1160
1150
# The old API returned a list, should this actually be a set?
1161
1151
return parent_map.keys()
1163
def _check_inventories(self, checker):
1164
"""Check the inventories found from the revision scan.
1166
This is responsible for verifying the sha1 of inventories and
1167
creating a pending_keys set that covers data referenced by inventories.
1169
bar = ui.ui_factory.nested_progress_bar()
1171
self._do_check_inventories(checker, bar)
1175
def _do_check_inventories(self, checker, bar):
1176
"""Helper for _check_inventories."""
1178
keys = {'chk_bytes':set(), 'inventories':set(), 'texts':set()}
1179
kinds = ['chk_bytes', 'texts']
1180
count = len(checker.pending_keys)
1181
bar.update("inventories", 0, 2)
1182
current_keys = checker.pending_keys
1183
checker.pending_keys = {}
1184
# Accumulate current checks.
1185
for key in current_keys:
1186
if key[0] != 'inventories' and key[0] not in kinds:
1187
checker._report_items.append('unknown key type %r' % (key,))
1188
keys[key[0]].add(key[1:])
1189
if keys['inventories']:
1190
# NB: output order *should* be roughly sorted - topo or
1191
# inverse topo depending on repository - either way decent
1192
# to just delta against. However, pre-CHK formats didn't
1193
# try to optimise inventory layout on disk. As such the
1194
# pre-CHK code path does not use inventory deltas.
1196
for record in self.inventories.check(keys=keys['inventories']):
1197
if record.storage_kind == 'absent':
1198
checker._report_items.append(
1199
'Missing inventory {%s}' % (record.key,))
1201
last_object = self._check_record('inventories', record,
1202
checker, last_object,
1203
current_keys[('inventories',) + record.key])
1204
del keys['inventories']
1207
bar.update("texts", 1)
1208
while (checker.pending_keys or keys['chk_bytes']
1210
# Something to check.
1211
current_keys = checker.pending_keys
1212
checker.pending_keys = {}
1213
# Accumulate current checks.
1214
for key in current_keys:
1215
if key[0] not in kinds:
1216
checker._report_items.append('unknown key type %r' % (key,))
1217
keys[key[0]].add(key[1:])
1218
# Check the outermost kind only - inventories || chk_bytes || texts
1222
for record in getattr(self, kind).check(keys=keys[kind]):
1223
if record.storage_kind == 'absent':
1224
checker._report_items.append(
1225
'Missing %s {%s}' % (kind, record.key,))
1227
last_object = self._check_record(kind, record,
1228
checker, last_object, current_keys[(kind,) + record.key])
1232
def _check_record(self, kind, record, checker, last_object, item_data):
1233
"""Check a single text from this repository."""
1234
if kind == 'inventories':
1235
rev_id = record.key[0]
1236
inv = self.deserialise_inventory(rev_id,
1237
record.get_bytes_as('fulltext'))
1238
if last_object is not None:
1239
delta = inv._make_delta(last_object)
1240
for old_path, path, file_id, ie in delta:
1243
ie.check(checker, rev_id, inv)
1245
for path, ie in inv.iter_entries():
1246
ie.check(checker, rev_id, inv)
1247
if self._format.fast_deltas:
1249
elif kind == 'chk_bytes':
1250
# No code written to check chk_bytes for this repo format.
1251
checker._report_items.append(
1252
'unsupported key type chk_bytes for %s' % (record.key,))
1253
elif kind == 'texts':
1254
self._check_text(record, checker, item_data)
1256
checker._report_items.append(
1257
'unknown key type %s for %s' % (kind, record.key))
1259
def _check_text(self, record, checker, item_data):
1260
"""Check a single text."""
1261
# Check it is extractable.
1262
# TODO: check length.
1263
if record.storage_kind == 'chunked':
1264
chunks = record.get_bytes_as(record.storage_kind)
1265
sha1 = osutils.sha_strings(chunks)
1266
length = sum(map(len, chunks))
1268
content = record.get_bytes_as('fulltext')
1269
sha1 = osutils.sha_string(content)
1270
length = len(content)
1271
if item_data and sha1 != item_data[1]:
1272
checker._report_items.append(
1273
'sha1 mismatch: %s has sha1 %s expected %s referenced by %s' %
1274
(record.key, sha1, item_data[1], item_data[2]))
1277
1154
def create(a_bzrdir):
1278
1155
"""Construct the current default format repository in a_bzrdir."""
1848
1670
@needs_read_lock
1849
1671
def get_revisions(self, revision_ids):
1850
"""Get many revisions at once.
1852
Repositories that need to check data on every revision read should
1853
subclass this method.
1672
"""Get many revisions at once."""
1855
1673
return self._get_revisions(revision_ids)
1857
1675
@needs_read_lock
1858
1676
def _get_revisions(self, revision_ids):
1859
1677
"""Core work logic to get many revisions without sanity checks."""
1678
for rev_id in revision_ids:
1679
if not rev_id or not isinstance(rev_id, basestring):
1680
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1681
keys = [(key,) for key in revision_ids]
1682
stream = self.revisions.get_record_stream(keys, 'unordered', True)
1861
for revid, rev in self._iter_revisions(revision_ids):
1863
raise errors.NoSuchRevision(self, revid)
1684
for record in stream:
1685
if record.storage_kind == 'absent':
1686
raise errors.NoSuchRevision(self, record.key[0])
1687
text = record.get_bytes_as('fulltext')
1688
rev = self._serializer.read_revision_from_string(text)
1689
revs[record.key[0]] = rev
1865
1690
return [revs[revid] for revid in revision_ids]
1867
def _iter_revisions(self, revision_ids):
1868
"""Iterate over revision objects.
1870
:param revision_ids: An iterable of revisions to examine. None may be
1871
passed to request all revisions known to the repository. Note that
1872
not all repositories can find unreferenced revisions; for those
1873
repositories only referenced ones will be returned.
1874
:return: An iterator of (revid, revision) tuples. Absent revisions (
1875
those asked for but not available) are returned as (revid, None).
1877
if revision_ids is None:
1878
revision_ids = self.all_revision_ids()
1880
for rev_id in revision_ids:
1881
if not rev_id or not isinstance(rev_id, basestring):
1882
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1883
keys = [(key,) for key in revision_ids]
1884
stream = self.revisions.get_record_stream(keys, 'unordered', True)
1885
for record in stream:
1886
revid = record.key[0]
1887
if record.storage_kind == 'absent':
1890
text = record.get_bytes_as('fulltext')
1891
rev = self._serializer.read_revision_from_string(text)
1894
1692
@needs_read_lock
1895
1693
def get_revision_xml(self, revision_id):
1896
1694
# TODO: jam 20070210 This shouldn't be necessary since get_revision
2366
2142
:param revision_ids: The expected revision ids of the inventories.
2367
:param ordering: optional ordering, e.g. 'topological'. If not
2368
specified, the order of revision_ids will be preserved (by
2369
buffering if necessary).
2370
2143
:return: An iterator of inventories.
2372
2145
if ((None in revision_ids)
2373
2146
or (_mod_revision.NULL_REVISION in revision_ids)):
2374
2147
raise ValueError('cannot get null revision inventory')
2375
return self._iter_inventories(revision_ids, ordering)
2148
return self._iter_inventories(revision_ids)
2377
def _iter_inventories(self, revision_ids, ordering):
2150
def _iter_inventories(self, revision_ids):
2378
2151
"""single-document based inventory iteration."""
2379
inv_xmls = self._iter_inventory_xmls(revision_ids, ordering)
2380
for text, revision_id in inv_xmls:
2152
for text, revision_id in self._iter_inventory_xmls(revision_ids):
2381
2153
yield self.deserialise_inventory(revision_id, text)
2383
def _iter_inventory_xmls(self, revision_ids, ordering):
2384
if ordering is None:
2385
order_as_requested = True
2386
ordering = 'unordered'
2388
order_as_requested = False
2155
def _iter_inventory_xmls(self, revision_ids):
2389
2156
keys = [(revision_id,) for revision_id in revision_ids]
2392
if order_as_requested:
2393
key_iter = iter(keys)
2394
next_key = key_iter.next()
2395
stream = self.inventories.get_record_stream(keys, ordering, True)
2157
stream = self.inventories.get_record_stream(keys, 'unordered', True)
2396
2158
text_chunks = {}
2397
2159
for record in stream:
2398
2160
if record.storage_kind != 'absent':
2399
chunks = record.get_bytes_as('chunked')
2400
if order_as_requested:
2401
text_chunks[record.key] = chunks
2403
yield ''.join(chunks), record.key[-1]
2161
text_chunks[record.key] = record.get_bytes_as('chunked')
2405
2163
raise errors.NoSuchRevision(self, record.key)
2406
if order_as_requested:
2407
# Yield as many results as we can while preserving order.
2408
while next_key in text_chunks:
2409
chunks = text_chunks.pop(next_key)
2410
yield ''.join(chunks), next_key[-1]
2412
next_key = key_iter.next()
2413
except StopIteration:
2414
# We still want to fully consume the get_record_stream,
2415
# just in case it is not actually finished at this point
2165
chunks = text_chunks.pop(key)
2166
yield ''.join(chunks), key[-1]
2419
2168
def deserialise_inventory(self, revision_id, xml):
2420
2169
"""Transform the xml into an inventory object.
3705
3402
return self.source.revision_ids_to_search_result(result_set)
3405
class InterPackRepo(InterSameDataRepository):
3406
"""Optimised code paths between Pack based repositories."""
3409
def _get_repo_format_to_test(self):
3410
from bzrlib.repofmt import pack_repo
3411
return pack_repo.RepositoryFormatKnitPack6RichRoot()
3414
def is_compatible(source, target):
3415
"""Be compatible with known Pack formats.
3417
We don't test for the stores being of specific types because that
3418
could lead to confusing results, and there is no need to be
3421
InterPackRepo does not support CHK based repositories.
3423
from bzrlib.repofmt.pack_repo import RepositoryFormatPack
3424
from bzrlib.repofmt.groupcompress_repo import RepositoryFormatCHK1
3426
are_packs = (isinstance(source._format, RepositoryFormatPack) and
3427
isinstance(target._format, RepositoryFormatPack))
3428
not_packs = (isinstance(source._format, RepositoryFormatCHK1) or
3429
isinstance(target._format, RepositoryFormatCHK1))
3430
except AttributeError:
3432
if not_packs or not are_packs:
3434
return InterRepository._same_model(source, target)
3437
def fetch(self, revision_id=None, pb=None, find_ghosts=False,
3439
"""See InterRepository.fetch()."""
3440
if (len(self.source._fallback_repositories) > 0 or
3441
len(self.target._fallback_repositories) > 0):
3442
# The pack layer is not aware of fallback repositories, so when
3443
# fetching from a stacked repository or into a stacked repository
3444
# we use the generic fetch logic which uses the VersionedFiles
3445
# attributes on repository.
3446
from bzrlib.fetch import RepoFetcher
3447
fetcher = RepoFetcher(self.target, self.source, revision_id,
3448
pb, find_ghosts, fetch_spec=fetch_spec)
3449
if fetch_spec is not None:
3450
if len(list(fetch_spec.heads)) != 1:
3451
raise AssertionError(
3452
"InterPackRepo.fetch doesn't support "
3453
"fetching multiple heads yet.")
3454
revision_id = list(fetch_spec.heads)[0]
3456
if revision_id is None:
3458
# everything to do - use pack logic
3459
# to fetch from all packs to one without
3460
# inventory parsing etc, IFF nothing to be copied is in the target.
3462
source_revision_ids = frozenset(self.source.all_revision_ids())
3463
revision_ids = source_revision_ids - \
3464
frozenset(self.target.get_parent_map(source_revision_ids))
3465
revision_keys = [(revid,) for revid in revision_ids]
3466
index = self.target._pack_collection.revision_index.combined_index
3467
present_revision_ids = set(item[1][0] for item in
3468
index.iter_entries(revision_keys))
3469
revision_ids = set(revision_ids) - present_revision_ids
3470
# implementing the TODO will involve:
3471
# - detecting when all of a pack is selected
3472
# - avoiding as much as possible pre-selection, so the
3473
# more-core routines such as create_pack_from_packs can filter in
3474
# a just-in-time fashion. (though having a HEADS list on a
3475
# repository might make this a lot easier, because we could
3476
# sensibly detect 'new revisions' without doing a full index scan.
3477
elif _mod_revision.is_null(revision_id):
3481
revision_ids = self.search_missing_revision_ids(revision_id,
3482
find_ghosts=find_ghosts).get_keys()
3483
if len(revision_ids) == 0:
3485
return self._pack(self.source, self.target, revision_ids)
3487
def _pack(self, source, target, revision_ids):
3488
from bzrlib.repofmt.pack_repo import Packer
3489
packs = source._pack_collection.all_packs()
3490
pack = Packer(self.target._pack_collection, packs, '.fetch',
3491
revision_ids).pack()
3492
if pack is not None:
3493
self.target._pack_collection._save_pack_names()
3494
copied_revs = pack.get_revision_count()
3495
# Trigger an autopack. This may duplicate effort as we've just done
3496
# a pack creation, but for now it is simpler to think about as
3497
# 'upload data, then repack if needed'.
3498
self.target._pack_collection.autopack()
3499
return (copied_revs, [])
3504
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
3505
"""See InterRepository.missing_revision_ids().
3507
:param find_ghosts: Find ghosts throughout the ancestry of
3510
if not find_ghosts and revision_id is not None:
3511
return self._walk_to_common_revisions([revision_id])
3512
elif revision_id is not None:
3513
# Find ghosts: search for revisions pointing from one repository to
3514
# the other, and vice versa, anywhere in the history of revision_id.
3515
graph = self.target.get_graph(other_repository=self.source)
3516
searcher = graph._make_breadth_first_searcher([revision_id])
3520
next_revs, ghosts = searcher.next_with_ghosts()
3521
except StopIteration:
3523
if revision_id in ghosts:
3524
raise errors.NoSuchRevision(self.source, revision_id)
3525
found_ids.update(next_revs)
3526
found_ids.update(ghosts)
3527
found_ids = frozenset(found_ids)
3528
# Double query here: should be able to avoid this by changing the
3529
# graph api further.
3530
result_set = found_ids - frozenset(
3531
self.target.get_parent_map(found_ids))
3533
source_ids = self.source.all_revision_ids()
3534
# source_ids is the worst possible case we may need to pull.
3535
# now we want to filter source_ids against what we actually
3536
# have in target, but don't try to check for existence where we know
3537
# we do not have a revision as that would be pointless.
3538
target_ids = set(self.target.all_revision_ids())
3539
result_set = set(source_ids).difference(target_ids)
3540
return self.source.revision_ids_to_search_result(result_set)
3708
3543
class InterDifferingSerializer(InterRepository):
3717
3552
# This is redundant with format.check_conversion_target(), however that
3718
3553
# raises an exception, and we just want to say "False" as in we won't
3719
3554
# support converting between these formats.
3720
if 'IDS_never' in debug.debug_flags:
3722
3555
if source.supports_rich_root() and not target.supports_rich_root():
3724
3557
if (source._format.supports_tree_reference
3725
3558
and not target._format.supports_tree_reference):
3727
if target._fallback_repositories and target._format.supports_chks:
3728
# IDS doesn't know how to copy CHKs for the parent inventories it
3729
# adds to stacked repos.
3731
if 'IDS_always' in debug.debug_flags:
3733
# Only use this code path for local source and target. IDS does far
3734
# too much IO (both bandwidth and roundtrips) over a network.
3735
if not source.bzrdir.transport.base.startswith('file:///'):
3737
if not target.bzrdir.transport.base.startswith('file:///'):
3741
def _get_trees(self, revision_ids, cache):
3743
for rev_id in revision_ids:
3745
possible_trees.append((rev_id, cache[rev_id]))
3747
# Not cached, but inventory might be present anyway.
3749
tree = self.source.revision_tree(rev_id)
3750
except errors.NoSuchRevision:
3751
# Nope, parent is ghost.
3754
cache[rev_id] = tree
3755
possible_trees.append((rev_id, tree))
3756
return possible_trees
3758
def _get_delta_for_revision(self, tree, parent_ids, possible_trees):
3562
def _get_delta_for_revision(self, tree, parent_ids, basis_id, cache):
3759
3563
"""Get the best delta and base for this revision.
3761
3565
:return: (basis_id, delta)
3567
possible_trees = [(parent_id, cache[parent_id])
3568
for parent_id in parent_ids
3569
if parent_id in cache]
3570
if len(possible_trees) == 0:
3571
# There either aren't any parents, or the parents aren't in the
3572
# cache, so just use the last converted tree
3573
possible_trees.append((basis_id, cache[basis_id]))
3764
# Generate deltas against each tree, to find the shortest.
3765
texts_possibly_new_in_tree = set()
3766
3575
for basis_id, basis_tree in possible_trees:
3767
3576
delta = tree.inventory._make_delta(basis_tree.inventory)
3768
for old_path, new_path, file_id, new_entry in delta:
3769
if new_path is None:
3770
# This file_id isn't present in the new rev, so we don't
3774
# Rich roots are handled elsewhere...
3776
kind = new_entry.kind
3777
if kind != 'directory' and kind != 'file':
3778
# No text record associated with this inventory entry.
3780
# This is a directory or file that has changed somehow.
3781
texts_possibly_new_in_tree.add((file_id, new_entry.revision))
3782
3577
deltas.append((len(delta), basis_id, delta))
3784
3579
return deltas[0][1:]
3786
def _fetch_parent_invs_for_stacking(self, parent_map, cache):
3787
"""Find all parent revisions that are absent, but for which the
3788
inventory is present, and copy those inventories.
3581
def _get_parent_keys(self, root_key, parent_map):
3582
"""Get the parent keys for a given root id."""
3583
root_id, rev_id = root_key
3584
# Include direct parents of the revision, but only if they used
3587
for parent_id in parent_map[rev_id]:
3588
if parent_id == _mod_revision.NULL_REVISION:
3590
if parent_id not in self._revision_id_to_root_id:
3591
# We probably didn't read this revision, go spend the
3592
# extra effort to actually check
3594
tree = self.source.revision_tree(parent_id)
3595
except errors.NoSuchRevision:
3596
# Ghost, fill out _revision_id_to_root_id in case we
3597
# encounter this again.
3598
# But set parent_root_id to None since we don't really know
3599
parent_root_id = None
3601
parent_root_id = tree.get_root_id()
3602
self._revision_id_to_root_id[parent_id] = None
3604
parent_root_id = self._revision_id_to_root_id[parent_id]
3605
if root_id == parent_root_id or parent_root_id is None:
3606
parent_keys.append((root_id, parent_id))
3607
return tuple(parent_keys)
3790
This is necessary to preserve correctness when the source is stacked
3791
without fallbacks configured. (Note that in cases like upgrade the
3792
source may be not have _fallback_repositories even though it is
3796
for parents in parent_map.values():
3797
parent_revs.update(parents)
3798
present_parents = self.source.get_parent_map(parent_revs)
3799
absent_parents = set(parent_revs).difference(present_parents)
3800
parent_invs_keys_for_stacking = self.source.inventories.get_parent_map(
3801
(rev_id,) for rev_id in absent_parents)
3802
parent_inv_ids = [key[-1] for key in parent_invs_keys_for_stacking]
3803
for parent_tree in self.source.revision_trees(parent_inv_ids):
3804
current_revision_id = parent_tree.get_revision_id()
3805
parents_parents_keys = parent_invs_keys_for_stacking[
3806
(current_revision_id,)]
3807
parents_parents = [key[-1] for key in parents_parents_keys]
3808
basis_id = _mod_revision.NULL_REVISION
3809
basis_tree = self.source.revision_tree(basis_id)
3810
delta = parent_tree.inventory._make_delta(basis_tree.inventory)
3811
self.target.add_inventory_by_delta(
3812
basis_id, delta, current_revision_id, parents_parents)
3813
cache[current_revision_id] = parent_tree
3609
def _new_root_data_stream(self, root_keys_to_create, parent_map):
3610
for root_key in root_keys_to_create:
3611
parent_keys = self._get_parent_keys(root_key, parent_map)
3612
yield versionedfile.FulltextContentFactory(root_key,
3613
parent_keys, None, '')
3815
3615
def _fetch_batch(self, revision_ids, basis_id, cache):
3816
3616
"""Fetch across a few revisions.
3830
3630
pending_deltas = []
3831
3631
pending_revisions = []
3832
3632
parent_map = self.source.get_parent_map(revision_ids)
3833
self._fetch_parent_invs_for_stacking(parent_map, cache)
3834
3633
for tree in self.source.revision_trees(revision_ids):
3835
# Find a inventory delta for this revision.
3836
# Find text entries that need to be copied, too.
3837
3634
current_revision_id = tree.get_revision_id()
3838
3635
parent_ids = parent_map.get(current_revision_id, ())
3839
parent_trees = self._get_trees(parent_ids, cache)
3840
possible_trees = list(parent_trees)
3841
if len(possible_trees) == 0:
3842
# There either aren't any parents, or the parents are ghosts,
3843
# so just use the last converted tree.
3844
possible_trees.append((basis_id, cache[basis_id]))
3845
3636
basis_id, delta = self._get_delta_for_revision(tree, parent_ids,
3847
revision = self.source.get_revision(current_revision_id)
3848
pending_deltas.append((basis_id, delta,
3849
current_revision_id, revision.parent_ids))
3850
3638
if self._converting_to_rich_root:
3851
3639
self._revision_id_to_root_id[current_revision_id] = \
3852
3640
tree.get_root_id()
3853
# Determine which texts are in present in this revision but not in
3854
# any of the available parents.
3855
texts_possibly_new_in_tree = set()
3641
# Find text entries that need to be copied
3856
3642
for old_path, new_path, file_id, entry in delta:
3857
if new_path is None:
3858
# This file_id isn't present in the new rev
3862
if not self.target.supports_rich_root():
3863
# The target doesn't support rich root, so we don't
3866
if self._converting_to_rich_root:
3867
# This can't be copied normally, we have to insert
3869
root_keys_to_create.add((file_id, entry.revision))
3872
texts_possibly_new_in_tree.add((file_id, entry.revision))
3873
for basis_id, basis_tree in possible_trees:
3874
basis_inv = basis_tree.inventory
3875
for file_key in list(texts_possibly_new_in_tree):
3876
file_id, file_revision = file_key
3878
entry = basis_inv[file_id]
3879
except errors.NoSuchId:
3881
if entry.revision == file_revision:
3882
texts_possibly_new_in_tree.remove(file_key)
3883
text_keys.update(texts_possibly_new_in_tree)
3643
if new_path is not None:
3646
if not self.target.supports_rich_root():
3647
# The target doesn't support rich root, so we don't
3650
if self._converting_to_rich_root:
3651
# This can't be copied normally, we have to insert
3653
root_keys_to_create.add((file_id, entry.revision))
3655
text_keys.add((file_id, entry.revision))
3656
revision = self.source.get_revision(current_revision_id)
3657
pending_deltas.append((basis_id, delta,
3658
current_revision_id, revision.parent_ids))
3884
3659
pending_revisions.append(revision)
3885
3660
cache[current_revision_id] = tree
3886
3661
basis_id = current_revision_id
4329
4064
# missing keys can handle suspending a write group).
4330
4065
write_group_tokens = self.target_repo.suspend_write_group()
4331
4066
return write_group_tokens, missing_keys
4332
hint = self.target_repo.commit_write_group()
4333
if (to_serializer != src_serializer and
4334
self.target_repo._format.pack_compresses):
4335
self.target_repo.pack(hint=hint)
4067
self.target_repo.commit_write_group()
4336
4068
return [], set()
4338
def _extract_and_insert_inventory_deltas(self, substream, serializer):
4339
target_rich_root = self.target_repo._format.rich_root_data
4340
target_tree_refs = self.target_repo._format.supports_tree_reference
4341
for record in substream:
4342
# Insert the delta directly
4343
inventory_delta_bytes = record.get_bytes_as('fulltext')
4344
deserialiser = inventory_delta.InventoryDeltaDeserializer()
4346
parse_result = deserialiser.parse_text_bytes(
4347
inventory_delta_bytes)
4348
except inventory_delta.IncompatibleInventoryDelta, err:
4349
trace.mutter("Incompatible delta: %s", err.msg)
4350
raise errors.IncompatibleRevision(self.target_repo._format)
4351
basis_id, new_id, rich_root, tree_refs, inv_delta = parse_result
4352
revision_id = new_id
4353
parents = [key[0] for key in record.parents]
4354
self.target_repo.add_inventory_by_delta(
4355
basis_id, inv_delta, revision_id, parents)
4357
def _extract_and_insert_inventories(self, substream, serializer,
4070
def _extract_and_insert_inventories(self, substream, serializer):
4359
4071
"""Generate a new inventory versionedfile in target, converting data.
4361
4073
The inventory is retrieved from the source, (deserializing it), and
4362
4074
stored in the target (reserializing it in a different format).
4364
target_rich_root = self.target_repo._format.rich_root_data
4365
target_tree_refs = self.target_repo._format.supports_tree_reference
4366
4076
for record in substream:
4367
# It's not a delta, so it must be a fulltext in the source
4368
# serializer's format.
4369
4077
bytes = record.get_bytes_as('fulltext')
4370
4078
revision_id = record.key[0]
4371
4079
inv = serializer.read_inventory_from_string(bytes, revision_id)
4372
4080
parents = [key[0] for key in record.parents]
4373
4081
self.target_repo.add_inventory(revision_id, inv, parents)
4374
# No need to keep holding this full inv in memory when the rest of
4375
# the substream is likely to be all deltas.
4378
4083
def _extract_and_insert_revisions(self, substream, serializer):
4379
4084
for record in substream:
4543
4230
return (not self.from_repository._format.rich_root_data and
4544
4231
self.to_format.rich_root_data)
4546
def _get_inventory_stream(self, revision_ids, missing=False):
4233
def _get_inventory_stream(self, revision_ids):
4547
4234
from_format = self.from_repository._format
4548
if (from_format.supports_chks and self.to_format.supports_chks and
4549
from_format.network_name() == self.to_format.network_name()):
4550
raise AssertionError(
4551
"this case should be handled by GroupCHKStreamSource")
4552
elif 'forceinvdeltas' in debug.debug_flags:
4553
return self._get_convertable_inventory_stream(revision_ids,
4554
delta_versus_null=missing)
4555
elif from_format.network_name() == self.to_format.network_name():
4557
return self._get_simple_inventory_stream(revision_ids,
4559
elif (not from_format.supports_chks and not self.to_format.supports_chks
4560
and from_format._serializer == self.to_format._serializer):
4561
# Essentially the same format.
4562
return self._get_simple_inventory_stream(revision_ids,
4235
if (from_format.supports_chks and self.to_format.supports_chks
4236
and (from_format._serializer == self.to_format._serializer)):
4237
# Both sides support chks, and they use the same serializer, so it
4238
# is safe to transmit the chk pages and inventory pages across
4240
return self._get_chk_inventory_stream(revision_ids)
4241
elif (not from_format.supports_chks):
4242
# Source repository doesn't support chks. So we can transmit the
4243
# inventories 'as-is' and either they are just accepted on the
4244
# target, or the Sink will properly convert it.
4245
return self._get_simple_inventory_stream(revision_ids)
4565
# Any time we switch serializations, we want to use an
4566
# inventory-delta based approach.
4567
return self._get_convertable_inventory_stream(revision_ids,
4568
delta_versus_null=missing)
4247
# XXX: Hack to make not-chk->chk fetch: copy the inventories as
4248
# inventories. Note that this should probably be done somehow
4249
# as part of bzrlib.repository.StreamSink. Except JAM couldn't
4250
# figure out how a non-chk repository could possibly handle
4251
# deserializing an inventory stream from a chk repo, as it
4252
# doesn't have a way to understand individual pages.
4253
return self._get_convertable_inventory_stream(revision_ids)
4570
def _get_simple_inventory_stream(self, revision_ids, missing=False):
4571
# NB: This currently reopens the inventory weave in source;
4572
# using a single stream interface instead would avoid this.
4255
def _get_simple_inventory_stream(self, revision_ids):
4573
4256
from_weave = self.from_repository.inventories
4575
delta_closure = True
4577
delta_closure = not self.delta_on_metadata()
4578
4257
yield ('inventories', from_weave.get_record_stream(
4579
4258
[(rev_id,) for rev_id in revision_ids],
4580
self.inventory_fetch_order(), delta_closure))
4582
def _get_convertable_inventory_stream(self, revision_ids,
4583
delta_versus_null=False):
4584
# The source is using CHKs, but the target either doesn't or it has a
4585
# different serializer. The StreamSink code expects to be able to
4586
# convert on the target, so we need to put bytes-on-the-wire that can
4587
# be converted. That means inventory deltas (if the remote is <1.19,
4588
# RemoteStreamSink will fallback to VFS to insert the deltas).
4589
yield ('inventory-deltas',
4590
self._stream_invs_as_deltas(revision_ids,
4591
delta_versus_null=delta_versus_null))
4593
def _stream_invs_as_deltas(self, revision_ids, delta_versus_null=False):
4594
"""Return a stream of inventory-deltas for the given rev ids.
4596
:param revision_ids: The list of inventories to transmit
4597
:param delta_versus_null: Don't try to find a minimal delta for this
4598
entry, instead compute the delta versus the NULL_REVISION. This
4599
effectively streams a complete inventory. Used for stuff like
4600
filling in missing parents, etc.
4259
self.inventory_fetch_order(),
4260
not self.delta_on_metadata()))
4262
def _get_chk_inventory_stream(self, revision_ids):
4263
"""Fetch the inventory texts, along with the associated chk maps."""
4264
# We want an inventory outside of the search set, so that we can filter
4265
# out uninteresting chk pages. For now we use
4266
# _find_revision_outside_set, but if we had a Search with cut_revs, we
4267
# could use that instead.
4268
start_rev_id = self.from_repository._find_revision_outside_set(
4270
start_rev_key = (start_rev_id,)
4271
inv_keys_to_fetch = [(rev_id,) for rev_id in revision_ids]
4272
if start_rev_id != _mod_revision.NULL_REVISION:
4273
inv_keys_to_fetch.append((start_rev_id,))
4274
# Any repo that supports chk_bytes must also support out-of-order
4275
# insertion. At least, that is how we expect it to work
4276
# We use get_record_stream instead of iter_inventories because we want
4277
# to be able to insert the stream as well. We could instead fetch
4278
# allowing deltas, and then iter_inventories, but we don't know whether
4279
# source or target is more 'local' anway.
4280
inv_stream = self.from_repository.inventories.get_record_stream(
4281
inv_keys_to_fetch, 'unordered',
4282
True) # We need them as full-texts so we can find their references
4283
uninteresting_chk_roots = set()
4284
interesting_chk_roots = set()
4285
def filter_inv_stream(inv_stream):
4286
for idx, record in enumerate(inv_stream):
4287
### child_pb.update('fetch inv', idx, len(inv_keys_to_fetch))
4288
bytes = record.get_bytes_as('fulltext')
4289
chk_inv = inventory.CHKInventory.deserialise(
4290
self.from_repository.chk_bytes, bytes, record.key)
4291
if record.key == start_rev_key:
4292
uninteresting_chk_roots.add(chk_inv.id_to_entry.key())
4293
p_id_map = chk_inv.parent_id_basename_to_file_id
4294
if p_id_map is not None:
4295
uninteresting_chk_roots.add(p_id_map.key())
4298
interesting_chk_roots.add(chk_inv.id_to_entry.key())
4299
p_id_map = chk_inv.parent_id_basename_to_file_id
4300
if p_id_map is not None:
4301
interesting_chk_roots.add(p_id_map.key())
4302
### pb.update('fetch inventory', 0, 2)
4303
yield ('inventories', filter_inv_stream(inv_stream))
4304
# Now that we have worked out all of the interesting root nodes, grab
4305
# all of the interesting pages and insert them
4306
### pb.update('fetch inventory', 1, 2)
4307
interesting = chk_map.iter_interesting_nodes(
4308
self.from_repository.chk_bytes, interesting_chk_roots,
4309
uninteresting_chk_roots)
4310
def to_stream_adapter():
4311
"""Adapt the iter_interesting_nodes result to a single stream.
4313
iter_interesting_nodes returns records as it processes them, along
4314
with keys. However, we only want to return the records themselves.
4316
for record, items in interesting:
4317
if record is not None:
4319
# XXX: We could instead call get_record_stream(records.keys())
4320
# ATM, this will always insert the records as fulltexts, and
4321
# requires that you can hang on to records once you have gone
4322
# on to the next one. Further, it causes the target to
4323
# recompress the data. Testing shows it to be faster than
4324
# requesting the records again, though.
4325
yield ('chk_bytes', to_stream_adapter())
4326
### pb.update('fetch inventory', 2, 2)
4328
def _get_convertable_inventory_stream(self, revision_ids):
4329
# XXX: One of source or target is using chks, and they don't have
4330
# compatible serializations. The StreamSink code expects to be
4331
# able to convert on the target, so we need to put
4332
# bytes-on-the-wire that can be converted
4333
yield ('inventories', self._stream_invs_as_fulltexts(revision_ids))
4335
def _stream_invs_as_fulltexts(self, revision_ids):
4602
4336
from_repo = self.from_repository
4337
from_serializer = from_repo._format._serializer
4603
4338
revision_keys = [(rev_id,) for rev_id in revision_ids]
4604
4339
parent_map = from_repo.inventories.get_parent_map(revision_keys)
4605
# XXX: possibly repos could implement a more efficient iter_inv_deltas
4607
inventories = self.from_repository.iter_inventories(
4608
revision_ids, 'topological')
4609
format = from_repo._format
4610
invs_sent_so_far = set([_mod_revision.NULL_REVISION])
4611
inventory_cache = lru_cache.LRUCache(50)
4612
null_inventory = from_repo.revision_tree(
4613
_mod_revision.NULL_REVISION).inventory
4614
# XXX: ideally the rich-root/tree-refs flags would be per-revision, not
4615
# per-repo (e.g. streaming a non-rich-root revision out of a rich-root
4616
# repo back into a non-rich-root repo ought to be allowed)
4617
serializer = inventory_delta.InventoryDeltaSerializer(
4618
versioned_root=format.rich_root_data,
4619
tree_references=format.supports_tree_reference)
4620
for inv in inventories:
4340
for inv in self.from_repository.iter_inventories(revision_ids):
4341
# XXX: This is a bit hackish, but it works. Basically,
4342
# CHKSerializer 'accidentally' supports
4343
# read/write_inventory_to_string, even though that is never
4344
# the format that is stored on disk. It *does* give us a
4345
# single string representation for an inventory, so live with
4347
# This would be far better if we had a 'serialized inventory
4348
# delta' form. Then we could use 'inventory._make_delta', and
4349
# transmit that. This would both be faster to generate, and
4350
# result in fewer bytes-on-the-wire.
4351
as_bytes = from_serializer.write_inventory_to_string(inv)
4621
4352
key = (inv.revision_id,)
4622
4353
parent_keys = parent_map.get(key, ())
4624
if not delta_versus_null and parent_keys:
4625
# The caller did not ask for complete inventories and we have
4626
# some parents that we can delta against. Make a delta against
4627
# each parent so that we can find the smallest.
4628
parent_ids = [parent_key[0] for parent_key in parent_keys]
4629
for parent_id in parent_ids:
4630
if parent_id not in invs_sent_so_far:
4631
# We don't know that the remote side has this basis, so
4634
if parent_id == _mod_revision.NULL_REVISION:
4635
parent_inv = null_inventory
4637
parent_inv = inventory_cache.get(parent_id, None)
4638
if parent_inv is None:
4639
parent_inv = from_repo.get_inventory(parent_id)
4640
candidate_delta = inv._make_delta(parent_inv)
4641
if (delta is None or
4642
len(delta) > len(candidate_delta)):
4643
delta = candidate_delta
4644
basis_id = parent_id
4646
# Either none of the parents ended up being suitable, or we
4647
# were asked to delta against NULL
4648
basis_id = _mod_revision.NULL_REVISION
4649
delta = inv._make_delta(null_inventory)
4650
invs_sent_so_far.add(inv.revision_id)
4651
inventory_cache[inv.revision_id] = inv
4652
delta_serialized = ''.join(
4653
serializer.delta_to_lines(basis_id, key[-1], delta))
4654
4354
yield versionedfile.FulltextContentFactory(
4655
key, parent_keys, None, delta_serialized)
4658
def _iter_for_revno(repo, partial_history_cache, stop_index=None,
4659
stop_revision=None):
4660
"""Extend the partial history to include a given index
4662
If a stop_index is supplied, stop when that index has been reached.
4663
If a stop_revision is supplied, stop when that revision is
4664
encountered. Otherwise, stop when the beginning of history is
4667
:param stop_index: The index which should be present. When it is
4668
present, history extension will stop.
4669
:param stop_revision: The revision id which should be present. When
4670
it is encountered, history extension will stop.
4672
start_revision = partial_history_cache[-1]
4673
iterator = repo.iter_reverse_revision_history(start_revision)
4675
#skip the last revision in the list
4678
if (stop_index is not None and
4679
len(partial_history_cache) > stop_index):
4681
if partial_history_cache[-1] == stop_revision:
4683
revision_id = iterator.next()
4684
partial_history_cache.append(revision_id)
4685
except StopIteration:
4355
key, parent_keys, None, as_bytes)
added
removed
bzrlib/repository.py
