558
492
ie.revision = parent_entry.revision
559
493
return self._get_delta(ie, basis_inv, path), False, None
560
494
ie.reference_revision = content_summary[3]
561
if ie.reference_revision is None:
562
raise AssertionError("invalid content_summary for nested tree: %r"
563
% (content_summary,))
564
self._add_text_to_weave(ie.file_id, '', heads, None)
496
self._add_text_to_weave(ie.file_id, lines, heads, None)
566
498
raise NotImplementedError('unknown kind')
567
499
ie.revision = self._new_revision_id
568
self._any_changes = True
569
500
return self._get_delta(ie, basis_inv, path), True, fingerprint
571
def record_iter_changes(self, tree, basis_revision_id, iter_changes,
572
_entry_factory=entry_factory):
573
"""Record a new tree via iter_changes.
575
:param tree: The tree to obtain text contents from for changed objects.
576
:param basis_revision_id: The revision id of the tree the iter_changes
577
has been generated against. Currently assumed to be the same
578
as self.parents[0] - if it is not, errors may occur.
579
:param iter_changes: An iter_changes iterator with the changes to apply
580
to basis_revision_id. The iterator must not include any items with
581
a current kind of None - missing items must be either filtered out
582
or errored-on beefore record_iter_changes sees the item.
583
:param _entry_factory: Private method to bind entry_factory locally for
585
:return: A generator of (file_id, relpath, fs_hash) tuples for use with
588
# Create an inventory delta based on deltas between all the parents and
589
# deltas between all the parent inventories. We use inventory delta's
590
# between the inventory objects because iter_changes masks
591
# last-changed-field only changes.
593
# file_id -> change map, change is fileid, paths, changed, versioneds,
594
# parents, names, kinds, executables
596
# {file_id -> revision_id -> inventory entry, for entries in parent
597
# trees that are not parents[0]
601
revtrees = list(self.repository.revision_trees(self.parents))
602
except errors.NoSuchRevision:
603
# one or more ghosts, slow path.
605
for revision_id in self.parents:
607
revtrees.append(self.repository.revision_tree(revision_id))
608
except errors.NoSuchRevision:
610
basis_revision_id = _mod_revision.NULL_REVISION
612
revtrees.append(self.repository.revision_tree(
613
_mod_revision.NULL_REVISION))
614
# The basis inventory from a repository
616
basis_inv = revtrees[0].inventory
618
basis_inv = self.repository.revision_tree(
619
_mod_revision.NULL_REVISION).inventory
620
if len(self.parents) > 0:
621
if basis_revision_id != self.parents[0] and not ghost_basis:
623
"arbitrary basis parents not yet supported with merges")
624
for revtree in revtrees[1:]:
625
for change in revtree.inventory._make_delta(basis_inv):
626
if change[1] is None:
627
# Not present in this parent.
629
if change[2] not in merged_ids:
630
if change[0] is not None:
631
basis_entry = basis_inv[change[2]]
632
merged_ids[change[2]] = [
634
basis_entry.revision,
637
parent_entries[change[2]] = {
639
basis_entry.revision:basis_entry,
641
change[3].revision:change[3],
644
merged_ids[change[2]] = [change[3].revision]
645
parent_entries[change[2]] = {change[3].revision:change[3]}
647
merged_ids[change[2]].append(change[3].revision)
648
parent_entries[change[2]][change[3].revision] = change[3]
651
# Setup the changes from the tree:
652
# changes maps file_id -> (change, [parent revision_ids])
654
for change in iter_changes:
655
# This probably looks up in basis_inv way to much.
656
if change[1][0] is not None:
657
head_candidate = [basis_inv[change[0]].revision]
660
changes[change[0]] = change, merged_ids.get(change[0],
662
unchanged_merged = set(merged_ids) - set(changes)
663
# Extend the changes dict with synthetic changes to record merges of
665
for file_id in unchanged_merged:
666
# Record a merged version of these items that did not change vs the
667
# basis. This can be either identical parallel changes, or a revert
668
# of a specific file after a merge. The recorded content will be
669
# that of the current tree (which is the same as the basis), but
670
# the per-file graph will reflect a merge.
671
# NB:XXX: We are reconstructing path information we had, this
672
# should be preserved instead.
673
# inv delta change: (file_id, (path_in_source, path_in_target),
674
# changed_content, versioned, parent, name, kind,
677
basis_entry = basis_inv[file_id]
678
except errors.NoSuchId:
679
# a change from basis->some_parents but file_id isn't in basis
680
# so was new in the merge, which means it must have changed
681
# from basis -> current, and as it hasn't the add was reverted
682
# by the user. So we discard this change.
686
(basis_inv.id2path(file_id), tree.id2path(file_id)),
688
(basis_entry.parent_id, basis_entry.parent_id),
689
(basis_entry.name, basis_entry.name),
690
(basis_entry.kind, basis_entry.kind),
691
(basis_entry.executable, basis_entry.executable))
692
changes[file_id] = (change, merged_ids[file_id])
693
# changes contains tuples with the change and a set of inventory
694
# candidates for the file.
696
# old_path, new_path, file_id, new_inventory_entry
697
seen_root = False # Is the root in the basis delta?
698
inv_delta = self._basis_delta
699
modified_rev = self._new_revision_id
700
for change, head_candidates in changes.values():
701
if change[3][1]: # versioned in target.
702
# Several things may be happening here:
703
# We may have a fork in the per-file graph
704
# - record a change with the content from tree
705
# We may have a change against < all trees
706
# - carry over the tree that hasn't changed
707
# We may have a change against all trees
708
# - record the change with the content from tree
711
entry = _entry_factory[kind](file_id, change[5][1],
713
head_set = self._heads(change[0], set(head_candidates))
716
for head_candidate in head_candidates:
717
if head_candidate in head_set:
718
heads.append(head_candidate)
719
head_set.remove(head_candidate)
722
# Could be a carry-over situation:
723
parent_entry_revs = parent_entries.get(file_id, None)
724
if parent_entry_revs:
725
parent_entry = parent_entry_revs.get(heads[0], None)
728
if parent_entry is None:
729
# The parent iter_changes was called against is the one
730
# that is the per-file head, so any change is relevant
731
# iter_changes is valid.
732
carry_over_possible = False
734
# could be a carry over situation
735
# A change against the basis may just indicate a merge,
736
# we need to check the content against the source of the
737
# merge to determine if it was changed after the merge
739
if (parent_entry.kind != entry.kind or
740
parent_entry.parent_id != entry.parent_id or
741
parent_entry.name != entry.name):
742
# Metadata common to all entries has changed
743
# against per-file parent
744
carry_over_possible = False
746
carry_over_possible = True
747
# per-type checks for changes against the parent_entry
750
# Cannot be a carry-over situation
751
carry_over_possible = False
752
# Populate the entry in the delta
754
# XXX: There is still a small race here: If someone reverts the content of a file
755
# after iter_changes examines and decides it has changed,
756
# we will unconditionally record a new version even if some
757
# other process reverts it while commit is running (with
758
# the revert happening after iter_changes did it's
761
entry.executable = True
763
entry.executable = False
764
if (carry_over_possible and
765
parent_entry.executable == entry.executable):
766
# Check the file length, content hash after reading
768
nostore_sha = parent_entry.text_sha1
771
file_obj, stat_value = tree.get_file_with_stat(file_id, change[1][1])
773
text = file_obj.read()
777
entry.text_sha1, entry.text_size = self._add_text_to_weave(
778
file_id, text, heads, nostore_sha)
779
yield file_id, change[1][1], (entry.text_sha1, stat_value)
780
except errors.ExistingContent:
781
# No content change against a carry_over parent
782
# Perhaps this should also yield a fs hash update?
784
entry.text_size = parent_entry.text_size
785
entry.text_sha1 = parent_entry.text_sha1
786
elif kind == 'symlink':
788
entry.symlink_target = tree.get_symlink_target(file_id)
789
if (carry_over_possible and
790
parent_entry.symlink_target == entry.symlink_target):
793
self._add_text_to_weave(change[0], '', heads, None)
794
elif kind == 'directory':
795
if carry_over_possible:
798
# Nothing to set on the entry.
799
# XXX: split into the Root and nonRoot versions.
800
if change[1][1] != '' or self.repository.supports_rich_root():
801
self._add_text_to_weave(change[0], '', heads, None)
802
elif kind == 'tree-reference':
803
if not self.repository._format.supports_tree_reference:
804
# This isn't quite sane as an error, but we shouldn't
805
# ever see this code path in practice: tree's don't
806
# permit references when the repo doesn't support tree
808
raise errors.UnsupportedOperation(tree.add_reference,
810
reference_revision = tree.get_reference_revision(change[0])
811
entry.reference_revision = reference_revision
812
if (carry_over_possible and
813
parent_entry.reference_revision == reference_revision):
816
self._add_text_to_weave(change[0], '', heads, None)
818
raise AssertionError('unknown kind %r' % kind)
820
entry.revision = modified_rev
822
entry.revision = parent_entry.revision
825
new_path = change[1][1]
826
inv_delta.append((change[1][0], new_path, change[0], entry))
829
self.new_inventory = None
831
# This should perhaps be guarded by a check that the basis we
832
# commit against is the basis for the commit and if not do a delta
834
self._any_changes = True
836
# housekeeping root entry changes do not affect no-change commits.
837
self._require_root_change(tree)
838
self.basis_delta_revision = basis_revision_id
840
def _add_text_to_weave(self, file_id, new_text, parents, nostore_sha):
841
parent_keys = tuple([(file_id, parent) for parent in parents])
842
return self.repository.texts._add_text(
843
(file_id, self._new_revision_id), parent_keys, new_text,
844
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]
847
515
class RootCommitBuilder(CommitBuilder):
1195
797
# The old API returned a list, should this actually be a set?
1196
798
return parent_map.keys()
1198
def _check_inventories(self, checker):
1199
"""Check the inventories found from the revision scan.
1201
This is responsible for verifying the sha1 of inventories and
1202
creating a pending_keys set that covers data referenced by inventories.
1204
bar = ui.ui_factory.nested_progress_bar()
1206
self._do_check_inventories(checker, bar)
1210
def _do_check_inventories(self, checker, bar):
1211
"""Helper for _check_inventories."""
1213
keys = {'chk_bytes':set(), 'inventories':set(), 'texts':set()}
1214
kinds = ['chk_bytes', 'texts']
1215
count = len(checker.pending_keys)
1216
bar.update("inventories", 0, 2)
1217
current_keys = checker.pending_keys
1218
checker.pending_keys = {}
1219
# Accumulate current checks.
1220
for key in current_keys:
1221
if key[0] != 'inventories' and key[0] not in kinds:
1222
checker._report_items.append('unknown key type %r' % (key,))
1223
keys[key[0]].add(key[1:])
1224
if keys['inventories']:
1225
# NB: output order *should* be roughly sorted - topo or
1226
# inverse topo depending on repository - either way decent
1227
# to just delta against. However, pre-CHK formats didn't
1228
# try to optimise inventory layout on disk. As such the
1229
# pre-CHK code path does not use inventory deltas.
1231
for record in self.inventories.check(keys=keys['inventories']):
1232
if record.storage_kind == 'absent':
1233
checker._report_items.append(
1234
'Missing inventory {%s}' % (record.key,))
1236
last_object = self._check_record('inventories', record,
1237
checker, last_object,
1238
current_keys[('inventories',) + record.key])
1239
del keys['inventories']
1242
bar.update("texts", 1)
1243
while (checker.pending_keys or keys['chk_bytes']
1245
# Something to check.
1246
current_keys = checker.pending_keys
1247
checker.pending_keys = {}
1248
# Accumulate current checks.
1249
for key in current_keys:
1250
if key[0] not in kinds:
1251
checker._report_items.append('unknown key type %r' % (key,))
1252
keys[key[0]].add(key[1:])
1253
# Check the outermost kind only - inventories || chk_bytes || texts
1257
for record in getattr(self, kind).check(keys=keys[kind]):
1258
if record.storage_kind == 'absent':
1259
checker._report_items.append(
1260
'Missing %s {%s}' % (kind, record.key,))
1262
last_object = self._check_record(kind, record,
1263
checker, last_object, current_keys[(kind,) + record.key])
1267
def _check_record(self, kind, record, checker, last_object, item_data):
1268
"""Check a single text from this repository."""
1269
if kind == 'inventories':
1270
rev_id = record.key[0]
1271
inv = self._deserialise_inventory(rev_id,
1272
record.get_bytes_as('fulltext'))
1273
if last_object is not None:
1274
delta = inv._make_delta(last_object)
1275
for old_path, path, file_id, ie in delta:
1278
ie.check(checker, rev_id, inv)
1280
for path, ie in inv.iter_entries():
1281
ie.check(checker, rev_id, inv)
1282
if self._format.fast_deltas:
1284
elif kind == 'chk_bytes':
1285
# No code written to check chk_bytes for this repo format.
1286
checker._report_items.append(
1287
'unsupported key type chk_bytes for %s' % (record.key,))
1288
elif kind == 'texts':
1289
self._check_text(record, checker, item_data)
1291
checker._report_items.append(
1292
'unknown key type %s for %s' % (kind, record.key))
1294
def _check_text(self, record, checker, item_data):
1295
"""Check a single text."""
1296
# Check it is extractable.
1297
# TODO: check length.
1298
if record.storage_kind == 'chunked':
1299
chunks = record.get_bytes_as(record.storage_kind)
1300
sha1 = osutils.sha_strings(chunks)
1301
length = sum(map(len, chunks))
1303
content = record.get_bytes_as('fulltext')
1304
sha1 = osutils.sha_string(content)
1305
length = len(content)
1306
if item_data and sha1 != item_data[1]:
1307
checker._report_items.append(
1308
'sha1 mismatch: %s has sha1 %s expected %s referenced by %s' %
1309
(record.key, sha1, item_data[1], item_data[2]))
1312
801
def create(a_bzrdir):
1313
802
"""Construct the current default format repository in a_bzrdir."""
1333
823
self._reconcile_does_inventory_gc = True
1334
824
self._reconcile_fixes_text_parents = False
1335
825
self._reconcile_backsup_inventory = True
826
# not right yet - should be more semantically clear ?
828
# TODO: make sure to construct the right store classes, etc, depending
829
# on whether escaping is required.
830
self._warn_if_deprecated()
1336
831
self._write_group = None
1337
832
# Additional places to query for data.
1338
833
self._fallback_repositories = []
834
# What order should fetch operations request streams in?
835
# The default is unordered as that is the cheapest for an origin to
837
self._fetch_order = 'unordered'
838
# Does this repository use deltas that can be fetched as-deltas ?
839
# (E.g. knits, where the knit deltas can be transplanted intact.
840
# We default to False, which will ensure that enough data to get
841
# a full text out of any fetch stream will be grabbed.
842
self._fetch_uses_deltas = False
843
# Should fetch trigger a reconcile after the fetch? Only needed for
844
# some repository formats that can suffer internal inconsistencies.
845
self._fetch_reconcile = False
1339
846
# An InventoryEntry cache, used during deserialization
1340
847
self._inventory_entry_cache = fifo_cache.FIFOCache(10*1024)
1341
# Is it safe to return inventory entries directly from the entry cache,
1342
# rather copying them?
1343
self._safe_to_return_from_cache = False
1346
def user_transport(self):
1347
return self.bzrdir.user_transport
1350
def control_transport(self):
1351
return self._transport
1353
849
def __repr__(self):
1354
if self._fallback_repositories:
1355
return '%s(%r, fallback_repositories=%r)' % (
1356
self.__class__.__name__,
1358
self._fallback_repositories)
1360
return '%s(%r)' % (self.__class__.__name__,
1363
def _has_same_fallbacks(self, other_repo):
1364
"""Returns true if the repositories have the same fallbacks."""
1365
my_fb = self._fallback_repositories
1366
other_fb = other_repo._fallback_repositories
1367
if len(my_fb) != len(other_fb):
1369
for f, g in zip(my_fb, other_fb):
1370
if not f.has_same_location(g):
850
return '%s(%r)' % (self.__class__.__name__,
1374
853
def has_same_location(self, other):
1375
854
"""Returns a boolean indicating if this repository is at the same
1612
1097
def suspend_write_group(self):
1613
1098
raise errors.UnsuspendableWriteGroup(self)
1615
def get_missing_parent_inventories(self, check_for_missing_texts=True):
1616
"""Return the keys of missing inventory parents for revisions added in
1619
A revision is not complete if the inventory delta for that revision
1620
cannot be calculated. Therefore if the parent inventories of a
1621
revision are not present, the revision is incomplete, and e.g. cannot
1622
be streamed by a smart server. This method finds missing inventory
1623
parents for revisions added in this write group.
1625
if not self._format.supports_external_lookups:
1626
# This is only an issue for stacked repositories
1628
if not self.is_in_write_group():
1629
raise AssertionError('not in a write group')
1631
# XXX: We assume that every added revision already has its
1632
# corresponding inventory, so we only check for parent inventories that
1633
# might be missing, rather than all inventories.
1634
parents = set(self.revisions._index.get_missing_parents())
1635
parents.discard(_mod_revision.NULL_REVISION)
1636
unstacked_inventories = self.inventories._index
1637
present_inventories = unstacked_inventories.get_parent_map(
1638
key[-1:] for key in parents)
1639
parents.difference_update(present_inventories)
1640
if len(parents) == 0:
1641
# No missing parent inventories.
1643
if not check_for_missing_texts:
1644
return set(('inventories', rev_id) for (rev_id,) in parents)
1645
# Ok, now we have a list of missing inventories. But these only matter
1646
# if the inventories that reference them are missing some texts they
1647
# appear to introduce.
1648
# XXX: Texts referenced by all added inventories need to be present,
1649
# but at the moment we're only checking for texts referenced by
1650
# inventories at the graph's edge.
1651
key_deps = self.revisions._index._key_dependencies
1652
key_deps.satisfy_refs_for_keys(present_inventories)
1653
referrers = frozenset(r[0] for r in key_deps.get_referrers())
1654
file_ids = self.fileids_altered_by_revision_ids(referrers)
1655
missing_texts = set()
1656
for file_id, version_ids in file_ids.iteritems():
1657
missing_texts.update(
1658
(file_id, version_id) for version_id in version_ids)
1659
present_texts = self.texts.get_parent_map(missing_texts)
1660
missing_texts.difference_update(present_texts)
1661
if not missing_texts:
1662
# No texts are missing, so all revisions and their deltas are
1665
# Alternatively the text versions could be returned as the missing
1666
# keys, but this is likely to be less data.
1667
missing_keys = set(('inventories', rev_id) for (rev_id,) in parents)
1670
def refresh_data(self):
1671
"""Re-read any data needed to synchronise with disk.
1673
This method is intended to be called after another repository instance
1674
(such as one used by a smart server) has inserted data into the
1675
repository. On all repositories this will work outside of write groups.
1676
Some repository formats (pack and newer for bzrlib native formats)
1677
support refresh_data inside write groups. If called inside a write
1678
group on a repository that does not support refreshing in a write group
1679
IsInWriteGroupError will be raised.
1681
self._refresh_data()
1683
1100
def resume_write_group(self, tokens):
1684
1101
if not self.is_write_locked():
1685
1102
raise errors.NotWriteLocked(self)
1894
1278
@needs_read_lock
1895
1279
def get_revisions(self, revision_ids):
1896
"""Get many revisions at once.
1898
Repositories that need to check data on every revision read should
1899
subclass this method.
1280
"""Get many revisions at once."""
1901
1281
return self._get_revisions(revision_ids)
1903
1283
@needs_read_lock
1904
1284
def _get_revisions(self, revision_ids):
1905
1285
"""Core work logic to get many revisions without sanity checks."""
1286
for rev_id in revision_ids:
1287
if not rev_id or not isinstance(rev_id, basestring):
1288
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1289
keys = [(key,) for key in revision_ids]
1290
stream = self.revisions.get_record_stream(keys, 'unordered', True)
1907
for revid, rev in self._iter_revisions(revision_ids):
1909
raise errors.NoSuchRevision(self, revid)
1292
for record in stream:
1293
if record.storage_kind == 'absent':
1294
raise errors.NoSuchRevision(self, record.key[0])
1295
text = record.get_bytes_as('fulltext')
1296
rev = self._serializer.read_revision_from_string(text)
1297
revs[record.key[0]] = rev
1911
1298
return [revs[revid] for revid in revision_ids]
1913
def _iter_revisions(self, revision_ids):
1914
"""Iterate over revision objects.
1916
:param revision_ids: An iterable of revisions to examine. None may be
1917
passed to request all revisions known to the repository. Note that
1918
not all repositories can find unreferenced revisions; for those
1919
repositories only referenced ones will be returned.
1920
:return: An iterator of (revid, revision) tuples. Absent revisions (
1921
those asked for but not available) are returned as (revid, None).
1923
if revision_ids is None:
1924
revision_ids = self.all_revision_ids()
1926
for rev_id in revision_ids:
1927
if not rev_id or not isinstance(rev_id, basestring):
1928
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1929
keys = [(key,) for key in revision_ids]
1930
stream = self.revisions.get_record_stream(keys, 'unordered', True)
1931
for record in stream:
1932
revid = record.key[0]
1933
if record.storage_kind == 'absent':
1936
text = record.get_bytes_as('fulltext')
1937
rev = self._serializer.read_revision_from_string(text)
1940
def get_deltas_for_revisions(self, revisions, specific_fileids=None):
1301
def get_revision_xml(self, revision_id):
1302
# TODO: jam 20070210 This shouldn't be necessary since get_revision
1303
# would have already do it.
1304
# TODO: jam 20070210 Just use _serializer.write_revision_to_string()
1305
rev = self.get_revision(revision_id)
1306
rev_tmp = cStringIO.StringIO()
1307
# the current serializer..
1308
self._serializer.write_revision(rev, rev_tmp)
1310
return rev_tmp.getvalue()
1312
def get_deltas_for_revisions(self, revisions):
1941
1313
"""Produce a generator of revision deltas.
1943
1315
Note that the input is a sequence of REVISIONS, not revision_ids.
1944
1316
Trees will be held in memory until the generator exits.
1945
1317
Each delta is relative to the revision's lefthand predecessor.
1947
:param specific_fileids: if not None, the result is filtered
1948
so that only those file-ids, their parents and their
1949
children are included.
1951
# Get the revision-ids of interest
1952
1319
required_trees = set()
1953
1320
for revision in revisions:
1954
1321
required_trees.add(revision.revision_id)
1955
1322
required_trees.update(revision.parent_ids[:1])
1957
# Get the matching filtered trees. Note that it's more
1958
# efficient to pass filtered trees to changes_from() rather
1959
# than doing the filtering afterwards. changes_from() could
1960
# arguably do the filtering itself but it's path-based, not
1961
# file-id based, so filtering before or afterwards is
1963
if specific_fileids is None:
1964
trees = dict((t.get_revision_id(), t) for
1965
t in self.revision_trees(required_trees))
1967
trees = dict((t.get_revision_id(), t) for
1968
t in self._filtered_revision_trees(required_trees,
1971
# Calculate the deltas
1323
trees = dict((t.get_revision_id(), t) for
1324
t in self.revision_trees(required_trees))
1972
1325
for revision in revisions:
1973
1326
if not revision.parent_ids:
1974
1327
old_tree = self.revision_tree(_mod_revision.NULL_REVISION)
2098
1449
result[key] = True
2101
def _inventory_xml_lines_for_keys(self, keys):
2102
"""Get a line iterator of the sort needed for findind references.
2104
Not relevant for non-xml inventory repositories.
2106
Ghosts in revision_keys are ignored.
2108
:param revision_keys: The revision keys for the inventories to inspect.
2109
:return: An iterator over (inventory line, revid) for the fulltexts of
2110
all of the xml inventories specified by revision_keys.
2112
stream = self.inventories.get_record_stream(keys, 'unordered', True)
2113
for record in stream:
2114
if record.storage_kind != 'absent':
2115
chunks = record.get_bytes_as('chunked')
2116
revid = record.key[-1]
2117
lines = osutils.chunks_to_lines(chunks)
2121
1452
def _find_file_ids_from_xml_inventory_lines(self, line_iterator,
2123
1454
"""Helper routine for fileids_altered_by_revision_ids.
2125
1456
This performs the translation of xml lines to revision ids.
2127
1458
:param line_iterator: An iterator of lines, origin_version_id
2128
:param revision_keys: The revision ids to filter for. This should be a
1459
:param revision_ids: The revision ids to filter for. This should be a
2129
1460
set or other type which supports efficient __contains__ lookups, as
2130
the revision key from each parsed line will be looked up in the
2131
revision_keys filter.
1461
the revision id from each parsed line will be looked up in the
1462
revision_ids filter.
2132
1463
:return: a dictionary mapping altered file-ids to an iterable of
2133
1464
revision_ids. Each altered file-ids has the exact revision_ids that
2134
1465
altered it listed explicitly.
2136
seen = set(self._find_text_key_references_from_xml_inventory_lines(
2137
line_iterator).iterkeys())
2138
parent_keys = self._find_parent_keys_of_revisions(revision_keys)
2139
parent_seen = set(self._find_text_key_references_from_xml_inventory_lines(
2140
self._inventory_xml_lines_for_keys(parent_keys)))
2141
new_keys = seen - parent_seen
2143
1468
setdefault = result.setdefault
2144
for key in new_keys:
2145
setdefault(key[0], set()).add(key[-1])
1470
self._find_text_key_references_from_xml_inventory_lines(
1471
line_iterator).iterkeys():
1472
# once data is all ensured-consistent; then this is
1473
# if revision_id == version_id
1474
if key[-1:] in revision_ids:
1475
setdefault(key[0], set()).add(key[-1])
2148
def _find_parent_ids_of_revisions(self, revision_ids):
2149
"""Find all parent ids that are mentioned in the revision graph.
2151
:return: set of revisions that are parents of revision_ids which are
2152
not part of revision_ids themselves
2154
parent_map = self.get_parent_map(revision_ids)
2156
map(parent_ids.update, parent_map.itervalues())
2157
parent_ids.difference_update(revision_ids)
2158
parent_ids.discard(_mod_revision.NULL_REVISION)
2161
def _find_parent_keys_of_revisions(self, revision_keys):
2162
"""Similar to _find_parent_ids_of_revisions, but used with keys.
2164
:param revision_keys: An iterable of revision_keys.
2165
:return: The parents of all revision_keys that are not already in
2168
parent_map = self.revisions.get_parent_map(revision_keys)
2170
map(parent_keys.update, parent_map.itervalues())
2171
parent_keys.difference_update(revision_keys)
2172
parent_keys.discard(_mod_revision.NULL_REVISION)
2175
1478
def fileids_altered_by_revision_ids(self, revision_ids, _inv_weave=None):
2176
1479
"""Find the file ids and versions affected by revisions.
2381
1681
"""Get Inventory object by revision id."""
2382
1682
return self.iter_inventories([revision_id]).next()
2384
def iter_inventories(self, revision_ids, ordering=None):
1684
def iter_inventories(self, revision_ids):
2385
1685
"""Get many inventories by revision_ids.
2387
1687
This will buffer some or all of the texts used in constructing the
2388
1688
inventories in memory, but will only parse a single inventory at a
2391
:param revision_ids: The expected revision ids of the inventories.
2392
:param ordering: optional ordering, e.g. 'topological'. If not
2393
specified, the order of revision_ids will be preserved (by
2394
buffering if necessary).
2395
1691
:return: An iterator of inventories.
2397
1693
if ((None in revision_ids)
2398
1694
or (_mod_revision.NULL_REVISION in revision_ids)):
2399
1695
raise ValueError('cannot get null revision inventory')
2400
return self._iter_inventories(revision_ids, ordering)
1696
return self._iter_inventories(revision_ids)
2402
def _iter_inventories(self, revision_ids, ordering):
1698
def _iter_inventories(self, revision_ids):
2403
1699
"""single-document based inventory iteration."""
2404
inv_xmls = self._iter_inventory_xmls(revision_ids, ordering)
2405
for text, revision_id in inv_xmls:
2406
yield self._deserialise_inventory(revision_id, text)
1700
for text, revision_id in self._iter_inventory_xmls(revision_ids):
1701
yield self.deserialise_inventory(revision_id, text)
2408
def _iter_inventory_xmls(self, revision_ids, ordering):
2409
if ordering is None:
2410
order_as_requested = True
2411
ordering = 'unordered'
2413
order_as_requested = False
1703
def _iter_inventory_xmls(self, revision_ids):
2414
1704
keys = [(revision_id,) for revision_id in revision_ids]
2417
if order_as_requested:
2418
key_iter = iter(keys)
2419
next_key = key_iter.next()
2420
stream = self.inventories.get_record_stream(keys, ordering, True)
1705
stream = self.inventories.get_record_stream(keys, 'unordered', True)
2421
1706
text_chunks = {}
2422
1707
for record in stream:
2423
1708
if record.storage_kind != 'absent':
2424
chunks = record.get_bytes_as('chunked')
2425
if order_as_requested:
2426
text_chunks[record.key] = chunks
2428
yield ''.join(chunks), record.key[-1]
1709
text_chunks[record.key] = record.get_bytes_as('chunked')
2430
1711
raise errors.NoSuchRevision(self, record.key)
2431
if order_as_requested:
2432
# Yield as many results as we can while preserving order.
2433
while next_key in text_chunks:
2434
chunks = text_chunks.pop(next_key)
2435
yield ''.join(chunks), next_key[-1]
2437
next_key = key_iter.next()
2438
except StopIteration:
2439
# We still want to fully consume the get_record_stream,
2440
# just in case it is not actually finished at this point
1713
chunks = text_chunks.pop(key)
1714
yield ''.join(chunks), key[-1]
2444
def _deserialise_inventory(self, revision_id, xml):
1716
def deserialise_inventory(self, revision_id, xml):
2445
1717
"""Transform the xml into an inventory object.
2447
1719
:param revision_id: The expected revision id of the inventory.
2448
1720
:param xml: A serialised inventory.
2450
1722
result = self._serializer.read_inventory_from_string(xml, revision_id,
2451
entry_cache=self._inventory_entry_cache,
2452
return_from_cache=self._safe_to_return_from_cache)
1723
entry_cache=self._inventory_entry_cache)
2453
1724
if result.revision_id != revision_id:
2454
1725
raise AssertionError('revision id mismatch %s != %s' % (
2455
1726
result.revision_id, revision_id))
1729
def serialise_inventory(self, inv):
1730
return self._serializer.write_inventory_to_string(inv)
1732
def _serialise_inventory_to_lines(self, inv):
1733
return self._serializer.write_inventory_to_lines(inv)
2458
1735
def get_serializer_format(self):
2459
1736
return self._serializer.format_num
2461
1738
@needs_read_lock
2462
def _get_inventory_xml(self, revision_id):
2463
"""Get serialized inventory as a string."""
2464
texts = self._iter_inventory_xmls([revision_id], 'unordered')
1739
def get_inventory_xml(self, revision_id):
1740
"""Get inventory XML as a file object."""
1741
texts = self._iter_inventory_xmls([revision_id])
2466
1743
text, revision_id = texts.next()
2467
1744
except StopIteration:
2468
1745
raise errors.HistoryMissing(self, 'inventory', revision_id)
2471
def get_rev_id_for_revno(self, revno, known_pair):
2472
"""Return the revision id of a revno, given a later (revno, revid)
2473
pair in the same history.
2475
:return: if found (True, revid). If the available history ran out
2476
before reaching the revno, then this returns
2477
(False, (closest_revno, closest_revid)).
1749
def get_inventory_sha1(self, revision_id):
1750
"""Return the sha1 hash of the inventory entry
2479
known_revno, known_revid = known_pair
2480
partial_history = [known_revid]
2481
distance_from_known = known_revno - revno
2482
if distance_from_known < 0:
2484
'requested revno (%d) is later than given known revno (%d)'
2485
% (revno, known_revno))
2488
self, partial_history, stop_index=distance_from_known)
2489
except errors.RevisionNotPresent, err:
2490
if err.revision_id == known_revid:
2491
# The start revision (known_revid) wasn't found.
2493
# This is a stacked repository with no fallbacks, or a there's a
2494
# left-hand ghost. Either way, even though the revision named in
2495
# the error isn't in this repo, we know it's the next step in this
2496
# left-hand history.
2497
partial_history.append(err.revision_id)
2498
if len(partial_history) <= distance_from_known:
2499
# Didn't find enough history to get a revid for the revno.
2500
earliest_revno = known_revno - len(partial_history) + 1
2501
return (False, (earliest_revno, partial_history[-1]))
2502
if len(partial_history) - 1 > distance_from_known:
2503
raise AssertionError('_iter_for_revno returned too much history')
2504
return (True, partial_history[-1])
1752
return self.get_revision(revision_id).inventory_sha1
2506
1754
def iter_reverse_revision_history(self, revision_id):
2507
1755
"""Iterate backwards through revision ids in the lefthand history
3402
2574
InterRepository.get(other).method_name(parameters).
3405
_walk_to_common_revisions_batch_size = 50
2577
_walk_to_common_revisions_batch_size = 1
3406
2578
_optimisers = []
3407
2579
"""The available optimised InterRepository types."""
2581
def __init__(self, source, target):
2582
InterObject.__init__(self, source, target)
2583
# These two attributes may be overridden by e.g. InterOtherToRemote to
2584
# provide a faster implementation.
2585
self.target_get_graph = self.target.get_graph
2586
self.target_get_parent_map = self.target.get_parent_map
3410
2588
def copy_content(self, revision_id=None):
3411
"""Make a complete copy of the content in self into destination.
3413
This is a destructive operation! Do not use it on existing
3416
:param revision_id: Only copy the content needed to construct
3417
revision_id and its parents.
3420
self.target.set_make_working_trees(self.source.make_working_trees())
3421
except NotImplementedError:
3423
self.target.fetch(self.source, revision_id=revision_id)
3426
def fetch(self, revision_id=None, pb=None, find_ghosts=False,
2589
raise NotImplementedError(self.copy_content)
2591
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3428
2592
"""Fetch the content required to construct revision_id.
3430
2594
The content is copied from self.source to self.target.
3432
2596
:param revision_id: if None all content is copied, if NULL_REVISION no
3433
2597
content is copied.
2598
:param pb: optional progress bar to use for progress reports. If not
2599
provided a default one will be created.
2601
:returns: (copied_revision_count, failures).
3437
ui.ui_factory.warn_experimental_format_fetch(self)
3438
from bzrlib.fetch import RepoFetcher
3439
# See <https://launchpad.net/bugs/456077> asking for a warning here
3440
if self.source._format.network_name() != self.target._format.network_name():
3441
ui.ui_factory.show_user_warning('cross_format_fetch',
3442
from_format=self.source._format,
3443
to_format=self.target._format)
3444
f = RepoFetcher(to_repository=self.target,
3445
from_repository=self.source,
3446
last_revision=revision_id,
3447
fetch_spec=fetch_spec,
3448
find_ghosts=find_ghosts)
2603
# Normally we should find a specific InterRepository subclass to do
2604
# the fetch; if nothing else then at least InterSameDataRepository.
2605
# If none of them is suitable it looks like fetching is not possible;
2606
# we try to give a good message why. _assert_same_model will probably
2607
# give a helpful message; otherwise a generic one.
2608
self._assert_same_model(self.source, self.target)
2609
raise errors.IncompatibleRepositories(self.source, self.target,
2610
"no suitableInterRepository found")
3450
2612
def _walk_to_common_revisions(self, revision_ids):
3451
2613
"""Walk out from revision_ids in source to revisions target has.
3740
2984
return self.source.revision_ids_to_search_result(result_set)
3743
class InterDifferingSerializer(InterRepository):
2987
class InterPackRepo(InterSameDataRepository):
2988
"""Optimised code paths between Pack based repositories."""
2991
def _get_repo_format_to_test(self):
2992
from bzrlib.repofmt import pack_repo
2993
return pack_repo.RepositoryFormatKnitPack1()
2996
def is_compatible(source, target):
2997
"""Be compatible with known Pack formats.
2999
We don't test for the stores being of specific types because that
3000
could lead to confusing results, and there is no need to be
3003
from bzrlib.repofmt.pack_repo import RepositoryFormatPack
3005
are_packs = (isinstance(source._format, RepositoryFormatPack) and
3006
isinstance(target._format, RepositoryFormatPack))
3007
except AttributeError:
3009
return are_packs and InterRepository._same_model(source, target)
3012
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3013
"""See InterRepository.fetch()."""
3014
if (len(self.source._fallback_repositories) > 0 or
3015
len(self.target._fallback_repositories) > 0):
3016
# The pack layer is not aware of fallback repositories, so when
3017
# fetching from a stacked repository or into a stacked repository
3018
# we use the generic fetch logic which uses the VersionedFiles
3019
# attributes on repository.
3020
from bzrlib.fetch import RepoFetcher
3021
fetcher = RepoFetcher(self.target, self.source, revision_id,
3023
return fetcher.count_copied, fetcher.failed_revisions
3024
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
3025
self.source, self.source._format, self.target, self.target._format)
3026
self.count_copied = 0
3027
if revision_id is None:
3029
# everything to do - use pack logic
3030
# to fetch from all packs to one without
3031
# inventory parsing etc, IFF nothing to be copied is in the target.
3033
source_revision_ids = frozenset(self.source.all_revision_ids())
3034
revision_ids = source_revision_ids - \
3035
frozenset(self.target_get_parent_map(source_revision_ids))
3036
revision_keys = [(revid,) for revid in revision_ids]
3037
target_pack_collection = self._get_target_pack_collection()
3038
index = target_pack_collection.revision_index.combined_index
3039
present_revision_ids = set(item[1][0] for item in
3040
index.iter_entries(revision_keys))
3041
revision_ids = set(revision_ids) - present_revision_ids
3042
# implementing the TODO will involve:
3043
# - detecting when all of a pack is selected
3044
# - avoiding as much as possible pre-selection, so the
3045
# more-core routines such as create_pack_from_packs can filter in
3046
# a just-in-time fashion. (though having a HEADS list on a
3047
# repository might make this a lot easier, because we could
3048
# sensibly detect 'new revisions' without doing a full index scan.
3049
elif _mod_revision.is_null(revision_id):
3054
revision_ids = self.search_missing_revision_ids(revision_id,
3055
find_ghosts=find_ghosts).get_keys()
3056
except errors.NoSuchRevision:
3057
raise errors.InstallFailed([revision_id])
3058
if len(revision_ids) == 0:
3060
return self._pack(self.source, self.target, revision_ids)
3062
def _pack(self, source, target, revision_ids):
3063
from bzrlib.repofmt.pack_repo import Packer
3064
target_pack_collection = self._get_target_pack_collection()
3065
packs = source._pack_collection.all_packs()
3066
pack = Packer(target_pack_collection, packs, '.fetch',
3067
revision_ids).pack()
3068
if pack is not None:
3069
target_pack_collection._save_pack_names()
3070
copied_revs = pack.get_revision_count()
3071
# Trigger an autopack. This may duplicate effort as we've just done
3072
# a pack creation, but for now it is simpler to think about as
3073
# 'upload data, then repack if needed'.
3075
return (copied_revs, [])
3079
def _autopack(self):
3080
self.target._pack_collection.autopack()
3082
def _get_target_pack_collection(self):
3083
return self.target._pack_collection
3086
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
3087
"""See InterRepository.missing_revision_ids().
3089
:param find_ghosts: Find ghosts throughout the ancestry of
3092
if not find_ghosts and revision_id is not None:
3093
return self._walk_to_common_revisions([revision_id])
3094
elif revision_id is not None:
3095
# Find ghosts: search for revisions pointing from one repository to
3096
# the other, and vice versa, anywhere in the history of revision_id.
3097
graph = self.target_get_graph(other_repository=self.source)
3098
searcher = graph._make_breadth_first_searcher([revision_id])
3102
next_revs, ghosts = searcher.next_with_ghosts()
3103
except StopIteration:
3105
if revision_id in ghosts:
3106
raise errors.NoSuchRevision(self.source, revision_id)
3107
found_ids.update(next_revs)
3108
found_ids.update(ghosts)
3109
found_ids = frozenset(found_ids)
3110
# Double query here: should be able to avoid this by changing the
3111
# graph api further.
3112
result_set = found_ids - frozenset(
3113
self.target_get_parent_map(found_ids))
3115
source_ids = self.source.all_revision_ids()
3116
# source_ids is the worst possible case we may need to pull.
3117
# now we want to filter source_ids against what we actually
3118
# have in target, but don't try to check for existence where we know
3119
# we do not have a revision as that would be pointless.
3120
target_ids = set(self.target.all_revision_ids())
3121
result_set = set(source_ids).difference(target_ids)
3122
return self.source.revision_ids_to_search_result(result_set)
3125
class InterModel1and2(InterRepository):
3128
def _get_repo_format_to_test(self):
3132
def is_compatible(source, target):
3133
if not source.supports_rich_root() and target.supports_rich_root():
3139
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3140
"""See InterRepository.fetch()."""
3141
from bzrlib.fetch import Model1toKnit2Fetcher
3142
f = Model1toKnit2Fetcher(to_repository=self.target,
3143
from_repository=self.source,
3144
last_revision=revision_id,
3145
pb=pb, find_ghosts=find_ghosts)
3146
return f.count_copied, f.failed_revisions
3149
def copy_content(self, revision_id=None):
3150
"""Make a complete copy of the content in self into destination.
3152
This is a destructive operation! Do not use it on existing
3155
:param revision_id: Only copy the content needed to construct
3156
revision_id and its parents.
3159
self.target.set_make_working_trees(self.source.make_working_trees())
3160
except NotImplementedError:
3162
# but don't bother fetching if we have the needed data now.
3163
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
3164
self.target.has_revision(revision_id)):
3166
self.target.fetch(self.source, revision_id=revision_id)
3169
class InterKnit1and2(InterKnitRepo):
3172
def _get_repo_format_to_test(self):
3176
def is_compatible(source, target):
3177
"""Be compatible with Knit1 source and Knit3 target"""
3179
from bzrlib.repofmt.knitrepo import (
3180
RepositoryFormatKnit1,
3181
RepositoryFormatKnit3,
3183
from bzrlib.repofmt.pack_repo import (
3184
RepositoryFormatKnitPack1,
3185
RepositoryFormatKnitPack3,
3186
RepositoryFormatKnitPack4,
3187
RepositoryFormatKnitPack5,
3188
RepositoryFormatKnitPack5RichRoot,
3189
RepositoryFormatKnitPack6,
3190
RepositoryFormatKnitPack6RichRoot,
3191
RepositoryFormatPackDevelopment2,
3192
RepositoryFormatPackDevelopment2Subtree,
3195
RepositoryFormatKnit1, # no rr, no subtree
3196
RepositoryFormatKnitPack1, # no rr, no subtree
3197
RepositoryFormatPackDevelopment2, # no rr, no subtree
3198
RepositoryFormatKnitPack5, # no rr, no subtree
3199
RepositoryFormatKnitPack6, # no rr, no subtree
3202
RepositoryFormatKnit3, # rr, subtree
3203
RepositoryFormatKnitPack3, # rr, subtree
3204
RepositoryFormatKnitPack4, # rr, no subtree
3205
RepositoryFormatKnitPack5RichRoot,# rr, no subtree
3206
RepositoryFormatKnitPack6RichRoot,# rr, no subtree
3207
RepositoryFormatPackDevelopment2Subtree, # rr, subtree
3209
for format in norichroot:
3210
if format.rich_root_data:
3211
raise AssertionError('Format %s is a rich-root format'
3212
' but is included in the non-rich-root list'
3214
for format in richroot:
3215
if not format.rich_root_data:
3216
raise AssertionError('Format %s is not a rich-root format'
3217
' but is included in the rich-root list'
3219
# TODO: One alternative is to just check format.rich_root_data,
3220
# instead of keeping membership lists. However, the formats
3221
# *also* have to use the same 'Knit' style of storage
3222
# (line-deltas, fulltexts, etc.)
3223
return (isinstance(source._format, norichroot) and
3224
isinstance(target._format, richroot))
3225
except AttributeError:
3229
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3230
"""See InterRepository.fetch()."""
3231
from bzrlib.fetch import Knit1to2Fetcher
3232
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
3233
self.source, self.source._format, self.target,
3234
self.target._format)
3235
f = Knit1to2Fetcher(to_repository=self.target,
3236
from_repository=self.source,
3237
last_revision=revision_id,
3238
pb=pb, find_ghosts=find_ghosts)
3239
return f.count_copied, f.failed_revisions
3242
class InterDifferingSerializer(InterKnitRepo):
3746
3245
def _get_repo_format_to_test(self):
3750
3249
def is_compatible(source, target):
3751
3250
"""Be compatible with Knit2 source and Knit3 target"""
3752
# This is redundant with format.check_conversion_target(), however that
3753
# raises an exception, and we just want to say "False" as in we won't
3754
# support converting between these formats.
3755
if 'IDS_never' in debug.debug_flags:
3757
if source.supports_rich_root() and not target.supports_rich_root():
3759
if (source._format.supports_tree_reference
3760
and not target._format.supports_tree_reference):
3762
if target._fallback_repositories and target._format.supports_chks:
3763
# IDS doesn't know how to copy CHKs for the parent inventories it
3764
# adds to stacked repos.
3766
if 'IDS_always' in debug.debug_flags:
3768
# Only use this code path for local source and target. IDS does far
3769
# too much IO (both bandwidth and roundtrips) over a network.
3770
if not source.bzrdir.transport.base.startswith('file:///'):
3772
if not target.bzrdir.transport.base.startswith('file:///'):
3251
if source.supports_rich_root() != target.supports_rich_root():
3253
# Ideally, we'd support fetching if the source had no tree references
3254
# even if it supported them...
3255
if (getattr(source, '_format.supports_tree_reference', False) and
3256
not getattr(target, '_format.supports_tree_reference', False)):
3776
def _get_trees(self, revision_ids, cache):
3778
for rev_id in revision_ids:
3780
possible_trees.append((rev_id, cache[rev_id]))
3782
# Not cached, but inventory might be present anyway.
3784
tree = self.source.revision_tree(rev_id)
3785
except errors.NoSuchRevision:
3786
# Nope, parent is ghost.
3789
cache[rev_id] = tree
3790
possible_trees.append((rev_id, tree))
3791
return possible_trees
3793
def _get_delta_for_revision(self, tree, parent_ids, possible_trees):
3794
"""Get the best delta and base for this revision.
3796
:return: (basis_id, delta)
3799
# Generate deltas against each tree, to find the shortest.
3800
texts_possibly_new_in_tree = set()
3801
for basis_id, basis_tree in possible_trees:
3802
delta = tree.inventory._make_delta(basis_tree.inventory)
3803
for old_path, new_path, file_id, new_entry in delta:
3804
if new_path is None:
3805
# This file_id isn't present in the new rev, so we don't
3809
# Rich roots are handled elsewhere...
3811
kind = new_entry.kind
3812
if kind != 'directory' and kind != 'file':
3813
# No text record associated with this inventory entry.
3815
# This is a directory or file that has changed somehow.
3816
texts_possibly_new_in_tree.add((file_id, new_entry.revision))
3817
deltas.append((len(delta), basis_id, delta))
3819
return deltas[0][1:]
3821
def _fetch_parent_invs_for_stacking(self, parent_map, cache):
3822
"""Find all parent revisions that are absent, but for which the
3823
inventory is present, and copy those inventories.
3825
This is necessary to preserve correctness when the source is stacked
3826
without fallbacks configured. (Note that in cases like upgrade the
3827
source may be not have _fallback_repositories even though it is
3831
for parents in parent_map.values():
3832
parent_revs.update(parents)
3833
present_parents = self.source.get_parent_map(parent_revs)
3834
absent_parents = set(parent_revs).difference(present_parents)
3835
parent_invs_keys_for_stacking = self.source.inventories.get_parent_map(
3836
(rev_id,) for rev_id in absent_parents)
3837
parent_inv_ids = [key[-1] for key in parent_invs_keys_for_stacking]
3838
for parent_tree in self.source.revision_trees(parent_inv_ids):
3839
current_revision_id = parent_tree.get_revision_id()
3840
parents_parents_keys = parent_invs_keys_for_stacking[
3841
(current_revision_id,)]
3842
parents_parents = [key[-1] for key in parents_parents_keys]
3843
basis_id = _mod_revision.NULL_REVISION
3844
basis_tree = self.source.revision_tree(basis_id)
3845
delta = parent_tree.inventory._make_delta(basis_tree.inventory)
3846
self.target.add_inventory_by_delta(
3847
basis_id, delta, current_revision_id, parents_parents)
3848
cache[current_revision_id] = parent_tree
3850
def _fetch_batch(self, revision_ids, basis_id, cache, a_graph=None):
3260
def _fetch_batch(self, revision_ids, basis_id, basis_tree):
3851
3261
"""Fetch across a few revisions.
3853
3263
:param revision_ids: The revisions to copy
3854
:param basis_id: The revision_id of a tree that must be in cache, used
3855
as a basis for delta when no other base is available
3856
:param cache: A cache of RevisionTrees that we can use.
3857
:param a_graph: A Graph object to determine the heads() of the
3858
rich-root data stream.
3859
:return: The revision_id of the last converted tree. The RevisionTree
3860
for it will be in cache
3264
:param basis_id: The revision_id of basis_tree
3265
:param basis_tree: A tree that is not in revision_ids which should
3266
already exist in the target.
3267
:return: (basis_id, basis_tree) A new basis to use now that these trees
3862
3270
# Walk though all revisions; get inventory deltas, copy referenced
3863
3271
# texts that delta references, insert the delta, revision and
3865
root_keys_to_create = set()
3866
3273
text_keys = set()
3867
3274
pending_deltas = []
3868
3275
pending_revisions = []
3869
parent_map = self.source.get_parent_map(revision_ids)
3870
self._fetch_parent_invs_for_stacking(parent_map, cache)
3871
self.source._safe_to_return_from_cache = True
3872
3276
for tree in self.source.revision_trees(revision_ids):
3873
# Find a inventory delta for this revision.
3874
# Find text entries that need to be copied, too.
3875
3277
current_revision_id = tree.get_revision_id()
3876
parent_ids = parent_map.get(current_revision_id, ())
3877
parent_trees = self._get_trees(parent_ids, cache)
3878
possible_trees = list(parent_trees)
3879
if len(possible_trees) == 0:
3880
# There either aren't any parents, or the parents are ghosts,
3881
# so just use the last converted tree.
3882
possible_trees.append((basis_id, cache[basis_id]))
3883
basis_id, delta = self._get_delta_for_revision(tree, parent_ids,
3278
delta = tree.inventory._make_delta(basis_tree.inventory)
3279
for old_path, new_path, file_id, entry in delta:
3280
if new_path is not None:
3281
if not (new_path or self.target.supports_rich_root()):
3282
# We leave the inventory delta in, because that
3283
# will have the deserialised inventory root
3287
# "if entry.revision == current_revision_id" ?
3288
if entry.revision == current_revision_id:
3289
text_keys.add((file_id, entry.revision))
3885
3290
revision = self.source.get_revision(current_revision_id)
3886
3291
pending_deltas.append((basis_id, delta,
3887
3292
current_revision_id, revision.parent_ids))
3888
if self._converting_to_rich_root:
3889
self._revision_id_to_root_id[current_revision_id] = \
3891
# Determine which texts are in present in this revision but not in
3892
# any of the available parents.
3893
texts_possibly_new_in_tree = set()
3894
for old_path, new_path, file_id, entry in delta:
3895
if new_path is None:
3896
# This file_id isn't present in the new rev
3900
if not self.target.supports_rich_root():
3901
# The target doesn't support rich root, so we don't
3904
if self._converting_to_rich_root:
3905
# This can't be copied normally, we have to insert
3907
root_keys_to_create.add((file_id, entry.revision))
3910
texts_possibly_new_in_tree.add((file_id, entry.revision))
3911
for basis_id, basis_tree in possible_trees:
3912
basis_inv = basis_tree.inventory
3913
for file_key in list(texts_possibly_new_in_tree):
3914
file_id, file_revision = file_key
3916
entry = basis_inv[file_id]
3917
except errors.NoSuchId:
3919
if entry.revision == file_revision:
3920
texts_possibly_new_in_tree.remove(file_key)
3921
text_keys.update(texts_possibly_new_in_tree)
3922
3293
pending_revisions.append(revision)
3923
cache[current_revision_id] = tree
3924
3294
basis_id = current_revision_id
3925
self.source._safe_to_return_from_cache = False
3926
3296
# Copy file texts
3927
3297
from_texts = self.source.texts
3928
3298
to_texts = self.target.texts
3929
if root_keys_to_create:
3930
root_stream = _mod_fetch._new_root_data_stream(
3931
root_keys_to_create, self._revision_id_to_root_id, parent_map,
3932
self.source, graph=a_graph)
3933
to_texts.insert_record_stream(root_stream)
3934
3299
to_texts.insert_record_stream(from_texts.get_record_stream(
3935
text_keys, self.target._format._fetch_order,
3936
not self.target._format._fetch_uses_deltas))
3937
# insert inventory deltas
3300
text_keys, self.target._fetch_order,
3301
not self.target._fetch_uses_deltas))
3938
3303
for delta in pending_deltas:
3939
3304
self.target.add_inventory_by_delta(*delta)
3940
if self.target._fallback_repositories:
3941
# Make sure this stacked repository has all the parent inventories
3942
# for the new revisions that we are about to insert. We do this
3943
# before adding the revisions so that no revision is added until
3944
# all the inventories it may depend on are added.
3945
# Note that this is overzealous, as we may have fetched these in an
3948
revision_ids = set()
3949
for revision in pending_revisions:
3950
revision_ids.add(revision.revision_id)
3951
parent_ids.update(revision.parent_ids)
3952
parent_ids.difference_update(revision_ids)
3953
parent_ids.discard(_mod_revision.NULL_REVISION)
3954
parent_map = self.source.get_parent_map(parent_ids)
3955
# we iterate over parent_map and not parent_ids because we don't
3956
# want to try copying any revision which is a ghost
3957
for parent_tree in self.source.revision_trees(parent_map):
3958
current_revision_id = parent_tree.get_revision_id()
3959
parents_parents = parent_map[current_revision_id]
3960
possible_trees = self._get_trees(parents_parents, cache)
3961
if len(possible_trees) == 0:
3962
# There either aren't any parents, or the parents are
3963
# ghosts, so just use the last converted tree.
3964
possible_trees.append((basis_id, cache[basis_id]))
3965
basis_id, delta = self._get_delta_for_revision(parent_tree,
3966
parents_parents, possible_trees)
3967
self.target.add_inventory_by_delta(
3968
basis_id, delta, current_revision_id, parents_parents)
3969
3305
# insert signatures and revisions
3970
3306
for revision in pending_revisions:
3976
3312
except errors.NoSuchRevision:
3978
3314
self.target.add_revision(revision.revision_id, revision)
3315
return basis_id, basis_tree
3981
3317
def _fetch_all_revisions(self, revision_ids, pb):
3982
3318
"""Fetch everything for the list of revisions.
3984
3320
:param revision_ids: The list of revisions to fetch. Must be in
3985
3321
topological order.
3986
:param pb: A ProgressTask
3322
:param pb: A ProgressBar
3989
3325
basis_id, basis_tree = self._get_basis(revision_ids[0])
3990
3326
batch_size = 100
3991
cache = lru_cache.LRUCache(100)
3992
cache[basis_id] = basis_tree
3993
del basis_tree # We don't want to hang on to it here
3995
if self._converting_to_rich_root and len(revision_ids) > 100:
3996
a_graph = _mod_fetch._get_rich_root_heads_graph(self.source,
4001
3327
for offset in range(0, len(revision_ids), batch_size):
4002
3328
self.target.start_write_group()
4004
3330
pb.update('Transferring revisions', offset,
4005
3331
len(revision_ids))
4006
3332
batch = revision_ids[offset:offset+batch_size]
4007
basis_id = self._fetch_batch(batch, basis_id, cache,
3333
basis_id, basis_tree = self._fetch_batch(batch,
3334
basis_id, basis_tree)
4010
self.source._safe_to_return_from_cache = False
4011
3336
self.target.abort_write_group()
4014
hint = self.target.commit_write_group()
4017
if hints and self.target._format.pack_compresses:
4018
self.target.pack(hint=hints)
3339
self.target.commit_write_group()
4019
3340
pb.update('Transferring revisions', len(revision_ids),
4020
3341
len(revision_ids))
4022
3343
@needs_write_lock
4023
def fetch(self, revision_id=None, pb=None, find_ghosts=False,
3344
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
4025
3345
"""See InterRepository.fetch()."""
4026
if fetch_spec is not None:
4027
raise AssertionError("Not implemented yet...")
4028
ui.ui_factory.warn_experimental_format_fetch(self)
4029
if (not self.source.supports_rich_root()
4030
and self.target.supports_rich_root()):
4031
self._converting_to_rich_root = True
4032
self._revision_id_to_root_id = {}
4034
self._converting_to_rich_root = False
4035
# See <https://launchpad.net/bugs/456077> asking for a warning here
4036
if self.source._format.network_name() != self.target._format.network_name():
4037
ui.ui_factory.show_user_warning('cross_format_fetch',
4038
from_format=self.source._format,
4039
to_format=self.target._format)
4040
3346
revision_ids = self.target.search_missing_revision_ids(self.source,
4041
3347
revision_id, find_ghosts=find_ghosts).get_keys()
4042
3348
if not revision_ids:
4044
3350
revision_ids = tsort.topo_sort(
4045
3351
self.source.get_graph().get_parent_map(revision_ids))
4046
if not revision_ids:
4048
# Walk though all revisions; get inventory deltas, copy referenced
4049
# texts that delta references, insert the delta, revision and
4052
3353
my_pb = ui.ui_factory.nested_progress_bar()
4055
symbol_versioning.warn(
4056
symbol_versioning.deprecated_in((1, 14, 0))
4057
% "pb parameter to fetch()")
4060
3358
self._fetch_all_revisions(revision_ids, pb)
4086
3384
return basis_id, basis_tree
3387
class InterOtherToRemote(InterRepository):
3388
"""An InterRepository that simply delegates to the 'real' InterRepository
3389
calculated for (source, target._real_repository).
3392
_walk_to_common_revisions_batch_size = 50
3394
def __init__(self, source, target):
3395
InterRepository.__init__(self, source, target)
3396
self._real_inter = None
3399
def is_compatible(source, target):
3400
if isinstance(target, remote.RemoteRepository):
3404
def _ensure_real_inter(self):
3405
if self._real_inter is None:
3406
self.target._ensure_real()
3407
real_target = self.target._real_repository
3408
self._real_inter = InterRepository.get(self.source, real_target)
3409
# Make _real_inter use the RemoteRepository for get_parent_map
3410
self._real_inter.target_get_graph = self.target.get_graph
3411
self._real_inter.target_get_parent_map = self.target.get_parent_map
3413
def copy_content(self, revision_id=None):
3414
self._ensure_real_inter()
3415
self._real_inter.copy_content(revision_id=revision_id)
3417
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3418
self._ensure_real_inter()
3419
return self._real_inter.fetch(revision_id=revision_id, pb=pb,
3420
find_ghosts=find_ghosts)
3423
def _get_repo_format_to_test(self):
3427
class InterRemoteToOther(InterRepository):
3429
def __init__(self, source, target):
3430
InterRepository.__init__(self, source, target)
3431
self._real_inter = None
3434
def is_compatible(source, target):
3435
if not isinstance(source, remote.RemoteRepository):
3437
# Is source's model compatible with target's model?
3438
source._ensure_real()
3439
real_source = source._real_repository
3440
if isinstance(real_source, remote.RemoteRepository):
3441
raise NotImplementedError(
3442
"We don't support remote repos backed by remote repos yet.")
3443
return InterRepository._same_model(real_source, target)
3445
def _ensure_real_inter(self):
3446
if self._real_inter is None:
3447
self.source._ensure_real()
3448
real_source = self.source._real_repository
3449
self._real_inter = InterRepository.get(real_source, self.target)
3451
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3452
self._ensure_real_inter()
3453
return self._real_inter.fetch(revision_id=revision_id, pb=pb,
3454
find_ghosts=find_ghosts)
3456
def copy_content(self, revision_id=None):
3457
self._ensure_real_inter()
3458
self._real_inter.copy_content(revision_id=revision_id)
3461
def _get_repo_format_to_test(self):
3466
class InterPackToRemotePack(InterPackRepo):
3467
"""A specialisation of InterPackRepo for a target that is a
3470
This will use the get_parent_map RPC rather than plain readvs, and also
3471
uses an RPC for autopacking.
3474
_walk_to_common_revisions_batch_size = 50
3477
def is_compatible(source, target):
3478
from bzrlib.repofmt.pack_repo import RepositoryFormatPack
3479
if isinstance(source._format, RepositoryFormatPack):
3480
if isinstance(target, remote.RemoteRepository):
3481
target._ensure_real()
3482
if isinstance(target._real_repository._format,
3483
RepositoryFormatPack):
3484
if InterRepository._same_model(source, target):
3488
def _autopack(self):
3489
self.target.autopack()
3492
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3493
"""See InterRepository.fetch()."""
3494
# Always fetch using the generic streaming fetch code, to allow
3495
# streaming fetching into remote servers.
3496
from bzrlib.fetch import RepoFetcher
3497
fetcher = RepoFetcher(self.target, self.source, revision_id,
3499
return fetcher.count_copied, fetcher.failed_revisions
3501
def _get_target_pack_collection(self):
3502
return self.target._real_repository._pack_collection
3505
def _get_repo_format_to_test(self):
4089
3509
InterRepository.register_optimiser(InterDifferingSerializer)
4090
3510
InterRepository.register_optimiser(InterSameDataRepository)
4091
3511
InterRepository.register_optimiser(InterWeaveRepo)
4092
3512
InterRepository.register_optimiser(InterKnitRepo)
3513
InterRepository.register_optimiser(InterModel1and2)
3514
InterRepository.register_optimiser(InterKnit1and2)
3515
InterRepository.register_optimiser(InterPackRepo)
3516
InterRepository.register_optimiser(InterOtherToRemote)
3517
InterRepository.register_optimiser(InterRemoteToOther)
3518
InterRepository.register_optimiser(InterPackToRemotePack)
4095
3521
class CopyConverter(object):
4345
3721
self.target_repo.signatures.insert_record_stream(substream)
4347
3723
raise AssertionError('kaboom! %s' % (substream_type,))
4348
# Done inserting data, and the missing_keys calculations will try to
4349
# read back from the inserted data, so flush the writes to the new pack
4350
# (if this is pack format).
4351
if new_pack is not None:
4352
new_pack._write_data('', flush=True)
4353
# Find all the new revisions (including ones from resume_tokens)
4354
missing_keys = self.target_repo.get_missing_parent_inventories(
4355
check_for_missing_texts=is_resume)
4357
for prefix, versioned_file in (
4358
('texts', self.target_repo.texts),
4359
('inventories', self.target_repo.inventories),
4360
('revisions', self.target_repo.revisions),
4361
('signatures', self.target_repo.signatures),
4362
('chk_bytes', self.target_repo.chk_bytes),
4364
if versioned_file is None:
4366
# TODO: key is often going to be a StaticTuple object
4367
# I don't believe we can define a method by which
4368
# (prefix,) + StaticTuple will work, though we could
4369
# define a StaticTuple.sq_concat that would allow you to
4370
# pass in either a tuple or a StaticTuple as the second
4371
# object, so instead we could have:
4372
# StaticTuple(prefix) + key here...
4373
missing_keys.update((prefix,) + key for key in
4374
versioned_file.get_missing_compression_parent_keys())
4375
except NotImplementedError:
4376
# cannot even attempt suspending, and missing would have failed
4377
# during stream insertion.
4378
missing_keys = set()
4381
# suspend the write group and tell the caller what we is
4382
# missing. We know we can suspend or else we would not have
4383
# entered this code path. (All repositories that can handle
4384
# missing keys can handle suspending a write group).
4385
write_group_tokens = self.target_repo.suspend_write_group()
4386
return write_group_tokens, missing_keys
4387
hint = self.target_repo.commit_write_group()
4388
if (to_serializer != src_serializer and
4389
self.target_repo._format.pack_compresses):
4390
self.target_repo.pack(hint=hint)
4393
def _extract_and_insert_inventory_deltas(self, substream, serializer):
4394
target_rich_root = self.target_repo._format.rich_root_data
4395
target_tree_refs = self.target_repo._format.supports_tree_reference
4396
for record in substream:
4397
# Insert the delta directly
4398
inventory_delta_bytes = record.get_bytes_as('fulltext')
4399
deserialiser = inventory_delta.InventoryDeltaDeserializer()
4401
parse_result = deserialiser.parse_text_bytes(
4402
inventory_delta_bytes)
4403
except inventory_delta.IncompatibleInventoryDelta, err:
4404
trace.mutter("Incompatible delta: %s", err.msg)
4405
raise errors.IncompatibleRevision(self.target_repo._format)
4406
basis_id, new_id, rich_root, tree_refs, inv_delta = parse_result
4407
revision_id = new_id
4408
parents = [key[0] for key in record.parents]
4409
self.target_repo.add_inventory_by_delta(
4410
basis_id, inv_delta, revision_id, parents)
4412
def _extract_and_insert_inventories(self, substream, serializer,
3726
def _extract_and_insert_inventories(self, substream, serializer):
4414
3727
"""Generate a new inventory versionedfile in target, converting data.
4416
3729
The inventory is retrieved from the source, (deserializing it), and
4417
3730
stored in the target (reserializing it in a different format).
4419
target_rich_root = self.target_repo._format.rich_root_data
4420
target_tree_refs = self.target_repo._format.supports_tree_reference
4421
3732
for record in substream:
4422
# It's not a delta, so it must be a fulltext in the source
4423
# serializer's format.
4424
3733
bytes = record.get_bytes_as('fulltext')
4425
3734
revision_id = record.key[0]
4426
3735
inv = serializer.read_inventory_from_string(bytes, revision_id)
4427
3736
parents = [key[0] for key in record.parents]
4428
3737
self.target_repo.add_inventory(revision_id, inv, parents)
4429
# No need to keep holding this full inv in memory when the rest of
4430
# the substream is likely to be all deltas.
4433
3739
def _extract_and_insert_revisions(self, substream, serializer):
4434
3740
for record in substream:
4440
3746
self.target_repo.add_revision(revision_id, rev)
4442
3748
def finished(self):
4443
if self.target_repo._format._fetch_reconcile:
3749
if self.target_repo._fetch_reconcile:
4444
3750
self.target_repo.reconcile()
4447
class StreamSource(object):
4448
"""A source of a stream for fetching between repositories."""
4450
def __init__(self, from_repository, to_format):
4451
"""Create a StreamSource streaming from from_repository."""
4452
self.from_repository = from_repository
4453
self.to_format = to_format
4455
def delta_on_metadata(self):
4456
"""Return True if delta's are permitted on metadata streams.
4458
That is on revisions and signatures.
4460
src_serializer = self.from_repository._format._serializer
4461
target_serializer = self.to_format._serializer
4462
return (self.to_format._fetch_uses_deltas and
4463
src_serializer == target_serializer)
4465
def _fetch_revision_texts(self, revs):
4466
# fetch signatures first and then the revision texts
4467
# may need to be a InterRevisionStore call here.
4468
from_sf = self.from_repository.signatures
4469
# A missing signature is just skipped.
4470
keys = [(rev_id,) for rev_id in revs]
4471
signatures = versionedfile.filter_absent(from_sf.get_record_stream(
4473
self.to_format._fetch_order,
4474
not self.to_format._fetch_uses_deltas))
4475
# If a revision has a delta, this is actually expanded inside the
4476
# insert_record_stream code now, which is an alternate fix for
4478
from_rf = self.from_repository.revisions
4479
revisions = from_rf.get_record_stream(
4481
self.to_format._fetch_order,
4482
not self.delta_on_metadata())
4483
return [('signatures', signatures), ('revisions', revisions)]
4485
def _generate_root_texts(self, revs):
4486
"""This will be called by get_stream between fetching weave texts and
4487
fetching the inventory weave.
4489
if self._rich_root_upgrade():
4490
return _mod_fetch.Inter1and2Helper(
4491
self.from_repository).generate_root_texts(revs)
4495
def get_stream(self, search):
4497
revs = search.get_keys()
4498
graph = self.from_repository.get_graph()
4499
revs = tsort.topo_sort(graph.get_parent_map(revs))
4500
data_to_fetch = self.from_repository.item_keys_introduced_by(revs)
4502
for knit_kind, file_id, revisions in data_to_fetch:
4503
if knit_kind != phase:
4505
# Make a new progress bar for this phase
4506
if knit_kind == "file":
4507
# Accumulate file texts
4508
text_keys.extend([(file_id, revision) for revision in
4510
elif knit_kind == "inventory":
4511
# Now copy the file texts.
4512
from_texts = self.from_repository.texts
4513
yield ('texts', from_texts.get_record_stream(
4514
text_keys, self.to_format._fetch_order,
4515
not self.to_format._fetch_uses_deltas))
4516
# Cause an error if a text occurs after we have done the
4519
# Before we process the inventory we generate the root
4520
# texts (if necessary) so that the inventories references
4522
for _ in self._generate_root_texts(revs):
4524
# we fetch only the referenced inventories because we do not
4525
# know for unselected inventories whether all their required
4526
# texts are present in the other repository - it could be
4528
for info in self._get_inventory_stream(revs):
4530
elif knit_kind == "signatures":
4531
# Nothing to do here; this will be taken care of when
4532
# _fetch_revision_texts happens.
4534
elif knit_kind == "revisions":
4535
for record in self._fetch_revision_texts(revs):
4538
raise AssertionError("Unknown knit kind %r" % knit_kind)
4540
def get_stream_for_missing_keys(self, missing_keys):
4541
# missing keys can only occur when we are byte copying and not
4542
# translating (because translation means we don't send
4543
# unreconstructable deltas ever).
4545
keys['texts'] = set()
4546
keys['revisions'] = set()
4547
keys['inventories'] = set()
4548
keys['chk_bytes'] = set()
4549
keys['signatures'] = set()
4550
for key in missing_keys:
4551
keys[key[0]].add(key[1:])
4552
if len(keys['revisions']):
4553
# If we allowed copying revisions at this point, we could end up
4554
# copying a revision without copying its required texts: a
4555
# violation of the requirements for repository integrity.
4556
raise AssertionError(
4557
'cannot copy revisions to fill in missing deltas %s' % (
4558
keys['revisions'],))
4559
for substream_kind, keys in keys.iteritems():
4560
vf = getattr(self.from_repository, substream_kind)
4561
if vf is None and keys:
4562
raise AssertionError(
4563
"cannot fill in keys for a versioned file we don't"
4564
" have: %s needs %s" % (substream_kind, keys))
4566
# No need to stream something we don't have
4568
if substream_kind == 'inventories':
4569
# Some missing keys are genuinely ghosts, filter those out.
4570
present = self.from_repository.inventories.get_parent_map(keys)
4571
revs = [key[0] for key in present]
4572
# Get the inventory stream more-or-less as we do for the
4573
# original stream; there's no reason to assume that records
4574
# direct from the source will be suitable for the sink. (Think
4575
# e.g. 2a -> 1.9-rich-root).
4576
for info in self._get_inventory_stream(revs, missing=True):
4580
# Ask for full texts always so that we don't need more round trips
4581
# after this stream.
4582
# Some of the missing keys are genuinely ghosts, so filter absent
4583
# records. The Sink is responsible for doing another check to
4584
# ensure that ghosts don't introduce missing data for future
4586
stream = versionedfile.filter_absent(vf.get_record_stream(keys,
4587
self.to_format._fetch_order, True))
4588
yield substream_kind, stream
4590
def inventory_fetch_order(self):
4591
if self._rich_root_upgrade():
4592
return 'topological'
4594
return self.to_format._fetch_order
4596
def _rich_root_upgrade(self):
4597
return (not self.from_repository._format.rich_root_data and
4598
self.to_format.rich_root_data)
4600
def _get_inventory_stream(self, revision_ids, missing=False):
4601
from_format = self.from_repository._format
4602
if (from_format.supports_chks and self.to_format.supports_chks and
4603
from_format.network_name() == self.to_format.network_name()):
4604
raise AssertionError(
4605
"this case should be handled by GroupCHKStreamSource")
4606
elif 'forceinvdeltas' in debug.debug_flags:
4607
return self._get_convertable_inventory_stream(revision_ids,
4608
delta_versus_null=missing)
4609
elif from_format.network_name() == self.to_format.network_name():
4611
return self._get_simple_inventory_stream(revision_ids,
4613
elif (not from_format.supports_chks and not self.to_format.supports_chks
4614
and from_format._serializer == self.to_format._serializer):
4615
# Essentially the same format.
4616
return self._get_simple_inventory_stream(revision_ids,
4619
# Any time we switch serializations, we want to use an
4620
# inventory-delta based approach.
4621
return self._get_convertable_inventory_stream(revision_ids,
4622
delta_versus_null=missing)
4624
def _get_simple_inventory_stream(self, revision_ids, missing=False):
4625
# NB: This currently reopens the inventory weave in source;
4626
# using a single stream interface instead would avoid this.
4627
from_weave = self.from_repository.inventories
4629
delta_closure = True
4631
delta_closure = not self.delta_on_metadata()
4632
yield ('inventories', from_weave.get_record_stream(
4633
[(rev_id,) for rev_id in revision_ids],
4634
self.inventory_fetch_order(), delta_closure))
4636
def _get_convertable_inventory_stream(self, revision_ids,
4637
delta_versus_null=False):
4638
# The two formats are sufficiently different that there is no fast
4639
# path, so we need to send just inventorydeltas, which any
4640
# sufficiently modern client can insert into any repository.
4641
# The StreamSink code expects to be able to
4642
# convert on the target, so we need to put bytes-on-the-wire that can
4643
# be converted. That means inventory deltas (if the remote is <1.19,
4644
# RemoteStreamSink will fallback to VFS to insert the deltas).
4645
yield ('inventory-deltas',
4646
self._stream_invs_as_deltas(revision_ids,
4647
delta_versus_null=delta_versus_null))
4649
def _stream_invs_as_deltas(self, revision_ids, delta_versus_null=False):
4650
"""Return a stream of inventory-deltas for the given rev ids.
4652
:param revision_ids: The list of inventories to transmit
4653
:param delta_versus_null: Don't try to find a minimal delta for this
4654
entry, instead compute the delta versus the NULL_REVISION. This
4655
effectively streams a complete inventory. Used for stuff like
4656
filling in missing parents, etc.
4658
from_repo = self.from_repository
4659
revision_keys = [(rev_id,) for rev_id in revision_ids]
4660
parent_map = from_repo.inventories.get_parent_map(revision_keys)
4661
# XXX: possibly repos could implement a more efficient iter_inv_deltas
4663
inventories = self.from_repository.iter_inventories(
4664
revision_ids, 'topological')
4665
format = from_repo._format
4666
invs_sent_so_far = set([_mod_revision.NULL_REVISION])
4667
inventory_cache = lru_cache.LRUCache(50)
4668
null_inventory = from_repo.revision_tree(
4669
_mod_revision.NULL_REVISION).inventory
4670
# XXX: ideally the rich-root/tree-refs flags would be per-revision, not
4671
# per-repo (e.g. streaming a non-rich-root revision out of a rich-root
4672
# repo back into a non-rich-root repo ought to be allowed)
4673
serializer = inventory_delta.InventoryDeltaSerializer(
4674
versioned_root=format.rich_root_data,
4675
tree_references=format.supports_tree_reference)
4676
for inv in inventories:
4677
key = (inv.revision_id,)
4678
parent_keys = parent_map.get(key, ())
4680
if not delta_versus_null and parent_keys:
4681
# The caller did not ask for complete inventories and we have
4682
# some parents that we can delta against. Make a delta against
4683
# each parent so that we can find the smallest.
4684
parent_ids = [parent_key[0] for parent_key in parent_keys]
4685
for parent_id in parent_ids:
4686
if parent_id not in invs_sent_so_far:
4687
# We don't know that the remote side has this basis, so
4690
if parent_id == _mod_revision.NULL_REVISION:
4691
parent_inv = null_inventory
4693
parent_inv = inventory_cache.get(parent_id, None)
4694
if parent_inv is None:
4695
parent_inv = from_repo.get_inventory(parent_id)
4696
candidate_delta = inv._make_delta(parent_inv)
4697
if (delta is None or
4698
len(delta) > len(candidate_delta)):
4699
delta = candidate_delta
4700
basis_id = parent_id
4702
# Either none of the parents ended up being suitable, or we
4703
# were asked to delta against NULL
4704
basis_id = _mod_revision.NULL_REVISION
4705
delta = inv._make_delta(null_inventory)
4706
invs_sent_so_far.add(inv.revision_id)
4707
inventory_cache[inv.revision_id] = inv
4708
delta_serialized = ''.join(
4709
serializer.delta_to_lines(basis_id, key[-1], delta))
4710
yield versionedfile.FulltextContentFactory(
4711
key, parent_keys, None, delta_serialized)
4714
def _iter_for_revno(repo, partial_history_cache, stop_index=None,
4715
stop_revision=None):
4716
"""Extend the partial history to include a given index
4718
If a stop_index is supplied, stop when that index has been reached.
4719
If a stop_revision is supplied, stop when that revision is
4720
encountered. Otherwise, stop when the beginning of history is
4723
:param stop_index: The index which should be present. When it is
4724
present, history extension will stop.
4725
:param stop_revision: The revision id which should be present. When
4726
it is encountered, history extension will stop.
4728
start_revision = partial_history_cache[-1]
4729
iterator = repo.iter_reverse_revision_history(start_revision)
4731
#skip the last revision in the list
4734
if (stop_index is not None and
4735
len(partial_history_cache) > stop_index):
4737
if partial_history_cache[-1] == stop_revision:
4739
revision_id = iterator.next()
4740
partial_history_cache.append(revision_id)
4741
except StopIteration:
added
removed
bzrlib/repository.py
