547
492
ie.revision = parent_entry.revision
548
493
return self._get_delta(ie, basis_inv, path), False, None
549
494
ie.reference_revision = content_summary[3]
550
if ie.reference_revision is None:
551
raise AssertionError("invalid content_summary for nested tree: %r"
552
% (content_summary,))
553
self._add_text_to_weave(ie.file_id, '', heads, None)
496
self._add_text_to_weave(ie.file_id, lines, heads, None)
555
498
raise NotImplementedError('unknown kind')
556
499
ie.revision = self._new_revision_id
557
self._any_changes = True
558
500
return self._get_delta(ie, basis_inv, path), True, fingerprint
560
def record_iter_changes(self, tree, basis_revision_id, iter_changes,
561
_entry_factory=entry_factory):
562
"""Record a new tree via iter_changes.
564
:param tree: The tree to obtain text contents from for changed objects.
565
:param basis_revision_id: The revision id of the tree the iter_changes
566
has been generated against. Currently assumed to be the same
567
as self.parents[0] - if it is not, errors may occur.
568
:param iter_changes: An iter_changes iterator with the changes to apply
569
to basis_revision_id. The iterator must not include any items with
570
a current kind of None - missing items must be either filtered out
571
or errored-on beefore record_iter_changes sees the item.
572
:param _entry_factory: Private method to bind entry_factory locally for
574
:return: A generator of (file_id, relpath, fs_hash) tuples for use with
577
# Create an inventory delta based on deltas between all the parents and
578
# deltas between all the parent inventories. We use inventory delta's
579
# between the inventory objects because iter_changes masks
580
# last-changed-field only changes.
582
# file_id -> change map, change is fileid, paths, changed, versioneds,
583
# parents, names, kinds, executables
585
# {file_id -> revision_id -> inventory entry, for entries in parent
586
# trees that are not parents[0]
590
revtrees = list(self.repository.revision_trees(self.parents))
591
except errors.NoSuchRevision:
592
# one or more ghosts, slow path.
594
for revision_id in self.parents:
596
revtrees.append(self.repository.revision_tree(revision_id))
597
except errors.NoSuchRevision:
599
basis_revision_id = _mod_revision.NULL_REVISION
601
revtrees.append(self.repository.revision_tree(
602
_mod_revision.NULL_REVISION))
603
# The basis inventory from a repository
605
basis_inv = revtrees[0].inventory
607
basis_inv = self.repository.revision_tree(
608
_mod_revision.NULL_REVISION).inventory
609
if len(self.parents) > 0:
610
if basis_revision_id != self.parents[0] and not ghost_basis:
612
"arbitrary basis parents not yet supported with merges")
613
for revtree in revtrees[1:]:
614
for change in revtree.inventory._make_delta(basis_inv):
615
if change[1] is None:
616
# Not present in this parent.
618
if change[2] not in merged_ids:
619
if change[0] is not None:
620
basis_entry = basis_inv[change[2]]
621
merged_ids[change[2]] = [
623
basis_entry.revision,
626
parent_entries[change[2]] = {
628
basis_entry.revision:basis_entry,
630
change[3].revision:change[3],
633
merged_ids[change[2]] = [change[3].revision]
634
parent_entries[change[2]] = {change[3].revision:change[3]}
636
merged_ids[change[2]].append(change[3].revision)
637
parent_entries[change[2]][change[3].revision] = change[3]
640
# Setup the changes from the tree:
641
# changes maps file_id -> (change, [parent revision_ids])
643
for change in iter_changes:
644
# This probably looks up in basis_inv way to much.
645
if change[1][0] is not None:
646
head_candidate = [basis_inv[change[0]].revision]
649
changes[change[0]] = change, merged_ids.get(change[0],
651
unchanged_merged = set(merged_ids) - set(changes)
652
# Extend the changes dict with synthetic changes to record merges of
654
for file_id in unchanged_merged:
655
# Record a merged version of these items that did not change vs the
656
# basis. This can be either identical parallel changes, or a revert
657
# of a specific file after a merge. The recorded content will be
658
# that of the current tree (which is the same as the basis), but
659
# the per-file graph will reflect a merge.
660
# NB:XXX: We are reconstructing path information we had, this
661
# should be preserved instead.
662
# inv delta change: (file_id, (path_in_source, path_in_target),
663
# changed_content, versioned, parent, name, kind,
666
basis_entry = basis_inv[file_id]
667
except errors.NoSuchId:
668
# a change from basis->some_parents but file_id isn't in basis
669
# so was new in the merge, which means it must have changed
670
# from basis -> current, and as it hasn't the add was reverted
671
# by the user. So we discard this change.
675
(basis_inv.id2path(file_id), tree.id2path(file_id)),
677
(basis_entry.parent_id, basis_entry.parent_id),
678
(basis_entry.name, basis_entry.name),
679
(basis_entry.kind, basis_entry.kind),
680
(basis_entry.executable, basis_entry.executable))
681
changes[file_id] = (change, merged_ids[file_id])
682
# changes contains tuples with the change and a set of inventory
683
# candidates for the file.
685
# old_path, new_path, file_id, new_inventory_entry
686
seen_root = False # Is the root in the basis delta?
687
inv_delta = self._basis_delta
688
modified_rev = self._new_revision_id
689
for change, head_candidates in changes.values():
690
if change[3][1]: # versioned in target.
691
# Several things may be happening here:
692
# We may have a fork in the per-file graph
693
# - record a change with the content from tree
694
# We may have a change against < all trees
695
# - carry over the tree that hasn't changed
696
# We may have a change against all trees
697
# - record the change with the content from tree
700
entry = _entry_factory[kind](file_id, change[5][1],
702
head_set = self._heads(change[0], set(head_candidates))
705
for head_candidate in head_candidates:
706
if head_candidate in head_set:
707
heads.append(head_candidate)
708
head_set.remove(head_candidate)
711
# Could be a carry-over situation:
712
parent_entry_revs = parent_entries.get(file_id, None)
713
if parent_entry_revs:
714
parent_entry = parent_entry_revs.get(heads[0], None)
717
if parent_entry is None:
718
# The parent iter_changes was called against is the one
719
# that is the per-file head, so any change is relevant
720
# iter_changes is valid.
721
carry_over_possible = False
723
# could be a carry over situation
724
# A change against the basis may just indicate a merge,
725
# we need to check the content against the source of the
726
# merge to determine if it was changed after the merge
728
if (parent_entry.kind != entry.kind or
729
parent_entry.parent_id != entry.parent_id or
730
parent_entry.name != entry.name):
731
# Metadata common to all entries has changed
732
# against per-file parent
733
carry_over_possible = False
735
carry_over_possible = True
736
# per-type checks for changes against the parent_entry
739
# Cannot be a carry-over situation
740
carry_over_possible = False
741
# Populate the entry in the delta
743
# XXX: There is still a small race here: If someone reverts the content of a file
744
# after iter_changes examines and decides it has changed,
745
# we will unconditionally record a new version even if some
746
# other process reverts it while commit is running (with
747
# the revert happening after iter_changes did it's
750
entry.executable = True
752
entry.executable = False
753
if (carry_over_possible and
754
parent_entry.executable == entry.executable):
755
# Check the file length, content hash after reading
757
nostore_sha = parent_entry.text_sha1
760
file_obj, stat_value = tree.get_file_with_stat(file_id, change[1][1])
762
text = file_obj.read()
766
entry.text_sha1, entry.text_size = self._add_text_to_weave(
767
file_id, text, heads, nostore_sha)
768
yield file_id, change[1][1], (entry.text_sha1, stat_value)
769
except errors.ExistingContent:
770
# No content change against a carry_over parent
771
# Perhaps this should also yield a fs hash update?
773
entry.text_size = parent_entry.text_size
774
entry.text_sha1 = parent_entry.text_sha1
775
elif kind == 'symlink':
777
entry.symlink_target = tree.get_symlink_target(file_id)
778
if (carry_over_possible and
779
parent_entry.symlink_target == entry.symlink_target):
782
self._add_text_to_weave(change[0], '', heads, None)
783
elif kind == 'directory':
784
if carry_over_possible:
787
# Nothing to set on the entry.
788
# XXX: split into the Root and nonRoot versions.
789
if change[1][1] != '' or self.repository.supports_rich_root():
790
self._add_text_to_weave(change[0], '', heads, None)
791
elif kind == 'tree-reference':
792
if not self.repository._format.supports_tree_reference:
793
# This isn't quite sane as an error, but we shouldn't
794
# ever see this code path in practice: tree's don't
795
# permit references when the repo doesn't support tree
797
raise errors.UnsupportedOperation(tree.add_reference,
799
reference_revision = tree.get_reference_revision(change[0])
800
entry.reference_revision = reference_revision
801
if (carry_over_possible and
802
parent_entry.reference_revision == reference_revision):
805
self._add_text_to_weave(change[0], '', heads, None)
807
raise AssertionError('unknown kind %r' % kind)
809
entry.revision = modified_rev
811
entry.revision = parent_entry.revision
814
new_path = change[1][1]
815
inv_delta.append((change[1][0], new_path, change[0], entry))
818
self.new_inventory = None
820
# This should perhaps be guarded by a check that the basis we
821
# commit against is the basis for the commit and if not do a delta
823
self._any_changes = True
825
# housekeeping root entry changes do not affect no-change commits.
826
self._require_root_change(tree)
827
self.basis_delta_revision = basis_revision_id
829
def _add_text_to_weave(self, file_id, new_text, parents, nostore_sha):
830
parent_keys = tuple([(file_id, parent) for parent in parents])
831
return self.repository.texts._add_text(
832
(file_id, self._new_revision_id), parent_keys, new_text,
833
nostore_sha=nostore_sha, random_id=self.random_revid)[0:2]
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]
836
515
class RootCommitBuilder(CommitBuilder):
1167
796
# The old API returned a list, should this actually be a set?
1168
797
return parent_map.keys()
1170
def _check_inventories(self, checker):
1171
"""Check the inventories found from the revision scan.
1173
This is responsible for verifying the sha1 of inventories and
1174
creating a pending_keys set that covers data referenced by inventories.
1176
bar = ui.ui_factory.nested_progress_bar()
1178
self._do_check_inventories(checker, bar)
1182
def _do_check_inventories(self, checker, bar):
1183
"""Helper for _check_inventories."""
1185
keys = {'chk_bytes':set(), 'inventories':set(), 'texts':set()}
1186
kinds = ['chk_bytes', 'texts']
1187
count = len(checker.pending_keys)
1188
bar.update("inventories", 0, 2)
1189
current_keys = checker.pending_keys
1190
checker.pending_keys = {}
1191
# Accumulate current checks.
1192
for key in current_keys:
1193
if key[0] != 'inventories' and key[0] not in kinds:
1194
checker._report_items.append('unknown key type %r' % (key,))
1195
keys[key[0]].add(key[1:])
1196
if keys['inventories']:
1197
# NB: output order *should* be roughly sorted - topo or
1198
# inverse topo depending on repository - either way decent
1199
# to just delta against. However, pre-CHK formats didn't
1200
# try to optimise inventory layout on disk. As such the
1201
# pre-CHK code path does not use inventory deltas.
1203
for record in self.inventories.check(keys=keys['inventories']):
1204
if record.storage_kind == 'absent':
1205
checker._report_items.append(
1206
'Missing inventory {%s}' % (record.key,))
1208
last_object = self._check_record('inventories', record,
1209
checker, last_object,
1210
current_keys[('inventories',) + record.key])
1211
del keys['inventories']
1214
bar.update("texts", 1)
1215
while (checker.pending_keys or keys['chk_bytes']
1217
# Something to check.
1218
current_keys = checker.pending_keys
1219
checker.pending_keys = {}
1220
# Accumulate current checks.
1221
for key in current_keys:
1222
if key[0] not in kinds:
1223
checker._report_items.append('unknown key type %r' % (key,))
1224
keys[key[0]].add(key[1:])
1225
# Check the outermost kind only - inventories || chk_bytes || texts
1229
for record in getattr(self, kind).check(keys=keys[kind]):
1230
if record.storage_kind == 'absent':
1231
checker._report_items.append(
1232
'Missing %s {%s}' % (kind, record.key,))
1234
last_object = self._check_record(kind, record,
1235
checker, last_object, current_keys[(kind,) + record.key])
1239
def _check_record(self, kind, record, checker, last_object, item_data):
1240
"""Check a single text from this repository."""
1241
if kind == 'inventories':
1242
rev_id = record.key[0]
1243
inv = self._deserialise_inventory(rev_id,
1244
record.get_bytes_as('fulltext'))
1245
if last_object is not None:
1246
delta = inv._make_delta(last_object)
1247
for old_path, path, file_id, ie in delta:
1250
ie.check(checker, rev_id, inv)
1252
for path, ie in inv.iter_entries():
1253
ie.check(checker, rev_id, inv)
1254
if self._format.fast_deltas:
1256
elif kind == 'chk_bytes':
1257
# No code written to check chk_bytes for this repo format.
1258
checker._report_items.append(
1259
'unsupported key type chk_bytes for %s' % (record.key,))
1260
elif kind == 'texts':
1261
self._check_text(record, checker, item_data)
1263
checker._report_items.append(
1264
'unknown key type %s for %s' % (kind, record.key))
1266
def _check_text(self, record, checker, item_data):
1267
"""Check a single text."""
1268
# Check it is extractable.
1269
# TODO: check length.
1270
if record.storage_kind == 'chunked':
1271
chunks = record.get_bytes_as(record.storage_kind)
1272
sha1 = osutils.sha_strings(chunks)
1273
length = sum(map(len, chunks))
1275
content = record.get_bytes_as('fulltext')
1276
sha1 = osutils.sha_string(content)
1277
length = len(content)
1278
if item_data and sha1 != item_data[1]:
1279
checker._report_items.append(
1280
'sha1 mismatch: %s has sha1 %s expected %s referenced by %s' %
1281
(record.key, sha1, item_data[1], item_data[2]))
1284
800
def create(a_bzrdir):
1285
801
"""Construct the current default format repository in a_bzrdir."""
1576
1084
def suspend_write_group(self):
1577
1085
raise errors.UnsuspendableWriteGroup(self)
1579
def get_missing_parent_inventories(self, check_for_missing_texts=True):
1580
"""Return the keys of missing inventory parents for revisions added in
1583
A revision is not complete if the inventory delta for that revision
1584
cannot be calculated. Therefore if the parent inventories of a
1585
revision are not present, the revision is incomplete, and e.g. cannot
1586
be streamed by a smart server. This method finds missing inventory
1587
parents for revisions added in this write group.
1589
if not self._format.supports_external_lookups:
1590
# This is only an issue for stacked repositories
1592
if not self.is_in_write_group():
1593
raise AssertionError('not in a write group')
1595
# XXX: We assume that every added revision already has its
1596
# corresponding inventory, so we only check for parent inventories that
1597
# might be missing, rather than all inventories.
1598
parents = set(self.revisions._index.get_missing_parents())
1599
parents.discard(_mod_revision.NULL_REVISION)
1600
unstacked_inventories = self.inventories._index
1601
present_inventories = unstacked_inventories.get_parent_map(
1602
key[-1:] for key in parents)
1603
parents.difference_update(present_inventories)
1604
if len(parents) == 0:
1605
# No missing parent inventories.
1607
if not check_for_missing_texts:
1608
return set(('inventories', rev_id) for (rev_id,) in parents)
1609
# Ok, now we have a list of missing inventories. But these only matter
1610
# if the inventories that reference them are missing some texts they
1611
# appear to introduce.
1612
# XXX: Texts referenced by all added inventories need to be present,
1613
# but at the moment we're only checking for texts referenced by
1614
# inventories at the graph's edge.
1615
key_deps = self.revisions._index._key_dependencies
1616
key_deps.satisfy_refs_for_keys(present_inventories)
1617
referrers = frozenset(r[0] for r in key_deps.get_referrers())
1618
file_ids = self.fileids_altered_by_revision_ids(referrers)
1619
missing_texts = set()
1620
for file_id, version_ids in file_ids.iteritems():
1621
missing_texts.update(
1622
(file_id, version_id) for version_id in version_ids)
1623
present_texts = self.texts.get_parent_map(missing_texts)
1624
missing_texts.difference_update(present_texts)
1625
if not missing_texts:
1626
# No texts are missing, so all revisions and their deltas are
1629
# Alternatively the text versions could be returned as the missing
1630
# keys, but this is likely to be less data.
1631
missing_keys = set(('inventories', rev_id) for (rev_id,) in parents)
1634
def refresh_data(self):
1635
"""Re-read any data needed to to synchronise with disk.
1637
This method is intended to be called after another repository instance
1638
(such as one used by a smart server) has inserted data into the
1639
repository. It may not be called during a write group, but may be
1640
called at any other time.
1642
if self.is_in_write_group():
1643
raise errors.InternalBzrError(
1644
"May not refresh_data while in a write group.")
1645
self._refresh_data()
1647
1087
def resume_write_group(self, tokens):
1648
1088
if not self.is_write_locked():
1649
1089
raise errors.NotWriteLocked(self)
1858
1269
@needs_read_lock
1859
1270
def get_revisions(self, revision_ids):
1860
"""Get many revisions at once.
1862
Repositories that need to check data on every revision read should
1863
subclass this method.
1271
"""Get many revisions at once."""
1865
1272
return self._get_revisions(revision_ids)
1867
1274
@needs_read_lock
1868
1275
def _get_revisions(self, revision_ids):
1869
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)
1871
for revid, rev in self._iter_revisions(revision_ids):
1873
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
1875
1289
return [revs[revid] for revid in revision_ids]
1877
def _iter_revisions(self, revision_ids):
1878
"""Iterate over revision objects.
1880
:param revision_ids: An iterable of revisions to examine. None may be
1881
passed to request all revisions known to the repository. Note that
1882
not all repositories can find unreferenced revisions; for those
1883
repositories only referenced ones will be returned.
1884
:return: An iterator of (revid, revision) tuples. Absent revisions (
1885
those asked for but not available) are returned as (revid, None).
1887
if revision_ids is None:
1888
revision_ids = self.all_revision_ids()
1890
for rev_id in revision_ids:
1891
if not rev_id or not isinstance(rev_id, basestring):
1892
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1893
keys = [(key,) for key in revision_ids]
1894
stream = self.revisions.get_record_stream(keys, 'unordered', True)
1895
for record in stream:
1896
revid = record.key[0]
1897
if record.storage_kind == 'absent':
1900
text = record.get_bytes_as('fulltext')
1901
rev = self._serializer.read_revision_from_string(text)
1904
1291
@needs_read_lock
1905
1292
def get_revision_xml(self, revision_id):
1906
1293
# TODO: jam 20070210 This shouldn't be necessary since get_revision
1907
1294
# would have already do it.
1908
1295
# TODO: jam 20070210 Just use _serializer.write_revision_to_string()
1909
# TODO: this can't just be replaced by:
1910
# return self._serializer.write_revision_to_string(
1911
# self.get_revision(revision_id))
1912
# as cStringIO preservers the encoding unlike write_revision_to_string
1913
# or some other call down the path.
1914
1296
rev = self.get_revision(revision_id)
1915
1297
rev_tmp = cStringIO.StringIO()
1916
1298
# the current serializer..
2079
1440
result[key] = True
2082
def _inventory_xml_lines_for_keys(self, keys):
2083
"""Get a line iterator of the sort needed for findind references.
2085
Not relevant for non-xml inventory repositories.
2087
Ghosts in revision_keys are ignored.
2089
:param revision_keys: The revision keys for the inventories to inspect.
2090
:return: An iterator over (inventory line, revid) for the fulltexts of
2091
all of the xml inventories specified by revision_keys.
2093
stream = self.inventories.get_record_stream(keys, 'unordered', True)
2094
for record in stream:
2095
if record.storage_kind != 'absent':
2096
chunks = record.get_bytes_as('chunked')
2097
revid = record.key[-1]
2098
lines = osutils.chunks_to_lines(chunks)
2102
1443
def _find_file_ids_from_xml_inventory_lines(self, line_iterator,
2104
1445
"""Helper routine for fileids_altered_by_revision_ids.
2106
1447
This performs the translation of xml lines to revision ids.
2108
1449
:param line_iterator: An iterator of lines, origin_version_id
2109
: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
2110
1451
set or other type which supports efficient __contains__ lookups, as
2111
the revision key from each parsed line will be looked up in the
2112
revision_keys filter.
1452
the revision id from each parsed line will be looked up in the
1453
revision_ids filter.
2113
1454
:return: a dictionary mapping altered file-ids to an iterable of
2114
1455
revision_ids. Each altered file-ids has the exact revision_ids that
2115
1456
altered it listed explicitly.
2117
seen = set(self._find_text_key_references_from_xml_inventory_lines(
2118
line_iterator).iterkeys())
2119
parent_keys = self._find_parent_keys_of_revisions(revision_keys)
2120
parent_seen = set(self._find_text_key_references_from_xml_inventory_lines(
2121
self._inventory_xml_lines_for_keys(parent_keys)))
2122
new_keys = seen - parent_seen
2124
1459
setdefault = result.setdefault
2125
for key in new_keys:
2126
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])
2129
def _find_parent_ids_of_revisions(self, revision_ids):
2130
"""Find all parent ids that are mentioned in the revision graph.
2132
:return: set of revisions that are parents of revision_ids which are
2133
not part of revision_ids themselves
2135
parent_map = self.get_parent_map(revision_ids)
2137
map(parent_ids.update, parent_map.itervalues())
2138
parent_ids.difference_update(revision_ids)
2139
parent_ids.discard(_mod_revision.NULL_REVISION)
2142
def _find_parent_keys_of_revisions(self, revision_keys):
2143
"""Similar to _find_parent_ids_of_revisions, but used with keys.
2145
:param revision_keys: An iterable of revision_keys.
2146
:return: The parents of all revision_keys that are not already in
2149
parent_map = self.revisions.get_parent_map(revision_keys)
2151
map(parent_keys.update, parent_map.itervalues())
2152
parent_keys.difference_update(revision_keys)
2153
parent_keys.discard(_mod_revision.NULL_REVISION)
2156
1469
def fileids_altered_by_revision_ids(self, revision_ids, _inv_weave=None):
2157
1470
"""Find the file ids and versions affected by revisions.
2362
1672
"""Get Inventory object by revision id."""
2363
1673
return self.iter_inventories([revision_id]).next()
2365
def iter_inventories(self, revision_ids, ordering=None):
1675
def iter_inventories(self, revision_ids):
2366
1676
"""Get many inventories by revision_ids.
2368
1678
This will buffer some or all of the texts used in constructing the
2369
1679
inventories in memory, but will only parse a single inventory at a
2372
:param revision_ids: The expected revision ids of the inventories.
2373
:param ordering: optional ordering, e.g. 'topological'. If not
2374
specified, the order of revision_ids will be preserved (by
2375
buffering if necessary).
2376
1682
:return: An iterator of inventories.
2378
1684
if ((None in revision_ids)
2379
1685
or (_mod_revision.NULL_REVISION in revision_ids)):
2380
1686
raise ValueError('cannot get null revision inventory')
2381
return self._iter_inventories(revision_ids, ordering)
1687
return self._iter_inventories(revision_ids)
2383
def _iter_inventories(self, revision_ids, ordering):
1689
def _iter_inventories(self, revision_ids):
2384
1690
"""single-document based inventory iteration."""
2385
inv_xmls = self._iter_inventory_xmls(revision_ids, ordering)
2386
for text, revision_id in inv_xmls:
2387
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)
2389
def _iter_inventory_xmls(self, revision_ids, ordering):
2390
if ordering is None:
2391
order_as_requested = True
2392
ordering = 'unordered'
2394
order_as_requested = False
1694
def _iter_inventory_xmls(self, revision_ids):
2395
1695
keys = [(revision_id,) for revision_id in revision_ids]
2398
if order_as_requested:
2399
key_iter = iter(keys)
2400
next_key = key_iter.next()
2401
stream = self.inventories.get_record_stream(keys, ordering, True)
1696
stream = self.inventories.get_record_stream(keys, 'unordered', True)
2402
1697
text_chunks = {}
2403
1698
for record in stream:
2404
1699
if record.storage_kind != 'absent':
2405
chunks = record.get_bytes_as('chunked')
2406
if order_as_requested:
2407
text_chunks[record.key] = chunks
2409
yield ''.join(chunks), record.key[-1]
1700
text_chunks[record.key] = record.get_bytes_as('chunked')
2411
1702
raise errors.NoSuchRevision(self, record.key)
2412
if order_as_requested:
2413
# Yield as many results as we can while preserving order.
2414
while next_key in text_chunks:
2415
chunks = text_chunks.pop(next_key)
2416
yield ''.join(chunks), next_key[-1]
2418
next_key = key_iter.next()
2419
except StopIteration:
2420
# We still want to fully consume the get_record_stream,
2421
# just in case it is not actually finished at this point
1704
chunks = text_chunks.pop(key)
1705
yield ''.join(chunks), key[-1]
2425
def _deserialise_inventory(self, revision_id, xml):
1707
def deserialise_inventory(self, revision_id, xml):
2426
1708
"""Transform the xml into an inventory object.
2428
1710
:param revision_id: The expected revision id of the inventory.
2429
1711
:param xml: A serialised inventory.
2431
1713
result = self._serializer.read_inventory_from_string(xml, revision_id,
2432
entry_cache=self._inventory_entry_cache,
2433
return_from_cache=self._safe_to_return_from_cache)
1714
entry_cache=self._inventory_entry_cache)
2434
1715
if result.revision_id != revision_id:
2435
1716
raise AssertionError('revision id mismatch %s != %s' % (
2436
1717
result.revision_id, revision_id))
2439
def _serialise_inventory(self, inv):
1720
def serialise_inventory(self, inv):
2440
1721
return self._serializer.write_inventory_to_string(inv)
2442
1723
def _serialise_inventory_to_lines(self, inv):
2446
1727
return self._serializer.format_num
2448
1729
@needs_read_lock
2449
def _get_inventory_xml(self, revision_id):
2450
"""Get serialized inventory as a string."""
2451
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])
2453
1734
text, revision_id = texts.next()
2454
1735
except StopIteration:
2455
1736
raise errors.HistoryMissing(self, 'inventory', revision_id)
2458
def get_rev_id_for_revno(self, revno, known_pair):
2459
"""Return the revision id of a revno, given a later (revno, revid)
2460
pair in the same history.
2462
:return: if found (True, revid). If the available history ran out
2463
before reaching the revno, then this returns
2464
(False, (closest_revno, closest_revid)).
1740
def get_inventory_sha1(self, revision_id):
1741
"""Return the sha1 hash of the inventory entry
2466
known_revno, known_revid = known_pair
2467
partial_history = [known_revid]
2468
distance_from_known = known_revno - revno
2469
if distance_from_known < 0:
2471
'requested revno (%d) is later than given known revno (%d)'
2472
% (revno, known_revno))
2475
self, partial_history, stop_index=distance_from_known)
2476
except errors.RevisionNotPresent, err:
2477
if err.revision_id == known_revid:
2478
# The start revision (known_revid) wasn't found.
2480
# This is a stacked repository with no fallbacks, or a there's a
2481
# left-hand ghost. Either way, even though the revision named in
2482
# the error isn't in this repo, we know it's the next step in this
2483
# left-hand history.
2484
partial_history.append(err.revision_id)
2485
if len(partial_history) <= distance_from_known:
2486
# Didn't find enough history to get a revid for the revno.
2487
earliest_revno = known_revno - len(partial_history) + 1
2488
return (False, (earliest_revno, partial_history[-1]))
2489
if len(partial_history) - 1 > distance_from_known:
2490
raise AssertionError('_iter_for_revno returned too much history')
2491
return (True, partial_history[-1])
1743
return self.get_revision(revision_id).inventory_sha1
2493
1745
def iter_reverse_revision_history(self, revision_id):
2494
1746
"""Iterate backwards through revision ids in the lefthand history
3713
2936
return self.source.revision_ids_to_search_result(result_set)
3716
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):
3719
3078
def _get_repo_format_to_test(self):
3723
3082
def is_compatible(source, target):
3724
3083
"""Be compatible with Knit2 source and Knit3 target"""
3725
# This is redundant with format.check_conversion_target(), however that
3726
# raises an exception, and we just want to say "False" as in we won't
3727
# support converting between these formats.
3728
if 'IDS_never' in debug.debug_flags:
3730
if source.supports_rich_root() and not target.supports_rich_root():
3732
if (source._format.supports_tree_reference
3733
and not target._format.supports_tree_reference):
3735
if target._fallback_repositories and target._format.supports_chks:
3736
# IDS doesn't know how to copy CHKs for the parent inventories it
3737
# adds to stacked repos.
3739
if 'IDS_always' in debug.debug_flags:
3741
# Only use this code path for local source and target. IDS does far
3742
# too much IO (both bandwidth and roundtrips) over a network.
3743
if not source.bzrdir.transport.base.startswith('file:///'):
3745
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)):
3749
def _get_trees(self, revision_ids, cache):
3751
for rev_id in revision_ids:
3753
possible_trees.append((rev_id, cache[rev_id]))
3755
# Not cached, but inventory might be present anyway.
3757
tree = self.source.revision_tree(rev_id)
3758
except errors.NoSuchRevision:
3759
# Nope, parent is ghost.
3762
cache[rev_id] = tree
3763
possible_trees.append((rev_id, tree))
3764
return possible_trees
3766
def _get_delta_for_revision(self, tree, parent_ids, possible_trees):
3093
def _get_delta_for_revision(self, tree, parent_ids, basis_id, cache):
3767
3094
"""Get the best delta and base for this revision.
3769
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]))
3772
# Generate deltas against each tree, to find the shortest.
3773
texts_possibly_new_in_tree = set()
3774
3106
for basis_id, basis_tree in possible_trees:
3775
3107
delta = tree.inventory._make_delta(basis_tree.inventory)
3776
for old_path, new_path, file_id, new_entry in delta:
3777
if new_path is None:
3778
# This file_id isn't present in the new rev, so we don't
3782
# Rich roots are handled elsewhere...
3784
kind = new_entry.kind
3785
if kind != 'directory' and kind != 'file':
3786
# No text record associated with this inventory entry.
3788
# This is a directory or file that has changed somehow.
3789
texts_possibly_new_in_tree.add((file_id, new_entry.revision))
3790
3108
deltas.append((len(delta), basis_id, delta))
3792
3110
return deltas[0][1:]
3794
def _fetch_parent_invs_for_stacking(self, parent_map, cache):
3795
"""Find all parent revisions that are absent, but for which the
3796
inventory is present, and copy those inventories.
3798
This is necessary to preserve correctness when the source is stacked
3799
without fallbacks configured. (Note that in cases like upgrade the
3800
source may be not have _fallback_repositories even though it is
3804
for parents in parent_map.values():
3805
parent_revs.update(parents)
3806
present_parents = self.source.get_parent_map(parent_revs)
3807
absent_parents = set(parent_revs).difference(present_parents)
3808
parent_invs_keys_for_stacking = self.source.inventories.get_parent_map(
3809
(rev_id,) for rev_id in absent_parents)
3810
parent_inv_ids = [key[-1] for key in parent_invs_keys_for_stacking]
3811
for parent_tree in self.source.revision_trees(parent_inv_ids):
3812
current_revision_id = parent_tree.get_revision_id()
3813
parents_parents_keys = parent_invs_keys_for_stacking[
3814
(current_revision_id,)]
3815
parents_parents = [key[-1] for key in parents_parents_keys]
3816
basis_id = _mod_revision.NULL_REVISION
3817
basis_tree = self.source.revision_tree(basis_id)
3818
delta = parent_tree.inventory._make_delta(basis_tree.inventory)
3819
self.target.add_inventory_by_delta(
3820
basis_id, delta, current_revision_id, parents_parents)
3821
cache[current_revision_id] = parent_tree
3823
def _fetch_batch(self, revision_ids, basis_id, cache, a_graph=None):
3112
def _fetch_batch(self, revision_ids, basis_id, cache):
3824
3113
"""Fetch across a few revisions.
3826
3115
:param revision_ids: The revisions to copy
3827
3116
:param basis_id: The revision_id of a tree that must be in cache, used
3828
3117
as a basis for delta when no other base is available
3829
3118
:param cache: A cache of RevisionTrees that we can use.
3830
:param a_graph: A Graph object to determine the heads() of the
3831
rich-root data stream.
3832
3119
:return: The revision_id of the last converted tree. The RevisionTree
3833
3120
for it will be in cache
3835
3122
# Walk though all revisions; get inventory deltas, copy referenced
3836
3123
# texts that delta references, insert the delta, revision and
3838
root_keys_to_create = set()
3839
3125
text_keys = set()
3840
3126
pending_deltas = []
3841
3127
pending_revisions = []
3842
3128
parent_map = self.source.get_parent_map(revision_ids)
3843
self._fetch_parent_invs_for_stacking(parent_map, cache)
3844
self.source._safe_to_return_from_cache = True
3845
3129
for tree in self.source.revision_trees(revision_ids):
3846
# Find a inventory delta for this revision.
3847
# Find text entries that need to be copied, too.
3848
3130
current_revision_id = tree.get_revision_id()
3849
3131
parent_ids = parent_map.get(current_revision_id, ())
3850
parent_trees = self._get_trees(parent_ids, cache)
3851
possible_trees = list(parent_trees)
3852
if len(possible_trees) == 0:
3853
# There either aren't any parents, or the parents are ghosts,
3854
# so just use the last converted tree.
3855
possible_trees.append((basis_id, cache[basis_id]))
3856
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))
3858
3145
revision = self.source.get_revision(current_revision_id)
3859
3146
pending_deltas.append((basis_id, delta,
3860
3147
current_revision_id, revision.parent_ids))
3861
if self._converting_to_rich_root:
3862
self._revision_id_to_root_id[current_revision_id] = \
3864
# Determine which texts are in present in this revision but not in
3865
# any of the available parents.
3866
texts_possibly_new_in_tree = set()
3867
for old_path, new_path, file_id, entry in delta:
3868
if new_path is None:
3869
# This file_id isn't present in the new rev
3873
if not self.target.supports_rich_root():
3874
# The target doesn't support rich root, so we don't
3877
if self._converting_to_rich_root:
3878
# This can't be copied normally, we have to insert
3880
root_keys_to_create.add((file_id, entry.revision))
3883
texts_possibly_new_in_tree.add((file_id, entry.revision))
3884
for basis_id, basis_tree in possible_trees:
3885
basis_inv = basis_tree.inventory
3886
for file_key in list(texts_possibly_new_in_tree):
3887
file_id, file_revision = file_key
3889
entry = basis_inv[file_id]
3890
except errors.NoSuchId:
3892
if entry.revision == file_revision:
3893
texts_possibly_new_in_tree.remove(file_key)
3894
text_keys.update(texts_possibly_new_in_tree)
3895
3148
pending_revisions.append(revision)
3896
3149
cache[current_revision_id] = tree
3897
3150
basis_id = current_revision_id
3898
self.source._safe_to_return_from_cache = False
3899
3151
# Copy file texts
3900
3152
from_texts = self.source.texts
3901
3153
to_texts = self.target.texts
3902
if root_keys_to_create:
3903
root_stream = _mod_fetch._new_root_data_stream(
3904
root_keys_to_create, self._revision_id_to_root_id, parent_map,
3905
self.source, graph=a_graph)
3906
to_texts.insert_record_stream(root_stream)
3907
3154
to_texts.insert_record_stream(from_texts.get_record_stream(
3908
3155
text_keys, self.target._format._fetch_order,
3909
3156
not self.target._format._fetch_uses_deltas))
3910
# insert inventory deltas
3911
3158
for delta in pending_deltas:
3912
3159
self.target.add_inventory_by_delta(*delta)
3913
if self.target._fallback_repositories:
3914
# Make sure this stacked repository has all the parent inventories
3915
# for the new revisions that we are about to insert. We do this
3916
# before adding the revisions so that no revision is added until
3917
# all the inventories it may depend on are added.
3918
# Note that this is overzealous, as we may have fetched these in an
3921
revision_ids = set()
3922
for revision in pending_revisions:
3923
revision_ids.add(revision.revision_id)
3924
parent_ids.update(revision.parent_ids)
3925
parent_ids.difference_update(revision_ids)
3926
parent_ids.discard(_mod_revision.NULL_REVISION)
3927
parent_map = self.source.get_parent_map(parent_ids)
3928
# we iterate over parent_map and not parent_ids because we don't
3929
# want to try copying any revision which is a ghost
3930
for parent_tree in self.source.revision_trees(parent_map):
3931
current_revision_id = parent_tree.get_revision_id()
3932
parents_parents = parent_map[current_revision_id]
3933
possible_trees = self._get_trees(parents_parents, cache)
3934
if len(possible_trees) == 0:
3935
# There either aren't any parents, or the parents are
3936
# ghosts, so just use the last converted tree.
3937
possible_trees.append((basis_id, cache[basis_id]))
3938
basis_id, delta = self._get_delta_for_revision(parent_tree,
3939
parents_parents, possible_trees)
3940
self.target.add_inventory_by_delta(
3941
basis_id, delta, current_revision_id, parents_parents)
3942
3160
# insert signatures and revisions
3943
3161
for revision in pending_revisions:
3964
3182
cache = lru_cache.LRUCache(100)
3965
3183
cache[basis_id] = basis_tree
3966
3184
del basis_tree # We don't want to hang on to it here
3968
if self._converting_to_rich_root and len(revision_ids) > 100:
3969
a_graph = _mod_fetch._get_rich_root_heads_graph(self.source,
3974
3185
for offset in range(0, len(revision_ids), batch_size):
3975
3186
self.target.start_write_group()
3977
3188
pb.update('Transferring revisions', offset,
3978
3189
len(revision_ids))
3979
3190
batch = revision_ids[offset:offset+batch_size]
3980
basis_id = self._fetch_batch(batch, basis_id, cache,
3191
basis_id = self._fetch_batch(batch, basis_id, cache)
3983
self.source._safe_to_return_from_cache = False
3984
3193
self.target.abort_write_group()
3987
hint = self.target.commit_write_group()
3990
if hints and self.target._format.pack_compresses:
3991
self.target.pack(hint=hints)
3196
self.target.commit_write_group()
3992
3197
pb.update('Transferring revisions', len(revision_ids),
3993
3198
len(revision_ids))
3995
3200
@needs_write_lock
3996
def fetch(self, revision_id=None, pb=None, find_ghosts=False,
3201
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3998
3202
"""See InterRepository.fetch()."""
3999
if fetch_spec is not None:
4000
raise AssertionError("Not implemented yet...")
4001
# See <https://launchpad.net/bugs/456077> asking for a warning here
4003
# nb this is only active for local-local fetches; other things using
4005
ui.ui_factory.warn_cross_format_fetch(self.source._format,
4006
self.target._format)
4007
if (not self.source.supports_rich_root()
4008
and self.target.supports_rich_root()):
4009
self._converting_to_rich_root = True
4010
self._revision_id_to_root_id = {}
4012
self._converting_to_rich_root = False
4013
3203
revision_ids = self.target.search_missing_revision_ids(self.source,
4014
3204
revision_id, find_ghosts=find_ghosts).get_keys()
4015
3205
if not revision_ids:
4017
3207
revision_ids = tsort.topo_sort(
4018
3208
self.source.get_graph().get_parent_map(revision_ids))
4019
if not revision_ids:
4021
# Walk though all revisions; get inventory deltas, copy referenced
4022
# texts that delta references, insert the delta, revision and
4025
3210
my_pb = ui.ui_factory.nested_progress_bar()
4028
symbol_versioning.warn(
4029
symbol_versioning.deprecated_in((1, 14, 0))
4030
% "pb parameter to fetch()")
4033
3215
self._fetch_all_revisions(revision_ids, pb)
4059
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):
4062
3363
InterRepository.register_optimiser(InterDifferingSerializer)
4063
3364
InterRepository.register_optimiser(InterSameDataRepository)
4064
3365
InterRepository.register_optimiser(InterWeaveRepo)
4065
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)
4068
3373
class CopyConverter(object):
4359
3610
# missing keys can handle suspending a write group).
4360
3611
write_group_tokens = self.target_repo.suspend_write_group()
4361
3612
return write_group_tokens, missing_keys
4362
hint = self.target_repo.commit_write_group()
4363
if (to_serializer != src_serializer and
4364
self.target_repo._format.pack_compresses):
4365
self.target_repo.pack(hint=hint)
3613
self.target_repo.commit_write_group()
4366
3614
return [], set()
4368
def _extract_and_insert_inventory_deltas(self, substream, serializer):
4369
target_rich_root = self.target_repo._format.rich_root_data
4370
target_tree_refs = self.target_repo._format.supports_tree_reference
4371
for record in substream:
4372
# Insert the delta directly
4373
inventory_delta_bytes = record.get_bytes_as('fulltext')
4374
deserialiser = inventory_delta.InventoryDeltaDeserializer()
4376
parse_result = deserialiser.parse_text_bytes(
4377
inventory_delta_bytes)
4378
except inventory_delta.IncompatibleInventoryDelta, err:
4379
trace.mutter("Incompatible delta: %s", err.msg)
4380
raise errors.IncompatibleRevision(self.target_repo._format)
4381
basis_id, new_id, rich_root, tree_refs, inv_delta = parse_result
4382
revision_id = new_id
4383
parents = [key[0] for key in record.parents]
4384
self.target_repo.add_inventory_by_delta(
4385
basis_id, inv_delta, revision_id, parents)
4387
def _extract_and_insert_inventories(self, substream, serializer,
3616
def _extract_and_insert_inventories(self, substream, serializer):
4389
3617
"""Generate a new inventory versionedfile in target, converting data.
4391
3619
The inventory is retrieved from the source, (deserializing it), and
4392
3620
stored in the target (reserializing it in a different format).
4394
target_rich_root = self.target_repo._format.rich_root_data
4395
target_tree_refs = self.target_repo._format.supports_tree_reference
4396
3622
for record in substream:
4397
# It's not a delta, so it must be a fulltext in the source
4398
# serializer's format.
4399
3623
bytes = record.get_bytes_as('fulltext')
4400
3624
revision_id = record.key[0]
4401
3625
inv = serializer.read_inventory_from_string(bytes, revision_id)
4402
3626
parents = [key[0] for key in record.parents]
4403
3627
self.target_repo.add_inventory(revision_id, inv, parents)
4404
# No need to keep holding this full inv in memory when the rest of
4405
# the substream is likely to be all deltas.
4408
3629
def _extract_and_insert_revisions(self, substream, serializer):
4409
3630
for record in substream:
4572
3778
return (not self.from_repository._format.rich_root_data and
4573
3779
self.to_format.rich_root_data)
4575
def _get_inventory_stream(self, revision_ids, missing=False):
4576
from_format = self.from_repository._format
4577
if (from_format.supports_chks and self.to_format.supports_chks and
4578
from_format.network_name() == self.to_format.network_name()):
4579
raise AssertionError(
4580
"this case should be handled by GroupCHKStreamSource")
4581
elif 'forceinvdeltas' in debug.debug_flags:
4582
return self._get_convertable_inventory_stream(revision_ids,
4583
delta_versus_null=missing)
4584
elif from_format.network_name() == self.to_format.network_name():
4586
return self._get_simple_inventory_stream(revision_ids,
4588
elif (not from_format.supports_chks and not self.to_format.supports_chks
4589
and from_format._serializer == self.to_format._serializer):
4590
# Essentially the same format.
4591
return self._get_simple_inventory_stream(revision_ids,
4594
# Any time we switch serializations, we want to use an
4595
# inventory-delta based approach.
4596
return self._get_convertable_inventory_stream(revision_ids,
4597
delta_versus_null=missing)
4599
def _get_simple_inventory_stream(self, revision_ids, missing=False):
4600
# NB: This currently reopens the inventory weave in source;
4601
# using a single stream interface instead would avoid this.
4602
from_weave = self.from_repository.inventories
4604
delta_closure = True
4606
delta_closure = not self.delta_on_metadata()
4607
yield ('inventories', from_weave.get_record_stream(
4608
[(rev_id,) for rev_id in revision_ids],
4609
self.inventory_fetch_order(), delta_closure))
4611
def _get_convertable_inventory_stream(self, revision_ids,
4612
delta_versus_null=False):
4613
# The two formats are sufficiently different that there is no fast
4614
# path, so we need to send just inventorydeltas, which any
4615
# sufficiently modern client can insert into any repository.
4616
# The StreamSink code expects to be able to
4617
# convert on the target, so we need to put bytes-on-the-wire that can
4618
# be converted. That means inventory deltas (if the remote is <1.19,
4619
# RemoteStreamSink will fallback to VFS to insert the deltas).
4620
yield ('inventory-deltas',
4621
self._stream_invs_as_deltas(revision_ids,
4622
delta_versus_null=delta_versus_null))
4624
def _stream_invs_as_deltas(self, revision_ids, delta_versus_null=False):
4625
"""Return a stream of inventory-deltas for the given rev ids.
4627
:param revision_ids: The list of inventories to transmit
4628
:param delta_versus_null: Don't try to find a minimal delta for this
4629
entry, instead compute the delta versus the NULL_REVISION. This
4630
effectively streams a complete inventory. Used for stuff like
4631
filling in missing parents, etc.
4633
from_repo = self.from_repository
4634
revision_keys = [(rev_id,) for rev_id in revision_ids]
4635
parent_map = from_repo.inventories.get_parent_map(revision_keys)
4636
# XXX: possibly repos could implement a more efficient iter_inv_deltas
4638
inventories = self.from_repository.iter_inventories(
4639
revision_ids, 'topological')
4640
format = from_repo._format
4641
invs_sent_so_far = set([_mod_revision.NULL_REVISION])
4642
inventory_cache = lru_cache.LRUCache(50)
4643
null_inventory = from_repo.revision_tree(
4644
_mod_revision.NULL_REVISION).inventory
4645
# XXX: ideally the rich-root/tree-refs flags would be per-revision, not
4646
# per-repo (e.g. streaming a non-rich-root revision out of a rich-root
4647
# repo back into a non-rich-root repo ought to be allowed)
4648
serializer = inventory_delta.InventoryDeltaSerializer(
4649
versioned_root=format.rich_root_data,
4650
tree_references=format.supports_tree_reference)
4651
for inv in inventories:
4652
key = (inv.revision_id,)
4653
parent_keys = parent_map.get(key, ())
4655
if not delta_versus_null and parent_keys:
4656
# The caller did not ask for complete inventories and we have
4657
# some parents that we can delta against. Make a delta against
4658
# each parent so that we can find the smallest.
4659
parent_ids = [parent_key[0] for parent_key in parent_keys]
4660
for parent_id in parent_ids:
4661
if parent_id not in invs_sent_so_far:
4662
# We don't know that the remote side has this basis, so
4665
if parent_id == _mod_revision.NULL_REVISION:
4666
parent_inv = null_inventory
4668
parent_inv = inventory_cache.get(parent_id, None)
4669
if parent_inv is None:
4670
parent_inv = from_repo.get_inventory(parent_id)
4671
candidate_delta = inv._make_delta(parent_inv)
4672
if (delta is None or
4673
len(delta) > len(candidate_delta)):
4674
delta = candidate_delta
4675
basis_id = parent_id
4677
# Either none of the parents ended up being suitable, or we
4678
# were asked to delta against NULL
4679
basis_id = _mod_revision.NULL_REVISION
4680
delta = inv._make_delta(null_inventory)
4681
invs_sent_so_far.add(inv.revision_id)
4682
inventory_cache[inv.revision_id] = inv
4683
delta_serialized = ''.join(
4684
serializer.delta_to_lines(basis_id, key[-1], delta))
4685
yield versionedfile.FulltextContentFactory(
4686
key, parent_keys, None, delta_serialized)
4689
def _iter_for_revno(repo, partial_history_cache, stop_index=None,
4690
stop_revision=None):
4691
"""Extend the partial history to include a given index
4693
If a stop_index is supplied, stop when that index has been reached.
4694
If a stop_revision is supplied, stop when that revision is
4695
encountered. Otherwise, stop when the beginning of history is
4698
:param stop_index: The index which should be present. When it is
4699
present, history extension will stop.
4700
:param stop_revision: The revision id which should be present. When
4701
it is encountered, history extension will stop.
4703
start_revision = partial_history_cache[-1]
4704
iterator = repo.iter_reverse_revision_history(start_revision)
4706
#skip the last revision in the list
4709
if (stop_index is not None and
4710
len(partial_history_cache) > stop_index):
4712
if partial_history_cache[-1] == stop_revision:
4714
revision_id = iterator.next()
4715
partial_history_cache.append(revision_id)
4716
except StopIteration:
added
removed
bzrlib/repository.py
