1196
1154
# The old API returned a list, should this actually be a set?
1197
1155
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
1158
def create(a_bzrdir):
1314
1159
"""Construct the current default format repository in a_bzrdir."""
1895
1701
@needs_read_lock
1896
1702
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.
1703
"""Get many revisions at once."""
1902
1704
return self._get_revisions(revision_ids)
1904
1706
@needs_read_lock
1905
1707
def _get_revisions(self, revision_ids):
1906
1708
"""Core work logic to get many revisions without sanity checks."""
1709
for rev_id in revision_ids:
1710
if not rev_id or not isinstance(rev_id, basestring):
1711
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1712
keys = [(key,) for key in revision_ids]
1713
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)
1715
for record in stream:
1716
if record.storage_kind == 'absent':
1717
raise errors.NoSuchRevision(self, record.key[0])
1718
text = record.get_bytes_as('fulltext')
1719
rev = self._serializer.read_revision_from_string(text)
1720
revs[record.key[0]] = rev
1912
1721
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)
1724
def get_revision_xml(self, revision_id):
1725
# TODO: jam 20070210 This shouldn't be necessary since get_revision
1726
# would have already do it.
1727
# TODO: jam 20070210 Just use _serializer.write_revision_to_string()
1728
# TODO: this can't just be replaced by:
1729
# return self._serializer.write_revision_to_string(
1730
# self.get_revision(revision_id))
1731
# as cStringIO preservers the encoding unlike write_revision_to_string
1732
# or some other call down the path.
1733
rev = self.get_revision(revision_id)
1734
rev_tmp = cStringIO.StringIO()
1735
# the current serializer..
1736
self._serializer.write_revision(rev, rev_tmp)
1738
return rev_tmp.getvalue()
1941
1740
def get_deltas_for_revisions(self, revisions, specific_fileids=None):
1942
1741
"""Produce a generator of revision deltas.
2392
2173
: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
2174
:return: An iterator of inventories.
2398
2176
if ((None in revision_ids)
2399
2177
or (_mod_revision.NULL_REVISION in revision_ids)):
2400
2178
raise ValueError('cannot get null revision inventory')
2401
return self._iter_inventories(revision_ids, ordering)
2179
return self._iter_inventories(revision_ids)
2403
def _iter_inventories(self, revision_ids, ordering):
2181
def _iter_inventories(self, revision_ids):
2404
2182
"""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)
2183
for text, revision_id in self._iter_inventory_xmls(revision_ids):
2184
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
2186
def _iter_inventory_xmls(self, revision_ids):
2415
2187
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)
2188
stream = self.inventories.get_record_stream(keys, 'unordered', True)
2422
2189
text_chunks = {}
2423
2190
for record in stream:
2424
2191
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]
2192
text_chunks[record.key] = record.get_bytes_as('chunked')
2431
2194
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
2196
chunks = text_chunks.pop(key)
2197
yield ''.join(chunks), key[-1]
2445
def _deserialise_inventory(self, revision_id, xml):
2199
def deserialise_inventory(self, revision_id, xml):
2446
2200
"""Transform the xml into an inventory object.
2448
2202
:param revision_id: The expected revision id of the inventory.
2449
2203
:param xml: A serialised inventory.
2451
2205
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)
2206
entry_cache=self._inventory_entry_cache)
2454
2207
if result.revision_id != revision_id:
2455
2208
raise AssertionError('revision id mismatch %s != %s' % (
2456
2209
result.revision_id, revision_id))
2212
def serialise_inventory(self, inv):
2213
return self._serializer.write_inventory_to_string(inv)
2215
def _serialise_inventory_to_lines(self, inv):
2216
return self._serializer.write_inventory_to_lines(inv)
2459
2218
def get_serializer_format(self):
2460
2219
return self._serializer.format_num
2462
2221
@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')
2222
def get_inventory_xml(self, revision_id):
2223
"""Get inventory XML as a file object."""
2224
texts = self._iter_inventory_xmls([revision_id])
2467
2226
text, revision_id = texts.next()
2468
2227
except StopIteration:
2469
2228
raise errors.HistoryMissing(self, 'inventory', revision_id)
2232
def get_inventory_sha1(self, revision_id):
2233
"""Return the sha1 hash of the inventory entry
2235
return self.get_revision(revision_id).inventory_sha1
2472
2237
def get_rev_id_for_revno(self, revno, known_pair):
2473
2238
"""Return the revision id of a revno, given a later (revno, revid)
2474
2239
pair in the same history.
3735
3453
return self.source.revision_ids_to_search_result(result_set)
3456
class InterPackRepo(InterSameDataRepository):
3457
"""Optimised code paths between Pack based repositories."""
3460
def _get_repo_format_to_test(self):
3461
from bzrlib.repofmt import pack_repo
3462
return pack_repo.RepositoryFormatKnitPack6RichRoot()
3465
def is_compatible(source, target):
3466
"""Be compatible with known Pack formats.
3468
We don't test for the stores being of specific types because that
3469
could lead to confusing results, and there is no need to be
3472
InterPackRepo does not support CHK based repositories.
3474
from bzrlib.repofmt.pack_repo import RepositoryFormatPack
3475
from bzrlib.repofmt.groupcompress_repo import RepositoryFormatCHK1
3477
are_packs = (isinstance(source._format, RepositoryFormatPack) and
3478
isinstance(target._format, RepositoryFormatPack))
3479
not_packs = (isinstance(source._format, RepositoryFormatCHK1) or
3480
isinstance(target._format, RepositoryFormatCHK1))
3481
except AttributeError:
3483
if not_packs or not are_packs:
3485
return InterRepository._same_model(source, target)
3488
def fetch(self, revision_id=None, pb=None, find_ghosts=False,
3490
"""See InterRepository.fetch()."""
3491
if (len(self.source._fallback_repositories) > 0 or
3492
len(self.target._fallback_repositories) > 0):
3493
# The pack layer is not aware of fallback repositories, so when
3494
# fetching from a stacked repository or into a stacked repository
3495
# we use the generic fetch logic which uses the VersionedFiles
3496
# attributes on repository.
3497
from bzrlib.fetch import RepoFetcher
3498
fetcher = RepoFetcher(self.target, self.source, revision_id,
3499
pb, find_ghosts, fetch_spec=fetch_spec)
3500
if fetch_spec is not None:
3501
if len(list(fetch_spec.heads)) != 1:
3502
raise AssertionError(
3503
"InterPackRepo.fetch doesn't support "
3504
"fetching multiple heads yet.")
3505
revision_id = list(fetch_spec.heads)[0]
3507
if revision_id is None:
3509
# everything to do - use pack logic
3510
# to fetch from all packs to one without
3511
# inventory parsing etc, IFF nothing to be copied is in the target.
3513
source_revision_ids = frozenset(self.source.all_revision_ids())
3514
revision_ids = source_revision_ids - \
3515
frozenset(self.target.get_parent_map(source_revision_ids))
3516
revision_keys = [(revid,) for revid in revision_ids]
3517
index = self.target._pack_collection.revision_index.combined_index
3518
present_revision_ids = set(item[1][0] for item in
3519
index.iter_entries(revision_keys))
3520
revision_ids = set(revision_ids) - present_revision_ids
3521
# implementing the TODO will involve:
3522
# - detecting when all of a pack is selected
3523
# - avoiding as much as possible pre-selection, so the
3524
# more-core routines such as create_pack_from_packs can filter in
3525
# a just-in-time fashion. (though having a HEADS list on a
3526
# repository might make this a lot easier, because we could
3527
# sensibly detect 'new revisions' without doing a full index scan.
3528
elif _mod_revision.is_null(revision_id):
3532
revision_ids = self.search_missing_revision_ids(revision_id,
3533
find_ghosts=find_ghosts).get_keys()
3534
if len(revision_ids) == 0:
3536
return self._pack(self.source, self.target, revision_ids)
3538
def _pack(self, source, target, revision_ids):
3539
from bzrlib.repofmt.pack_repo import Packer
3540
packs = source._pack_collection.all_packs()
3541
pack = Packer(self.target._pack_collection, packs, '.fetch',
3542
revision_ids).pack()
3543
if pack is not None:
3544
self.target._pack_collection._save_pack_names()
3545
copied_revs = pack.get_revision_count()
3546
# Trigger an autopack. This may duplicate effort as we've just done
3547
# a pack creation, but for now it is simpler to think about as
3548
# 'upload data, then repack if needed'.
3549
self.target._pack_collection.autopack()
3550
return (copied_revs, [])
3555
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
3556
"""See InterRepository.missing_revision_ids().
3558
:param find_ghosts: Find ghosts throughout the ancestry of
3561
if not find_ghosts and revision_id is not None:
3562
return self._walk_to_common_revisions([revision_id])
3563
elif revision_id is not None:
3564
# Find ghosts: search for revisions pointing from one repository to
3565
# the other, and vice versa, anywhere in the history of revision_id.
3566
graph = self.target.get_graph(other_repository=self.source)
3567
searcher = graph._make_breadth_first_searcher([revision_id])
3571
next_revs, ghosts = searcher.next_with_ghosts()
3572
except StopIteration:
3574
if revision_id in ghosts:
3575
raise errors.NoSuchRevision(self.source, revision_id)
3576
found_ids.update(next_revs)
3577
found_ids.update(ghosts)
3578
found_ids = frozenset(found_ids)
3579
# Double query here: should be able to avoid this by changing the
3580
# graph api further.
3581
result_set = found_ids - frozenset(
3582
self.target.get_parent_map(found_ids))
3584
source_ids = self.source.all_revision_ids()
3585
# source_ids is the worst possible case we may need to pull.
3586
# now we want to filter source_ids against what we actually
3587
# have in target, but don't try to check for existence where we know
3588
# we do not have a revision as that would be pointless.
3589
target_ids = set(self.target.all_revision_ids())
3590
result_set = set(source_ids).difference(target_ids)
3591
return self.source.revision_ids_to_search_result(result_set)
3738
3594
class InterDifferingSerializer(InterRepository):
3747
3603
# This is redundant with format.check_conversion_target(), however that
3748
3604
# raises an exception, and we just want to say "False" as in we won't
3749
3605
# support converting between these formats.
3750
if 'IDS_never' in debug.debug_flags:
3752
3606
if source.supports_rich_root() and not target.supports_rich_root():
3754
3608
if (source._format.supports_tree_reference
3755
3609
and not target._format.supports_tree_reference):
3757
if target._fallback_repositories and target._format.supports_chks:
3758
# IDS doesn't know how to copy CHKs for the parent inventories it
3759
# adds to stacked repos.
3761
if 'IDS_always' in debug.debug_flags:
3763
# Only use this code path for local source and target. IDS does far
3764
# too much IO (both bandwidth and roundtrips) over a network.
3765
if not source.bzrdir.transport.base.startswith('file:///'):
3767
if not target.bzrdir.transport.base.startswith('file:///'):
3771
def _get_trees(self, revision_ids, cache):
3773
for rev_id in revision_ids:
3775
possible_trees.append((rev_id, cache[rev_id]))
3777
# Not cached, but inventory might be present anyway.
3779
tree = self.source.revision_tree(rev_id)
3780
except errors.NoSuchRevision:
3781
# Nope, parent is ghost.
3784
cache[rev_id] = tree
3785
possible_trees.append((rev_id, tree))
3786
return possible_trees
3788
def _get_delta_for_revision(self, tree, parent_ids, possible_trees):
3613
def _get_delta_for_revision(self, tree, parent_ids, basis_id, cache):
3789
3614
"""Get the best delta and base for this revision.
3791
3616
:return: (basis_id, delta)
3618
possible_trees = [(parent_id, cache[parent_id])
3619
for parent_id in parent_ids
3620
if parent_id in cache]
3621
if len(possible_trees) == 0:
3622
# There either aren't any parents, or the parents aren't in the
3623
# cache, so just use the last converted tree
3624
possible_trees.append((basis_id, cache[basis_id]))
3794
# Generate deltas against each tree, to find the shortest.
3795
texts_possibly_new_in_tree = set()
3796
3626
for basis_id, basis_tree in possible_trees:
3797
3627
delta = tree.inventory._make_delta(basis_tree.inventory)
3798
for old_path, new_path, file_id, new_entry in delta:
3799
if new_path is None:
3800
# This file_id isn't present in the new rev, so we don't
3804
# Rich roots are handled elsewhere...
3806
kind = new_entry.kind
3807
if kind != 'directory' and kind != 'file':
3808
# No text record associated with this inventory entry.
3810
# This is a directory or file that has changed somehow.
3811
texts_possibly_new_in_tree.add((file_id, new_entry.revision))
3812
3628
deltas.append((len(delta), basis_id, delta))
3814
3630
return deltas[0][1:]
3816
def _fetch_parent_invs_for_stacking(self, parent_map, cache):
3817
"""Find all parent revisions that are absent, but for which the
3818
inventory is present, and copy those inventories.
3632
def _get_parent_keys(self, root_key, parent_map):
3633
"""Get the parent keys for a given root id."""
3634
root_id, rev_id = root_key
3635
# Include direct parents of the revision, but only if they used
3636
# the same root_id and are heads.
3638
for parent_id in parent_map[rev_id]:
3639
if parent_id == _mod_revision.NULL_REVISION:
3641
if parent_id not in self._revision_id_to_root_id:
3642
# We probably didn't read this revision, go spend the
3643
# extra effort to actually check
3645
tree = self.source.revision_tree(parent_id)
3646
except errors.NoSuchRevision:
3647
# Ghost, fill out _revision_id_to_root_id in case we
3648
# encounter this again.
3649
# But set parent_root_id to None since we don't really know
3650
parent_root_id = None
3652
parent_root_id = tree.get_root_id()
3653
self._revision_id_to_root_id[parent_id] = None
3655
parent_root_id = self._revision_id_to_root_id[parent_id]
3656
if root_id == parent_root_id:
3657
# With stacking we _might_ want to refer to a non-local
3658
# revision, but this code path only applies when we have the
3659
# full content available, so ghosts really are ghosts, not just
3660
# the edge of local data.
3661
parent_keys.append((parent_id,))
3663
# root_id may be in the parent anyway.
3665
tree = self.source.revision_tree(parent_id)
3666
except errors.NoSuchRevision:
3667
# ghost, can't refer to it.
3671
parent_keys.append((tree.inventory[root_id].revision,))
3672
except errors.NoSuchId:
3675
g = graph.Graph(self.source.revisions)
3676
heads = g.heads(parent_keys)
3678
for key in parent_keys:
3679
if key in heads and key not in selected_keys:
3680
selected_keys.append(key)
3681
return tuple([(root_id,)+ key for key in selected_keys])
3820
This is necessary to preserve correctness when the source is stacked
3821
without fallbacks configured. (Note that in cases like upgrade the
3822
source may be not have _fallback_repositories even though it is
3826
for parents in parent_map.values():
3827
parent_revs.update(parents)
3828
present_parents = self.source.get_parent_map(parent_revs)
3829
absent_parents = set(parent_revs).difference(present_parents)
3830
parent_invs_keys_for_stacking = self.source.inventories.get_parent_map(
3831
(rev_id,) for rev_id in absent_parents)
3832
parent_inv_ids = [key[-1] for key in parent_invs_keys_for_stacking]
3833
for parent_tree in self.source.revision_trees(parent_inv_ids):
3834
current_revision_id = parent_tree.get_revision_id()
3835
parents_parents_keys = parent_invs_keys_for_stacking[
3836
(current_revision_id,)]
3837
parents_parents = [key[-1] for key in parents_parents_keys]
3838
basis_id = _mod_revision.NULL_REVISION
3839
basis_tree = self.source.revision_tree(basis_id)
3840
delta = parent_tree.inventory._make_delta(basis_tree.inventory)
3841
self.target.add_inventory_by_delta(
3842
basis_id, delta, current_revision_id, parents_parents)
3843
cache[current_revision_id] = parent_tree
3683
def _new_root_data_stream(self, root_keys_to_create, parent_map):
3684
for root_key in root_keys_to_create:
3685
parent_keys = self._get_parent_keys(root_key, parent_map)
3686
yield versionedfile.FulltextContentFactory(root_key,
3687
parent_keys, None, '')
3845
3689
def _fetch_batch(self, revision_ids, basis_id, cache):
3846
3690
"""Fetch across a few revisions.
3860
3704
pending_deltas = []
3861
3705
pending_revisions = []
3862
3706
parent_map = self.source.get_parent_map(revision_ids)
3863
self._fetch_parent_invs_for_stacking(parent_map, cache)
3864
self.source._safe_to_return_from_cache = True
3865
3707
for tree in self.source.revision_trees(revision_ids):
3866
# Find a inventory delta for this revision.
3867
# Find text entries that need to be copied, too.
3868
3708
current_revision_id = tree.get_revision_id()
3869
3709
parent_ids = parent_map.get(current_revision_id, ())
3870
parent_trees = self._get_trees(parent_ids, cache)
3871
possible_trees = list(parent_trees)
3872
if len(possible_trees) == 0:
3873
# There either aren't any parents, or the parents are ghosts,
3874
# so just use the last converted tree.
3875
possible_trees.append((basis_id, cache[basis_id]))
3876
3710
basis_id, delta = self._get_delta_for_revision(tree, parent_ids,
3878
revision = self.source.get_revision(current_revision_id)
3879
pending_deltas.append((basis_id, delta,
3880
current_revision_id, revision.parent_ids))
3881
3712
if self._converting_to_rich_root:
3882
3713
self._revision_id_to_root_id[current_revision_id] = \
3883
3714
tree.get_root_id()
3884
# Determine which texts are in present in this revision but not in
3885
# any of the available parents.
3886
texts_possibly_new_in_tree = set()
3715
# Find text entries that need to be copied
3887
3716
for old_path, new_path, file_id, entry in delta:
3888
if new_path is None:
3889
# This file_id isn't present in the new rev
3893
if not self.target.supports_rich_root():
3894
# The target doesn't support rich root, so we don't
3897
if self._converting_to_rich_root:
3898
# This can't be copied normally, we have to insert
3900
root_keys_to_create.add((file_id, entry.revision))
3903
texts_possibly_new_in_tree.add((file_id, entry.revision))
3904
for basis_id, basis_tree in possible_trees:
3905
basis_inv = basis_tree.inventory
3906
for file_key in list(texts_possibly_new_in_tree):
3907
file_id, file_revision = file_key
3909
entry = basis_inv[file_id]
3910
except errors.NoSuchId:
3912
if entry.revision == file_revision:
3913
texts_possibly_new_in_tree.remove(file_key)
3914
text_keys.update(texts_possibly_new_in_tree)
3717
if new_path is not None:
3720
if not self.target.supports_rich_root():
3721
# The target doesn't support rich root, so we don't
3724
if self._converting_to_rich_root:
3725
# This can't be copied normally, we have to insert
3727
root_keys_to_create.add((file_id, entry.revision))
3729
text_keys.add((file_id, entry.revision))
3730
revision = self.source.get_revision(current_revision_id)
3731
pending_deltas.append((basis_id, delta,
3732
current_revision_id, revision.parent_ids))
3915
3733
pending_revisions.append(revision)
3916
3734
cache[current_revision_id] = tree
3917
3735
basis_id = current_revision_id
3918
self.source._safe_to_return_from_cache = False
3919
3736
# Copy file texts
3920
3737
from_texts = self.source.texts
3921
3738
to_texts = self.target.texts
3922
3739
if root_keys_to_create:
3923
root_stream = _mod_fetch._new_root_data_stream(
3924
root_keys_to_create, self._revision_id_to_root_id, parent_map,
3740
root_stream = self._new_root_data_stream(root_keys_to_create,
3926
3742
to_texts.insert_record_stream(root_stream)
3927
3743
to_texts.insert_record_stream(from_texts.get_record_stream(
3928
3744
text_keys, self.target._format._fetch_order,
4372
4150
# missing keys can handle suspending a write group).
4373
4151
write_group_tokens = self.target_repo.suspend_write_group()
4374
4152
return write_group_tokens, missing_keys
4375
hint = self.target_repo.commit_write_group()
4376
if (to_serializer != src_serializer and
4377
self.target_repo._format.pack_compresses):
4378
self.target_repo.pack(hint=hint)
4153
self.target_repo.commit_write_group()
4379
4154
return [], set()
4381
def _extract_and_insert_inventory_deltas(self, substream, serializer):
4382
target_rich_root = self.target_repo._format.rich_root_data
4383
target_tree_refs = self.target_repo._format.supports_tree_reference
4384
for record in substream:
4385
# Insert the delta directly
4386
inventory_delta_bytes = record.get_bytes_as('fulltext')
4387
deserialiser = inventory_delta.InventoryDeltaDeserializer()
4389
parse_result = deserialiser.parse_text_bytes(
4390
inventory_delta_bytes)
4391
except inventory_delta.IncompatibleInventoryDelta, err:
4392
trace.mutter("Incompatible delta: %s", err.msg)
4393
raise errors.IncompatibleRevision(self.target_repo._format)
4394
basis_id, new_id, rich_root, tree_refs, inv_delta = parse_result
4395
revision_id = new_id
4396
parents = [key[0] for key in record.parents]
4397
self.target_repo.add_inventory_by_delta(
4398
basis_id, inv_delta, revision_id, parents)
4400
def _extract_and_insert_inventories(self, substream, serializer,
4156
def _extract_and_insert_inventories(self, substream, serializer):
4402
4157
"""Generate a new inventory versionedfile in target, converting data.
4404
4159
The inventory is retrieved from the source, (deserializing it), and
4405
4160
stored in the target (reserializing it in a different format).
4407
target_rich_root = self.target_repo._format.rich_root_data
4408
target_tree_refs = self.target_repo._format.supports_tree_reference
4409
4162
for record in substream:
4410
# It's not a delta, so it must be a fulltext in the source
4411
# serializer's format.
4412
4163
bytes = record.get_bytes_as('fulltext')
4413
4164
revision_id = record.key[0]
4414
4165
inv = serializer.read_inventory_from_string(bytes, revision_id)
4415
4166
parents = [key[0] for key in record.parents]
4416
4167
self.target_repo.add_inventory(revision_id, inv, parents)
4417
# No need to keep holding this full inv in memory when the rest of
4418
# the substream is likely to be all deltas.
4421
4169
def _extract_and_insert_revisions(self, substream, serializer):
4422
4170
for record in substream:
4586
4328
return (not self.from_repository._format.rich_root_data and
4587
4329
self.to_format.rich_root_data)
4589
def _get_inventory_stream(self, revision_ids, missing=False):
4331
def _get_inventory_stream(self, revision_ids):
4590
4332
from_format = self.from_repository._format
4591
if (from_format.supports_chks and self.to_format.supports_chks and
4592
from_format.network_name() == self.to_format.network_name()):
4593
raise AssertionError(
4594
"this case should be handled by GroupCHKStreamSource")
4595
elif 'forceinvdeltas' in debug.debug_flags:
4596
return self._get_convertable_inventory_stream(revision_ids,
4597
delta_versus_null=missing)
4598
elif from_format.network_name() == self.to_format.network_name():
4600
return self._get_simple_inventory_stream(revision_ids,
4602
elif (not from_format.supports_chks and not self.to_format.supports_chks
4603
and from_format._serializer == self.to_format._serializer):
4604
# Essentially the same format.
4605
return self._get_simple_inventory_stream(revision_ids,
4333
if (from_format.supports_chks and self.to_format.supports_chks
4334
and (from_format._serializer == self.to_format._serializer)):
4335
# Both sides support chks, and they use the same serializer, so it
4336
# is safe to transmit the chk pages and inventory pages across
4338
return self._get_chk_inventory_stream(revision_ids)
4339
elif (not from_format.supports_chks):
4340
# Source repository doesn't support chks. So we can transmit the
4341
# inventories 'as-is' and either they are just accepted on the
4342
# target, or the Sink will properly convert it.
4343
return self._get_simple_inventory_stream(revision_ids)
4608
# Any time we switch serializations, we want to use an
4609
# inventory-delta based approach.
4610
return self._get_convertable_inventory_stream(revision_ids,
4611
delta_versus_null=missing)
4345
# XXX: Hack to make not-chk->chk fetch: copy the inventories as
4346
# inventories. Note that this should probably be done somehow
4347
# as part of bzrlib.repository.StreamSink. Except JAM couldn't
4348
# figure out how a non-chk repository could possibly handle
4349
# deserializing an inventory stream from a chk repo, as it
4350
# doesn't have a way to understand individual pages.
4351
return self._get_convertable_inventory_stream(revision_ids)
4613
def _get_simple_inventory_stream(self, revision_ids, missing=False):
4614
# NB: This currently reopens the inventory weave in source;
4615
# using a single stream interface instead would avoid this.
4353
def _get_simple_inventory_stream(self, revision_ids):
4616
4354
from_weave = self.from_repository.inventories
4618
delta_closure = True
4620
delta_closure = not self.delta_on_metadata()
4621
4355
yield ('inventories', from_weave.get_record_stream(
4622
4356
[(rev_id,) for rev_id in revision_ids],
4623
self.inventory_fetch_order(), delta_closure))
4625
def _get_convertable_inventory_stream(self, revision_ids,
4626
delta_versus_null=False):
4627
# The two formats are sufficiently different that there is no fast
4628
# path, so we need to send just inventorydeltas, which any
4629
# sufficiently modern client can insert into any repository.
4630
# The StreamSink code expects to be able to
4631
# convert on the target, so we need to put bytes-on-the-wire that can
4632
# be converted. That means inventory deltas (if the remote is <1.19,
4633
# RemoteStreamSink will fallback to VFS to insert the deltas).
4634
yield ('inventory-deltas',
4635
self._stream_invs_as_deltas(revision_ids,
4636
delta_versus_null=delta_versus_null))
4638
def _stream_invs_as_deltas(self, revision_ids, delta_versus_null=False):
4639
"""Return a stream of inventory-deltas for the given rev ids.
4641
:param revision_ids: The list of inventories to transmit
4642
:param delta_versus_null: Don't try to find a minimal delta for this
4643
entry, instead compute the delta versus the NULL_REVISION. This
4644
effectively streams a complete inventory. Used for stuff like
4645
filling in missing parents, etc.
4357
self.inventory_fetch_order(),
4358
not self.delta_on_metadata()))
4360
def _get_chk_inventory_stream(self, revision_ids):
4361
"""Fetch the inventory texts, along with the associated chk maps."""
4362
# We want an inventory outside of the search set, so that we can filter
4363
# out uninteresting chk pages. For now we use
4364
# _find_revision_outside_set, but if we had a Search with cut_revs, we
4365
# could use that instead.
4366
start_rev_id = self.from_repository._find_revision_outside_set(
4368
start_rev_key = (start_rev_id,)
4369
inv_keys_to_fetch = [(rev_id,) for rev_id in revision_ids]
4370
if start_rev_id != _mod_revision.NULL_REVISION:
4371
inv_keys_to_fetch.append((start_rev_id,))
4372
# Any repo that supports chk_bytes must also support out-of-order
4373
# insertion. At least, that is how we expect it to work
4374
# We use get_record_stream instead of iter_inventories because we want
4375
# to be able to insert the stream as well. We could instead fetch
4376
# allowing deltas, and then iter_inventories, but we don't know whether
4377
# source or target is more 'local' anway.
4378
inv_stream = self.from_repository.inventories.get_record_stream(
4379
inv_keys_to_fetch, 'unordered',
4380
True) # We need them as full-texts so we can find their references
4381
uninteresting_chk_roots = set()
4382
interesting_chk_roots = set()
4383
def filter_inv_stream(inv_stream):
4384
for idx, record in enumerate(inv_stream):
4385
### child_pb.update('fetch inv', idx, len(inv_keys_to_fetch))
4386
bytes = record.get_bytes_as('fulltext')
4387
chk_inv = inventory.CHKInventory.deserialise(
4388
self.from_repository.chk_bytes, bytes, record.key)
4389
if record.key == start_rev_key:
4390
uninteresting_chk_roots.add(chk_inv.id_to_entry.key())
4391
p_id_map = chk_inv.parent_id_basename_to_file_id
4392
if p_id_map is not None:
4393
uninteresting_chk_roots.add(p_id_map.key())
4396
interesting_chk_roots.add(chk_inv.id_to_entry.key())
4397
p_id_map = chk_inv.parent_id_basename_to_file_id
4398
if p_id_map is not None:
4399
interesting_chk_roots.add(p_id_map.key())
4400
### pb.update('fetch inventory', 0, 2)
4401
yield ('inventories', filter_inv_stream(inv_stream))
4402
# Now that we have worked out all of the interesting root nodes, grab
4403
# all of the interesting pages and insert them
4404
### pb.update('fetch inventory', 1, 2)
4405
interesting = chk_map.iter_interesting_nodes(
4406
self.from_repository.chk_bytes, interesting_chk_roots,
4407
uninteresting_chk_roots)
4408
def to_stream_adapter():
4409
"""Adapt the iter_interesting_nodes result to a single stream.
4411
iter_interesting_nodes returns records as it processes them, along
4412
with keys. However, we only want to return the records themselves.
4414
for record, items in interesting:
4415
if record is not None:
4417
# XXX: We could instead call get_record_stream(records.keys())
4418
# ATM, this will always insert the records as fulltexts, and
4419
# requires that you can hang on to records once you have gone
4420
# on to the next one. Further, it causes the target to
4421
# recompress the data. Testing shows it to be faster than
4422
# requesting the records again, though.
4423
yield ('chk_bytes', to_stream_adapter())
4424
### pb.update('fetch inventory', 2, 2)
4426
def _get_convertable_inventory_stream(self, revision_ids):
4427
# XXX: One of source or target is using chks, and they don't have
4428
# compatible serializations. The StreamSink code expects to be
4429
# able to convert on the target, so we need to put
4430
# bytes-on-the-wire that can be converted
4431
yield ('inventories', self._stream_invs_as_fulltexts(revision_ids))
4433
def _stream_invs_as_fulltexts(self, revision_ids):
4647
4434
from_repo = self.from_repository
4435
from_serializer = from_repo._format._serializer
4648
4436
revision_keys = [(rev_id,) for rev_id in revision_ids]
4649
4437
parent_map = from_repo.inventories.get_parent_map(revision_keys)
4650
# XXX: possibly repos could implement a more efficient iter_inv_deltas
4652
inventories = self.from_repository.iter_inventories(
4653
revision_ids, 'topological')
4654
format = from_repo._format
4655
invs_sent_so_far = set([_mod_revision.NULL_REVISION])
4656
inventory_cache = lru_cache.LRUCache(50)
4657
null_inventory = from_repo.revision_tree(
4658
_mod_revision.NULL_REVISION).inventory
4659
# XXX: ideally the rich-root/tree-refs flags would be per-revision, not
4660
# per-repo (e.g. streaming a non-rich-root revision out of a rich-root
4661
# repo back into a non-rich-root repo ought to be allowed)
4662
serializer = inventory_delta.InventoryDeltaSerializer(
4663
versioned_root=format.rich_root_data,
4664
tree_references=format.supports_tree_reference)
4665
for inv in inventories:
4438
for inv in self.from_repository.iter_inventories(revision_ids):
4439
# XXX: This is a bit hackish, but it works. Basically,
4440
# CHKSerializer 'accidentally' supports
4441
# read/write_inventory_to_string, even though that is never
4442
# the format that is stored on disk. It *does* give us a
4443
# single string representation for an inventory, so live with
4445
# This would be far better if we had a 'serialized inventory
4446
# delta' form. Then we could use 'inventory._make_delta', and
4447
# transmit that. This would both be faster to generate, and
4448
# result in fewer bytes-on-the-wire.
4449
as_bytes = from_serializer.write_inventory_to_string(inv)
4666
4450
key = (inv.revision_id,)
4667
4451
parent_keys = parent_map.get(key, ())
4669
if not delta_versus_null and parent_keys:
4670
# The caller did not ask for complete inventories and we have
4671
# some parents that we can delta against. Make a delta against
4672
# each parent so that we can find the smallest.
4673
parent_ids = [parent_key[0] for parent_key in parent_keys]
4674
for parent_id in parent_ids:
4675
if parent_id not in invs_sent_so_far:
4676
# We don't know that the remote side has this basis, so
4679
if parent_id == _mod_revision.NULL_REVISION:
4680
parent_inv = null_inventory
4682
parent_inv = inventory_cache.get(parent_id, None)
4683
if parent_inv is None:
4684
parent_inv = from_repo.get_inventory(parent_id)
4685
candidate_delta = inv._make_delta(parent_inv)
4686
if (delta is None or
4687
len(delta) > len(candidate_delta)):
4688
delta = candidate_delta
4689
basis_id = parent_id
4691
# Either none of the parents ended up being suitable, or we
4692
# were asked to delta against NULL
4693
basis_id = _mod_revision.NULL_REVISION
4694
delta = inv._make_delta(null_inventory)
4695
invs_sent_so_far.add(inv.revision_id)
4696
inventory_cache[inv.revision_id] = inv
4697
delta_serialized = ''.join(
4698
serializer.delta_to_lines(basis_id, key[-1], delta))
4699
4452
yield versionedfile.FulltextContentFactory(
4700
key, parent_keys, None, delta_serialized)
4453
key, parent_keys, None, as_bytes)
4703
4456
def _iter_for_revno(repo, partial_history_cache, stop_index=None,
added
removed
bzrlib/repository.py
