1111
1195
# The old API returned a list, should this actually be a set?
1112
1196
return parent_map.keys()
1198
def _check_inventories(self, checker):
1199
"""Check the inventories found from the revision scan.
1201
This is responsible for verifying the sha1 of inventories and
1202
creating a pending_keys set that covers data referenced by inventories.
1204
bar = ui.ui_factory.nested_progress_bar()
1206
self._do_check_inventories(checker, bar)
1210
def _do_check_inventories(self, checker, bar):
1211
"""Helper for _check_inventories."""
1213
keys = {'chk_bytes':set(), 'inventories':set(), 'texts':set()}
1214
kinds = ['chk_bytes', 'texts']
1215
count = len(checker.pending_keys)
1216
bar.update("inventories", 0, 2)
1217
current_keys = checker.pending_keys
1218
checker.pending_keys = {}
1219
# Accumulate current checks.
1220
for key in current_keys:
1221
if key[0] != 'inventories' and key[0] not in kinds:
1222
checker._report_items.append('unknown key type %r' % (key,))
1223
keys[key[0]].add(key[1:])
1224
if keys['inventories']:
1225
# NB: output order *should* be roughly sorted - topo or
1226
# inverse topo depending on repository - either way decent
1227
# to just delta against. However, pre-CHK formats didn't
1228
# try to optimise inventory layout on disk. As such the
1229
# pre-CHK code path does not use inventory deltas.
1231
for record in self.inventories.check(keys=keys['inventories']):
1232
if record.storage_kind == 'absent':
1233
checker._report_items.append(
1234
'Missing inventory {%s}' % (record.key,))
1236
last_object = self._check_record('inventories', record,
1237
checker, last_object,
1238
current_keys[('inventories',) + record.key])
1239
del keys['inventories']
1242
bar.update("texts", 1)
1243
while (checker.pending_keys or keys['chk_bytes']
1245
# Something to check.
1246
current_keys = checker.pending_keys
1247
checker.pending_keys = {}
1248
# Accumulate current checks.
1249
for key in current_keys:
1250
if key[0] not in kinds:
1251
checker._report_items.append('unknown key type %r' % (key,))
1252
keys[key[0]].add(key[1:])
1253
# Check the outermost kind only - inventories || chk_bytes || texts
1257
for record in getattr(self, kind).check(keys=keys[kind]):
1258
if record.storage_kind == 'absent':
1259
checker._report_items.append(
1260
'Missing %s {%s}' % (kind, record.key,))
1262
last_object = self._check_record(kind, record,
1263
checker, last_object, current_keys[(kind,) + record.key])
1267
def _check_record(self, kind, record, checker, last_object, item_data):
1268
"""Check a single text from this repository."""
1269
if kind == 'inventories':
1270
rev_id = record.key[0]
1271
inv = self._deserialise_inventory(rev_id,
1272
record.get_bytes_as('fulltext'))
1273
if last_object is not None:
1274
delta = inv._make_delta(last_object)
1275
for old_path, path, file_id, ie in delta:
1278
ie.check(checker, rev_id, inv)
1280
for path, ie in inv.iter_entries():
1281
ie.check(checker, rev_id, inv)
1282
if self._format.fast_deltas:
1284
elif kind == 'chk_bytes':
1285
# No code written to check chk_bytes for this repo format.
1286
checker._report_items.append(
1287
'unsupported key type chk_bytes for %s' % (record.key,))
1288
elif kind == 'texts':
1289
self._check_text(record, checker, item_data)
1291
checker._report_items.append(
1292
'unknown key type %s for %s' % (kind, record.key))
1294
def _check_text(self, record, checker, item_data):
1295
"""Check a single text."""
1296
# Check it is extractable.
1297
# TODO: check length.
1298
if record.storage_kind == 'chunked':
1299
chunks = record.get_bytes_as(record.storage_kind)
1300
sha1 = osutils.sha_strings(chunks)
1301
length = sum(map(len, chunks))
1303
content = record.get_bytes_as('fulltext')
1304
sha1 = osutils.sha_string(content)
1305
length = len(content)
1306
if item_data and sha1 != item_data[1]:
1307
checker._report_items.append(
1308
'sha1 mismatch: %s has sha1 %s expected %s referenced by %s' %
1309
(record.key, sha1, item_data[1], item_data[2]))
1115
1312
def create(a_bzrdir):
1116
1313
"""Construct the current default format repository in a_bzrdir."""
1385
1612
def suspend_write_group(self):
1386
1613
raise errors.UnsuspendableWriteGroup(self)
1615
def get_missing_parent_inventories(self, check_for_missing_texts=True):
1616
"""Return the keys of missing inventory parents for revisions added in
1619
A revision is not complete if the inventory delta for that revision
1620
cannot be calculated. Therefore if the parent inventories of a
1621
revision are not present, the revision is incomplete, and e.g. cannot
1622
be streamed by a smart server. This method finds missing inventory
1623
parents for revisions added in this write group.
1625
if not self._format.supports_external_lookups:
1626
# This is only an issue for stacked repositories
1628
if not self.is_in_write_group():
1629
raise AssertionError('not in a write group')
1631
# XXX: We assume that every added revision already has its
1632
# corresponding inventory, so we only check for parent inventories that
1633
# might be missing, rather than all inventories.
1634
parents = set(self.revisions._index.get_missing_parents())
1635
parents.discard(_mod_revision.NULL_REVISION)
1636
unstacked_inventories = self.inventories._index
1637
present_inventories = unstacked_inventories.get_parent_map(
1638
key[-1:] for key in parents)
1639
parents.difference_update(present_inventories)
1640
if len(parents) == 0:
1641
# No missing parent inventories.
1643
if not check_for_missing_texts:
1644
return set(('inventories', rev_id) for (rev_id,) in parents)
1645
# Ok, now we have a list of missing inventories. But these only matter
1646
# if the inventories that reference them are missing some texts they
1647
# appear to introduce.
1648
# XXX: Texts referenced by all added inventories need to be present,
1649
# but at the moment we're only checking for texts referenced by
1650
# inventories at the graph's edge.
1651
key_deps = self.revisions._index._key_dependencies
1652
key_deps.satisfy_refs_for_keys(present_inventories)
1653
referrers = frozenset(r[0] for r in key_deps.get_referrers())
1654
file_ids = self.fileids_altered_by_revision_ids(referrers)
1655
missing_texts = set()
1656
for file_id, version_ids in file_ids.iteritems():
1657
missing_texts.update(
1658
(file_id, version_id) for version_id in version_ids)
1659
present_texts = self.texts.get_parent_map(missing_texts)
1660
missing_texts.difference_update(present_texts)
1661
if not missing_texts:
1662
# No texts are missing, so all revisions and their deltas are
1665
# Alternatively the text versions could be returned as the missing
1666
# keys, but this is likely to be less data.
1667
missing_keys = set(('inventories', rev_id) for (rev_id,) in parents)
1388
1670
def refresh_data(self):
1389
"""Re-read any data needed to to synchronise with disk.
1671
"""Re-read any data needed to synchronise with disk.
1391
1673
This method is intended to be called after another repository instance
1392
1674
(such as one used by a smart server) has inserted data into the
1393
repository. It may not be called during a write group, but may be
1394
called at any other time.
1675
repository. On all repositories this will work outside of write groups.
1676
Some repository formats (pack and newer for bzrlib native formats)
1677
support refresh_data inside write groups. If called inside a write
1678
group on a repository that does not support refreshing in a write group
1679
IsInWriteGroupError will be raised.
1396
if self.is_in_write_group():
1397
raise errors.InternalBzrError(
1398
"May not refresh_data while in a write group.")
1399
1681
self._refresh_data()
1401
1683
def resume_write_group(self, tokens):
1603
1894
@needs_read_lock
1604
1895
def get_revisions(self, revision_ids):
1605
"""Get many revisions at once."""
1896
"""Get many revisions at once.
1898
Repositories that need to check data on every revision read should
1899
subclass this method.
1606
1901
return self._get_revisions(revision_ids)
1608
1903
@needs_read_lock
1609
1904
def _get_revisions(self, revision_ids):
1610
1905
"""Core work logic to get many revisions without sanity checks."""
1611
for rev_id in revision_ids:
1612
if not rev_id or not isinstance(rev_id, basestring):
1613
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1907
for revid, rev in self._iter_revisions(revision_ids):
1909
raise errors.NoSuchRevision(self, revid)
1911
return [revs[revid] for revid in revision_ids]
1913
def _iter_revisions(self, revision_ids):
1914
"""Iterate over revision objects.
1916
:param revision_ids: An iterable of revisions to examine. None may be
1917
passed to request all revisions known to the repository. Note that
1918
not all repositories can find unreferenced revisions; for those
1919
repositories only referenced ones will be returned.
1920
:return: An iterator of (revid, revision) tuples. Absent revisions (
1921
those asked for but not available) are returned as (revid, None).
1923
if revision_ids is None:
1924
revision_ids = self.all_revision_ids()
1926
for rev_id in revision_ids:
1927
if not rev_id or not isinstance(rev_id, basestring):
1928
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1614
1929
keys = [(key,) for key in revision_ids]
1615
1930
stream = self.revisions.get_record_stream(keys, 'unordered', True)
1617
1931
for record in stream:
1932
revid = record.key[0]
1618
1933
if record.storage_kind == 'absent':
1619
raise errors.NoSuchRevision(self, record.key[0])
1620
text = record.get_bytes_as('fulltext')
1621
rev = self._serializer.read_revision_from_string(text)
1622
revs[record.key[0]] = rev
1623
return [revs[revid] for revid in revision_ids]
1626
def get_revision_xml(self, revision_id):
1627
# TODO: jam 20070210 This shouldn't be necessary since get_revision
1628
# would have already do it.
1629
# TODO: jam 20070210 Just use _serializer.write_revision_to_string()
1630
rev = self.get_revision(revision_id)
1631
rev_tmp = cStringIO.StringIO()
1632
# the current serializer..
1633
self._serializer.write_revision(rev, rev_tmp)
1635
return rev_tmp.getvalue()
1936
text = record.get_bytes_as('fulltext')
1937
rev = self._serializer.read_revision_from_string(text)
1637
1940
def get_deltas_for_revisions(self, revisions, specific_fileids=None):
1638
1941
"""Produce a generator of revision deltas.
2070
2391
:param revision_ids: The expected revision ids of the inventories.
2392
:param ordering: optional ordering, e.g. 'topological'. If not
2393
specified, the order of revision_ids will be preserved (by
2394
buffering if necessary).
2071
2395
:return: An iterator of inventories.
2073
2397
if ((None in revision_ids)
2074
2398
or (_mod_revision.NULL_REVISION in revision_ids)):
2075
2399
raise ValueError('cannot get null revision inventory')
2076
return self._iter_inventories(revision_ids)
2400
return self._iter_inventories(revision_ids, ordering)
2078
def _iter_inventories(self, revision_ids):
2402
def _iter_inventories(self, revision_ids, ordering):
2079
2403
"""single-document based inventory iteration."""
2080
for text, revision_id in self._iter_inventory_xmls(revision_ids):
2081
yield self.deserialise_inventory(revision_id, text)
2404
inv_xmls = self._iter_inventory_xmls(revision_ids, ordering)
2405
for text, revision_id in inv_xmls:
2406
yield self._deserialise_inventory(revision_id, text)
2083
def _iter_inventory_xmls(self, revision_ids):
2408
def _iter_inventory_xmls(self, revision_ids, ordering):
2409
if ordering is None:
2410
order_as_requested = True
2411
ordering = 'unordered'
2413
order_as_requested = False
2084
2414
keys = [(revision_id,) for revision_id in revision_ids]
2085
stream = self.inventories.get_record_stream(keys, 'unordered', True)
2417
if order_as_requested:
2418
key_iter = iter(keys)
2419
next_key = key_iter.next()
2420
stream = self.inventories.get_record_stream(keys, ordering, True)
2086
2421
text_chunks = {}
2087
2422
for record in stream:
2088
2423
if record.storage_kind != 'absent':
2089
text_chunks[record.key] = record.get_bytes_as('chunked')
2424
chunks = record.get_bytes_as('chunked')
2425
if order_as_requested:
2426
text_chunks[record.key] = chunks
2428
yield ''.join(chunks), record.key[-1]
2091
2430
raise errors.NoSuchRevision(self, record.key)
2093
chunks = text_chunks.pop(key)
2094
yield ''.join(chunks), key[-1]
2431
if order_as_requested:
2432
# Yield as many results as we can while preserving order.
2433
while next_key in text_chunks:
2434
chunks = text_chunks.pop(next_key)
2435
yield ''.join(chunks), next_key[-1]
2437
next_key = key_iter.next()
2438
except StopIteration:
2439
# We still want to fully consume the get_record_stream,
2440
# just in case it is not actually finished at this point
2096
def deserialise_inventory(self, revision_id, xml):
2444
def _deserialise_inventory(self, revision_id, xml):
2097
2445
"""Transform the xml into an inventory object.
2099
2447
:param revision_id: The expected revision id of the inventory.
2100
2448
:param xml: A serialised inventory.
2102
2450
result = self._serializer.read_inventory_from_string(xml, revision_id,
2103
entry_cache=self._inventory_entry_cache)
2451
entry_cache=self._inventory_entry_cache,
2452
return_from_cache=self._safe_to_return_from_cache)
2104
2453
if result.revision_id != revision_id:
2105
2454
raise AssertionError('revision id mismatch %s != %s' % (
2106
2455
result.revision_id, revision_id))
2109
def serialise_inventory(self, inv):
2110
return self._serializer.write_inventory_to_string(inv)
2112
def _serialise_inventory_to_lines(self, inv):
2113
return self._serializer.write_inventory_to_lines(inv)
2115
2458
def get_serializer_format(self):
2116
2459
return self._serializer.format_num
2118
2461
@needs_read_lock
2119
def get_inventory_xml(self, revision_id):
2120
"""Get inventory XML as a file object."""
2121
texts = self._iter_inventory_xmls([revision_id])
2462
def _get_inventory_xml(self, revision_id):
2463
"""Get serialized inventory as a string."""
2464
texts = self._iter_inventory_xmls([revision_id], 'unordered')
2123
2466
text, revision_id = texts.next()
2124
2467
except StopIteration:
2125
2468
raise errors.HistoryMissing(self, 'inventory', revision_id)
2129
def get_inventory_sha1(self, revision_id):
2130
"""Return the sha1 hash of the inventory entry
2471
def get_rev_id_for_revno(self, revno, known_pair):
2472
"""Return the revision id of a revno, given a later (revno, revid)
2473
pair in the same history.
2475
:return: if found (True, revid). If the available history ran out
2476
before reaching the revno, then this returns
2477
(False, (closest_revno, closest_revid)).
2132
return self.get_revision(revision_id).inventory_sha1
2479
known_revno, known_revid = known_pair
2480
partial_history = [known_revid]
2481
distance_from_known = known_revno - revno
2482
if distance_from_known < 0:
2484
'requested revno (%d) is later than given known revno (%d)'
2485
% (revno, known_revno))
2488
self, partial_history, stop_index=distance_from_known)
2489
except errors.RevisionNotPresent, err:
2490
if err.revision_id == known_revid:
2491
# The start revision (known_revid) wasn't found.
2493
# This is a stacked repository with no fallbacks, or a there's a
2494
# left-hand ghost. Either way, even though the revision named in
2495
# the error isn't in this repo, we know it's the next step in this
2496
# left-hand history.
2497
partial_history.append(err.revision_id)
2498
if len(partial_history) <= distance_from_known:
2499
# Didn't find enough history to get a revid for the revno.
2500
earliest_revno = known_revno - len(partial_history) + 1
2501
return (False, (earliest_revno, partial_history[-1]))
2502
if len(partial_history) - 1 > distance_from_known:
2503
raise AssertionError('_iter_for_revno returned too much history')
2504
return (True, partial_history[-1])
2134
2506
def iter_reverse_revision_history(self, revision_id):
2135
2507
"""Iterate backwards through revision ids in the lefthand history
2945
3363
# Development formats.
2946
# 1.7->1.8 go below here
2947
format_registry.register_lazy(
2948
"Bazaar development format 2 (needs bzr.dev from before 1.8)\n",
2949
'bzrlib.repofmt.pack_repo',
2950
'RepositoryFormatPackDevelopment2',
3364
# Obsolete but kept pending a CHK based subtree format.
2952
3365
format_registry.register_lazy(
2953
3366
("Bazaar development format 2 with subtree support "
2954
3367
"(needs bzr.dev from before 1.8)\n"),
2955
3368
'bzrlib.repofmt.pack_repo',
2956
3369
'RepositoryFormatPackDevelopment2Subtree',
2958
# 1.9->1.110 go below here
2959
format_registry.register_lazy(
2960
# merge-bbc-dev4-to-bzr.dev
2961
"Bazaar development format 5 (needs bzr.dev from before 1.13)\n",
2962
'bzrlib.repofmt.pack_repo',
2963
'RepositoryFormatPackDevelopment5',
2965
format_registry.register_lazy(
2966
# merge-bbc-dev4-to-bzr.dev
2967
("Bazaar development format 5 with subtree support"
2968
" (needs bzr.dev from before 1.13)\n"),
2969
'bzrlib.repofmt.pack_repo',
2970
'RepositoryFormatPackDevelopment5Subtree',
2972
format_registry.register_lazy(
2973
# merge-bbc-dev4-to-bzr.dev
2974
('Bazaar development format 5 hash 16'
2975
' (needs bzr.dev from before 1.13)\n'),
2976
'bzrlib.repofmt.pack_repo',
2977
'RepositoryFormatPackDevelopment5Hash16',
2979
format_registry.register_lazy(
2980
# merge-bbc-dev4-to-bzr.dev
2981
('Bazaar development format 5 hash 255'
2982
' (needs bzr.dev from before 1.13)\n'),
2983
'bzrlib.repofmt.pack_repo',
2984
'RepositoryFormatPackDevelopment5Hash255',
2986
format_registry.register_lazy(
2987
'Bazaar development format - hash16chk+gc rich-root (needs bzr.dev from 1.14)\n',
2988
'bzrlib.repofmt.groupcompress_repo',
2989
'RepositoryFormatPackGCCHK16',
2991
format_registry.register_lazy(
2992
'Bazaar development format - hash255chk+gc rich-root (needs bzr.dev from 1.14)\n',
2993
'bzrlib.repofmt.groupcompress_repo',
2994
'RepositoryFormatPackGCCHK255',
2996
format_registry.register_lazy(
2997
'Bazaar development format - hash255chk+gc rich-root bigpage (needs bzr.dev from 1.14)\n',
2998
'bzrlib.repofmt.groupcompress_repo',
2999
'RepositoryFormatPackGCCHK255Big',
3372
# 1.14->1.16 go below here
3373
format_registry.register_lazy(
3374
'Bazaar development format - group compression and chk inventory'
3375
' (needs bzr.dev from 1.14)\n',
3376
'bzrlib.repofmt.groupcompress_repo',
3377
'RepositoryFormatCHK1',
3380
format_registry.register_lazy(
3381
'Bazaar development format - chk repository with bencode revision '
3382
'serialization (needs bzr.dev from 1.16)\n',
3383
'bzrlib.repofmt.groupcompress_repo',
3384
'RepositoryFormatCHK2',
3386
format_registry.register_lazy(
3387
'Bazaar repository format 2a (needs bzr 1.16 or later)\n',
3388
'bzrlib.repofmt.groupcompress_repo',
3389
'RepositoryFormat2a',
3352
3740
return self.source.revision_ids_to_search_result(result_set)
3355
class InterPackRepo(InterSameDataRepository):
3356
"""Optimised code paths between Pack based repositories."""
3359
def _get_repo_format_to_test(self):
3360
from bzrlib.repofmt import pack_repo
3361
return pack_repo.RepositoryFormatKnitPack1()
3364
def is_compatible(source, target):
3365
"""Be compatible with known Pack formats.
3367
We don't test for the stores being of specific types because that
3368
could lead to confusing results, and there is no need to be
3371
Do not support CHK based repositories at this point.
3373
from bzrlib.repofmt.pack_repo import RepositoryFormatPack
3374
# XXX: This format is scheduled for termination
3375
# from bzrlib.repofmt.groupcompress_repo import (
3376
# RepositoryFormatPackGCPlain,
3379
are_packs = (isinstance(source._format, RepositoryFormatPack) and
3380
isinstance(target._format, RepositoryFormatPack))
3381
except AttributeError:
3385
# if (isinstance(source._format, RepositoryFormatPackGCPlain)
3386
# or isinstance(target._format, RepositoryFormatPackGCPlain)):
3388
return (InterRepository._same_model(source, target) and
3389
not source._format.supports_chks)
3392
def fetch(self, revision_id=None, pb=None, find_ghosts=False,
3394
"""See InterRepository.fetch()."""
3395
if (len(self.source._fallback_repositories) > 0 or
3396
len(self.target._fallback_repositories) > 0):
3397
# The pack layer is not aware of fallback repositories, so when
3398
# fetching from a stacked repository or into a stacked repository
3399
# we use the generic fetch logic which uses the VersionedFiles
3400
# attributes on repository.
3401
from bzrlib.fetch import RepoFetcher
3402
fetcher = RepoFetcher(self.target, self.source, revision_id,
3403
pb, find_ghosts, fetch_spec=fetch_spec)
3404
if fetch_spec is not None:
3405
if len(list(fetch_spec.heads)) != 1:
3406
raise AssertionError(
3407
"InterPackRepo.fetch doesn't support "
3408
"fetching multiple heads yet.")
3409
revision_id = list(fetch_spec.heads)[0]
3411
if revision_id is None:
3413
# everything to do - use pack logic
3414
# to fetch from all packs to one without
3415
# inventory parsing etc, IFF nothing to be copied is in the target.
3417
source_revision_ids = frozenset(self.source.all_revision_ids())
3418
revision_ids = source_revision_ids - \
3419
frozenset(self.target.get_parent_map(source_revision_ids))
3420
revision_keys = [(revid,) for revid in revision_ids]
3421
index = self.target._pack_collection.revision_index.combined_index
3422
present_revision_ids = set(item[1][0] for item in
3423
index.iter_entries(revision_keys))
3424
revision_ids = set(revision_ids) - present_revision_ids
3425
# implementing the TODO will involve:
3426
# - detecting when all of a pack is selected
3427
# - avoiding as much as possible pre-selection, so the
3428
# more-core routines such as create_pack_from_packs can filter in
3429
# a just-in-time fashion. (though having a HEADS list on a
3430
# repository might make this a lot easier, because we could
3431
# sensibly detect 'new revisions' without doing a full index scan.
3432
elif _mod_revision.is_null(revision_id):
3437
revision_ids = self.search_missing_revision_ids(revision_id,
3438
find_ghosts=find_ghosts).get_keys()
3439
except errors.NoSuchRevision:
3440
raise errors.InstallFailed([revision_id])
3441
if len(revision_ids) == 0:
3443
return self._pack(self.source, self.target, revision_ids)
3445
def _pack(self, source, target, revision_ids):
3446
from bzrlib.repofmt.pack_repo import Packer
3447
packs = source._pack_collection.all_packs()
3448
pack = Packer(self.target._pack_collection, packs, '.fetch',
3449
revision_ids).pack()
3450
if pack is not None:
3451
self.target._pack_collection._save_pack_names()
3452
copied_revs = pack.get_revision_count()
3453
# Trigger an autopack. This may duplicate effort as we've just done
3454
# a pack creation, but for now it is simpler to think about as
3455
# 'upload data, then repack if needed'.
3456
self.target._pack_collection.autopack()
3457
return (copied_revs, [])
3462
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
3463
"""See InterRepository.missing_revision_ids().
3465
:param find_ghosts: Find ghosts throughout the ancestry of
3468
if not find_ghosts and revision_id is not None:
3469
return self._walk_to_common_revisions([revision_id])
3470
elif revision_id is not None:
3471
# Find ghosts: search for revisions pointing from one repository to
3472
# the other, and vice versa, anywhere in the history of revision_id.
3473
graph = self.target.get_graph(other_repository=self.source)
3474
searcher = graph._make_breadth_first_searcher([revision_id])
3478
next_revs, ghosts = searcher.next_with_ghosts()
3479
except StopIteration:
3481
if revision_id in ghosts:
3482
raise errors.NoSuchRevision(self.source, revision_id)
3483
found_ids.update(next_revs)
3484
found_ids.update(ghosts)
3485
found_ids = frozenset(found_ids)
3486
# Double query here: should be able to avoid this by changing the
3487
# graph api further.
3488
result_set = found_ids - frozenset(
3489
self.target.get_parent_map(found_ids))
3491
source_ids = self.source.all_revision_ids()
3492
# source_ids is the worst possible case we may need to pull.
3493
# now we want to filter source_ids against what we actually
3494
# have in target, but don't try to check for existence where we know
3495
# we do not have a revision as that would be pointless.
3496
target_ids = set(self.target.all_revision_ids())
3497
result_set = set(source_ids).difference(target_ids)
3498
return self.source.revision_ids_to_search_result(result_set)
3501
class InterDifferingSerializer(InterKnitRepo):
3743
class InterDifferingSerializer(InterRepository):
3504
3746
def _get_repo_format_to_test(self):
3508
3750
def is_compatible(source, target):
3509
3751
"""Be compatible with Knit2 source and Knit3 target"""
3510
# XXX: What do we need to do to support fetching them?
3511
# if source.supports_rich_root() != target.supports_rich_root():
3513
# Ideally, we'd support fetching if the source had no tree references
3514
# even if it supported them...
3515
# XXX: What do we need to do to support fetching them?
3516
# if (getattr(source._format, 'supports_tree_reference', False) and
3517
# not getattr(target._format, 'supports_tree_reference', False)):
3752
# This is redundant with format.check_conversion_target(), however that
3753
# raises an exception, and we just want to say "False" as in we won't
3754
# support converting between these formats.
3755
if 'IDS_never' in debug.debug_flags:
3757
if source.supports_rich_root() and not target.supports_rich_root():
3759
if (source._format.supports_tree_reference
3760
and not target._format.supports_tree_reference):
3762
if target._fallback_repositories and target._format.supports_chks:
3763
# IDS doesn't know how to copy CHKs for the parent inventories it
3764
# adds to stacked repos.
3766
if 'IDS_always' in debug.debug_flags:
3768
# Only use this code path for local source and target. IDS does far
3769
# too much IO (both bandwidth and roundtrips) over a network.
3770
if not source.bzrdir.transport.base.startswith('file:///'):
3772
if not target.bzrdir.transport.base.startswith('file:///'):
3521
def _get_delta_for_revision(self, tree, parent_ids, basis_id, cache):
3776
def _get_trees(self, revision_ids, cache):
3778
for rev_id in revision_ids:
3780
possible_trees.append((rev_id, cache[rev_id]))
3782
# Not cached, but inventory might be present anyway.
3784
tree = self.source.revision_tree(rev_id)
3785
except errors.NoSuchRevision:
3786
# Nope, parent is ghost.
3789
cache[rev_id] = tree
3790
possible_trees.append((rev_id, tree))
3791
return possible_trees
3793
def _get_delta_for_revision(self, tree, parent_ids, possible_trees):
3522
3794
"""Get the best delta and base for this revision.
3524
3796
:return: (basis_id, delta)
3526
possible_trees = [(parent_id, cache[parent_id])
3527
for parent_id in parent_ids
3528
if parent_id in cache]
3529
if len(possible_trees) == 0:
3530
# There either aren't any parents, or the parents aren't in the
3531
# cache, so just use the last converted tree
3532
possible_trees.append((basis_id, cache[basis_id]))
3799
# Generate deltas against each tree, to find the shortest.
3800
texts_possibly_new_in_tree = set()
3534
3801
for basis_id, basis_tree in possible_trees:
3535
3802
delta = tree.inventory._make_delta(basis_tree.inventory)
3803
for old_path, new_path, file_id, new_entry in delta:
3804
if new_path is None:
3805
# This file_id isn't present in the new rev, so we don't
3809
# Rich roots are handled elsewhere...
3811
kind = new_entry.kind
3812
if kind != 'directory' and kind != 'file':
3813
# No text record associated with this inventory entry.
3815
# This is a directory or file that has changed somehow.
3816
texts_possibly_new_in_tree.add((file_id, new_entry.revision))
3536
3817
deltas.append((len(delta), basis_id, delta))
3538
3819
return deltas[0][1:]
3540
def _fetch_batch(self, revision_ids, basis_id, cache):
3821
def _fetch_parent_invs_for_stacking(self, parent_map, cache):
3822
"""Find all parent revisions that are absent, but for which the
3823
inventory is present, and copy those inventories.
3825
This is necessary to preserve correctness when the source is stacked
3826
without fallbacks configured. (Note that in cases like upgrade the
3827
source may be not have _fallback_repositories even though it is
3831
for parents in parent_map.values():
3832
parent_revs.update(parents)
3833
present_parents = self.source.get_parent_map(parent_revs)
3834
absent_parents = set(parent_revs).difference(present_parents)
3835
parent_invs_keys_for_stacking = self.source.inventories.get_parent_map(
3836
(rev_id,) for rev_id in absent_parents)
3837
parent_inv_ids = [key[-1] for key in parent_invs_keys_for_stacking]
3838
for parent_tree in self.source.revision_trees(parent_inv_ids):
3839
current_revision_id = parent_tree.get_revision_id()
3840
parents_parents_keys = parent_invs_keys_for_stacking[
3841
(current_revision_id,)]
3842
parents_parents = [key[-1] for key in parents_parents_keys]
3843
basis_id = _mod_revision.NULL_REVISION
3844
basis_tree = self.source.revision_tree(basis_id)
3845
delta = parent_tree.inventory._make_delta(basis_tree.inventory)
3846
self.target.add_inventory_by_delta(
3847
basis_id, delta, current_revision_id, parents_parents)
3848
cache[current_revision_id] = parent_tree
3850
def _fetch_batch(self, revision_ids, basis_id, cache, a_graph=None):
3541
3851
"""Fetch across a few revisions.
3543
3853
:param revision_ids: The revisions to copy
3544
3854
:param basis_id: The revision_id of a tree that must be in cache, used
3545
3855
as a basis for delta when no other base is available
3546
3856
:param cache: A cache of RevisionTrees that we can use.
3857
:param a_graph: A Graph object to determine the heads() of the
3858
rich-root data stream.
3547
3859
:return: The revision_id of the last converted tree. The RevisionTree
3548
3860
for it will be in cache
3555
3867
pending_deltas = []
3556
3868
pending_revisions = []
3557
3869
parent_map = self.source.get_parent_map(revision_ids)
3558
# NB: This fails with dubious inventory data (when inv A has rev OLD
3559
# for file F, and in B, after A, has rev A for file F) when A and B are
3560
# in different groups.
3870
self._fetch_parent_invs_for_stacking(parent_map, cache)
3871
self.source._safe_to_return_from_cache = True
3561
3872
for tree in self.source.revision_trees(revision_ids):
3873
# Find a inventory delta for this revision.
3874
# Find text entries that need to be copied, too.
3562
3875
current_revision_id = tree.get_revision_id()
3563
3876
parent_ids = parent_map.get(current_revision_id, ())
3877
parent_trees = self._get_trees(parent_ids, cache)
3878
possible_trees = list(parent_trees)
3879
if len(possible_trees) == 0:
3880
# There either aren't any parents, or the parents are ghosts,
3881
# so just use the last converted tree.
3882
possible_trees.append((basis_id, cache[basis_id]))
3564
3883
basis_id, delta = self._get_delta_for_revision(tree, parent_ids,
3885
revision = self.source.get_revision(current_revision_id)
3886
pending_deltas.append((basis_id, delta,
3887
current_revision_id, revision.parent_ids))
3566
3888
if self._converting_to_rich_root:
3567
3889
self._revision_id_to_root_id[current_revision_id] = \
3568
3890
tree.get_root_id()
3569
# Find text entries that need to be copied
3891
# Determine which texts are in present in this revision but not in
3892
# any of the available parents.
3893
texts_possibly_new_in_tree = set()
3570
3894
for old_path, new_path, file_id, entry in delta:
3571
if new_path is not None:
3574
if not self.target.supports_rich_root():
3575
# The target doesn't support rich root, so we don't
3578
if self._converting_to_rich_root:
3579
# This can't be copied normally, we have to insert
3581
root_keys_to_create.add((file_id, entry.revision))
3583
text_keys.add((file_id, entry.revision))
3584
revision = self.source.get_revision(current_revision_id)
3585
pending_deltas.append((basis_id, delta,
3586
current_revision_id, revision.parent_ids))
3895
if new_path is None:
3896
# This file_id isn't present in the new rev
3900
if not self.target.supports_rich_root():
3901
# The target doesn't support rich root, so we don't
3904
if self._converting_to_rich_root:
3905
# This can't be copied normally, we have to insert
3907
root_keys_to_create.add((file_id, entry.revision))
3910
texts_possibly_new_in_tree.add((file_id, entry.revision))
3911
for basis_id, basis_tree in possible_trees:
3912
basis_inv = basis_tree.inventory
3913
for file_key in list(texts_possibly_new_in_tree):
3914
file_id, file_revision = file_key
3916
entry = basis_inv[file_id]
3917
except errors.NoSuchId:
3919
if entry.revision == file_revision:
3920
texts_possibly_new_in_tree.remove(file_key)
3921
text_keys.update(texts_possibly_new_in_tree)
3587
3922
pending_revisions.append(revision)
3588
3923
cache[current_revision_id] = tree
3589
3924
basis_id = current_revision_id
3925
self.source._safe_to_return_from_cache = False
3590
3926
# Copy file texts
3591
3927
from_texts = self.source.texts
3592
3928
to_texts = self.target.texts
3593
3929
if root_keys_to_create:
3594
NULL_REVISION = _mod_revision.NULL_REVISION
3595
def _get_parent_keys(root_key):
3596
root_id, rev_id = root_key
3597
# Include direct parents of the revision, but only if they used
3599
parent_keys = tuple([(root_id, parent_id)
3600
for parent_id in parent_map[rev_id]
3601
if parent_id != NULL_REVISION and
3602
self._revision_id_to_root_id.get(parent_id, root_id) == root_id])
3604
def new_root_data_stream():
3605
for root_key in root_keys_to_create:
3606
parent_keys = _get_parent_keys(root_key)
3607
yield versionedfile.FulltextContentFactory(root_key,
3608
parent_keys, None, '')
3609
to_texts.insert_record_stream(new_root_data_stream())
3930
root_stream = _mod_fetch._new_root_data_stream(
3931
root_keys_to_create, self._revision_id_to_root_id, parent_map,
3932
self.source, graph=a_graph)
3933
to_texts.insert_record_stream(root_stream)
3610
3934
to_texts.insert_record_stream(from_texts.get_record_stream(
3611
3935
text_keys, self.target._format._fetch_order,
3612
3936
not self.target._format._fetch_uses_deltas))
3937
# insert inventory deltas
3614
3938
for delta in pending_deltas:
3615
3939
self.target.add_inventory_by_delta(*delta)
3940
if self.target._fallback_repositories:
3941
# Make sure this stacked repository has all the parent inventories
3942
# for the new revisions that we are about to insert. We do this
3943
# before adding the revisions so that no revision is added until
3944
# all the inventories it may depend on are added.
3945
# Note that this is overzealous, as we may have fetched these in an
3948
revision_ids = set()
3949
for revision in pending_revisions:
3950
revision_ids.add(revision.revision_id)
3951
parent_ids.update(revision.parent_ids)
3952
parent_ids.difference_update(revision_ids)
3953
parent_ids.discard(_mod_revision.NULL_REVISION)
3954
parent_map = self.source.get_parent_map(parent_ids)
3955
# we iterate over parent_map and not parent_ids because we don't
3956
# want to try copying any revision which is a ghost
3957
for parent_tree in self.source.revision_trees(parent_map):
3958
current_revision_id = parent_tree.get_revision_id()
3959
parents_parents = parent_map[current_revision_id]
3960
possible_trees = self._get_trees(parents_parents, cache)
3961
if len(possible_trees) == 0:
3962
# There either aren't any parents, or the parents are
3963
# ghosts, so just use the last converted tree.
3964
possible_trees.append((basis_id, cache[basis_id]))
3965
basis_id, delta = self._get_delta_for_revision(parent_tree,
3966
parents_parents, possible_trees)
3967
self.target.add_inventory_by_delta(
3968
basis_id, delta, current_revision_id, parents_parents)
3616
3969
# insert signatures and revisions
3617
3970
for revision in pending_revisions:
3985
4384
# missing keys can handle suspending a write group).
3986
4385
write_group_tokens = self.target_repo.suspend_write_group()
3987
4386
return write_group_tokens, missing_keys
3988
self.target_repo.commit_write_group()
4387
hint = self.target_repo.commit_write_group()
4388
if (to_serializer != src_serializer and
4389
self.target_repo._format.pack_compresses):
4390
self.target_repo.pack(hint=hint)
3989
4391
return [], set()
3991
def _extract_and_insert_inventories(self, substream, serializer):
4393
def _extract_and_insert_inventory_deltas(self, substream, serializer):
4394
target_rich_root = self.target_repo._format.rich_root_data
4395
target_tree_refs = self.target_repo._format.supports_tree_reference
4396
for record in substream:
4397
# Insert the delta directly
4398
inventory_delta_bytes = record.get_bytes_as('fulltext')
4399
deserialiser = inventory_delta.InventoryDeltaDeserializer()
4401
parse_result = deserialiser.parse_text_bytes(
4402
inventory_delta_bytes)
4403
except inventory_delta.IncompatibleInventoryDelta, err:
4404
trace.mutter("Incompatible delta: %s", err.msg)
4405
raise errors.IncompatibleRevision(self.target_repo._format)
4406
basis_id, new_id, rich_root, tree_refs, inv_delta = parse_result
4407
revision_id = new_id
4408
parents = [key[0] for key in record.parents]
4409
self.target_repo.add_inventory_by_delta(
4410
basis_id, inv_delta, revision_id, parents)
4412
def _extract_and_insert_inventories(self, substream, serializer,
3992
4414
"""Generate a new inventory versionedfile in target, converting data.
3994
4416
The inventory is retrieved from the source, (deserializing it), and
3995
4417
stored in the target (reserializing it in a different format).
4419
target_rich_root = self.target_repo._format.rich_root_data
4420
target_tree_refs = self.target_repo._format.supports_tree_reference
3997
4421
for record in substream:
4422
# It's not a delta, so it must be a fulltext in the source
4423
# serializer's format.
3998
4424
bytes = record.get_bytes_as('fulltext')
3999
4425
revision_id = record.key[0]
4000
4426
inv = serializer.read_inventory_from_string(bytes, revision_id)
4001
4427
parents = [key[0] for key in record.parents]
4002
4428
self.target_repo.add_inventory(revision_id, inv, parents)
4429
# No need to keep holding this full inv in memory when the rest of
4430
# the substream is likely to be all deltas.
4004
4433
def _extract_and_insert_revisions(self, substream, serializer):
4005
4434
for record in substream:
4151
4597
return (not self.from_repository._format.rich_root_data and
4152
4598
self.to_format.rich_root_data)
4154
def _get_inventory_stream(self, revision_ids):
4600
def _get_inventory_stream(self, revision_ids, missing=False):
4155
4601
from_format = self.from_repository._format
4156
if (from_format.supports_chks and self.to_format.supports_chks
4157
and (from_format._serializer == self.to_format._serializer)):
4158
# Both sides support chks, and they use the same serializer, so it
4159
# is safe to transmit the chk pages and inventory pages across
4161
return self._get_chk_inventory_stream(revision_ids)
4162
elif (not from_format.supports_chks):
4163
# Source repository doesn't support chks. So we can transmit the
4164
# inventories 'as-is' and either they are just accepted on the
4165
# target, or the Sink will properly convert it.
4166
return self._get_simple_inventory_stream(revision_ids)
4602
if (from_format.supports_chks and self.to_format.supports_chks and
4603
from_format.network_name() == self.to_format.network_name()):
4604
raise AssertionError(
4605
"this case should be handled by GroupCHKStreamSource")
4606
elif 'forceinvdeltas' in debug.debug_flags:
4607
return self._get_convertable_inventory_stream(revision_ids,
4608
delta_versus_null=missing)
4609
elif from_format.network_name() == self.to_format.network_name():
4611
return self._get_simple_inventory_stream(revision_ids,
4613
elif (not from_format.supports_chks and not self.to_format.supports_chks
4614
and from_format._serializer == self.to_format._serializer):
4615
# Essentially the same format.
4616
return self._get_simple_inventory_stream(revision_ids,
4168
# XXX: Hack to make not-chk->chk fetch: copy the inventories as
4169
# inventories. Note that this should probably be done somehow
4170
# as part of bzrlib.repository.StreamSink. Except JAM couldn't
4171
# figure out how a non-chk repository could possibly handle
4172
# deserializing an inventory stream from a chk repo, as it
4173
# doesn't have a way to understand individual pages.
4174
return self._get_convertable_inventory_stream(revision_ids)
4619
# Any time we switch serializations, we want to use an
4620
# inventory-delta based approach.
4621
return self._get_convertable_inventory_stream(revision_ids,
4622
delta_versus_null=missing)
4176
def _get_simple_inventory_stream(self, revision_ids):
4624
def _get_simple_inventory_stream(self, revision_ids, missing=False):
4625
# NB: This currently reopens the inventory weave in source;
4626
# using a single stream interface instead would avoid this.
4177
4627
from_weave = self.from_repository.inventories
4629
delta_closure = True
4631
delta_closure = not self.delta_on_metadata()
4178
4632
yield ('inventories', from_weave.get_record_stream(
4179
4633
[(rev_id,) for rev_id in revision_ids],
4180
self.inventory_fetch_order(),
4181
not self.delta_on_metadata()))
4183
def _get_chk_inventory_stream(self, revision_ids):
4184
"""Fetch the inventory texts, along with the associated chk maps."""
4185
# We want an inventory outside of the search set, so that we can filter
4186
# out uninteresting chk pages. For now we use
4187
# _find_revision_outside_set, but if we had a Search with cut_revs, we
4188
# could use that instead.
4189
start_rev_id = self.from_repository._find_revision_outside_set(
4191
start_rev_key = (start_rev_id,)
4192
inv_keys_to_fetch = [(rev_id,) for rev_id in revision_ids]
4193
if start_rev_id != _mod_revision.NULL_REVISION:
4194
inv_keys_to_fetch.append((start_rev_id,))
4195
# Any repo that supports chk_bytes must also support out-of-order
4196
# insertion. At least, that is how we expect it to work
4197
# We use get_record_stream instead of iter_inventories because we want
4198
# to be able to insert the stream as well. We could instead fetch
4199
# allowing deltas, and then iter_inventories, but we don't know whether
4200
# source or target is more 'local' anway.
4201
inv_stream = self.from_repository.inventories.get_record_stream(
4202
inv_keys_to_fetch, 'unordered',
4203
True) # We need them as full-texts so we can find their references
4204
uninteresting_chk_roots = set()
4205
interesting_chk_roots = set()
4206
def filter_inv_stream(inv_stream):
4207
for idx, record in enumerate(inv_stream):
4208
### child_pb.update('fetch inv', idx, len(inv_keys_to_fetch))
4209
bytes = record.get_bytes_as('fulltext')
4210
chk_inv = inventory.CHKInventory.deserialise(
4211
self.from_repository.chk_bytes, bytes, record.key)
4212
if record.key == start_rev_key:
4213
uninteresting_chk_roots.add(chk_inv.id_to_entry.key())
4214
p_id_map = chk_inv.parent_id_basename_to_file_id
4215
if p_id_map is not None:
4216
uninteresting_chk_roots.add(p_id_map.key())
4219
interesting_chk_roots.add(chk_inv.id_to_entry.key())
4220
p_id_map = chk_inv.parent_id_basename_to_file_id
4221
if p_id_map is not None:
4222
interesting_chk_roots.add(p_id_map.key())
4223
### pb.update('fetch inventory', 0, 2)
4224
yield ('inventories', filter_inv_stream(inv_stream))
4225
# Now that we have worked out all of the interesting root nodes, grab
4226
# all of the interesting pages and insert them
4227
### pb.update('fetch inventory', 1, 2)
4228
interesting = chk_map.iter_interesting_nodes(
4229
self.from_repository.chk_bytes, interesting_chk_roots,
4230
uninteresting_chk_roots)
4231
def to_stream_adapter():
4232
"""Adapt the iter_interesting_nodes result to a single stream.
4234
iter_interesting_nodes returns records as it processes them, which
4235
can be in batches. But we only want a single stream to be inserted.
4237
for record, items in interesting:
4238
for value in record.itervalues():
4240
# XXX: We could instead call get_record_stream(records.keys())
4241
# ATM, this will always insert the records as fulltexts, and
4242
# requires that you can hang on to records once you have gone
4243
# on to the next one. Further, it causes the target to
4244
# recompress the data. Testing shows it to be faster than
4245
# requesting the records again, though.
4246
yield ('chk_bytes', to_stream_adapter())
4247
### pb.update('fetch inventory', 2, 2)
4249
def _get_convertable_inventory_stream(self, revision_ids):
4250
# XXX: One of source or target is using chks, and they don't have
4251
# compatible serializations. The StreamSink code expects to be
4252
# able to convert on the target, so we need to put
4253
# bytes-on-the-wire that can be converted
4254
yield ('inventories', self._stream_invs_as_fulltexts(revision_ids))
4256
def _stream_invs_as_fulltexts(self, revision_ids):
4634
self.inventory_fetch_order(), delta_closure))
4636
def _get_convertable_inventory_stream(self, revision_ids,
4637
delta_versus_null=False):
4638
# The two formats are sufficiently different that there is no fast
4639
# path, so we need to send just inventorydeltas, which any
4640
# sufficiently modern client can insert into any repository.
4641
# The StreamSink code expects to be able to
4642
# convert on the target, so we need to put bytes-on-the-wire that can
4643
# be converted. That means inventory deltas (if the remote is <1.19,
4644
# RemoteStreamSink will fallback to VFS to insert the deltas).
4645
yield ('inventory-deltas',
4646
self._stream_invs_as_deltas(revision_ids,
4647
delta_versus_null=delta_versus_null))
4649
def _stream_invs_as_deltas(self, revision_ids, delta_versus_null=False):
4650
"""Return a stream of inventory-deltas for the given rev ids.
4652
:param revision_ids: The list of inventories to transmit
4653
:param delta_versus_null: Don't try to find a minimal delta for this
4654
entry, instead compute the delta versus the NULL_REVISION. This
4655
effectively streams a complete inventory. Used for stuff like
4656
filling in missing parents, etc.
4257
4658
from_repo = self.from_repository
4258
from_serializer = from_repo._format._serializer
4259
4659
revision_keys = [(rev_id,) for rev_id in revision_ids]
4260
4660
parent_map = from_repo.inventories.get_parent_map(revision_keys)
4261
for inv in self.from_repository.iter_inventories(revision_ids):
4262
# XXX: This is a bit hackish, but it works. Basically,
4263
# CHKSerializer 'accidentally' supports
4264
# read/write_inventory_to_string, even though that is never
4265
# the format that is stored on disk. It *does* give us a
4266
# single string representation for an inventory, so live with
4268
# This would be far better if we had a 'serialized inventory
4269
# delta' form. Then we could use 'inventory._make_delta', and
4270
# transmit that. This would both be faster to generate, and
4271
# result in fewer bytes-on-the-wire.
4272
as_bytes = from_serializer.write_inventory_to_string(inv)
4661
# XXX: possibly repos could implement a more efficient iter_inv_deltas
4663
inventories = self.from_repository.iter_inventories(
4664
revision_ids, 'topological')
4665
format = from_repo._format
4666
invs_sent_so_far = set([_mod_revision.NULL_REVISION])
4667
inventory_cache = lru_cache.LRUCache(50)
4668
null_inventory = from_repo.revision_tree(
4669
_mod_revision.NULL_REVISION).inventory
4670
# XXX: ideally the rich-root/tree-refs flags would be per-revision, not
4671
# per-repo (e.g. streaming a non-rich-root revision out of a rich-root
4672
# repo back into a non-rich-root repo ought to be allowed)
4673
serializer = inventory_delta.InventoryDeltaSerializer(
4674
versioned_root=format.rich_root_data,
4675
tree_references=format.supports_tree_reference)
4676
for inv in inventories:
4273
4677
key = (inv.revision_id,)
4274
4678
parent_keys = parent_map.get(key, ())
4680
if not delta_versus_null and parent_keys:
4681
# The caller did not ask for complete inventories and we have
4682
# some parents that we can delta against. Make a delta against
4683
# each parent so that we can find the smallest.
4684
parent_ids = [parent_key[0] for parent_key in parent_keys]
4685
for parent_id in parent_ids:
4686
if parent_id not in invs_sent_so_far:
4687
# We don't know that the remote side has this basis, so
4690
if parent_id == _mod_revision.NULL_REVISION:
4691
parent_inv = null_inventory
4693
parent_inv = inventory_cache.get(parent_id, None)
4694
if parent_inv is None:
4695
parent_inv = from_repo.get_inventory(parent_id)
4696
candidate_delta = inv._make_delta(parent_inv)
4697
if (delta is None or
4698
len(delta) > len(candidate_delta)):
4699
delta = candidate_delta
4700
basis_id = parent_id
4702
# Either none of the parents ended up being suitable, or we
4703
# were asked to delta against NULL
4704
basis_id = _mod_revision.NULL_REVISION
4705
delta = inv._make_delta(null_inventory)
4706
invs_sent_so_far.add(inv.revision_id)
4707
inventory_cache[inv.revision_id] = inv
4708
delta_serialized = ''.join(
4709
serializer.delta_to_lines(basis_id, key[-1], delta))
4275
4710
yield versionedfile.FulltextContentFactory(
4276
key, parent_keys, None, as_bytes)
4711
key, parent_keys, None, delta_serialized)
4714
def _iter_for_revno(repo, partial_history_cache, stop_index=None,
4715
stop_revision=None):
4716
"""Extend the partial history to include a given index
4718
If a stop_index is supplied, stop when that index has been reached.
4719
If a stop_revision is supplied, stop when that revision is
4720
encountered. Otherwise, stop when the beginning of history is
4723
:param stop_index: The index which should be present. When it is
4724
present, history extension will stop.
4725
:param stop_revision: The revision id which should be present. When
4726
it is encountered, history extension will stop.
4728
start_revision = partial_history_cache[-1]
4729
iterator = repo.iter_reverse_revision_history(start_revision)
4731
#skip the last revision in the list
4734
if (stop_index is not None and
4735
len(partial_history_cache) > stop_index):
4737
if partial_history_cache[-1] == stop_revision:
4739
revision_id = iterator.next()
4740
partial_history_cache.append(revision_id)
4741
except StopIteration:
added
removed
bzrlib/repository.py
