1146
1154
# The old API returned a list, should this actually be a set?
1147
1155
return parent_map.keys()
1157
def _check_inventories(self, checker):
1158
"""Check the inventories found from the revision scan.
1160
This is responsible for verifying the sha1 of inventories and
1161
creating a pending_keys set that covers data referenced by inventories.
1163
bar = ui.ui_factory.nested_progress_bar()
1165
self._do_check_inventories(checker, bar)
1169
def _do_check_inventories(self, checker, bar):
1170
"""Helper for _check_inventories."""
1172
keys = {'chk_bytes':set(), 'inventories':set(), 'texts':set()}
1173
kinds = ['chk_bytes', 'texts']
1174
count = len(checker.pending_keys)
1175
bar.update("inventories", 0, 2)
1176
current_keys = checker.pending_keys
1177
checker.pending_keys = {}
1178
# Accumulate current checks.
1179
for key in current_keys:
1180
if key[0] != 'inventories' and key[0] not in kinds:
1181
checker._report_items.append('unknown key type %r' % (key,))
1182
keys[key[0]].add(key[1:])
1183
if keys['inventories']:
1184
# NB: output order *should* be roughly sorted - topo or
1185
# inverse topo depending on repository - either way decent
1186
# to just delta against. However, pre-CHK formats didn't
1187
# try to optimise inventory layout on disk. As such the
1188
# pre-CHK code path does not use inventory deltas.
1190
for record in self.inventories.check(keys=keys['inventories']):
1191
if record.storage_kind == 'absent':
1192
checker._report_items.append(
1193
'Missing inventory {%s}' % (record.key,))
1195
last_object = self._check_record('inventories', record,
1196
checker, last_object,
1197
current_keys[('inventories',) + record.key])
1198
del keys['inventories']
1201
bar.update("texts", 1)
1202
while (checker.pending_keys or keys['chk_bytes']
1204
# Something to check.
1205
current_keys = checker.pending_keys
1206
checker.pending_keys = {}
1207
# Accumulate current checks.
1208
for key in current_keys:
1209
if key[0] not in kinds:
1210
checker._report_items.append('unknown key type %r' % (key,))
1211
keys[key[0]].add(key[1:])
1212
# Check the outermost kind only - inventories || chk_bytes || texts
1216
for record in getattr(self, kind).check(keys=keys[kind]):
1217
if record.storage_kind == 'absent':
1218
checker._report_items.append(
1219
'Missing inventory {%s}' % (record.key,))
1221
last_object = self._check_record(kind, record,
1222
checker, last_object, current_keys[(kind,) + record.key])
1226
def _check_record(self, kind, record, checker, last_object, item_data):
1227
"""Check a single text from this repository."""
1228
if kind == 'inventories':
1229
rev_id = record.key[0]
1230
inv = self.deserialise_inventory(rev_id,
1231
record.get_bytes_as('fulltext'))
1232
if last_object is not None:
1233
delta = inv._make_delta(last_object)
1234
for old_path, path, file_id, ie in delta:
1237
ie.check(checker, rev_id, inv)
1239
for path, ie in inv.iter_entries():
1240
ie.check(checker, rev_id, inv)
1241
if self._format.fast_deltas:
1243
elif kind == 'chk_bytes':
1244
# No code written to check chk_bytes for this repo format.
1245
checker._report_items.append(
1246
'unsupported key type chk_bytes for %s' % (record.key,))
1247
elif kind == 'texts':
1248
self._check_text(record, checker, item_data)
1250
checker._report_items.append(
1251
'unknown key type %s for %s' % (kind, record.key))
1253
def _check_text(self, record, checker, item_data):
1254
"""Check a single text."""
1255
# Check it is extractable.
1256
# TODO: check length.
1257
if record.storage_kind == 'chunked':
1258
chunks = record.get_bytes_as(record.storage_kind)
1259
sha1 = osutils.sha_strings(chunks)
1260
length = sum(map(len, chunks))
1262
content = record.get_bytes_as('fulltext')
1263
sha1 = osutils.sha_string(content)
1264
length = len(content)
1265
if item_data and sha1 != item_data[1]:
1266
checker._report_items.append(
1267
'sha1 mismatch: %s has sha1 %s expected %s referenced by %s' %
1268
(record.key, sha1, item_data[1], item_data[2]))
1150
1271
def create(a_bzrdir):
1151
1272
"""Construct the current default format repository in a_bzrdir."""
1715
1842
@needs_read_lock
1716
1843
def get_revisions(self, revision_ids):
1717
"""Get many revisions at once."""
1844
"""Get many revisions at once.
1846
Repositories that need to check data on every revision read should
1847
subclass this method.
1718
1849
return self._get_revisions(revision_ids)
1720
1851
@needs_read_lock
1721
1852
def _get_revisions(self, revision_ids):
1722
1853
"""Core work logic to get many revisions without sanity checks."""
1723
for rev_id in revision_ids:
1724
if not rev_id or not isinstance(rev_id, basestring):
1725
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1855
for revid, rev in self._iter_revisions(revision_ids):
1857
raise errors.NoSuchRevision(self, revid)
1859
return [revs[revid] for revid in revision_ids]
1861
def _iter_revisions(self, revision_ids):
1862
"""Iterate over revision objects.
1864
:param revision_ids: An iterable of revisions to examine. None may be
1865
passed to request all revisions known to the repository. Note that
1866
not all repositories can find unreferenced revisions; for those
1867
repositories only referenced ones will be returned.
1868
:return: An iterator of (revid, revision) tuples. Absent revisions (
1869
those asked for but not available) are returned as (revid, None).
1871
if revision_ids is None:
1872
revision_ids = self.all_revision_ids()
1874
for rev_id in revision_ids:
1875
if not rev_id or not isinstance(rev_id, basestring):
1876
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1726
1877
keys = [(key,) for key in revision_ids]
1727
1878
stream = self.revisions.get_record_stream(keys, 'unordered', True)
1729
1879
for record in stream:
1880
revid = record.key[0]
1730
1881
if record.storage_kind == 'absent':
1731
raise errors.NoSuchRevision(self, record.key[0])
1732
text = record.get_bytes_as('fulltext')
1733
rev = self._serializer.read_revision_from_string(text)
1734
revs[record.key[0]] = rev
1735
return [revs[revid] for revid in revision_ids]
1884
text = record.get_bytes_as('fulltext')
1885
rev = self._serializer.read_revision_from_string(text)
1737
1888
@needs_read_lock
1738
1889
def get_revision_xml(self, revision_id):
2210
2360
:param revision_ids: The expected revision ids of the inventories.
2361
:param ordering: optional ordering, e.g. 'topological'. If not
2362
specified, the order of revision_ids will be preserved (by
2363
buffering if necessary).
2211
2364
:return: An iterator of inventories.
2213
2366
if ((None in revision_ids)
2214
2367
or (_mod_revision.NULL_REVISION in revision_ids)):
2215
2368
raise ValueError('cannot get null revision inventory')
2216
return self._iter_inventories(revision_ids)
2369
return self._iter_inventories(revision_ids, ordering)
2218
def _iter_inventories(self, revision_ids):
2371
def _iter_inventories(self, revision_ids, ordering):
2219
2372
"""single-document based inventory iteration."""
2220
for text, revision_id in self._iter_inventory_xmls(revision_ids):
2373
inv_xmls = self._iter_inventory_xmls(revision_ids, ordering)
2374
for text, revision_id in inv_xmls:
2221
2375
yield self.deserialise_inventory(revision_id, text)
2223
def _iter_inventory_xmls(self, revision_ids):
2377
def _iter_inventory_xmls(self, revision_ids, ordering):
2378
if ordering is None:
2379
order_as_requested = True
2380
ordering = 'unordered'
2382
order_as_requested = False
2224
2383
keys = [(revision_id,) for revision_id in revision_ids]
2225
stream = self.inventories.get_record_stream(keys, 'unordered', True)
2386
if order_as_requested:
2387
key_iter = iter(keys)
2388
next_key = key_iter.next()
2389
stream = self.inventories.get_record_stream(keys, ordering, True)
2226
2390
text_chunks = {}
2227
2391
for record in stream:
2228
2392
if record.storage_kind != 'absent':
2229
text_chunks[record.key] = record.get_bytes_as('chunked')
2393
chunks = record.get_bytes_as('chunked')
2394
if order_as_requested:
2395
text_chunks[record.key] = chunks
2397
yield ''.join(chunks), record.key[-1]
2231
2399
raise errors.NoSuchRevision(self, record.key)
2233
chunks = text_chunks.pop(key)
2234
yield ''.join(chunks), key[-1]
2400
if order_as_requested:
2401
# Yield as many results as we can while preserving order.
2402
while next_key in text_chunks:
2403
chunks = text_chunks.pop(next_key)
2404
yield ''.join(chunks), next_key[-1]
2406
next_key = key_iter.next()
2407
except StopIteration:
2408
# We still want to fully consume the get_record_stream,
2409
# just in case it is not actually finished at this point
2236
2413
def deserialise_inventory(self, revision_id, xml):
2237
2414
"""Transform the xml into an inventory object.
3541
3750
return deltas[0][1:]
3543
def _get_parent_keys(self, root_key, parent_map):
3544
"""Get the parent keys for a given root id."""
3545
root_id, rev_id = root_key
3546
# Include direct parents of the revision, but only if they used
3547
# the same root_id and are heads.
3549
for parent_id in parent_map[rev_id]:
3550
if parent_id == _mod_revision.NULL_REVISION:
3552
if parent_id not in self._revision_id_to_root_id:
3553
# We probably didn't read this revision, go spend the
3554
# extra effort to actually check
3556
tree = self.source.revision_tree(parent_id)
3557
except errors.NoSuchRevision:
3558
# Ghost, fill out _revision_id_to_root_id in case we
3559
# encounter this again.
3560
# But set parent_root_id to None since we don't really know
3561
parent_root_id = None
3563
parent_root_id = tree.get_root_id()
3564
self._revision_id_to_root_id[parent_id] = None
3566
parent_root_id = self._revision_id_to_root_id[parent_id]
3567
if root_id == parent_root_id:
3568
# With stacking we _might_ want to refer to a non-local
3569
# revision, but this code path only applies when we have the
3570
# full content available, so ghosts really are ghosts, not just
3571
# the edge of local data.
3572
parent_keys.append((parent_id,))
3574
# root_id may be in the parent anyway.
3576
tree = self.source.revision_tree(parent_id)
3577
except errors.NoSuchRevision:
3578
# ghost, can't refer to it.
3582
parent_keys.append((tree.inventory[root_id].revision,))
3583
except errors.NoSuchId:
3586
g = graph.Graph(self.source.revisions)
3587
heads = g.heads(parent_keys)
3589
for key in parent_keys:
3590
if key in heads and key not in selected_keys:
3591
selected_keys.append(key)
3592
return tuple([(root_id,)+ key for key in selected_keys])
3594
def _new_root_data_stream(self, root_keys_to_create, parent_map):
3595
for root_key in root_keys_to_create:
3596
parent_keys = self._get_parent_keys(root_key, parent_map)
3597
yield versionedfile.FulltextContentFactory(root_key,
3598
parent_keys, None, '')
3600
3752
def _fetch_batch(self, revision_ids, basis_id, cache):
3601
3753
"""Fetch across a few revisions.
4071
4242
self.target_repo.pack(hint=hint)
4072
4243
return [], set()
4074
def _extract_and_insert_inventories(self, substream, serializer):
4245
def _extract_and_insert_inventory_deltas(self, substream, serializer):
4246
target_rich_root = self.target_repo._format.rich_root_data
4247
target_tree_refs = self.target_repo._format.supports_tree_reference
4248
for record in substream:
4249
# Insert the delta directly
4250
inventory_delta_bytes = record.get_bytes_as('fulltext')
4251
deserialiser = inventory_delta.InventoryDeltaDeserializer()
4253
parse_result = deserialiser.parse_text_bytes(
4254
inventory_delta_bytes)
4255
except inventory_delta.IncompatibleInventoryDelta, err:
4256
trace.mutter("Incompatible delta: %s", err.msg)
4257
raise errors.IncompatibleRevision(self.target_repo._format)
4258
basis_id, new_id, rich_root, tree_refs, inv_delta = parse_result
4259
revision_id = new_id
4260
parents = [key[0] for key in record.parents]
4261
self.target_repo.add_inventory_by_delta(
4262
basis_id, inv_delta, revision_id, parents)
4264
def _extract_and_insert_inventories(self, substream, serializer,
4075
4266
"""Generate a new inventory versionedfile in target, converting data.
4077
4268
The inventory is retrieved from the source, (deserializing it), and
4078
4269
stored in the target (reserializing it in a different format).
4271
target_rich_root = self.target_repo._format.rich_root_data
4272
target_tree_refs = self.target_repo._format.supports_tree_reference
4080
4273
for record in substream:
4274
# It's not a delta, so it must be a fulltext in the source
4275
# serializer's format.
4081
4276
bytes = record.get_bytes_as('fulltext')
4082
4277
revision_id = record.key[0]
4083
4278
inv = serializer.read_inventory_from_string(bytes, revision_id)
4084
4279
parents = [key[0] for key in record.parents]
4085
4280
self.target_repo.add_inventory(revision_id, inv, parents)
4281
# No need to keep holding this full inv in memory when the rest of
4282
# the substream is likely to be all deltas.
4087
4285
def _extract_and_insert_revisions(self, substream, serializer):
4088
4286
for record in substream:
4246
4450
return (not self.from_repository._format.rich_root_data and
4247
4451
self.to_format.rich_root_data)
4249
def _get_inventory_stream(self, revision_ids):
4453
def _get_inventory_stream(self, revision_ids, missing=False):
4250
4454
from_format = self.from_repository._format
4251
if (from_format.supports_chks and self.to_format.supports_chks
4252
and (from_format._serializer == self.to_format._serializer)):
4253
# Both sides support chks, and they use the same serializer, so it
4254
# is safe to transmit the chk pages and inventory pages across
4256
return self._get_chk_inventory_stream(revision_ids)
4257
elif (not from_format.supports_chks):
4258
# Source repository doesn't support chks. So we can transmit the
4259
# inventories 'as-is' and either they are just accepted on the
4260
# target, or the Sink will properly convert it.
4261
return self._get_simple_inventory_stream(revision_ids)
4455
if (from_format.supports_chks and self.to_format.supports_chks and
4456
from_format.network_name() == self.to_format.network_name()):
4457
raise AssertionError(
4458
"this case should be handled by GroupCHKStreamSource")
4459
elif 'forceinvdeltas' in debug.debug_flags:
4460
return self._get_convertable_inventory_stream(revision_ids,
4461
delta_versus_null=missing)
4462
elif from_format.network_name() == self.to_format.network_name():
4464
return self._get_simple_inventory_stream(revision_ids,
4466
elif (not from_format.supports_chks and not self.to_format.supports_chks
4467
and from_format._serializer == self.to_format._serializer):
4468
# Essentially the same format.
4469
return self._get_simple_inventory_stream(revision_ids,
4263
# XXX: Hack to make not-chk->chk fetch: copy the inventories as
4264
# inventories. Note that this should probably be done somehow
4265
# as part of bzrlib.repository.StreamSink. Except JAM couldn't
4266
# figure out how a non-chk repository could possibly handle
4267
# deserializing an inventory stream from a chk repo, as it
4268
# doesn't have a way to understand individual pages.
4269
return self._get_convertable_inventory_stream(revision_ids)
4472
# Any time we switch serializations, we want to use an
4473
# inventory-delta based approach.
4474
return self._get_convertable_inventory_stream(revision_ids,
4475
delta_versus_null=missing)
4271
def _get_simple_inventory_stream(self, revision_ids):
4477
def _get_simple_inventory_stream(self, revision_ids, missing=False):
4478
# NB: This currently reopens the inventory weave in source;
4479
# using a single stream interface instead would avoid this.
4272
4480
from_weave = self.from_repository.inventories
4482
delta_closure = True
4484
delta_closure = not self.delta_on_metadata()
4273
4485
yield ('inventories', from_weave.get_record_stream(
4274
4486
[(rev_id,) for rev_id in revision_ids],
4275
self.inventory_fetch_order(),
4276
not self.delta_on_metadata()))
4278
def _get_chk_inventory_stream(self, revision_ids):
4279
"""Fetch the inventory texts, along with the associated chk maps."""
4280
# We want an inventory outside of the search set, so that we can filter
4281
# out uninteresting chk pages. For now we use
4282
# _find_revision_outside_set, but if we had a Search with cut_revs, we
4283
# could use that instead.
4284
start_rev_id = self.from_repository._find_revision_outside_set(
4286
start_rev_key = (start_rev_id,)
4287
inv_keys_to_fetch = [(rev_id,) for rev_id in revision_ids]
4288
if start_rev_id != _mod_revision.NULL_REVISION:
4289
inv_keys_to_fetch.append((start_rev_id,))
4290
# Any repo that supports chk_bytes must also support out-of-order
4291
# insertion. At least, that is how we expect it to work
4292
# We use get_record_stream instead of iter_inventories because we want
4293
# to be able to insert the stream as well. We could instead fetch
4294
# allowing deltas, and then iter_inventories, but we don't know whether
4295
# source or target is more 'local' anway.
4296
inv_stream = self.from_repository.inventories.get_record_stream(
4297
inv_keys_to_fetch, 'unordered',
4298
True) # We need them as full-texts so we can find their references
4299
uninteresting_chk_roots = set()
4300
interesting_chk_roots = set()
4301
def filter_inv_stream(inv_stream):
4302
for idx, record in enumerate(inv_stream):
4303
### child_pb.update('fetch inv', idx, len(inv_keys_to_fetch))
4304
bytes = record.get_bytes_as('fulltext')
4305
chk_inv = inventory.CHKInventory.deserialise(
4306
self.from_repository.chk_bytes, bytes, record.key)
4307
if record.key == start_rev_key:
4308
uninteresting_chk_roots.add(chk_inv.id_to_entry.key())
4309
p_id_map = chk_inv.parent_id_basename_to_file_id
4310
if p_id_map is not None:
4311
uninteresting_chk_roots.add(p_id_map.key())
4314
interesting_chk_roots.add(chk_inv.id_to_entry.key())
4315
p_id_map = chk_inv.parent_id_basename_to_file_id
4316
if p_id_map is not None:
4317
interesting_chk_roots.add(p_id_map.key())
4318
### pb.update('fetch inventory', 0, 2)
4319
yield ('inventories', filter_inv_stream(inv_stream))
4320
# Now that we have worked out all of the interesting root nodes, grab
4321
# all of the interesting pages and insert them
4322
### pb.update('fetch inventory', 1, 2)
4323
interesting = chk_map.iter_interesting_nodes(
4324
self.from_repository.chk_bytes, interesting_chk_roots,
4325
uninteresting_chk_roots)
4326
def to_stream_adapter():
4327
"""Adapt the iter_interesting_nodes result to a single stream.
4329
iter_interesting_nodes returns records as it processes them, along
4330
with keys. However, we only want to return the records themselves.
4332
for record, items in interesting:
4333
if record is not None:
4335
# XXX: We could instead call get_record_stream(records.keys())
4336
# ATM, this will always insert the records as fulltexts, and
4337
# requires that you can hang on to records once you have gone
4338
# on to the next one. Further, it causes the target to
4339
# recompress the data. Testing shows it to be faster than
4340
# requesting the records again, though.
4341
yield ('chk_bytes', to_stream_adapter())
4342
### pb.update('fetch inventory', 2, 2)
4344
def _get_convertable_inventory_stream(self, revision_ids):
4345
# XXX: One of source or target is using chks, and they don't have
4346
# compatible serializations. The StreamSink code expects to be
4347
# able to convert on the target, so we need to put
4348
# bytes-on-the-wire that can be converted
4349
yield ('inventories', self._stream_invs_as_fulltexts(revision_ids))
4351
def _stream_invs_as_fulltexts(self, revision_ids):
4487
self.inventory_fetch_order(), delta_closure))
4489
def _get_convertable_inventory_stream(self, revision_ids,
4490
delta_versus_null=False):
4491
# The source is using CHKs, but the target either doesn't or it has a
4492
# different serializer. The StreamSink code expects to be able to
4493
# convert on the target, so we need to put bytes-on-the-wire that can
4494
# be converted. That means inventory deltas (if the remote is <1.19,
4495
# RemoteStreamSink will fallback to VFS to insert the deltas).
4496
yield ('inventory-deltas',
4497
self._stream_invs_as_deltas(revision_ids,
4498
delta_versus_null=delta_versus_null))
4500
def _stream_invs_as_deltas(self, revision_ids, delta_versus_null=False):
4501
"""Return a stream of inventory-deltas for the given rev ids.
4503
:param revision_ids: The list of inventories to transmit
4504
:param delta_versus_null: Don't try to find a minimal delta for this
4505
entry, instead compute the delta versus the NULL_REVISION. This
4506
effectively streams a complete inventory. Used for stuff like
4507
filling in missing parents, etc.
4352
4509
from_repo = self.from_repository
4353
from_serializer = from_repo._format._serializer
4354
4510
revision_keys = [(rev_id,) for rev_id in revision_ids]
4355
4511
parent_map = from_repo.inventories.get_parent_map(revision_keys)
4356
for inv in self.from_repository.iter_inventories(revision_ids):
4357
# XXX: This is a bit hackish, but it works. Basically,
4358
# CHKSerializer 'accidentally' supports
4359
# read/write_inventory_to_string, even though that is never
4360
# the format that is stored on disk. It *does* give us a
4361
# single string representation for an inventory, so live with
4363
# This would be far better if we had a 'serialized inventory
4364
# delta' form. Then we could use 'inventory._make_delta', and
4365
# transmit that. This would both be faster to generate, and
4366
# result in fewer bytes-on-the-wire.
4367
as_bytes = from_serializer.write_inventory_to_string(inv)
4512
# XXX: possibly repos could implement a more efficient iter_inv_deltas
4514
inventories = self.from_repository.iter_inventories(
4515
revision_ids, 'topological')
4516
format = from_repo._format
4517
invs_sent_so_far = set([_mod_revision.NULL_REVISION])
4518
inventory_cache = lru_cache.LRUCache(50)
4519
null_inventory = from_repo.revision_tree(
4520
_mod_revision.NULL_REVISION).inventory
4521
# XXX: ideally the rich-root/tree-refs flags would be per-revision, not
4522
# per-repo (e.g. streaming a non-rich-root revision out of a rich-root
4523
# repo back into a non-rich-root repo ought to be allowed)
4524
serializer = inventory_delta.InventoryDeltaSerializer(
4525
versioned_root=format.rich_root_data,
4526
tree_references=format.supports_tree_reference)
4527
for inv in inventories:
4368
4528
key = (inv.revision_id,)
4369
4529
parent_keys = parent_map.get(key, ())
4531
if not delta_versus_null and parent_keys:
4532
# The caller did not ask for complete inventories and we have
4533
# some parents that we can delta against. Make a delta against
4534
# each parent so that we can find the smallest.
4535
parent_ids = [parent_key[0] for parent_key in parent_keys]
4536
for parent_id in parent_ids:
4537
if parent_id not in invs_sent_so_far:
4538
# We don't know that the remote side has this basis, so
4541
if parent_id == _mod_revision.NULL_REVISION:
4542
parent_inv = null_inventory
4544
parent_inv = inventory_cache.get(parent_id, None)
4545
if parent_inv is None:
4546
parent_inv = from_repo.get_inventory(parent_id)
4547
candidate_delta = inv._make_delta(parent_inv)
4548
if (delta is None or
4549
len(delta) > len(candidate_delta)):
4550
delta = candidate_delta
4551
basis_id = parent_id
4553
# Either none of the parents ended up being suitable, or we
4554
# were asked to delta against NULL
4555
basis_id = _mod_revision.NULL_REVISION
4556
delta = inv._make_delta(null_inventory)
4557
invs_sent_so_far.add(inv.revision_id)
4558
inventory_cache[inv.revision_id] = inv
4559
delta_serialized = ''.join(
4560
serializer.delta_to_lines(basis_id, key[-1], delta))
4370
4561
yield versionedfile.FulltextContentFactory(
4371
key, parent_keys, None, as_bytes)
4562
key, parent_keys, None, delta_serialized)
4374
4565
def _iter_for_revno(repo, partial_history_cache, stop_index=None,