1146
1167
# The old API returned a list, should this actually be a set?
1147
1168
return parent_map.keys()
1170
def _check_inventories(self, checker):
1171
"""Check the inventories found from the revision scan.
1173
This is responsible for verifying the sha1 of inventories and
1174
creating a pending_keys set that covers data referenced by inventories.
1176
bar = ui.ui_factory.nested_progress_bar()
1178
self._do_check_inventories(checker, bar)
1182
def _do_check_inventories(self, checker, bar):
1183
"""Helper for _check_inventories."""
1185
keys = {'chk_bytes':set(), 'inventories':set(), 'texts':set()}
1186
kinds = ['chk_bytes', 'texts']
1187
count = len(checker.pending_keys)
1188
bar.update("inventories", 0, 2)
1189
current_keys = checker.pending_keys
1190
checker.pending_keys = {}
1191
# Accumulate current checks.
1192
for key in current_keys:
1193
if key[0] != 'inventories' and key[0] not in kinds:
1194
checker._report_items.append('unknown key type %r' % (key,))
1195
keys[key[0]].add(key[1:])
1196
if keys['inventories']:
1197
# NB: output order *should* be roughly sorted - topo or
1198
# inverse topo depending on repository - either way decent
1199
# to just delta against. However, pre-CHK formats didn't
1200
# try to optimise inventory layout on disk. As such the
1201
# pre-CHK code path does not use inventory deltas.
1203
for record in self.inventories.check(keys=keys['inventories']):
1204
if record.storage_kind == 'absent':
1205
checker._report_items.append(
1206
'Missing inventory {%s}' % (record.key,))
1208
last_object = self._check_record('inventories', record,
1209
checker, last_object,
1210
current_keys[('inventories',) + record.key])
1211
del keys['inventories']
1214
bar.update("texts", 1)
1215
while (checker.pending_keys or keys['chk_bytes']
1217
# Something to check.
1218
current_keys = checker.pending_keys
1219
checker.pending_keys = {}
1220
# Accumulate current checks.
1221
for key in current_keys:
1222
if key[0] not in kinds:
1223
checker._report_items.append('unknown key type %r' % (key,))
1224
keys[key[0]].add(key[1:])
1225
# Check the outermost kind only - inventories || chk_bytes || texts
1229
for record in getattr(self, kind).check(keys=keys[kind]):
1230
if record.storage_kind == 'absent':
1231
checker._report_items.append(
1232
'Missing %s {%s}' % (kind, record.key,))
1234
last_object = self._check_record(kind, record,
1235
checker, last_object, current_keys[(kind,) + record.key])
1239
def _check_record(self, kind, record, checker, last_object, item_data):
1240
"""Check a single text from this repository."""
1241
if kind == 'inventories':
1242
rev_id = record.key[0]
1243
inv = self.deserialise_inventory(rev_id,
1244
record.get_bytes_as('fulltext'))
1245
if last_object is not None:
1246
delta = inv._make_delta(last_object)
1247
for old_path, path, file_id, ie in delta:
1250
ie.check(checker, rev_id, inv)
1252
for path, ie in inv.iter_entries():
1253
ie.check(checker, rev_id, inv)
1254
if self._format.fast_deltas:
1256
elif kind == 'chk_bytes':
1257
# No code written to check chk_bytes for this repo format.
1258
checker._report_items.append(
1259
'unsupported key type chk_bytes for %s' % (record.key,))
1260
elif kind == 'texts':
1261
self._check_text(record, checker, item_data)
1263
checker._report_items.append(
1264
'unknown key type %s for %s' % (kind, record.key))
1266
def _check_text(self, record, checker, item_data):
1267
"""Check a single text."""
1268
# Check it is extractable.
1269
# TODO: check length.
1270
if record.storage_kind == 'chunked':
1271
chunks = record.get_bytes_as(record.storage_kind)
1272
sha1 = osutils.sha_strings(chunks)
1273
length = sum(map(len, chunks))
1275
content = record.get_bytes_as('fulltext')
1276
sha1 = osutils.sha_string(content)
1277
length = len(content)
1278
if item_data and sha1 != item_data[1]:
1279
checker._report_items.append(
1280
'sha1 mismatch: %s has sha1 %s expected %s referenced by %s' %
1281
(record.key, sha1, item_data[1], item_data[2]))
1150
1284
def create(a_bzrdir):
1151
1285
"""Construct the current default format repository in a_bzrdir."""
1715
1858
@needs_read_lock
1716
1859
def get_revisions(self, revision_ids):
1717
"""Get many revisions at once."""
1860
"""Get many revisions at once.
1862
Repositories that need to check data on every revision read should
1863
subclass this method.
1718
1865
return self._get_revisions(revision_ids)
1720
1867
@needs_read_lock
1721
1868
def _get_revisions(self, revision_ids):
1722
1869
"""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)
1871
for revid, rev in self._iter_revisions(revision_ids):
1873
raise errors.NoSuchRevision(self, revid)
1875
return [revs[revid] for revid in revision_ids]
1877
def _iter_revisions(self, revision_ids):
1878
"""Iterate over revision objects.
1880
:param revision_ids: An iterable of revisions to examine. None may be
1881
passed to request all revisions known to the repository. Note that
1882
not all repositories can find unreferenced revisions; for those
1883
repositories only referenced ones will be returned.
1884
:return: An iterator of (revid, revision) tuples. Absent revisions (
1885
those asked for but not available) are returned as (revid, None).
1887
if revision_ids is None:
1888
revision_ids = self.all_revision_ids()
1890
for rev_id in revision_ids:
1891
if not rev_id or not isinstance(rev_id, basestring):
1892
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1726
1893
keys = [(key,) for key in revision_ids]
1727
1894
stream = self.revisions.get_record_stream(keys, 'unordered', True)
1729
1895
for record in stream:
1896
revid = record.key[0]
1730
1897
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]
1900
text = record.get_bytes_as('fulltext')
1901
rev = self._serializer.read_revision_from_string(text)
1737
1904
@needs_read_lock
1738
1905
def get_revision_xml(self, revision_id):
2210
2376
:param revision_ids: The expected revision ids of the inventories.
2377
:param ordering: optional ordering, e.g. 'topological'. If not
2378
specified, the order of revision_ids will be preserved (by
2379
buffering if necessary).
2211
2380
:return: An iterator of inventories.
2213
2382
if ((None in revision_ids)
2214
2383
or (_mod_revision.NULL_REVISION in revision_ids)):
2215
2384
raise ValueError('cannot get null revision inventory')
2216
return self._iter_inventories(revision_ids)
2385
return self._iter_inventories(revision_ids, ordering)
2218
def _iter_inventories(self, revision_ids):
2387
def _iter_inventories(self, revision_ids, ordering):
2219
2388
"""single-document based inventory iteration."""
2220
for text, revision_id in self._iter_inventory_xmls(revision_ids):
2389
inv_xmls = self._iter_inventory_xmls(revision_ids, ordering)
2390
for text, revision_id in inv_xmls:
2221
2391
yield self.deserialise_inventory(revision_id, text)
2223
def _iter_inventory_xmls(self, revision_ids):
2393
def _iter_inventory_xmls(self, revision_ids, ordering):
2394
if ordering is None:
2395
order_as_requested = True
2396
ordering = 'unordered'
2398
order_as_requested = False
2224
2399
keys = [(revision_id,) for revision_id in revision_ids]
2225
stream = self.inventories.get_record_stream(keys, 'unordered', True)
2402
if order_as_requested:
2403
key_iter = iter(keys)
2404
next_key = key_iter.next()
2405
stream = self.inventories.get_record_stream(keys, ordering, True)
2226
2406
text_chunks = {}
2227
2407
for record in stream:
2228
2408
if record.storage_kind != 'absent':
2229
text_chunks[record.key] = record.get_bytes_as('chunked')
2409
chunks = record.get_bytes_as('chunked')
2410
if order_as_requested:
2411
text_chunks[record.key] = chunks
2413
yield ''.join(chunks), record.key[-1]
2231
2415
raise errors.NoSuchRevision(self, record.key)
2233
chunks = text_chunks.pop(key)
2234
yield ''.join(chunks), key[-1]
2416
if order_as_requested:
2417
# Yield as many results as we can while preserving order.
2418
while next_key in text_chunks:
2419
chunks = text_chunks.pop(next_key)
2420
yield ''.join(chunks), next_key[-1]
2422
next_key = key_iter.next()
2423
except StopIteration:
2424
# We still want to fully consume the get_record_stream,
2425
# just in case it is not actually finished at this point
2236
2429
def deserialise_inventory(self, revision_id, xml):
2237
2430
"""Transform the xml into an inventory object.
3514
3736
# This is redundant with format.check_conversion_target(), however that
3515
3737
# raises an exception, and we just want to say "False" as in we won't
3516
3738
# support converting between these formats.
3739
if 'IDS_never' in debug.debug_flags:
3517
3741
if source.supports_rich_root() and not target.supports_rich_root():
3519
3743
if (source._format.supports_tree_reference
3520
3744
and not target._format.supports_tree_reference):
3746
if target._fallback_repositories and target._format.supports_chks:
3747
# IDS doesn't know how to copy CHKs for the parent inventories it
3748
# adds to stacked repos.
3750
if 'IDS_always' in debug.debug_flags:
3752
# Only use this code path for local source and target. IDS does far
3753
# too much IO (both bandwidth and roundtrips) over a network.
3754
if not source.bzrdir.transport.base.startswith('file:///'):
3756
if not target.bzrdir.transport.base.startswith('file:///'):
3524
def _get_delta_for_revision(self, tree, parent_ids, basis_id, cache):
3760
def _get_trees(self, revision_ids, cache):
3762
for rev_id in revision_ids:
3764
possible_trees.append((rev_id, cache[rev_id]))
3766
# Not cached, but inventory might be present anyway.
3768
tree = self.source.revision_tree(rev_id)
3769
except errors.NoSuchRevision:
3770
# Nope, parent is ghost.
3773
cache[rev_id] = tree
3774
possible_trees.append((rev_id, tree))
3775
return possible_trees
3777
def _get_delta_for_revision(self, tree, parent_ids, possible_trees):
3525
3778
"""Get the best delta and base for this revision.
3527
3780
:return: (basis_id, delta)
3529
possible_trees = [(parent_id, cache[parent_id])
3530
for parent_id in parent_ids
3531
if parent_id in cache]
3532
if len(possible_trees) == 0:
3533
# There either aren't any parents, or the parents aren't in the
3534
# cache, so just use the last converted tree
3535
possible_trees.append((basis_id, cache[basis_id]))
3783
# Generate deltas against each tree, to find the shortest.
3784
texts_possibly_new_in_tree = set()
3537
3785
for basis_id, basis_tree in possible_trees:
3538
3786
delta = tree.inventory._make_delta(basis_tree.inventory)
3787
for old_path, new_path, file_id, new_entry in delta:
3788
if new_path is None:
3789
# This file_id isn't present in the new rev, so we don't
3793
# Rich roots are handled elsewhere...
3795
kind = new_entry.kind
3796
if kind != 'directory' and kind != 'file':
3797
# No text record associated with this inventory entry.
3799
# This is a directory or file that has changed somehow.
3800
texts_possibly_new_in_tree.add((file_id, new_entry.revision))
3539
3801
deltas.append((len(delta), basis_id, delta))
3541
3803
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
def _fetch_batch(self, revision_ids, basis_id, cache):
3805
def _fetch_parent_invs_for_stacking(self, parent_map, cache):
3806
"""Find all parent revisions that are absent, but for which the
3807
inventory is present, and copy those inventories.
3809
This is necessary to preserve correctness when the source is stacked
3810
without fallbacks configured. (Note that in cases like upgrade the
3811
source may be not have _fallback_repositories even though it is
3815
for parents in parent_map.values():
3816
parent_revs.update(parents)
3817
present_parents = self.source.get_parent_map(parent_revs)
3818
absent_parents = set(parent_revs).difference(present_parents)
3819
parent_invs_keys_for_stacking = self.source.inventories.get_parent_map(
3820
(rev_id,) for rev_id in absent_parents)
3821
parent_inv_ids = [key[-1] for key in parent_invs_keys_for_stacking]
3822
for parent_tree in self.source.revision_trees(parent_inv_ids):
3823
current_revision_id = parent_tree.get_revision_id()
3824
parents_parents_keys = parent_invs_keys_for_stacking[
3825
(current_revision_id,)]
3826
parents_parents = [key[-1] for key in parents_parents_keys]
3827
basis_id = _mod_revision.NULL_REVISION
3828
basis_tree = self.source.revision_tree(basis_id)
3829
delta = parent_tree.inventory._make_delta(basis_tree.inventory)
3830
self.target.add_inventory_by_delta(
3831
basis_id, delta, current_revision_id, parents_parents)
3832
cache[current_revision_id] = parent_tree
3834
def _fetch_batch(self, revision_ids, basis_id, cache, a_graph=None):
3601
3835
"""Fetch across a few revisions.
3603
3837
:param revision_ids: The revisions to copy
3604
3838
:param basis_id: The revision_id of a tree that must be in cache, used
3605
3839
as a basis for delta when no other base is available
3606
3840
:param cache: A cache of RevisionTrees that we can use.
3841
:param a_graph: A Graph object to determine the heads() of the
3842
rich-root data stream.
3607
3843
:return: The revision_id of the last converted tree. The RevisionTree
3608
3844
for it will be in cache
3615
3851
pending_deltas = []
3616
3852
pending_revisions = []
3617
3853
parent_map = self.source.get_parent_map(revision_ids)
3854
self._fetch_parent_invs_for_stacking(parent_map, cache)
3855
self.source._safe_to_return_from_cache = True
3618
3856
for tree in self.source.revision_trees(revision_ids):
3857
# Find a inventory delta for this revision.
3858
# Find text entries that need to be copied, too.
3619
3859
current_revision_id = tree.get_revision_id()
3620
3860
parent_ids = parent_map.get(current_revision_id, ())
3861
parent_trees = self._get_trees(parent_ids, cache)
3862
possible_trees = list(parent_trees)
3863
if len(possible_trees) == 0:
3864
# There either aren't any parents, or the parents are ghosts,
3865
# so just use the last converted tree.
3866
possible_trees.append((basis_id, cache[basis_id]))
3621
3867
basis_id, delta = self._get_delta_for_revision(tree, parent_ids,
3869
revision = self.source.get_revision(current_revision_id)
3870
pending_deltas.append((basis_id, delta,
3871
current_revision_id, revision.parent_ids))
3623
3872
if self._converting_to_rich_root:
3624
3873
self._revision_id_to_root_id[current_revision_id] = \
3625
3874
tree.get_root_id()
3626
# Find text entries that need to be copied
3875
# Determine which texts are in present in this revision but not in
3876
# any of the available parents.
3877
texts_possibly_new_in_tree = set()
3627
3878
for old_path, new_path, file_id, entry in delta:
3628
if new_path is not None:
3631
if not self.target.supports_rich_root():
3632
# The target doesn't support rich root, so we don't
3635
if self._converting_to_rich_root:
3636
# This can't be copied normally, we have to insert
3638
root_keys_to_create.add((file_id, entry.revision))
3640
text_keys.add((file_id, entry.revision))
3641
revision = self.source.get_revision(current_revision_id)
3642
pending_deltas.append((basis_id, delta,
3643
current_revision_id, revision.parent_ids))
3879
if new_path is None:
3880
# This file_id isn't present in the new rev
3884
if not self.target.supports_rich_root():
3885
# The target doesn't support rich root, so we don't
3888
if self._converting_to_rich_root:
3889
# This can't be copied normally, we have to insert
3891
root_keys_to_create.add((file_id, entry.revision))
3894
texts_possibly_new_in_tree.add((file_id, entry.revision))
3895
for basis_id, basis_tree in possible_trees:
3896
basis_inv = basis_tree.inventory
3897
for file_key in list(texts_possibly_new_in_tree):
3898
file_id, file_revision = file_key
3900
entry = basis_inv[file_id]
3901
except errors.NoSuchId:
3903
if entry.revision == file_revision:
3904
texts_possibly_new_in_tree.remove(file_key)
3905
text_keys.update(texts_possibly_new_in_tree)
3644
3906
pending_revisions.append(revision)
3645
3907
cache[current_revision_id] = tree
3646
3908
basis_id = current_revision_id
3909
self.source._safe_to_return_from_cache = False
3647
3910
# Copy file texts
3648
3911
from_texts = self.source.texts
3649
3912
to_texts = self.target.texts
3650
3913
if root_keys_to_create:
3651
root_stream = self._new_root_data_stream(root_keys_to_create,
3914
root_stream = _mod_fetch._new_root_data_stream(
3915
root_keys_to_create, self._revision_id_to_root_id, parent_map,
3916
self.source, graph=a_graph)
3653
3917
to_texts.insert_record_stream(root_stream)
3654
3918
to_texts.insert_record_stream(from_texts.get_record_stream(
3655
3919
text_keys, self.target._format._fetch_order,
4246
4587
return (not self.from_repository._format.rich_root_data and
4247
4588
self.to_format.rich_root_data)
4249
def _get_inventory_stream(self, revision_ids):
4590
def _get_inventory_stream(self, revision_ids, missing=False):
4250
4591
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)
4592
if (from_format.supports_chks and self.to_format.supports_chks and
4593
from_format.network_name() == self.to_format.network_name()):
4594
raise AssertionError(
4595
"this case should be handled by GroupCHKStreamSource")
4596
elif 'forceinvdeltas' in debug.debug_flags:
4597
return self._get_convertable_inventory_stream(revision_ids,
4598
delta_versus_null=missing)
4599
elif from_format.network_name() == self.to_format.network_name():
4601
return self._get_simple_inventory_stream(revision_ids,
4603
elif (not from_format.supports_chks and not self.to_format.supports_chks
4604
and from_format._serializer == self.to_format._serializer):
4605
# Essentially the same format.
4606
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)
4609
# Any time we switch serializations, we want to use an
4610
# inventory-delta based approach.
4611
return self._get_convertable_inventory_stream(revision_ids,
4612
delta_versus_null=missing)
4271
def _get_simple_inventory_stream(self, revision_ids):
4614
def _get_simple_inventory_stream(self, revision_ids, missing=False):
4615
# NB: This currently reopens the inventory weave in source;
4616
# using a single stream interface instead would avoid this.
4272
4617
from_weave = self.from_repository.inventories
4619
delta_closure = True
4621
delta_closure = not self.delta_on_metadata()
4273
4622
yield ('inventories', from_weave.get_record_stream(
4274
4623
[(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):
4624
self.inventory_fetch_order(), delta_closure))
4626
def _get_convertable_inventory_stream(self, revision_ids,
4627
delta_versus_null=False):
4628
# The two formats are sufficiently different that there is no fast
4629
# path, so we need to send just inventorydeltas, which any
4630
# sufficiently modern client can insert into any repository.
4631
# The StreamSink code expects to be able to
4632
# convert on the target, so we need to put bytes-on-the-wire that can
4633
# be converted. That means inventory deltas (if the remote is <1.19,
4634
# RemoteStreamSink will fallback to VFS to insert the deltas).
4635
yield ('inventory-deltas',
4636
self._stream_invs_as_deltas(revision_ids,
4637
delta_versus_null=delta_versus_null))
4639
def _stream_invs_as_deltas(self, revision_ids, delta_versus_null=False):
4640
"""Return a stream of inventory-deltas for the given rev ids.
4642
:param revision_ids: The list of inventories to transmit
4643
:param delta_versus_null: Don't try to find a minimal delta for this
4644
entry, instead compute the delta versus the NULL_REVISION. This
4645
effectively streams a complete inventory. Used for stuff like
4646
filling in missing parents, etc.
4352
4648
from_repo = self.from_repository
4353
from_serializer = from_repo._format._serializer
4354
4649
revision_keys = [(rev_id,) for rev_id in revision_ids]
4355
4650
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)
4651
# XXX: possibly repos could implement a more efficient iter_inv_deltas
4653
inventories = self.from_repository.iter_inventories(
4654
revision_ids, 'topological')
4655
format = from_repo._format
4656
invs_sent_so_far = set([_mod_revision.NULL_REVISION])
4657
inventory_cache = lru_cache.LRUCache(50)
4658
null_inventory = from_repo.revision_tree(
4659
_mod_revision.NULL_REVISION).inventory
4660
# XXX: ideally the rich-root/tree-refs flags would be per-revision, not
4661
# per-repo (e.g. streaming a non-rich-root revision out of a rich-root
4662
# repo back into a non-rich-root repo ought to be allowed)
4663
serializer = inventory_delta.InventoryDeltaSerializer(
4664
versioned_root=format.rich_root_data,
4665
tree_references=format.supports_tree_reference)
4666
for inv in inventories:
4368
4667
key = (inv.revision_id,)
4369
4668
parent_keys = parent_map.get(key, ())
4670
if not delta_versus_null and parent_keys:
4671
# The caller did not ask for complete inventories and we have
4672
# some parents that we can delta against. Make a delta against
4673
# each parent so that we can find the smallest.
4674
parent_ids = [parent_key[0] for parent_key in parent_keys]
4675
for parent_id in parent_ids:
4676
if parent_id not in invs_sent_so_far:
4677
# We don't know that the remote side has this basis, so
4680
if parent_id == _mod_revision.NULL_REVISION:
4681
parent_inv = null_inventory
4683
parent_inv = inventory_cache.get(parent_id, None)
4684
if parent_inv is None:
4685
parent_inv = from_repo.get_inventory(parent_id)
4686
candidate_delta = inv._make_delta(parent_inv)
4687
if (delta is None or
4688
len(delta) > len(candidate_delta)):
4689
delta = candidate_delta
4690
basis_id = parent_id
4692
# Either none of the parents ended up being suitable, or we
4693
# were asked to delta against NULL
4694
basis_id = _mod_revision.NULL_REVISION
4695
delta = inv._make_delta(null_inventory)
4696
invs_sent_so_far.add(inv.revision_id)
4697
inventory_cache[inv.revision_id] = inv
4698
delta_serialized = ''.join(
4699
serializer.delta_to_lines(basis_id, key[-1], delta))
4370
4700
yield versionedfile.FulltextContentFactory(
4371
key, parent_keys, None, as_bytes)
4701
key, parent_keys, None, delta_serialized)
4374
4704
def _iter_for_revno(repo, partial_history_cache, stop_index=None,