (content_summary,))
554
self._add_text_to_weave(ie.file_id, '', heads, None)
496
self._add_text_to_weave(ie.file_id, lines, heads, None)
556
498
raise NotImplementedError('unknown kind')
557
499
ie.revision = self._new_revision_id
558
self._any_changes = True
559
500
return self._get_delta(ie, basis_inv, path), True, fingerprint
561
def record_iter_changes(self, tree, basis_revision_id, iter_changes,
562
_entry_factory=entry_factory):
563
"""Record a new tree via iter_changes.
565
:param tree: The tree to obtain text contents from for changed objects.
566
:param basis_revision_id: The revision id of the tree the iter_changes
567
has been generated against. Currently assumed to be the same
568
as self.parents[0] - if it is not, errors may occur.
569
:param iter_changes: An iter_changes iterator with the changes to apply
570
to basis_revision_id. The iterator must not include any items with
571
a current kind of None - missing items must be either filtered out
572
or errored-on beefore record_iter_changes sees the item.
573
:param _entry_factory: Private method to bind entry_factory locally for
575
:return: A generator of (file_id, relpath, fs_hash) tuples for use with
578
# Create an inventory delta based on deltas between all the parents and
579
# deltas between all the parent inventories. We use inventory delta's
580
# between the inventory objects because iter_changes masks
581
# last-changed-field only changes.
583
# file_id -> change map, change is fileid, paths, changed, versioneds,
584
# parents, names, kinds, executables
586
# {file_id -> revision_id -> inventory entry, for entries in parent
587
# trees that are not parents[0]
591
revtrees = list(self.repository.revision_trees(self.parents))
592
except errors.NoSuchRevision:
593
# one or more ghosts, slow path.
595
for revision_id in self.parents:
597
revtrees.append(self.repository.revision_tree(revision_id))
598
except errors.NoSuchRevision:
600
basis_revision_id = _mod_revision.NULL_REVISION
602
revtrees.append(self.repository.revision_tree(
603
_mod_revision.NULL_REVISION))
604
# The basis inventory from a repository
606
basis_inv = revtrees[0].inventory
608
basis_inv = self.repository.revision_tree(
609
_mod_revision.NULL_REVISION).inventory
610
if len(self.parents) > 0:
611
if basis_revision_id != self.parents[0] and not ghost_basis:
613
"arbitrary basis parents not yet supported with merges")
614
for revtree in revtrees[1:]:
615
for change in revtree.inventory._make_delta(basis_inv):
616
if change[1] is None:
617
# Not present in this parent.
619
if change[2] not in merged_ids:
620
if change[0] is not None:
621
basis_entry = basis_inv[change[2]]
622
merged_ids[change[2]] = [
624
basis_entry.revision,
627
parent_entries[change[2]] = {
629
basis_entry.revision:basis_entry,
631
change[3].revision:change[3],
634
merged_ids[change[2]] = [change[3].revision]
635
parent_entries[change[2]] = {change[3].revision:change[3]}
637
merged_ids[change[2]].append(change[3].revision)
638
parent_entries[change[2]][change[3].revision] = change[3]
641
# Setup the changes from the tree:
642
# changes maps file_id -> (change, [parent revision_ids])
644
for change in iter_changes:
645
# This probably looks up in basis_inv way to much.
646
if change[1][0] is not None:
647
head_candidate = [basis_inv[change[0]].revision]
650
changes[change[0]] = change, merged_ids.get(change[0],
652
unchanged_merged = set(merged_ids) - set(changes)
653
# Extend the changes dict with synthetic changes to record merges of
655
for file_id in unchanged_merged:
656
# Record a merged version of these items that did not change vs the
657
# basis. This can be either identical parallel changes, or a revert
658
# of a specific file after a merge. The recorded content will be
659
# that of the current tree (which is the same as the basis), but
660
# the per-file graph will reflect a merge.
661
# NB:XXX: We are reconstructing path information we had, this
662
# should be preserved instead.
663
# inv delta change: (file_id, (path_in_source, path_in_target),
664
# changed_content, versioned, parent, name, kind,
667
basis_entry = basis_inv[file_id]
668
except errors.NoSuchId:
669
# a change from basis->some_parents but file_id isn't in basis
670
# so was new in the merge, which means it must have changed
671
# from basis -> current, and as it hasn't the add was reverted
672
# by the user. So we discard this change.
676
(basis_inv.id2path(file_id), tree.id2path(file_id)),
678
(basis_entry.parent_id, basis_entry.parent_id),
679
(basis_entry.name, basis_entry.name),
680
(basis_entry.kind, basis_entry.kind),
681
(basis_entry.executable, basis_entry.executable))
682
changes[file_id] = (change, merged_ids[file_id])
683
# changes contains tuples with the change and a set of inventory
684
# candidates for the file.
686
# old_path, new_path, file_id, new_inventory_entry
687
seen_root = False # Is the root in the basis delta?
688
inv_delta = self._basis_delta
689
modified_rev = self._new_revision_id
690
for change, head_candidates in changes.values():
691
if change[3][1]: # versioned in target.
692
# Several things may be happening here:
693
# We may have a fork in the per-file graph
694
# - record a change with the content from tree
695
# We may have a change against < all trees
696
# - carry over the tree that hasn't changed
697
# We may have a change against all trees
698
# - record the change with the content from tree
701
entry = _entry_factory[kind](file_id, change[5][1],
703
head_set = self._heads(change[0], set(head_candidates))
706
for head_candidate in head_candidates:
707
if head_candidate in head_set:
708
heads.append(head_candidate)
709
head_set.remove(head_candidate)
712
# Could be a carry-over situation:
713
parent_entry_revs = parent_entries.get(file_id, None)
714
if parent_entry_revs:
715
parent_entry = parent_entry_revs.get(heads[0], None)
718
if parent_entry is None:
719
# The parent iter_changes was called against is the one
720
# that is the per-file head, so any change is relevant
721
# iter_changes is valid.
722
carry_over_possible = False
724
# could be a carry over situation
725
# A change against the basis may just indicate a merge,
726
# we need to check the content against the source of the
727
# merge to determine if it was changed after the merge
729
if (parent_entry.kind != entry.kind or
730
parent_entry.parent_id != entry.parent_id or
731
parent_entry.name != entry.name):
732
# Metadata common to all entries has changed
733
# against per-file parent
734
carry_over_possible = False
736
carry_over_possible = True
737
# per-type checks for changes against the parent_entry
740
# Cannot be a carry-over situation
741
carry_over_possible = False
742
# Populate the entry in the delta
744
# XXX: There is still a small race here: If someone reverts the content of a file
745
# after iter_changes examines and decides it has changed,
746
# we will unconditionally record a new version even if some
747
# other process reverts it while commit is running (with
748
# the revert happening after iter_changes did it's
751
entry.executable = True
753
entry.executable = False
754
if (carry_over_possible and
755
parent_entry.executable == entry.executable):
756
# Check the file length, content hash after reading
758
nostore_sha = parent_entry.text_sha1
761
file_obj, stat_value = tree.get_file_with_stat(file_id, change[1][1])
763
text = file_obj.read()
767
entry.text_sha1, entry.text_size = self._add_text_to_weave(
768
file_id, text, heads, nostore_sha)
769
yield file_id, change[1][1], (entry.text_sha1, stat_value)
770
except errors.ExistingContent:
771
# No content change against a carry_over parent
772
# Perhaps this should also yield a fs hash update?
774
entry.text_size = parent_entry.text_size
775
entry.text_sha1 = parent_entry.text_sha1
776
elif kind == 'symlink':
778
entry.symlink_target = tree.get_symlink_target(file_id)
779
if (carry_over_possible and
780
parent_entry.symlink_target == entry.symlink_target):
783
self._add_text_to_weave(change[0], '', heads, None)
784
elif kind == 'directory':
785
if carry_over_possible:
788
# Nothing to set on the entry.
789
# XXX: split into the Root and nonRoot versions.
790
if change[1][1] != '' or self.repository.supports_rich_root():
791
self._add_text_to_weave(change[0], '', heads, None)
792
elif kind == 'tree-reference':
793
if not self.repository._format.supports_tree_reference:
794
# This isn't quite sane as an error, but we shouldn't
795
# ever see this code path in practice: tree's don't
796
# permit references when the repo doesn't support tree
798
raise errors.UnsupportedOperation(tree.add_reference,
800
reference_revision = tree.get_reference_revision(change[0])
801
entry.reference_revision = reference_revision
802
if (carry_over_possible and
803
parent_entry.reference_revision == reference_revision):
806
self._add_text_to_weave(change[0], '', heads, None)
808
raise AssertionError('unknown kind %r' % kind)
810
entry.revision = modified_rev
812
entry.revision = parent_entry.revision
815
new_path = change[1][1]
816
inv_delta.append((change[1][0], new_path, change[0], entry))
819
self.new_inventory = None
821
# This should perhaps be guarded by a check that the basis we
822
# commit against is the basis for the commit and if not do a delta
824
self._any_changes = True
826
# housekeeping root entry changes do not affect no-change commits.
827
self._require_root_change(tree)
828
self.basis_delta_revision = basis_revision_id
830
def _add_text_to_weave(self, file_id, new_text, parents, nostore_sha):
831
parent_keys = tuple([(file_id, parent) for parent in parents])
832
return self.repository.texts._add_text(
833
(file_id, self._new_revision_id), parent_keys, new_text,
834
nostore_sha=nostore_sha, random_id=self.random_revid)[0:2]
502
def _add_text_to_weave(self, file_id, new_lines, parents, nostore_sha):
503
# Note: as we read the content directly from the tree, we know its not
504
# been turned into unicode or badly split - but a broken tree
505
# implementation could give us bad output from readlines() so this is
506
# not a guarantee of safety. What would be better is always checking
507
# the content during test suite execution. RBC 20070912
508
parent_keys = tuple((file_id, parent) for parent in parents)
509
return self.repository.texts.add_lines(
510
(file_id, self._new_revision_id), parent_keys, new_lines,
511
nostore_sha=nostore_sha, random_id=self.random_revid,
512
check_content=False)[0:2]
837
515
class RootCommitBuilder(CommitBuilder):
1168
796
# The old API returned a list, should this actually be a set?
1169
797
return parent_map.keys()
1171
def _check_inventories(self, checker):
1172
"""Check the inventories found from the revision scan.
1174
This is responsible for verifying the sha1 of inventories and
1175
creating a pending_keys set that covers data referenced by inventories.
1177
bar = ui.ui_factory.nested_progress_bar()
1179
self._do_check_inventories(checker, bar)
1183
def _do_check_inventories(self, checker, bar):
1184
"""Helper for _check_inventories."""
1186
keys = {'chk_bytes':set(), 'inventories':set(), 'texts':set()}
1187
kinds = ['chk_bytes', 'texts']
1188
count = len(checker.pending_keys)
1189
bar.update("inventories", 0, 2)
1190
current_keys = checker.pending_keys
1191
checker.pending_keys = {}
1192
# Accumulate current checks.
1193
for key in current_keys:
1194
if key[0] != 'inventories' and key[0] not in kinds:
1195
checker._report_items.append('unknown key type %r' % (key,))
1196
keys[key[0]].add(key[1:])
1197
if keys['inventories']:
1198
# NB: output order *should* be roughly sorted - topo or
1199
# inverse topo depending on repository - either way decent
1200
# to just delta against. However, pre-CHK formats didn't
1201
# try to optimise inventory layout on disk. As such the
1202
# pre-CHK code path does not use inventory deltas.
1204
for record in self.inventories.check(keys=keys['inventories']):
1205
if record.storage_kind == 'absent':
1206
checker._report_items.append(
1207
'Missing inventory {%s}' % (record.key,))
1209
last_object = self._check_record('inventories', record,
1210
checker, last_object,
1211
current_keys[('inventories',) + record.key])
1212
del keys['inventories']
1215
bar.update("texts", 1)
1216
while (checker.pending_keys or keys['chk_bytes']
1218
# Something to check.
1219
current_keys = checker.pending_keys
1220
checker.pending_keys = {}
1221
# Accumulate current checks.
1222
for key in current_keys:
1223
if key[0] not in kinds:
1224
checker._report_items.append('unknown key type %r' % (key,))
1225
keys[key[0]].add(key[1:])
1226
# Check the outermost kind only - inventories || chk_bytes || texts
1230
for record in getattr(self, kind).check(keys=keys[kind]):
1231
if record.storage_kind == 'absent':
1232
checker._report_items.append(
1233
'Missing %s {%s}' % (kind, record.key,))
1235
last_object = self._check_record(kind, record,
1236
checker, last_object, current_keys[(kind,) + record.key])
1240
def _check_record(self, kind, record, checker, last_object, item_data):
1241
"""Check a single text from this repository."""
1242
if kind == 'inventories':
1243
rev_id = record.key[0]
1244
inv = self._deserialise_inventory(rev_id,
1245
record.get_bytes_as('fulltext'))
1246
if last_object is not None:
1247
delta = inv._make_delta(last_object)
1248
for old_path, path, file_id, ie in delta:
1251
ie.check(checker, rev_id, inv)
1253
for path, ie in inv.iter_entries():
1254
ie.check(checker, rev_id, inv)
1255
if self._format.fast_deltas:
1257
elif kind == 'chk_bytes':
1258
# No code written to check chk_bytes for this repo format.
1259
checker._report_items.append(
1260
'unsupported key type chk_bytes for %s' % (record.key,))
1261
elif kind == 'texts':
1262
self._check_text(record, checker, item_data)
1264
checker._report_items.append(
1265
'unknown key type %s for %s' % (kind, record.key))
1267
def _check_text(self, record, checker, item_data):
1268
"""Check a single text."""
1269
# Check it is extractable.
1270
# TODO: check length.
1271
if record.storage_kind == 'chunked':
1272
chunks = record.get_bytes_as(record.storage_kind)
1273
sha1 = osutils.sha_strings(chunks)
1274
length = sum(map(len, chunks))
1276
content = record.get_bytes_as('fulltext')
1277
sha1 = osutils.sha_string(content)
1278
length = len(content)
1279
if item_data and sha1 != item_data[1]:
1280
checker._report_items.append(
1281
'sha1 mismatch: %s has sha1 %s expected %s referenced by %s' %
1282
(record.key, sha1, item_data[1], item_data[2]))
1285
800
def create(a_bzrdir):
1286
801
"""Construct the current default format repository in a_bzrdir."""
1578
1084
def suspend_write_group(self):
1579
1085
raise errors.UnsuspendableWriteGroup(self)
1581
def get_missing_parent_inventories(self, check_for_missing_texts=True):
1582
"""Return the keys of missing inventory parents for revisions added in
1585
A revision is not complete if the inventory delta for that revision
1586
cannot be calculated. Therefore if the parent inventories of a
1587
revision are not present, the revision is incomplete, and e.g. cannot
1588
be streamed by a smart server. This method finds missing inventory
1589
parents for revisions added in this write group.
1591
if not self._format.supports_external_lookups:
1592
# This is only an issue for stacked repositories
1594
if not self.is_in_write_group():
1595
raise AssertionError('not in a write group')
1597
# XXX: We assume that every added revision already has its
1598
# corresponding inventory, so we only check for parent inventories that
1599
# might be missing, rather than all inventories.
1600
parents = set(self.revisions._index.get_missing_parents())
1601
parents.discard(_mod_revision.NULL_REVISION)
1602
unstacked_inventories = self.inventories._index
1603
present_inventories = unstacked_inventories.get_parent_map(
1604
key[-1:] for key in parents)
1605
parents.difference_update(present_inventories)
1606
if len(parents) == 0:
1607
# No missing parent inventories.
1609
if not check_for_missing_texts:
1610
return set(('inventories', rev_id) for (rev_id,) in parents)
1611
# Ok, now we have a list of missing inventories. But these only matter
1612
# if the inventories that reference them are missing some texts they
1613
# appear to introduce.
1614
# XXX: Texts referenced by all added inventories need to be present,
1615
# but at the moment we're only checking for texts referenced by
1616
# inventories at the graph's edge.
1617
key_deps = self.revisions._index._key_dependencies
1618
key_deps.satisfy_refs_for_keys(present_inventories)
1619
referrers = frozenset(r[0] for r in key_deps.get_referrers())
1620
file_ids = self.fileids_altered_by_revision_ids(referrers)
1621
missing_texts = set()
1622
for file_id, version_ids in file_ids.iteritems():
1623
missing_texts.update(
1624
(file_id, version_id) for version_id in version_ids)
1625
present_texts = self.texts.get_parent_map(missing_texts)
1626
missing_texts.difference_update(present_texts)
1627
if not missing_texts:
1628
# No texts are missing, so all revisions and their deltas are
1631
# Alternatively the text versions could be returned as the missing
1632
# keys, but this is likely to be less data.
1633
missing_keys = set(('inventories', rev_id) for (rev_id,) in parents)
1636
def refresh_data(self):
1637
"""Re-read any data needed to to synchronise with disk.
1639
This method is intended to be called after another repository instance
1640
(such as one used by a smart server) has inserted data into the
1641
repository. It may not be called during a write group, but may be
1642
called at any other time.
1644
if self.is_in_write_group():
1645
raise errors.InternalBzrError(
1646
"May not refresh_data while in a write group.")
1647
self._refresh_data()
1649
1087
def resume_write_group(self, tokens):
1650
1088
if not self.is_write_locked():
1651
1089
raise errors.NotWriteLocked(self)
1860
1269
@needs_read_lock
1861
1270
def get_revisions(self, revision_ids):
1862
"""Get many revisions at once.
1864
Repositories that need to check data on every revision read should
1865
subclass this method.
1271
"""Get many revisions at once."""
1867
1272
return self._get_revisions(revision_ids)
1869
1274
@needs_read_lock
1870
1275
def _get_revisions(self, revision_ids):
1871
1276
"""Core work logic to get many revisions without sanity checks."""
1277
for rev_id in revision_ids:
1278
if not rev_id or not isinstance(rev_id, basestring):
1279
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1280
keys = [(key,) for key in revision_ids]
1281
stream = self.revisions.get_record_stream(keys, 'unordered', True)
1873
for revid, rev in self._iter_revisions(revision_ids):
1875
raise errors.NoSuchRevision(self, revid)
1283
for record in stream:
1284
if record.storage_kind == 'absent':
1285
raise errors.NoSuchRevision(self, record.key[0])
1286
text = record.get_bytes_as('fulltext')
1287
rev = self._serializer.read_revision_from_string(text)
1288
revs[record.key[0]] = rev
1877
1289
return [revs[revid] for revid in revision_ids]
1879
def _iter_revisions(self, revision_ids):
1880
"""Iterate over revision objects.
1882
:param revision_ids: An iterable of revisions to examine. None may be
1883
passed to request all revisions known to the repository. Note that
1884
not all repositories can find unreferenced revisions; for those
1885
repositories only referenced ones will be returned.
1886
:return: An iterator of (revid, revision) tuples. Absent revisions (
1887
those asked for but not available) are returned as (revid, None).
1889
if revision_ids is None:
1890
revision_ids = self.all_revision_ids()
1892
for rev_id in revision_ids:
1893
if not rev_id or not isinstance(rev_id, basestring):
1894
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1895
keys = [(key,) for key in revision_ids]
1896
stream = self.revisions.get_record_stream(keys, 'unordered', True)
1897
for record in stream:
1898
revid = record.key[0]
1899
if record.storage_kind == 'absent':
1902
text = record.get_bytes_as('fulltext')
1903
rev = self._serializer.read_revision_from_string(text)
1906
def get_deltas_for_revisions(self, revisions, specific_fileids=None):
1292
def get_revision_xml(self, revision_id):
1293
# TODO: jam 20070210 This shouldn't be necessary since get_revision
1294
# would have already do it.
1295
# TODO: jam 20070210 Just use _serializer.write_revision_to_string()
1296
rev = self.get_revision(revision_id)
1297
rev_tmp = cStringIO.StringIO()
1298
# the current serializer..
1299
self._serializer.write_revision(rev, rev_tmp)
1301
return rev_tmp.getvalue()
1303
def get_deltas_for_revisions(self, revisions):
1907
1304
"""Produce a generator of revision deltas.
1909
1306
Note that the input is a sequence of REVISIONS, not revision_ids.
1910
1307
Trees will be held in memory until the generator exits.
1911
1308
Each delta is relative to the revision's lefthand predecessor.
1913
:param specific_fileids: if not None, the result is filtered
1914
so that only those file-ids, their parents and their
1915
children are included.
1917
# Get the revision-ids of interest
1918
1310
required_trees = set()
1919
1311
for revision in revisions:
1920
1312
required_trees.add(revision.revision_id)
1921
1313
required_trees.update(revision.parent_ids[:1])
1923
# Get the matching filtered trees. Note that it's more
1924
# efficient to pass filtered trees to changes_from() rather
1925
# than doing the filtering afterwards. changes_from() could
1926
# arguably do the filtering itself but it's path-based, not
1927
# file-id based, so filtering before or afterwards is
1929
if specific_fileids is None:
1930
trees = dict((t.get_revision_id(), t) for
1931
t in self.revision_trees(required_trees))
1933
trees = dict((t.get_revision_id(), t) for
1934
t in self._filtered_revision_trees(required_trees,
1937
# Calculate the deltas
1314
trees = dict((t.get_revision_id(), t) for
1315
t in self.revision_trees(required_trees))
1938
1316
for revision in revisions:
1939
1317
if not revision.parent_ids:
1940
1318
old_tree = self.revision_tree(_mod_revision.NULL_REVISION)
2064
1440
result[key] = True
2067
def _inventory_xml_lines_for_keys(self, keys):
2068
"""Get a line iterator of the sort needed for findind references.
2070
Not relevant for non-xml inventory repositories.
2072
Ghosts in revision_keys are ignored.
2074
:param revision_keys: The revision keys for the inventories to inspect.
2075
:return: An iterator over (inventory line, revid) for the fulltexts of
2076
all of the xml inventories specified by revision_keys.
2078
stream = self.inventories.get_record_stream(keys, 'unordered', True)
2079
for record in stream:
2080
if record.storage_kind != 'absent':
2081
chunks = record.get_bytes_as('chunked')
2082
revid = record.key[-1]
2083
lines = osutils.chunks_to_lines(chunks)
2087
1443
def _find_file_ids_from_xml_inventory_lines(self, line_iterator,
2089
1445
"""Helper routine for fileids_altered_by_revision_ids.
2091
1447
This performs the translation of xml lines to revision ids.
2093
1449
:param line_iterator: An iterator of lines, origin_version_id
2094
:param revision_keys: The revision ids to filter for. This should be a
1450
:param revision_ids: The revision ids to filter for. This should be a
2095
1451
set or other type which supports efficient __contains__ lookups, as
2096
the revision key from each parsed line will be looked up in the
2097
revision_keys filter.
1452
the revision id from each parsed line will be looked up in the
1453
revision_ids filter.
2098
1454
:return: a dictionary mapping altered file-ids to an iterable of
2099
1455
revision_ids. Each altered file-ids has the exact revision_ids that
2100
1456
altered it listed explicitly.
2102
seen = set(self._find_text_key_references_from_xml_inventory_lines(
2103
line_iterator).iterkeys())
2104
parent_keys = self._find_parent_keys_of_revisions(revision_keys)
2105
parent_seen = set(self._find_text_key_references_from_xml_inventory_lines(
2106
self._inventory_xml_lines_for_keys(parent_keys)))
2107
new_keys = seen - parent_seen
2109
1459
setdefault = result.setdefault
2110
for key in new_keys:
2111
setdefault(key[0], set()).add(key[-1])
1461
self._find_text_key_references_from_xml_inventory_lines(
1462
line_iterator).iterkeys():
1463
# once data is all ensured-consistent; then this is
1464
# if revision_id == version_id
1465
if key[-1:] in revision_ids:
1466
setdefault(key[0], set()).add(key[-1])
2114
def _find_parent_ids_of_revisions(self, revision_ids):
2115
"""Find all parent ids that are mentioned in the revision graph.
2117
:return: set of revisions that are parents of revision_ids which are
2118
not part of revision_ids themselves
2120
parent_map = self.get_parent_map(revision_ids)
2122
map(parent_ids.update, parent_map.itervalues())
2123
parent_ids.difference_update(revision_ids)
2124
parent_ids.discard(_mod_revision.NULL_REVISION)
2127
def _find_parent_keys_of_revisions(self, revision_keys):
2128
"""Similar to _find_parent_ids_of_revisions, but used with keys.
2130
:param revision_keys: An iterable of revision_keys.
2131
:return: The parents of all revision_keys that are not already in
2134
parent_map = self.revisions.get_parent_map(revision_keys)
2136
map(parent_keys.update, parent_map.itervalues())
2137
parent_keys.difference_update(revision_keys)
2138
parent_keys.discard(_mod_revision.NULL_REVISION)
2141
1469
def fileids_altered_by_revision_ids(self, revision_ids, _inv_weave=None):
2142
1470
"""Find the file ids and versions affected by revisions.
2347
1672
"""Get Inventory object by revision id."""
2348
1673
return self.iter_inventories([revision_id]).next()
2350
def iter_inventories(self, revision_ids, ordering=None):
1675
def iter_inventories(self, revision_ids):
2351
1676
"""Get many inventories by revision_ids.
2353
1678
This will buffer some or all of the texts used in constructing the
2354
1679
inventories in memory, but will only parse a single inventory at a
2357
:param revision_ids: The expected revision ids of the inventories.
2358
:param ordering: optional ordering, e.g. 'topological'. If not
2359
specified, the order of revision_ids will be preserved (by
2360
buffering if necessary).
2361
1682
:return: An iterator of inventories.
2363
1684
if ((None in revision_ids)
2364
1685
or (_mod_revision.NULL_REVISION in revision_ids)):
2365
1686
raise ValueError('cannot get null revision inventory')
2366
return self._iter_inventories(revision_ids, ordering)
1687
return self._iter_inventories(revision_ids)
2368
def _iter_inventories(self, revision_ids, ordering):
1689
def _iter_inventories(self, revision_ids):
2369
1690
"""single-document based inventory iteration."""
2370
inv_xmls = self._iter_inventory_xmls(revision_ids, ordering)
2371
for text, revision_id in inv_xmls:
2372
yield self._deserialise_inventory(revision_id, text)
1691
for text, revision_id in self._iter_inventory_xmls(revision_ids):
1692
yield self.deserialise_inventory(revision_id, text)
2374
def _iter_inventory_xmls(self, revision_ids, ordering):
2375
if ordering is None:
2376
order_as_requested = True
2377
ordering = 'unordered'
2379
order_as_requested = False
1694
def _iter_inventory_xmls(self, revision_ids):
2380
1695
keys = [(revision_id,) for revision_id in revision_ids]
2383
if order_as_requested:
2384
key_iter = iter(keys)
2385
next_key = key_iter.next()
2386
stream = self.inventories.get_record_stream(keys, ordering, True)
1696
stream = self.inventories.get_record_stream(keys, 'unordered', True)
2387
1697
text_chunks = {}
2388
1698
for record in stream:
2389
1699
if record.storage_kind != 'absent':
2390
chunks = record.get_bytes_as('chunked')
2391
if order_as_requested:
2392
text_chunks[record.key] = chunks
2394
yield ''.join(chunks), record.key[-1]
1700
text_chunks[record.key] = record.get_bytes_as('chunked')
2396
1702
raise errors.NoSuchRevision(self, record.key)
2397
if order_as_requested:
2398
# Yield as many results as we can while preserving order.
2399
while next_key in text_chunks:
2400
chunks = text_chunks.pop(next_key)
2401
yield ''.join(chunks), next_key[-1]
2403
next_key = key_iter.next()
2404
except StopIteration:
2405
# We still want to fully consume the get_record_stream,
2406
# just in case it is not actually finished at this point
1704
chunks = text_chunks.pop(key)
1705
yield ''.join(chunks), key[-1]
2410
def _deserialise_inventory(self, revision_id, xml):
1707
def deserialise_inventory(self, revision_id, xml):
2411
1708
"""Transform the xml into an inventory object.
2413
1710
:param revision_id: The expected revision id of the inventory.
2414
1711
:param xml: A serialised inventory.
2416
1713
result = self._serializer.read_inventory_from_string(xml, revision_id,
2417
entry_cache=self._inventory_entry_cache,
2418
return_from_cache=self._safe_to_return_from_cache)
1714
entry_cache=self._inventory_entry_cache)
2419
1715
if result.revision_id != revision_id:
2420
1716
raise AssertionError('revision id mismatch %s != %s' % (
2421
1717
result.revision_id, revision_id))
1720
def serialise_inventory(self, inv):
1721
return self._serializer.write_inventory_to_string(inv)
1723
def _serialise_inventory_to_lines(self, inv):
1724
return self._serializer.write_inventory_to_lines(inv)
2424
1726
def get_serializer_format(self):
2425
1727
return self._serializer.format_num
2427
1729
@needs_read_lock
2428
def _get_inventory_xml(self, revision_id):
2429
"""Get serialized inventory as a string."""
2430
texts = self._iter_inventory_xmls([revision_id], 'unordered')
1730
def get_inventory_xml(self, revision_id):
1731
"""Get inventory XML as a file object."""
1732
texts = self._iter_inventory_xmls([revision_id])
2432
1734
text, revision_id = texts.next()
2433
1735
except StopIteration:
2434
1736
raise errors.HistoryMissing(self, 'inventory', revision_id)
2437
def get_rev_id_for_revno(self, revno, known_pair):
2438
"""Return the revision id of a revno, given a later (revno, revid)
2439
pair in the same history.
2441
:return: if found (True, revid). If the available history ran out
2442
before reaching the revno, then this returns
2443
(False, (closest_revno, closest_revid)).
1740
def get_inventory_sha1(self, revision_id):
1741
"""Return the sha1 hash of the inventory entry
2445
known_revno, known_revid = known_pair
2446
partial_history = [known_revid]
2447
distance_from_known = known_revno - revno
2448
if distance_from_known < 0:
2450
'requested revno (%d) is later than given known revno (%d)'
2451
% (revno, known_revno))
2454
self, partial_history, stop_index=distance_from_known)
2455
except errors.RevisionNotPresent, err:
2456
if err.revision_id == known_revid:
2457
# The start revision (known_revid) wasn't found.
2459
# This is a stacked repository with no fallbacks, or a there's a
2460
# left-hand ghost. Either way, even though the revision named in
2461
# the error isn't in this repo, we know it's the next step in this
2462
# left-hand history.
2463
partial_history.append(err.revision_id)
2464
if len(partial_history) <= distance_from_known:
2465
# Didn't find enough history to get a revid for the revno.
2466
earliest_revno = known_revno - len(partial_history) + 1
2467
return (False, (earliest_revno, partial_history[-1]))
2468
if len(partial_history) - 1 > distance_from_known:
2469
raise AssertionError('_iter_for_revno returned too much history')
2470
return (True, partial_history[-1])
1743
return self.get_revision(revision_id).inventory_sha1
2472
1745
def iter_reverse_revision_history(self, revision_id):
2473
1746
"""Iterate backwards through revision ids in the lefthand history
3706
2936
return self.source.revision_ids_to_search_result(result_set)
3709
class InterDifferingSerializer(InterRepository):
2939
class InterPackRepo(InterSameDataRepository):
2940
"""Optimised code paths between Pack based repositories."""
2943
def _get_repo_format_to_test(self):
2944
from bzrlib.repofmt import pack_repo
2945
return pack_repo.RepositoryFormatKnitPack1()
2948
def is_compatible(source, target):
2949
"""Be compatible with known Pack formats.
2951
We don't test for the stores being of specific types because that
2952
could lead to confusing results, and there is no need to be
2955
from bzrlib.repofmt.pack_repo import RepositoryFormatPack
2957
are_packs = (isinstance(source._format, RepositoryFormatPack) and
2958
isinstance(target._format, RepositoryFormatPack))
2959
except AttributeError:
2961
return are_packs and InterRepository._same_model(source, target)
2964
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2965
"""See InterRepository.fetch()."""
2966
if (len(self.source._fallback_repositories) > 0 or
2967
len(self.target._fallback_repositories) > 0):
2968
# The pack layer is not aware of fallback repositories, so when
2969
# fetching from a stacked repository or into a stacked repository
2970
# we use the generic fetch logic which uses the VersionedFiles
2971
# attributes on repository.
2972
from bzrlib.fetch import RepoFetcher
2973
fetcher = RepoFetcher(self.target, self.source, revision_id,
2975
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2976
self.source, self.source._format, self.target, self.target._format)
2977
if revision_id is None:
2979
# everything to do - use pack logic
2980
# to fetch from all packs to one without
2981
# inventory parsing etc, IFF nothing to be copied is in the target.
2983
source_revision_ids = frozenset(self.source.all_revision_ids())
2984
revision_ids = source_revision_ids - \
2985
frozenset(self.target_get_parent_map(source_revision_ids))
2986
revision_keys = [(revid,) for revid in revision_ids]
2987
target_pack_collection = self._get_target_pack_collection()
2988
index = target_pack_collection.revision_index.combined_index
2989
present_revision_ids = set(item[1][0] for item in
2990
index.iter_entries(revision_keys))
2991
revision_ids = set(revision_ids) - present_revision_ids
2992
# implementing the TODO will involve:
2993
# - detecting when all of a pack is selected
2994
# - avoiding as much as possible pre-selection, so the
2995
# more-core routines such as create_pack_from_packs can filter in
2996
# a just-in-time fashion. (though having a HEADS list on a
2997
# repository might make this a lot easier, because we could
2998
# sensibly detect 'new revisions' without doing a full index scan.
2999
elif _mod_revision.is_null(revision_id):
3004
revision_ids = self.search_missing_revision_ids(revision_id,
3005
find_ghosts=find_ghosts).get_keys()
3006
except errors.NoSuchRevision:
3007
raise errors.InstallFailed([revision_id])
3008
if len(revision_ids) == 0:
3010
return self._pack(self.source, self.target, revision_ids)
3012
def _pack(self, source, target, revision_ids):
3013
from bzrlib.repofmt.pack_repo import Packer
3014
target_pack_collection = self._get_target_pack_collection()
3015
packs = source._pack_collection.all_packs()
3016
pack = Packer(target_pack_collection, packs, '.fetch',
3017
revision_ids).pack()
3018
if pack is not None:
3019
target_pack_collection._save_pack_names()
3020
copied_revs = pack.get_revision_count()
3021
# Trigger an autopack. This may duplicate effort as we've just done
3022
# a pack creation, but for now it is simpler to think about as
3023
# 'upload data, then repack if needed'.
3025
return (copied_revs, [])
3029
def _autopack(self):
3030
self.target._pack_collection.autopack()
3032
def _get_target_pack_collection(self):
3033
return self.target._pack_collection
3036
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
3037
"""See InterRepository.missing_revision_ids().
3039
:param find_ghosts: Find ghosts throughout the ancestry of
3042
if not find_ghosts and revision_id is not None:
3043
return self._walk_to_common_revisions([revision_id])
3044
elif revision_id is not None:
3045
# Find ghosts: search for revisions pointing from one repository to
3046
# the other, and vice versa, anywhere in the history of revision_id.
3047
graph = self.target_get_graph(other_repository=self.source)
3048
searcher = graph._make_breadth_first_searcher([revision_id])
3052
next_revs, ghosts = searcher.next_with_ghosts()
3053
except StopIteration:
3055
if revision_id in ghosts:
3056
raise errors.NoSuchRevision(self.source, revision_id)
3057
found_ids.update(next_revs)
3058
found_ids.update(ghosts)
3059
found_ids = frozenset(found_ids)
3060
# Double query here: should be able to avoid this by changing the
3061
# graph api further.
3062
result_set = found_ids - frozenset(
3063
self.target_get_parent_map(found_ids))
3065
source_ids = self.source.all_revision_ids()
3066
# source_ids is the worst possible case we may need to pull.
3067
# now we want to filter source_ids against what we actually
3068
# have in target, but don't try to check for existence where we know
3069
# we do not have a revision as that would be pointless.
3070
target_ids = set(self.target.all_revision_ids())
3071
result_set = set(source_ids).difference(target_ids)
3072
return self.source.revision_ids_to_search_result(result_set)
3075
class InterDifferingSerializer(InterKnitRepo):
3712
3078
def _get_repo_format_to_test(self):
3716
3082
def is_compatible(source, target):
3717
3083
"""Be compatible with Knit2 source and Knit3 target"""
3718
# This is redundant with format.check_conversion_target(), however that
3719
# raises an exception, and we just want to say "False" as in we won't
3720
# support converting between these formats.
3721
if 'IDS_never' in debug.debug_flags:
3723
if source.supports_rich_root() and not target.supports_rich_root():
3725
if (source._format.supports_tree_reference
3726
and not target._format.supports_tree_reference):
3728
if target._fallback_repositories and target._format.supports_chks:
3729
# IDS doesn't know how to copy CHKs for the parent inventories it
3730
# adds to stacked repos.
3732
if 'IDS_always' in debug.debug_flags:
3734
# Only use this code path for local source and target. IDS does far
3735
# too much IO (both bandwidth and roundtrips) over a network.
3736
if not source.bzrdir.transport.base.startswith('file:///'):
3738
if not target.bzrdir.transport.base.startswith('file:///'):
3084
if source.supports_rich_root() != target.supports_rich_root():
3086
# Ideally, we'd support fetching if the source had no tree references
3087
# even if it supported them...
3088
if (getattr(source, '_format.supports_tree_reference', False) and
3089
not getattr(target, '_format.supports_tree_reference', False)):
3742
def _get_trees(self, revision_ids, cache):
3744
for rev_id in revision_ids:
3746
possible_trees.append((rev_id, cache[rev_id]))
3748
# Not cached, but inventory might be present anyway.
3750
tree = self.source.revision_tree(rev_id)
3751
except errors.NoSuchRevision:
3752
# Nope, parent is ghost.
3755
cache[rev_id] = tree
3756
possible_trees.append((rev_id, tree))
3757
return possible_trees
3759
def _get_delta_for_revision(self, tree, parent_ids, possible_trees):
3093
def _get_delta_for_revision(self, tree, parent_ids, basis_id, cache):
3760
3094
"""Get the best delta and base for this revision.
3762
3096
:return: (basis_id, delta)
3098
possible_trees = [(parent_id, cache[parent_id])
3099
for parent_id in parent_ids
3100
if parent_id in cache]
3101
if len(possible_trees) == 0:
3102
# There either aren't any parents, or the parents aren't in the
3103
# cache, so just use the last converted tree
3104
possible_trees.append((basis_id, cache[basis_id]))
3765
# Generate deltas against each tree, to find the shortest.
3766
texts_possibly_new_in_tree = set()
3767
3106
for basis_id, basis_tree in possible_trees:
3768
3107
delta = tree.inventory._make_delta(basis_tree.inventory)
3769
for old_path, new_path, file_id, new_entry in delta:
3770
if new_path is None:
3771
# This file_id isn't present in the new rev, so we don't
3775
# Rich roots are handled elsewhere...
3777
kind = new_entry.kind
3778
if kind != 'directory' and kind != 'file':
3779
# No text record associated with this inventory entry.
3781
# This is a directory or file that has changed somehow.
3782
texts_possibly_new_in_tree.add((file_id, new_entry.revision))
3783
3108
deltas.append((len(delta), basis_id, delta))
3785
3110
return deltas[0][1:]
3787
def _fetch_parent_invs_for_stacking(self, parent_map, cache):
3788
"""Find all parent revisions that are absent, but for which the
3789
inventory is present, and copy those inventories.
3791
This is necessary to preserve correctness when the source is stacked
3792
without fallbacks configured. (Note that in cases like upgrade the
3793
source may be not have _fallback_repositories even though it is
3797
for parents in parent_map.values():
3798
parent_revs.update(parents)
3799
present_parents = self.source.get_parent_map(parent_revs)
3800
absent_parents = set(parent_revs).difference(present_parents)
3801
parent_invs_keys_for_stacking = self.source.inventories.get_parent_map(
3802
(rev_id,) for rev_id in absent_parents)
3803
parent_inv_ids = [key[-1] for key in parent_invs_keys_for_stacking]
3804
for parent_tree in self.source.revision_trees(parent_inv_ids):
3805
current_revision_id = parent_tree.get_revision_id()
3806
parents_parents_keys = parent_invs_keys_for_stacking[
3807
(current_revision_id,)]
3808
parents_parents = [key[-1] for key in parents_parents_keys]
3809
basis_id = _mod_revision.NULL_REVISION
3810
basis_tree = self.source.revision_tree(basis_id)
3811
delta = parent_tree.inventory._make_delta(basis_tree.inventory)
3812
self.target.add_inventory_by_delta(
3813
basis_id, delta, current_revision_id, parents_parents)
3814
cache[current_revision_id] = parent_tree
3816
def _fetch_batch(self, revision_ids, basis_id, cache, a_graph=None):
3112
def _fetch_batch(self, revision_ids, basis_id, cache):
3817
3113
"""Fetch across a few revisions.
3819
3115
:param revision_ids: The revisions to copy
3820
3116
:param basis_id: The revision_id of a tree that must be in cache, used
3821
3117
as a basis for delta when no other base is available
3822
3118
:param cache: A cache of RevisionTrees that we can use.
3823
:param a_graph: A Graph object to determine the heads() of the
3824
rich-root data stream.
3825
3119
:return: The revision_id of the last converted tree. The RevisionTree
3826
3120
for it will be in cache
3828
3122
# Walk though all revisions; get inventory deltas, copy referenced
3829
3123
# texts that delta references, insert the delta, revision and
3831
root_keys_to_create = set()
3832
3125
text_keys = set()
3833
3126
pending_deltas = []
3834
3127
pending_revisions = []
3835
3128
parent_map = self.source.get_parent_map(revision_ids)
3836
self._fetch_parent_invs_for_stacking(parent_map, cache)
3837
self.source._safe_to_return_from_cache = True
3838
3129
for tree in self.source.revision_trees(revision_ids):
3839
# Find a inventory delta for this revision.
3840
# Find text entries that need to be copied, too.
3841
3130
current_revision_id = tree.get_revision_id()
3842
3131
parent_ids = parent_map.get(current_revision_id, ())
3843
parent_trees = self._get_trees(parent_ids, cache)
3844
possible_trees = list(parent_trees)
3845
if len(possible_trees) == 0:
3846
# There either aren't any parents, or the parents are ghosts,
3847
# so just use the last converted tree.
3848
possible_trees.append((basis_id, cache[basis_id]))
3849
3132
basis_id, delta = self._get_delta_for_revision(tree, parent_ids,
3134
# Find text entries that need to be copied
3135
for old_path, new_path, file_id, entry in delta:
3136
if new_path is not None:
3137
if not (new_path or self.target.supports_rich_root()):
3138
# We don't copy the text for the root node unless the
3139
# target supports_rich_root.
3142
# "if entry.revision == current_revision_id" ?
3143
if entry.revision == current_revision_id:
3144
text_keys.add((file_id, entry.revision))
3851
3145
revision = self.source.get_revision(current_revision_id)
3852
3146
pending_deltas.append((basis_id, delta,
3853
3147
current_revision_id, revision.parent_ids))
3854
if self._converting_to_rich_root:
3855
self._revision_id_to_root_id[current_revision_id] = \
3857
# Determine which texts are in present in this revision but not in
3858
# any of the available parents.
3859
texts_possibly_new_in_tree = set()
3860
for old_path, new_path, file_id, entry in delta:
3861
if new_path is None:
3862
# This file_id isn't present in the new rev
3866
if not self.target.supports_rich_root():
3867
# The target doesn't support rich root, so we don't
3870
if self._converting_to_rich_root:
3871
# This can't be copied normally, we have to insert
3873
root_keys_to_create.add((file_id, entry.revision))
3876
texts_possibly_new_in_tree.add((file_id, entry.revision))
3877
for basis_id, basis_tree in possible_trees:
3878
basis_inv = basis_tree.inventory
3879
for file_key in list(texts_possibly_new_in_tree):
3880
file_id, file_revision = file_key
3882
entry = basis_inv[file_id]
3883
except errors.NoSuchId:
3885
if entry.revision == file_revision:
3886
texts_possibly_new_in_tree.remove(file_key)
3887
text_keys.update(texts_possibly_new_in_tree)
3888
3148
pending_revisions.append(revision)
3889
3149
cache[current_revision_id] = tree
3890
3150
basis_id = current_revision_id
3891
self.source._safe_to_return_from_cache = False
3892
3151
# Copy file texts
3893
3152
from_texts = self.source.texts
3894
3153
to_texts = self.target.texts
3895
if root_keys_to_create:
3896
root_stream = _mod_fetch._new_root_data_stream(
3897
root_keys_to_create, self._revision_id_to_root_id, parent_map,
3898
self.source, graph=a_graph)
3899
to_texts.insert_record_stream(root_stream)
3900
3154
to_texts.insert_record_stream(from_texts.get_record_stream(
3901
3155
text_keys, self.target._format._fetch_order,
3902
3156
not self.target._format._fetch_uses_deltas))
3903
# insert inventory deltas
3904
3158
for delta in pending_deltas:
3905
3159
self.target.add_inventory_by_delta(*delta)
3906
if self.target._fallback_repositories:
3907
# Make sure this stacked repository has all the parent inventories
3908
# for the new revisions that we are about to insert. We do this
3909
# before adding the revisions so that no revision is added until
3910
# all the inventories it may depend on are added.
3911
# Note that this is overzealous, as we may have fetched these in an
3914
revision_ids = set()
3915
for revision in pending_revisions:
3916
revision_ids.add(revision.revision_id)
3917
parent_ids.update(revision.parent_ids)
3918
parent_ids.difference_update(revision_ids)
3919
parent_ids.discard(_mod_revision.NULL_REVISION)
3920
parent_map = self.source.get_parent_map(parent_ids)
3921
# we iterate over parent_map and not parent_ids because we don't
3922
# want to try copying any revision which is a ghost
3923
for parent_tree in self.source.revision_trees(parent_map):
3924
current_revision_id = parent_tree.get_revision_id()
3925
parents_parents = parent_map[current_revision_id]
3926
possible_trees = self._get_trees(parents_parents, cache)
3927
if len(possible_trees) == 0:
3928
# There either aren't any parents, or the parents are
3929
# ghosts, so just use the last converted tree.
3930
possible_trees.append((basis_id, cache[basis_id]))
3931
basis_id, delta = self._get_delta_for_revision(parent_tree,
3932
parents_parents, possible_trees)
3933
self.target.add_inventory_by_delta(
3934
basis_id, delta, current_revision_id, parents_parents)
3935
3160
# insert signatures and revisions
3936
3161
for revision in pending_revisions:
3957
3182
cache = lru_cache.LRUCache(100)
3958
3183
cache[basis_id] = basis_tree
3959
3184
del basis_tree # We don't want to hang on to it here
3961
if self._converting_to_rich_root and len(revision_ids) > 100:
3962
a_graph = _mod_fetch._get_rich_root_heads_graph(self.source,
3967
3185
for offset in range(0, len(revision_ids), batch_size):
3968
3186
self.target.start_write_group()
3970
3188
pb.update('Transferring revisions', offset,
3971
3189
len(revision_ids))
3972
3190
batch = revision_ids[offset:offset+batch_size]
3973
basis_id = self._fetch_batch(batch, basis_id, cache,
3191
basis_id = self._fetch_batch(batch, basis_id, cache)
3976
self.source._safe_to_return_from_cache = False
3977
3193
self.target.abort_write_group()
3980
hint = self.target.commit_write_group()
3983
if hints and self.target._format.pack_compresses:
3984
self.target.pack(hint=hints)
3196
self.target.commit_write_group()
3985
3197
pb.update('Transferring revisions', len(revision_ids),
3986
3198
len(revision_ids))
3988
3200
@needs_write_lock
3989
def fetch(self, revision_id=None, pb=None, find_ghosts=False,
3201
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3991
3202
"""See InterRepository.fetch()."""
3992
if fetch_spec is not None:
3993
raise AssertionError("Not implemented yet...")
3994
ui.ui_factory.warn_experimental_format_fetch(self)
3995
if (not self.source.supports_rich_root()
3996
and self.target.supports_rich_root()):
3997
self._converting_to_rich_root = True
3998
self._revision_id_to_root_id = {}
4000
self._converting_to_rich_root = False
4001
# See <https://launchpad.net/bugs/456077> asking for a warning here
4002
if self.source._format.network_name() != self.target._format.network_name():
4003
ui.ui_factory.show_user_warning('cross_format_fetch',
4004
from_format=self.source._format,
4005
to_format=self.target._format)
4006
3203
revision_ids = self.target.search_missing_revision_ids(self.source,
4007
3204
revision_id, find_ghosts=find_ghosts).get_keys()
4008
3205
if not revision_ids:
4010
3207
revision_ids = tsort.topo_sort(
4011
3208
self.source.get_graph().get_parent_map(revision_ids))
4012
if not revision_ids:
4014
# Walk though all revisions; get inventory deltas, copy referenced
4015
# texts that delta references, insert the delta, revision and
4018
3210
my_pb = ui.ui_factory.nested_progress_bar()
4021
symbol_versioning.warn(
4022
symbol_versioning.deprecated_in((1, 14, 0))
4023
% "pb parameter to fetch()")
4026
3215
self._fetch_all_revisions(revision_ids, pb)
4052
3241
return basis_id, basis_tree
3244
class InterOtherToRemote(InterRepository):
3245
"""An InterRepository that simply delegates to the 'real' InterRepository
3246
calculated for (source, target._real_repository).
3249
_walk_to_common_revisions_batch_size = 50
3251
def __init__(self, source, target):
3252
InterRepository.__init__(self, source, target)
3253
self._real_inter = None
3256
def is_compatible(source, target):
3257
if isinstance(target, remote.RemoteRepository):
3261
def _ensure_real_inter(self):
3262
if self._real_inter is None:
3263
self.target._ensure_real()
3264
real_target = self.target._real_repository
3265
self._real_inter = InterRepository.get(self.source, real_target)
3266
# Make _real_inter use the RemoteRepository for get_parent_map
3267
self._real_inter.target_get_graph = self.target.get_graph
3268
self._real_inter.target_get_parent_map = self.target.get_parent_map
3270
def copy_content(self, revision_id=None):
3271
self._ensure_real_inter()
3272
self._real_inter.copy_content(revision_id=revision_id)
3274
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3275
self._ensure_real_inter()
3276
return self._real_inter.fetch(revision_id=revision_id, pb=pb,
3277
find_ghosts=find_ghosts)
3280
def _get_repo_format_to_test(self):
3284
class InterRemoteToOther(InterRepository):
3286
def __init__(self, source, target):
3287
InterRepository.__init__(self, source, target)
3288
self._real_inter = None
3291
def is_compatible(source, target):
3292
if not isinstance(source, remote.RemoteRepository):
3294
return InterRepository._same_model(source, target)
3296
def _ensure_real_inter(self):
3297
if self._real_inter is None:
3298
self.source._ensure_real()
3299
real_source = self.source._real_repository
3300
self._real_inter = InterRepository.get(real_source, self.target)
3303
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3304
"""See InterRepository.fetch()."""
3305
# Always fetch using the generic streaming fetch code, to allow
3306
# streaming fetching from remote servers.
3307
from bzrlib.fetch import RepoFetcher
3308
fetcher = RepoFetcher(self.target, self.source, revision_id,
3311
def copy_content(self, revision_id=None):
3312
self._ensure_real_inter()
3313
self._real_inter.copy_content(revision_id=revision_id)
3316
def _get_repo_format_to_test(self):
3321
class InterPackToRemotePack(InterPackRepo):
3322
"""A specialisation of InterPackRepo for a target that is a
3325
This will use the get_parent_map RPC rather than plain readvs, and also
3326
uses an RPC for autopacking.
3329
_walk_to_common_revisions_batch_size = 50
3332
def is_compatible(source, target):
3333
from bzrlib.repofmt.pack_repo import RepositoryFormatPack
3334
if isinstance(source._format, RepositoryFormatPack):
3335
if isinstance(target, remote.RemoteRepository):
3336
target._format._ensure_real()
3337
if isinstance(target._format._custom_format,
3338
RepositoryFormatPack):
3339
if InterRepository._same_model(source, target):
3343
def _autopack(self):
3344
self.target.autopack()
3347
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3348
"""See InterRepository.fetch()."""
3349
# Always fetch using the generic streaming fetch code, to allow
3350
# streaming fetching into remote servers.
3351
from bzrlib.fetch import RepoFetcher
3352
fetcher = RepoFetcher(self.target, self.source, revision_id,
3355
def _get_target_pack_collection(self):
3356
return self.target._real_repository._pack_collection
3359
def _get_repo_format_to_test(self):
4055
3363
InterRepository.register_optimiser(InterDifferingSerializer)
4056
3364
InterRepository.register_optimiser(InterSameDataRepository)
4057
3365
InterRepository.register_optimiser(InterWeaveRepo)
4058
3366
InterRepository.register_optimiser(InterKnitRepo)
3367
InterRepository.register_optimiser(InterPackRepo)
3368
InterRepository.register_optimiser(InterOtherToRemote)
3369
InterRepository.register_optimiser(InterRemoteToOther)
3370
InterRepository.register_optimiser(InterPackToRemotePack)
4061
3373
class CopyConverter(object):
4350
3610
# missing keys can handle suspending a write group).
4351
3611
write_group_tokens = self.target_repo.suspend_write_group()
4352
3612
return write_group_tokens, missing_keys
4353
hint = self.target_repo.commit_write_group()
4354
if (to_serializer != src_serializer and
4355
self.target_repo._format.pack_compresses):
4356
self.target_repo.pack(hint=hint)
3613
self.target_repo.commit_write_group()
4357
3614
return [], set()
4359
def _extract_and_insert_inventory_deltas(self, substream, serializer):
4360
target_rich_root = self.target_repo._format.rich_root_data
4361
target_tree_refs = self.target_repo._format.supports_tree_reference
4362
for record in substream:
4363
# Insert the delta directly
4364
inventory_delta_bytes = record.get_bytes_as('fulltext')
4365
deserialiser = inventory_delta.InventoryDeltaDeserializer()
4367
parse_result = deserialiser.parse_text_bytes(
4368
inventory_delta_bytes)
4369
except inventory_delta.IncompatibleInventoryDelta, err:
4370
trace.mutter("Incompatible delta: %s", err.msg)
4371
raise errors.IncompatibleRevision(self.target_repo._format)
4372
basis_id, new_id, rich_root, tree_refs, inv_delta = parse_result
4373
revision_id = new_id
4374
parents = [key[0] for key in record.parents]
4375
self.target_repo.add_inventory_by_delta(
4376
basis_id, inv_delta, revision_id, parents)
4378
def _extract_and_insert_inventories(self, substream, serializer,
3616
def _extract_and_insert_inventories(self, substream, serializer):
4380
3617
"""Generate a new inventory versionedfile in target, converting data.
4382
3619
The inventory is retrieved from the source, (deserializing it), and
4383
3620
stored in the target (reserializing it in a different format).
4385
target_rich_root = self.target_repo._format.rich_root_data
4386
target_tree_refs = self.target_repo._format.supports_tree_reference
4387
3622
for record in substream:
4388
# It's not a delta, so it must be a fulltext in the source
4389
# serializer's format.
4390
3623
bytes = record.get_bytes_as('fulltext')
4391
3624
revision_id = record.key[0]
4392
3625
inv = serializer.read_inventory_from_string(bytes, revision_id)
4393
3626
parents = [key[0] for key in record.parents]
4394
3627
self.target_repo.add_inventory(revision_id, inv, parents)
4395
# No need to keep holding this full inv in memory when the rest of
4396
# the substream is likely to be all deltas.
4399
3629
def _extract_and_insert_revisions(self, substream, serializer):
4400
3630
for record in substream:
4563
3778
return (not self.from_repository._format.rich_root_data and
4564
3779
self.to_format.rich_root_data)
4566
def _get_inventory_stream(self, revision_ids, missing=False):
4567
from_format = self.from_repository._format
4568
if (from_format.supports_chks and self.to_format.supports_chks and
4569
from_format.network_name() == self.to_format.network_name()):
4570
raise AssertionError(
4571
"this case should be handled by GroupCHKStreamSource")
4572
elif 'forceinvdeltas' in debug.debug_flags:
4573
return self._get_convertable_inventory_stream(revision_ids,
4574
delta_versus_null=missing)
4575
elif from_format.network_name() == self.to_format.network_name():
4577
return self._get_simple_inventory_stream(revision_ids,
4579
elif (not from_format.supports_chks and not self.to_format.supports_chks
4580
and from_format._serializer == self.to_format._serializer):
4581
# Essentially the same format.
4582
return self._get_simple_inventory_stream(revision_ids,
4585
# Any time we switch serializations, we want to use an
4586
# inventory-delta based approach.
4587
return self._get_convertable_inventory_stream(revision_ids,
4588
delta_versus_null=missing)
4590
def _get_simple_inventory_stream(self, revision_ids, missing=False):
4591
# NB: This currently reopens the inventory weave in source;
4592
# using a single stream interface instead would avoid this.
4593
from_weave = self.from_repository.inventories
4595
delta_closure = True
4597
delta_closure = not self.delta_on_metadata()
4598
yield ('inventories', from_weave.get_record_stream(
4599
[(rev_id,) for rev_id in revision_ids],
4600
self.inventory_fetch_order(), delta_closure))
4602
def _get_convertable_inventory_stream(self, revision_ids,
4603
delta_versus_null=False):
4604
# The two formats are sufficiently different that there is no fast
4605
# path, so we need to send just inventorydeltas, which any
4606
# sufficiently modern client can insert into any repository.
4607
# The StreamSink code expects to be able to
4608
# convert on the target, so we need to put bytes-on-the-wire that can
4609
# be converted. That means inventory deltas (if the remote is <1.19,
4610
# RemoteStreamSink will fallback to VFS to insert the deltas).
4611
yield ('inventory-deltas',
4612
self._stream_invs_as_deltas(revision_ids,
4613
delta_versus_null=delta_versus_null))
4615
def _stream_invs_as_deltas(self, revision_ids, delta_versus_null=False):
4616
"""Return a stream of inventory-deltas for the given rev ids.
4618
:param revision_ids: The list of inventories to transmit
4619
:param delta_versus_null: Don't try to find a minimal delta for this
4620
entry, instead compute the delta versus the NULL_REVISION. This
4621
effectively streams a complete inventory. Used for stuff like
4622
filling in missing parents, etc.
4624
from_repo = self.from_repository
4625
revision_keys = [(rev_id,) for rev_id in revision_ids]
4626
parent_map = from_repo.inventories.get_parent_map(revision_keys)
4627
# XXX: possibly repos could implement a more efficient iter_inv_deltas
4629
inventories = self.from_repository.iter_inventories(
4630
revision_ids, 'topological')
4631
format = from_repo._format
4632
invs_sent_so_far = set([_mod_revision.NULL_REVISION])
4633
inventory_cache = lru_cache.LRUCache(50)
4634
null_inventory = from_repo.revision_tree(
4635
_mod_revision.NULL_REVISION).inventory
4636
# XXX: ideally the rich-root/tree-refs flags would be per-revision, not
4637
# per-repo (e.g. streaming a non-rich-root revision out of a rich-root
4638
# repo back into a non-rich-root repo ought to be allowed)
4639
serializer = inventory_delta.InventoryDeltaSerializer(
4640
versioned_root=format.rich_root_data,
4641
tree_references=format.supports_tree_reference)
4642
for inv in inventories:
4643
key = (inv.revision_id,)
4644
parent_keys = parent_map.get(key, ())
4646
if not delta_versus_null and parent_keys:
4647
# The caller did not ask for complete inventories and we have
4648
# some parents that we can delta against. Make a delta against
4649
# each parent so that we can find the smallest.
4650
parent_ids = [parent_key[0] for parent_key in parent_keys]
4651
for parent_id in parent_ids:
4652
if parent_id not in invs_sent_so_far:
4653
# We don't know that the remote side has this basis, so
4656
if parent_id == _mod_revision.NULL_REVISION:
4657
parent_inv = null_inventory
4659
parent_inv = inventory_cache.get(parent_id, None)
4660
if parent_inv is None:
4661
parent_inv = from_repo.get_inventory(parent_id)
4662
candidate_delta = inv._make_delta(parent_inv)
4663
if (delta is None or
4664
len(delta) > len(candidate_delta)):
4665
delta = candidate_delta
4666
basis_id = parent_id
4668
# Either none of the parents ended up being suitable, or we
4669
# were asked to delta against NULL
4670
basis_id = _mod_revision.NULL_REVISION
4671
delta = inv._make_delta(null_inventory)
4672
invs_sent_so_far.add(inv.revision_id)
4673
inventory_cache[inv.revision_id] = inv
4674
delta_serialized = ''.join(
4675
serializer.delta_to_lines(basis_id, key[-1], delta))
4676
yield versionedfile.FulltextContentFactory(
4677
key, parent_keys, None, delta_serialized)
4680
def _iter_for_revno(repo, partial_history_cache, stop_index=None,
4681
stop_revision=None):
4682
"""Extend the partial history to include a given index
4684
If a stop_index is supplied, stop when that index has been reached.
4685
If a stop_revision is supplied, stop when that revision is
4686
encountered. Otherwise, stop when the beginning of history is
4689
:param stop_index: The index which should be present. When it is
4690
present, history extension will stop.
4691
:param stop_revision: The revision id which should be present. When
4692
it is encountered, history extension will stop.
4694
start_revision = partial_history_cache[-1]
4695
iterator = repo.iter_reverse_revision_history(start_revision)
4697
#skip the last revision in the list
4700
if (stop_index is not None and
4701
len(partial_history_cache) > stop_index):
4703
if partial_history_cache[-1] == stop_revision:
4705
revision_id = iterator.next()
4706
partial_history_cache.append(revision_id)
4707
except StopIteration:
added
removed
bzrlib/repository.py
