830
811
self.new_inventory = None
831
812
if len(inv_delta):
832
# This should perhaps be guarded by a check that the basis we
833
# commit against is the basis for the commit and if not do a delta
835
813
self._any_changes = True
836
814
if not seen_root:
837
815
# housekeeping root entry changes do not affect no-change commits.
838
816
self._require_root_change(tree)
839
817
self.basis_delta_revision = basis_revision_id
841
def _add_text_to_weave(self, file_id, new_text, parents, nostore_sha):
842
parent_keys = tuple([(file_id, parent) for parent in parents])
843
return self.repository.texts._add_text(
844
(file_id, self._new_revision_id), parent_keys, new_text,
845
nostore_sha=nostore_sha, random_id=self.random_revid)[0:2]
819
def _add_text_to_weave(self, file_id, new_lines, parents, nostore_sha):
820
# Note: as we read the content directly from the tree, we know its not
821
# been turned into unicode or badly split - but a broken tree
822
# implementation could give us bad output from readlines() so this is
823
# not a guarantee of safety. What would be better is always checking
824
# the content during test suite execution. RBC 20070912
825
parent_keys = tuple((file_id, parent) for parent in parents)
826
return self.repository.texts.add_lines(
827
(file_id, self._new_revision_id), parent_keys, new_lines,
828
nostore_sha=nostore_sha, random_id=self.random_revid,
829
check_content=False)[0:2]
848
832
class RootCommitBuilder(CommitBuilder):
900
866
which views a particular line of development through that history.
902
868
The Repository builds on top of some byte storage facilies (the revisions,
903
signatures, inventories, texts and chk_bytes attributes) and a Transport,
904
which respectively provide byte storage and a means to access the (possibly
869
signatures, inventories and texts attributes) and a Transport, which
870
respectively provide byte storage and a means to access the (possibly
907
873
The byte storage facilities are addressed via tuples, which we refer to
908
874
as 'keys' throughout the code base. Revision_keys, inventory_keys and
909
875
signature_keys are all 1-tuples: (revision_id,). text_keys are two-tuples:
910
(file_id, revision_id). chk_bytes uses CHK keys - a 1-tuple with a single
911
byte string made up of a hash identifier and a hash value.
912
We use this interface because it allows low friction with the underlying
913
code that implements disk indices, network encoding and other parts of
876
(file_id, revision_id). We use this interface because it allows low
877
friction with the underlying code that implements disk indices, network
878
encoding and other parts of bzrlib.
916
880
:ivar revisions: A bzrlib.versionedfile.VersionedFiles instance containing
917
881
the serialised revisions for the repository. This can be used to obtain
1046
992
" id and insertion revid (%r, %r)"
1047
993
% (inv.revision_id, revision_id))
1048
994
if inv.root is None:
1049
raise errors.RootMissing()
1050
return self._add_inventory_checked(revision_id, inv, parents)
1052
def _add_inventory_checked(self, revision_id, inv, parents):
1053
"""Add inv to the repository after checking the inputs.
1055
This function can be overridden to allow different inventory styles.
1057
:seealso: add_inventory, for the contract.
1059
inv_lines = self._serializer.write_inventory_to_lines(inv)
995
raise AssertionError()
996
inv_lines = self._serialise_inventory_to_lines(inv)
1060
997
return self._inventory_add_lines(revision_id, parents,
1061
998
inv_lines, check_content=False)
1063
1000
def add_inventory_by_delta(self, basis_revision_id, delta, new_revision_id,
1064
parents, basis_inv=None, propagate_caches=False):
1065
1002
"""Add a new inventory expressed as a delta against another revision.
1067
See the inventory developers documentation for the theory behind
1070
1004
:param basis_revision_id: The inventory id the delta was created
1071
1005
against. (This does not have to be a direct parent.)
1072
1006
:param delta: The inventory delta (see Inventory.apply_delta for
1196
1123
# The old API returned a list, should this actually be a set?
1197
1124
return parent_map.keys()
1199
def _check_inventories(self, checker):
1200
"""Check the inventories found from the revision scan.
1202
This is responsible for verifying the sha1 of inventories and
1203
creating a pending_keys set that covers data referenced by inventories.
1205
bar = ui.ui_factory.nested_progress_bar()
1207
self._do_check_inventories(checker, bar)
1211
def _do_check_inventories(self, checker, bar):
1212
"""Helper for _check_inventories."""
1214
keys = {'chk_bytes':set(), 'inventories':set(), 'texts':set()}
1215
kinds = ['chk_bytes', 'texts']
1216
count = len(checker.pending_keys)
1217
bar.update("inventories", 0, 2)
1218
current_keys = checker.pending_keys
1219
checker.pending_keys = {}
1220
# Accumulate current checks.
1221
for key in current_keys:
1222
if key[0] != 'inventories' and key[0] not in kinds:
1223
checker._report_items.append('unknown key type %r' % (key,))
1224
keys[key[0]].add(key[1:])
1225
if keys['inventories']:
1226
# NB: output order *should* be roughly sorted - topo or
1227
# inverse topo depending on repository - either way decent
1228
# to just delta against. However, pre-CHK formats didn't
1229
# try to optimise inventory layout on disk. As such the
1230
# pre-CHK code path does not use inventory deltas.
1232
for record in self.inventories.check(keys=keys['inventories']):
1233
if record.storage_kind == 'absent':
1234
checker._report_items.append(
1235
'Missing inventory {%s}' % (record.key,))
1237
last_object = self._check_record('inventories', record,
1238
checker, last_object,
1239
current_keys[('inventories',) + record.key])
1240
del keys['inventories']
1243
bar.update("texts", 1)
1244
while (checker.pending_keys or keys['chk_bytes']
1246
# Something to check.
1247
current_keys = checker.pending_keys
1248
checker.pending_keys = {}
1249
# Accumulate current checks.
1250
for key in current_keys:
1251
if key[0] not in kinds:
1252
checker._report_items.append('unknown key type %r' % (key,))
1253
keys[key[0]].add(key[1:])
1254
# Check the outermost kind only - inventories || chk_bytes || texts
1258
for record in getattr(self, kind).check(keys=keys[kind]):
1259
if record.storage_kind == 'absent':
1260
checker._report_items.append(
1261
'Missing %s {%s}' % (kind, record.key,))
1263
last_object = self._check_record(kind, record,
1264
checker, last_object, current_keys[(kind,) + record.key])
1268
def _check_record(self, kind, record, checker, last_object, item_data):
1269
"""Check a single text from this repository."""
1270
if kind == 'inventories':
1271
rev_id = record.key[0]
1272
inv = self._deserialise_inventory(rev_id,
1273
record.get_bytes_as('fulltext'))
1274
if last_object is not None:
1275
delta = inv._make_delta(last_object)
1276
for old_path, path, file_id, ie in delta:
1279
ie.check(checker, rev_id, inv)
1281
for path, ie in inv.iter_entries():
1282
ie.check(checker, rev_id, inv)
1283
if self._format.fast_deltas:
1285
elif kind == 'chk_bytes':
1286
# No code written to check chk_bytes for this repo format.
1287
checker._report_items.append(
1288
'unsupported key type chk_bytes for %s' % (record.key,))
1289
elif kind == 'texts':
1290
self._check_text(record, checker, item_data)
1292
checker._report_items.append(
1293
'unknown key type %s for %s' % (kind, record.key))
1295
def _check_text(self, record, checker, item_data):
1296
"""Check a single text."""
1297
# Check it is extractable.
1298
# TODO: check length.
1299
if record.storage_kind == 'chunked':
1300
chunks = record.get_bytes_as(record.storage_kind)
1301
sha1 = osutils.sha_strings(chunks)
1302
length = sum(map(len, chunks))
1304
content = record.get_bytes_as('fulltext')
1305
sha1 = osutils.sha_string(content)
1306
length = len(content)
1307
if item_data and sha1 != item_data[1]:
1308
checker._report_items.append(
1309
'sha1 mismatch: %s has sha1 %s expected %s referenced by %s' %
1310
(record.key, sha1, item_data[1], item_data[2]))
1313
1127
def create(a_bzrdir):
1314
1128
"""Construct the current default format repository in a_bzrdir."""
1334
1149
self._reconcile_does_inventory_gc = True
1335
1150
self._reconcile_fixes_text_parents = False
1336
1151
self._reconcile_backsup_inventory = True
1152
# not right yet - should be more semantically clear ?
1154
# TODO: make sure to construct the right store classes, etc, depending
1155
# on whether escaping is required.
1156
self._warn_if_deprecated()
1337
1157
self._write_group = None
1338
1158
# Additional places to query for data.
1339
1159
self._fallback_repositories = []
1340
1160
# An InventoryEntry cache, used during deserialization
1341
1161
self._inventory_entry_cache = fifo_cache.FIFOCache(10*1024)
1342
# Is it safe to return inventory entries directly from the entry cache,
1343
# rather copying them?
1344
self._safe_to_return_from_cache = False
1347
def user_transport(self):
1348
return self.bzrdir.user_transport
1351
def control_transport(self):
1352
return self._transport
1354
1163
def __repr__(self):
1355
if self._fallback_repositories:
1356
return '%s(%r, fallback_repositories=%r)' % (
1357
self.__class__.__name__,
1359
self._fallback_repositories)
1361
return '%s(%r)' % (self.__class__.__name__,
1364
def _has_same_fallbacks(self, other_repo):
1365
"""Returns true if the repositories have the same fallbacks."""
1366
my_fb = self._fallback_repositories
1367
other_fb = other_repo._fallback_repositories
1368
if len(my_fb) != len(other_fb):
1370
for f, g in zip(my_fb, other_fb):
1371
if not f.has_same_location(g):
1164
return '%s(%r)' % (self.__class__.__name__,
1375
1167
def has_same_location(self, other):
1376
1168
"""Returns a boolean indicating if this repository is at the same
1418
1204
:raises MismatchedToken: if the specified token doesn't match the token
1419
1205
of the existing lock.
1420
1206
:seealso: start_write_group.
1421
:return: A RepositoryWriteLockResult.
1208
A token should be passed in if you know that you have locked the object
1209
some other way, and need to synchronise this object's state with that
1212
XXX: this docstring is duplicated in many places, e.g. lockable_files.py
1423
1214
locked = self.is_locked()
1424
token = self.control_files.lock_write(token=token)
1215
result = self.control_files.lock_write(token=token)
1216
for repo in self._fallback_repositories:
1217
# Writes don't affect fallback repos
1426
self._warn_if_deprecated()
1427
self._note_lock('w')
1428
for repo in self._fallback_repositories:
1429
# Writes don't affect fallback repos
1431
1220
self._refresh_data()
1432
return RepositoryWriteLockResult(self.unlock, token)
1434
1223
def lock_read(self):
1435
"""Lock the repository for read operations.
1437
:return: An object with an unlock method which will release the lock
1440
1224
locked = self.is_locked()
1441
1225
self.control_files.lock_read()
1226
for repo in self._fallback_repositories:
1443
self._warn_if_deprecated()
1444
self._note_lock('r')
1445
for repo in self._fallback_repositories:
1447
1229
self._refresh_data()
1448
return LogicalLockResult(self.unlock)
1450
1231
def get_physical_lock_status(self):
1451
1232
return self.control_files.get_physical_lock_status()
1613
1397
def suspend_write_group(self):
1614
1398
raise errors.UnsuspendableWriteGroup(self)
1616
def get_missing_parent_inventories(self, check_for_missing_texts=True):
1617
"""Return the keys of missing inventory parents for revisions added in
1620
A revision is not complete if the inventory delta for that revision
1621
cannot be calculated. Therefore if the parent inventories of a
1622
revision are not present, the revision is incomplete, and e.g. cannot
1623
be streamed by a smart server. This method finds missing inventory
1624
parents for revisions added in this write group.
1626
if not self._format.supports_external_lookups:
1627
# This is only an issue for stacked repositories
1629
if not self.is_in_write_group():
1630
raise AssertionError('not in a write group')
1632
# XXX: We assume that every added revision already has its
1633
# corresponding inventory, so we only check for parent inventories that
1634
# might be missing, rather than all inventories.
1635
parents = set(self.revisions._index.get_missing_parents())
1636
parents.discard(_mod_revision.NULL_REVISION)
1637
unstacked_inventories = self.inventories._index
1638
present_inventories = unstacked_inventories.get_parent_map(
1639
key[-1:] for key in parents)
1640
parents.difference_update(present_inventories)
1641
if len(parents) == 0:
1642
# No missing parent inventories.
1644
if not check_for_missing_texts:
1645
return set(('inventories', rev_id) for (rev_id,) in parents)
1646
# Ok, now we have a list of missing inventories. But these only matter
1647
# if the inventories that reference them are missing some texts they
1648
# appear to introduce.
1649
# XXX: Texts referenced by all added inventories need to be present,
1650
# but at the moment we're only checking for texts referenced by
1651
# inventories at the graph's edge.
1652
key_deps = self.revisions._index._key_dependencies
1653
key_deps.satisfy_refs_for_keys(present_inventories)
1654
referrers = frozenset(r[0] for r in key_deps.get_referrers())
1655
file_ids = self.fileids_altered_by_revision_ids(referrers)
1656
missing_texts = set()
1657
for file_id, version_ids in file_ids.iteritems():
1658
missing_texts.update(
1659
(file_id, version_id) for version_id in version_ids)
1660
present_texts = self.texts.get_parent_map(missing_texts)
1661
missing_texts.difference_update(present_texts)
1662
if not missing_texts:
1663
# No texts are missing, so all revisions and their deltas are
1666
# Alternatively the text versions could be returned as the missing
1667
# keys, but this is likely to be less data.
1668
missing_keys = set(('inventories', rev_id) for (rev_id,) in parents)
1671
1400
def refresh_data(self):
1672
"""Re-read any data needed to synchronise with disk.
1401
"""Re-read any data needed to to synchronise with disk.
1674
1403
This method is intended to be called after another repository instance
1675
1404
(such as one used by a smart server) has inserted data into the
1676
repository. On all repositories this will work outside of write groups.
1677
Some repository formats (pack and newer for bzrlib native formats)
1678
support refresh_data inside write groups. If called inside a write
1679
group on a repository that does not support refreshing in a write group
1680
IsInWriteGroupError will be raised.
1405
repository. It may not be called during a write group, but may be
1406
called at any other time.
1408
if self.is_in_write_group():
1409
raise errors.InternalBzrError(
1410
"May not refresh_data while in a write group.")
1682
1411
self._refresh_data()
1684
1413
def resume_write_group(self, tokens):
1895
1615
@needs_read_lock
1896
1616
def get_revisions(self, revision_ids):
1897
"""Get many revisions at once.
1899
Repositories that need to check data on every revision read should
1900
subclass this method.
1617
"""Get many revisions at once."""
1902
1618
return self._get_revisions(revision_ids)
1904
1620
@needs_read_lock
1905
1621
def _get_revisions(self, revision_ids):
1906
1622
"""Core work logic to get many revisions without sanity checks."""
1623
for rev_id in revision_ids:
1624
if not rev_id or not isinstance(rev_id, basestring):
1625
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1626
keys = [(key,) for key in revision_ids]
1627
stream = self.revisions.get_record_stream(keys, 'unordered', True)
1908
for revid, rev in self._iter_revisions(revision_ids):
1910
raise errors.NoSuchRevision(self, revid)
1629
for record in stream:
1630
if record.storage_kind == 'absent':
1631
raise errors.NoSuchRevision(self, record.key[0])
1632
text = record.get_bytes_as('fulltext')
1633
rev = self._serializer.read_revision_from_string(text)
1634
revs[record.key[0]] = rev
1912
1635
return [revs[revid] for revid in revision_ids]
1914
def _iter_revisions(self, revision_ids):
1915
"""Iterate over revision objects.
1917
:param revision_ids: An iterable of revisions to examine. None may be
1918
passed to request all revisions known to the repository. Note that
1919
not all repositories can find unreferenced revisions; for those
1920
repositories only referenced ones will be returned.
1921
:return: An iterator of (revid, revision) tuples. Absent revisions (
1922
those asked for but not available) are returned as (revid, None).
1924
if revision_ids is None:
1925
revision_ids = self.all_revision_ids()
1927
for rev_id in revision_ids:
1928
if not rev_id or not isinstance(rev_id, basestring):
1929
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1930
keys = [(key,) for key in revision_ids]
1931
stream = self.revisions.get_record_stream(keys, 'unordered', True)
1932
for record in stream:
1933
revid = record.key[0]
1934
if record.storage_kind == 'absent':
1937
text = record.get_bytes_as('fulltext')
1938
rev = self._serializer.read_revision_from_string(text)
1638
def get_revision_xml(self, revision_id):
1639
# TODO: jam 20070210 This shouldn't be necessary since get_revision
1640
# would have already do it.
1641
# TODO: jam 20070210 Just use _serializer.write_revision_to_string()
1642
# TODO: this can't just be replaced by:
1643
# return self._serializer.write_revision_to_string(
1644
# self.get_revision(revision_id))
1645
# as cStringIO preservers the encoding unlike write_revision_to_string
1646
# or some other call down the path.
1647
rev = self.get_revision(revision_id)
1648
rev_tmp = cStringIO.StringIO()
1649
# the current serializer..
1650
self._serializer.write_revision(rev, rev_tmp)
1652
return rev_tmp.getvalue()
1941
1654
def get_deltas_for_revisions(self, revisions, specific_fileids=None):
1942
1655
"""Produce a generator of revision deltas.
2120
1833
yield line, revid
2122
1835
def _find_file_ids_from_xml_inventory_lines(self, line_iterator,
2124
1837
"""Helper routine for fileids_altered_by_revision_ids.
2126
1839
This performs the translation of xml lines to revision ids.
2128
1841
:param line_iterator: An iterator of lines, origin_version_id
2129
:param revision_keys: The revision ids to filter for. This should be a
1842
:param revision_ids: The revision ids to filter for. This should be a
2130
1843
set or other type which supports efficient __contains__ lookups, as
2131
the revision key from each parsed line will be looked up in the
2132
revision_keys filter.
1844
the revision id from each parsed line will be looked up in the
1845
revision_ids filter.
2133
1846
:return: a dictionary mapping altered file-ids to an iterable of
2134
1847
revision_ids. Each altered file-ids has the exact revision_ids that
2135
1848
altered it listed explicitly.
2137
1850
seen = set(self._find_text_key_references_from_xml_inventory_lines(
2138
1851
line_iterator).iterkeys())
2139
parent_keys = self._find_parent_keys_of_revisions(revision_keys)
1852
# Note that revision_ids are revision keys.
1853
parent_maps = self.revisions.get_parent_map(revision_ids)
1855
map(parents.update, parent_maps.itervalues())
1856
parents.difference_update(revision_ids)
2140
1857
parent_seen = set(self._find_text_key_references_from_xml_inventory_lines(
2141
self._inventory_xml_lines_for_keys(parent_keys)))
1858
self._inventory_xml_lines_for_keys(parents)))
2142
1859
new_keys = seen - parent_seen
2144
1861
setdefault = result.setdefault
2146
1863
setdefault(key[0], set()).add(key[-1])
2149
def _find_parent_ids_of_revisions(self, revision_ids):
2150
"""Find all parent ids that are mentioned in the revision graph.
2152
:return: set of revisions that are parents of revision_ids which are
2153
not part of revision_ids themselves
2155
parent_map = self.get_parent_map(revision_ids)
2157
map(parent_ids.update, parent_map.itervalues())
2158
parent_ids.difference_update(revision_ids)
2159
parent_ids.discard(_mod_revision.NULL_REVISION)
2162
def _find_parent_keys_of_revisions(self, revision_keys):
2163
"""Similar to _find_parent_ids_of_revisions, but used with keys.
2165
:param revision_keys: An iterable of revision_keys.
2166
:return: The parents of all revision_keys that are not already in
2169
parent_map = self.revisions.get_parent_map(revision_keys)
2171
map(parent_keys.update, parent_map.itervalues())
2172
parent_keys.difference_update(revision_keys)
2173
parent_keys.discard(_mod_revision.NULL_REVISION)
2176
1866
def fileids_altered_by_revision_ids(self, revision_ids, _inv_weave=None):
2177
1867
"""Find the file ids and versions affected by revisions.
2392
2079
:param revision_ids: The expected revision ids of the inventories.
2393
:param ordering: optional ordering, e.g. 'topological'. If not
2394
specified, the order of revision_ids will be preserved (by
2395
buffering if necessary).
2396
2080
:return: An iterator of inventories.
2398
2082
if ((None in revision_ids)
2399
2083
or (_mod_revision.NULL_REVISION in revision_ids)):
2400
2084
raise ValueError('cannot get null revision inventory')
2401
return self._iter_inventories(revision_ids, ordering)
2085
return self._iter_inventories(revision_ids)
2403
def _iter_inventories(self, revision_ids, ordering):
2087
def _iter_inventories(self, revision_ids):
2404
2088
"""single-document based inventory iteration."""
2405
inv_xmls = self._iter_inventory_xmls(revision_ids, ordering)
2406
for text, revision_id in inv_xmls:
2407
yield self._deserialise_inventory(revision_id, text)
2089
for text, revision_id in self._iter_inventory_xmls(revision_ids):
2090
yield self.deserialise_inventory(revision_id, text)
2409
def _iter_inventory_xmls(self, revision_ids, ordering):
2410
if ordering is None:
2411
order_as_requested = True
2412
ordering = 'unordered'
2414
order_as_requested = False
2092
def _iter_inventory_xmls(self, revision_ids):
2415
2093
keys = [(revision_id,) for revision_id in revision_ids]
2418
if order_as_requested:
2419
key_iter = iter(keys)
2420
next_key = key_iter.next()
2421
stream = self.inventories.get_record_stream(keys, ordering, True)
2094
stream = self.inventories.get_record_stream(keys, 'unordered', True)
2422
2095
text_chunks = {}
2423
2096
for record in stream:
2424
2097
if record.storage_kind != 'absent':
2425
chunks = record.get_bytes_as('chunked')
2426
if order_as_requested:
2427
text_chunks[record.key] = chunks
2429
yield ''.join(chunks), record.key[-1]
2098
text_chunks[record.key] = record.get_bytes_as('chunked')
2431
2100
raise errors.NoSuchRevision(self, record.key)
2432
if order_as_requested:
2433
# Yield as many results as we can while preserving order.
2434
while next_key in text_chunks:
2435
chunks = text_chunks.pop(next_key)
2436
yield ''.join(chunks), next_key[-1]
2438
next_key = key_iter.next()
2439
except StopIteration:
2440
# We still want to fully consume the get_record_stream,
2441
# just in case it is not actually finished at this point
2102
chunks = text_chunks.pop(key)
2103
yield ''.join(chunks), key[-1]
2445
def _deserialise_inventory(self, revision_id, xml):
2105
def deserialise_inventory(self, revision_id, xml):
2446
2106
"""Transform the xml into an inventory object.
2448
2108
:param revision_id: The expected revision id of the inventory.
2449
2109
:param xml: A serialised inventory.
2451
2111
result = self._serializer.read_inventory_from_string(xml, revision_id,
2452
entry_cache=self._inventory_entry_cache,
2453
return_from_cache=self._safe_to_return_from_cache)
2112
entry_cache=self._inventory_entry_cache)
2454
2113
if result.revision_id != revision_id:
2455
2114
raise AssertionError('revision id mismatch %s != %s' % (
2456
2115
result.revision_id, revision_id))
2118
def serialise_inventory(self, inv):
2119
return self._serializer.write_inventory_to_string(inv)
2121
def _serialise_inventory_to_lines(self, inv):
2122
return self._serializer.write_inventory_to_lines(inv)
2459
2124
def get_serializer_format(self):
2460
2125
return self._serializer.format_num
2462
2127
@needs_read_lock
2463
def _get_inventory_xml(self, revision_id):
2464
"""Get serialized inventory as a string."""
2465
texts = self._iter_inventory_xmls([revision_id], 'unordered')
2128
def get_inventory_xml(self, revision_id):
2129
"""Get inventory XML as a file object."""
2130
texts = self._iter_inventory_xmls([revision_id])
2467
2132
text, revision_id = texts.next()
2468
2133
except StopIteration:
2469
2134
raise errors.HistoryMissing(self, 'inventory', revision_id)
2472
def get_rev_id_for_revno(self, revno, known_pair):
2473
"""Return the revision id of a revno, given a later (revno, revid)
2474
pair in the same history.
2476
:return: if found (True, revid). If the available history ran out
2477
before reaching the revno, then this returns
2478
(False, (closest_revno, closest_revid)).
2138
def get_inventory_sha1(self, revision_id):
2139
"""Return the sha1 hash of the inventory entry
2480
known_revno, known_revid = known_pair
2481
partial_history = [known_revid]
2482
distance_from_known = known_revno - revno
2483
if distance_from_known < 0:
2485
'requested revno (%d) is later than given known revno (%d)'
2486
% (revno, known_revno))
2489
self, partial_history, stop_index=distance_from_known)
2490
except errors.RevisionNotPresent, err:
2491
if err.revision_id == known_revid:
2492
# The start revision (known_revid) wasn't found.
2494
# This is a stacked repository with no fallbacks, or a there's a
2495
# left-hand ghost. Either way, even though the revision named in
2496
# the error isn't in this repo, we know it's the next step in this
2497
# left-hand history.
2498
partial_history.append(err.revision_id)
2499
if len(partial_history) <= distance_from_known:
2500
# Didn't find enough history to get a revid for the revno.
2501
earliest_revno = known_revno - len(partial_history) + 1
2502
return (False, (earliest_revno, partial_history[-1]))
2503
if len(partial_history) - 1 > distance_from_known:
2504
raise AssertionError('_iter_for_revno returned too much history')
2505
return (True, partial_history[-1])
2141
return self.get_revision(revision_id).inventory_sha1
2507
2143
def iter_reverse_revision_history(self, revision_id):
2508
2144
"""Iterate backwards through revision ids in the lefthand history
2511
2147
ancestors will be traversed.
2513
2149
graph = self.get_graph()
2514
stop_revisions = (None, _mod_revision.NULL_REVISION)
2515
return graph.iter_lefthand_ancestry(revision_id, stop_revisions)
2150
next_id = revision_id
2152
if next_id in (None, _mod_revision.NULL_REVISION):
2155
# Note: The following line may raise KeyError in the event of
2156
# truncated history. We decided not to have a try:except:raise
2157
# RevisionNotPresent here until we see a use for it, because of the
2158
# cost in an inner loop that is by its very nature O(history).
2159
# Robert Collins 20080326
2160
parents = graph.get_parent_map([next_id])[next_id]
2161
if len(parents) == 0:
2164
next_id = parents[0]
2167
def get_revision_inventory(self, revision_id):
2168
"""Return inventory of a past revision."""
2169
# TODO: Unify this with get_inventory()
2170
# bzr 0.0.6 and later imposes the constraint that the inventory_id
2171
# must be the same as its revision, so this is trivial.
2172
if revision_id is None:
2173
# This does not make sense: if there is no revision,
2174
# then it is the current tree inventory surely ?!
2175
# and thus get_root_id() is something that looks at the last
2176
# commit on the branch, and the get_root_id is an inventory check.
2177
raise NotImplementedError
2178
# return Inventory(self.get_root_id())
2180
return self.get_inventory(revision_id)
2517
2182
def is_shared(self):
2518
2183
"""Return True if this repository is flagged as a shared repository."""
2612
2277
keys = tsort.topo_sort(parent_map)
2613
2278
return [None] + list(keys)
2615
def pack(self, hint=None, clean_obsolete_packs=False):
2616
2281
"""Compress the data within the repository.
2618
2283
This operation only makes sense for some repository types. For other
2619
2284
types it should be a no-op that just returns.
2621
2286
This stub method does not require a lock, but subclasses should use
2622
@needs_write_lock as this is a long running call it's reasonable to
2287
@needs_write_lock as this is a long running call its reasonable to
2623
2288
implicitly lock for the user.
2625
:param hint: If not supplied, the whole repository is packed.
2626
If supplied, the repository may use the hint parameter as a
2627
hint for the parts of the repository to pack. A hint can be
2628
obtained from the result of commit_write_group(). Out of
2629
date hints are simply ignored, because concurrent operations
2630
can obsolete them rapidly.
2632
:param clean_obsolete_packs: Clean obsolete packs immediately after
2636
2291
def get_transaction(self):
2637
2292
return self.control_files.get_transaction()
2639
2294
def get_parent_map(self, revision_ids):
2640
"""See graph.StackedParentsProvider.get_parent_map"""
2295
"""See graph._StackedParentsProvider.get_parent_map"""
2641
2296
# revisions index works in keys; this just works in revisions
2642
2297
# therefore wrap and unwrap
2643
2298
query_keys = []
2661
2316
def _make_parents_provider(self):
2665
def get_known_graph_ancestry(self, revision_ids):
2666
"""Return the known graph for a set of revision ids and their ancestors.
2668
st = static_tuple.StaticTuple
2669
revision_keys = [st(r_id).intern() for r_id in revision_ids]
2670
known_graph = self.revisions.get_known_graph_ancestry(revision_keys)
2671
return graph.GraphThunkIdsToKeys(known_graph)
2673
2319
def get_graph(self, other_repository=None):
2674
2320
"""Return the graph walker for this repository format"""
2675
2321
parents_provider = self._make_parents_provider()
2676
2322
if (other_repository is not None and
2677
2323
not self.has_same_location(other_repository)):
2678
parents_provider = graph.StackedParentsProvider(
2324
parents_provider = graph._StackedParentsProvider(
2679
2325
[parents_provider, other_repository._make_parents_provider()])
2680
2326
return graph.Graph(parents_provider)
2682
def _get_versioned_file_checker(self, text_key_references=None,
2328
def _get_versioned_file_checker(self, text_key_references=None):
2684
2329
"""Return an object suitable for checking versioned files.
2686
2331
:param text_key_references: if non-None, an already built
2749
2391
return record.get_bytes_as('fulltext')
2751
2393
@needs_read_lock
2752
def check(self, revision_ids=None, callback_refs=None, check_repo=True):
2394
def check(self, revision_ids=None):
2753
2395
"""Check consistency of all history of given revision_ids.
2755
2397
Different repository implementations should override _check().
2757
2399
:param revision_ids: A non-empty list of revision_ids whose ancestry
2758
2400
will be checked. Typically the last revision_id of a branch.
2759
:param callback_refs: A dict of check-refs to resolve and callback
2760
the check/_check method on the items listed as wanting the ref.
2762
:param check_repo: If False do not check the repository contents, just
2763
calculate the data callback_refs requires and call them back.
2765
return self._check(revision_ids, callback_refs=callback_refs,
2766
check_repo=check_repo)
2402
return self._check(revision_ids)
2768
def _check(self, revision_ids, callback_refs, check_repo):
2769
result = check.Check(self, check_repo=check_repo)
2770
result.check(callback_refs)
2404
def _check(self, revision_ids):
2405
result = check.Check(self)
2773
def _warn_if_deprecated(self, branch=None):
2409
def _warn_if_deprecated(self):
2774
2410
global _deprecation_warning_done
2775
2411
if _deprecation_warning_done:
2779
conf = config.GlobalConfig()
2781
conf = branch.get_config()
2782
if conf.suppress_warning('format_deprecation'):
2784
warning("Format %s for %s is deprecated -"
2785
" please use 'bzr upgrade' to get better performance"
2786
% (self._format, self.bzrdir.transport.base))
2788
_deprecation_warning_done = True
2413
_deprecation_warning_done = True
2414
warning("Format %s for %s is deprecated - please use 'bzr upgrade' to get better performance"
2415
% (self._format, self.bzrdir.transport.base))
2790
2417
def supports_rich_root(self):
2791
2418
return self._format.rich_root_data
3359
2945
'RepositoryFormatPackDevelopment2Subtree',
3362
# 1.14->1.16 go below here
3363
format_registry.register_lazy(
3364
'Bazaar development format - group compression and chk inventory'
3365
' (needs bzr.dev from 1.14)\n',
3366
'bzrlib.repofmt.groupcompress_repo',
3367
'RepositoryFormatCHK1',
3370
format_registry.register_lazy(
3371
'Bazaar development format - chk repository with bencode revision '
3372
'serialization (needs bzr.dev from 1.16)\n',
3373
'bzrlib.repofmt.groupcompress_repo',
3374
'RepositoryFormatCHK2',
3376
format_registry.register_lazy(
3377
'Bazaar repository format 2a (needs bzr 1.16 or later)\n',
3378
'bzrlib.repofmt.groupcompress_repo',
3379
'RepositoryFormat2a',
3381
format_registry.register_lazy(
3382
'Bazaar development format 8\n',
3383
'bzrlib.repofmt.groupcompress_repo',
3384
'RepositoryFormat2aSubtree',
3388
2949
class InterRepository(InterObject):
3389
2950
"""This class represents operations taking place between two repositories.
3735
3298
return self.source.revision_ids_to_search_result(result_set)
3738
class InterDifferingSerializer(InterRepository):
3301
class InterPackRepo(InterSameDataRepository):
3302
"""Optimised code paths between Pack based repositories."""
3305
def _get_repo_format_to_test(self):
3306
from bzrlib.repofmt import pack_repo
3307
return pack_repo.RepositoryFormatKnitPack1()
3310
def is_compatible(source, target):
3311
"""Be compatible with known Pack formats.
3313
We don't test for the stores being of specific types because that
3314
could lead to confusing results, and there is no need to be
3317
from bzrlib.repofmt.pack_repo import RepositoryFormatPack
3319
are_packs = (isinstance(source._format, RepositoryFormatPack) and
3320
isinstance(target._format, RepositoryFormatPack))
3321
except AttributeError:
3323
return are_packs and InterRepository._same_model(source, target)
3326
def fetch(self, revision_id=None, pb=None, find_ghosts=False,
3328
"""See InterRepository.fetch()."""
3329
if (len(self.source._fallback_repositories) > 0 or
3330
len(self.target._fallback_repositories) > 0):
3331
# The pack layer is not aware of fallback repositories, so when
3332
# fetching from a stacked repository or into a stacked repository
3333
# we use the generic fetch logic which uses the VersionedFiles
3334
# attributes on repository.
3335
from bzrlib.fetch import RepoFetcher
3336
fetcher = RepoFetcher(self.target, self.source, revision_id,
3337
pb, find_ghosts, fetch_spec=fetch_spec)
3338
if fetch_spec is not None:
3339
if len(list(fetch_spec.heads)) != 1:
3340
raise AssertionError(
3341
"InterPackRepo.fetch doesn't support "
3342
"fetching multiple heads yet.")
3343
revision_id = list(fetch_spec.heads)[0]
3345
if revision_id is None:
3347
# everything to do - use pack logic
3348
# to fetch from all packs to one without
3349
# inventory parsing etc, IFF nothing to be copied is in the target.
3351
source_revision_ids = frozenset(self.source.all_revision_ids())
3352
revision_ids = source_revision_ids - \
3353
frozenset(self.target.get_parent_map(source_revision_ids))
3354
revision_keys = [(revid,) for revid in revision_ids]
3355
index = self.target._pack_collection.revision_index.combined_index
3356
present_revision_ids = set(item[1][0] for item in
3357
index.iter_entries(revision_keys))
3358
revision_ids = set(revision_ids) - present_revision_ids
3359
# implementing the TODO will involve:
3360
# - detecting when all of a pack is selected
3361
# - avoiding as much as possible pre-selection, so the
3362
# more-core routines such as create_pack_from_packs can filter in
3363
# a just-in-time fashion. (though having a HEADS list on a
3364
# repository might make this a lot easier, because we could
3365
# sensibly detect 'new revisions' without doing a full index scan.
3366
elif _mod_revision.is_null(revision_id):
3371
revision_ids = self.search_missing_revision_ids(revision_id,
3372
find_ghosts=find_ghosts).get_keys()
3373
except errors.NoSuchRevision:
3374
raise errors.InstallFailed([revision_id])
3375
if len(revision_ids) == 0:
3377
return self._pack(self.source, self.target, revision_ids)
3379
def _pack(self, source, target, revision_ids):
3380
from bzrlib.repofmt.pack_repo import Packer
3381
packs = source._pack_collection.all_packs()
3382
pack = Packer(self.target._pack_collection, packs, '.fetch',
3383
revision_ids).pack()
3384
if pack is not None:
3385
self.target._pack_collection._save_pack_names()
3386
copied_revs = pack.get_revision_count()
3387
# Trigger an autopack. This may duplicate effort as we've just done
3388
# a pack creation, but for now it is simpler to think about as
3389
# 'upload data, then repack if needed'.
3390
self.target._pack_collection.autopack()
3391
return (copied_revs, [])
3396
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
3397
"""See InterRepository.missing_revision_ids().
3399
:param find_ghosts: Find ghosts throughout the ancestry of
3402
if not find_ghosts and revision_id is not None:
3403
return self._walk_to_common_revisions([revision_id])
3404
elif revision_id is not None:
3405
# Find ghosts: search for revisions pointing from one repository to
3406
# the other, and vice versa, anywhere in the history of revision_id.
3407
graph = self.target.get_graph(other_repository=self.source)
3408
searcher = graph._make_breadth_first_searcher([revision_id])
3412
next_revs, ghosts = searcher.next_with_ghosts()
3413
except StopIteration:
3415
if revision_id in ghosts:
3416
raise errors.NoSuchRevision(self.source, revision_id)
3417
found_ids.update(next_revs)
3418
found_ids.update(ghosts)
3419
found_ids = frozenset(found_ids)
3420
# Double query here: should be able to avoid this by changing the
3421
# graph api further.
3422
result_set = found_ids - frozenset(
3423
self.target.get_parent_map(found_ids))
3425
source_ids = self.source.all_revision_ids()
3426
# source_ids is the worst possible case we may need to pull.
3427
# now we want to filter source_ids against what we actually
3428
# have in target, but don't try to check for existence where we know
3429
# we do not have a revision as that would be pointless.
3430
target_ids = set(self.target.all_revision_ids())
3431
result_set = set(source_ids).difference(target_ids)
3432
return self.source.revision_ids_to_search_result(result_set)
3435
class InterDifferingSerializer(InterKnitRepo):
3741
3438
def _get_repo_format_to_test(self):
3745
3442
def is_compatible(source, target):
3746
3443
"""Be compatible with Knit2 source and Knit3 target"""
3747
# This is redundant with format.check_conversion_target(), however that
3748
# raises an exception, and we just want to say "False" as in we won't
3749
# support converting between these formats.
3750
if 'IDS_never' in debug.debug_flags:
3752
if source.supports_rich_root() and not target.supports_rich_root():
3754
if (source._format.supports_tree_reference
3755
and not target._format.supports_tree_reference):
3757
if target._fallback_repositories and target._format.supports_chks:
3758
# IDS doesn't know how to copy CHKs for the parent inventories it
3759
# adds to stacked repos.
3761
if 'IDS_always' in debug.debug_flags:
3763
# Only use this code path for local source and target. IDS does far
3764
# too much IO (both bandwidth and roundtrips) over a network.
3765
if not source.bzrdir.transport.base.startswith('file:///'):
3767
if not target.bzrdir.transport.base.startswith('file:///'):
3444
if source.supports_rich_root() != target.supports_rich_root():
3446
# Ideally, we'd support fetching if the source had no tree references
3447
# even if it supported them...
3448
if (getattr(source, '_format.supports_tree_reference', False) and
3449
not getattr(target, '_format.supports_tree_reference', False)):
3771
def _get_trees(self, revision_ids, cache):
3773
for rev_id in revision_ids:
3775
possible_trees.append((rev_id, cache[rev_id]))
3777
# Not cached, but inventory might be present anyway.
3779
tree = self.source.revision_tree(rev_id)
3780
except errors.NoSuchRevision:
3781
# Nope, parent is ghost.
3784
cache[rev_id] = tree
3785
possible_trees.append((rev_id, tree))
3786
return possible_trees
3788
def _get_delta_for_revision(self, tree, parent_ids, possible_trees):
3453
def _get_delta_for_revision(self, tree, parent_ids, basis_id, cache):
3789
3454
"""Get the best delta and base for this revision.
3791
3456
:return: (basis_id, delta)
3458
possible_trees = [(parent_id, cache[parent_id])
3459
for parent_id in parent_ids
3460
if parent_id in cache]
3461
if len(possible_trees) == 0:
3462
# There either aren't any parents, or the parents aren't in the
3463
# cache, so just use the last converted tree
3464
possible_trees.append((basis_id, cache[basis_id]))
3794
# Generate deltas against each tree, to find the shortest.
3795
texts_possibly_new_in_tree = set()
3796
3466
for basis_id, basis_tree in possible_trees:
3797
3467
delta = tree.inventory._make_delta(basis_tree.inventory)
3798
for old_path, new_path, file_id, new_entry in delta:
3799
if new_path is None:
3800
# This file_id isn't present in the new rev, so we don't
3804
# Rich roots are handled elsewhere...
3806
kind = new_entry.kind
3807
if kind != 'directory' and kind != 'file':
3808
# No text record associated with this inventory entry.
3810
# This is a directory or file that has changed somehow.
3811
texts_possibly_new_in_tree.add((file_id, new_entry.revision))
3812
3468
deltas.append((len(delta), basis_id, delta))
3814
3470
return deltas[0][1:]
3816
def _fetch_parent_invs_for_stacking(self, parent_map, cache):
3817
"""Find all parent revisions that are absent, but for which the
3818
inventory is present, and copy those inventories.
3820
This is necessary to preserve correctness when the source is stacked
3821
without fallbacks configured. (Note that in cases like upgrade the
3822
source may be not have _fallback_repositories even though it is
3826
for parents in parent_map.values():
3827
parent_revs.update(parents)
3828
present_parents = self.source.get_parent_map(parent_revs)
3829
absent_parents = set(parent_revs).difference(present_parents)
3830
parent_invs_keys_for_stacking = self.source.inventories.get_parent_map(
3831
(rev_id,) for rev_id in absent_parents)
3832
parent_inv_ids = [key[-1] for key in parent_invs_keys_for_stacking]
3833
for parent_tree in self.source.revision_trees(parent_inv_ids):
3834
current_revision_id = parent_tree.get_revision_id()
3835
parents_parents_keys = parent_invs_keys_for_stacking[
3836
(current_revision_id,)]
3837
parents_parents = [key[-1] for key in parents_parents_keys]
3838
basis_id = _mod_revision.NULL_REVISION
3839
basis_tree = self.source.revision_tree(basis_id)
3840
delta = parent_tree.inventory._make_delta(basis_tree.inventory)
3841
self.target.add_inventory_by_delta(
3842
basis_id, delta, current_revision_id, parents_parents)
3843
cache[current_revision_id] = parent_tree
3845
3472
def _fetch_batch(self, revision_ids, basis_id, cache):
3846
3473
"""Fetch across a few revisions.
3855
3482
# Walk though all revisions; get inventory deltas, copy referenced
3856
3483
# texts that delta references, insert the delta, revision and
3858
root_keys_to_create = set()
3859
3485
text_keys = set()
3860
3486
pending_deltas = []
3861
3487
pending_revisions = []
3862
3488
parent_map = self.source.get_parent_map(revision_ids)
3863
self._fetch_parent_invs_for_stacking(parent_map, cache)
3864
self.source._safe_to_return_from_cache = True
3865
3489
for tree in self.source.revision_trees(revision_ids):
3866
# Find a inventory delta for this revision.
3867
# Find text entries that need to be copied, too.
3868
3490
current_revision_id = tree.get_revision_id()
3869
3491
parent_ids = parent_map.get(current_revision_id, ())
3870
parent_trees = self._get_trees(parent_ids, cache)
3871
possible_trees = list(parent_trees)
3872
if len(possible_trees) == 0:
3873
# There either aren't any parents, or the parents are ghosts,
3874
# so just use the last converted tree.
3875
possible_trees.append((basis_id, cache[basis_id]))
3876
3492
basis_id, delta = self._get_delta_for_revision(tree, parent_ids,
3494
# Find text entries that need to be copied
3495
for old_path, new_path, file_id, entry in delta:
3496
if new_path is not None:
3497
if not (new_path or self.target.supports_rich_root()):
3498
# We don't copy the text for the root node unless the
3499
# target supports_rich_root.
3501
text_keys.add((file_id, entry.revision))
3878
3502
revision = self.source.get_revision(current_revision_id)
3879
3503
pending_deltas.append((basis_id, delta,
3880
3504
current_revision_id, revision.parent_ids))
3881
if self._converting_to_rich_root:
3882
self._revision_id_to_root_id[current_revision_id] = \
3884
# Determine which texts are in present in this revision but not in
3885
# any of the available parents.
3886
texts_possibly_new_in_tree = set()
3887
for old_path, new_path, file_id, entry in delta:
3888
if new_path is None:
3889
# This file_id isn't present in the new rev
3893
if not self.target.supports_rich_root():
3894
# The target doesn't support rich root, so we don't
3897
if self._converting_to_rich_root:
3898
# This can't be copied normally, we have to insert
3900
root_keys_to_create.add((file_id, entry.revision))
3903
texts_possibly_new_in_tree.add((file_id, entry.revision))
3904
for basis_id, basis_tree in possible_trees:
3905
basis_inv = basis_tree.inventory
3906
for file_key in list(texts_possibly_new_in_tree):
3907
file_id, file_revision = file_key
3909
entry = basis_inv[file_id]
3910
except errors.NoSuchId:
3912
if entry.revision == file_revision:
3913
texts_possibly_new_in_tree.remove(file_key)
3914
text_keys.update(texts_possibly_new_in_tree)
3915
3505
pending_revisions.append(revision)
3916
3506
cache[current_revision_id] = tree
3917
3507
basis_id = current_revision_id
3918
self.source._safe_to_return_from_cache = False
3919
3508
# Copy file texts
3920
3509
from_texts = self.source.texts
3921
3510
to_texts = self.target.texts
3922
if root_keys_to_create:
3923
root_stream = _mod_fetch._new_root_data_stream(
3924
root_keys_to_create, self._revision_id_to_root_id, parent_map,
3926
to_texts.insert_record_stream(root_stream)
3927
3511
to_texts.insert_record_stream(from_texts.get_record_stream(
3928
3512
text_keys, self.target._format._fetch_order,
3929
3513
not self.target._format._fetch_uses_deltas))
3930
# insert inventory deltas
3931
3515
for delta in pending_deltas:
3932
3516
self.target.add_inventory_by_delta(*delta)
3933
if self.target._fallback_repositories:
3934
# Make sure this stacked repository has all the parent inventories
3935
# for the new revisions that we are about to insert. We do this
3936
# before adding the revisions so that no revision is added until
3937
# all the inventories it may depend on are added.
3938
# Note that this is overzealous, as we may have fetched these in an
3941
revision_ids = set()
3942
for revision in pending_revisions:
3943
revision_ids.add(revision.revision_id)
3944
parent_ids.update(revision.parent_ids)
3945
parent_ids.difference_update(revision_ids)
3946
parent_ids.discard(_mod_revision.NULL_REVISION)
3947
parent_map = self.source.get_parent_map(parent_ids)
3948
# we iterate over parent_map and not parent_ids because we don't
3949
# want to try copying any revision which is a ghost
3950
for parent_tree in self.source.revision_trees(parent_map):
3951
current_revision_id = parent_tree.get_revision_id()
3952
parents_parents = parent_map[current_revision_id]
3953
possible_trees = self._get_trees(parents_parents, cache)
3954
if len(possible_trees) == 0:
3955
# There either aren't any parents, or the parents are
3956
# ghosts, so just use the last converted tree.
3957
possible_trees.append((basis_id, cache[basis_id]))
3958
basis_id, delta = self._get_delta_for_revision(parent_tree,
3959
parents_parents, possible_trees)
3960
self.target.add_inventory_by_delta(
3961
basis_id, delta, current_revision_id, parents_parents)
3962
3517
# insert signatures and revisions
3963
3518
for revision in pending_revisions:
4013
3560
"""See InterRepository.fetch()."""
4014
3561
if fetch_spec is not None:
4015
3562
raise AssertionError("Not implemented yet...")
4016
ui.ui_factory.warn_experimental_format_fetch(self)
4017
if (not self.source.supports_rich_root()
4018
and self.target.supports_rich_root()):
4019
self._converting_to_rich_root = True
4020
self._revision_id_to_root_id = {}
4022
self._converting_to_rich_root = False
4023
# See <https://launchpad.net/bugs/456077> asking for a warning here
4024
if self.source._format.network_name() != self.target._format.network_name():
4025
ui.ui_factory.show_user_warning('cross_format_fetch',
4026
from_format=self.source._format,
4027
to_format=self.target._format)
4028
3563
revision_ids = self.target.search_missing_revision_ids(self.source,
4029
3564
revision_id, find_ghosts=find_ghosts).get_keys()
4030
3565
if not revision_ids:
4032
3567
revision_ids = tsort.topo_sort(
4033
3568
self.source.get_graph().get_parent_map(revision_ids))
4034
if not revision_ids:
4036
# Walk though all revisions; get inventory deltas, copy referenced
4037
# texts that delta references, insert the delta, revision and
4040
3570
my_pb = ui.ui_factory.nested_progress_bar()
4099
3630
:param to_convert: The disk object to convert.
4100
3631
:param pb: a progress bar to use for progress information.
4102
pb = ui.ui_factory.nested_progress_bar()
4105
3636
# this is only useful with metadir layouts - separated repo content.
4106
3637
# trigger an assertion if not such
4107
3638
repo._format.get_format_string()
4108
3639
self.repo_dir = repo.bzrdir
4109
pb.update('Moving repository to repository.backup')
3640
self.step('Moving repository to repository.backup')
4110
3641
self.repo_dir.transport.move('repository', 'repository.backup')
4111
3642
backup_transport = self.repo_dir.transport.clone('repository.backup')
4112
3643
repo._format.check_conversion_target(self.target_format)
4113
3644
self.source_repo = repo._format.open(self.repo_dir,
4115
3646
_override_transport=backup_transport)
4116
pb.update('Creating new repository')
3647
self.step('Creating new repository')
4117
3648
converted = self.target_format.initialize(self.repo_dir,
4118
3649
self.source_repo.is_shared())
4119
3650
converted.lock_write()
4121
pb.update('Copying content')
3652
self.step('Copying content into repository.')
4122
3653
self.source_repo.copy_content_into(converted)
4124
3655
converted.unlock()
4125
pb.update('Deleting old repository content')
3656
self.step('Deleting old repository content.')
4126
3657
self.repo_dir.transport.delete_tree('repository.backup')
4127
ui.ui_factory.note('repository converted')
3658
self.pb.note('repository converted')
3660
def step(self, message):
3661
"""Update the pb by a step."""
3663
self.pb.update(message, self.count, self.total)
4131
3666
_unescape_map = {
4333
3848
self.target_repo.signatures.insert_record_stream(substream)
4335
3850
raise AssertionError('kaboom! %s' % (substream_type,))
4336
# Done inserting data, and the missing_keys calculations will try to
4337
# read back from the inserted data, so flush the writes to the new pack
4338
# (if this is pack format).
4339
if new_pack is not None:
4340
new_pack._write_data('', flush=True)
4341
# Find all the new revisions (including ones from resume_tokens)
4342
missing_keys = self.target_repo.get_missing_parent_inventories(
4343
check_for_missing_texts=is_resume)
3852
missing_keys = set()
4345
3853
for prefix, versioned_file in (
4346
3854
('texts', self.target_repo.texts),
4347
3855
('inventories', self.target_repo.inventories),
4348
3856
('revisions', self.target_repo.revisions),
4349
3857
('signatures', self.target_repo.signatures),
4350
('chk_bytes', self.target_repo.chk_bytes),
4352
if versioned_file is None:
4354
# TODO: key is often going to be a StaticTuple object
4355
# I don't believe we can define a method by which
4356
# (prefix,) + StaticTuple will work, though we could
4357
# define a StaticTuple.sq_concat that would allow you to
4358
# pass in either a tuple or a StaticTuple as the second
4359
# object, so instead we could have:
4360
# StaticTuple(prefix) + key here...
4361
3859
missing_keys.update((prefix,) + key for key in
4362
3860
versioned_file.get_missing_compression_parent_keys())
4363
3861
except NotImplementedError:
4372
3870
# missing keys can handle suspending a write group).
4373
3871
write_group_tokens = self.target_repo.suspend_write_group()
4374
3872
return write_group_tokens, missing_keys
4375
hint = self.target_repo.commit_write_group()
4376
if (to_serializer != src_serializer and
4377
self.target_repo._format.pack_compresses):
4378
self.target_repo.pack(hint=hint)
3873
self.target_repo.commit_write_group()
4379
3874
return [], set()
4381
def _extract_and_insert_inventory_deltas(self, substream, serializer):
4382
target_rich_root = self.target_repo._format.rich_root_data
4383
target_tree_refs = self.target_repo._format.supports_tree_reference
4384
for record in substream:
4385
# Insert the delta directly
4386
inventory_delta_bytes = record.get_bytes_as('fulltext')
4387
deserialiser = inventory_delta.InventoryDeltaDeserializer()
4389
parse_result = deserialiser.parse_text_bytes(
4390
inventory_delta_bytes)
4391
except inventory_delta.IncompatibleInventoryDelta, err:
4392
trace.mutter("Incompatible delta: %s", err.msg)
4393
raise errors.IncompatibleRevision(self.target_repo._format)
4394
basis_id, new_id, rich_root, tree_refs, inv_delta = parse_result
4395
revision_id = new_id
4396
parents = [key[0] for key in record.parents]
4397
self.target_repo.add_inventory_by_delta(
4398
basis_id, inv_delta, revision_id, parents)
4400
def _extract_and_insert_inventories(self, substream, serializer,
3876
def _extract_and_insert_inventories(self, substream, serializer):
4402
3877
"""Generate a new inventory versionedfile in target, converting data.
4404
3879
The inventory is retrieved from the source, (deserializing it), and
4405
3880
stored in the target (reserializing it in a different format).
4407
target_rich_root = self.target_repo._format.rich_root_data
4408
target_tree_refs = self.target_repo._format.supports_tree_reference
4409
3882
for record in substream:
4410
# It's not a delta, so it must be a fulltext in the source
4411
# serializer's format.
4412
3883
bytes = record.get_bytes_as('fulltext')
4413
3884
revision_id = record.key[0]
4414
3885
inv = serializer.read_inventory_from_string(bytes, revision_id)
4415
3886
parents = [key[0] for key in record.parents]
4416
3887
self.target_repo.add_inventory(revision_id, inv, parents)
4417
# No need to keep holding this full inv in memory when the rest of
4418
# the substream is likely to be all deltas.
4421
3889
def _extract_and_insert_revisions(self, substream, serializer):
4422
3890
for record in substream:
4547
4022
keys['revisions'],))
4548
4023
for substream_kind, keys in keys.iteritems():
4549
4024
vf = getattr(self.from_repository, substream_kind)
4550
if vf is None and keys:
4551
raise AssertionError(
4552
"cannot fill in keys for a versioned file we don't"
4553
" have: %s needs %s" % (substream_kind, keys))
4555
# No need to stream something we don't have
4557
if substream_kind == 'inventories':
4558
# Some missing keys are genuinely ghosts, filter those out.
4559
present = self.from_repository.inventories.get_parent_map(keys)
4560
revs = [key[0] for key in present]
4561
# Get the inventory stream more-or-less as we do for the
4562
# original stream; there's no reason to assume that records
4563
# direct from the source will be suitable for the sink. (Think
4564
# e.g. 2a -> 1.9-rich-root).
4565
for info in self._get_inventory_stream(revs, missing=True):
4569
4025
# Ask for full texts always so that we don't need more round trips
4570
4026
# after this stream.
4571
# Some of the missing keys are genuinely ghosts, so filter absent
4572
# records. The Sink is responsible for doing another check to
4573
# ensure that ghosts don't introduce missing data for future
4575
stream = versionedfile.filter_absent(vf.get_record_stream(keys,
4576
self.to_format._fetch_order, True))
4027
stream = vf.get_record_stream(keys,
4028
self.to_format._fetch_order, True)
4577
4029
yield substream_kind, stream
4579
4031
def inventory_fetch_order(self):
4586
4038
return (not self.from_repository._format.rich_root_data and
4587
4039
self.to_format.rich_root_data)
4589
def _get_inventory_stream(self, revision_ids, missing=False):
4590
from_format = self.from_repository._format
4591
if (from_format.supports_chks and self.to_format.supports_chks and
4592
from_format.network_name() == self.to_format.network_name()):
4593
raise AssertionError(
4594
"this case should be handled by GroupCHKStreamSource")
4595
elif 'forceinvdeltas' in debug.debug_flags:
4596
return self._get_convertable_inventory_stream(revision_ids,
4597
delta_versus_null=missing)
4598
elif from_format.network_name() == self.to_format.network_name():
4600
return self._get_simple_inventory_stream(revision_ids,
4602
elif (not from_format.supports_chks and not self.to_format.supports_chks
4603
and from_format._serializer == self.to_format._serializer):
4604
# Essentially the same format.
4605
return self._get_simple_inventory_stream(revision_ids,
4608
# Any time we switch serializations, we want to use an
4609
# inventory-delta based approach.
4610
return self._get_convertable_inventory_stream(revision_ids,
4611
delta_versus_null=missing)
4613
def _get_simple_inventory_stream(self, revision_ids, missing=False):
4614
# NB: This currently reopens the inventory weave in source;
4615
# using a single stream interface instead would avoid this.
4616
from_weave = self.from_repository.inventories
4618
delta_closure = True
4620
delta_closure = not self.delta_on_metadata()
4621
yield ('inventories', from_weave.get_record_stream(
4622
[(rev_id,) for rev_id in revision_ids],
4623
self.inventory_fetch_order(), delta_closure))
4625
def _get_convertable_inventory_stream(self, revision_ids,
4626
delta_versus_null=False):
4627
# The two formats are sufficiently different that there is no fast
4628
# path, so we need to send just inventorydeltas, which any
4629
# sufficiently modern client can insert into any repository.
4630
# The StreamSink code expects to be able to
4631
# convert on the target, so we need to put bytes-on-the-wire that can
4632
# be converted. That means inventory deltas (if the remote is <1.19,
4633
# RemoteStreamSink will fallback to VFS to insert the deltas).
4634
yield ('inventory-deltas',
4635
self._stream_invs_as_deltas(revision_ids,
4636
delta_versus_null=delta_versus_null))
4638
def _stream_invs_as_deltas(self, revision_ids, delta_versus_null=False):
4639
"""Return a stream of inventory-deltas for the given rev ids.
4641
:param revision_ids: The list of inventories to transmit
4642
:param delta_versus_null: Don't try to find a minimal delta for this
4643
entry, instead compute the delta versus the NULL_REVISION. This
4644
effectively streams a complete inventory. Used for stuff like
4645
filling in missing parents, etc.
4647
from_repo = self.from_repository
4648
revision_keys = [(rev_id,) for rev_id in revision_ids]
4649
parent_map = from_repo.inventories.get_parent_map(revision_keys)
4650
# XXX: possibly repos could implement a more efficient iter_inv_deltas
4652
inventories = self.from_repository.iter_inventories(
4653
revision_ids, 'topological')
4654
format = from_repo._format
4655
invs_sent_so_far = set([_mod_revision.NULL_REVISION])
4656
inventory_cache = lru_cache.LRUCache(50)
4657
null_inventory = from_repo.revision_tree(
4658
_mod_revision.NULL_REVISION).inventory
4659
# XXX: ideally the rich-root/tree-refs flags would be per-revision, not
4660
# per-repo (e.g. streaming a non-rich-root revision out of a rich-root
4661
# repo back into a non-rich-root repo ought to be allowed)
4662
serializer = inventory_delta.InventoryDeltaSerializer(
4663
versioned_root=format.rich_root_data,
4664
tree_references=format.supports_tree_reference)
4665
for inv in inventories:
4666
key = (inv.revision_id,)
4667
parent_keys = parent_map.get(key, ())
4669
if not delta_versus_null and parent_keys:
4670
# The caller did not ask for complete inventories and we have
4671
# some parents that we can delta against. Make a delta against
4672
# each parent so that we can find the smallest.
4673
parent_ids = [parent_key[0] for parent_key in parent_keys]
4674
for parent_id in parent_ids:
4675
if parent_id not in invs_sent_so_far:
4676
# We don't know that the remote side has this basis, so
4679
if parent_id == _mod_revision.NULL_REVISION:
4680
parent_inv = null_inventory
4682
parent_inv = inventory_cache.get(parent_id, None)
4683
if parent_inv is None:
4684
parent_inv = from_repo.get_inventory(parent_id)
4685
candidate_delta = inv._make_delta(parent_inv)
4686
if (delta is None or
4687
len(delta) > len(candidate_delta)):
4688
delta = candidate_delta
4689
basis_id = parent_id
4691
# Either none of the parents ended up being suitable, or we
4692
# were asked to delta against NULL
4693
basis_id = _mod_revision.NULL_REVISION
4694
delta = inv._make_delta(null_inventory)
4695
invs_sent_so_far.add(inv.revision_id)
4696
inventory_cache[inv.revision_id] = inv
4697
delta_serialized = ''.join(
4698
serializer.delta_to_lines(basis_id, key[-1], delta))
4699
yield versionedfile.FulltextContentFactory(
4700
key, parent_keys, None, delta_serialized)
4703
def _iter_for_revno(repo, partial_history_cache, stop_index=None,
4704
stop_revision=None):
4705
"""Extend the partial history to include a given index
4707
If a stop_index is supplied, stop when that index has been reached.
4708
If a stop_revision is supplied, stop when that revision is
4709
encountered. Otherwise, stop when the beginning of history is
4712
:param stop_index: The index which should be present. When it is
4713
present, history extension will stop.
4714
:param stop_revision: The revision id which should be present. When
4715
it is encountered, history extension will stop.
4717
start_revision = partial_history_cache[-1]
4718
iterator = repo.iter_reverse_revision_history(start_revision)
4720
#skip the last revision in the list
4723
if (stop_index is not None and
4724
len(partial_history_cache) > stop_index):
4726
if partial_history_cache[-1] == stop_revision:
4728
revision_id = iterator.next()
4729
partial_history_cache.append(revision_id)
4730
except StopIteration:
added
removed
bzrlib/repository.py
