1168
1146
# The old API returned a list, should this actually be a set?
1169
1147
return parent_map.keys()
1171
def _check_inventories(self, checker):
1172
"""Check the inventories found from the revision scan.
1174
This is responsible for verifying the sha1 of inventories and
1175
creating a pending_keys set that covers data referenced by inventories.
1177
bar = ui.ui_factory.nested_progress_bar()
1179
self._do_check_inventories(checker, bar)
1183
def _do_check_inventories(self, checker, bar):
1184
"""Helper for _check_inventories."""
1186
keys = {'chk_bytes':set(), 'inventories':set(), 'texts':set()}
1187
kinds = ['chk_bytes', 'texts']
1188
count = len(checker.pending_keys)
1189
bar.update("inventories", 0, 2)
1190
current_keys = checker.pending_keys
1191
checker.pending_keys = {}
1192
# Accumulate current checks.
1193
for key in current_keys:
1194
if key[0] != 'inventories' and key[0] not in kinds:
1195
checker._report_items.append('unknown key type %r' % (key,))
1196
keys[key[0]].add(key[1:])
1197
if keys['inventories']:
1198
# NB: output order *should* be roughly sorted - topo or
1199
# inverse topo depending on repository - either way decent
1200
# to just delta against. However, pre-CHK formats didn't
1201
# try to optimise inventory layout on disk. As such the
1202
# pre-CHK code path does not use inventory deltas.
1204
for record in self.inventories.check(keys=keys['inventories']):
1205
if record.storage_kind == 'absent':
1206
checker._report_items.append(
1207
'Missing inventory {%s}' % (record.key,))
1209
last_object = self._check_record('inventories', record,
1210
checker, last_object,
1211
current_keys[('inventories',) + record.key])
1212
del keys['inventories']
1215
bar.update("texts", 1)
1216
while (checker.pending_keys or keys['chk_bytes']
1218
# Something to check.
1219
current_keys = checker.pending_keys
1220
checker.pending_keys = {}
1221
# Accumulate current checks.
1222
for key in current_keys:
1223
if key[0] not in kinds:
1224
checker._report_items.append('unknown key type %r' % (key,))
1225
keys[key[0]].add(key[1:])
1226
# Check the outermost kind only - inventories || chk_bytes || texts
1230
for record in getattr(self, kind).check(keys=keys[kind]):
1231
if record.storage_kind == 'absent':
1232
checker._report_items.append(
1233
'Missing %s {%s}' % (kind, record.key,))
1235
last_object = self._check_record(kind, record,
1236
checker, last_object, current_keys[(kind,) + record.key])
1240
def _check_record(self, kind, record, checker, last_object, item_data):
1241
"""Check a single text from this repository."""
1242
if kind == 'inventories':
1243
rev_id = record.key[0]
1244
inv = self._deserialise_inventory(rev_id,
1245
record.get_bytes_as('fulltext'))
1246
if last_object is not None:
1247
delta = inv._make_delta(last_object)
1248
for old_path, path, file_id, ie in delta:
1251
ie.check(checker, rev_id, inv)
1253
for path, ie in inv.iter_entries():
1254
ie.check(checker, rev_id, inv)
1255
if self._format.fast_deltas:
1257
elif kind == 'chk_bytes':
1258
# No code written to check chk_bytes for this repo format.
1259
checker._report_items.append(
1260
'unsupported key type chk_bytes for %s' % (record.key,))
1261
elif kind == 'texts':
1262
self._check_text(record, checker, item_data)
1264
checker._report_items.append(
1265
'unknown key type %s for %s' % (kind, record.key))
1267
def _check_text(self, record, checker, item_data):
1268
"""Check a single text."""
1269
# Check it is extractable.
1270
# TODO: check length.
1271
if record.storage_kind == 'chunked':
1272
chunks = record.get_bytes_as(record.storage_kind)
1273
sha1 = osutils.sha_strings(chunks)
1274
length = sum(map(len, chunks))
1276
content = record.get_bytes_as('fulltext')
1277
sha1 = osutils.sha_string(content)
1278
length = len(content)
1279
if item_data and sha1 != item_data[1]:
1280
checker._report_items.append(
1281
'sha1 mismatch: %s has sha1 %s expected %s referenced by %s' %
1282
(record.key, sha1, item_data[1], item_data[2]))
1285
1150
def create(a_bzrdir):
1286
1151
"""Construct the current default format repository in a_bzrdir."""
1306
1172
self._reconcile_does_inventory_gc = True
1307
1173
self._reconcile_fixes_text_parents = False
1308
1174
self._reconcile_backsup_inventory = True
1175
# not right yet - should be more semantically clear ?
1177
# TODO: make sure to construct the right store classes, etc, depending
1178
# on whether escaping is required.
1179
self._warn_if_deprecated()
1309
1180
self._write_group = None
1310
1181
# Additional places to query for data.
1311
1182
self._fallback_repositories = []
1312
1183
# An InventoryEntry cache, used during deserialization
1313
1184
self._inventory_entry_cache = fifo_cache.FIFOCache(10*1024)
1314
# Is it safe to return inventory entries directly from the entry cache,
1315
# rather copying them?
1316
self._safe_to_return_from_cache = False
1319
def user_transport(self):
1320
return self.bzrdir.user_transport
1323
def control_transport(self):
1324
return self._transport
1326
1186
def __repr__(self):
1327
if self._fallback_repositories:
1328
return '%s(%r, fallback_repositories=%r)' % (
1329
self.__class__.__name__,
1331
self._fallback_repositories)
1333
return '%s(%r)' % (self.__class__.__name__,
1336
def _has_same_fallbacks(self, other_repo):
1337
"""Returns true if the repositories have the same fallbacks."""
1338
my_fb = self._fallback_repositories
1339
other_fb = other_repo._fallback_repositories
1340
if len(my_fb) != len(other_fb):
1342
for f, g in zip(my_fb, other_fb):
1343
if not f.has_same_location(g):
1187
return '%s(%r)' % (self.__class__.__name__,
1347
1190
def has_same_location(self, other):
1348
1191
"""Returns a boolean indicating if this repository is at the same
1860
1694
@needs_read_lock
1861
1695
def get_revisions(self, revision_ids):
1862
"""Get many revisions at once.
1864
Repositories that need to check data on every revision read should
1865
subclass this method.
1696
"""Get many revisions at once."""
1867
1697
return self._get_revisions(revision_ids)
1869
1699
@needs_read_lock
1870
1700
def _get_revisions(self, revision_ids):
1871
1701
"""Core work logic to get many revisions without sanity checks."""
1702
for rev_id in revision_ids:
1703
if not rev_id or not isinstance(rev_id, basestring):
1704
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1705
keys = [(key,) for key in revision_ids]
1706
stream = self.revisions.get_record_stream(keys, 'unordered', True)
1873
for revid, rev in self._iter_revisions(revision_ids):
1875
raise errors.NoSuchRevision(self, revid)
1708
for record in stream:
1709
if record.storage_kind == 'absent':
1710
raise errors.NoSuchRevision(self, record.key[0])
1711
text = record.get_bytes_as('fulltext')
1712
rev = self._serializer.read_revision_from_string(text)
1713
revs[record.key[0]] = rev
1877
1714
return [revs[revid] for revid in revision_ids]
1879
def _iter_revisions(self, revision_ids):
1880
"""Iterate over revision objects.
1882
:param revision_ids: An iterable of revisions to examine. None may be
1883
passed to request all revisions known to the repository. Note that
1884
not all repositories can find unreferenced revisions; for those
1885
repositories only referenced ones will be returned.
1886
:return: An iterator of (revid, revision) tuples. Absent revisions (
1887
those asked for but not available) are returned as (revid, None).
1889
if revision_ids is None:
1890
revision_ids = self.all_revision_ids()
1892
for rev_id in revision_ids:
1893
if not rev_id or not isinstance(rev_id, basestring):
1894
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1895
keys = [(key,) for key in revision_ids]
1896
stream = self.revisions.get_record_stream(keys, 'unordered', True)
1897
for record in stream:
1898
revid = record.key[0]
1899
if record.storage_kind == 'absent':
1902
text = record.get_bytes_as('fulltext')
1903
rev = self._serializer.read_revision_from_string(text)
1717
def get_revision_xml(self, revision_id):
1718
# TODO: jam 20070210 This shouldn't be necessary since get_revision
1719
# would have already do it.
1720
# TODO: jam 20070210 Just use _serializer.write_revision_to_string()
1721
# TODO: this can't just be replaced by:
1722
# return self._serializer.write_revision_to_string(
1723
# self.get_revision(revision_id))
1724
# as cStringIO preservers the encoding unlike write_revision_to_string
1725
# or some other call down the path.
1726
rev = self.get_revision(revision_id)
1727
rev_tmp = cStringIO.StringIO()
1728
# the current serializer..
1729
self._serializer.write_revision(rev, rev_tmp)
1731
return rev_tmp.getvalue()
1906
1733
def get_deltas_for_revisions(self, revisions, specific_fileids=None):
1907
1734
"""Produce a generator of revision deltas.
2369
2199
"""single-document based inventory iteration."""
2370
2200
inv_xmls = self._iter_inventory_xmls(revision_ids, ordering)
2371
2201
for text, revision_id in inv_xmls:
2372
yield self._deserialise_inventory(revision_id, text)
2202
yield self.deserialise_inventory(revision_id, text)
2374
def _iter_inventory_xmls(self, revision_ids, ordering):
2375
if ordering is None:
2376
order_as_requested = True
2377
ordering = 'unordered'
2379
order_as_requested = False
2204
def _iter_inventory_xmls(self, revision_ids, ordering='unordered'):
2380
2205
keys = [(revision_id,) for revision_id in revision_ids]
2383
if order_as_requested:
2384
key_iter = iter(keys)
2385
next_key = key_iter.next()
2386
2206
stream = self.inventories.get_record_stream(keys, ordering, True)
2387
2207
text_chunks = {}
2388
2208
for record in stream:
2389
2209
if record.storage_kind != 'absent':
2390
chunks = record.get_bytes_as('chunked')
2391
if order_as_requested:
2392
text_chunks[record.key] = chunks
2394
yield ''.join(chunks), record.key[-1]
2210
text_chunks[record.key] = record.get_bytes_as('chunked')
2396
2212
raise errors.NoSuchRevision(self, record.key)
2397
if order_as_requested:
2398
# Yield as many results as we can while preserving order.
2399
while next_key in text_chunks:
2400
chunks = text_chunks.pop(next_key)
2401
yield ''.join(chunks), next_key[-1]
2403
next_key = key_iter.next()
2404
except StopIteration:
2405
# We still want to fully consume the get_record_stream,
2406
# just in case it is not actually finished at this point
2214
chunks = text_chunks.pop(key)
2215
yield ''.join(chunks), key[-1]
2410
def _deserialise_inventory(self, revision_id, xml):
2217
def deserialise_inventory(self, revision_id, xml):
2411
2218
"""Transform the xml into an inventory object.
2413
2220
:param revision_id: The expected revision id of the inventory.
2414
2221
:param xml: A serialised inventory.
2416
2223
result = self._serializer.read_inventory_from_string(xml, revision_id,
2417
entry_cache=self._inventory_entry_cache,
2418
return_from_cache=self._safe_to_return_from_cache)
2224
entry_cache=self._inventory_entry_cache)
2419
2225
if result.revision_id != revision_id:
2420
2226
raise AssertionError('revision id mismatch %s != %s' % (
2421
2227
result.revision_id, revision_id))
2230
def serialise_inventory(self, inv):
2231
return self._serializer.write_inventory_to_string(inv)
2233
def _serialise_inventory_to_lines(self, inv):
2234
return self._serializer.write_inventory_to_lines(inv)
2424
2236
def get_serializer_format(self):
2425
2237
return self._serializer.format_num
2427
2239
@needs_read_lock
2428
def _get_inventory_xml(self, revision_id):
2429
"""Get serialized inventory as a string."""
2240
def get_inventory_xml(self, revision_id):
2241
"""Get inventory XML as a file object."""
2430
2242
texts = self._iter_inventory_xmls([revision_id], 'unordered')
2432
2244
text, revision_id = texts.next()
2725
2543
return record.get_bytes_as('fulltext')
2727
2545
@needs_read_lock
2728
def check(self, revision_ids=None, callback_refs=None, check_repo=True):
2546
def check(self, revision_ids=None):
2729
2547
"""Check consistency of all history of given revision_ids.
2731
2549
Different repository implementations should override _check().
2733
2551
:param revision_ids: A non-empty list of revision_ids whose ancestry
2734
2552
will be checked. Typically the last revision_id of a branch.
2735
:param callback_refs: A dict of check-refs to resolve and callback
2736
the check/_check method on the items listed as wanting the ref.
2738
:param check_repo: If False do not check the repository contents, just
2739
calculate the data callback_refs requires and call them back.
2741
return self._check(revision_ids, callback_refs=callback_refs,
2742
check_repo=check_repo)
2554
return self._check(revision_ids)
2744
def _check(self, revision_ids, callback_refs, check_repo):
2745
result = check.Check(self, check_repo=check_repo)
2746
result.check(callback_refs)
2556
def _check(self, revision_ids):
2557
result = check.Check(self)
2749
def _warn_if_deprecated(self, branch=None):
2561
def _warn_if_deprecated(self):
2750
2562
global _deprecation_warning_done
2751
2563
if _deprecation_warning_done:
2755
conf = config.GlobalConfig()
2757
conf = branch.get_config()
2758
if conf.suppress_warning('format_deprecation'):
2760
warning("Format %s for %s is deprecated -"
2761
" please use 'bzr upgrade' to get better performance"
2762
% (self._format, self.bzrdir.transport.base))
2764
_deprecation_warning_done = True
2565
_deprecation_warning_done = True
2566
warning("Format %s for %s is deprecated - please use 'bzr upgrade' to get better performance"
2567
% (self._format, self.bzrdir.transport.base))
2766
2569
def supports_rich_root(self):
2767
2570
return self._format.rich_root_data
3398
3180
:param revision_id: if None all content is copied, if NULL_REVISION no
3399
3181
content is copied.
3182
:param pb: optional progress bar to use for progress reports. If not
3183
provided a default one will be created.
3403
ui.ui_factory.warn_experimental_format_fetch(self)
3404
3186
from bzrlib.fetch import RepoFetcher
3405
# See <https://launchpad.net/bugs/456077> asking for a warning here
3406
if self.source._format.network_name() != self.target._format.network_name():
3407
ui.ui_factory.show_user_warning('cross_format_fetch',
3408
from_format=self.source._format,
3409
to_format=self.target._format)
3410
3187
f = RepoFetcher(to_repository=self.target,
3411
3188
from_repository=self.source,
3412
3189
last_revision=revision_id,
3413
3190
fetch_spec=fetch_spec,
3414
find_ghosts=find_ghosts)
3191
pb=pb, find_ghosts=find_ghosts)
3416
3193
def _walk_to_common_revisions(self, revision_ids):
3417
3194
"""Walk out from revision_ids in source to revisions target has.
3718
3495
# This is redundant with format.check_conversion_target(), however that
3719
3496
# raises an exception, and we just want to say "False" as in we won't
3720
3497
# support converting between these formats.
3721
if 'IDS_never' in debug.debug_flags:
3723
3498
if source.supports_rich_root() and not target.supports_rich_root():
3725
3500
if (source._format.supports_tree_reference
3726
3501
and not target._format.supports_tree_reference):
3728
if target._fallback_repositories and target._format.supports_chks:
3729
# IDS doesn't know how to copy CHKs for the parent inventories it
3730
# adds to stacked repos.
3732
if 'IDS_always' in debug.debug_flags:
3734
# Only use this code path for local source and target. IDS does far
3735
# too much IO (both bandwidth and roundtrips) over a network.
3736
if not source.bzrdir.transport.base.startswith('file:///'):
3738
if not target.bzrdir.transport.base.startswith('file:///'):
3742
def _get_trees(self, revision_ids, cache):
3744
for rev_id in revision_ids:
3746
possible_trees.append((rev_id, cache[rev_id]))
3748
# Not cached, but inventory might be present anyway.
3750
tree = self.source.revision_tree(rev_id)
3751
except errors.NoSuchRevision:
3752
# Nope, parent is ghost.
3755
cache[rev_id] = tree
3756
possible_trees.append((rev_id, tree))
3757
return possible_trees
3759
def _get_delta_for_revision(self, tree, parent_ids, possible_trees):
3505
def _get_delta_for_revision(self, tree, parent_ids, basis_id, cache):
3760
3506
"""Get the best delta and base for this revision.
3762
3508
:return: (basis_id, delta)
3510
possible_trees = [(parent_id, cache[parent_id])
3511
for parent_id in parent_ids
3512
if parent_id in cache]
3513
if len(possible_trees) == 0:
3514
# There either aren't any parents, or the parents aren't in the
3515
# cache, so just use the last converted tree
3516
possible_trees.append((basis_id, cache[basis_id]))
3765
# Generate deltas against each tree, to find the shortest.
3766
texts_possibly_new_in_tree = set()
3767
3518
for basis_id, basis_tree in possible_trees:
3768
3519
delta = tree.inventory._make_delta(basis_tree.inventory)
3769
for old_path, new_path, file_id, new_entry in delta:
3770
if new_path is None:
3771
# This file_id isn't present in the new rev, so we don't
3775
# Rich roots are handled elsewhere...
3777
kind = new_entry.kind
3778
if kind != 'directory' and kind != 'file':
3779
# No text record associated with this inventory entry.
3781
# This is a directory or file that has changed somehow.
3782
texts_possibly_new_in_tree.add((file_id, new_entry.revision))
3783
3520
deltas.append((len(delta), basis_id, delta))
3785
3522
return deltas[0][1:]
3787
def _fetch_parent_invs_for_stacking(self, parent_map, cache):
3788
"""Find all parent revisions that are absent, but for which the
3789
inventory is present, and copy those inventories.
3791
This is necessary to preserve correctness when the source is stacked
3792
without fallbacks configured. (Note that in cases like upgrade the
3793
source may be not have _fallback_repositories even though it is
3797
for parents in parent_map.values():
3798
parent_revs.update(parents)
3799
present_parents = self.source.get_parent_map(parent_revs)
3800
absent_parents = set(parent_revs).difference(present_parents)
3801
parent_invs_keys_for_stacking = self.source.inventories.get_parent_map(
3802
(rev_id,) for rev_id in absent_parents)
3803
parent_inv_ids = [key[-1] for key in parent_invs_keys_for_stacking]
3804
for parent_tree in self.source.revision_trees(parent_inv_ids):
3805
current_revision_id = parent_tree.get_revision_id()
3806
parents_parents_keys = parent_invs_keys_for_stacking[
3807
(current_revision_id,)]
3808
parents_parents = [key[-1] for key in parents_parents_keys]
3809
basis_id = _mod_revision.NULL_REVISION
3810
basis_tree = self.source.revision_tree(basis_id)
3811
delta = parent_tree.inventory._make_delta(basis_tree.inventory)
3812
self.target.add_inventory_by_delta(
3813
basis_id, delta, current_revision_id, parents_parents)
3814
cache[current_revision_id] = parent_tree
3816
def _fetch_batch(self, revision_ids, basis_id, cache, a_graph=None):
3524
def _fetch_batch(self, revision_ids, basis_id, cache):
3817
3525
"""Fetch across a few revisions.
3819
3527
:param revision_ids: The revisions to copy
3820
3528
:param basis_id: The revision_id of a tree that must be in cache, used
3821
3529
as a basis for delta when no other base is available
3822
3530
:param cache: A cache of RevisionTrees that we can use.
3823
:param a_graph: A Graph object to determine the heads() of the
3824
rich-root data stream.
3825
3531
:return: The revision_id of the last converted tree. The RevisionTree
3826
3532
for it will be in cache
3833
3539
pending_deltas = []
3834
3540
pending_revisions = []
3835
3541
parent_map = self.source.get_parent_map(revision_ids)
3836
self._fetch_parent_invs_for_stacking(parent_map, cache)
3837
self.source._safe_to_return_from_cache = True
3838
3542
for tree in self.source.revision_trees(revision_ids):
3839
# Find a inventory delta for this revision.
3840
# Find text entries that need to be copied, too.
3841
3543
current_revision_id = tree.get_revision_id()
3842
3544
parent_ids = parent_map.get(current_revision_id, ())
3843
parent_trees = self._get_trees(parent_ids, cache)
3844
possible_trees = list(parent_trees)
3845
if len(possible_trees) == 0:
3846
# There either aren't any parents, or the parents are ghosts,
3847
# so just use the last converted tree.
3848
possible_trees.append((basis_id, cache[basis_id]))
3849
3545
basis_id, delta = self._get_delta_for_revision(tree, parent_ids,
3851
revision = self.source.get_revision(current_revision_id)
3852
pending_deltas.append((basis_id, delta,
3853
current_revision_id, revision.parent_ids))
3854
3547
if self._converting_to_rich_root:
3855
3548
self._revision_id_to_root_id[current_revision_id] = \
3856
3549
tree.get_root_id()
3857
# Determine which texts are in present in this revision but not in
3858
# any of the available parents.
3859
texts_possibly_new_in_tree = set()
3550
# Find text entries that need to be copied
3860
3551
for old_path, new_path, file_id, entry in delta:
3861
if new_path is None:
3862
# This file_id isn't present in the new rev
3866
if not self.target.supports_rich_root():
3867
# The target doesn't support rich root, so we don't
3870
if self._converting_to_rich_root:
3871
# This can't be copied normally, we have to insert
3873
root_keys_to_create.add((file_id, entry.revision))
3876
texts_possibly_new_in_tree.add((file_id, entry.revision))
3877
for basis_id, basis_tree in possible_trees:
3878
basis_inv = basis_tree.inventory
3879
for file_key in list(texts_possibly_new_in_tree):
3880
file_id, file_revision = file_key
3882
entry = basis_inv[file_id]
3883
except errors.NoSuchId:
3885
if entry.revision == file_revision:
3886
texts_possibly_new_in_tree.remove(file_key)
3887
text_keys.update(texts_possibly_new_in_tree)
3552
if new_path is not None:
3555
if not self.target.supports_rich_root():
3556
# The target doesn't support rich root, so we don't
3559
if self._converting_to_rich_root:
3560
# This can't be copied normally, we have to insert
3562
root_keys_to_create.add((file_id, entry.revision))
3564
text_keys.add((file_id, entry.revision))
3565
revision = self.source.get_revision(current_revision_id)
3566
pending_deltas.append((basis_id, delta,
3567
current_revision_id, revision.parent_ids))
3888
3568
pending_revisions.append(revision)
3889
3569
cache[current_revision_id] = tree
3890
3570
basis_id = current_revision_id
3891
self.source._safe_to_return_from_cache = False
3892
3571
# Copy file texts
3893
3572
from_texts = self.source.texts
3894
3573
to_texts = self.target.texts
3895
3574
if root_keys_to_create:
3896
root_stream = _mod_fetch._new_root_data_stream(
3575
from bzrlib.fetch import _new_root_data_stream
3576
root_stream = _new_root_data_stream(
3897
3577
root_keys_to_create, self._revision_id_to_root_id, parent_map,
3898
self.source, graph=a_graph)
3899
3579
to_texts.insert_record_stream(root_stream)
3900
3580
to_texts.insert_record_stream(from_texts.get_record_stream(
3901
3581
text_keys, self.target._format._fetch_order,
3918
3596
parent_ids.difference_update(revision_ids)
3919
3597
parent_ids.discard(_mod_revision.NULL_REVISION)
3920
3598
parent_map = self.source.get_parent_map(parent_ids)
3921
# we iterate over parent_map and not parent_ids because we don't
3922
# want to try copying any revision which is a ghost
3923
for parent_tree in self.source.revision_trees(parent_map):
3599
for parent_tree in self.source.revision_trees(parent_ids):
3600
basis_id, delta = self._get_delta_for_revision(tree, parent_ids, basis_id, cache)
3924
3601
current_revision_id = parent_tree.get_revision_id()
3925
3602
parents_parents = parent_map[current_revision_id]
3926
possible_trees = self._get_trees(parents_parents, cache)
3927
if len(possible_trees) == 0:
3928
# There either aren't any parents, or the parents are
3929
# ghosts, so just use the last converted tree.
3930
possible_trees.append((basis_id, cache[basis_id]))
3931
basis_id, delta = self._get_delta_for_revision(parent_tree,
3932
parents_parents, possible_trees)
3933
3603
self.target.add_inventory_by_delta(
3934
3604
basis_id, delta, current_revision_id, parents_parents)
3935
3605
# insert signatures and revisions
4077
3733
:param to_convert: The disk object to convert.
4078
3734
:param pb: a progress bar to use for progress information.
4080
pb = ui.ui_factory.nested_progress_bar()
4083
3739
# this is only useful with metadir layouts - separated repo content.
4084
3740
# trigger an assertion if not such
4085
3741
repo._format.get_format_string()
4086
3742
self.repo_dir = repo.bzrdir
4087
pb.update('Moving repository to repository.backup')
3743
self.step('Moving repository to repository.backup')
4088
3744
self.repo_dir.transport.move('repository', 'repository.backup')
4089
3745
backup_transport = self.repo_dir.transport.clone('repository.backup')
4090
3746
repo._format.check_conversion_target(self.target_format)
4091
3747
self.source_repo = repo._format.open(self.repo_dir,
4093
3749
_override_transport=backup_transport)
4094
pb.update('Creating new repository')
3750
self.step('Creating new repository')
4095
3751
converted = self.target_format.initialize(self.repo_dir,
4096
3752
self.source_repo.is_shared())
4097
3753
converted.lock_write()
4099
pb.update('Copying content')
3755
self.step('Copying content into repository.')
4100
3756
self.source_repo.copy_content_into(converted)
4102
3758
converted.unlock()
4103
pb.update('Deleting old repository content')
3759
self.step('Deleting old repository content.')
4104
3760
self.repo_dir.transport.delete_tree('repository.backup')
4105
ui.ui_factory.note('repository converted')
3761
self.pb.note('repository converted')
3763
def step(self, message):
3764
"""Update the pb by a step."""
3766
self.pb.update(message, self.count, self.total)
4109
3769
_unescape_map = {
4356
3995
self.target_repo.pack(hint=hint)
4357
3996
return [], set()
4359
def _extract_and_insert_inventory_deltas(self, substream, serializer):
4360
target_rich_root = self.target_repo._format.rich_root_data
4361
target_tree_refs = self.target_repo._format.supports_tree_reference
4362
for record in substream:
4363
# Insert the delta directly
4364
inventory_delta_bytes = record.get_bytes_as('fulltext')
4365
deserialiser = inventory_delta.InventoryDeltaDeserializer()
4367
parse_result = deserialiser.parse_text_bytes(
4368
inventory_delta_bytes)
4369
except inventory_delta.IncompatibleInventoryDelta, err:
4370
trace.mutter("Incompatible delta: %s", err.msg)
4371
raise errors.IncompatibleRevision(self.target_repo._format)
4372
basis_id, new_id, rich_root, tree_refs, inv_delta = parse_result
4373
revision_id = new_id
4374
parents = [key[0] for key in record.parents]
4375
self.target_repo.add_inventory_by_delta(
4376
basis_id, inv_delta, revision_id, parents)
4378
3998
def _extract_and_insert_inventories(self, substream, serializer,
4379
3999
parse_delta=None):
4380
4000
"""Generate a new inventory versionedfile in target, converting data.
4566
4205
def _get_inventory_stream(self, revision_ids, missing=False):
4567
4206
from_format = self.from_repository._format
4568
if (from_format.supports_chks and self.to_format.supports_chks and
4569
from_format.network_name() == self.to_format.network_name()):
4570
raise AssertionError(
4571
"this case should be handled by GroupCHKStreamSource")
4572
elif 'forceinvdeltas' in debug.debug_flags:
4573
return self._get_convertable_inventory_stream(revision_ids,
4574
delta_versus_null=missing)
4575
elif from_format.network_name() == self.to_format.network_name():
4577
return self._get_simple_inventory_stream(revision_ids,
4579
elif (not from_format.supports_chks and not self.to_format.supports_chks
4580
and from_format._serializer == self.to_format._serializer):
4581
# Essentially the same format.
4207
if (from_format.supports_chks and self.to_format.supports_chks
4208
and (from_format._serializer == self.to_format._serializer)):
4209
# Both sides support chks, and they use the same serializer, so it
4210
# is safe to transmit the chk pages and inventory pages across
4212
# XXX: does this case need to take 'missing' into account?
4213
return self._get_chk_inventory_stream(revision_ids)
4214
elif (not from_format.supports_chks):
4215
# Source repository doesn't support chks. So we can transmit the
4216
# inventories 'as-is' and either they are just accepted on the
4217
# target, or the Sink will properly convert it.
4582
4218
return self._get_simple_inventory_stream(revision_ids,
4583
4219
missing=missing)
4585
# Any time we switch serializations, we want to use an
4586
# inventory-delta based approach.
4221
# Make not-chk->chk fetch: copy the inventories as (format-neutral)
4587
4223
return self._get_convertable_inventory_stream(revision_ids,
4588
delta_versus_null=missing)
4590
4226
def _get_simple_inventory_stream(self, revision_ids, missing=False):
4591
4227
# NB: This currently reopens the inventory weave in source;
4599
4235
[(rev_id,) for rev_id in revision_ids],
4600
4236
self.inventory_fetch_order(), delta_closure))
4602
def _get_convertable_inventory_stream(self, revision_ids,
4603
delta_versus_null=False):
4604
# The two formats are sufficiently different that there is no fast
4605
# path, so we need to send just inventorydeltas, which any
4606
# sufficiently modern client can insert into any repository.
4607
# The StreamSink code expects to be able to
4608
# convert on the target, so we need to put bytes-on-the-wire that can
4609
# be converted. That means inventory deltas (if the remote is <1.19,
4610
# RemoteStreamSink will fallback to VFS to insert the deltas).
4611
yield ('inventory-deltas',
4612
self._stream_invs_as_deltas(revision_ids,
4613
delta_versus_null=delta_versus_null))
4615
def _stream_invs_as_deltas(self, revision_ids, delta_versus_null=False):
4616
"""Return a stream of inventory-deltas for the given rev ids.
4618
:param revision_ids: The list of inventories to transmit
4619
:param delta_versus_null: Don't try to find a minimal delta for this
4620
entry, instead compute the delta versus the NULL_REVISION. This
4621
effectively streams a complete inventory. Used for stuff like
4622
filling in missing parents, etc.
4238
def _get_chk_inventory_stream(self, revision_ids):
4239
"""Fetch the inventory texts, along with the associated chk maps."""
4240
# We want an inventory outside of the search set, so that we can filter
4241
# out uninteresting chk pages. For now we use
4242
# _find_revision_outside_set, but if we had a Search with cut_revs, we
4243
# could use that instead.
4244
start_rev_id = self.from_repository._find_revision_outside_set(
4246
start_rev_key = (start_rev_id,)
4247
inv_keys_to_fetch = [(rev_id,) for rev_id in revision_ids]
4248
if start_rev_id != _mod_revision.NULL_REVISION:
4249
inv_keys_to_fetch.append((start_rev_id,))
4250
# Any repo that supports chk_bytes must also support out-of-order
4251
# insertion. At least, that is how we expect it to work
4252
# We use get_record_stream instead of iter_inventories because we want
4253
# to be able to insert the stream as well. We could instead fetch
4254
# allowing deltas, and then iter_inventories, but we don't know whether
4255
# source or target is more 'local' anway.
4256
inv_stream = self.from_repository.inventories.get_record_stream(
4257
inv_keys_to_fetch, 'unordered',
4258
True) # We need them as full-texts so we can find their references
4259
uninteresting_chk_roots = set()
4260
interesting_chk_roots = set()
4261
def filter_inv_stream(inv_stream):
4262
for idx, record in enumerate(inv_stream):
4263
### child_pb.update('fetch inv', idx, len(inv_keys_to_fetch))
4264
bytes = record.get_bytes_as('fulltext')
4265
chk_inv = inventory.CHKInventory.deserialise(
4266
self.from_repository.chk_bytes, bytes, record.key)
4267
if record.key == start_rev_key:
4268
uninteresting_chk_roots.add(chk_inv.id_to_entry.key())
4269
p_id_map = chk_inv.parent_id_basename_to_file_id
4270
if p_id_map is not None:
4271
uninteresting_chk_roots.add(p_id_map.key())
4274
interesting_chk_roots.add(chk_inv.id_to_entry.key())
4275
p_id_map = chk_inv.parent_id_basename_to_file_id
4276
if p_id_map is not None:
4277
interesting_chk_roots.add(p_id_map.key())
4278
### pb.update('fetch inventory', 0, 2)
4279
yield ('inventories', filter_inv_stream(inv_stream))
4280
# Now that we have worked out all of the interesting root nodes, grab
4281
# all of the interesting pages and insert them
4282
### pb.update('fetch inventory', 1, 2)
4283
interesting = chk_map.iter_interesting_nodes(
4284
self.from_repository.chk_bytes, interesting_chk_roots,
4285
uninteresting_chk_roots)
4286
def to_stream_adapter():
4287
"""Adapt the iter_interesting_nodes result to a single stream.
4289
iter_interesting_nodes returns records as it processes them, along
4290
with keys. However, we only want to return the records themselves.
4292
for record, items in interesting:
4293
if record is not None:
4295
# XXX: We could instead call get_record_stream(records.keys())
4296
# ATM, this will always insert the records as fulltexts, and
4297
# requires that you can hang on to records once you have gone
4298
# on to the next one. Further, it causes the target to
4299
# recompress the data. Testing shows it to be faster than
4300
# requesting the records again, though.
4301
yield ('chk_bytes', to_stream_adapter())
4302
### pb.update('fetch inventory', 2, 2)
4304
def _get_convertable_inventory_stream(self, revision_ids, fulltexts=False):
4305
# XXX: One of source or target is using chks, and they don't have
4306
# compatible serializations. The StreamSink code expects to be
4307
# able to convert on the target, so we need to put
4308
# bytes-on-the-wire that can be converted
4309
# XXX: choose between fulltexts (for compat) or deltas (for efficiency)
4310
yield ('inventories',
4311
self._stream_invs_as_deltas(revision_ids, fulltexts=fulltexts))
4313
def _stream_invs_as_deltas(self, revision_ids, fulltexts=False):
4624
4314
from_repo = self.from_repository
4625
4315
revision_keys = [(rev_id,) for rev_id in revision_ids]
4626
4316
parent_map = from_repo.inventories.get_parent_map(revision_keys)
4629
4319
inventories = self.from_repository.iter_inventories(
4630
4320
revision_ids, 'topological')
4321
# XXX: ideally these flags would be per-revision, not per-repo (e.g.
4322
# streaming a non-rich-root revision out of a rich-root repo back into
4323
# a non-rich-root repo ought to be allowed)
4631
4324
format = from_repo._format
4632
invs_sent_so_far = set([_mod_revision.NULL_REVISION])
4633
inventory_cache = lru_cache.LRUCache(50)
4634
null_inventory = from_repo.revision_tree(
4635
_mod_revision.NULL_REVISION).inventory
4636
# XXX: ideally the rich-root/tree-refs flags would be per-revision, not
4637
# per-repo (e.g. streaming a non-rich-root revision out of a rich-root
4638
# repo back into a non-rich-root repo ought to be allowed)
4639
serializer = inventory_delta.InventoryDeltaSerializer(
4640
versioned_root=format.rich_root_data,
4641
tree_references=format.supports_tree_reference)
4325
flags = (format.rich_root_data, format.supports_tree_reference)
4642
4326
for inv in inventories:
4643
4327
key = (inv.revision_id,)
4644
parent_keys = parent_map.get(key, ())
4646
if not delta_versus_null and parent_keys:
4647
# The caller did not ask for complete inventories and we have
4648
# some parents that we can delta against. Make a delta against
4649
# each parent so that we can find the smallest.
4650
parent_ids = [parent_key[0] for parent_key in parent_keys]
4651
for parent_id in parent_ids:
4652
if parent_id not in invs_sent_so_far:
4653
# We don't know that the remote side has this basis, so
4656
if parent_id == _mod_revision.NULL_REVISION:
4657
parent_inv = null_inventory
4659
parent_inv = inventory_cache.get(parent_id, None)
4660
if parent_inv is None:
4661
parent_inv = from_repo.get_inventory(parent_id)
4328
parents = parent_map.get(key, ())
4329
if fulltexts or parents == ():
4330
# Either the caller asked for fulltexts, or there is no parent,
4331
# so, stream as a delta from null:.
4332
basis_id = _mod_revision.NULL_REVISION
4333
parent_inv = Inventory()
4334
delta = inv._make_delta(parent_inv)
4336
# Make a delta against each parent so that we can find the
4339
for parent_key in parents:
4340
parent_id = parent_key[0]
4341
parent_inv = from_repo.get_inventory(parent_id)
4662
4342
candidate_delta = inv._make_delta(parent_inv)
4663
if (delta is None or
4664
len(delta) > len(candidate_delta)):
4665
delta = candidate_delta
4343
if (best_delta is None or
4344
len(best_delta) > len(candidate_delta)):
4345
best_delta = candidate_delta
4666
4346
basis_id = parent_id
4668
# Either none of the parents ended up being suitable, or we
4669
# were asked to delta against NULL
4670
basis_id = _mod_revision.NULL_REVISION
4671
delta = inv._make_delta(null_inventory)
4672
invs_sent_so_far.add(inv.revision_id)
4673
inventory_cache[inv.revision_id] = inv
4674
delta_serialized = ''.join(
4675
serializer.delta_to_lines(basis_id, key[-1], delta))
4676
yield versionedfile.FulltextContentFactory(
4677
key, parent_keys, None, delta_serialized)
4348
yield versionedfile.InventoryDeltaContentFactory(
4349
key, parents, None, delta, basis_id, flags)
4680
4352
def _iter_for_revno(repo, partial_history_cache, stop_index=None,
added
removed
bzrlib/repository.py
