1181
1150
# The old API returned a list, should this actually be a set?
1182
1151
return parent_map.keys()
1184
def _check_inventories(self, checker):
1185
"""Check the inventories found from the revision scan.
1187
This is responsible for verifying the sha1 of inventories and
1188
creating a pending_keys set that covers data referenced by inventories.
1190
bar = ui.ui_factory.nested_progress_bar()
1192
self._do_check_inventories(checker, bar)
1196
def _do_check_inventories(self, checker, bar):
1197
"""Helper for _check_inventories."""
1199
keys = {'chk_bytes':set(), 'inventories':set(), 'texts':set()}
1200
kinds = ['chk_bytes', 'texts']
1201
count = len(checker.pending_keys)
1202
bar.update("inventories", 0, 2)
1203
current_keys = checker.pending_keys
1204
checker.pending_keys = {}
1205
# Accumulate current checks.
1206
for key in current_keys:
1207
if key[0] != 'inventories' and key[0] not in kinds:
1208
checker._report_items.append('unknown key type %r' % (key,))
1209
keys[key[0]].add(key[1:])
1210
if keys['inventories']:
1211
# NB: output order *should* be roughly sorted - topo or
1212
# inverse topo depending on repository - either way decent
1213
# to just delta against. However, pre-CHK formats didn't
1214
# try to optimise inventory layout on disk. As such the
1215
# pre-CHK code path does not use inventory deltas.
1217
for record in self.inventories.check(keys=keys['inventories']):
1218
if record.storage_kind == 'absent':
1219
checker._report_items.append(
1220
'Missing inventory {%s}' % (record.key,))
1222
last_object = self._check_record('inventories', record,
1223
checker, last_object,
1224
current_keys[('inventories',) + record.key])
1225
del keys['inventories']
1228
bar.update("texts", 1)
1229
while (checker.pending_keys or keys['chk_bytes']
1231
# Something to check.
1232
current_keys = checker.pending_keys
1233
checker.pending_keys = {}
1234
# Accumulate current checks.
1235
for key in current_keys:
1236
if key[0] not in kinds:
1237
checker._report_items.append('unknown key type %r' % (key,))
1238
keys[key[0]].add(key[1:])
1239
# Check the outermost kind only - inventories || chk_bytes || texts
1243
for record in getattr(self, kind).check(keys=keys[kind]):
1244
if record.storage_kind == 'absent':
1245
checker._report_items.append(
1246
'Missing %s {%s}' % (kind, record.key,))
1248
last_object = self._check_record(kind, record,
1249
checker, last_object, current_keys[(kind,) + record.key])
1253
def _check_record(self, kind, record, checker, last_object, item_data):
1254
"""Check a single text from this repository."""
1255
if kind == 'inventories':
1256
rev_id = record.key[0]
1257
inv = self._deserialise_inventory(rev_id,
1258
record.get_bytes_as('fulltext'))
1259
if last_object is not None:
1260
delta = inv._make_delta(last_object)
1261
for old_path, path, file_id, ie in delta:
1264
ie.check(checker, rev_id, inv)
1266
for path, ie in inv.iter_entries():
1267
ie.check(checker, rev_id, inv)
1268
if self._format.fast_deltas:
1270
elif kind == 'chk_bytes':
1271
# No code written to check chk_bytes for this repo format.
1272
checker._report_items.append(
1273
'unsupported key type chk_bytes for %s' % (record.key,))
1274
elif kind == 'texts':
1275
self._check_text(record, checker, item_data)
1277
checker._report_items.append(
1278
'unknown key type %s for %s' % (kind, record.key))
1280
def _check_text(self, record, checker, item_data):
1281
"""Check a single text."""
1282
# Check it is extractable.
1283
# TODO: check length.
1284
if record.storage_kind == 'chunked':
1285
chunks = record.get_bytes_as(record.storage_kind)
1286
sha1 = osutils.sha_strings(chunks)
1287
length = sum(map(len, chunks))
1289
content = record.get_bytes_as('fulltext')
1290
sha1 = osutils.sha_string(content)
1291
length = len(content)
1292
if item_data and sha1 != item_data[1]:
1293
checker._report_items.append(
1294
'sha1 mismatch: %s has sha1 %s expected %s referenced by %s' %
1295
(record.key, sha1, item_data[1], item_data[2]))
1298
1154
def create(a_bzrdir):
1299
1155
"""Construct the current default format repository in a_bzrdir."""
1598
1424
def suspend_write_group(self):
1599
1425
raise errors.UnsuspendableWriteGroup(self)
1601
def get_missing_parent_inventories(self, check_for_missing_texts=True):
1602
"""Return the keys of missing inventory parents for revisions added in
1605
A revision is not complete if the inventory delta for that revision
1606
cannot be calculated. Therefore if the parent inventories of a
1607
revision are not present, the revision is incomplete, and e.g. cannot
1608
be streamed by a smart server. This method finds missing inventory
1609
parents for revisions added in this write group.
1611
if not self._format.supports_external_lookups:
1612
# This is only an issue for stacked repositories
1614
if not self.is_in_write_group():
1615
raise AssertionError('not in a write group')
1617
# XXX: We assume that every added revision already has its
1618
# corresponding inventory, so we only check for parent inventories that
1619
# might be missing, rather than all inventories.
1620
parents = set(self.revisions._index.get_missing_parents())
1621
parents.discard(_mod_revision.NULL_REVISION)
1622
unstacked_inventories = self.inventories._index
1623
present_inventories = unstacked_inventories.get_parent_map(
1624
key[-1:] for key in parents)
1625
parents.difference_update(present_inventories)
1626
if len(parents) == 0:
1627
# No missing parent inventories.
1629
if not check_for_missing_texts:
1630
return set(('inventories', rev_id) for (rev_id,) in parents)
1631
# Ok, now we have a list of missing inventories. But these only matter
1632
# if the inventories that reference them are missing some texts they
1633
# appear to introduce.
1634
# XXX: Texts referenced by all added inventories need to be present,
1635
# but at the moment we're only checking for texts referenced by
1636
# inventories at the graph's edge.
1637
key_deps = self.revisions._index._key_dependencies
1638
key_deps.satisfy_refs_for_keys(present_inventories)
1639
referrers = frozenset(r[0] for r in key_deps.get_referrers())
1640
file_ids = self.fileids_altered_by_revision_ids(referrers)
1641
missing_texts = set()
1642
for file_id, version_ids in file_ids.iteritems():
1643
missing_texts.update(
1644
(file_id, version_id) for version_id in version_ids)
1645
present_texts = self.texts.get_parent_map(missing_texts)
1646
missing_texts.difference_update(present_texts)
1647
if not missing_texts:
1648
# No texts are missing, so all revisions and their deltas are
1651
# Alternatively the text versions could be returned as the missing
1652
# keys, but this is likely to be less data.
1653
missing_keys = set(('inventories', rev_id) for (rev_id,) in parents)
1656
1427
def refresh_data(self):
1657
1428
"""Re-read any data needed to to synchronise with disk.
1880
1642
@needs_read_lock
1881
1643
def get_revisions(self, revision_ids):
1882
"""Get many revisions at once.
1884
Repositories that need to check data on every revision read should
1885
subclass this method.
1644
"""Get many revisions at once."""
1887
1645
return self._get_revisions(revision_ids)
1889
1647
@needs_read_lock
1890
1648
def _get_revisions(self, revision_ids):
1891
1649
"""Core work logic to get many revisions without sanity checks."""
1650
for rev_id in revision_ids:
1651
if not rev_id or not isinstance(rev_id, basestring):
1652
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1653
keys = [(key,) for key in revision_ids]
1654
stream = self.revisions.get_record_stream(keys, 'unordered', True)
1893
for revid, rev in self._iter_revisions(revision_ids):
1895
raise errors.NoSuchRevision(self, revid)
1656
for record in stream:
1657
if record.storage_kind == 'absent':
1658
raise errors.NoSuchRevision(self, record.key[0])
1659
text = record.get_bytes_as('fulltext')
1660
rev = self._serializer.read_revision_from_string(text)
1661
revs[record.key[0]] = rev
1897
1662
return [revs[revid] for revid in revision_ids]
1899
def _iter_revisions(self, revision_ids):
1900
"""Iterate over revision objects.
1902
:param revision_ids: An iterable of revisions to examine. None may be
1903
passed to request all revisions known to the repository. Note that
1904
not all repositories can find unreferenced revisions; for those
1905
repositories only referenced ones will be returned.
1906
:return: An iterator of (revid, revision) tuples. Absent revisions (
1907
those asked for but not available) are returned as (revid, None).
1909
if revision_ids is None:
1910
revision_ids = self.all_revision_ids()
1912
for rev_id in revision_ids:
1913
if not rev_id or not isinstance(rev_id, basestring):
1914
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1915
keys = [(key,) for key in revision_ids]
1916
stream = self.revisions.get_record_stream(keys, 'unordered', True)
1917
for record in stream:
1918
revid = record.key[0]
1919
if record.storage_kind == 'absent':
1922
text = record.get_bytes_as('fulltext')
1923
rev = self._serializer.read_revision_from_string(text)
1665
def get_revision_xml(self, revision_id):
1666
# TODO: jam 20070210 This shouldn't be necessary since get_revision
1667
# would have already do it.
1668
# TODO: jam 20070210 Just use _serializer.write_revision_to_string()
1669
# TODO: this can't just be replaced by:
1670
# return self._serializer.write_revision_to_string(
1671
# self.get_revision(revision_id))
1672
# as cStringIO preservers the encoding unlike write_revision_to_string
1673
# or some other call down the path.
1674
rev = self.get_revision(revision_id)
1675
rev_tmp = cStringIO.StringIO()
1676
# the current serializer..
1677
self._serializer.write_revision(rev, rev_tmp)
1679
return rev_tmp.getvalue()
1926
1681
def get_deltas_for_revisions(self, revisions, specific_fileids=None):
1927
1682
"""Produce a generator of revision deltas.
2377
2114
:param revision_ids: The expected revision ids of the inventories.
2378
:param ordering: optional ordering, e.g. 'topological'. If not
2379
specified, the order of revision_ids will be preserved (by
2380
buffering if necessary).
2381
2115
:return: An iterator of inventories.
2383
2117
if ((None in revision_ids)
2384
2118
or (_mod_revision.NULL_REVISION in revision_ids)):
2385
2119
raise ValueError('cannot get null revision inventory')
2386
return self._iter_inventories(revision_ids, ordering)
2120
return self._iter_inventories(revision_ids)
2388
def _iter_inventories(self, revision_ids, ordering):
2122
def _iter_inventories(self, revision_ids):
2389
2123
"""single-document based inventory iteration."""
2390
inv_xmls = self._iter_inventory_xmls(revision_ids, ordering)
2391
for text, revision_id in inv_xmls:
2392
yield self._deserialise_inventory(revision_id, text)
2124
for text, revision_id in self._iter_inventory_xmls(revision_ids):
2125
yield self.deserialise_inventory(revision_id, text)
2394
def _iter_inventory_xmls(self, revision_ids, ordering):
2395
if ordering is None:
2396
order_as_requested = True
2397
ordering = 'unordered'
2399
order_as_requested = False
2127
def _iter_inventory_xmls(self, revision_ids):
2400
2128
keys = [(revision_id,) for revision_id in revision_ids]
2403
if order_as_requested:
2404
key_iter = iter(keys)
2405
next_key = key_iter.next()
2406
stream = self.inventories.get_record_stream(keys, ordering, True)
2129
stream = self.inventories.get_record_stream(keys, 'unordered', True)
2407
2130
text_chunks = {}
2408
2131
for record in stream:
2409
2132
if record.storage_kind != 'absent':
2410
chunks = record.get_bytes_as('chunked')
2411
if order_as_requested:
2412
text_chunks[record.key] = chunks
2414
yield ''.join(chunks), record.key[-1]
2133
text_chunks[record.key] = record.get_bytes_as('chunked')
2416
2135
raise errors.NoSuchRevision(self, record.key)
2417
if order_as_requested:
2418
# Yield as many results as we can while preserving order.
2419
while next_key in text_chunks:
2420
chunks = text_chunks.pop(next_key)
2421
yield ''.join(chunks), next_key[-1]
2423
next_key = key_iter.next()
2424
except StopIteration:
2425
# We still want to fully consume the get_record_stream,
2426
# just in case it is not actually finished at this point
2137
chunks = text_chunks.pop(key)
2138
yield ''.join(chunks), key[-1]
2430
def _deserialise_inventory(self, revision_id, xml):
2140
def deserialise_inventory(self, revision_id, xml):
2431
2141
"""Transform the xml into an inventory object.
2433
2143
:param revision_id: The expected revision id of the inventory.
2434
2144
:param xml: A serialised inventory.
2436
2146
result = self._serializer.read_inventory_from_string(xml, revision_id,
2437
entry_cache=self._inventory_entry_cache,
2438
return_from_cache=self._safe_to_return_from_cache)
2147
entry_cache=self._inventory_entry_cache)
2439
2148
if result.revision_id != revision_id:
2440
2149
raise AssertionError('revision id mismatch %s != %s' % (
2441
2150
result.revision_id, revision_id))
2153
def serialise_inventory(self, inv):
2154
return self._serializer.write_inventory_to_string(inv)
2156
def _serialise_inventory_to_lines(self, inv):
2157
return self._serializer.write_inventory_to_lines(inv)
2444
2159
def get_serializer_format(self):
2445
2160
return self._serializer.format_num
2447
2162
@needs_read_lock
2448
def _get_inventory_xml(self, revision_id):
2449
"""Get serialized inventory as a string."""
2450
texts = self._iter_inventory_xmls([revision_id], 'unordered')
2163
def get_inventory_xml(self, revision_id):
2164
"""Get inventory XML as a file object."""
2165
texts = self._iter_inventory_xmls([revision_id])
2452
2167
text, revision_id = texts.next()
2453
2168
except StopIteration:
2454
2169
raise errors.HistoryMissing(self, 'inventory', revision_id)
2457
def get_rev_id_for_revno(self, revno, known_pair):
2458
"""Return the revision id of a revno, given a later (revno, revid)
2459
pair in the same history.
2461
:return: if found (True, revid). If the available history ran out
2462
before reaching the revno, then this returns
2463
(False, (closest_revno, closest_revid)).
2173
def get_inventory_sha1(self, revision_id):
2174
"""Return the sha1 hash of the inventory entry
2465
known_revno, known_revid = known_pair
2466
partial_history = [known_revid]
2467
distance_from_known = known_revno - revno
2468
if distance_from_known < 0:
2470
'requested revno (%d) is later than given known revno (%d)'
2471
% (revno, known_revno))
2474
self, partial_history, stop_index=distance_from_known)
2475
except errors.RevisionNotPresent, err:
2476
if err.revision_id == known_revid:
2477
# The start revision (known_revid) wasn't found.
2479
# This is a stacked repository with no fallbacks, or a there's a
2480
# left-hand ghost. Either way, even though the revision named in
2481
# the error isn't in this repo, we know it's the next step in this
2482
# left-hand history.
2483
partial_history.append(err.revision_id)
2484
if len(partial_history) <= distance_from_known:
2485
# Didn't find enough history to get a revid for the revno.
2486
earliest_revno = known_revno - len(partial_history) + 1
2487
return (False, (earliest_revno, partial_history[-1]))
2488
if len(partial_history) - 1 > distance_from_known:
2489
raise AssertionError('_iter_for_revno returned too much history')
2490
return (True, partial_history[-1])
2176
return self.get_revision(revision_id).inventory_sha1
2492
2178
def iter_reverse_revision_history(self, revision_id):
2493
2179
"""Iterate backwards through revision ids in the lefthand history
3726
3356
return self.source.revision_ids_to_search_result(result_set)
3359
class InterPackRepo(InterSameDataRepository):
3360
"""Optimised code paths between Pack based repositories."""
3363
def _get_repo_format_to_test(self):
3364
from bzrlib.repofmt import pack_repo
3365
return pack_repo.RepositoryFormatKnitPack6RichRoot()
3368
def is_compatible(source, target):
3369
"""Be compatible with known Pack formats.
3371
We don't test for the stores being of specific types because that
3372
could lead to confusing results, and there is no need to be
3375
InterPackRepo does not support CHK based repositories.
3377
from bzrlib.repofmt.pack_repo import RepositoryFormatPack
3378
from bzrlib.repofmt.groupcompress_repo import RepositoryFormatCHK1
3380
are_packs = (isinstance(source._format, RepositoryFormatPack) and
3381
isinstance(target._format, RepositoryFormatPack))
3382
not_packs = (isinstance(source._format, RepositoryFormatCHK1) or
3383
isinstance(target._format, RepositoryFormatCHK1))
3384
except AttributeError:
3386
if not_packs or not are_packs:
3388
return InterRepository._same_model(source, target)
3391
def fetch(self, revision_id=None, pb=None, find_ghosts=False,
3393
"""See InterRepository.fetch()."""
3394
if (len(self.source._fallback_repositories) > 0 or
3395
len(self.target._fallback_repositories) > 0):
3396
# The pack layer is not aware of fallback repositories, so when
3397
# fetching from a stacked repository or into a stacked repository
3398
# we use the generic fetch logic which uses the VersionedFiles
3399
# attributes on repository.
3400
from bzrlib.fetch import RepoFetcher
3401
fetcher = RepoFetcher(self.target, self.source, revision_id,
3402
pb, find_ghosts, fetch_spec=fetch_spec)
3403
if fetch_spec is not None:
3404
if len(list(fetch_spec.heads)) != 1:
3405
raise AssertionError(
3406
"InterPackRepo.fetch doesn't support "
3407
"fetching multiple heads yet.")
3408
revision_id = list(fetch_spec.heads)[0]
3410
if revision_id is None:
3412
# everything to do - use pack logic
3413
# to fetch from all packs to one without
3414
# inventory parsing etc, IFF nothing to be copied is in the target.
3416
source_revision_ids = frozenset(self.source.all_revision_ids())
3417
revision_ids = source_revision_ids - \
3418
frozenset(self.target.get_parent_map(source_revision_ids))
3419
revision_keys = [(revid,) for revid in revision_ids]
3420
index = self.target._pack_collection.revision_index.combined_index
3421
present_revision_ids = set(item[1][0] for item in
3422
index.iter_entries(revision_keys))
3423
revision_ids = set(revision_ids) - present_revision_ids
3424
# implementing the TODO will involve:
3425
# - detecting when all of a pack is selected
3426
# - avoiding as much as possible pre-selection, so the
3427
# more-core routines such as create_pack_from_packs can filter in
3428
# a just-in-time fashion. (though having a HEADS list on a
3429
# repository might make this a lot easier, because we could
3430
# sensibly detect 'new revisions' without doing a full index scan.
3431
elif _mod_revision.is_null(revision_id):
3436
revision_ids = self.search_missing_revision_ids(revision_id,
3437
find_ghosts=find_ghosts).get_keys()
3438
except errors.NoSuchRevision:
3439
raise errors.InstallFailed([revision_id])
3440
if len(revision_ids) == 0:
3442
return self._pack(self.source, self.target, revision_ids)
3444
def _pack(self, source, target, revision_ids):
3445
from bzrlib.repofmt.pack_repo import Packer
3446
packs = source._pack_collection.all_packs()
3447
pack = Packer(self.target._pack_collection, packs, '.fetch',
3448
revision_ids).pack()
3449
if pack is not None:
3450
self.target._pack_collection._save_pack_names()
3451
copied_revs = pack.get_revision_count()
3452
# Trigger an autopack. This may duplicate effort as we've just done
3453
# a pack creation, but for now it is simpler to think about as
3454
# 'upload data, then repack if needed'.
3455
self.target._pack_collection.autopack()
3456
return (copied_revs, [])
3461
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
3462
"""See InterRepository.missing_revision_ids().
3464
:param find_ghosts: Find ghosts throughout the ancestry of
3467
if not find_ghosts and revision_id is not None:
3468
return self._walk_to_common_revisions([revision_id])
3469
elif revision_id is not None:
3470
# Find ghosts: search for revisions pointing from one repository to
3471
# the other, and vice versa, anywhere in the history of revision_id.
3472
graph = self.target.get_graph(other_repository=self.source)
3473
searcher = graph._make_breadth_first_searcher([revision_id])
3477
next_revs, ghosts = searcher.next_with_ghosts()
3478
except StopIteration:
3480
if revision_id in ghosts:
3481
raise errors.NoSuchRevision(self.source, revision_id)
3482
found_ids.update(next_revs)
3483
found_ids.update(ghosts)
3484
found_ids = frozenset(found_ids)
3485
# Double query here: should be able to avoid this by changing the
3486
# graph api further.
3487
result_set = found_ids - frozenset(
3488
self.target.get_parent_map(found_ids))
3490
source_ids = self.source.all_revision_ids()
3491
# source_ids is the worst possible case we may need to pull.
3492
# now we want to filter source_ids against what we actually
3493
# have in target, but don't try to check for existence where we know
3494
# we do not have a revision as that would be pointless.
3495
target_ids = set(self.target.all_revision_ids())
3496
result_set = set(source_ids).difference(target_ids)
3497
return self.source.revision_ids_to_search_result(result_set)
3729
3500
class InterDifferingSerializer(InterRepository):
3738
3509
# This is redundant with format.check_conversion_target(), however that
3739
3510
# raises an exception, and we just want to say "False" as in we won't
3740
3511
# support converting between these formats.
3741
if 'IDS_never' in debug.debug_flags:
3743
3512
if source.supports_rich_root() and not target.supports_rich_root():
3745
3514
if (source._format.supports_tree_reference
3746
3515
and not target._format.supports_tree_reference):
3748
if target._fallback_repositories and target._format.supports_chks:
3749
# IDS doesn't know how to copy CHKs for the parent inventories it
3750
# adds to stacked repos.
3752
if 'IDS_always' in debug.debug_flags:
3754
# Only use this code path for local source and target. IDS does far
3755
# too much IO (both bandwidth and roundtrips) over a network.
3756
if not source.bzrdir.transport.base.startswith('file:///'):
3758
if not target.bzrdir.transport.base.startswith('file:///'):
3762
def _get_trees(self, revision_ids, cache):
3764
for rev_id in revision_ids:
3766
possible_trees.append((rev_id, cache[rev_id]))
3768
# Not cached, but inventory might be present anyway.
3770
tree = self.source.revision_tree(rev_id)
3771
except errors.NoSuchRevision:
3772
# Nope, parent is ghost.
3775
cache[rev_id] = tree
3776
possible_trees.append((rev_id, tree))
3777
return possible_trees
3779
def _get_delta_for_revision(self, tree, parent_ids, possible_trees):
3519
def _get_delta_for_revision(self, tree, parent_ids, basis_id, cache):
3780
3520
"""Get the best delta and base for this revision.
3782
3522
:return: (basis_id, delta)
3524
possible_trees = [(parent_id, cache[parent_id])
3525
for parent_id in parent_ids
3526
if parent_id in cache]
3527
if len(possible_trees) == 0:
3528
# There either aren't any parents, or the parents aren't in the
3529
# cache, so just use the last converted tree
3530
possible_trees.append((basis_id, cache[basis_id]))
3785
# Generate deltas against each tree, to find the shortest.
3786
texts_possibly_new_in_tree = set()
3787
3532
for basis_id, basis_tree in possible_trees:
3788
3533
delta = tree.inventory._make_delta(basis_tree.inventory)
3789
for old_path, new_path, file_id, new_entry in delta:
3790
if new_path is None:
3791
# This file_id isn't present in the new rev, so we don't
3795
# Rich roots are handled elsewhere...
3797
kind = new_entry.kind
3798
if kind != 'directory' and kind != 'file':
3799
# No text record associated with this inventory entry.
3801
# This is a directory or file that has changed somehow.
3802
texts_possibly_new_in_tree.add((file_id, new_entry.revision))
3803
3534
deltas.append((len(delta), basis_id, delta))
3805
3536
return deltas[0][1:]
3807
def _fetch_parent_invs_for_stacking(self, parent_map, cache):
3808
"""Find all parent revisions that are absent, but for which the
3809
inventory is present, and copy those inventories.
3811
This is necessary to preserve correctness when the source is stacked
3812
without fallbacks configured. (Note that in cases like upgrade the
3813
source may be not have _fallback_repositories even though it is
3817
for parents in parent_map.values():
3818
parent_revs.update(parents)
3819
present_parents = self.source.get_parent_map(parent_revs)
3820
absent_parents = set(parent_revs).difference(present_parents)
3821
parent_invs_keys_for_stacking = self.source.inventories.get_parent_map(
3822
(rev_id,) for rev_id in absent_parents)
3823
parent_inv_ids = [key[-1] for key in parent_invs_keys_for_stacking]
3824
for parent_tree in self.source.revision_trees(parent_inv_ids):
3825
current_revision_id = parent_tree.get_revision_id()
3826
parents_parents_keys = parent_invs_keys_for_stacking[
3827
(current_revision_id,)]
3828
parents_parents = [key[-1] for key in parents_parents_keys]
3829
basis_id = _mod_revision.NULL_REVISION
3830
basis_tree = self.source.revision_tree(basis_id)
3831
delta = parent_tree.inventory._make_delta(basis_tree.inventory)
3832
self.target.add_inventory_by_delta(
3833
basis_id, delta, current_revision_id, parents_parents)
3834
cache[current_revision_id] = parent_tree
3836
def _fetch_batch(self, revision_ids, basis_id, cache, a_graph=None):
3538
def _get_parent_keys(self, root_key, parent_map):
3539
"""Get the parent keys for a given root id."""
3540
root_id, rev_id = root_key
3541
# Include direct parents of the revision, but only if they used
3544
for parent_id in parent_map[rev_id]:
3545
if parent_id == _mod_revision.NULL_REVISION:
3547
if parent_id not in self._revision_id_to_root_id:
3548
# We probably didn't read this revision, go spend the
3549
# extra effort to actually check
3551
tree = self.source.revision_tree(parent_id)
3552
except errors.NoSuchRevision:
3553
# Ghost, fill out _revision_id_to_root_id in case we
3554
# encounter this again.
3555
# But set parent_root_id to None since we don't really know
3556
parent_root_id = None
3558
parent_root_id = tree.get_root_id()
3559
self._revision_id_to_root_id[parent_id] = None
3561
parent_root_id = self._revision_id_to_root_id[parent_id]
3562
if root_id == parent_root_id or parent_root_id is None:
3563
parent_keys.append((root_id, parent_id))
3564
return tuple(parent_keys)
3566
def _new_root_data_stream(self, root_keys_to_create, parent_map):
3567
for root_key in root_keys_to_create:
3568
parent_keys = self._get_parent_keys(root_key, parent_map)
3569
yield versionedfile.FulltextContentFactory(root_key,
3570
parent_keys, None, '')
3572
def _fetch_batch(self, revision_ids, basis_id, cache):
3837
3573
"""Fetch across a few revisions.
3839
3575
:param revision_ids: The revisions to copy
3840
3576
:param basis_id: The revision_id of a tree that must be in cache, used
3841
3577
as a basis for delta when no other base is available
3842
3578
:param cache: A cache of RevisionTrees that we can use.
3843
:param a_graph: A Graph object to determine the heads() of the
3844
rich-root data stream.
3845
3579
:return: The revision_id of the last converted tree. The RevisionTree
3846
3580
for it will be in cache
3853
3587
pending_deltas = []
3854
3588
pending_revisions = []
3855
3589
parent_map = self.source.get_parent_map(revision_ids)
3856
self._fetch_parent_invs_for_stacking(parent_map, cache)
3857
self.source._safe_to_return_from_cache = True
3858
3590
for tree in self.source.revision_trees(revision_ids):
3859
# Find a inventory delta for this revision.
3860
# Find text entries that need to be copied, too.
3861
3591
current_revision_id = tree.get_revision_id()
3862
3592
parent_ids = parent_map.get(current_revision_id, ())
3863
parent_trees = self._get_trees(parent_ids, cache)
3864
possible_trees = list(parent_trees)
3865
if len(possible_trees) == 0:
3866
# There either aren't any parents, or the parents are ghosts,
3867
# so just use the last converted tree.
3868
possible_trees.append((basis_id, cache[basis_id]))
3869
3593
basis_id, delta = self._get_delta_for_revision(tree, parent_ids,
3871
revision = self.source.get_revision(current_revision_id)
3872
pending_deltas.append((basis_id, delta,
3873
current_revision_id, revision.parent_ids))
3874
3595
if self._converting_to_rich_root:
3875
3596
self._revision_id_to_root_id[current_revision_id] = \
3876
3597
tree.get_root_id()
3877
# Determine which texts are in present in this revision but not in
3878
# any of the available parents.
3879
texts_possibly_new_in_tree = set()
3598
# Find text entries that need to be copied
3880
3599
for old_path, new_path, file_id, entry in delta:
3881
if new_path is None:
3882
# This file_id isn't present in the new rev
3886
if not self.target.supports_rich_root():
3887
# The target doesn't support rich root, so we don't
3890
if self._converting_to_rich_root:
3891
# This can't be copied normally, we have to insert
3893
root_keys_to_create.add((file_id, entry.revision))
3896
texts_possibly_new_in_tree.add((file_id, entry.revision))
3897
for basis_id, basis_tree in possible_trees:
3898
basis_inv = basis_tree.inventory
3899
for file_key in list(texts_possibly_new_in_tree):
3900
file_id, file_revision = file_key
3902
entry = basis_inv[file_id]
3903
except errors.NoSuchId:
3905
if entry.revision == file_revision:
3906
texts_possibly_new_in_tree.remove(file_key)
3907
text_keys.update(texts_possibly_new_in_tree)
3600
if new_path is not None:
3603
if not self.target.supports_rich_root():
3604
# The target doesn't support rich root, so we don't
3607
if self._converting_to_rich_root:
3608
# This can't be copied normally, we have to insert
3610
root_keys_to_create.add((file_id, entry.revision))
3612
text_keys.add((file_id, entry.revision))
3613
revision = self.source.get_revision(current_revision_id)
3614
pending_deltas.append((basis_id, delta,
3615
current_revision_id, revision.parent_ids))
3908
3616
pending_revisions.append(revision)
3909
3617
cache[current_revision_id] = tree
3910
3618
basis_id = current_revision_id
3911
self.source._safe_to_return_from_cache = False
3912
3619
# Copy file texts
3913
3620
from_texts = self.source.texts
3914
3621
to_texts = self.target.texts
3915
3622
if root_keys_to_create:
3916
root_stream = _mod_fetch._new_root_data_stream(
3917
root_keys_to_create, self._revision_id_to_root_id, parent_map,
3918
self.source, graph=a_graph)
3623
root_stream = self._new_root_data_stream(root_keys_to_create,
3919
3625
to_texts.insert_record_stream(root_stream)
3920
3626
to_texts.insert_record_stream(from_texts.get_record_stream(
3921
3627
text_keys, self.target._format._fetch_order,
4370
4020
# missing keys can handle suspending a write group).
4371
4021
write_group_tokens = self.target_repo.suspend_write_group()
4372
4022
return write_group_tokens, missing_keys
4373
hint = self.target_repo.commit_write_group()
4374
if (to_serializer != src_serializer and
4375
self.target_repo._format.pack_compresses):
4376
self.target_repo.pack(hint=hint)
4023
self.target_repo.commit_write_group()
4377
4024
return [], set()
4379
def _extract_and_insert_inventory_deltas(self, substream, serializer):
4380
target_rich_root = self.target_repo._format.rich_root_data
4381
target_tree_refs = self.target_repo._format.supports_tree_reference
4382
for record in substream:
4383
# Insert the delta directly
4384
inventory_delta_bytes = record.get_bytes_as('fulltext')
4385
deserialiser = inventory_delta.InventoryDeltaDeserializer()
4387
parse_result = deserialiser.parse_text_bytes(
4388
inventory_delta_bytes)
4389
except inventory_delta.IncompatibleInventoryDelta, err:
4390
trace.mutter("Incompatible delta: %s", err.msg)
4391
raise errors.IncompatibleRevision(self.target_repo._format)
4392
basis_id, new_id, rich_root, tree_refs, inv_delta = parse_result
4393
revision_id = new_id
4394
parents = [key[0] for key in record.parents]
4395
self.target_repo.add_inventory_by_delta(
4396
basis_id, inv_delta, revision_id, parents)
4398
def _extract_and_insert_inventories(self, substream, serializer,
4026
def _extract_and_insert_inventories(self, substream, serializer):
4400
4027
"""Generate a new inventory versionedfile in target, converting data.
4402
4029
The inventory is retrieved from the source, (deserializing it), and
4403
4030
stored in the target (reserializing it in a different format).
4405
target_rich_root = self.target_repo._format.rich_root_data
4406
target_tree_refs = self.target_repo._format.supports_tree_reference
4407
4032
for record in substream:
4408
# It's not a delta, so it must be a fulltext in the source
4409
# serializer's format.
4410
4033
bytes = record.get_bytes_as('fulltext')
4411
4034
revision_id = record.key[0]
4412
4035
inv = serializer.read_inventory_from_string(bytes, revision_id)
4413
4036
parents = [key[0] for key in record.parents]
4414
4037
self.target_repo.add_inventory(revision_id, inv, parents)
4415
# No need to keep holding this full inv in memory when the rest of
4416
# the substream is likely to be all deltas.
4419
4039
def _extract_and_insert_revisions(self, substream, serializer):
4420
4040
for record in substream:
4583
4186
return (not self.from_repository._format.rich_root_data and
4584
4187
self.to_format.rich_root_data)
4586
def _get_inventory_stream(self, revision_ids, missing=False):
4189
def _get_inventory_stream(self, revision_ids):
4587
4190
from_format = self.from_repository._format
4588
if (from_format.supports_chks and self.to_format.supports_chks and
4589
from_format.network_name() == self.to_format.network_name()):
4590
raise AssertionError(
4591
"this case should be handled by GroupCHKStreamSource")
4592
elif 'forceinvdeltas' in debug.debug_flags:
4593
return self._get_convertable_inventory_stream(revision_ids,
4594
delta_versus_null=missing)
4595
elif from_format.network_name() == self.to_format.network_name():
4597
return self._get_simple_inventory_stream(revision_ids,
4599
elif (not from_format.supports_chks and not self.to_format.supports_chks
4600
and from_format._serializer == self.to_format._serializer):
4601
# Essentially the same format.
4602
return self._get_simple_inventory_stream(revision_ids,
4191
if (from_format.supports_chks and self.to_format.supports_chks
4192
and (from_format._serializer == self.to_format._serializer)):
4193
# Both sides support chks, and they use the same serializer, so it
4194
# is safe to transmit the chk pages and inventory pages across
4196
return self._get_chk_inventory_stream(revision_ids)
4197
elif (not from_format.supports_chks):
4198
# Source repository doesn't support chks. So we can transmit the
4199
# inventories 'as-is' and either they are just accepted on the
4200
# target, or the Sink will properly convert it.
4201
return self._get_simple_inventory_stream(revision_ids)
4605
# Any time we switch serializations, we want to use an
4606
# inventory-delta based approach.
4607
return self._get_convertable_inventory_stream(revision_ids,
4608
delta_versus_null=missing)
4203
# XXX: Hack to make not-chk->chk fetch: copy the inventories as
4204
# inventories. Note that this should probably be done somehow
4205
# as part of bzrlib.repository.StreamSink. Except JAM couldn't
4206
# figure out how a non-chk repository could possibly handle
4207
# deserializing an inventory stream from a chk repo, as it
4208
# doesn't have a way to understand individual pages.
4209
return self._get_convertable_inventory_stream(revision_ids)
4610
def _get_simple_inventory_stream(self, revision_ids, missing=False):
4611
# NB: This currently reopens the inventory weave in source;
4612
# using a single stream interface instead would avoid this.
4211
def _get_simple_inventory_stream(self, revision_ids):
4613
4212
from_weave = self.from_repository.inventories
4615
delta_closure = True
4617
delta_closure = not self.delta_on_metadata()
4618
4213
yield ('inventories', from_weave.get_record_stream(
4619
4214
[(rev_id,) for rev_id in revision_ids],
4620
self.inventory_fetch_order(), delta_closure))
4622
def _get_convertable_inventory_stream(self, revision_ids,
4623
delta_versus_null=False):
4624
# The two formats are sufficiently different that there is no fast
4625
# path, so we need to send just inventorydeltas, which any
4626
# sufficiently modern client can insert into any repository.
4627
# The StreamSink code expects to be able to
4628
# convert on the target, so we need to put bytes-on-the-wire that can
4629
# be converted. That means inventory deltas (if the remote is <1.19,
4630
# RemoteStreamSink will fallback to VFS to insert the deltas).
4631
yield ('inventory-deltas',
4632
self._stream_invs_as_deltas(revision_ids,
4633
delta_versus_null=delta_versus_null))
4635
def _stream_invs_as_deltas(self, revision_ids, delta_versus_null=False):
4636
"""Return a stream of inventory-deltas for the given rev ids.
4638
:param revision_ids: The list of inventories to transmit
4639
:param delta_versus_null: Don't try to find a minimal delta for this
4640
entry, instead compute the delta versus the NULL_REVISION. This
4641
effectively streams a complete inventory. Used for stuff like
4642
filling in missing parents, etc.
4215
self.inventory_fetch_order(),
4216
not self.delta_on_metadata()))
4218
def _get_chk_inventory_stream(self, revision_ids):
4219
"""Fetch the inventory texts, along with the associated chk maps."""
4220
# We want an inventory outside of the search set, so that we can filter
4221
# out uninteresting chk pages. For now we use
4222
# _find_revision_outside_set, but if we had a Search with cut_revs, we
4223
# could use that instead.
4224
start_rev_id = self.from_repository._find_revision_outside_set(
4226
start_rev_key = (start_rev_id,)
4227
inv_keys_to_fetch = [(rev_id,) for rev_id in revision_ids]
4228
if start_rev_id != _mod_revision.NULL_REVISION:
4229
inv_keys_to_fetch.append((start_rev_id,))
4230
# Any repo that supports chk_bytes must also support out-of-order
4231
# insertion. At least, that is how we expect it to work
4232
# We use get_record_stream instead of iter_inventories because we want
4233
# to be able to insert the stream as well. We could instead fetch
4234
# allowing deltas, and then iter_inventories, but we don't know whether
4235
# source or target is more 'local' anway.
4236
inv_stream = self.from_repository.inventories.get_record_stream(
4237
inv_keys_to_fetch, 'unordered',
4238
True) # We need them as full-texts so we can find their references
4239
uninteresting_chk_roots = set()
4240
interesting_chk_roots = set()
4241
def filter_inv_stream(inv_stream):
4242
for idx, record in enumerate(inv_stream):
4243
### child_pb.update('fetch inv', idx, len(inv_keys_to_fetch))
4244
bytes = record.get_bytes_as('fulltext')
4245
chk_inv = inventory.CHKInventory.deserialise(
4246
self.from_repository.chk_bytes, bytes, record.key)
4247
if record.key == start_rev_key:
4248
uninteresting_chk_roots.add(chk_inv.id_to_entry.key())
4249
p_id_map = chk_inv.parent_id_basename_to_file_id
4250
if p_id_map is not None:
4251
uninteresting_chk_roots.add(p_id_map.key())
4254
interesting_chk_roots.add(chk_inv.id_to_entry.key())
4255
p_id_map = chk_inv.parent_id_basename_to_file_id
4256
if p_id_map is not None:
4257
interesting_chk_roots.add(p_id_map.key())
4258
### pb.update('fetch inventory', 0, 2)
4259
yield ('inventories', filter_inv_stream(inv_stream))
4260
# Now that we have worked out all of the interesting root nodes, grab
4261
# all of the interesting pages and insert them
4262
### pb.update('fetch inventory', 1, 2)
4263
interesting = chk_map.iter_interesting_nodes(
4264
self.from_repository.chk_bytes, interesting_chk_roots,
4265
uninteresting_chk_roots)
4266
def to_stream_adapter():
4267
"""Adapt the iter_interesting_nodes result to a single stream.
4269
iter_interesting_nodes returns records as it processes them, along
4270
with keys. However, we only want to return the records themselves.
4272
for record, items in interesting:
4273
if record is not None:
4275
# XXX: We could instead call get_record_stream(records.keys())
4276
# ATM, this will always insert the records as fulltexts, and
4277
# requires that you can hang on to records once you have gone
4278
# on to the next one. Further, it causes the target to
4279
# recompress the data. Testing shows it to be faster than
4280
# requesting the records again, though.
4281
yield ('chk_bytes', to_stream_adapter())
4282
### pb.update('fetch inventory', 2, 2)
4284
def _get_convertable_inventory_stream(self, revision_ids):
4285
# XXX: One of source or target is using chks, and they don't have
4286
# compatible serializations. The StreamSink code expects to be
4287
# able to convert on the target, so we need to put
4288
# bytes-on-the-wire that can be converted
4289
yield ('inventories', self._stream_invs_as_fulltexts(revision_ids))
4291
def _stream_invs_as_fulltexts(self, revision_ids):
4644
4292
from_repo = self.from_repository
4293
from_serializer = from_repo._format._serializer
4645
4294
revision_keys = [(rev_id,) for rev_id in revision_ids]
4646
4295
parent_map = from_repo.inventories.get_parent_map(revision_keys)
4647
# XXX: possibly repos could implement a more efficient iter_inv_deltas
4649
inventories = self.from_repository.iter_inventories(
4650
revision_ids, 'topological')
4651
format = from_repo._format
4652
invs_sent_so_far = set([_mod_revision.NULL_REVISION])
4653
inventory_cache = lru_cache.LRUCache(50)
4654
null_inventory = from_repo.revision_tree(
4655
_mod_revision.NULL_REVISION).inventory
4656
# XXX: ideally the rich-root/tree-refs flags would be per-revision, not
4657
# per-repo (e.g. streaming a non-rich-root revision out of a rich-root
4658
# repo back into a non-rich-root repo ought to be allowed)
4659
serializer = inventory_delta.InventoryDeltaSerializer(
4660
versioned_root=format.rich_root_data,
4661
tree_references=format.supports_tree_reference)
4662
for inv in inventories:
4296
for inv in self.from_repository.iter_inventories(revision_ids):
4297
# XXX: This is a bit hackish, but it works. Basically,
4298
# CHKSerializer 'accidentally' supports
4299
# read/write_inventory_to_string, even though that is never
4300
# the format that is stored on disk. It *does* give us a
4301
# single string representation for an inventory, so live with
4303
# This would be far better if we had a 'serialized inventory
4304
# delta' form. Then we could use 'inventory._make_delta', and
4305
# transmit that. This would both be faster to generate, and
4306
# result in fewer bytes-on-the-wire.
4307
as_bytes = from_serializer.write_inventory_to_string(inv)
4663
4308
key = (inv.revision_id,)
4664
4309
parent_keys = parent_map.get(key, ())
4666
if not delta_versus_null and parent_keys:
4667
# The caller did not ask for complete inventories and we have
4668
# some parents that we can delta against. Make a delta against
4669
# each parent so that we can find the smallest.
4670
parent_ids = [parent_key[0] for parent_key in parent_keys]
4671
for parent_id in parent_ids:
4672
if parent_id not in invs_sent_so_far:
4673
# We don't know that the remote side has this basis, so
4676
if parent_id == _mod_revision.NULL_REVISION:
4677
parent_inv = null_inventory
4679
parent_inv = inventory_cache.get(parent_id, None)
4680
if parent_inv is None:
4681
parent_inv = from_repo.get_inventory(parent_id)
4682
candidate_delta = inv._make_delta(parent_inv)
4683
if (delta is None or
4684
len(delta) > len(candidate_delta)):
4685
delta = candidate_delta
4686
basis_id = parent_id
4688
# Either none of the parents ended up being suitable, or we
4689
# were asked to delta against NULL
4690
basis_id = _mod_revision.NULL_REVISION
4691
delta = inv._make_delta(null_inventory)
4692
invs_sent_so_far.add(inv.revision_id)
4693
inventory_cache[inv.revision_id] = inv
4694
delta_serialized = ''.join(
4695
serializer.delta_to_lines(basis_id, key[-1], delta))
4696
4310
yield versionedfile.FulltextContentFactory(
4697
key, parent_keys, None, delta_serialized)
4700
def _iter_for_revno(repo, partial_history_cache, stop_index=None,
4701
stop_revision=None):
4702
"""Extend the partial history to include a given index
4704
If a stop_index is supplied, stop when that index has been reached.
4705
If a stop_revision is supplied, stop when that revision is
4706
encountered. Otherwise, stop when the beginning of history is
4709
:param stop_index: The index which should be present. When it is
4710
present, history extension will stop.
4711
:param stop_revision: The revision id which should be present. When
4712
it is encountered, history extension will stop.
4714
start_revision = partial_history_cache[-1]
4715
iterator = repo.iter_reverse_revision_history(start_revision)
4717
#skip the last revision in the list
4720
if (stop_index is not None and
4721
len(partial_history_cache) > stop_index):
4723
if partial_history_cache[-1] == stop_revision:
4725
revision_id = iterator.next()
4726
partial_history_cache.append(revision_id)
4727
except StopIteration:
4311
key, parent_keys, None, as_bytes)
added
removed
bzrlib/repository.py
