1196
1150
# The old API returned a list, should this actually be a set?
1197
1151
return parent_map.keys()
1199
def _check_inventories(self, checker):
1200
"""Check the inventories found from the revision scan.
1202
This is responsible for verifying the sha1 of inventories and
1203
creating a pending_keys set that covers data referenced by inventories.
1205
bar = ui.ui_factory.nested_progress_bar()
1207
self._do_check_inventories(checker, bar)
1211
def _do_check_inventories(self, checker, bar):
1212
"""Helper for _check_inventories."""
1214
keys = {'chk_bytes':set(), 'inventories':set(), 'texts':set()}
1215
kinds = ['chk_bytes', 'texts']
1216
count = len(checker.pending_keys)
1217
bar.update("inventories", 0, 2)
1218
current_keys = checker.pending_keys
1219
checker.pending_keys = {}
1220
# Accumulate current checks.
1221
for key in current_keys:
1222
if key[0] != 'inventories' and key[0] not in kinds:
1223
checker._report_items.append('unknown key type %r' % (key,))
1224
keys[key[0]].add(key[1:])
1225
if keys['inventories']:
1226
# NB: output order *should* be roughly sorted - topo or
1227
# inverse topo depending on repository - either way decent
1228
# to just delta against. However, pre-CHK formats didn't
1229
# try to optimise inventory layout on disk. As such the
1230
# pre-CHK code path does not use inventory deltas.
1232
for record in self.inventories.check(keys=keys['inventories']):
1233
if record.storage_kind == 'absent':
1234
checker._report_items.append(
1235
'Missing inventory {%s}' % (record.key,))
1237
last_object = self._check_record('inventories', record,
1238
checker, last_object,
1239
current_keys[('inventories',) + record.key])
1240
del keys['inventories']
1243
bar.update("texts", 1)
1244
while (checker.pending_keys or keys['chk_bytes']
1246
# Something to check.
1247
current_keys = checker.pending_keys
1248
checker.pending_keys = {}
1249
# Accumulate current checks.
1250
for key in current_keys:
1251
if key[0] not in kinds:
1252
checker._report_items.append('unknown key type %r' % (key,))
1253
keys[key[0]].add(key[1:])
1254
# Check the outermost kind only - inventories || chk_bytes || texts
1258
for record in getattr(self, kind).check(keys=keys[kind]):
1259
if record.storage_kind == 'absent':
1260
checker._report_items.append(
1261
'Missing %s {%s}' % (kind, record.key,))
1263
last_object = self._check_record(kind, record,
1264
checker, last_object, current_keys[(kind,) + record.key])
1268
def _check_record(self, kind, record, checker, last_object, item_data):
1269
"""Check a single text from this repository."""
1270
if kind == 'inventories':
1271
rev_id = record.key[0]
1272
inv = self._deserialise_inventory(rev_id,
1273
record.get_bytes_as('fulltext'))
1274
if last_object is not None:
1275
delta = inv._make_delta(last_object)
1276
for old_path, path, file_id, ie in delta:
1279
ie.check(checker, rev_id, inv)
1281
for path, ie in inv.iter_entries():
1282
ie.check(checker, rev_id, inv)
1283
if self._format.fast_deltas:
1285
elif kind == 'chk_bytes':
1286
# No code written to check chk_bytes for this repo format.
1287
checker._report_items.append(
1288
'unsupported key type chk_bytes for %s' % (record.key,))
1289
elif kind == 'texts':
1290
self._check_text(record, checker, item_data)
1292
checker._report_items.append(
1293
'unknown key type %s for %s' % (kind, record.key))
1295
def _check_text(self, record, checker, item_data):
1296
"""Check a single text."""
1297
# Check it is extractable.
1298
# TODO: check length.
1299
if record.storage_kind == 'chunked':
1300
chunks = record.get_bytes_as(record.storage_kind)
1301
sha1 = osutils.sha_strings(chunks)
1302
length = sum(map(len, chunks))
1304
content = record.get_bytes_as('fulltext')
1305
sha1 = osutils.sha_string(content)
1306
length = len(content)
1307
if item_data and sha1 != item_data[1]:
1308
checker._report_items.append(
1309
'sha1 mismatch: %s has sha1 %s expected %s referenced by %s' %
1310
(record.key, sha1, item_data[1], item_data[2]))
1313
1154
def create(a_bzrdir):
1314
1155
"""Construct the current default format repository in a_bzrdir."""
1613
1424
def suspend_write_group(self):
1614
1425
raise errors.UnsuspendableWriteGroup(self)
1616
def get_missing_parent_inventories(self, check_for_missing_texts=True):
1617
"""Return the keys of missing inventory parents for revisions added in
1620
A revision is not complete if the inventory delta for that revision
1621
cannot be calculated. Therefore if the parent inventories of a
1622
revision are not present, the revision is incomplete, and e.g. cannot
1623
be streamed by a smart server. This method finds missing inventory
1624
parents for revisions added in this write group.
1626
if not self._format.supports_external_lookups:
1627
# This is only an issue for stacked repositories
1629
if not self.is_in_write_group():
1630
raise AssertionError('not in a write group')
1632
# XXX: We assume that every added revision already has its
1633
# corresponding inventory, so we only check for parent inventories that
1634
# might be missing, rather than all inventories.
1635
parents = set(self.revisions._index.get_missing_parents())
1636
parents.discard(_mod_revision.NULL_REVISION)
1637
unstacked_inventories = self.inventories._index
1638
present_inventories = unstacked_inventories.get_parent_map(
1639
key[-1:] for key in parents)
1640
parents.difference_update(present_inventories)
1641
if len(parents) == 0:
1642
# No missing parent inventories.
1644
if not check_for_missing_texts:
1645
return set(('inventories', rev_id) for (rev_id,) in parents)
1646
# Ok, now we have a list of missing inventories. But these only matter
1647
# if the inventories that reference them are missing some texts they
1648
# appear to introduce.
1649
# XXX: Texts referenced by all added inventories need to be present,
1650
# but at the moment we're only checking for texts referenced by
1651
# inventories at the graph's edge.
1652
key_deps = self.revisions._index._key_dependencies
1653
key_deps.satisfy_refs_for_keys(present_inventories)
1654
referrers = frozenset(r[0] for r in key_deps.get_referrers())
1655
file_ids = self.fileids_altered_by_revision_ids(referrers)
1656
missing_texts = set()
1657
for file_id, version_ids in file_ids.iteritems():
1658
missing_texts.update(
1659
(file_id, version_id) for version_id in version_ids)
1660
present_texts = self.texts.get_parent_map(missing_texts)
1661
missing_texts.difference_update(present_texts)
1662
if not missing_texts:
1663
# No texts are missing, so all revisions and their deltas are
1666
# Alternatively the text versions could be returned as the missing
1667
# keys, but this is likely to be less data.
1668
missing_keys = set(('inventories', rev_id) for (rev_id,) in parents)
1671
1427
def refresh_data(self):
1672
"""Re-read any data needed to synchronise with disk.
1428
"""Re-read any data needed to to synchronise with disk.
1674
1430
This method is intended to be called after another repository instance
1675
1431
(such as one used by a smart server) has inserted data into the
1676
repository. On all repositories this will work outside of write groups.
1677
Some repository formats (pack and newer for bzrlib native formats)
1678
support refresh_data inside write groups. If called inside a write
1679
group on a repository that does not support refreshing in a write group
1680
IsInWriteGroupError will be raised.
1432
repository. It may not be called during a write group, but may be
1433
called at any other time.
1435
if self.is_in_write_group():
1436
raise errors.InternalBzrError(
1437
"May not refresh_data while in a write group.")
1682
1438
self._refresh_data()
1684
1440
def resume_write_group(self, tokens):
1895
1642
@needs_read_lock
1896
1643
def get_revisions(self, revision_ids):
1897
"""Get many revisions at once.
1899
Repositories that need to check data on every revision read should
1900
subclass this method.
1644
"""Get many revisions at once."""
1902
1645
return self._get_revisions(revision_ids)
1904
1647
@needs_read_lock
1905
1648
def _get_revisions(self, revision_ids):
1906
1649
"""Core work logic to get many revisions without sanity checks."""
1650
for rev_id in revision_ids:
1651
if not rev_id or not isinstance(rev_id, basestring):
1652
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1653
keys = [(key,) for key in revision_ids]
1654
stream = self.revisions.get_record_stream(keys, 'unordered', True)
1908
for revid, rev in self._iter_revisions(revision_ids):
1910
raise errors.NoSuchRevision(self, revid)
1656
for record in stream:
1657
if record.storage_kind == 'absent':
1658
raise errors.NoSuchRevision(self, record.key[0])
1659
text = record.get_bytes_as('fulltext')
1660
rev = self._serializer.read_revision_from_string(text)
1661
revs[record.key[0]] = rev
1912
1662
return [revs[revid] for revid in revision_ids]
1914
def _iter_revisions(self, revision_ids):
1915
"""Iterate over revision objects.
1917
:param revision_ids: An iterable of revisions to examine. None may be
1918
passed to request all revisions known to the repository. Note that
1919
not all repositories can find unreferenced revisions; for those
1920
repositories only referenced ones will be returned.
1921
:return: An iterator of (revid, revision) tuples. Absent revisions (
1922
those asked for but not available) are returned as (revid, None).
1924
if revision_ids is None:
1925
revision_ids = self.all_revision_ids()
1927
for rev_id in revision_ids:
1928
if not rev_id or not isinstance(rev_id, basestring):
1929
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1930
keys = [(key,) for key in revision_ids]
1931
stream = self.revisions.get_record_stream(keys, 'unordered', True)
1932
for record in stream:
1933
revid = record.key[0]
1934
if record.storage_kind == 'absent':
1937
text = record.get_bytes_as('fulltext')
1938
rev = self._serializer.read_revision_from_string(text)
1665
def get_revision_xml(self, revision_id):
1666
# TODO: jam 20070210 This shouldn't be necessary since get_revision
1667
# would have already do it.
1668
# TODO: jam 20070210 Just use _serializer.write_revision_to_string()
1669
# TODO: this can't just be replaced by:
1670
# return self._serializer.write_revision_to_string(
1671
# self.get_revision(revision_id))
1672
# as cStringIO preservers the encoding unlike write_revision_to_string
1673
# or some other call down the path.
1674
rev = self.get_revision(revision_id)
1675
rev_tmp = cStringIO.StringIO()
1676
# the current serializer..
1677
self._serializer.write_revision(rev, rev_tmp)
1679
return rev_tmp.getvalue()
1941
1681
def get_deltas_for_revisions(self, revisions, specific_fileids=None):
1942
1682
"""Produce a generator of revision deltas.
2392
2114
:param revision_ids: The expected revision ids of the inventories.
2393
:param ordering: optional ordering, e.g. 'topological'. If not
2394
specified, the order of revision_ids will be preserved (by
2395
buffering if necessary).
2396
2115
:return: An iterator of inventories.
2398
2117
if ((None in revision_ids)
2399
2118
or (_mod_revision.NULL_REVISION in revision_ids)):
2400
2119
raise ValueError('cannot get null revision inventory')
2401
return self._iter_inventories(revision_ids, ordering)
2120
return self._iter_inventories(revision_ids)
2403
def _iter_inventories(self, revision_ids, ordering):
2122
def _iter_inventories(self, revision_ids):
2404
2123
"""single-document based inventory iteration."""
2405
inv_xmls = self._iter_inventory_xmls(revision_ids, ordering)
2406
for text, revision_id in inv_xmls:
2407
yield self._deserialise_inventory(revision_id, text)
2124
for text, revision_id in self._iter_inventory_xmls(revision_ids):
2125
yield self.deserialise_inventory(revision_id, text)
2409
def _iter_inventory_xmls(self, revision_ids, ordering):
2410
if ordering is None:
2411
order_as_requested = True
2412
ordering = 'unordered'
2414
order_as_requested = False
2127
def _iter_inventory_xmls(self, revision_ids):
2415
2128
keys = [(revision_id,) for revision_id in revision_ids]
2418
if order_as_requested:
2419
key_iter = iter(keys)
2420
next_key = key_iter.next()
2421
stream = self.inventories.get_record_stream(keys, ordering, True)
2129
stream = self.inventories.get_record_stream(keys, 'unordered', True)
2422
2130
text_chunks = {}
2423
2131
for record in stream:
2424
2132
if record.storage_kind != 'absent':
2425
chunks = record.get_bytes_as('chunked')
2426
if order_as_requested:
2427
text_chunks[record.key] = chunks
2429
yield ''.join(chunks), record.key[-1]
2133
text_chunks[record.key] = record.get_bytes_as('chunked')
2431
2135
raise errors.NoSuchRevision(self, record.key)
2432
if order_as_requested:
2433
# Yield as many results as we can while preserving order.
2434
while next_key in text_chunks:
2435
chunks = text_chunks.pop(next_key)
2436
yield ''.join(chunks), next_key[-1]
2438
next_key = key_iter.next()
2439
except StopIteration:
2440
# We still want to fully consume the get_record_stream,
2441
# just in case it is not actually finished at this point
2137
chunks = text_chunks.pop(key)
2138
yield ''.join(chunks), key[-1]
2445
def _deserialise_inventory(self, revision_id, xml):
2140
def deserialise_inventory(self, revision_id, xml):
2446
2141
"""Transform the xml into an inventory object.
2448
2143
:param revision_id: The expected revision id of the inventory.
2449
2144
:param xml: A serialised inventory.
2451
2146
result = self._serializer.read_inventory_from_string(xml, revision_id,
2452
entry_cache=self._inventory_entry_cache,
2453
return_from_cache=self._safe_to_return_from_cache)
2147
entry_cache=self._inventory_entry_cache)
2454
2148
if result.revision_id != revision_id:
2455
2149
raise AssertionError('revision id mismatch %s != %s' % (
2456
2150
result.revision_id, revision_id))
2153
def serialise_inventory(self, inv):
2154
return self._serializer.write_inventory_to_string(inv)
2156
def _serialise_inventory_to_lines(self, inv):
2157
return self._serializer.write_inventory_to_lines(inv)
2459
2159
def get_serializer_format(self):
2460
2160
return self._serializer.format_num
2462
2162
@needs_read_lock
2463
def _get_inventory_xml(self, revision_id):
2464
"""Get serialized inventory as a string."""
2465
texts = self._iter_inventory_xmls([revision_id], 'unordered')
2163
def get_inventory_xml(self, revision_id):
2164
"""Get inventory XML as a file object."""
2165
texts = self._iter_inventory_xmls([revision_id])
2467
2167
text, revision_id = texts.next()
2468
2168
except StopIteration:
2469
2169
raise errors.HistoryMissing(self, 'inventory', revision_id)
2472
def get_rev_id_for_revno(self, revno, known_pair):
2473
"""Return the revision id of a revno, given a later (revno, revid)
2474
pair in the same history.
2476
:return: if found (True, revid). If the available history ran out
2477
before reaching the revno, then this returns
2478
(False, (closest_revno, closest_revid)).
2173
def get_inventory_sha1(self, revision_id):
2174
"""Return the sha1 hash of the inventory entry
2480
known_revno, known_revid = known_pair
2481
partial_history = [known_revid]
2482
distance_from_known = known_revno - revno
2483
if distance_from_known < 0:
2485
'requested revno (%d) is later than given known revno (%d)'
2486
% (revno, known_revno))
2489
self, partial_history, stop_index=distance_from_known)
2490
except errors.RevisionNotPresent, err:
2491
if err.revision_id == known_revid:
2492
# The start revision (known_revid) wasn't found.
2494
# This is a stacked repository with no fallbacks, or a there's a
2495
# left-hand ghost. Either way, even though the revision named in
2496
# the error isn't in this repo, we know it's the next step in this
2497
# left-hand history.
2498
partial_history.append(err.revision_id)
2499
if len(partial_history) <= distance_from_known:
2500
# Didn't find enough history to get a revid for the revno.
2501
earliest_revno = known_revno - len(partial_history) + 1
2502
return (False, (earliest_revno, partial_history[-1]))
2503
if len(partial_history) - 1 > distance_from_known:
2504
raise AssertionError('_iter_for_revno returned too much history')
2505
return (True, partial_history[-1])
2176
return self.get_revision(revision_id).inventory_sha1
2507
2178
def iter_reverse_revision_history(self, revision_id):
2508
2179
"""Iterate backwards through revision ids in the lefthand history
3741
3356
return self.source.revision_ids_to_search_result(result_set)
3359
class InterPackRepo(InterSameDataRepository):
3360
"""Optimised code paths between Pack based repositories."""
3363
def _get_repo_format_to_test(self):
3364
from bzrlib.repofmt import pack_repo
3365
return pack_repo.RepositoryFormatKnitPack6RichRoot()
3368
def is_compatible(source, target):
3369
"""Be compatible with known Pack formats.
3371
We don't test for the stores being of specific types because that
3372
could lead to confusing results, and there is no need to be
3375
InterPackRepo does not support CHK based repositories.
3377
from bzrlib.repofmt.pack_repo import RepositoryFormatPack
3378
from bzrlib.repofmt.groupcompress_repo import RepositoryFormatCHK1
3380
are_packs = (isinstance(source._format, RepositoryFormatPack) and
3381
isinstance(target._format, RepositoryFormatPack))
3382
not_packs = (isinstance(source._format, RepositoryFormatCHK1) or
3383
isinstance(target._format, RepositoryFormatCHK1))
3384
except AttributeError:
3386
if not_packs or not are_packs:
3388
return InterRepository._same_model(source, target)
3391
def fetch(self, revision_id=None, pb=None, find_ghosts=False,
3393
"""See InterRepository.fetch()."""
3394
if (len(self.source._fallback_repositories) > 0 or
3395
len(self.target._fallback_repositories) > 0):
3396
# The pack layer is not aware of fallback repositories, so when
3397
# fetching from a stacked repository or into a stacked repository
3398
# we use the generic fetch logic which uses the VersionedFiles
3399
# attributes on repository.
3400
from bzrlib.fetch import RepoFetcher
3401
fetcher = RepoFetcher(self.target, self.source, revision_id,
3402
pb, find_ghosts, fetch_spec=fetch_spec)
3403
if fetch_spec is not None:
3404
if len(list(fetch_spec.heads)) != 1:
3405
raise AssertionError(
3406
"InterPackRepo.fetch doesn't support "
3407
"fetching multiple heads yet.")
3408
revision_id = list(fetch_spec.heads)[0]
3410
if revision_id is None:
3412
# everything to do - use pack logic
3413
# to fetch from all packs to one without
3414
# inventory parsing etc, IFF nothing to be copied is in the target.
3416
source_revision_ids = frozenset(self.source.all_revision_ids())
3417
revision_ids = source_revision_ids - \
3418
frozenset(self.target.get_parent_map(source_revision_ids))
3419
revision_keys = [(revid,) for revid in revision_ids]
3420
index = self.target._pack_collection.revision_index.combined_index
3421
present_revision_ids = set(item[1][0] for item in
3422
index.iter_entries(revision_keys))
3423
revision_ids = set(revision_ids) - present_revision_ids
3424
# implementing the TODO will involve:
3425
# - detecting when all of a pack is selected
3426
# - avoiding as much as possible pre-selection, so the
3427
# more-core routines such as create_pack_from_packs can filter in
3428
# a just-in-time fashion. (though having a HEADS list on a
3429
# repository might make this a lot easier, because we could
3430
# sensibly detect 'new revisions' without doing a full index scan.
3431
elif _mod_revision.is_null(revision_id):
3436
revision_ids = self.search_missing_revision_ids(revision_id,
3437
find_ghosts=find_ghosts).get_keys()
3438
except errors.NoSuchRevision:
3439
raise errors.InstallFailed([revision_id])
3440
if len(revision_ids) == 0:
3442
return self._pack(self.source, self.target, revision_ids)
3444
def _pack(self, source, target, revision_ids):
3445
from bzrlib.repofmt.pack_repo import Packer
3446
packs = source._pack_collection.all_packs()
3447
pack = Packer(self.target._pack_collection, packs, '.fetch',
3448
revision_ids).pack()
3449
if pack is not None:
3450
self.target._pack_collection._save_pack_names()
3451
copied_revs = pack.get_revision_count()
3452
# Trigger an autopack. This may duplicate effort as we've just done
3453
# a pack creation, but for now it is simpler to think about as
3454
# 'upload data, then repack if needed'.
3455
self.target._pack_collection.autopack()
3456
return (copied_revs, [])
3461
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
3462
"""See InterRepository.missing_revision_ids().
3464
:param find_ghosts: Find ghosts throughout the ancestry of
3467
if not find_ghosts and revision_id is not None:
3468
return self._walk_to_common_revisions([revision_id])
3469
elif revision_id is not None:
3470
# Find ghosts: search for revisions pointing from one repository to
3471
# the other, and vice versa, anywhere in the history of revision_id.
3472
graph = self.target.get_graph(other_repository=self.source)
3473
searcher = graph._make_breadth_first_searcher([revision_id])
3477
next_revs, ghosts = searcher.next_with_ghosts()
3478
except StopIteration:
3480
if revision_id in ghosts:
3481
raise errors.NoSuchRevision(self.source, revision_id)
3482
found_ids.update(next_revs)
3483
found_ids.update(ghosts)
3484
found_ids = frozenset(found_ids)
3485
# Double query here: should be able to avoid this by changing the
3486
# graph api further.
3487
result_set = found_ids - frozenset(
3488
self.target.get_parent_map(found_ids))
3490
source_ids = self.source.all_revision_ids()
3491
# source_ids is the worst possible case we may need to pull.
3492
# now we want to filter source_ids against what we actually
3493
# have in target, but don't try to check for existence where we know
3494
# we do not have a revision as that would be pointless.
3495
target_ids = set(self.target.all_revision_ids())
3496
result_set = set(source_ids).difference(target_ids)
3497
return self.source.revision_ids_to_search_result(result_set)
3744
3500
class InterDifferingSerializer(InterRepository):
3753
3509
# This is redundant with format.check_conversion_target(), however that
3754
3510
# raises an exception, and we just want to say "False" as in we won't
3755
3511
# support converting between these formats.
3756
if 'IDS_never' in debug.debug_flags:
3758
3512
if source.supports_rich_root() and not target.supports_rich_root():
3760
3514
if (source._format.supports_tree_reference
3761
3515
and not target._format.supports_tree_reference):
3763
if target._fallback_repositories and target._format.supports_chks:
3764
# IDS doesn't know how to copy CHKs for the parent inventories it
3765
# adds to stacked repos.
3767
if 'IDS_always' in debug.debug_flags:
3769
# Only use this code path for local source and target. IDS does far
3770
# too much IO (both bandwidth and roundtrips) over a network.
3771
if not source.bzrdir.transport.base.startswith('file:///'):
3773
if not target.bzrdir.transport.base.startswith('file:///'):
3777
def _get_trees(self, revision_ids, cache):
3779
for rev_id in revision_ids:
3781
possible_trees.append((rev_id, cache[rev_id]))
3783
# Not cached, but inventory might be present anyway.
3785
tree = self.source.revision_tree(rev_id)
3786
except errors.NoSuchRevision:
3787
# Nope, parent is ghost.
3790
cache[rev_id] = tree
3791
possible_trees.append((rev_id, tree))
3792
return possible_trees
3794
def _get_delta_for_revision(self, tree, parent_ids, possible_trees):
3519
def _get_delta_for_revision(self, tree, parent_ids, basis_id, cache):
3795
3520
"""Get the best delta and base for this revision.
3797
3522
:return: (basis_id, delta)
3524
possible_trees = [(parent_id, cache[parent_id])
3525
for parent_id in parent_ids
3526
if parent_id in cache]
3527
if len(possible_trees) == 0:
3528
# There either aren't any parents, or the parents aren't in the
3529
# cache, so just use the last converted tree
3530
possible_trees.append((basis_id, cache[basis_id]))
3800
# Generate deltas against each tree, to find the shortest.
3801
texts_possibly_new_in_tree = set()
3802
3532
for basis_id, basis_tree in possible_trees:
3803
3533
delta = tree.inventory._make_delta(basis_tree.inventory)
3804
for old_path, new_path, file_id, new_entry in delta:
3805
if new_path is None:
3806
# This file_id isn't present in the new rev, so we don't
3810
# Rich roots are handled elsewhere...
3812
kind = new_entry.kind
3813
if kind != 'directory' and kind != 'file':
3814
# No text record associated with this inventory entry.
3816
# This is a directory or file that has changed somehow.
3817
texts_possibly_new_in_tree.add((file_id, new_entry.revision))
3818
3534
deltas.append((len(delta), basis_id, delta))
3820
3536
return deltas[0][1:]
3822
def _fetch_parent_invs_for_stacking(self, parent_map, cache):
3823
"""Find all parent revisions that are absent, but for which the
3824
inventory is present, and copy those inventories.
3826
This is necessary to preserve correctness when the source is stacked
3827
without fallbacks configured. (Note that in cases like upgrade the
3828
source may be not have _fallback_repositories even though it is
3832
for parents in parent_map.values():
3833
parent_revs.update(parents)
3834
present_parents = self.source.get_parent_map(parent_revs)
3835
absent_parents = set(parent_revs).difference(present_parents)
3836
parent_invs_keys_for_stacking = self.source.inventories.get_parent_map(
3837
(rev_id,) for rev_id in absent_parents)
3838
parent_inv_ids = [key[-1] for key in parent_invs_keys_for_stacking]
3839
for parent_tree in self.source.revision_trees(parent_inv_ids):
3840
current_revision_id = parent_tree.get_revision_id()
3841
parents_parents_keys = parent_invs_keys_for_stacking[
3842
(current_revision_id,)]
3843
parents_parents = [key[-1] for key in parents_parents_keys]
3844
basis_id = _mod_revision.NULL_REVISION
3845
basis_tree = self.source.revision_tree(basis_id)
3846
delta = parent_tree.inventory._make_delta(basis_tree.inventory)
3847
self.target.add_inventory_by_delta(
3848
basis_id, delta, current_revision_id, parents_parents)
3849
cache[current_revision_id] = parent_tree
3851
def _fetch_batch(self, revision_ids, basis_id, cache, a_graph=None):
3538
def _get_parent_keys(self, root_key, parent_map):
3539
"""Get the parent keys for a given root id."""
3540
root_id, rev_id = root_key
3541
# Include direct parents of the revision, but only if they used
3544
for parent_id in parent_map[rev_id]:
3545
if parent_id == _mod_revision.NULL_REVISION:
3547
if parent_id not in self._revision_id_to_root_id:
3548
# We probably didn't read this revision, go spend the
3549
# extra effort to actually check
3551
tree = self.source.revision_tree(parent_id)
3552
except errors.NoSuchRevision:
3553
# Ghost, fill out _revision_id_to_root_id in case we
3554
# encounter this again.
3555
# But set parent_root_id to None since we don't really know
3556
parent_root_id = None
3558
parent_root_id = tree.get_root_id()
3559
self._revision_id_to_root_id[parent_id] = None
3561
parent_root_id = self._revision_id_to_root_id[parent_id]
3562
if root_id == parent_root_id or parent_root_id is None:
3563
parent_keys.append((root_id, parent_id))
3564
return tuple(parent_keys)
3566
def _new_root_data_stream(self, root_keys_to_create, parent_map):
3567
for root_key in root_keys_to_create:
3568
parent_keys = self._get_parent_keys(root_key, parent_map)
3569
yield versionedfile.FulltextContentFactory(root_key,
3570
parent_keys, None, '')
3572
def _fetch_batch(self, revision_ids, basis_id, cache):
3852
3573
"""Fetch across a few revisions.
3854
3575
:param revision_ids: The revisions to copy
3855
3576
:param basis_id: The revision_id of a tree that must be in cache, used
3856
3577
as a basis for delta when no other base is available
3857
3578
:param cache: A cache of RevisionTrees that we can use.
3858
:param a_graph: A Graph object to determine the heads() of the
3859
rich-root data stream.
3860
3579
:return: The revision_id of the last converted tree. The RevisionTree
3861
3580
for it will be in cache
3868
3587
pending_deltas = []
3869
3588
pending_revisions = []
3870
3589
parent_map = self.source.get_parent_map(revision_ids)
3871
self._fetch_parent_invs_for_stacking(parent_map, cache)
3872
self.source._safe_to_return_from_cache = True
3873
3590
for tree in self.source.revision_trees(revision_ids):
3874
# Find a inventory delta for this revision.
3875
# Find text entries that need to be copied, too.
3876
3591
current_revision_id = tree.get_revision_id()
3877
3592
parent_ids = parent_map.get(current_revision_id, ())
3878
parent_trees = self._get_trees(parent_ids, cache)
3879
possible_trees = list(parent_trees)
3880
if len(possible_trees) == 0:
3881
# There either aren't any parents, or the parents are ghosts,
3882
# so just use the last converted tree.
3883
possible_trees.append((basis_id, cache[basis_id]))
3884
3593
basis_id, delta = self._get_delta_for_revision(tree, parent_ids,
3886
revision = self.source.get_revision(current_revision_id)
3887
pending_deltas.append((basis_id, delta,
3888
current_revision_id, revision.parent_ids))
3889
3595
if self._converting_to_rich_root:
3890
3596
self._revision_id_to_root_id[current_revision_id] = \
3891
3597
tree.get_root_id()
3892
# Determine which texts are in present in this revision but not in
3893
# any of the available parents.
3894
texts_possibly_new_in_tree = set()
3598
# Find text entries that need to be copied
3895
3599
for old_path, new_path, file_id, entry in delta:
3896
if new_path is None:
3897
# This file_id isn't present in the new rev
3901
if not self.target.supports_rich_root():
3902
# The target doesn't support rich root, so we don't
3905
if self._converting_to_rich_root:
3906
# This can't be copied normally, we have to insert
3908
root_keys_to_create.add((file_id, entry.revision))
3911
texts_possibly_new_in_tree.add((file_id, entry.revision))
3912
for basis_id, basis_tree in possible_trees:
3913
basis_inv = basis_tree.inventory
3914
for file_key in list(texts_possibly_new_in_tree):
3915
file_id, file_revision = file_key
3917
entry = basis_inv[file_id]
3918
except errors.NoSuchId:
3920
if entry.revision == file_revision:
3921
texts_possibly_new_in_tree.remove(file_key)
3922
text_keys.update(texts_possibly_new_in_tree)
3600
if new_path is not None:
3603
if not self.target.supports_rich_root():
3604
# The target doesn't support rich root, so we don't
3607
if self._converting_to_rich_root:
3608
# This can't be copied normally, we have to insert
3610
root_keys_to_create.add((file_id, entry.revision))
3612
text_keys.add((file_id, entry.revision))
3613
revision = self.source.get_revision(current_revision_id)
3614
pending_deltas.append((basis_id, delta,
3615
current_revision_id, revision.parent_ids))
3923
3616
pending_revisions.append(revision)
3924
3617
cache[current_revision_id] = tree
3925
3618
basis_id = current_revision_id
3926
self.source._safe_to_return_from_cache = False
3927
3619
# Copy file texts
3928
3620
from_texts = self.source.texts
3929
3621
to_texts = self.target.texts
3930
3622
if root_keys_to_create:
3931
root_stream = _mod_fetch._new_root_data_stream(
3932
root_keys_to_create, self._revision_id_to_root_id, parent_map,
3933
self.source, graph=a_graph)
3623
root_stream = self._new_root_data_stream(root_keys_to_create,
3934
3625
to_texts.insert_record_stream(root_stream)
3935
3626
to_texts.insert_record_stream(from_texts.get_record_stream(
3936
3627
text_keys, self.target._format._fetch_order,
4385
4020
# missing keys can handle suspending a write group).
4386
4021
write_group_tokens = self.target_repo.suspend_write_group()
4387
4022
return write_group_tokens, missing_keys
4388
hint = self.target_repo.commit_write_group()
4389
if (to_serializer != src_serializer and
4390
self.target_repo._format.pack_compresses):
4391
self.target_repo.pack(hint=hint)
4023
self.target_repo.commit_write_group()
4392
4024
return [], set()
4394
def _extract_and_insert_inventory_deltas(self, substream, serializer):
4395
target_rich_root = self.target_repo._format.rich_root_data
4396
target_tree_refs = self.target_repo._format.supports_tree_reference
4397
for record in substream:
4398
# Insert the delta directly
4399
inventory_delta_bytes = record.get_bytes_as('fulltext')
4400
deserialiser = inventory_delta.InventoryDeltaDeserializer()
4402
parse_result = deserialiser.parse_text_bytes(
4403
inventory_delta_bytes)
4404
except inventory_delta.IncompatibleInventoryDelta, err:
4405
trace.mutter("Incompatible delta: %s", err.msg)
4406
raise errors.IncompatibleRevision(self.target_repo._format)
4407
basis_id, new_id, rich_root, tree_refs, inv_delta = parse_result
4408
revision_id = new_id
4409
parents = [key[0] for key in record.parents]
4410
self.target_repo.add_inventory_by_delta(
4411
basis_id, inv_delta, revision_id, parents)
4413
def _extract_and_insert_inventories(self, substream, serializer,
4026
def _extract_and_insert_inventories(self, substream, serializer):
4415
4027
"""Generate a new inventory versionedfile in target, converting data.
4417
4029
The inventory is retrieved from the source, (deserializing it), and
4418
4030
stored in the target (reserializing it in a different format).
4420
target_rich_root = self.target_repo._format.rich_root_data
4421
target_tree_refs = self.target_repo._format.supports_tree_reference
4422
4032
for record in substream:
4423
# It's not a delta, so it must be a fulltext in the source
4424
# serializer's format.
4425
4033
bytes = record.get_bytes_as('fulltext')
4426
4034
revision_id = record.key[0]
4427
4035
inv = serializer.read_inventory_from_string(bytes, revision_id)
4428
4036
parents = [key[0] for key in record.parents]
4429
4037
self.target_repo.add_inventory(revision_id, inv, parents)
4430
# No need to keep holding this full inv in memory when the rest of
4431
# the substream is likely to be all deltas.
4434
4039
def _extract_and_insert_revisions(self, substream, serializer):
4435
4040
for record in substream:
4599
4186
return (not self.from_repository._format.rich_root_data and
4600
4187
self.to_format.rich_root_data)
4602
def _get_inventory_stream(self, revision_ids, missing=False):
4189
def _get_inventory_stream(self, revision_ids):
4603
4190
from_format = self.from_repository._format
4604
if (from_format.supports_chks and self.to_format.supports_chks and
4605
from_format.network_name() == self.to_format.network_name()):
4606
raise AssertionError(
4607
"this case should be handled by GroupCHKStreamSource")
4608
elif 'forceinvdeltas' in debug.debug_flags:
4609
return self._get_convertable_inventory_stream(revision_ids,
4610
delta_versus_null=missing)
4611
elif from_format.network_name() == self.to_format.network_name():
4613
return self._get_simple_inventory_stream(revision_ids,
4615
elif (not from_format.supports_chks and not self.to_format.supports_chks
4616
and from_format._serializer == self.to_format._serializer):
4617
# Essentially the same format.
4618
return self._get_simple_inventory_stream(revision_ids,
4191
if (from_format.supports_chks and self.to_format.supports_chks
4192
and (from_format._serializer == self.to_format._serializer)):
4193
# Both sides support chks, and they use the same serializer, so it
4194
# is safe to transmit the chk pages and inventory pages across
4196
return self._get_chk_inventory_stream(revision_ids)
4197
elif (not from_format.supports_chks):
4198
# Source repository doesn't support chks. So we can transmit the
4199
# inventories 'as-is' and either they are just accepted on the
4200
# target, or the Sink will properly convert it.
4201
return self._get_simple_inventory_stream(revision_ids)
4621
# Any time we switch serializations, we want to use an
4622
# inventory-delta based approach.
4623
return self._get_convertable_inventory_stream(revision_ids,
4624
delta_versus_null=missing)
4203
# XXX: Hack to make not-chk->chk fetch: copy the inventories as
4204
# inventories. Note that this should probably be done somehow
4205
# as part of bzrlib.repository.StreamSink. Except JAM couldn't
4206
# figure out how a non-chk repository could possibly handle
4207
# deserializing an inventory stream from a chk repo, as it
4208
# doesn't have a way to understand individual pages.
4209
return self._get_convertable_inventory_stream(revision_ids)
4626
def _get_simple_inventory_stream(self, revision_ids, missing=False):
4627
# NB: This currently reopens the inventory weave in source;
4628
# using a single stream interface instead would avoid this.
4211
def _get_simple_inventory_stream(self, revision_ids):
4629
4212
from_weave = self.from_repository.inventories
4631
delta_closure = True
4633
delta_closure = not self.delta_on_metadata()
4634
4213
yield ('inventories', from_weave.get_record_stream(
4635
4214
[(rev_id,) for rev_id in revision_ids],
4636
self.inventory_fetch_order(), delta_closure))
4638
def _get_convertable_inventory_stream(self, revision_ids,
4639
delta_versus_null=False):
4640
# The two formats are sufficiently different that there is no fast
4641
# path, so we need to send just inventorydeltas, which any
4642
# sufficiently modern client can insert into any repository.
4643
# The StreamSink code expects to be able to
4644
# convert on the target, so we need to put bytes-on-the-wire that can
4645
# be converted. That means inventory deltas (if the remote is <1.19,
4646
# RemoteStreamSink will fallback to VFS to insert the deltas).
4647
yield ('inventory-deltas',
4648
self._stream_invs_as_deltas(revision_ids,
4649
delta_versus_null=delta_versus_null))
4651
def _stream_invs_as_deltas(self, revision_ids, delta_versus_null=False):
4652
"""Return a stream of inventory-deltas for the given rev ids.
4654
:param revision_ids: The list of inventories to transmit
4655
:param delta_versus_null: Don't try to find a minimal delta for this
4656
entry, instead compute the delta versus the NULL_REVISION. This
4657
effectively streams a complete inventory. Used for stuff like
4658
filling in missing parents, etc.
4215
self.inventory_fetch_order(),
4216
not self.delta_on_metadata()))
4218
def _get_chk_inventory_stream(self, revision_ids):
4219
"""Fetch the inventory texts, along with the associated chk maps."""
4220
# We want an inventory outside of the search set, so that we can filter
4221
# out uninteresting chk pages. For now we use
4222
# _find_revision_outside_set, but if we had a Search with cut_revs, we
4223
# could use that instead.
4224
start_rev_id = self.from_repository._find_revision_outside_set(
4226
start_rev_key = (start_rev_id,)
4227
inv_keys_to_fetch = [(rev_id,) for rev_id in revision_ids]
4228
if start_rev_id != _mod_revision.NULL_REVISION:
4229
inv_keys_to_fetch.append((start_rev_id,))
4230
# Any repo that supports chk_bytes must also support out-of-order
4231
# insertion. At least, that is how we expect it to work
4232
# We use get_record_stream instead of iter_inventories because we want
4233
# to be able to insert the stream as well. We could instead fetch
4234
# allowing deltas, and then iter_inventories, but we don't know whether
4235
# source or target is more 'local' anway.
4236
inv_stream = self.from_repository.inventories.get_record_stream(
4237
inv_keys_to_fetch, 'unordered',
4238
True) # We need them as full-texts so we can find their references
4239
uninteresting_chk_roots = set()
4240
interesting_chk_roots = set()
4241
def filter_inv_stream(inv_stream):
4242
for idx, record in enumerate(inv_stream):
4243
### child_pb.update('fetch inv', idx, len(inv_keys_to_fetch))
4244
bytes = record.get_bytes_as('fulltext')
4245
chk_inv = inventory.CHKInventory.deserialise(
4246
self.from_repository.chk_bytes, bytes, record.key)
4247
if record.key == start_rev_key:
4248
uninteresting_chk_roots.add(chk_inv.id_to_entry.key())
4249
p_id_map = chk_inv.parent_id_basename_to_file_id
4250
if p_id_map is not None:
4251
uninteresting_chk_roots.add(p_id_map.key())
4254
interesting_chk_roots.add(chk_inv.id_to_entry.key())
4255
p_id_map = chk_inv.parent_id_basename_to_file_id
4256
if p_id_map is not None:
4257
interesting_chk_roots.add(p_id_map.key())
4258
### pb.update('fetch inventory', 0, 2)
4259
yield ('inventories', filter_inv_stream(inv_stream))
4260
# Now that we have worked out all of the interesting root nodes, grab
4261
# all of the interesting pages and insert them
4262
### pb.update('fetch inventory', 1, 2)
4263
interesting = chk_map.iter_interesting_nodes(
4264
self.from_repository.chk_bytes, interesting_chk_roots,
4265
uninteresting_chk_roots)
4266
def to_stream_adapter():
4267
"""Adapt the iter_interesting_nodes result to a single stream.
4269
iter_interesting_nodes returns records as it processes them, along
4270
with keys. However, we only want to return the records themselves.
4272
for record, items in interesting:
4273
if record is not None:
4275
# XXX: We could instead call get_record_stream(records.keys())
4276
# ATM, this will always insert the records as fulltexts, and
4277
# requires that you can hang on to records once you have gone
4278
# on to the next one. Further, it causes the target to
4279
# recompress the data. Testing shows it to be faster than
4280
# requesting the records again, though.
4281
yield ('chk_bytes', to_stream_adapter())
4282
### pb.update('fetch inventory', 2, 2)
4284
def _get_convertable_inventory_stream(self, revision_ids):
4285
# XXX: One of source or target is using chks, and they don't have
4286
# compatible serializations. The StreamSink code expects to be
4287
# able to convert on the target, so we need to put
4288
# bytes-on-the-wire that can be converted
4289
yield ('inventories', self._stream_invs_as_fulltexts(revision_ids))
4291
def _stream_invs_as_fulltexts(self, revision_ids):
4660
4292
from_repo = self.from_repository
4293
from_serializer = from_repo._format._serializer
4661
4294
revision_keys = [(rev_id,) for rev_id in revision_ids]
4662
4295
parent_map = from_repo.inventories.get_parent_map(revision_keys)
4663
# XXX: possibly repos could implement a more efficient iter_inv_deltas
4665
inventories = self.from_repository.iter_inventories(
4666
revision_ids, 'topological')
4667
format = from_repo._format
4668
invs_sent_so_far = set([_mod_revision.NULL_REVISION])
4669
inventory_cache = lru_cache.LRUCache(50)
4670
null_inventory = from_repo.revision_tree(
4671
_mod_revision.NULL_REVISION).inventory
4672
# XXX: ideally the rich-root/tree-refs flags would be per-revision, not
4673
# per-repo (e.g. streaming a non-rich-root revision out of a rich-root
4674
# repo back into a non-rich-root repo ought to be allowed)
4675
serializer = inventory_delta.InventoryDeltaSerializer(
4676
versioned_root=format.rich_root_data,
4677
tree_references=format.supports_tree_reference)
4678
for inv in inventories:
4296
for inv in self.from_repository.iter_inventories(revision_ids):
4297
# XXX: This is a bit hackish, but it works. Basically,
4298
# CHKSerializer 'accidentally' supports
4299
# read/write_inventory_to_string, even though that is never
4300
# the format that is stored on disk. It *does* give us a
4301
# single string representation for an inventory, so live with
4303
# This would be far better if we had a 'serialized inventory
4304
# delta' form. Then we could use 'inventory._make_delta', and
4305
# transmit that. This would both be faster to generate, and
4306
# result in fewer bytes-on-the-wire.
4307
as_bytes = from_serializer.write_inventory_to_string(inv)
4679
4308
key = (inv.revision_id,)
4680
4309
parent_keys = parent_map.get(key, ())
4682
if not delta_versus_null and parent_keys:
4683
# The caller did not ask for complete inventories and we have
4684
# some parents that we can delta against. Make a delta against
4685
# each parent so that we can find the smallest.
4686
parent_ids = [parent_key[0] for parent_key in parent_keys]
4687
for parent_id in parent_ids:
4688
if parent_id not in invs_sent_so_far:
4689
# We don't know that the remote side has this basis, so
4692
if parent_id == _mod_revision.NULL_REVISION:
4693
parent_inv = null_inventory
4695
parent_inv = inventory_cache.get(parent_id, None)
4696
if parent_inv is None:
4697
parent_inv = from_repo.get_inventory(parent_id)
4698
candidate_delta = inv._make_delta(parent_inv)
4699
if (delta is None or
4700
len(delta) > len(candidate_delta)):
4701
delta = candidate_delta
4702
basis_id = parent_id
4704
# Either none of the parents ended up being suitable, or we
4705
# were asked to delta against NULL
4706
basis_id = _mod_revision.NULL_REVISION
4707
delta = inv._make_delta(null_inventory)
4708
invs_sent_so_far.add(inv.revision_id)
4709
inventory_cache[inv.revision_id] = inv
4710
delta_serialized = ''.join(
4711
serializer.delta_to_lines(basis_id, key[-1], delta))
4712
4310
yield versionedfile.FulltextContentFactory(
4713
key, parent_keys, None, delta_serialized)
4716
def _iter_for_revno(repo, partial_history_cache, stop_index=None,
4717
stop_revision=None):
4718
"""Extend the partial history to include a given index
4720
If a stop_index is supplied, stop when that index has been reached.
4721
If a stop_revision is supplied, stop when that revision is
4722
encountered. Otherwise, stop when the beginning of history is
4725
:param stop_index: The index which should be present. When it is
4726
present, history extension will stop.
4727
:param stop_revision: The revision id which should be present. When
4728
it is encountered, history extension will stop.
4730
start_revision = partial_history_cache[-1]
4731
iterator = repo.iter_reverse_revision_history(start_revision)
4733
#skip the last revision in the list
4736
if (stop_index is not None and
4737
len(partial_history_cache) > stop_index):
4739
if partial_history_cache[-1] == stop_revision:
4741
revision_id = iterator.next()
4742
partial_history_cache.append(revision_id)
4743
except StopIteration:
4311
key, parent_keys, None, as_bytes)
added
removed
bzrlib/repository.py
