« back to all changes in this revision
Viewing changes to bzrlib/repository.py
- browse files at revision 4608
- compare with another revision
- compare with revision 4476.3.35
- download diff
- download tarball
- view history from revision 4608
- Committer: Canonical.com Patch Queue Manager
- Date: 2009-08-14 05:49:27 UTC
- mfrom: (4476.3.86 inventory-delta)
- Revision ID: pqm@pqm.ubuntu.com-20090814054927-k0k18dn46ax4b91f
(andrew) Add inventory-delta streaming for cross-format fetch.
- files modified:
- NEWS
added
removed
bzrlib/repository.py
31
31
gpg,
32
32
graph,
33
33
inventory,
34
inventory_delta,
34
35
lazy_regex,
35
36
lockable_files,
36
37
lockdir,
924
925
"""
925
926
if self._write_group is not self.get_transaction():
926
927
# has an unlock or relock occured ?
928
if suppress_errors:
929
mutter(
930
'(suppressed) mismatched lock context and write group. %r, %r',
931
self._write_group, self.get_transaction())
932
return
927
933
raise errors.BzrError(
928
934
'mismatched lock context and write group. %r, %r' %
929
935
(self._write_group, self.get_transaction()))
1063
1069
check_content=True):
1064
1070
"""Store lines in inv_vf and return the sha1 of the inventory."""
1065
1071
parents = [(parent,) for parent in parents]
1066
return self.inventories.add_lines((revision_id,), parents, lines,
1072
result = self.inventories.add_lines((revision_id,), parents, lines,
1067
1073
check_content=check_content)[0]
1074
self.inventories._access.flush()
1075
return result
1068
1076
1069
1077
def add_revision(self, revision_id, rev, inv=None, config=None):
1070
1078
"""Add rev to the revision store as revision_id.
1529
1537
"""Commit the contents accrued within the current write group.
1530
1538
1531
1539
:seealso: start_write_group.
1540
1541
:return: it may return an opaque hint that can be passed to 'pack'.
1532
1542
"""
1533
1543
if self._write_group is not self.get_transaction():
1534
1544
# has an unlock or relock occured ?
2340
2350
"""Get Inventory object by revision id."""
2341
2351
return self.iter_inventories([revision_id]).next()
2342
2352
2343
def iter_inventories(self, revision_ids):
2353
def iter_inventories(self, revision_ids, ordering=None):
2344
2354
"""Get many inventories by revision_ids.
2345
2355
2346
2356
This will buffer some or all of the texts used in constructing the
2348
2358
time.
2349
2359
2350
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).
2351
2364
:return: An iterator of inventories.
2352
2365
"""
2353
2366
if ((None in revision_ids)
2354
2367
or (_mod_revision.NULL_REVISION in revision_ids)):
2355
2368
raise ValueError('cannot get null revision inventory')
2356
return self._iter_inventories(revision_ids)
2369
return self._iter_inventories(revision_ids, ordering)
2357
2370
2358
def _iter_inventories(self, revision_ids):
2371
def _iter_inventories(self, revision_ids, ordering):
2359
2372
"""single-document based inventory iteration."""
2360
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:
2361
2375
yield self.deserialise_inventory(revision_id, text)
2362
2376
2363
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'
2381
else:
2382
order_as_requested = False
2364
2383
keys = [(revision_id,) for revision_id in revision_ids]
2365
stream = self.inventories.get_record_stream(keys, 'unordered', True)
2384
if not keys:
2385
return
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)
2366
2390
text_chunks = {}
2367
2391
for record in stream:
2368
2392
if record.storage_kind != 'absent':
2369
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
2396
else:
2397
yield ''.join(chunks), record.key[-1]
2370
2398
else:
2371
2399
raise errors.NoSuchRevision(self, record.key)
2372
for key in keys:
2373
chunks = text_chunks.pop(key)
2374
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]
2405
try:
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
2410
next_key = None
2411
break
2375
2412
2376
2413
def deserialise_inventory(self, revision_id, xml):
2377
2414
"""Transform the xml into an inventory object.
2398
2435
@needs_read_lock
2399
2436
def get_inventory_xml(self, revision_id):
2400
2437
"""Get inventory XML as a file object."""
2401
texts = self._iter_inventory_xmls([revision_id])
2438
texts = self._iter_inventory_xmls([revision_id], 'unordered')
2402
2439
try:
2403
2440
text, revision_id = texts.next()
2404
2441
except StopIteration:
3664
3701
# This is redundant with format.check_conversion_target(), however that
3665
3702
# raises an exception, and we just want to say "False" as in we won't
3666
3703
# support converting between these formats.
3704
if 'IDS_never' in debug.debug_flags:
3705
return False
3667
3706
if source.supports_rich_root() and not target.supports_rich_root():
3668
3707
return False
3669
3708
if (source._format.supports_tree_reference
3670
3709
and not target._format.supports_tree_reference):
3671
3710
return False
3711
if target._fallback_repositories and target._format.supports_chks:
3712
# IDS doesn't know how to copy CHKs for the parent inventories it
3713
# adds to stacked repos.
3714
return False
3715
if 'IDS_always' in debug.debug_flags:
3716
return True
3717
# Only use this code path for local source and target. IDS does far
3718
# too much IO (both bandwidth and roundtrips) over a network.
3719
if not source.bzrdir.transport.base.startswith('file:///'):
3720
return False
3721
if not target.bzrdir.transport.base.startswith('file:///'):
3722
return False
3672
3723
return True
3673
3724
3674
3725
def _get_delta_for_revision(self, tree, parent_ids, basis_id, cache):
3690
3741
deltas.sort()
3691
3742
return deltas[0][1:]
3692
3743
3693
def _get_parent_keys(self, root_key, parent_map):
3694
"""Get the parent keys for a given root id."""
3695
root_id, rev_id = root_key
3696
# Include direct parents of the revision, but only if they used
3697
# the same root_id and are heads.
3698
parent_keys = []
3699
for parent_id in parent_map[rev_id]:
3700
if parent_id == _mod_revision.NULL_REVISION:
3701
continue
3702
if parent_id not in self._revision_id_to_root_id:
3703
# We probably didn't read this revision, go spend the
3704
# extra effort to actually check
3705
try:
3706
tree = self.source.revision_tree(parent_id)
3707
except errors.NoSuchRevision:
3708
# Ghost, fill out _revision_id_to_root_id in case we
3709
# encounter this again.
3710
# But set parent_root_id to None since we don't really know
3711
parent_root_id = None
3712
else:
3713
parent_root_id = tree.get_root_id()
3714
self._revision_id_to_root_id[parent_id] = None
3715
else:
3716
parent_root_id = self._revision_id_to_root_id[parent_id]
3717
if root_id == parent_root_id:
3718
# With stacking we _might_ want to refer to a non-local
3719
# revision, but this code path only applies when we have the
3720
# full content available, so ghosts really are ghosts, not just
3721
# the edge of local data.
3722
parent_keys.append((parent_id,))
3723
else:
3724
# root_id may be in the parent anyway.
3725
try:
3726
tree = self.source.revision_tree(parent_id)
3727
except errors.NoSuchRevision:
3728
# ghost, can't refer to it.
3729
pass
3730
else:
3731
try:
3732
parent_keys.append((tree.inventory[root_id].revision,))
3733
except errors.NoSuchId:
3734
# not in the tree
3735
pass
3736
g = graph.Graph(self.source.revisions)
3737
heads = g.heads(parent_keys)
3738
selected_keys = []
3739
for key in parent_keys:
3740
if key in heads and key not in selected_keys:
3741
selected_keys.append(key)
3742
return tuple([(root_id,)+ key for key in selected_keys])
3743
3744
def _new_root_data_stream(self, root_keys_to_create, parent_map):
3745
for root_key in root_keys_to_create:
3746
parent_keys = self._get_parent_keys(root_key, parent_map)
3747
yield versionedfile.FulltextContentFactory(root_key,
3748
parent_keys, None, '')
3749
3750
3744
def _fetch_batch(self, revision_ids, basis_id, cache):
3751
3745
"""Fetch across a few revisions.
3752
3746
3798
3792
from_texts = self.source.texts
3799
3793
to_texts = self.target.texts
3800
3794
if root_keys_to_create:
3801
root_stream = self._new_root_data_stream(root_keys_to_create,
3802
parent_map)
3795
from bzrlib.fetch import _new_root_data_stream
3796
root_stream = _new_root_data_stream(
3797
root_keys_to_create, self._revision_id_to_root_id, parent_map,
3798
self.source)
3803
3799
to_texts.insert_record_stream(root_stream)
3804
3800
to_texts.insert_record_stream(from_texts.get_record_stream(
3805
3801
text_keys, self.target._format._fetch_order,
3899
3895
# Walk though all revisions; get inventory deltas, copy referenced
3900
3896
# texts that delta references, insert the delta, revision and
3901
3897
# signature.
3902
first_rev = self.source.get_revision(revision_ids[0])
3903
3898
if pb is None:
3904
3899
my_pb = ui.ui_factory.nested_progress_bar()
3905
3900
pb = my_pb
4071
4066
self.file_ids = set([file_id for file_id, _ in
4072
4067
self.text_index.iterkeys()])
4073
4068
# text keys is now grouped by file_id
4074
n_weaves = len(self.file_ids)
4075
files_in_revisions = {}
4076
revisions_of_files = {}
4077
4069
n_versions = len(self.text_index)
4078
4070
progress_bar.update('loading text store', 0, n_versions)
4079
4071
parent_map = self.repository.texts.get_parent_map(self.text_index)
4172
4164
else:
4173
4165
new_pack.set_write_cache_size(1024*1024)
4174
4166
for substream_type, substream in stream:
4167
if 'stream' in debug.debug_flags:
4168
mutter('inserting substream: %s', substream_type)
4175
4169
if substream_type == 'texts':
4176
4170
self.target_repo.texts.insert_record_stream(substream)
4177
4171
elif substream_type == 'inventories':
4181
4175
else:
4182
4176
self._extract_and_insert_inventories(
4183
4177
substream, src_serializer)
4178
elif substream_type == 'inventory-deltas':
4179
self._extract_and_insert_inventory_deltas(
4180
substream, src_serializer)
4184
4181
elif substream_type == 'chk_bytes':
4185
4182
# XXX: This doesn't support conversions, as it assumes the
4186
4183
# conversion was done in the fetch code.
4237
4234
self.target_repo.pack(hint=hint)
4238
4235
return [], set()
4239
4236
4240
def _extract_and_insert_inventories(self, substream, serializer):
4237
def _extract_and_insert_inventory_deltas(self, substream, serializer):
4238
target_rich_root = self.target_repo._format.rich_root_data
4239
target_tree_refs = self.target_repo._format.supports_tree_reference
4240
for record in substream:
4241
# Insert the delta directly
4242
inventory_delta_bytes = record.get_bytes_as('fulltext')
4243
deserialiser = inventory_delta.InventoryDeltaDeserializer()
4244
try:
4245
parse_result = deserialiser.parse_text_bytes(
4246
inventory_delta_bytes)
4247
except inventory_delta.IncompatibleInventoryDelta, err:
4248
trace.mutter("Incompatible delta: %s", err.msg)
4249
raise errors.IncompatibleRevision(self.target_repo._format)
4250
basis_id, new_id, rich_root, tree_refs, inv_delta = parse_result
4251
revision_id = new_id
4252
parents = [key[0] for key in record.parents]
4253
self.target_repo.add_inventory_by_delta(
4254
basis_id, inv_delta, revision_id, parents)
4255
4256
def _extract_and_insert_inventories(self, substream, serializer,
4257
parse_delta=None):
4241
4258
"""Generate a new inventory versionedfile in target, converting data.
4242
4259
4243
4260
The inventory is retrieved from the source, (deserializing it), and
4244
4261
stored in the target (reserializing it in a different format).
4245
4262
"""
4263
target_rich_root = self.target_repo._format.rich_root_data
4264
target_tree_refs = self.target_repo._format.supports_tree_reference
4246
4265
for record in substream:
4266
# It's not a delta, so it must be a fulltext in the source
4267
# serializer's format.
4247
4268
bytes = record.get_bytes_as('fulltext')
4248
4269
revision_id = record.key[0]
4249
4270
inv = serializer.read_inventory_from_string(bytes, revision_id)
4250
4271
parents = [key[0] for key in record.parents]
4251
4272
self.target_repo.add_inventory(revision_id, inv, parents)
4273
# No need to keep holding this full inv in memory when the rest of
4274
# the substream is likely to be all deltas.
4275
del inv
4252
4276
4253
4277
def _extract_and_insert_revisions(self, substream, serializer):
4254
4278
for record in substream:
4303
4327
return [('signatures', signatures), ('revisions', revisions)]
4304
4328
4305
4329
def _generate_root_texts(self, revs):
4306
"""This will be called by __fetch between fetching weave texts and
4330
"""This will be called by get_stream between fetching weave texts and
4307
4331
fetching the inventory weave.
4308
4309
Subclasses should override this if they need to generate root texts
4310
after fetching weave texts.
4311
4332
"""
4312
4333
if self._rich_root_upgrade():
4313
4334
import bzrlib.fetch
4345
4366
# will be valid.
4346
4367
for _ in self._generate_root_texts(revs):
4347
4368
yield _
4348
# NB: This currently reopens the inventory weave in source;
4349
# using a single stream interface instead would avoid this.
4350
from_weave = self.from_repository.inventories
4351
4369
# we fetch only the referenced inventories because we do not
4352
4370
# know for unselected inventories whether all their required
4353
4371
# texts are present in the other repository - it could be
4392
4410
if not keys:
4393
4411
# No need to stream something we don't have
4394
4412
continue
4413
if substream_kind == 'inventories':
4414
# Some missing keys are genuinely ghosts, filter those out.
4415
present = self.from_repository.inventories.get_parent_map(keys)
4416
revs = [key[0] for key in present]
4417
# Get the inventory stream more-or-less as we do for the
4418
# original stream; there's no reason to assume that records
4419
# direct from the source will be suitable for the sink. (Think
4420
# e.g. 2a -> 1.9-rich-root).
4421
for info in self._get_inventory_stream(revs, missing=True):
4422
yield info
4423
continue
4424
4395
4425
# Ask for full texts always so that we don't need more round trips
4396
4426
# after this stream.
4397
4427
# Some of the missing keys are genuinely ghosts, so filter absent
4412
4442
return (not self.from_repository._format.rich_root_data and
4413
4443
self.to_format.rich_root_data)
4414
4444
4415
def _get_inventory_stream(self, revision_ids):
4445
def _get_inventory_stream(self, revision_ids, missing=False):
4416
4446
from_format = self.from_repository._format
4417
if (from_format.supports_chks and self.to_format.supports_chks
4418
and (from_format._serializer == self.to_format._serializer)):
4419
# Both sides support chks, and they use the same serializer, so it
4420
# is safe to transmit the chk pages and inventory pages across
4421
# as-is.
4422
return self._get_chk_inventory_stream(revision_ids)
4423
elif (not from_format.supports_chks):
4424
# Source repository doesn't support chks. So we can transmit the
4425
# inventories 'as-is' and either they are just accepted on the
4426
# target, or the Sink will properly convert it.
4427
return self._get_simple_inventory_stream(revision_ids)
4447
if (from_format.supports_chks and self.to_format.supports_chks and
4448
from_format.network_name() == self.to_format.network_name()):
4449
raise AssertionError(
4450
"this case should be handled by GroupCHKStreamSource")
4451
elif 'forceinvdeltas' in debug.debug_flags:
4452
return self._get_convertable_inventory_stream(revision_ids,
4453
delta_versus_null=missing)
4454
elif from_format.network_name() == self.to_format.network_name():
4455
# Same format.
4456
return self._get_simple_inventory_stream(revision_ids,
4457
missing=missing)
4458
elif (not from_format.supports_chks and not self.to_format.supports_chks
4459
and from_format._serializer == self.to_format._serializer):
4460
# Essentially the same format.
4461
return self._get_simple_inventory_stream(revision_ids,
4462
missing=missing)
4428
4463
else:
4429
# XXX: Hack to make not-chk->chk fetch: copy the inventories as
4430
# inventories. Note that this should probably be done somehow
4431
# as part of bzrlib.repository.StreamSink. Except JAM couldn't
4432
# figure out how a non-chk repository could possibly handle
4433
# deserializing an inventory stream from a chk repo, as it
4434
# doesn't have a way to understand individual pages.
4435
return self._get_convertable_inventory_stream(revision_ids)
4464
# Any time we switch serializations, we want to use an
4465
# inventory-delta based approach.
4466
return self._get_convertable_inventory_stream(revision_ids,
4467
delta_versus_null=missing)
4436
4468
4437
def _get_simple_inventory_stream(self, revision_ids):
4469
def _get_simple_inventory_stream(self, revision_ids, missing=False):
4470
# NB: This currently reopens the inventory weave in source;
4471
# using a single stream interface instead would avoid this.
4438
4472
from_weave = self.from_repository.inventories
4473
if missing:
4474
delta_closure = True
4475
else:
4476
delta_closure = not self.delta_on_metadata()
4439
4477
yield ('inventories', from_weave.get_record_stream(
4440
4478
[(rev_id,) for rev_id in revision_ids],
4441
self.inventory_fetch_order(),
4442
not self.delta_on_metadata()))
4443
4444
def _get_chk_inventory_stream(self, revision_ids):
4445
"""Fetch the inventory texts, along with the associated chk maps."""
4446
# We want an inventory outside of the search set, so that we can filter
4447
# out uninteresting chk pages. For now we use
4448
# _find_revision_outside_set, but if we had a Search with cut_revs, we
4449
# could use that instead.
4450
start_rev_id = self.from_repository._find_revision_outside_set(
4451
revision_ids)
4452
start_rev_key = (start_rev_id,)
4453
inv_keys_to_fetch = [(rev_id,) for rev_id in revision_ids]
4454
if start_rev_id != _mod_revision.NULL_REVISION:
4455
inv_keys_to_fetch.append((start_rev_id,))
4456
# Any repo that supports chk_bytes must also support out-of-order
4457
# insertion. At least, that is how we expect it to work
4458
# We use get_record_stream instead of iter_inventories because we want
4459
# to be able to insert the stream as well. We could instead fetch
4460
# allowing deltas, and then iter_inventories, but we don't know whether
4461
# source or target is more 'local' anway.
4462
inv_stream = self.from_repository.inventories.get_record_stream(
4463
inv_keys_to_fetch, 'unordered',
4464
True) # We need them as full-texts so we can find their references
4465
uninteresting_chk_roots = set()
4466
interesting_chk_roots = set()
4467
def filter_inv_stream(inv_stream):
4468
for idx, record in enumerate(inv_stream):
4469
### child_pb.update('fetch inv', idx, len(inv_keys_to_fetch))
4470
bytes = record.get_bytes_as('fulltext')
4471
chk_inv = inventory.CHKInventory.deserialise(
4472
self.from_repository.chk_bytes, bytes, record.key)
4473
if record.key == start_rev_key:
4474
uninteresting_chk_roots.add(chk_inv.id_to_entry.key())
4475
p_id_map = chk_inv.parent_id_basename_to_file_id
4476
if p_id_map is not None:
4477
uninteresting_chk_roots.add(p_id_map.key())
4478
else:
4479
yield record
4480
interesting_chk_roots.add(chk_inv.id_to_entry.key())
4481
p_id_map = chk_inv.parent_id_basename_to_file_id
4482
if p_id_map is not None:
4483
interesting_chk_roots.add(p_id_map.key())
4484
### pb.update('fetch inventory', 0, 2)
4485
yield ('inventories', filter_inv_stream(inv_stream))
4486
# Now that we have worked out all of the interesting root nodes, grab
4487
# all of the interesting pages and insert them
4488
### pb.update('fetch inventory', 1, 2)
4489
interesting = chk_map.iter_interesting_nodes(
4490
self.from_repository.chk_bytes, interesting_chk_roots,
4491
uninteresting_chk_roots)
4492
def to_stream_adapter():
4493
"""Adapt the iter_interesting_nodes result to a single stream.
4494
4495
iter_interesting_nodes returns records as it processes them, along
4496
with keys. However, we only want to return the records themselves.
4497
"""
4498
for record, items in interesting:
4499
if record is not None:
4500
yield record
4501
# XXX: We could instead call get_record_stream(records.keys())
4502
# ATM, this will always insert the records as fulltexts, and
4503
# requires that you can hang on to records once you have gone
4504
# on to the next one. Further, it causes the target to
4505
# recompress the data. Testing shows it to be faster than
4506
# requesting the records again, though.
4507
yield ('chk_bytes', to_stream_adapter())
4508
### pb.update('fetch inventory', 2, 2)
4509
4510
def _get_convertable_inventory_stream(self, revision_ids):
4511
# XXX: One of source or target is using chks, and they don't have
4512
# compatible serializations. The StreamSink code expects to be
4513
# able to convert on the target, so we need to put
4514
# bytes-on-the-wire that can be converted
4515
yield ('inventories', self._stream_invs_as_fulltexts(revision_ids))
4516
4517
def _stream_invs_as_fulltexts(self, revision_ids):
4479
self.inventory_fetch_order(), delta_closure))
4480
4481
def _get_convertable_inventory_stream(self, revision_ids,
4482
delta_versus_null=False):
4483
# The source is using CHKs, but the target either doesn't or it has a
4484
# different serializer. The StreamSink code expects to be able to
4485
# convert on the target, so we need to put bytes-on-the-wire that can
4486
# be converted. That means inventory deltas (if the remote is <1.19,
4487
# RemoteStreamSink will fallback to VFS to insert the deltas).
4488
yield ('inventory-deltas',
4489
self._stream_invs_as_deltas(revision_ids,
4490
delta_versus_null=delta_versus_null))
4491
4492
def _stream_invs_as_deltas(self, revision_ids, delta_versus_null=False):
4493
"""Return a stream of inventory-deltas for the given rev ids.
4494
4495
:param revision_ids: The list of inventories to transmit
4496
:param delta_versus_null: Don't try to find a minimal delta for this
4497
entry, instead compute the delta versus the NULL_REVISION. This
4498
effectively streams a complete inventory. Used for stuff like
4499
filling in missing parents, etc.
4500
"""
4518
4501
from_repo = self.from_repository
4519
from_serializer = from_repo._format._serializer
4520
4502
revision_keys = [(rev_id,) for rev_id in revision_ids]
4521
4503
parent_map = from_repo.inventories.get_parent_map(revision_keys)
4522
for inv in self.from_repository.iter_inventories(revision_ids):
4523
# XXX: This is a bit hackish, but it works. Basically,
4524
# CHKSerializer 'accidentally' supports
4525
# read/write_inventory_to_string, even though that is never
4526
# the format that is stored on disk. It *does* give us a
4527
# single string representation for an inventory, so live with
4528
# it for now.
4529
# This would be far better if we had a 'serialized inventory
4530
# delta' form. Then we could use 'inventory._make_delta', and
4531
# transmit that. This would both be faster to generate, and
4532
# result in fewer bytes-on-the-wire.
4533
as_bytes = from_serializer.write_inventory_to_string(inv)
4504
# XXX: possibly repos could implement a more efficient iter_inv_deltas
4505
# method...
4506
inventories = self.from_repository.iter_inventories(
4507
revision_ids, 'topological')
4508
format = from_repo._format
4509
invs_sent_so_far = set([_mod_revision.NULL_REVISION])
4510
inventory_cache = lru_cache.LRUCache(50)
4511
null_inventory = from_repo.revision_tree(
4512
_mod_revision.NULL_REVISION).inventory
4513
# XXX: ideally the rich-root/tree-refs flags would be per-revision, not
4514
# per-repo (e.g. streaming a non-rich-root revision out of a rich-root
4515
# repo back into a non-rich-root repo ought to be allowed)
4516
serializer = inventory_delta.InventoryDeltaSerializer(
4517
versioned_root=format.rich_root_data,
4518
tree_references=format.supports_tree_reference)
4519
for inv in inventories:
4534
4520
key = (inv.revision_id,)
4535
4521
parent_keys = parent_map.get(key, ())
4522
delta = None
4523
if not delta_versus_null and parent_keys:
4524
# The caller did not ask for complete inventories and we have
4525
# some parents that we can delta against. Make a delta against
4526
# each parent so that we can find the smallest.
4527
parent_ids = [parent_key[0] for parent_key in parent_keys]
4528
for parent_id in parent_ids:
4529
if parent_id not in invs_sent_so_far:
4530
# We don't know that the remote side has this basis, so
4531
# we can't use it.
4532
continue
4533
if parent_id == _mod_revision.NULL_REVISION:
4534
parent_inv = null_inventory
4535
else:
4536
parent_inv = inventory_cache.get(parent_id, None)
4537
if parent_inv is None:
4538
parent_inv = from_repo.get_inventory(parent_id)
4539
candidate_delta = inv._make_delta(parent_inv)
4540
if (delta is None or
4541
len(delta) > len(candidate_delta)):
4542
delta = candidate_delta
4543
basis_id = parent_id
4544
if delta is None:
4545
# Either none of the parents ended up being suitable, or we
4546
# were asked to delta against NULL
4547
basis_id = _mod_revision.NULL_REVISION
4548
delta = inv._make_delta(null_inventory)
4549
invs_sent_so_far.add(inv.revision_id)
4550
inventory_cache[inv.revision_id] = inv
4551
delta_serialized = ''.join(
4552
serializer.delta_to_lines(basis_id, key[-1], delta))
4536
4553
yield versionedfile.FulltextContentFactory(
4537
key, parent_keys, None, as_bytes)
4554
key, parent_keys, None, delta_serialized)
4538
4555
4539
4556
4540
4557
def _iter_for_revno(repo, partial_history_cache, stop_index=None,
Loggerhead is a web-based interface for Breezy
Version: 2.0.1