709
711
return compatible
712
class Inventory(object):
713
"""Inventory of versioned files in a tree.
715
This describes which file_id is present at each point in the tree,
716
and possibly the SHA-1 or other information about the file.
714
class CommonInventory(object):
715
"""Basic inventory logic, defined in terms of primitives like has_id.
717
An inventory is the metadata about the contents of a tree.
719
This is broadly a map from file_id to entries such as directories, files,
720
symlinks and tree references. Each entry maintains its own metadata like
721
SHA1 and length for files, or children for a directory.
717
723
Entries can be looked up either by path or by file_id.
719
The inventory represents a typical unix file tree, with
720
directories containing files and subdirectories. We never store
721
the full path to a file, because renaming a directory implicitly
722
moves all of its contents. This class internally maintains a
723
lookup tree that allows the children under a directory to be
726
725
InventoryEntry objects must not be modified after they are
727
726
inserted, other than through the Inventory API.
729
>>> inv = Inventory()
730
>>> inv.add(InventoryFile('123-123', 'hello.c', ROOT_ID))
731
InventoryFile('123-123', 'hello.c', parent_id='TREE_ROOT', sha1=None, len=None)
732
>>> inv['123-123'].name
735
May be treated as an iterator or set to look up file ids:
737
>>> bool(inv.path2id('hello.c'))
742
May also look up by name:
744
>>> [x[0] for x in inv.iter_entries()]
746
>>> inv = Inventory('TREE_ROOT-12345678-12345678')
747
>>> inv.add(InventoryFile('123-123', 'hello.c', ROOT_ID))
748
Traceback (most recent call last):
749
BzrError: parent_id {TREE_ROOT} not in inventory
750
>>> inv.add(InventoryFile('123-123', 'hello.c', 'TREE_ROOT-12345678-12345678'))
751
InventoryFile('123-123', 'hello.c', parent_id='TREE_ROOT-12345678-12345678', sha1=None, len=None)
753
def __init__(self, root_id=ROOT_ID, revision_id=None):
754
"""Create or read an inventory.
756
If a working directory is specified, the inventory is read
757
from there. If the file is specified, read from that. If not,
758
the inventory is created empty.
760
The inventory is created with a default root directory, with
763
if root_id is not None:
764
self._set_root(InventoryDirectory(root_id, u'', None))
768
self.revision_id = revision_id
771
return "<Inventory object at %x, contents=%r>" % (id(self), self._byid)
773
def apply_delta(self, delta):
774
"""Apply a delta to this inventory.
776
:param delta: A list of changes to apply. After all the changes are
777
applied the final inventory must be internally consistent, but it
778
is ok to supply changes which, if only half-applied would have an
779
invalid result - such as supplying two changes which rename two
780
files, 'A' and 'B' with each other : [('A', 'B', 'A-id', a_entry),
781
('B', 'A', 'B-id', b_entry)].
783
Each change is a tuple, of the form (old_path, new_path, file_id,
786
When new_path is None, the change indicates the removal of an entry
787
from the inventory and new_entry will be ignored (using None is
788
appropriate). If new_path is not None, then new_entry must be an
789
InventoryEntry instance, which will be incorporated into the
790
inventory (and replace any existing entry with the same file id).
792
When old_path is None, the change indicates the addition of
793
a new entry to the inventory.
795
When neither new_path nor old_path are None, the change is a
796
modification to an entry, such as a rename, reparent, kind change
799
The children attribute of new_entry is ignored. This is because
800
this method preserves children automatically across alterations to
801
the parent of the children, and cases where the parent id of a
802
child is changing require the child to be passed in as a separate
803
change regardless. E.g. in the recursive deletion of a directory -
804
the directory's children must be included in the delta, or the
805
final inventory will be invalid.
808
# Remove all affected items which were in the original inventory,
809
# starting with the longest paths, thus ensuring parents are examined
810
# after their children, which means that everything we examine has no
811
# modified children remaining by the time we examine it.
812
for old_path, file_id in sorted(((op, f) for op, np, f, e in delta
813
if op is not None), reverse=True):
814
if file_id not in self:
817
# Preserve unaltered children of file_id for later reinsertion.
818
file_id_children = getattr(self[file_id], 'children', {})
819
if len(file_id_children):
820
children[file_id] = file_id_children
821
# Remove file_id and the unaltered children. If file_id is not
822
# being deleted it will be reinserted back later.
823
self.remove_recursive_id(file_id)
824
# Insert all affected which should be in the new inventory, reattaching
825
# their children if they had any. This is done from shortest path to
826
# longest, ensuring that items which were modified and whose parents in
827
# the resulting inventory were also modified, are inserted after their
829
for new_path, new_entry in sorted((np, e) for op, np, f, e in
830
delta if np is not None):
831
if new_entry.kind == 'directory':
832
# Pop the child which to allow detection of children whose
833
# parents were deleted and which were not reattached to a new
835
new_entry.children = children.pop(new_entry.file_id, {})
838
# Get the parent id that was deleted
839
parent_id, children = children.popitem()
840
raise errors.InconsistentDelta("<deleted>", parent_id,
841
"The file id was deleted but its children were not deleted.")
843
def _set_root(self, ie):
845
self._byid = {self.root.file_id: self.root}
848
# TODO: jam 20051218 Should copy also copy the revision_id?
849
entries = self.iter_entries()
850
if self.root is None:
851
return Inventory(root_id=None)
852
other = Inventory(entries.next()[1].file_id)
853
other.root.revision = self.root.revision
854
# copy recursively so we know directories will be added before
855
# their children. There are more efficient ways than this...
856
for path, entry in entries:
857
other.add(entry.copy())
861
return iter(self._byid)
864
"""Returns number of entries."""
865
return len(self._byid)
867
def iter_entries(self, from_dir=None):
868
"""Return (path, entry) pairs, in order by name."""
729
def __contains__(self, file_id):
730
"""True if this entry contains a file with given id.
732
>>> inv = Inventory()
733
>>> inv.add(InventoryFile('123', 'foo.c', ROOT_ID))
734
InventoryFile('123', 'foo.c', parent_id='TREE_ROOT', sha1=None, len=None, revision=None)
740
Note that this method along with __iter__ are not encouraged for use as
741
they are less clear than specific query methods - they may be rmeoved
744
return self.has_id(file_id)
746
def has_filename(self, filename):
747
return bool(self.path2id(filename))
749
def id2path(self, file_id):
750
"""Return as a string the path to file_id.
753
>>> e = i.add(InventoryDirectory('src-id', 'src', ROOT_ID))
754
>>> e = i.add(InventoryFile('foo-id', 'foo.c', parent_id='src-id'))
755
>>> print i.id2path('foo-id')
758
# get all names, skipping root
759
return '/'.join(reversed(
760
[parent.name for parent in
761
self._iter_file_id_parents(file_id)][:-1]))
763
def iter_entries(self, from_dir=None, recursive=True):
764
"""Return (path, entry) pairs, in order by name.
766
:param from_dir: if None, start from the root,
767
otherwise start from this directory (either file-id or entry)
768
:param recursive: recurse into directories or not
869
770
if from_dir is None:
870
771
if self.root is None:
872
773
from_dir = self.root
873
774
yield '', self.root
874
775
elif isinstance(from_dir, basestring):
875
from_dir = self._byid[from_dir]
776
from_dir = self[from_dir]
877
778
# unrolling the recursive called changed the time from
878
779
# 440ms/663ms (inline/total) to 116ms/116ms
879
780
children = from_dir.children.items()
783
for name, ie in children:
881
786
children = collections.deque(children)
882
787
stack = [(u'', children)]
1018
949
descend(child_ie, child_path)
1019
950
descend(self.root, u'')
953
def path2id(self, name):
954
"""Walk down through directories to return entry of last component.
956
names may be either a list of path components, or a single
957
string, in which case it is automatically split.
959
This returns the entry of the last component in the path,
960
which may be either a file or a directory.
962
Returns None IFF the path is not found.
964
if isinstance(name, basestring):
965
name = osutils.splitpath(name)
967
# mutter("lookup path %r" % name)
971
except errors.NoSuchId:
972
# root doesn't exist yet so nothing else can
978
children = getattr(parent, 'children', None)
987
return parent.file_id
989
def filter(self, specific_fileids):
990
"""Get an inventory view filtered against a set of file-ids.
992
Children of directories and parents are included.
994
The result may or may not reference the underlying inventory
995
so it should be treated as immutable.
997
interesting_parents = set()
998
for fileid in specific_fileids:
1000
interesting_parents.update(self.get_idpath(fileid))
1001
except errors.NoSuchId:
1002
# This fileid is not in the inventory - that's ok
1004
entries = self.iter_entries()
1005
if self.root is None:
1006
return Inventory(root_id=None)
1007
other = Inventory(entries.next()[1].file_id)
1008
other.root.revision = self.root.revision
1009
other.revision_id = self.revision_id
1010
directories_to_expand = set()
1011
for path, entry in entries:
1012
file_id = entry.file_id
1013
if (file_id in specific_fileids
1014
or entry.parent_id in directories_to_expand):
1015
if entry.kind == 'directory':
1016
directories_to_expand.add(file_id)
1017
elif file_id not in interesting_parents:
1019
other.add(entry.copy())
1022
def get_idpath(self, file_id):
1023
"""Return a list of file_ids for the path to an entry.
1025
The list contains one element for each directory followed by
1026
the id of the file itself. So the length of the returned list
1027
is equal to the depth of the file in the tree, counting the
1028
root directory as depth 1.
1031
for parent in self._iter_file_id_parents(file_id):
1032
p.insert(0, parent.file_id)
1036
class Inventory(CommonInventory):
1037
"""Mutable dict based in-memory inventory.
1039
We never store the full path to a file, because renaming a directory
1040
implicitly moves all of its contents. This class internally maintains a
1041
lookup tree that allows the children under a directory to be
1044
>>> inv = Inventory()
1045
>>> inv.add(InventoryFile('123-123', 'hello.c', ROOT_ID))
1046
InventoryFile('123-123', 'hello.c', parent_id='TREE_ROOT', sha1=None, len=None, revision=None)
1047
>>> inv['123-123'].name
1050
Id's may be looked up from paths:
1052
>>> inv.path2id('hello.c')
1054
>>> '123-123' in inv
1057
There are iterators over the contents:
1059
>>> [entry[0] for entry in inv.iter_entries()]
1063
def __init__(self, root_id=ROOT_ID, revision_id=None):
1064
"""Create or read an inventory.
1066
If a working directory is specified, the inventory is read
1067
from there. If the file is specified, read from that. If not,
1068
the inventory is created empty.
1070
The inventory is created with a default root directory, with
1073
if root_id is not None:
1074
self._set_root(InventoryDirectory(root_id, u'', None))
1078
self.revision_id = revision_id
1081
# More than one page of ouput is not useful anymore to debug
1084
contents = repr(self._byid)
1085
if len(contents) > max_len:
1086
contents = contents[:(max_len-len(closing))] + closing
1087
return "<Inventory object at %x, contents=%r>" % (id(self), contents)
1089
def apply_delta(self, delta):
1090
"""Apply a delta to this inventory.
1092
See the inventory developers documentation for the theory behind
1095
If delta application fails the inventory is left in an indeterminate
1096
state and must not be used.
1098
:param delta: A list of changes to apply. After all the changes are
1099
applied the final inventory must be internally consistent, but it
1100
is ok to supply changes which, if only half-applied would have an
1101
invalid result - such as supplying two changes which rename two
1102
files, 'A' and 'B' with each other : [('A', 'B', 'A-id', a_entry),
1103
('B', 'A', 'B-id', b_entry)].
1105
Each change is a tuple, of the form (old_path, new_path, file_id,
1108
When new_path is None, the change indicates the removal of an entry
1109
from the inventory and new_entry will be ignored (using None is
1110
appropriate). If new_path is not None, then new_entry must be an
1111
InventoryEntry instance, which will be incorporated into the
1112
inventory (and replace any existing entry with the same file id).
1114
When old_path is None, the change indicates the addition of
1115
a new entry to the inventory.
1117
When neither new_path nor old_path are None, the change is a
1118
modification to an entry, such as a rename, reparent, kind change
1121
The children attribute of new_entry is ignored. This is because
1122
this method preserves children automatically across alterations to
1123
the parent of the children, and cases where the parent id of a
1124
child is changing require the child to be passed in as a separate
1125
change regardless. E.g. in the recursive deletion of a directory -
1126
the directory's children must be included in the delta, or the
1127
final inventory will be invalid.
1129
Note that a file_id must only appear once within a given delta.
1130
An AssertionError is raised otherwise.
1132
# Check that the delta is legal. It would be nice if this could be
1133
# done within the loops below but it's safer to validate the delta
1134
# before starting to mutate the inventory, as there isn't a rollback
1136
list(_check_delta_unique_ids(_check_delta_unique_new_paths(
1137
_check_delta_unique_old_paths(_check_delta_ids_match_entry(
1138
_check_delta_ids_are_valid(
1139
_check_delta_new_path_entry_both_or_None(
1143
# Remove all affected items which were in the original inventory,
1144
# starting with the longest paths, thus ensuring parents are examined
1145
# after their children, which means that everything we examine has no
1146
# modified children remaining by the time we examine it.
1147
for old_path, file_id in sorted(((op, f) for op, np, f, e in delta
1148
if op is not None), reverse=True):
1149
# Preserve unaltered children of file_id for later reinsertion.
1150
file_id_children = getattr(self[file_id], 'children', {})
1151
if len(file_id_children):
1152
children[file_id] = file_id_children
1153
if self.id2path(file_id) != old_path:
1154
raise errors.InconsistentDelta(old_path, file_id,
1155
"Entry was at wrong other path %r." % self.id2path(file_id))
1156
# Remove file_id and the unaltered children. If file_id is not
1157
# being deleted it will be reinserted back later.
1158
self.remove_recursive_id(file_id)
1159
# Insert all affected which should be in the new inventory, reattaching
1160
# their children if they had any. This is done from shortest path to
1161
# longest, ensuring that items which were modified and whose parents in
1162
# the resulting inventory were also modified, are inserted after their
1164
for new_path, f, new_entry in sorted((np, f, e) for op, np, f, e in
1165
delta if np is not None):
1166
if new_entry.kind == 'directory':
1167
# Pop the child which to allow detection of children whose
1168
# parents were deleted and which were not reattached to a new
1170
replacement = InventoryDirectory(new_entry.file_id,
1171
new_entry.name, new_entry.parent_id)
1172
replacement.revision = new_entry.revision
1173
replacement.children = children.pop(replacement.file_id, {})
1174
new_entry = replacement
1177
except errors.DuplicateFileId:
1178
raise errors.InconsistentDelta(new_path, new_entry.file_id,
1179
"New id is already present in target.")
1180
except AttributeError:
1181
raise errors.InconsistentDelta(new_path, new_entry.file_id,
1182
"Parent is not a directory.")
1183
if self.id2path(new_entry.file_id) != new_path:
1184
raise errors.InconsistentDelta(new_path, new_entry.file_id,
1185
"New path is not consistent with parent path.")
1187
# Get the parent id that was deleted
1188
parent_id, children = children.popitem()
1189
raise errors.InconsistentDelta("<deleted>", parent_id,
1190
"The file id was deleted but its children were not deleted.")
1192
def _set_root(self, ie):
1194
self._byid = {self.root.file_id: self.root}
1197
# TODO: jam 20051218 Should copy also copy the revision_id?
1198
entries = self.iter_entries()
1199
if self.root is None:
1200
return Inventory(root_id=None)
1201
other = Inventory(entries.next()[1].file_id)
1202
other.root.revision = self.root.revision
1203
# copy recursively so we know directories will be added before
1204
# their children. There are more efficient ways than this...
1205
for path, entry in entries:
1206
other.add(entry.copy())
1209
def _get_mutable_inventory(self):
1210
"""See CommonInventory._get_mutable_inventory."""
1211
return copy.deepcopy(self)
1214
"""Iterate over all file-ids."""
1215
return iter(self._byid)
1217
def iter_just_entries(self):
1218
"""Iterate over all entries.
1022
def __contains__(self, file_id):
1023
"""True if this entry contains a file with given id.
1220
Unlike iter_entries(), just the entries are returned (not (path, ie))
1221
and the order of entries is undefined.
1025
>>> inv = Inventory()
1026
>>> inv.add(InventoryFile('123', 'foo.c', ROOT_ID))
1027
InventoryFile('123', 'foo.c', parent_id='TREE_ROOT', sha1=None, len=None)
1223
XXX: We may not want to merge this into bzr.dev.
1033
return (file_id in self._byid)
1225
if self.root is None:
1227
for _, ie in self._byid.iteritems():
1231
"""Returns number of entries."""
1232
return len(self._byid)
1035
1234
def __getitem__(self, file_id):
1036
1235
"""Return the entry for given file_id.
1038
1237
>>> inv = Inventory()
1039
1238
>>> inv.add(InventoryFile('123123', 'hello.c', ROOT_ID))
1040
InventoryFile('123123', 'hello.c', parent_id='TREE_ROOT', sha1=None, len=None)
1239
InventoryFile('123123', 'hello.c', parent_id='TREE_ROOT', sha1=None, len=None, revision=None)
1041
1240
>>> inv['123123'].name
1288
1455
return self.root is not None and file_id == self.root.file_id
1458
class CHKInventory(CommonInventory):
1459
"""An inventory persisted in a CHK store.
1461
By design, a CHKInventory is immutable so many of the methods
1462
supported by Inventory - add, rename, apply_delta, etc - are *not*
1463
supported. To create a new CHKInventory, use create_by_apply_delta()
1464
or from_inventory(), say.
1466
Internally, a CHKInventory has one or two CHKMaps:
1468
* id_to_entry - a map from (file_id,) => InventoryEntry as bytes
1469
* parent_id_basename_to_file_id - a map from (parent_id, basename_utf8)
1472
The second map is optional and not present in early CHkRepository's.
1474
No caching is performed: every method call or item access will perform
1475
requests to the storage layer. As such, keep references to objects you
1479
def __init__(self, search_key_name):
1480
CommonInventory.__init__(self)
1481
self._fileid_to_entry_cache = {}
1482
self._path_to_fileid_cache = {}
1483
self._search_key_name = search_key_name
1486
def __eq__(self, other):
1487
"""Compare two sets by comparing their contents."""
1488
if not isinstance(other, CHKInventory):
1489
return NotImplemented
1491
this_key = self.id_to_entry.key()
1492
other_key = other.id_to_entry.key()
1493
this_pid_key = self.parent_id_basename_to_file_id.key()
1494
other_pid_key = other.parent_id_basename_to_file_id.key()
1495
if None in (this_key, this_pid_key, other_key, other_pid_key):
1497
return this_key == other_key and this_pid_key == other_pid_key
1499
def _entry_to_bytes(self, entry):
1500
"""Serialise entry as a single bytestring.
1502
:param Entry: An inventory entry.
1503
:return: A bytestring for the entry.
1506
ENTRY ::= FILE | DIR | SYMLINK | TREE
1507
FILE ::= "file: " COMMON SEP SHA SEP SIZE SEP EXECUTABLE
1508
DIR ::= "dir: " COMMON
1509
SYMLINK ::= "symlink: " COMMON SEP TARGET_UTF8
1510
TREE ::= "tree: " COMMON REFERENCE_REVISION
1511
COMMON ::= FILE_ID SEP PARENT_ID SEP NAME_UTF8 SEP REVISION
1514
if entry.parent_id is not None:
1515
parent_str = entry.parent_id
1518
name_str = entry.name.encode("utf8")
1519
if entry.kind == 'file':
1520
if entry.executable:
1524
return "file: %s\n%s\n%s\n%s\n%s\n%d\n%s" % (
1525
entry.file_id, parent_str, name_str, entry.revision,
1526
entry.text_sha1, entry.text_size, exec_str)
1527
elif entry.kind == 'directory':
1528
return "dir: %s\n%s\n%s\n%s" % (
1529
entry.file_id, parent_str, name_str, entry.revision)
1530
elif entry.kind == 'symlink':
1531
return "symlink: %s\n%s\n%s\n%s\n%s" % (
1532
entry.file_id, parent_str, name_str, entry.revision,
1533
entry.symlink_target.encode("utf8"))
1534
elif entry.kind == 'tree-reference':
1535
return "tree: %s\n%s\n%s\n%s\n%s" % (
1536
entry.file_id, parent_str, name_str, entry.revision,
1537
entry.reference_revision)
1539
raise ValueError("unknown kind %r" % entry.kind)
1542
def _bytes_to_utf8name_key(bytes):
1543
"""Get the file_id, revision_id key out of bytes."""
1544
# We don't normally care about name, except for times when we want
1545
# to filter out empty names because of non rich-root...
1546
sections = bytes.split('\n')
1547
kind, file_id = sections[0].split(': ')
1548
return (sections[2], file_id, sections[3])
1550
def _bytes_to_entry(self, bytes):
1551
"""Deserialise a serialised entry."""
1552
sections = bytes.split('\n')
1553
if sections[0].startswith("file: "):
1554
result = InventoryFile(sections[0][6:],
1555
sections[2].decode('utf8'),
1557
result.text_sha1 = sections[4]
1558
result.text_size = int(sections[5])
1559
result.executable = sections[6] == "Y"
1560
elif sections[0].startswith("dir: "):
1561
result = CHKInventoryDirectory(sections[0][5:],
1562
sections[2].decode('utf8'),
1564
elif sections[0].startswith("symlink: "):
1565
result = InventoryLink(sections[0][9:],
1566
sections[2].decode('utf8'),
1568
result.symlink_target = sections[4].decode('utf8')
1569
elif sections[0].startswith("tree: "):
1570
result = TreeReference(sections[0][6:],
1571
sections[2].decode('utf8'),
1573
result.reference_revision = sections[4]
1575
raise ValueError("Not a serialised entry %r" % bytes)
1576
result.revision = sections[3]
1577
if result.parent_id == '':
1578
result.parent_id = None
1579
self._fileid_to_entry_cache[result.file_id] = result
1582
def _get_mutable_inventory(self):
1583
"""See CommonInventory._get_mutable_inventory."""
1584
entries = self.iter_entries()
1585
inv = Inventory(None, self.revision_id)
1586
for path, inv_entry in entries:
1587
inv.add(inv_entry.copy())
1590
def create_by_apply_delta(self, inventory_delta, new_revision_id,
1591
propagate_caches=False):
1592
"""Create a new CHKInventory by applying inventory_delta to this one.
1594
See the inventory developers documentation for the theory behind
1597
:param inventory_delta: The inventory delta to apply. See
1598
Inventory.apply_delta for details.
1599
:param new_revision_id: The revision id of the resulting CHKInventory.
1600
:param propagate_caches: If True, the caches for this inventory are
1601
copied to and updated for the result.
1602
:return: The new CHKInventory.
1604
split = osutils.split
1605
result = CHKInventory(self._search_key_name)
1606
if propagate_caches:
1607
# Just propagate the path-to-fileid cache for now
1608
result._path_to_fileid_cache = dict(self._path_to_fileid_cache.iteritems())
1609
search_key_func = chk_map.search_key_registry.get(self._search_key_name)
1610
self.id_to_entry._ensure_root()
1611
maximum_size = self.id_to_entry._root_node.maximum_size
1612
result.revision_id = new_revision_id
1613
result.id_to_entry = chk_map.CHKMap(
1614
self.id_to_entry._store,
1615
self.id_to_entry.key(),
1616
search_key_func=search_key_func)
1617
result.id_to_entry._ensure_root()
1618
result.id_to_entry._root_node.set_maximum_size(maximum_size)
1619
# Change to apply to the parent_id_basename delta. The dict maps
1620
# (parent_id, basename) -> (old_key, new_value). We use a dict because
1621
# when a path has its id replaced (e.g. the root is changed, or someone
1622
# does bzr mv a b, bzr mv c a, we should output a single change to this
1623
# map rather than two.
1624
parent_id_basename_delta = {}
1625
if self.parent_id_basename_to_file_id is not None:
1626
result.parent_id_basename_to_file_id = chk_map.CHKMap(
1627
self.parent_id_basename_to_file_id._store,
1628
self.parent_id_basename_to_file_id.key(),
1629
search_key_func=search_key_func)
1630
result.parent_id_basename_to_file_id._ensure_root()
1631
self.parent_id_basename_to_file_id._ensure_root()
1632
result_p_id_root = result.parent_id_basename_to_file_id._root_node
1633
p_id_root = self.parent_id_basename_to_file_id._root_node
1634
result_p_id_root.set_maximum_size(p_id_root.maximum_size)
1635
result_p_id_root._key_width = p_id_root._key_width
1637
result.parent_id_basename_to_file_id = None
1638
result.root_id = self.root_id
1639
id_to_entry_delta = []
1640
# inventory_delta is only traversed once, so we just update the
1642
# Check for repeated file ids
1643
inventory_delta = _check_delta_unique_ids(inventory_delta)
1644
# Repeated old paths
1645
inventory_delta = _check_delta_unique_old_paths(inventory_delta)
1646
# Check for repeated new paths
1647
inventory_delta = _check_delta_unique_new_paths(inventory_delta)
1648
# Check for entries that don't match the fileid
1649
inventory_delta = _check_delta_ids_match_entry(inventory_delta)
1650
# Check for nonsense fileids
1651
inventory_delta = _check_delta_ids_are_valid(inventory_delta)
1652
# Check for new_path <-> entry consistency
1653
inventory_delta = _check_delta_new_path_entry_both_or_None(
1655
# All changed entries need to have their parents be directories and be
1656
# at the right path. This set contains (path, id) tuples.
1658
# When we delete an item, all the children of it must be either deleted
1659
# or altered in their own right. As we batch process the change via
1660
# CHKMap.apply_delta, we build a set of things to use to validate the
1664
for old_path, new_path, file_id, entry in inventory_delta:
1667
result.root_id = file_id
1668
if new_path is None:
1673
if propagate_caches:
1675
del result._path_to_fileid_cache[old_path]
1678
deletes.add(file_id)
1680
new_key = (file_id,)
1681
new_value = result._entry_to_bytes(entry)
1682
# Update caches. It's worth doing this whether
1683
# we're propagating the old caches or not.
1684
result._path_to_fileid_cache[new_path] = file_id
1685
parents.add((split(new_path)[0], entry.parent_id))
1686
if old_path is None:
1689
old_key = (file_id,)
1690
if self.id2path(file_id) != old_path:
1691
raise errors.InconsistentDelta(old_path, file_id,
1692
"Entry was at wrong other path %r." %
1693
self.id2path(file_id))
1694
altered.add(file_id)
1695
id_to_entry_delta.append((old_key, new_key, new_value))
1696
if result.parent_id_basename_to_file_id is not None:
1697
# parent_id, basename changes
1698
if old_path is None:
1701
old_entry = self[file_id]
1702
old_key = self._parent_id_basename_key(old_entry)
1703
if new_path is None:
1707
new_key = self._parent_id_basename_key(entry)
1709
# If the two keys are the same, the value will be unchanged
1710
# as its always the file id for this entry.
1711
if old_key != new_key:
1712
# Transform a change into explicit delete/add preserving
1713
# a possible match on the key from a different file id.
1714
if old_key is not None:
1715
parent_id_basename_delta.setdefault(
1716
old_key, [None, None])[0] = old_key
1717
if new_key is not None:
1718
parent_id_basename_delta.setdefault(
1719
new_key, [None, None])[1] = new_value
1720
# validate that deletes are complete.
1721
for file_id in deletes:
1722
entry = self[file_id]
1723
if entry.kind != 'directory':
1725
# This loop could potentially be better by using the id_basename
1726
# map to just get the child file ids.
1727
for child in entry.children.values():
1728
if child.file_id not in altered:
1729
raise errors.InconsistentDelta(self.id2path(child.file_id),
1730
child.file_id, "Child not deleted or reparented when "
1732
result.id_to_entry.apply_delta(id_to_entry_delta)
1733
if parent_id_basename_delta:
1734
# Transform the parent_id_basename delta data into a linear delta
1735
# with only one record for a given key. Optimally this would allow
1736
# re-keying, but its simpler to just output that as a delete+add
1737
# to spend less time calculating the delta.
1739
for key, (old_key, value) in parent_id_basename_delta.iteritems():
1740
if value is not None:
1741
delta_list.append((old_key, key, value))
1743
delta_list.append((old_key, None, None))
1744
result.parent_id_basename_to_file_id.apply_delta(delta_list)
1745
parents.discard(('', None))
1746
for parent_path, parent in parents:
1748
if result[parent].kind != 'directory':
1749
raise errors.InconsistentDelta(result.id2path(parent), parent,
1750
'Not a directory, but given children')
1751
except errors.NoSuchId:
1752
raise errors.InconsistentDelta("<unknown>", parent,
1753
"Parent is not present in resulting inventory.")
1754
if result.path2id(parent_path) != parent:
1755
raise errors.InconsistentDelta(parent_path, parent,
1756
"Parent has wrong path %r." % result.path2id(parent_path))
1760
def deserialise(klass, chk_store, bytes, expected_revision_id):
1761
"""Deserialise a CHKInventory.
1763
:param chk_store: A CHK capable VersionedFiles instance.
1764
:param bytes: The serialised bytes.
1765
:param expected_revision_id: The revision ID we think this inventory is
1767
:return: A CHKInventory
1769
lines = bytes.split('\n')
1771
raise AssertionError('bytes to deserialize must end with an eol')
1773
if lines[0] != 'chkinventory:':
1774
raise ValueError("not a serialised CHKInventory: %r" % bytes)
1776
allowed_keys = frozenset(['root_id', 'revision_id', 'search_key_name',
1777
'parent_id_basename_to_file_id',
1779
for line in lines[1:]:
1780
key, value = line.split(': ', 1)
1781
if key not in allowed_keys:
1782
raise errors.BzrError('Unknown key in inventory: %r\n%r'
1785
raise errors.BzrError('Duplicate key in inventory: %r\n%r'
1788
revision_id = info['revision_id']
1789
root_id = info['root_id']
1790
search_key_name = info.get('search_key_name', 'plain')
1791
parent_id_basename_to_file_id = info.get(
1792
'parent_id_basename_to_file_id', None)
1793
id_to_entry = info['id_to_entry']
1795
result = CHKInventory(search_key_name)
1796
result.revision_id = revision_id
1797
result.root_id = root_id
1798
search_key_func = chk_map.search_key_registry.get(
1799
result._search_key_name)
1800
if parent_id_basename_to_file_id is not None:
1801
result.parent_id_basename_to_file_id = chk_map.CHKMap(
1802
chk_store, (parent_id_basename_to_file_id,),
1803
search_key_func=search_key_func)
1805
result.parent_id_basename_to_file_id = None
1807
result.id_to_entry = chk_map.CHKMap(chk_store, (id_to_entry,),
1808
search_key_func=search_key_func)
1809
if (result.revision_id,) != expected_revision_id:
1810
raise ValueError("Mismatched revision id and expected: %r, %r" %
1811
(result.revision_id, expected_revision_id))
1815
def from_inventory(klass, chk_store, inventory, maximum_size=0, search_key_name='plain'):
1816
"""Create a CHKInventory from an existing inventory.
1818
The content of inventory is copied into the chk_store, and a
1819
CHKInventory referencing that is returned.
1821
:param chk_store: A CHK capable VersionedFiles instance.
1822
:param inventory: The inventory to copy.
1823
:param maximum_size: The CHKMap node size limit.
1824
:param search_key_name: The identifier for the search key function
1826
result = klass(search_key_name)
1827
result.revision_id = inventory.revision_id
1828
result.root_id = inventory.root.file_id
1830
entry_to_bytes = result._entry_to_bytes
1831
parent_id_basename_key = result._parent_id_basename_key
1832
id_to_entry_dict = {}
1833
parent_id_basename_dict = {}
1834
for path, entry in inventory.iter_entries():
1835
id_to_entry_dict[(entry.file_id,)] = entry_to_bytes(entry)
1836
p_id_key = parent_id_basename_key(entry)
1837
parent_id_basename_dict[p_id_key] = entry.file_id
1839
result._populate_from_dicts(chk_store, id_to_entry_dict,
1840
parent_id_basename_dict, maximum_size=maximum_size)
1843
def _populate_from_dicts(self, chk_store, id_to_entry_dict,
1844
parent_id_basename_dict, maximum_size):
1845
search_key_func = chk_map.search_key_registry.get(self._search_key_name)
1846
root_key = chk_map.CHKMap.from_dict(chk_store, id_to_entry_dict,
1847
maximum_size=maximum_size, key_width=1,
1848
search_key_func=search_key_func)
1849
self.id_to_entry = chk_map.CHKMap(chk_store, root_key,
1851
root_key = chk_map.CHKMap.from_dict(chk_store,
1852
parent_id_basename_dict,
1853
maximum_size=maximum_size, key_width=2,
1854
search_key_func=search_key_func)
1855
self.parent_id_basename_to_file_id = chk_map.CHKMap(chk_store,
1856
root_key, search_key_func)
1858
def _parent_id_basename_key(self, entry):
1859
"""Create a key for a entry in a parent_id_basename_to_file_id index."""
1860
if entry.parent_id is not None:
1861
parent_id = entry.parent_id
1864
return parent_id, entry.name.encode('utf8')
1866
def __getitem__(self, file_id):
1867
"""map a single file_id -> InventoryEntry."""
1869
raise errors.NoSuchId(self, file_id)
1870
result = self._fileid_to_entry_cache.get(file_id, None)
1871
if result is not None:
1874
return self._bytes_to_entry(
1875
self.id_to_entry.iteritems([(file_id,)]).next()[1])
1876
except StopIteration:
1877
# really we're passing an inventory, not a tree...
1878
raise errors.NoSuchId(self, file_id)
1880
def has_id(self, file_id):
1881
# Perhaps have an explicit 'contains' method on CHKMap ?
1882
if self._fileid_to_entry_cache.get(file_id, None) is not None:
1884
return len(list(self.id_to_entry.iteritems([(file_id,)]))) == 1
1886
def is_root(self, file_id):
1887
return file_id == self.root_id
1889
def _iter_file_id_parents(self, file_id):
1890
"""Yield the parents of file_id up to the root."""
1891
while file_id is not None:
1895
raise errors.NoSuchId(tree=self, file_id=file_id)
1897
file_id = ie.parent_id
1900
"""Iterate over all file-ids."""
1901
for key, _ in self.id_to_entry.iteritems():
1904
def iter_just_entries(self):
1905
"""Iterate over all entries.
1907
Unlike iter_entries(), just the entries are returned (not (path, ie))
1908
and the order of entries is undefined.
1910
XXX: We may not want to merge this into bzr.dev.
1912
for key, entry in self.id_to_entry.iteritems():
1914
ie = self._fileid_to_entry_cache.get(file_id, None)
1916
ie = self._bytes_to_entry(entry)
1917
self._fileid_to_entry_cache[file_id] = ie
1920
def iter_changes(self, basis):
1921
"""Generate a Tree.iter_changes change list between this and basis.
1923
:param basis: Another CHKInventory.
1924
:return: An iterator over the changes between self and basis, as per
1925
tree.iter_changes().
1927
# We want: (file_id, (path_in_source, path_in_target),
1928
# changed_content, versioned, parent, name, kind,
1930
for key, basis_value, self_value in \
1931
self.id_to_entry.iter_changes(basis.id_to_entry):
1933
if basis_value is not None:
1934
basis_entry = basis._bytes_to_entry(basis_value)
1935
path_in_source = basis.id2path(file_id)
1936
basis_parent = basis_entry.parent_id
1937
basis_name = basis_entry.name
1938
basis_executable = basis_entry.executable
1940
path_in_source = None
1943
basis_executable = None
1944
if self_value is not None:
1945
self_entry = self._bytes_to_entry(self_value)
1946
path_in_target = self.id2path(file_id)
1947
self_parent = self_entry.parent_id
1948
self_name = self_entry.name
1949
self_executable = self_entry.executable
1951
path_in_target = None
1954
self_executable = None
1955
if basis_value is None:
1957
kind = (None, self_entry.kind)
1958
versioned = (False, True)
1959
elif self_value is None:
1961
kind = (basis_entry.kind, None)
1962
versioned = (True, False)
1964
kind = (basis_entry.kind, self_entry.kind)
1965
versioned = (True, True)
1966
changed_content = False
1967
if kind[0] != kind[1]:
1968
changed_content = True
1969
elif kind[0] == 'file':
1970
if (self_entry.text_size != basis_entry.text_size or
1971
self_entry.text_sha1 != basis_entry.text_sha1):
1972
changed_content = True
1973
elif kind[0] == 'symlink':
1974
if self_entry.symlink_target != basis_entry.symlink_target:
1975
changed_content = True
1976
elif kind[0] == 'tree-reference':
1977
if (self_entry.reference_revision !=
1978
basis_entry.reference_revision):
1979
changed_content = True
1980
parent = (basis_parent, self_parent)
1981
name = (basis_name, self_name)
1982
executable = (basis_executable, self_executable)
1983
if (not changed_content
1984
and parent[0] == parent[1]
1985
and name[0] == name[1]
1986
and executable[0] == executable[1]):
1987
# Could happen when only the revision changed for a directory
1990
yield (file_id, (path_in_source, path_in_target), changed_content,
1991
versioned, parent, name, kind, executable)
1994
"""Return the number of entries in the inventory."""
1995
return len(self.id_to_entry)
1997
def _make_delta(self, old):
1998
"""Make an inventory delta from two inventories."""
1999
if type(old) != CHKInventory:
2000
return CommonInventory._make_delta(self, old)
2002
for key, old_value, self_value in \
2003
self.id_to_entry.iter_changes(old.id_to_entry):
2005
if old_value is not None:
2006
old_path = old.id2path(file_id)
2009
if self_value is not None:
2010
entry = self._bytes_to_entry(self_value)
2011
self._fileid_to_entry_cache[file_id] = entry
2012
new_path = self.id2path(file_id)
2016
delta.append((old_path, new_path, file_id, entry))
2019
def path2id(self, name):
2020
"""See CommonInventory.path2id()."""
2021
# TODO: perhaps support negative hits?
2022
result = self._path_to_fileid_cache.get(name, None)
2023
if result is not None:
2025
if isinstance(name, basestring):
2026
names = osutils.splitpath(name)
2029
current_id = self.root_id
2030
if current_id is None:
2032
parent_id_index = self.parent_id_basename_to_file_id
2033
for basename in names:
2034
# TODO: Cache each path we figure out in this function.
2035
basename_utf8 = basename.encode('utf8')
2036
key_filter = [(current_id, basename_utf8)]
2038
for (parent_id, name_utf8), file_id in parent_id_index.iteritems(
2039
key_filter=key_filter):
2040
if parent_id != current_id or name_utf8 != basename_utf8:
2041
raise errors.BzrError("corrupt inventory lookup! "
2042
"%r %r %r %r" % (parent_id, current_id, name_utf8,
2046
current_id = file_id
2047
self._path_to_fileid_cache[name] = current_id
2051
"""Serialise the inventory to lines."""
2052
lines = ["chkinventory:\n"]
2053
if self._search_key_name != 'plain':
2054
# custom ordering grouping things that don't change together
2055
lines.append('search_key_name: %s\n' % (self._search_key_name,))
2056
lines.append("root_id: %s\n" % self.root_id)
2057
lines.append('parent_id_basename_to_file_id: %s\n' %
2058
self.parent_id_basename_to_file_id.key())
2059
lines.append("revision_id: %s\n" % self.revision_id)
2060
lines.append("id_to_entry: %s\n" % self.id_to_entry.key())
2062
lines.append("revision_id: %s\n" % self.revision_id)
2063
lines.append("root_id: %s\n" % self.root_id)
2064
if self.parent_id_basename_to_file_id is not None:
2065
lines.append('parent_id_basename_to_file_id: %s\n' %
2066
self.parent_id_basename_to_file_id.key())
2067
lines.append("id_to_entry: %s\n" % self.id_to_entry.key())
2072
"""Get the root entry."""
2073
return self[self.root_id]
2076
class CHKInventoryDirectory(InventoryDirectory):
2077
"""A directory in an inventory."""
2079
__slots__ = ['text_sha1', 'text_size', 'file_id', 'name', 'kind',
2080
'text_id', 'parent_id', '_children', 'executable',
2081
'revision', 'symlink_target', 'reference_revision',
2084
def __init__(self, file_id, name, parent_id, chk_inventory):
2085
# Don't call InventoryDirectory.__init__ - it isn't right for this
2087
InventoryEntry.__init__(self, file_id, name, parent_id)
2088
self._children = None
2089
self.kind = 'directory'
2090
self._chk_inventory = chk_inventory
2094
"""Access the list of children of this directory.
2096
With a parent_id_basename_to_file_id index, loads all the children,
2097
without loads the entire index. Without is bad. A more sophisticated
2098
proxy object might be nice, to allow partial loading of children as
2099
well when specific names are accessed. (So path traversal can be
2100
written in the obvious way but not examine siblings.).
2102
if self._children is not None:
2103
return self._children
2104
# No longer supported
2105
if self._chk_inventory.parent_id_basename_to_file_id is None:
2106
raise AssertionError("Inventories without"
2107
" parent_id_basename_to_file_id are no longer supported")
2109
# XXX: Todo - use proxy objects for the children rather than loading
2110
# all when the attribute is referenced.
2111
parent_id_index = self._chk_inventory.parent_id_basename_to_file_id
2113
for (parent_id, name_utf8), file_id in parent_id_index.iteritems(
2114
key_filter=[(self.file_id,)]):
2115
child_keys.add((file_id,))
2117
for file_id_key in child_keys:
2118
entry = self._chk_inventory._fileid_to_entry_cache.get(
2119
file_id_key[0], None)
2120
if entry is not None:
2121
result[entry.name] = entry
2122
cached.add(file_id_key)
2123
child_keys.difference_update(cached)
2124
# populate; todo: do by name
2125
id_to_entry = self._chk_inventory.id_to_entry
2126
for file_id_key, bytes in id_to_entry.iteritems(child_keys):
2127
entry = self._chk_inventory._bytes_to_entry(bytes)
2128
result[entry.name] = entry
2129
self._chk_inventory._fileid_to_entry_cache[file_id_key[0]] = entry
2130
self._children = result
1291
2133
entry_factory = {
1292
2134
'directory': InventoryDirectory,
1293
2135
'file': InventoryFile,
added
removed
bzrlib/inventory.py
