709
614
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.
617
class CommonInventory(object):
618
"""Basic inventory logic, defined in terms of primitives like has_id.
620
An inventory is the metadata about the contents of a tree.
622
This is broadly a map from file_id to entries such as directories, files,
623
symlinks and tree references. Each entry maintains its own metadata like
624
SHA1 and length for files, or children for a directory.
717
626
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
628
InventoryEntry objects must not be modified after they are
727
629
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."""
632
def __contains__(self, file_id):
633
"""True if this entry contains a file with given id.
635
>>> inv = Inventory()
636
>>> inv.add(InventoryFile('123', 'foo.c', ROOT_ID))
637
InventoryFile('123', 'foo.c', parent_id='TREE_ROOT', sha1=None, len=None, revision=None)
643
Note that this method along with __iter__ are not encouraged for use as
644
they are less clear than specific query methods - they may be rmeoved
647
return self.has_id(file_id)
649
def has_filename(self, filename):
650
return bool(self.path2id(filename))
652
def id2path(self, file_id):
653
"""Return as a string the path to file_id.
656
>>> e = i.add(InventoryDirectory('src-id', 'src', ROOT_ID))
657
>>> e = i.add(InventoryFile('foo-id', 'foo.c', parent_id='src-id'))
658
>>> print i.id2path('foo-id')
661
:raises NoSuchId: If file_id is not present in the inventory.
663
# get all names, skipping root
664
return '/'.join(reversed(
665
[parent.name for parent in
666
self._iter_file_id_parents(file_id)][:-1]))
668
def iter_entries(self, from_dir=None, recursive=True):
669
"""Return (path, entry) pairs, in order by name.
671
:param from_dir: if None, start from the root,
672
otherwise start from this directory (either file-id or entry)
673
:param recursive: recurse into directories or not
869
675
if from_dir is None:
870
676
if self.root is None:
872
678
from_dir = self.root
873
679
yield '', self.root
874
680
elif isinstance(from_dir, basestring):
875
from_dir = self._byid[from_dir]
681
from_dir = self[from_dir]
877
683
# unrolling the recursive called changed the time from
878
684
# 440ms/663ms (inline/total) to 116ms/116ms
879
685
children = from_dir.children.items()
688
for name, ie in children:
881
691
children = collections.deque(children)
882
692
stack = [(u'', children)]
1018
860
descend(child_ie, child_path)
1019
861
descend(self.root, u'')
864
def path2id(self, relpath):
865
"""Walk down through directories to return entry of last component.
867
:param relpath: may be either a list of path components, or a single
868
string, in which case it is automatically split.
870
This returns the entry of the last component in the path,
871
which may be either a file or a directory.
873
Returns None IFF the path is not found.
875
if isinstance(relpath, basestring):
876
names = osutils.splitpath(relpath)
882
except errors.NoSuchId:
883
# root doesn't exist yet so nothing else can
889
children = getattr(parent, 'children', None)
898
return parent.file_id
900
def filter(self, specific_fileids):
901
"""Get an inventory view filtered against a set of file-ids.
903
Children of directories and parents are included.
905
The result may or may not reference the underlying inventory
906
so it should be treated as immutable.
908
interesting_parents = set()
909
for fileid in specific_fileids:
911
interesting_parents.update(self.get_idpath(fileid))
912
except errors.NoSuchId:
913
# This fileid is not in the inventory - that's ok
915
entries = self.iter_entries()
916
if self.root is None:
917
return Inventory(root_id=None)
918
other = Inventory(entries.next()[1].file_id)
919
other.root.revision = self.root.revision
920
other.revision_id = self.revision_id
921
directories_to_expand = set()
922
for path, entry in entries:
923
file_id = entry.file_id
924
if (file_id in specific_fileids
925
or entry.parent_id in directories_to_expand):
926
if entry.kind == 'directory':
927
directories_to_expand.add(file_id)
928
elif file_id not in interesting_parents:
930
other.add(entry.copy())
933
def get_idpath(self, file_id):
934
"""Return a list of file_ids for the path to an entry.
936
The list contains one element for each directory followed by
937
the id of the file itself. So the length of the returned list
938
is equal to the depth of the file in the tree, counting the
939
root directory as depth 1.
942
for parent in self._iter_file_id_parents(file_id):
943
p.insert(0, parent.file_id)
947
class Inventory(CommonInventory):
948
"""Mutable dict based in-memory inventory.
950
We never store the full path to a file, because renaming a directory
951
implicitly moves all of its contents. This class internally maintains a
952
lookup tree that allows the children under a directory to be
955
>>> inv = Inventory()
956
>>> inv.add(InventoryFile('123-123', 'hello.c', ROOT_ID))
957
InventoryFile('123-123', 'hello.c', parent_id='TREE_ROOT', sha1=None, len=None, revision=None)
958
>>> inv['123-123'].name
961
Id's may be looked up from paths:
963
>>> inv.path2id('hello.c')
968
There are iterators over the contents:
970
>>> [entry[0] for entry in inv.iter_entries()]
974
def __init__(self, root_id=ROOT_ID, revision_id=None):
975
"""Create or read an inventory.
977
If a working directory is specified, the inventory is read
978
from there. If the file is specified, read from that. If not,
979
the inventory is created empty.
981
The inventory is created with a default root directory, with
984
if root_id is not None:
985
self._set_root(InventoryDirectory(root_id, u'', None))
989
self.revision_id = revision_id
992
# More than one page of ouput is not useful anymore to debug
995
contents = repr(self._byid)
996
if len(contents) > max_len:
997
contents = contents[:(max_len-len(closing))] + closing
998
return "<Inventory object at %x, contents=%r>" % (id(self), contents)
1000
def apply_delta(self, delta):
1001
"""Apply a delta to this inventory.
1003
See the inventory developers documentation for the theory behind
1006
If delta application fails the inventory is left in an indeterminate
1007
state and must not be used.
1009
:param delta: A list of changes to apply. After all the changes are
1010
applied the final inventory must be internally consistent, but it
1011
is ok to supply changes which, if only half-applied would have an
1012
invalid result - such as supplying two changes which rename two
1013
files, 'A' and 'B' with each other : [('A', 'B', 'A-id', a_entry),
1014
('B', 'A', 'B-id', b_entry)].
1016
Each change is a tuple, of the form (old_path, new_path, file_id,
1019
When new_path is None, the change indicates the removal of an entry
1020
from the inventory and new_entry will be ignored (using None is
1021
appropriate). If new_path is not None, then new_entry must be an
1022
InventoryEntry instance, which will be incorporated into the
1023
inventory (and replace any existing entry with the same file id).
1025
When old_path is None, the change indicates the addition of
1026
a new entry to the inventory.
1028
When neither new_path nor old_path are None, the change is a
1029
modification to an entry, such as a rename, reparent, kind change
1032
The children attribute of new_entry is ignored. This is because
1033
this method preserves children automatically across alterations to
1034
the parent of the children, and cases where the parent id of a
1035
child is changing require the child to be passed in as a separate
1036
change regardless. E.g. in the recursive deletion of a directory -
1037
the directory's children must be included in the delta, or the
1038
final inventory will be invalid.
1040
Note that a file_id must only appear once within a given delta.
1041
An AssertionError is raised otherwise.
1043
# Check that the delta is legal. It would be nice if this could be
1044
# done within the loops below but it's safer to validate the delta
1045
# before starting to mutate the inventory, as there isn't a rollback
1047
list(_check_delta_unique_ids(_check_delta_unique_new_paths(
1048
_check_delta_unique_old_paths(_check_delta_ids_match_entry(
1049
_check_delta_ids_are_valid(
1050
_check_delta_new_path_entry_both_or_None(
1054
# Remove all affected items which were in the original inventory,
1055
# starting with the longest paths, thus ensuring parents are examined
1056
# after their children, which means that everything we examine has no
1057
# modified children remaining by the time we examine it.
1058
for old_path, file_id in sorted(((op, f) for op, np, f, e in delta
1059
if op is not None), reverse=True):
1060
# Preserve unaltered children of file_id for later reinsertion.
1061
file_id_children = getattr(self[file_id], 'children', {})
1062
if len(file_id_children):
1063
children[file_id] = file_id_children
1064
if self.id2path(file_id) != old_path:
1065
raise errors.InconsistentDelta(old_path, file_id,
1066
"Entry was at wrong other path %r." % self.id2path(file_id))
1067
# Remove file_id and the unaltered children. If file_id is not
1068
# being deleted it will be reinserted back later.
1069
self.remove_recursive_id(file_id)
1070
# Insert all affected which should be in the new inventory, reattaching
1071
# their children if they had any. This is done from shortest path to
1072
# longest, ensuring that items which were modified and whose parents in
1073
# the resulting inventory were also modified, are inserted after their
1075
for new_path, f, new_entry in sorted((np, f, e) for op, np, f, e in
1076
delta if np is not None):
1077
if new_entry.kind == 'directory':
1078
# Pop the child which to allow detection of children whose
1079
# parents were deleted and which were not reattached to a new
1081
replacement = InventoryDirectory(new_entry.file_id,
1082
new_entry.name, new_entry.parent_id)
1083
replacement.revision = new_entry.revision
1084
replacement.children = children.pop(replacement.file_id, {})
1085
new_entry = replacement
1088
except errors.DuplicateFileId:
1089
raise errors.InconsistentDelta(new_path, new_entry.file_id,
1090
"New id is already present in target.")
1091
except AttributeError:
1092
raise errors.InconsistentDelta(new_path, new_entry.file_id,
1093
"Parent is not a directory.")
1094
if self.id2path(new_entry.file_id) != new_path:
1095
raise errors.InconsistentDelta(new_path, new_entry.file_id,
1096
"New path is not consistent with parent path.")
1098
# Get the parent id that was deleted
1099
parent_id, children = children.popitem()
1100
raise errors.InconsistentDelta("<deleted>", parent_id,
1101
"The file id was deleted but its children were not deleted.")
1103
def create_by_apply_delta(self, inventory_delta, new_revision_id,
1104
propagate_caches=False):
1105
"""See CHKInventory.create_by_apply_delta()"""
1106
new_inv = self.copy()
1107
new_inv.apply_delta(inventory_delta)
1108
new_inv.revision_id = new_revision_id
1111
def _set_root(self, ie):
1113
self._byid = {self.root.file_id: self.root}
1116
# TODO: jam 20051218 Should copy also copy the revision_id?
1117
entries = self.iter_entries()
1118
if self.root is None:
1119
return Inventory(root_id=None)
1120
other = Inventory(entries.next()[1].file_id)
1121
other.root.revision = self.root.revision
1122
# copy recursively so we know directories will be added before
1123
# their children. There are more efficient ways than this...
1124
for path, entry in entries:
1125
other.add(entry.copy())
1129
"""Iterate over all file-ids."""
1130
return iter(self._byid)
1132
def iter_just_entries(self):
1133
"""Iterate over all entries.
1022
def __contains__(self, file_id):
1023
"""True if this entry contains a file with given id.
1135
Unlike iter_entries(), just the entries are returned (not (path, ie))
1136
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)
1138
XXX: We may not want to merge this into bzr.dev.
1033
return (file_id in self._byid)
1140
if self.root is None:
1142
for _, ie in self._byid.iteritems():
1146
"""Returns number of entries."""
1147
return len(self._byid)
1035
1149
def __getitem__(self, file_id):
1036
1150
"""Return the entry for given file_id.
1038
1152
>>> inv = Inventory()
1039
1153
>>> inv.add(InventoryFile('123123', 'hello.c', ROOT_ID))
1040
InventoryFile('123123', 'hello.c', parent_id='TREE_ROOT', sha1=None, len=None)
1154
InventoryFile('123123', 'hello.c', parent_id='TREE_ROOT', sha1=None, len=None, revision=None)
1041
1155
>>> inv['123123'].name
1288
1370
return self.root is not None and file_id == self.root.file_id
1373
class CHKInventory(CommonInventory):
1374
"""An inventory persisted in a CHK store.
1376
By design, a CHKInventory is immutable so many of the methods
1377
supported by Inventory - add, rename, apply_delta, etc - are *not*
1378
supported. To create a new CHKInventory, use create_by_apply_delta()
1379
or from_inventory(), say.
1381
Internally, a CHKInventory has one or two CHKMaps:
1383
* id_to_entry - a map from (file_id,) => InventoryEntry as bytes
1384
* parent_id_basename_to_file_id - a map from (parent_id, basename_utf8)
1387
The second map is optional and not present in early CHkRepository's.
1389
No caching is performed: every method call or item access will perform
1390
requests to the storage layer. As such, keep references to objects you
1394
def __init__(self, search_key_name):
1395
CommonInventory.__init__(self)
1396
self._fileid_to_entry_cache = {}
1397
self._fully_cached = False
1398
self._path_to_fileid_cache = {}
1399
self._search_key_name = search_key_name
1402
def __eq__(self, other):
1403
"""Compare two sets by comparing their contents."""
1404
if not isinstance(other, CHKInventory):
1405
return NotImplemented
1407
this_key = self.id_to_entry.key()
1408
other_key = other.id_to_entry.key()
1409
this_pid_key = self.parent_id_basename_to_file_id.key()
1410
other_pid_key = other.parent_id_basename_to_file_id.key()
1411
if None in (this_key, this_pid_key, other_key, other_pid_key):
1413
return this_key == other_key and this_pid_key == other_pid_key
1415
def _entry_to_bytes(self, entry):
1416
"""Serialise entry as a single bytestring.
1418
:param Entry: An inventory entry.
1419
:return: A bytestring for the entry.
1422
ENTRY ::= FILE | DIR | SYMLINK | TREE
1423
FILE ::= "file: " COMMON SEP SHA SEP SIZE SEP EXECUTABLE
1424
DIR ::= "dir: " COMMON
1425
SYMLINK ::= "symlink: " COMMON SEP TARGET_UTF8
1426
TREE ::= "tree: " COMMON REFERENCE_REVISION
1427
COMMON ::= FILE_ID SEP PARENT_ID SEP NAME_UTF8 SEP REVISION
1430
if entry.parent_id is not None:
1431
parent_str = entry.parent_id
1434
name_str = entry.name.encode("utf8")
1435
if entry.kind == 'file':
1436
if entry.executable:
1440
return "file: %s\n%s\n%s\n%s\n%s\n%d\n%s" % (
1441
entry.file_id, parent_str, name_str, entry.revision,
1442
entry.text_sha1, entry.text_size, exec_str)
1443
elif entry.kind == 'directory':
1444
return "dir: %s\n%s\n%s\n%s" % (
1445
entry.file_id, parent_str, name_str, entry.revision)
1446
elif entry.kind == 'symlink':
1447
return "symlink: %s\n%s\n%s\n%s\n%s" % (
1448
entry.file_id, parent_str, name_str, entry.revision,
1449
entry.symlink_target.encode("utf8"))
1450
elif entry.kind == 'tree-reference':
1451
return "tree: %s\n%s\n%s\n%s\n%s" % (
1452
entry.file_id, parent_str, name_str, entry.revision,
1453
entry.reference_revision)
1455
raise ValueError("unknown kind %r" % entry.kind)
1457
def _expand_fileids_to_parents_and_children(self, file_ids):
1458
"""Give a more wholistic view starting with the given file_ids.
1460
For any file_id which maps to a directory, we will include all children
1461
of that directory. We will also include all directories which are
1462
parents of the given file_ids, but we will not include their children.
1469
fringle # fringle-id
1473
if given [foo-id] we will include
1474
TREE_ROOT as interesting parents
1476
foo-id, baz-id, frob-id, fringle-id
1480
# TODO: Pre-pass over the list of fileids to see if anything is already
1481
# deserialized in self._fileid_to_entry_cache
1483
directories_to_expand = set()
1484
children_of_parent_id = {}
1485
# It is okay if some of the fileids are missing
1486
for entry in self._getitems(file_ids):
1487
if entry.kind == 'directory':
1488
directories_to_expand.add(entry.file_id)
1489
interesting.add(entry.parent_id)
1490
children_of_parent_id.setdefault(entry.parent_id, []
1491
).append(entry.file_id)
1493
# Now, interesting has all of the direct parents, but not the
1494
# parents of those parents. It also may have some duplicates with
1496
remaining_parents = interesting.difference(file_ids)
1497
# When we hit the TREE_ROOT, we'll get an interesting parent of None,
1498
# but we don't actually want to recurse into that
1499
interesting.add(None) # this will auto-filter it in the loop
1500
remaining_parents.discard(None)
1501
while remaining_parents:
1502
next_parents = set()
1503
for entry in self._getitems(remaining_parents):
1504
next_parents.add(entry.parent_id)
1505
children_of_parent_id.setdefault(entry.parent_id, []
1506
).append(entry.file_id)
1507
# Remove any search tips we've already processed
1508
remaining_parents = next_parents.difference(interesting)
1509
interesting.update(remaining_parents)
1510
# We should probably also .difference(directories_to_expand)
1511
interesting.update(file_ids)
1512
interesting.discard(None)
1513
while directories_to_expand:
1514
# Expand directories by looking in the
1515
# parent_id_basename_to_file_id map
1516
keys = [StaticTuple(f,).intern() for f in directories_to_expand]
1517
directories_to_expand = set()
1518
items = self.parent_id_basename_to_file_id.iteritems(keys)
1519
next_file_ids = set([item[1] for item in items])
1520
next_file_ids = next_file_ids.difference(interesting)
1521
interesting.update(next_file_ids)
1522
for entry in self._getitems(next_file_ids):
1523
if entry.kind == 'directory':
1524
directories_to_expand.add(entry.file_id)
1525
children_of_parent_id.setdefault(entry.parent_id, []
1526
).append(entry.file_id)
1527
return interesting, children_of_parent_id
1529
def filter(self, specific_fileids):
1530
"""Get an inventory view filtered against a set of file-ids.
1532
Children of directories and parents are included.
1534
The result may or may not reference the underlying inventory
1535
so it should be treated as immutable.
1538
parent_to_children) = self._expand_fileids_to_parents_and_children(
1540
# There is some overlap here, but we assume that all interesting items
1541
# are in the _fileid_to_entry_cache because we had to read them to
1542
# determine if they were a dir we wanted to recurse, or just a file
1543
# This should give us all the entries we'll want to add, so start
1545
other = Inventory(self.root_id)
1546
other.root.revision = self.root.revision
1547
other.revision_id = self.revision_id
1548
if not interesting or not parent_to_children:
1549
# empty filter, or filtering entrys that don't exist
1550
# (if even 1 existed, then we would have populated
1551
# parent_to_children with at least the tree root.)
1553
cache = self._fileid_to_entry_cache
1554
remaining_children = collections.deque(parent_to_children[self.root_id])
1555
while remaining_children:
1556
file_id = remaining_children.popleft()
1558
if ie.kind == 'directory':
1559
ie = ie.copy() # We create a copy to depopulate the .children attribute
1560
# TODO: depending on the uses of 'other' we should probably alwyas
1561
# '.copy()' to prevent someone from mutating other and
1562
# invaliding our internal cache
1564
if file_id in parent_to_children:
1565
remaining_children.extend(parent_to_children[file_id])
1569
def _bytes_to_utf8name_key(bytes):
1570
"""Get the file_id, revision_id key out of bytes."""
1571
# We don't normally care about name, except for times when we want
1572
# to filter out empty names because of non rich-root...
1573
sections = bytes.split('\n')
1574
kind, file_id = sections[0].split(': ')
1575
return (sections[2], intern(file_id), intern(sections[3]))
1577
def _bytes_to_entry(self, bytes):
1578
"""Deserialise a serialised entry."""
1579
sections = bytes.split('\n')
1580
if sections[0].startswith("file: "):
1581
result = InventoryFile(sections[0][6:],
1582
sections[2].decode('utf8'),
1584
result.text_sha1 = sections[4]
1585
result.text_size = int(sections[5])
1586
result.executable = sections[6] == "Y"
1587
elif sections[0].startswith("dir: "):
1588
result = CHKInventoryDirectory(sections[0][5:],
1589
sections[2].decode('utf8'),
1591
elif sections[0].startswith("symlink: "):
1592
result = InventoryLink(sections[0][9:],
1593
sections[2].decode('utf8'),
1595
result.symlink_target = sections[4].decode('utf8')
1596
elif sections[0].startswith("tree: "):
1597
result = TreeReference(sections[0][6:],
1598
sections[2].decode('utf8'),
1600
result.reference_revision = sections[4]
1602
raise ValueError("Not a serialised entry %r" % bytes)
1603
result.file_id = intern(result.file_id)
1604
result.revision = intern(sections[3])
1605
if result.parent_id == '':
1606
result.parent_id = None
1607
self._fileid_to_entry_cache[result.file_id] = result
1610
def create_by_apply_delta(self, inventory_delta, new_revision_id,
1611
propagate_caches=False):
1612
"""Create a new CHKInventory by applying inventory_delta to this one.
1614
See the inventory developers documentation for the theory behind
1617
:param inventory_delta: The inventory delta to apply. See
1618
Inventory.apply_delta for details.
1619
:param new_revision_id: The revision id of the resulting CHKInventory.
1620
:param propagate_caches: If True, the caches for this inventory are
1621
copied to and updated for the result.
1622
:return: The new CHKInventory.
1624
split = osutils.split
1625
result = CHKInventory(self._search_key_name)
1626
if propagate_caches:
1627
# Just propagate the path-to-fileid cache for now
1628
result._path_to_fileid_cache = dict(self._path_to_fileid_cache.iteritems())
1629
search_key_func = chk_map.search_key_registry.get(self._search_key_name)
1630
self.id_to_entry._ensure_root()
1631
maximum_size = self.id_to_entry._root_node.maximum_size
1632
result.revision_id = new_revision_id
1633
result.id_to_entry = chk_map.CHKMap(
1634
self.id_to_entry._store,
1635
self.id_to_entry.key(),
1636
search_key_func=search_key_func)
1637
result.id_to_entry._ensure_root()
1638
result.id_to_entry._root_node.set_maximum_size(maximum_size)
1639
# Change to apply to the parent_id_basename delta. The dict maps
1640
# (parent_id, basename) -> (old_key, new_value). We use a dict because
1641
# when a path has its id replaced (e.g. the root is changed, or someone
1642
# does bzr mv a b, bzr mv c a, we should output a single change to this
1643
# map rather than two.
1644
parent_id_basename_delta = {}
1645
if self.parent_id_basename_to_file_id is not None:
1646
result.parent_id_basename_to_file_id = chk_map.CHKMap(
1647
self.parent_id_basename_to_file_id._store,
1648
self.parent_id_basename_to_file_id.key(),
1649
search_key_func=search_key_func)
1650
result.parent_id_basename_to_file_id._ensure_root()
1651
self.parent_id_basename_to_file_id._ensure_root()
1652
result_p_id_root = result.parent_id_basename_to_file_id._root_node
1653
p_id_root = self.parent_id_basename_to_file_id._root_node
1654
result_p_id_root.set_maximum_size(p_id_root.maximum_size)
1655
result_p_id_root._key_width = p_id_root._key_width
1657
result.parent_id_basename_to_file_id = None
1658
result.root_id = self.root_id
1659
id_to_entry_delta = []
1660
# inventory_delta is only traversed once, so we just update the
1662
# Check for repeated file ids
1663
inventory_delta = _check_delta_unique_ids(inventory_delta)
1664
# Repeated old paths
1665
inventory_delta = _check_delta_unique_old_paths(inventory_delta)
1666
# Check for repeated new paths
1667
inventory_delta = _check_delta_unique_new_paths(inventory_delta)
1668
# Check for entries that don't match the fileid
1669
inventory_delta = _check_delta_ids_match_entry(inventory_delta)
1670
# Check for nonsense fileids
1671
inventory_delta = _check_delta_ids_are_valid(inventory_delta)
1672
# Check for new_path <-> entry consistency
1673
inventory_delta = _check_delta_new_path_entry_both_or_None(
1675
# All changed entries need to have their parents be directories and be
1676
# at the right path. This set contains (path, id) tuples.
1678
# When we delete an item, all the children of it must be either deleted
1679
# or altered in their own right. As we batch process the change via
1680
# CHKMap.apply_delta, we build a set of things to use to validate the
1684
for old_path, new_path, file_id, entry in inventory_delta:
1687
result.root_id = file_id
1688
if new_path is None:
1693
if propagate_caches:
1695
del result._path_to_fileid_cache[old_path]
1698
deletes.add(file_id)
1700
new_key = StaticTuple(file_id,)
1701
new_value = result._entry_to_bytes(entry)
1702
# Update caches. It's worth doing this whether
1703
# we're propagating the old caches or not.
1704
result._path_to_fileid_cache[new_path] = file_id
1705
parents.add((split(new_path)[0], entry.parent_id))
1706
if old_path is None:
1709
old_key = StaticTuple(file_id,)
1710
if self.id2path(file_id) != old_path:
1711
raise errors.InconsistentDelta(old_path, file_id,
1712
"Entry was at wrong other path %r." %
1713
self.id2path(file_id))
1714
altered.add(file_id)
1715
id_to_entry_delta.append(StaticTuple(old_key, new_key, new_value))
1716
if result.parent_id_basename_to_file_id is not None:
1717
# parent_id, basename changes
1718
if old_path is None:
1721
old_entry = self[file_id]
1722
old_key = self._parent_id_basename_key(old_entry)
1723
if new_path is None:
1727
new_key = self._parent_id_basename_key(entry)
1729
# If the two keys are the same, the value will be unchanged
1730
# as its always the file id for this entry.
1731
if old_key != new_key:
1732
# Transform a change into explicit delete/add preserving
1733
# a possible match on the key from a different file id.
1734
if old_key is not None:
1735
parent_id_basename_delta.setdefault(
1736
old_key, [None, None])[0] = old_key
1737
if new_key is not None:
1738
parent_id_basename_delta.setdefault(
1739
new_key, [None, None])[1] = new_value
1740
# validate that deletes are complete.
1741
for file_id in deletes:
1742
entry = self[file_id]
1743
if entry.kind != 'directory':
1745
# This loop could potentially be better by using the id_basename
1746
# map to just get the child file ids.
1747
for child in entry.children.values():
1748
if child.file_id not in altered:
1749
raise errors.InconsistentDelta(self.id2path(child.file_id),
1750
child.file_id, "Child not deleted or reparented when "
1752
result.id_to_entry.apply_delta(id_to_entry_delta)
1753
if parent_id_basename_delta:
1754
# Transform the parent_id_basename delta data into a linear delta
1755
# with only one record for a given key. Optimally this would allow
1756
# re-keying, but its simpler to just output that as a delete+add
1757
# to spend less time calculating the delta.
1759
for key, (old_key, value) in parent_id_basename_delta.iteritems():
1760
if value is not None:
1761
delta_list.append((old_key, key, value))
1763
delta_list.append((old_key, None, None))
1764
result.parent_id_basename_to_file_id.apply_delta(delta_list)
1765
parents.discard(('', None))
1766
for parent_path, parent in parents:
1768
if result[parent].kind != 'directory':
1769
raise errors.InconsistentDelta(result.id2path(parent), parent,
1770
'Not a directory, but given children')
1771
except errors.NoSuchId:
1772
raise errors.InconsistentDelta("<unknown>", parent,
1773
"Parent is not present in resulting inventory.")
1774
if result.path2id(parent_path) != parent:
1775
raise errors.InconsistentDelta(parent_path, parent,
1776
"Parent has wrong path %r." % result.path2id(parent_path))
1780
def deserialise(klass, chk_store, bytes, expected_revision_id):
1781
"""Deserialise a CHKInventory.
1783
:param chk_store: A CHK capable VersionedFiles instance.
1784
:param bytes: The serialised bytes.
1785
:param expected_revision_id: The revision ID we think this inventory is
1787
:return: A CHKInventory
1789
lines = bytes.split('\n')
1791
raise AssertionError('bytes to deserialize must end with an eol')
1793
if lines[0] != 'chkinventory:':
1794
raise ValueError("not a serialised CHKInventory: %r" % bytes)
1796
allowed_keys = frozenset(['root_id', 'revision_id', 'search_key_name',
1797
'parent_id_basename_to_file_id',
1799
for line in lines[1:]:
1800
key, value = line.split(': ', 1)
1801
if key not in allowed_keys:
1802
raise errors.BzrError('Unknown key in inventory: %r\n%r'
1805
raise errors.BzrError('Duplicate key in inventory: %r\n%r'
1808
revision_id = intern(info['revision_id'])
1809
root_id = intern(info['root_id'])
1810
search_key_name = intern(info.get('search_key_name', 'plain'))
1811
parent_id_basename_to_file_id = intern(info.get(
1812
'parent_id_basename_to_file_id', None))
1813
if not parent_id_basename_to_file_id.startswith('sha1:'):
1814
raise ValueError('parent_id_basename_to_file_id should be a sha1'
1815
' key not %r' % (parent_id_basename_to_file_id,))
1816
id_to_entry = info['id_to_entry']
1817
if not id_to_entry.startswith('sha1:'):
1818
raise ValueError('id_to_entry should be a sha1'
1819
' key not %r' % (id_to_entry,))
1821
result = CHKInventory(search_key_name)
1822
result.revision_id = revision_id
1823
result.root_id = root_id
1824
search_key_func = chk_map.search_key_registry.get(
1825
result._search_key_name)
1826
if parent_id_basename_to_file_id is not None:
1827
result.parent_id_basename_to_file_id = chk_map.CHKMap(
1828
chk_store, StaticTuple(parent_id_basename_to_file_id,),
1829
search_key_func=search_key_func)
1831
result.parent_id_basename_to_file_id = None
1833
result.id_to_entry = chk_map.CHKMap(chk_store,
1834
StaticTuple(id_to_entry,),
1835
search_key_func=search_key_func)
1836
if (result.revision_id,) != expected_revision_id:
1837
raise ValueError("Mismatched revision id and expected: %r, %r" %
1838
(result.revision_id, expected_revision_id))
1842
def from_inventory(klass, chk_store, inventory, maximum_size=0, search_key_name='plain'):
1843
"""Create a CHKInventory from an existing inventory.
1845
The content of inventory is copied into the chk_store, and a
1846
CHKInventory referencing that is returned.
1848
:param chk_store: A CHK capable VersionedFiles instance.
1849
:param inventory: The inventory to copy.
1850
:param maximum_size: The CHKMap node size limit.
1851
:param search_key_name: The identifier for the search key function
1853
result = klass(search_key_name)
1854
result.revision_id = inventory.revision_id
1855
result.root_id = inventory.root.file_id
1857
entry_to_bytes = result._entry_to_bytes
1858
parent_id_basename_key = result._parent_id_basename_key
1859
id_to_entry_dict = {}
1860
parent_id_basename_dict = {}
1861
for path, entry in inventory.iter_entries():
1862
key = StaticTuple(entry.file_id,).intern()
1863
id_to_entry_dict[key] = entry_to_bytes(entry)
1864
p_id_key = parent_id_basename_key(entry)
1865
parent_id_basename_dict[p_id_key] = entry.file_id
1867
result._populate_from_dicts(chk_store, id_to_entry_dict,
1868
parent_id_basename_dict, maximum_size=maximum_size)
1871
def _populate_from_dicts(self, chk_store, id_to_entry_dict,
1872
parent_id_basename_dict, maximum_size):
1873
search_key_func = chk_map.search_key_registry.get(self._search_key_name)
1874
root_key = chk_map.CHKMap.from_dict(chk_store, id_to_entry_dict,
1875
maximum_size=maximum_size, key_width=1,
1876
search_key_func=search_key_func)
1877
self.id_to_entry = chk_map.CHKMap(chk_store, root_key,
1879
root_key = chk_map.CHKMap.from_dict(chk_store,
1880
parent_id_basename_dict,
1881
maximum_size=maximum_size, key_width=2,
1882
search_key_func=search_key_func)
1883
self.parent_id_basename_to_file_id = chk_map.CHKMap(chk_store,
1884
root_key, search_key_func)
1886
def _parent_id_basename_key(self, entry):
1887
"""Create a key for a entry in a parent_id_basename_to_file_id index."""
1888
if entry.parent_id is not None:
1889
parent_id = entry.parent_id
1892
return StaticTuple(parent_id, entry.name.encode('utf8')).intern()
1894
def __getitem__(self, file_id):
1895
"""map a single file_id -> InventoryEntry."""
1897
raise errors.NoSuchId(self, file_id)
1898
result = self._fileid_to_entry_cache.get(file_id, None)
1899
if result is not None:
1902
return self._bytes_to_entry(
1903
self.id_to_entry.iteritems([StaticTuple(file_id,)]).next()[1])
1904
except StopIteration:
1905
# really we're passing an inventory, not a tree...
1906
raise errors.NoSuchId(self, file_id)
1908
def _getitems(self, file_ids):
1909
"""Similar to __getitem__, but lets you query for multiple.
1911
The returned order is undefined. And currently if an item doesn't
1912
exist, it isn't included in the output.
1916
for file_id in file_ids:
1917
entry = self._fileid_to_entry_cache.get(file_id, None)
1919
remaining.append(file_id)
1921
result.append(entry)
1922
file_keys = [StaticTuple(f,).intern() for f in remaining]
1923
for file_key, value in self.id_to_entry.iteritems(file_keys):
1924
entry = self._bytes_to_entry(value)
1925
result.append(entry)
1926
self._fileid_to_entry_cache[entry.file_id] = entry
1929
def has_id(self, file_id):
1930
# Perhaps have an explicit 'contains' method on CHKMap ?
1931
if self._fileid_to_entry_cache.get(file_id, None) is not None:
1934
self.id_to_entry.iteritems([StaticTuple(file_id,)]))) == 1
1936
def is_root(self, file_id):
1937
return file_id == self.root_id
1939
def _iter_file_id_parents(self, file_id):
1940
"""Yield the parents of file_id up to the root."""
1941
while file_id is not None:
1945
raise errors.NoSuchId(tree=self, file_id=file_id)
1947
file_id = ie.parent_id
1950
"""Iterate over all file-ids."""
1951
for key, _ in self.id_to_entry.iteritems():
1954
def iter_just_entries(self):
1955
"""Iterate over all entries.
1957
Unlike iter_entries(), just the entries are returned (not (path, ie))
1958
and the order of entries is undefined.
1960
XXX: We may not want to merge this into bzr.dev.
1962
for key, entry in self.id_to_entry.iteritems():
1964
ie = self._fileid_to_entry_cache.get(file_id, None)
1966
ie = self._bytes_to_entry(entry)
1967
self._fileid_to_entry_cache[file_id] = ie
1970
def _preload_cache(self):
1971
"""Make sure all file-ids are in _fileid_to_entry_cache"""
1972
if self._fully_cached:
1973
return # No need to do it again
1974
# The optimal sort order is to use iteritems() directly
1975
cache = self._fileid_to_entry_cache
1976
for key, entry in self.id_to_entry.iteritems():
1978
if file_id not in cache:
1979
ie = self._bytes_to_entry(entry)
1983
last_parent_id = last_parent_ie = None
1984
pid_items = self.parent_id_basename_to_file_id.iteritems()
1985
for key, child_file_id in pid_items:
1986
if key == ('', ''): # This is the root
1987
if child_file_id != self.root_id:
1988
raise ValueError('Data inconsistency detected.'
1989
' We expected data with key ("","") to match'
1990
' the root id, but %s != %s'
1991
% (child_file_id, self.root_id))
1993
parent_id, basename = key
1994
ie = cache[child_file_id]
1995
if parent_id == last_parent_id:
1996
parent_ie = last_parent_ie
1998
parent_ie = cache[parent_id]
1999
if parent_ie.kind != 'directory':
2000
raise ValueError('Data inconsistency detected.'
2001
' An entry in the parent_id_basename_to_file_id map'
2002
' has parent_id {%s} but the kind of that object'
2003
' is %r not "directory"' % (parent_id, parent_ie.kind))
2004
if parent_ie._children is None:
2005
parent_ie._children = {}
2006
basename = basename.decode('utf-8')
2007
if basename in parent_ie._children:
2008
existing_ie = parent_ie._children[basename]
2009
if existing_ie != ie:
2010
raise ValueError('Data inconsistency detected.'
2011
' Two entries with basename %r were found'
2012
' in the parent entry {%s}'
2013
% (basename, parent_id))
2014
if basename != ie.name:
2015
raise ValueError('Data inconsistency detected.'
2016
' In the parent_id_basename_to_file_id map, file_id'
2017
' {%s} is listed as having basename %r, but in the'
2018
' id_to_entry map it is %r'
2019
% (child_file_id, basename, ie.name))
2020
parent_ie._children[basename] = ie
2021
self._fully_cached = True
2023
def iter_changes(self, basis):
2024
"""Generate a Tree.iter_changes change list between this and basis.
2026
:param basis: Another CHKInventory.
2027
:return: An iterator over the changes between self and basis, as per
2028
tree.iter_changes().
2030
# We want: (file_id, (path_in_source, path_in_target),
2031
# changed_content, versioned, parent, name, kind,
2033
for key, basis_value, self_value in \
2034
self.id_to_entry.iter_changes(basis.id_to_entry):
2036
if basis_value is not None:
2037
basis_entry = basis._bytes_to_entry(basis_value)
2038
path_in_source = basis.id2path(file_id)
2039
basis_parent = basis_entry.parent_id
2040
basis_name = basis_entry.name
2041
basis_executable = basis_entry.executable
2043
path_in_source = None
2046
basis_executable = None
2047
if self_value is not None:
2048
self_entry = self._bytes_to_entry(self_value)
2049
path_in_target = self.id2path(file_id)
2050
self_parent = self_entry.parent_id
2051
self_name = self_entry.name
2052
self_executable = self_entry.executable
2054
path_in_target = None
2057
self_executable = None
2058
if basis_value is None:
2060
kind = (None, self_entry.kind)
2061
versioned = (False, True)
2062
elif self_value is None:
2064
kind = (basis_entry.kind, None)
2065
versioned = (True, False)
2067
kind = (basis_entry.kind, self_entry.kind)
2068
versioned = (True, True)
2069
changed_content = False
2070
if kind[0] != kind[1]:
2071
changed_content = True
2072
elif kind[0] == 'file':
2073
if (self_entry.text_size != basis_entry.text_size or
2074
self_entry.text_sha1 != basis_entry.text_sha1):
2075
changed_content = True
2076
elif kind[0] == 'symlink':
2077
if self_entry.symlink_target != basis_entry.symlink_target:
2078
changed_content = True
2079
elif kind[0] == 'tree-reference':
2080
if (self_entry.reference_revision !=
2081
basis_entry.reference_revision):
2082
changed_content = True
2083
parent = (basis_parent, self_parent)
2084
name = (basis_name, self_name)
2085
executable = (basis_executable, self_executable)
2086
if (not changed_content
2087
and parent[0] == parent[1]
2088
and name[0] == name[1]
2089
and executable[0] == executable[1]):
2090
# Could happen when only the revision changed for a directory
2093
yield (file_id, (path_in_source, path_in_target), changed_content,
2094
versioned, parent, name, kind, executable)
2097
"""Return the number of entries in the inventory."""
2098
return len(self.id_to_entry)
2100
def _make_delta(self, old):
2101
"""Make an inventory delta from two inventories."""
2102
if type(old) != CHKInventory:
2103
return CommonInventory._make_delta(self, old)
2105
for key, old_value, self_value in \
2106
self.id_to_entry.iter_changes(old.id_to_entry):
2108
if old_value is not None:
2109
old_path = old.id2path(file_id)
2112
if self_value is not None:
2113
entry = self._bytes_to_entry(self_value)
2114
self._fileid_to_entry_cache[file_id] = entry
2115
new_path = self.id2path(file_id)
2119
delta.append((old_path, new_path, file_id, entry))
2122
def path2id(self, relpath):
2123
"""See CommonInventory.path2id()."""
2124
# TODO: perhaps support negative hits?
2125
result = self._path_to_fileid_cache.get(relpath, None)
2126
if result is not None:
2128
if isinstance(relpath, basestring):
2129
names = osutils.splitpath(relpath)
2132
current_id = self.root_id
2133
if current_id is None:
2135
parent_id_index = self.parent_id_basename_to_file_id
2137
for basename in names:
2138
if cur_path is None:
2141
cur_path = cur_path + '/' + basename
2142
basename_utf8 = basename.encode('utf8')
2143
file_id = self._path_to_fileid_cache.get(cur_path, None)
2145
key_filter = [StaticTuple(current_id, basename_utf8)]
2146
items = parent_id_index.iteritems(key_filter)
2147
for (parent_id, name_utf8), file_id in items:
2148
if parent_id != current_id or name_utf8 != basename_utf8:
2149
raise errors.BzrError("corrupt inventory lookup! "
2150
"%r %r %r %r" % (parent_id, current_id, name_utf8,
2155
self._path_to_fileid_cache[cur_path] = file_id
2156
current_id = file_id
2160
"""Serialise the inventory to lines."""
2161
lines = ["chkinventory:\n"]
2162
if self._search_key_name != 'plain':
2163
# custom ordering grouping things that don't change together
2164
lines.append('search_key_name: %s\n' % (self._search_key_name,))
2165
lines.append("root_id: %s\n" % self.root_id)
2166
lines.append('parent_id_basename_to_file_id: %s\n' %
2167
(self.parent_id_basename_to_file_id.key()[0],))
2168
lines.append("revision_id: %s\n" % self.revision_id)
2169
lines.append("id_to_entry: %s\n" % (self.id_to_entry.key()[0],))
2171
lines.append("revision_id: %s\n" % self.revision_id)
2172
lines.append("root_id: %s\n" % self.root_id)
2173
if self.parent_id_basename_to_file_id is not None:
2174
lines.append('parent_id_basename_to_file_id: %s\n' %
2175
(self.parent_id_basename_to_file_id.key()[0],))
2176
lines.append("id_to_entry: %s\n" % (self.id_to_entry.key()[0],))
2181
"""Get the root entry."""
2182
return self[self.root_id]
2185
class CHKInventoryDirectory(InventoryDirectory):
2186
"""A directory in an inventory."""
2188
__slots__ = ['_children', '_chk_inventory']
2190
def __init__(self, file_id, name, parent_id, chk_inventory):
2191
# Don't call InventoryDirectory.__init__ - it isn't right for this
2193
InventoryEntry.__init__(self, file_id, name, parent_id)
2194
self._children = None
2195
self._chk_inventory = chk_inventory
2199
"""Access the list of children of this directory.
2201
With a parent_id_basename_to_file_id index, loads all the children,
2202
without loads the entire index. Without is bad. A more sophisticated
2203
proxy object might be nice, to allow partial loading of children as
2204
well when specific names are accessed. (So path traversal can be
2205
written in the obvious way but not examine siblings.).
2207
if self._children is not None:
2208
return self._children
2209
# No longer supported
2210
if self._chk_inventory.parent_id_basename_to_file_id is None:
2211
raise AssertionError("Inventories without"
2212
" parent_id_basename_to_file_id are no longer supported")
2214
# XXX: Todo - use proxy objects for the children rather than loading
2215
# all when the attribute is referenced.
2216
parent_id_index = self._chk_inventory.parent_id_basename_to_file_id
2218
for (parent_id, name_utf8), file_id in parent_id_index.iteritems(
2219
key_filter=[StaticTuple(self.file_id,)]):
2220
child_keys.add(StaticTuple(file_id,))
2222
for file_id_key in child_keys:
2223
entry = self._chk_inventory._fileid_to_entry_cache.get(
2224
file_id_key[0], None)
2225
if entry is not None:
2226
result[entry.name] = entry
2227
cached.add(file_id_key)
2228
child_keys.difference_update(cached)
2229
# populate; todo: do by name
2230
id_to_entry = self._chk_inventory.id_to_entry
2231
for file_id_key, bytes in id_to_entry.iteritems(child_keys):
2232
entry = self._chk_inventory._bytes_to_entry(bytes)
2233
result[entry.name] = entry
2234
self._chk_inventory._fileid_to_entry_cache[file_id_key[0]] = entry
2235
self._children = result
1291
2238
entry_factory = {
1292
2239
'directory': InventoryDirectory,
1293
2240
'file': InventoryFile,
added
removed
bzrlib/inventory.py
