1195
1073
# The old API returned a list, should this actually be a set?
1196
1074
return parent_map.keys()
1198
def _check_inventories(self, checker):
1199
"""Check the inventories found from the revision scan.
1201
This is responsible for verifying the sha1 of inventories and
1202
creating a pending_keys set that covers data referenced by inventories.
1204
bar = ui.ui_factory.nested_progress_bar()
1206
self._do_check_inventories(checker, bar)
1210
def _do_check_inventories(self, checker, bar):
1211
"""Helper for _check_inventories."""
1213
keys = {'chk_bytes':set(), 'inventories':set(), 'texts':set()}
1214
kinds = ['chk_bytes', 'texts']
1215
count = len(checker.pending_keys)
1216
bar.update("inventories", 0, 2)
1217
current_keys = checker.pending_keys
1218
checker.pending_keys = {}
1219
# Accumulate current checks.
1220
for key in current_keys:
1221
if key[0] != 'inventories' and key[0] not in kinds:
1222
checker._report_items.append('unknown key type %r' % (key,))
1223
keys[key[0]].add(key[1:])
1224
if keys['inventories']:
1225
# NB: output order *should* be roughly sorted - topo or
1226
# inverse topo depending on repository - either way decent
1227
# to just delta against. However, pre-CHK formats didn't
1228
# try to optimise inventory layout on disk. As such the
1229
# pre-CHK code path does not use inventory deltas.
1231
for record in self.inventories.check(keys=keys['inventories']):
1232
if record.storage_kind == 'absent':
1233
checker._report_items.append(
1234
'Missing inventory {%s}' % (record.key,))
1236
last_object = self._check_record('inventories', record,
1237
checker, last_object,
1238
current_keys[('inventories',) + record.key])
1239
del keys['inventories']
1242
bar.update("texts", 1)
1243
while (checker.pending_keys or keys['chk_bytes']
1245
# Something to check.
1246
current_keys = checker.pending_keys
1247
checker.pending_keys = {}
1248
# Accumulate current checks.
1249
for key in current_keys:
1250
if key[0] not in kinds:
1251
checker._report_items.append('unknown key type %r' % (key,))
1252
keys[key[0]].add(key[1:])
1253
# Check the outermost kind only - inventories || chk_bytes || texts
1257
for record in getattr(self, kind).check(keys=keys[kind]):
1258
if record.storage_kind == 'absent':
1259
checker._report_items.append(
1260
'Missing %s {%s}' % (kind, record.key,))
1262
last_object = self._check_record(kind, record,
1263
checker, last_object, current_keys[(kind,) + record.key])
1267
def _check_record(self, kind, record, checker, last_object, item_data):
1268
"""Check a single text from this repository."""
1269
if kind == 'inventories':
1270
rev_id = record.key[0]
1271
inv = self._deserialise_inventory(rev_id,
1272
record.get_bytes_as('fulltext'))
1273
if last_object is not None:
1274
delta = inv._make_delta(last_object)
1275
for old_path, path, file_id, ie in delta:
1278
ie.check(checker, rev_id, inv)
1280
for path, ie in inv.iter_entries():
1281
ie.check(checker, rev_id, inv)
1282
if self._format.fast_deltas:
1284
elif kind == 'chk_bytes':
1285
# No code written to check chk_bytes for this repo format.
1286
checker._report_items.append(
1287
'unsupported key type chk_bytes for %s' % (record.key,))
1288
elif kind == 'texts':
1289
self._check_text(record, checker, item_data)
1291
checker._report_items.append(
1292
'unknown key type %s for %s' % (kind, record.key))
1294
def _check_text(self, record, checker, item_data):
1295
"""Check a single text."""
1296
# Check it is extractable.
1297
# TODO: check length.
1298
if record.storage_kind == 'chunked':
1299
chunks = record.get_bytes_as(record.storage_kind)
1300
sha1 = osutils.sha_strings(chunks)
1301
length = sum(map(len, chunks))
1303
content = record.get_bytes_as('fulltext')
1304
sha1 = osutils.sha_string(content)
1305
length = len(content)
1306
if item_data and sha1 != item_data[1]:
1307
checker._report_items.append(
1308
'sha1 mismatch: %s has sha1 %s expected %s referenced by %s' %
1309
(record.key, sha1, item_data[1], item_data[2]))
1312
1077
def create(a_bzrdir):
1313
1078
"""Construct the current default format repository in a_bzrdir."""
1612
1347
def suspend_write_group(self):
1613
1348
raise errors.UnsuspendableWriteGroup(self)
1615
def get_missing_parent_inventories(self, check_for_missing_texts=True):
1616
"""Return the keys of missing inventory parents for revisions added in
1619
A revision is not complete if the inventory delta for that revision
1620
cannot be calculated. Therefore if the parent inventories of a
1621
revision are not present, the revision is incomplete, and e.g. cannot
1622
be streamed by a smart server. This method finds missing inventory
1623
parents for revisions added in this write group.
1625
if not self._format.supports_external_lookups:
1626
# This is only an issue for stacked repositories
1628
if not self.is_in_write_group():
1629
raise AssertionError('not in a write group')
1631
# XXX: We assume that every added revision already has its
1632
# corresponding inventory, so we only check for parent inventories that
1633
# might be missing, rather than all inventories.
1634
parents = set(self.revisions._index.get_missing_parents())
1635
parents.discard(_mod_revision.NULL_REVISION)
1636
unstacked_inventories = self.inventories._index
1637
present_inventories = unstacked_inventories.get_parent_map(
1638
key[-1:] for key in parents)
1639
parents.difference_update(present_inventories)
1640
if len(parents) == 0:
1641
# No missing parent inventories.
1643
if not check_for_missing_texts:
1644
return set(('inventories', rev_id) for (rev_id,) in parents)
1645
# Ok, now we have a list of missing inventories. But these only matter
1646
# if the inventories that reference them are missing some texts they
1647
# appear to introduce.
1648
# XXX: Texts referenced by all added inventories need to be present,
1649
# but at the moment we're only checking for texts referenced by
1650
# inventories at the graph's edge.
1651
key_deps = self.revisions._index._key_dependencies
1652
key_deps.satisfy_refs_for_keys(present_inventories)
1653
referrers = frozenset(r[0] for r in key_deps.get_referrers())
1654
file_ids = self.fileids_altered_by_revision_ids(referrers)
1655
missing_texts = set()
1656
for file_id, version_ids in file_ids.iteritems():
1657
missing_texts.update(
1658
(file_id, version_id) for version_id in version_ids)
1659
present_texts = self.texts.get_parent_map(missing_texts)
1660
missing_texts.difference_update(present_texts)
1661
if not missing_texts:
1662
# No texts are missing, so all revisions and their deltas are
1665
# Alternatively the text versions could be returned as the missing
1666
# keys, but this is likely to be less data.
1667
missing_keys = set(('inventories', rev_id) for (rev_id,) in parents)
1670
1350
def refresh_data(self):
1671
"""Re-read any data needed to synchronise with disk.
1351
"""Re-read any data needed to to synchronise with disk.
1673
1353
This method is intended to be called after another repository instance
1674
1354
(such as one used by a smart server) has inserted data into the
1675
repository. On all repositories this will work outside of write groups.
1676
Some repository formats (pack and newer for bzrlib native formats)
1677
support refresh_data inside write groups. If called inside a write
1678
group on a repository that does not support refreshing in a write group
1679
IsInWriteGroupError will be raised.
1355
repository. It may not be called during a write group, but may be
1356
called at any other time.
1358
if self.is_in_write_group():
1359
raise errors.InternalBzrError(
1360
"May not refresh_data while in a write group.")
1681
1361
self._refresh_data()
1683
1363
def resume_write_group(self, tokens):
1894
1565
@needs_read_lock
1895
1566
def get_revisions(self, revision_ids):
1896
"""Get many revisions at once.
1898
Repositories that need to check data on every revision read should
1899
subclass this method.
1567
"""Get many revisions at once."""
1901
1568
return self._get_revisions(revision_ids)
1903
1570
@needs_read_lock
1904
1571
def _get_revisions(self, revision_ids):
1905
1572
"""Core work logic to get many revisions without sanity checks."""
1573
for rev_id in revision_ids:
1574
if not rev_id or not isinstance(rev_id, basestring):
1575
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1576
keys = [(key,) for key in revision_ids]
1577
stream = self.revisions.get_record_stream(keys, 'unordered', True)
1907
for revid, rev in self._iter_revisions(revision_ids):
1909
raise errors.NoSuchRevision(self, revid)
1579
for record in stream:
1580
if record.storage_kind == 'absent':
1581
raise errors.NoSuchRevision(self, record.key[0])
1582
text = record.get_bytes_as('fulltext')
1583
rev = self._serializer.read_revision_from_string(text)
1584
revs[record.key[0]] = rev
1911
1585
return [revs[revid] for revid in revision_ids]
1913
def _iter_revisions(self, revision_ids):
1914
"""Iterate over revision objects.
1916
:param revision_ids: An iterable of revisions to examine. None may be
1917
passed to request all revisions known to the repository. Note that
1918
not all repositories can find unreferenced revisions; for those
1919
repositories only referenced ones will be returned.
1920
:return: An iterator of (revid, revision) tuples. Absent revisions (
1921
those asked for but not available) are returned as (revid, None).
1923
if revision_ids is None:
1924
revision_ids = self.all_revision_ids()
1926
for rev_id in revision_ids:
1927
if not rev_id or not isinstance(rev_id, basestring):
1928
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1929
keys = [(key,) for key in revision_ids]
1930
stream = self.revisions.get_record_stream(keys, 'unordered', True)
1931
for record in stream:
1932
revid = record.key[0]
1933
if record.storage_kind == 'absent':
1936
text = record.get_bytes_as('fulltext')
1937
rev = self._serializer.read_revision_from_string(text)
1588
def get_revision_xml(self, revision_id):
1589
# TODO: jam 20070210 This shouldn't be necessary since get_revision
1590
# would have already do it.
1591
# TODO: jam 20070210 Just use _serializer.write_revision_to_string()
1592
rev = self.get_revision(revision_id)
1593
rev_tmp = cStringIO.StringIO()
1594
# the current serializer..
1595
self._serializer.write_revision(rev, rev_tmp)
1597
return rev_tmp.getvalue()
1940
1599
def get_deltas_for_revisions(self, revisions, specific_fileids=None):
1941
1600
"""Produce a generator of revision deltas.
2381
2014
"""Get Inventory object by revision id."""
2382
2015
return self.iter_inventories([revision_id]).next()
2384
def iter_inventories(self, revision_ids, ordering=None):
2017
def iter_inventories(self, revision_ids):
2385
2018
"""Get many inventories by revision_ids.
2387
2020
This will buffer some or all of the texts used in constructing the
2388
2021
inventories in memory, but will only parse a single inventory at a
2391
:param revision_ids: The expected revision ids of the inventories.
2392
:param ordering: optional ordering, e.g. 'topological'. If not
2393
specified, the order of revision_ids will be preserved (by
2394
buffering if necessary).
2395
2024
:return: An iterator of inventories.
2397
2026
if ((None in revision_ids)
2398
2027
or (_mod_revision.NULL_REVISION in revision_ids)):
2399
2028
raise ValueError('cannot get null revision inventory')
2400
return self._iter_inventories(revision_ids, ordering)
2029
return self._iter_inventories(revision_ids)
2402
def _iter_inventories(self, revision_ids, ordering):
2031
def _iter_inventories(self, revision_ids):
2403
2032
"""single-document based inventory iteration."""
2404
inv_xmls = self._iter_inventory_xmls(revision_ids, ordering)
2405
for text, revision_id in inv_xmls:
2406
yield self._deserialise_inventory(revision_id, text)
2033
for text, revision_id in self._iter_inventory_xmls(revision_ids):
2034
yield self.deserialise_inventory(revision_id, text)
2408
def _iter_inventory_xmls(self, revision_ids, ordering):
2409
if ordering is None:
2410
order_as_requested = True
2411
ordering = 'unordered'
2413
order_as_requested = False
2036
def _iter_inventory_xmls(self, revision_ids):
2414
2037
keys = [(revision_id,) for revision_id in revision_ids]
2417
if order_as_requested:
2418
key_iter = iter(keys)
2419
next_key = key_iter.next()
2420
stream = self.inventories.get_record_stream(keys, ordering, True)
2038
stream = self.inventories.get_record_stream(keys, 'unordered', True)
2421
2039
text_chunks = {}
2422
2040
for record in stream:
2423
2041
if record.storage_kind != 'absent':
2424
chunks = record.get_bytes_as('chunked')
2425
if order_as_requested:
2426
text_chunks[record.key] = chunks
2428
yield ''.join(chunks), record.key[-1]
2042
text_chunks[record.key] = record.get_bytes_as('chunked')
2430
2044
raise errors.NoSuchRevision(self, record.key)
2431
if order_as_requested:
2432
# Yield as many results as we can while preserving order.
2433
while next_key in text_chunks:
2434
chunks = text_chunks.pop(next_key)
2435
yield ''.join(chunks), next_key[-1]
2437
next_key = key_iter.next()
2438
except StopIteration:
2439
# We still want to fully consume the get_record_stream,
2440
# just in case it is not actually finished at this point
2046
chunks = text_chunks.pop(key)
2047
yield ''.join(chunks), key[-1]
2444
def _deserialise_inventory(self, revision_id, xml):
2049
def deserialise_inventory(self, revision_id, xml):
2445
2050
"""Transform the xml into an inventory object.
2447
2052
:param revision_id: The expected revision id of the inventory.
2448
2053
:param xml: A serialised inventory.
2450
2055
result = self._serializer.read_inventory_from_string(xml, revision_id,
2451
entry_cache=self._inventory_entry_cache,
2452
return_from_cache=self._safe_to_return_from_cache)
2056
entry_cache=self._inventory_entry_cache)
2453
2057
if result.revision_id != revision_id:
2454
2058
raise AssertionError('revision id mismatch %s != %s' % (
2455
2059
result.revision_id, revision_id))
2062
def serialise_inventory(self, inv):
2063
return self._serializer.write_inventory_to_string(inv)
2065
def _serialise_inventory_to_lines(self, inv):
2066
return self._serializer.write_inventory_to_lines(inv)
2458
2068
def get_serializer_format(self):
2459
2069
return self._serializer.format_num
2461
2071
@needs_read_lock
2462
def _get_inventory_xml(self, revision_id):
2463
"""Get serialized inventory as a string."""
2464
texts = self._iter_inventory_xmls([revision_id], 'unordered')
2072
def get_inventory_xml(self, revision_id):
2073
"""Get inventory XML as a file object."""
2074
texts = self._iter_inventory_xmls([revision_id])
2466
2076
text, revision_id = texts.next()
2467
2077
except StopIteration:
2468
2078
raise errors.HistoryMissing(self, 'inventory', revision_id)
2471
def get_rev_id_for_revno(self, revno, known_pair):
2472
"""Return the revision id of a revno, given a later (revno, revid)
2473
pair in the same history.
2475
:return: if found (True, revid). If the available history ran out
2476
before reaching the revno, then this returns
2477
(False, (closest_revno, closest_revid)).
2082
def get_inventory_sha1(self, revision_id):
2083
"""Return the sha1 hash of the inventory entry
2479
known_revno, known_revid = known_pair
2480
partial_history = [known_revid]
2481
distance_from_known = known_revno - revno
2482
if distance_from_known < 0:
2484
'requested revno (%d) is later than given known revno (%d)'
2485
% (revno, known_revno))
2488
self, partial_history, stop_index=distance_from_known)
2489
except errors.RevisionNotPresent, err:
2490
if err.revision_id == known_revid:
2491
# The start revision (known_revid) wasn't found.
2493
# This is a stacked repository with no fallbacks, or a there's a
2494
# left-hand ghost. Either way, even though the revision named in
2495
# the error isn't in this repo, we know it's the next step in this
2496
# left-hand history.
2497
partial_history.append(err.revision_id)
2498
if len(partial_history) <= distance_from_known:
2499
# Didn't find enough history to get a revid for the revno.
2500
earliest_revno = known_revno - len(partial_history) + 1
2501
return (False, (earliest_revno, partial_history[-1]))
2502
if len(partial_history) - 1 > distance_from_known:
2503
raise AssertionError('_iter_for_revno returned too much history')
2504
return (True, partial_history[-1])
2085
return self.get_revision(revision_id).inventory_sha1
2506
2087
def iter_reverse_revision_history(self, revision_id):
2507
2088
"""Iterate backwards through revision ids in the lefthand history
2759
2335
return record.get_bytes_as('fulltext')
2761
2337
@needs_read_lock
2762
def check(self, revision_ids=None, callback_refs=None, check_repo=True):
2338
def check(self, revision_ids=None):
2763
2339
"""Check consistency of all history of given revision_ids.
2765
2341
Different repository implementations should override _check().
2767
2343
:param revision_ids: A non-empty list of revision_ids whose ancestry
2768
2344
will be checked. Typically the last revision_id of a branch.
2769
:param callback_refs: A dict of check-refs to resolve and callback
2770
the check/_check method on the items listed as wanting the ref.
2772
:param check_repo: If False do not check the repository contents, just
2773
calculate the data callback_refs requires and call them back.
2775
return self._check(revision_ids, callback_refs=callback_refs,
2776
check_repo=check_repo)
2346
return self._check(revision_ids)
2778
def _check(self, revision_ids, callback_refs, check_repo):
2779
result = check.Check(self, check_repo=check_repo)
2780
result.check(callback_refs)
2348
def _check(self, revision_ids):
2349
result = check.Check(self)
2783
def _warn_if_deprecated(self, branch=None):
2353
def _warn_if_deprecated(self):
2784
2354
global _deprecation_warning_done
2785
2355
if _deprecation_warning_done:
2789
conf = config.GlobalConfig()
2791
conf = branch.get_config()
2792
if conf.suppress_warning('format_deprecation'):
2794
warning("Format %s for %s is deprecated -"
2795
" please use 'bzr upgrade' to get better performance"
2796
% (self._format, self.bzrdir.transport.base))
2798
_deprecation_warning_done = True
2357
_deprecation_warning_done = True
2358
warning("Format %s for %s is deprecated - please use 'bzr upgrade' to get better performance"
2359
% (self._format, self.bzrdir.transport.base))
2800
2361
def supports_rich_root(self):
2801
2362
return self._format.rich_root_data
3740
3242
return self.source.revision_ids_to_search_result(result_set)
3743
class InterDifferingSerializer(InterRepository):
3245
class InterPackRepo(InterSameDataRepository):
3246
"""Optimised code paths between Pack based repositories."""
3249
def _get_repo_format_to_test(self):
3250
from bzrlib.repofmt import pack_repo
3251
return pack_repo.RepositoryFormatKnitPack1()
3254
def is_compatible(source, target):
3255
"""Be compatible with known Pack formats.
3257
We don't test for the stores being of specific types because that
3258
could lead to confusing results, and there is no need to be
3261
from bzrlib.repofmt.pack_repo import RepositoryFormatPack
3263
are_packs = (isinstance(source._format, RepositoryFormatPack) and
3264
isinstance(target._format, RepositoryFormatPack))
3265
except AttributeError:
3267
return are_packs and InterRepository._same_model(source, target)
3270
def fetch(self, revision_id=None, pb=None, find_ghosts=False,
3272
"""See InterRepository.fetch()."""
3273
if (len(self.source._fallback_repositories) > 0 or
3274
len(self.target._fallback_repositories) > 0):
3275
# The pack layer is not aware of fallback repositories, so when
3276
# fetching from a stacked repository or into a stacked repository
3277
# we use the generic fetch logic which uses the VersionedFiles
3278
# attributes on repository.
3279
from bzrlib.fetch import RepoFetcher
3280
fetcher = RepoFetcher(self.target, self.source, revision_id,
3281
pb, find_ghosts, fetch_spec=fetch_spec)
3282
if fetch_spec is not None:
3283
if len(list(fetch_spec.heads)) != 1:
3284
raise AssertionError(
3285
"InterPackRepo.fetch doesn't support "
3286
"fetching multiple heads yet.")
3287
revision_id = list(fetch_spec.heads)[0]
3289
if revision_id is None:
3291
# everything to do - use pack logic
3292
# to fetch from all packs to one without
3293
# inventory parsing etc, IFF nothing to be copied is in the target.
3295
source_revision_ids = frozenset(self.source.all_revision_ids())
3296
revision_ids = source_revision_ids - \
3297
frozenset(self.target.get_parent_map(source_revision_ids))
3298
revision_keys = [(revid,) for revid in revision_ids]
3299
index = self.target._pack_collection.revision_index.combined_index
3300
present_revision_ids = set(item[1][0] for item in
3301
index.iter_entries(revision_keys))
3302
revision_ids = set(revision_ids) - present_revision_ids
3303
# implementing the TODO will involve:
3304
# - detecting when all of a pack is selected
3305
# - avoiding as much as possible pre-selection, so the
3306
# more-core routines such as create_pack_from_packs can filter in
3307
# a just-in-time fashion. (though having a HEADS list on a
3308
# repository might make this a lot easier, because we could
3309
# sensibly detect 'new revisions' without doing a full index scan.
3310
elif _mod_revision.is_null(revision_id):
3315
revision_ids = self.search_missing_revision_ids(revision_id,
3316
find_ghosts=find_ghosts).get_keys()
3317
except errors.NoSuchRevision:
3318
raise errors.InstallFailed([revision_id])
3319
if len(revision_ids) == 0:
3321
return self._pack(self.source, self.target, revision_ids)
3323
def _pack(self, source, target, revision_ids):
3324
from bzrlib.repofmt.pack_repo import Packer
3325
packs = source._pack_collection.all_packs()
3326
pack = Packer(self.target._pack_collection, packs, '.fetch',
3327
revision_ids).pack()
3328
if pack is not None:
3329
self.target._pack_collection._save_pack_names()
3330
copied_revs = pack.get_revision_count()
3331
# Trigger an autopack. This may duplicate effort as we've just done
3332
# a pack creation, but for now it is simpler to think about as
3333
# 'upload data, then repack if needed'.
3334
self.target._pack_collection.autopack()
3335
return (copied_revs, [])
3340
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
3341
"""See InterRepository.missing_revision_ids().
3343
:param find_ghosts: Find ghosts throughout the ancestry of
3346
if not find_ghosts and revision_id is not None:
3347
return self._walk_to_common_revisions([revision_id])
3348
elif revision_id is not None:
3349
# Find ghosts: search for revisions pointing from one repository to
3350
# the other, and vice versa, anywhere in the history of revision_id.
3351
graph = self.target.get_graph(other_repository=self.source)
3352
searcher = graph._make_breadth_first_searcher([revision_id])
3356
next_revs, ghosts = searcher.next_with_ghosts()
3357
except StopIteration:
3359
if revision_id in ghosts:
3360
raise errors.NoSuchRevision(self.source, revision_id)
3361
found_ids.update(next_revs)
3362
found_ids.update(ghosts)
3363
found_ids = frozenset(found_ids)
3364
# Double query here: should be able to avoid this by changing the
3365
# graph api further.
3366
result_set = found_ids - frozenset(
3367
self.target.get_parent_map(found_ids))
3369
source_ids = self.source.all_revision_ids()
3370
# source_ids is the worst possible case we may need to pull.
3371
# now we want to filter source_ids against what we actually
3372
# have in target, but don't try to check for existence where we know
3373
# we do not have a revision as that would be pointless.
3374
target_ids = set(self.target.all_revision_ids())
3375
result_set = set(source_ids).difference(target_ids)
3376
return self.source.revision_ids_to_search_result(result_set)
3379
class InterDifferingSerializer(InterKnitRepo):
3746
3382
def _get_repo_format_to_test(self):
3750
3386
def is_compatible(source, target):
3751
3387
"""Be compatible with Knit2 source and Knit3 target"""
3752
# This is redundant with format.check_conversion_target(), however that
3753
# raises an exception, and we just want to say "False" as in we won't
3754
# support converting between these formats.
3755
if 'IDS_never' in debug.debug_flags:
3757
if source.supports_rich_root() and not target.supports_rich_root():
3759
if (source._format.supports_tree_reference
3760
and not target._format.supports_tree_reference):
3762
if target._fallback_repositories and target._format.supports_chks:
3763
# IDS doesn't know how to copy CHKs for the parent inventories it
3764
# adds to stacked repos.
3766
if 'IDS_always' in debug.debug_flags:
3768
# Only use this code path for local source and target. IDS does far
3769
# too much IO (both bandwidth and roundtrips) over a network.
3770
if not source.bzrdir.transport.base.startswith('file:///'):
3772
if not target.bzrdir.transport.base.startswith('file:///'):
3388
if source.supports_rich_root() != target.supports_rich_root():
3390
# Ideally, we'd support fetching if the source had no tree references
3391
# even if it supported them...
3392
if (getattr(source, '_format.supports_tree_reference', False) and
3393
not getattr(target, '_format.supports_tree_reference', False)):
3776
def _get_trees(self, revision_ids, cache):
3778
for rev_id in revision_ids:
3780
possible_trees.append((rev_id, cache[rev_id]))
3782
# Not cached, but inventory might be present anyway.
3784
tree = self.source.revision_tree(rev_id)
3785
except errors.NoSuchRevision:
3786
# Nope, parent is ghost.
3789
cache[rev_id] = tree
3790
possible_trees.append((rev_id, tree))
3791
return possible_trees
3793
def _get_delta_for_revision(self, tree, parent_ids, possible_trees):
3397
def _get_delta_for_revision(self, tree, parent_ids, basis_id, cache):
3794
3398
"""Get the best delta and base for this revision.
3796
3400
:return: (basis_id, delta)
3402
possible_trees = [(parent_id, cache[parent_id])
3403
for parent_id in parent_ids
3404
if parent_id in cache]
3405
if len(possible_trees) == 0:
3406
# There either aren't any parents, or the parents aren't in the
3407
# cache, so just use the last converted tree
3408
possible_trees.append((basis_id, cache[basis_id]))
3799
# Generate deltas against each tree, to find the shortest.
3800
texts_possibly_new_in_tree = set()
3801
3410
for basis_id, basis_tree in possible_trees:
3802
3411
delta = tree.inventory._make_delta(basis_tree.inventory)
3803
for old_path, new_path, file_id, new_entry in delta:
3804
if new_path is None:
3805
# This file_id isn't present in the new rev, so we don't
3809
# Rich roots are handled elsewhere...
3811
kind = new_entry.kind
3812
if kind != 'directory' and kind != 'file':
3813
# No text record associated with this inventory entry.
3815
# This is a directory or file that has changed somehow.
3816
texts_possibly_new_in_tree.add((file_id, new_entry.revision))
3817
3412
deltas.append((len(delta), basis_id, delta))
3819
3414
return deltas[0][1:]
3821
def _fetch_parent_invs_for_stacking(self, parent_map, cache):
3822
"""Find all parent revisions that are absent, but for which the
3823
inventory is present, and copy those inventories.
3825
This is necessary to preserve correctness when the source is stacked
3826
without fallbacks configured. (Note that in cases like upgrade the
3827
source may be not have _fallback_repositories even though it is
3831
for parents in parent_map.values():
3832
parent_revs.update(parents)
3833
present_parents = self.source.get_parent_map(parent_revs)
3834
absent_parents = set(parent_revs).difference(present_parents)
3835
parent_invs_keys_for_stacking = self.source.inventories.get_parent_map(
3836
(rev_id,) for rev_id in absent_parents)
3837
parent_inv_ids = [key[-1] for key in parent_invs_keys_for_stacking]
3838
for parent_tree in self.source.revision_trees(parent_inv_ids):
3839
current_revision_id = parent_tree.get_revision_id()
3840
parents_parents_keys = parent_invs_keys_for_stacking[
3841
(current_revision_id,)]
3842
parents_parents = [key[-1] for key in parents_parents_keys]
3843
basis_id = _mod_revision.NULL_REVISION
3844
basis_tree = self.source.revision_tree(basis_id)
3845
delta = parent_tree.inventory._make_delta(basis_tree.inventory)
3846
self.target.add_inventory_by_delta(
3847
basis_id, delta, current_revision_id, parents_parents)
3848
cache[current_revision_id] = parent_tree
3850
def _fetch_batch(self, revision_ids, basis_id, cache, a_graph=None):
3416
def _fetch_batch(self, revision_ids, basis_id, cache):
3851
3417
"""Fetch across a few revisions.
3853
3419
:param revision_ids: The revisions to copy
3854
3420
:param basis_id: The revision_id of a tree that must be in cache, used
3855
3421
as a basis for delta when no other base is available
3856
3422
:param cache: A cache of RevisionTrees that we can use.
3857
:param a_graph: A Graph object to determine the heads() of the
3858
rich-root data stream.
3859
3423
:return: The revision_id of the last converted tree. The RevisionTree
3860
3424
for it will be in cache
3862
3426
# Walk though all revisions; get inventory deltas, copy referenced
3863
3427
# texts that delta references, insert the delta, revision and
3865
root_keys_to_create = set()
3866
3429
text_keys = set()
3867
3430
pending_deltas = []
3868
3431
pending_revisions = []
3869
3432
parent_map = self.source.get_parent_map(revision_ids)
3870
self._fetch_parent_invs_for_stacking(parent_map, cache)
3871
self.source._safe_to_return_from_cache = True
3872
3433
for tree in self.source.revision_trees(revision_ids):
3873
# Find a inventory delta for this revision.
3874
# Find text entries that need to be copied, too.
3875
3434
current_revision_id = tree.get_revision_id()
3876
3435
parent_ids = parent_map.get(current_revision_id, ())
3877
parent_trees = self._get_trees(parent_ids, cache)
3878
possible_trees = list(parent_trees)
3879
if len(possible_trees) == 0:
3880
# There either aren't any parents, or the parents are ghosts,
3881
# so just use the last converted tree.
3882
possible_trees.append((basis_id, cache[basis_id]))
3883
3436
basis_id, delta = self._get_delta_for_revision(tree, parent_ids,
3438
# Find text entries that need to be copied
3439
for old_path, new_path, file_id, entry in delta:
3440
if new_path is not None:
3441
if not (new_path or self.target.supports_rich_root()):
3442
# We don't copy the text for the root node unless the
3443
# target supports_rich_root.
3445
text_keys.add((file_id, entry.revision))
3885
3446
revision = self.source.get_revision(current_revision_id)
3886
3447
pending_deltas.append((basis_id, delta,
3887
3448
current_revision_id, revision.parent_ids))
3888
if self._converting_to_rich_root:
3889
self._revision_id_to_root_id[current_revision_id] = \
3891
# Determine which texts are in present in this revision but not in
3892
# any of the available parents.
3893
texts_possibly_new_in_tree = set()
3894
for old_path, new_path, file_id, entry in delta:
3895
if new_path is None:
3896
# This file_id isn't present in the new rev
3900
if not self.target.supports_rich_root():
3901
# The target doesn't support rich root, so we don't
3904
if self._converting_to_rich_root:
3905
# This can't be copied normally, we have to insert
3907
root_keys_to_create.add((file_id, entry.revision))
3910
texts_possibly_new_in_tree.add((file_id, entry.revision))
3911
for basis_id, basis_tree in possible_trees:
3912
basis_inv = basis_tree.inventory
3913
for file_key in list(texts_possibly_new_in_tree):
3914
file_id, file_revision = file_key
3916
entry = basis_inv[file_id]
3917
except errors.NoSuchId:
3919
if entry.revision == file_revision:
3920
texts_possibly_new_in_tree.remove(file_key)
3921
text_keys.update(texts_possibly_new_in_tree)
3922
3449
pending_revisions.append(revision)
3923
3450
cache[current_revision_id] = tree
3924
3451
basis_id = current_revision_id
3925
self.source._safe_to_return_from_cache = False
3926
3452
# Copy file texts
3927
3453
from_texts = self.source.texts
3928
3454
to_texts = self.target.texts
3929
if root_keys_to_create:
3930
root_stream = _mod_fetch._new_root_data_stream(
3931
root_keys_to_create, self._revision_id_to_root_id, parent_map,
3932
self.source, graph=a_graph)
3933
to_texts.insert_record_stream(root_stream)
3934
3455
to_texts.insert_record_stream(from_texts.get_record_stream(
3935
3456
text_keys, self.target._format._fetch_order,
3936
3457
not self.target._format._fetch_uses_deltas))
3937
# insert inventory deltas
3938
3459
for delta in pending_deltas:
3939
3460
self.target.add_inventory_by_delta(*delta)
3940
if self.target._fallback_repositories:
3941
# Make sure this stacked repository has all the parent inventories
3942
# for the new revisions that we are about to insert. We do this
3943
# before adding the revisions so that no revision is added until
3944
# all the inventories it may depend on are added.
3945
# Note that this is overzealous, as we may have fetched these in an
3948
revision_ids = set()
3949
for revision in pending_revisions:
3950
revision_ids.add(revision.revision_id)
3951
parent_ids.update(revision.parent_ids)
3952
parent_ids.difference_update(revision_ids)
3953
parent_ids.discard(_mod_revision.NULL_REVISION)
3954
parent_map = self.source.get_parent_map(parent_ids)
3955
# we iterate over parent_map and not parent_ids because we don't
3956
# want to try copying any revision which is a ghost
3957
for parent_tree in self.source.revision_trees(parent_map):
3958
current_revision_id = parent_tree.get_revision_id()
3959
parents_parents = parent_map[current_revision_id]
3960
possible_trees = self._get_trees(parents_parents, cache)
3961
if len(possible_trees) == 0:
3962
# There either aren't any parents, or the parents are
3963
# ghosts, so just use the last converted tree.
3964
possible_trees.append((basis_id, cache[basis_id]))
3965
basis_id, delta = self._get_delta_for_revision(parent_tree,
3966
parents_parents, possible_trees)
3967
self.target.add_inventory_by_delta(
3968
basis_id, delta, current_revision_id, parents_parents)
3969
3461
# insert signatures and revisions
3970
3462
for revision in pending_revisions:
4384
3796
# missing keys can handle suspending a write group).
4385
3797
write_group_tokens = self.target_repo.suspend_write_group()
4386
3798
return write_group_tokens, missing_keys
4387
hint = self.target_repo.commit_write_group()
4388
if (to_serializer != src_serializer and
4389
self.target_repo._format.pack_compresses):
4390
self.target_repo.pack(hint=hint)
3799
self.target_repo.commit_write_group()
4391
3800
return [], set()
4393
def _extract_and_insert_inventory_deltas(self, substream, serializer):
4394
target_rich_root = self.target_repo._format.rich_root_data
4395
target_tree_refs = self.target_repo._format.supports_tree_reference
4396
for record in substream:
4397
# Insert the delta directly
4398
inventory_delta_bytes = record.get_bytes_as('fulltext')
4399
deserialiser = inventory_delta.InventoryDeltaDeserializer()
4401
parse_result = deserialiser.parse_text_bytes(
4402
inventory_delta_bytes)
4403
except inventory_delta.IncompatibleInventoryDelta, err:
4404
trace.mutter("Incompatible delta: %s", err.msg)
4405
raise errors.IncompatibleRevision(self.target_repo._format)
4406
basis_id, new_id, rich_root, tree_refs, inv_delta = parse_result
4407
revision_id = new_id
4408
parents = [key[0] for key in record.parents]
4409
self.target_repo.add_inventory_by_delta(
4410
basis_id, inv_delta, revision_id, parents)
4412
def _extract_and_insert_inventories(self, substream, serializer,
3802
def _extract_and_insert_inventories(self, substream, serializer):
4414
3803
"""Generate a new inventory versionedfile in target, converting data.
4416
3805
The inventory is retrieved from the source, (deserializing it), and
4417
3806
stored in the target (reserializing it in a different format).
4419
target_rich_root = self.target_repo._format.rich_root_data
4420
target_tree_refs = self.target_repo._format.supports_tree_reference
4421
3808
for record in substream:
4422
# It's not a delta, so it must be a fulltext in the source
4423
# serializer's format.
4424
3809
bytes = record.get_bytes_as('fulltext')
4425
3810
revision_id = record.key[0]
4426
3811
inv = serializer.read_inventory_from_string(bytes, revision_id)
4427
3812
parents = [key[0] for key in record.parents]
4428
3813
self.target_repo.add_inventory(revision_id, inv, parents)
4429
# No need to keep holding this full inv in memory when the rest of
4430
# the substream is likely to be all deltas.
4433
3815
def _extract_and_insert_revisions(self, substream, serializer):
4434
3816
for record in substream:
4597
3964
return (not self.from_repository._format.rich_root_data and
4598
3965
self.to_format.rich_root_data)
4600
def _get_inventory_stream(self, revision_ids, missing=False):
4601
from_format = self.from_repository._format
4602
if (from_format.supports_chks and self.to_format.supports_chks and
4603
from_format.network_name() == self.to_format.network_name()):
4604
raise AssertionError(
4605
"this case should be handled by GroupCHKStreamSource")
4606
elif 'forceinvdeltas' in debug.debug_flags:
4607
return self._get_convertable_inventory_stream(revision_ids,
4608
delta_versus_null=missing)
4609
elif from_format.network_name() == self.to_format.network_name():
4611
return self._get_simple_inventory_stream(revision_ids,
4613
elif (not from_format.supports_chks and not self.to_format.supports_chks
4614
and from_format._serializer == self.to_format._serializer):
4615
# Essentially the same format.
4616
return self._get_simple_inventory_stream(revision_ids,
4619
# Any time we switch serializations, we want to use an
4620
# inventory-delta based approach.
4621
return self._get_convertable_inventory_stream(revision_ids,
4622
delta_versus_null=missing)
4624
def _get_simple_inventory_stream(self, revision_ids, missing=False):
4625
# NB: This currently reopens the inventory weave in source;
4626
# using a single stream interface instead would avoid this.
4627
from_weave = self.from_repository.inventories
4629
delta_closure = True
4631
delta_closure = not self.delta_on_metadata()
4632
yield ('inventories', from_weave.get_record_stream(
4633
[(rev_id,) for rev_id in revision_ids],
4634
self.inventory_fetch_order(), delta_closure))
4636
def _get_convertable_inventory_stream(self, revision_ids,
4637
delta_versus_null=False):
4638
# The two formats are sufficiently different that there is no fast
4639
# path, so we need to send just inventorydeltas, which any
4640
# sufficiently modern client can insert into any repository.
4641
# The StreamSink code expects to be able to
4642
# convert on the target, so we need to put bytes-on-the-wire that can
4643
# be converted. That means inventory deltas (if the remote is <1.19,
4644
# RemoteStreamSink will fallback to VFS to insert the deltas).
4645
yield ('inventory-deltas',
4646
self._stream_invs_as_deltas(revision_ids,
4647
delta_versus_null=delta_versus_null))
4649
def _stream_invs_as_deltas(self, revision_ids, delta_versus_null=False):
4650
"""Return a stream of inventory-deltas for the given rev ids.
4652
:param revision_ids: The list of inventories to transmit
4653
:param delta_versus_null: Don't try to find a minimal delta for this
4654
entry, instead compute the delta versus the NULL_REVISION. This
4655
effectively streams a complete inventory. Used for stuff like
4656
filling in missing parents, etc.
4658
from_repo = self.from_repository
4659
revision_keys = [(rev_id,) for rev_id in revision_ids]
4660
parent_map = from_repo.inventories.get_parent_map(revision_keys)
4661
# XXX: possibly repos could implement a more efficient iter_inv_deltas
4663
inventories = self.from_repository.iter_inventories(
4664
revision_ids, 'topological')
4665
format = from_repo._format
4666
invs_sent_so_far = set([_mod_revision.NULL_REVISION])
4667
inventory_cache = lru_cache.LRUCache(50)
4668
null_inventory = from_repo.revision_tree(
4669
_mod_revision.NULL_REVISION).inventory
4670
# XXX: ideally the rich-root/tree-refs flags would be per-revision, not
4671
# per-repo (e.g. streaming a non-rich-root revision out of a rich-root
4672
# repo back into a non-rich-root repo ought to be allowed)
4673
serializer = inventory_delta.InventoryDeltaSerializer(
4674
versioned_root=format.rich_root_data,
4675
tree_references=format.supports_tree_reference)
4676
for inv in inventories:
4677
key = (inv.revision_id,)
4678
parent_keys = parent_map.get(key, ())
4680
if not delta_versus_null and parent_keys:
4681
# The caller did not ask for complete inventories and we have
4682
# some parents that we can delta against. Make a delta against
4683
# each parent so that we can find the smallest.
4684
parent_ids = [parent_key[0] for parent_key in parent_keys]
4685
for parent_id in parent_ids:
4686
if parent_id not in invs_sent_so_far:
4687
# We don't know that the remote side has this basis, so
4690
if parent_id == _mod_revision.NULL_REVISION:
4691
parent_inv = null_inventory
4693
parent_inv = inventory_cache.get(parent_id, None)
4694
if parent_inv is None:
4695
parent_inv = from_repo.get_inventory(parent_id)
4696
candidate_delta = inv._make_delta(parent_inv)
4697
if (delta is None or
4698
len(delta) > len(candidate_delta)):
4699
delta = candidate_delta
4700
basis_id = parent_id
4702
# Either none of the parents ended up being suitable, or we
4703
# were asked to delta against NULL
4704
basis_id = _mod_revision.NULL_REVISION
4705
delta = inv._make_delta(null_inventory)
4706
invs_sent_so_far.add(inv.revision_id)
4707
inventory_cache[inv.revision_id] = inv
4708
delta_serialized = ''.join(
4709
serializer.delta_to_lines(basis_id, key[-1], delta))
4710
yield versionedfile.FulltextContentFactory(
4711
key, parent_keys, None, delta_serialized)
4714
def _iter_for_revno(repo, partial_history_cache, stop_index=None,
4715
stop_revision=None):
4716
"""Extend the partial history to include a given index
4718
If a stop_index is supplied, stop when that index has been reached.
4719
If a stop_revision is supplied, stop when that revision is
4720
encountered. Otherwise, stop when the beginning of history is
4723
:param stop_index: The index which should be present. When it is
4724
present, history extension will stop.
4725
:param stop_revision: The revision id which should be present. When
4726
it is encountered, history extension will stop.
4728
start_revision = partial_history_cache[-1]
4729
iterator = repo.iter_reverse_revision_history(start_revision)
4731
#skip the last revision in the list
4734
if (stop_index is not None and
4735
len(partial_history_cache) > stop_index):
4737
if partial_history_cache[-1] == stop_revision:
4739
revision_id = iterator.next()
4740
partial_history_cache.append(revision_id)
4741
except StopIteration:
added
removed
bzrlib/repository.py
