27
27
# created, but it's not for now.
28
28
ROOT_ID = "TREE_ROOT"
34
from bzrlib.lazy_import import lazy_import
35
lazy_import(globals(), """
37
from warnings import warn
40
from bzrlib import errors, osutils
41
from bzrlib.osutils import (pumpfile, quotefn, splitpath, joinpath,
42
pathjoin, sha_strings)
43
from bzrlib.errors import (NotVersionedError, InvalidEntryName,
44
BzrError, BzrCheckError, BinaryFile)
49
from bzrlib.errors import (
53
from bzrlib.symbol_versioning import deprecated_in, deprecated_method
45
54
from bzrlib.trace import mutter
82
91
InventoryDirectory('123', 'src', parent_id='TREE_ROOT', revision=None)
83
92
>>> i.add(InventoryFile('2323', 'hello.c', parent_id='123'))
84
93
InventoryFile('2323', 'hello.c', parent_id='123', sha1=None, len=None)
85
>>> shouldbe = {0: '', 1: 'src', 2: pathjoin('src','hello.c')}
94
>>> shouldbe = {0: '', 1: 'src', 2: 'src/hello.c'}
86
95
>>> for ix, j in enumerate(i.iter_entries()):
87
96
... print (j[0] == shouldbe[ix], j[1])
89
(True, InventoryDirectory('TREE_ROOT', '', parent_id=None, revision=None))
98
(True, InventoryDirectory('TREE_ROOT', u'', parent_id=None, revision=None))
90
99
(True, InventoryDirectory('123', 'src', parent_id='TREE_ROOT', revision=None))
91
100
(True, InventoryFile('2323', 'hello.c', parent_id='123', sha1=None, len=None))
92
>>> i.add(InventoryFile('2323', 'bye.c', '123'))
93
Traceback (most recent call last):
95
BzrError: inventory already contains entry with id {2323}
96
101
>>> i.add(InventoryFile('2324', 'bye.c', '123'))
97
102
InventoryFile('2324', 'bye.c', parent_id='123', sha1=None, len=None)
98
103
>>> i.add(InventoryDirectory('2325', 'wibble', '123'))
138
142
return False, False
140
def diff(self, text_diff, from_label, tree, to_label, to_entry, to_tree,
141
output_to, reverse=False):
142
"""Perform a diff from this to to_entry.
144
text_diff will be used for textual difference calculation.
145
This is a template method, override _diff in child classes.
147
self._read_tree_state(tree.id2path(self.file_id), tree)
149
# cannot diff from one kind to another - you must do a removal
150
# and an addif they do not match.
151
assert self.kind == to_entry.kind
152
to_entry._read_tree_state(to_tree.id2path(to_entry.file_id),
154
self._diff(text_diff, from_label, tree, to_label, to_entry, to_tree,
157
144
def _diff(self, text_diff, from_label, tree, to_label, to_entry, to_tree,
158
145
output_to, reverse=False):
159
146
"""Perform a diff between two entries of the same kind."""
161
def find_previous_heads(self, previous_inventories,
162
versioned_file_store,
165
"""Return the revisions and entries that directly precede this.
167
Returned as a map from revision to inventory entry.
169
This is a map containing the file revisions in all parents
170
for which the file exists, and its revision is not a parent of
171
any other. If the file is new, the set will be empty.
173
:param versioned_file_store: A store where ancestry data on this
174
file id can be queried.
175
:param transaction: The transaction that queries to the versioned
176
file store should be completed under.
177
:param entry_vf: The entry versioned file, if its already available.
148
def parent_candidates(self, previous_inventories):
149
"""Find possible per-file graph parents.
151
This is currently defined by:
152
- Select the last changed revision in the parent inventory.
153
- Do deal with a short lived bug in bzr 0.8's development two entries
154
that have the same last changed but different 'x' bit settings are
179
def get_ancestors(weave, entry):
180
return set(weave.get_ancestry(entry.revision))
181
157
# revision:ie mapping for each ie found in previous_inventories.
183
# revision:ie mapping with one revision for each head.
185
# revision: ancestor list for each head
187
159
# identify candidate head revision ids.
188
160
for inv in previous_inventories:
189
161
if self.file_id in inv:
190
162
ie = inv[self.file_id]
191
assert ie.file_id == self.file_id
192
163
if ie.revision in candidates:
193
164
# same revision value in two different inventories:
194
165
# correct possible inconsistencies:
200
171
ie.executable = False
201
172
except AttributeError:
203
# must now be the same.
204
assert candidates[ie.revision] == ie
206
175
# add this revision as a candidate.
207
176
candidates[ie.revision] = ie
209
# common case optimisation
210
if len(candidates) == 1:
211
# if there is only one candidate revision found
212
# then we can opening the versioned file to access ancestry:
213
# there cannot be any ancestors to eliminate when there is
214
# only one revision available.
215
heads[ie.revision] = ie
218
# eliminate ancestors amongst the available candidates:
219
# heads are those that are not an ancestor of any other candidate
220
# - this provides convergence at a per-file level.
221
for ie in candidates.values():
222
# may be an ancestor of a known head:
223
already_present = 0 != len(
224
[head for head in heads
225
if ie.revision in head_ancestors[head]])
227
# an ancestor of an analyzed candidate.
229
# not an ancestor of a known head:
230
# load the versioned file for this file id if needed
232
entry_vf = versioned_file_store.get_weave_or_empty(
233
self.file_id, transaction)
234
ancestors = get_ancestors(entry_vf, ie)
235
# may knock something else out:
236
check_heads = list(heads.keys())
237
for head in check_heads:
238
if head in ancestors:
239
# this previously discovered 'head' is not
240
# really a head - its an ancestor of the newly
243
head_ancestors[ie.revision] = ancestors
244
heads[ie.revision] = ie
179
@deprecated_method(deprecated_in((1, 6, 0)))
247
180
def get_tar_item(self, root, dp, now, tree):
248
181
"""Get a tarfile item and a file stream for its content."""
249
item = tarfile.TarInfo(pathjoin(root, dp))
182
item = tarfile.TarInfo(osutils.pathjoin(root, dp).encode('utf8'))
250
183
# TODO: would be cool to actually set it to the timestamp of the
251
184
# revision it was last changed
306
239
raise BzrError("don't know how to export {%s} of kind %r" %
307
240
(self.file_id, self.kind))
242
@deprecated_method(deprecated_in((1, 6, 0)))
309
243
def put_on_disk(self, dest, dp, tree):
310
244
"""Create a representation of self on disk in the prefix dest.
312
246
This is a template method - implement _put_on_disk in subclasses.
314
fullpath = pathjoin(dest, dp)
248
fullpath = osutils.pathjoin(dest, dp)
315
249
self._put_on_disk(fullpath, tree)
316
250
# mutter(" export {%s} kind %s to %s", self.file_id,
317
251
# self.kind, fullpath)
407
def snapshot(self, revision, path, previous_entries,
408
work_tree, commit_builder):
409
"""Make a snapshot of this entry which may or may not have changed.
411
This means that all its fields are populated, that it has its
412
text stored in the text store or weave.
414
# mutter('new parents of %s are %r', path, previous_entries)
415
self._read_tree_state(path, work_tree)
416
# TODO: Where should we determine whether to reuse a
417
# previous revision id or create a new revision? 20060606
418
if len(previous_entries) == 1:
419
# cannot be unchanged unless there is only one parent file rev.
420
parent_ie = previous_entries.values()[0]
421
if self._unchanged(parent_ie):
422
# mutter("found unchanged entry")
423
self.revision = parent_ie.revision
425
return self._snapshot_into_revision(revision, previous_entries,
426
work_tree, commit_builder)
428
def _snapshot_into_revision(self, revision, previous_entries, work_tree,
430
"""Record this revision unconditionally into a store.
432
The entry's last-changed revision property (`revision`) is updated to
433
that of the new revision.
435
:param revision: id of the new revision that is being recorded.
437
:returns: String description of the commit (e.g. "merged", "modified"), etc.
439
# mutter('new revision {%s} for {%s}', revision, self.file_id)
440
self.revision = revision
441
self._snapshot_text(previous_entries, work_tree, commit_builder)
443
def _snapshot_text(self, file_parents, work_tree, commit_builder):
444
"""Record the 'text' of this entry, whatever form that takes.
446
This default implementation simply adds an empty text.
448
raise NotImplementedError(self._snapshot_text)
450
343
def __eq__(self, other):
451
344
if not isinstance(other, InventoryEntry):
452
345
return NotImplemented
501
397
class RootEntry(InventoryEntry):
503
399
__slots__ = ['text_sha1', 'text_size', 'file_id', 'name', 'kind',
504
'text_id', 'parent_id', 'children', 'executable',
505
'revision', 'symlink_target']
400
'text_id', 'parent_id', 'children', 'executable',
401
'revision', 'symlink_target', 'reference_revision']
507
403
def _check(self, checker, rev_id, tree):
508
404
"""See InventoryEntry._check"""
530
426
"""A directory in an inventory."""
532
428
__slots__ = ['text_sha1', 'text_size', 'file_id', 'name', 'kind',
533
'text_id', 'parent_id', 'children', 'executable',
534
'revision', 'symlink_target']
429
'text_id', 'parent_id', 'children', 'executable',
430
'revision', 'symlink_target', 'reference_revision']
536
432
def _check(self, checker, rev_id, tree):
537
433
"""See InventoryEntry._check"""
568
464
"""See InventoryEntry._put_on_disk."""
569
465
os.mkdir(fullpath)
571
def _snapshot_text(self, file_parents, work_tree, commit_builder):
572
"""See InventoryEntry._snapshot_text."""
573
commit_builder.modified_directory(self.file_id, file_parents)
576
468
class InventoryFile(InventoryEntry):
577
469
"""A file in an inventory."""
579
471
__slots__ = ['text_sha1', 'text_size', 'file_id', 'name', 'kind',
580
'text_id', 'parent_id', 'children', 'executable',
581
'revision', 'symlink_target']
472
'text_id', 'parent_id', 'children', 'executable',
473
'revision', 'symlink_target', 'reference_revision']
583
475
def _check(self, checker, tree_revision_id, tree):
584
476
"""See InventoryEntry._check"""
585
t = (self.file_id, self.revision)
586
if t in checker.checked_texts:
587
prev_sha = checker.checked_texts[t]
477
key = (self.file_id, self.revision)
478
if key in checker.checked_texts:
479
prev_sha = checker.checked_texts[key]
588
480
if prev_sha != self.text_sha1:
589
raise BzrCheckError('mismatched sha1 on {%s} in {%s}' %
590
(self.file_id, tree_revision_id))
482
'mismatched sha1 on {%s} in {%s} (%s != %s) %r' %
483
(self.file_id, tree_revision_id, prev_sha, self.text_sha1,
592
486
checker.repeated_text_cnt += 1
595
if self.file_id not in checker.checked_weaves:
596
mutter('check weave {%s}', self.file_id)
597
w = tree.get_weave(self.file_id)
598
# Not passing a progress bar, because it creates a new
599
# progress, which overwrites the current progress,
600
# and doesn't look nice
602
checker.checked_weaves[self.file_id] = True
604
w = tree.get_weave(self.file_id)
606
489
mutter('check version {%s} of {%s}', tree_revision_id, self.file_id)
607
490
checker.checked_text_cnt += 1
608
491
# We can't check the length, because Weave doesn't store that
609
492
# information, and the whole point of looking at the weave's
610
493
# sha1sum is that we don't have to extract the text.
611
if self.text_sha1 != w.get_sha1(self.revision):
612
raise BzrCheckError('text {%s} version {%s} wrong sha1'
613
% (self.file_id, self.revision))
614
checker.checked_texts[t] = self.text_sha1
494
if (self.text_sha1 != tree._repository.texts.get_sha1s([key])[key]):
495
raise BzrCheckError('text {%s} version {%s} wrong sha1' % key)
496
checker.checked_texts[key] = self.text_sha1
617
499
other = InventoryFile(self.file_id, self.name, self.parent_id)
633
513
def _diff(self, text_diff, from_label, tree, to_label, to_entry, to_tree,
634
514
output_to, reverse=False):
635
515
"""See InventoryEntry._diff."""
637
from_text = tree.get_file(self.file_id).readlines()
639
to_text = to_tree.get_file(to_entry.file_id).readlines()
643
text_diff(from_label, from_text,
644
to_label, to_text, output_to)
646
text_diff(to_label, to_text,
647
from_label, from_text, output_to)
650
label_pair = (to_label, from_label)
652
label_pair = (from_label, to_label)
653
print >> output_to, "Binary files %s and %s differ" % label_pair
516
from bzrlib.diff import DiffText
517
from_file_id = self.file_id
519
to_file_id = to_entry.file_id
523
to_file_id, from_file_id = from_file_id, to_file_id
524
tree, to_tree = to_tree, tree
525
from_label, to_label = to_label, from_label
526
differ = DiffText(tree, to_tree, output_to, 'utf-8', '', '',
528
return differ.diff_text(from_file_id, to_file_id, from_label, to_label)
655
530
def has_text(self):
656
531
"""See InventoryEntry.has_text."""
725
593
"""A file in an inventory."""
727
595
__slots__ = ['text_sha1', 'text_size', 'file_id', 'name', 'kind',
728
'text_id', 'parent_id', 'children', 'executable',
729
'revision', 'symlink_target']
596
'text_id', 'parent_id', 'children', 'executable',
597
'revision', 'symlink_target', 'reference_revision']
731
599
def _check(self, checker, rev_id, tree):
732
600
"""See InventoryEntry._check"""
755
623
def _diff(self, text_diff, from_label, tree, to_label, to_entry, to_tree,
756
624
output_to, reverse=False):
757
625
"""See InventoryEntry._diff."""
758
from_text = self.symlink_target
626
from bzrlib.diff import DiffSymlink
627
old_target = self.symlink_target
759
628
if to_entry is not None:
760
to_text = to_entry.symlink_target
765
print >>output_to, '=== target changed %r => %r' % (from_text, to_text)
768
print >>output_to, '=== target was %r' % self.symlink_target
770
print >>output_to, '=== target is %r' % self.symlink_target
629
new_target = to_entry.symlink_target
638
new_target, old_target = old_target, new_target
639
differ = DiffSymlink(old_tree, new_tree, output_to)
640
return differ.diff_symlink(old_target, new_target)
772
642
def __init__(self, file_id, name, parent_id):
773
643
super(InventoryLink, self).__init__(file_id, name, parent_id)
807
677
compatible = False
808
678
return compatible
810
def _snapshot_text(self, file_parents, work_tree, commit_builder):
811
"""See InventoryEntry._snapshot_text."""
812
commit_builder.modified_link(
813
self.file_id, file_parents, self.symlink_target)
681
class TreeReference(InventoryEntry):
683
kind = 'tree-reference'
685
def __init__(self, file_id, name, parent_id, revision=None,
686
reference_revision=None):
687
InventoryEntry.__init__(self, file_id, name, parent_id)
688
self.revision = revision
689
self.reference_revision = reference_revision
692
return TreeReference(self.file_id, self.name, self.parent_id,
693
self.revision, self.reference_revision)
695
def _read_tree_state(self, path, work_tree):
696
"""Populate fields in the inventory entry from the given tree.
698
self.reference_revision = work_tree.get_reference_revision(
701
def _forget_tree_state(self):
702
self.reference_revision = None
704
def _unchanged(self, previous_ie):
705
"""See InventoryEntry._unchanged."""
706
compatible = super(TreeReference, self)._unchanged(previous_ie)
707
if self.reference_revision != previous_ie.reference_revision:
816
712
class Inventory(object):
861
760
The inventory is created with a default root directory, with
864
# We are letting Branch.create() create a unique inventory
865
# root id. Rather than generating a random one here.
867
# root_id = bzrlib.branch.gen_file_id('TREE_ROOT')
868
763
if root_id is not None:
869
self._set_root(InventoryDirectory(root_id, '', None))
764
self._set_root(InventoryDirectory(root_id, u'', None))
873
# FIXME: this isn't ever used, changing it to self.revision may break
874
# things. TODO make everything use self.revision_id
875
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.")
877
843
def _set_root(self, ie):
879
845
self._byid = {self.root.file_id: self.root}
946
917
lexicographically sorted order, and is a hybrid between
947
918
depth-first and breadth-first.
920
:param yield_parents: If True, yield the parents from the root leading
921
down to specific_file_ids that have been requested. This has no
922
impact if specific_file_ids is None.
949
923
:return: This yields (path, entry) pairs
925
if specific_file_ids and not isinstance(specific_file_ids, set):
926
specific_file_ids = set(specific_file_ids)
951
927
# TODO? Perhaps this should return the from_dir so that the root is
952
928
# yielded? or maybe an option?
953
929
if from_dir is None:
930
if self.root is None:
932
# Optimize a common case
933
if (not yield_parents and specific_file_ids is not None and
934
len(specific_file_ids) == 1):
935
file_id = list(specific_file_ids)[0]
937
yield self.id2path(file_id), self[file_id]
955
939
from_dir = self.root
940
if (specific_file_ids is None or yield_parents or
941
self.root.file_id in specific_file_ids):
957
943
elif isinstance(from_dir, basestring):
958
944
from_dir = self._byid[from_dir]
946
if specific_file_ids is not None:
947
# TODO: jam 20070302 This could really be done as a loop rather
948
# than a bunch of recursive calls.
951
def add_ancestors(file_id):
952
if file_id not in byid:
954
parent_id = byid[file_id].parent_id
955
if parent_id is None:
957
if parent_id not in parents:
958
parents.add(parent_id)
959
add_ancestors(parent_id)
960
for file_id in specific_file_ids:
961
add_ancestors(file_id)
960
965
stack = [(u'', from_dir)]
967
972
child_relpath = cur_relpath + child_name
969
yield child_relpath, child_ie
974
if (specific_file_ids is None or
975
child_ie.file_id in specific_file_ids or
976
(yield_parents and child_ie.file_id in parents)):
977
yield child_relpath, child_ie
971
979
if child_ie.kind == 'directory':
972
child_dirs.append((child_relpath+'/', child_ie))
980
if parents is None or child_ie.file_id in parents:
981
child_dirs.append((child_relpath+'/', child_ie))
973
982
stack.extend(reversed(child_dirs))
984
def make_entry(self, kind, name, parent_id, file_id=None):
985
"""Simple thunk to bzrlib.inventory.make_entry."""
986
return make_entry(kind, name, parent_id, file_id)
975
988
def entries(self):
976
989
"""Return list of (path, ie) for all entries except the root.
1051
1072
Returns the new entry object.
1053
1074
if entry.file_id in self._byid:
1054
raise BzrError("inventory already contains entry with id {%s}" % entry.file_id)
1075
raise errors.DuplicateFileId(entry.file_id,
1076
self._byid[entry.file_id])
1056
1078
if entry.parent_id is None:
1057
assert self.root is None and len(self._byid) == 0
1058
self._set_root(entry)
1060
if entry.parent_id == ROOT_ID:
1061
assert self.root is not None, self
1062
entry.parent_id = self.root.file_id
1065
parent = self._byid[entry.parent_id]
1067
raise BzrError("parent_id {%s} not in inventory" % entry.parent_id)
1069
if entry.name in parent.children:
1070
raise BzrError("%s is already versioned" %
1071
pathjoin(self.id2path(parent.file_id), entry.name))
1073
self._byid[entry.file_id] = entry
1074
parent.children[entry.name] = entry
1082
parent = self._byid[entry.parent_id]
1084
raise BzrError("parent_id {%s} not in inventory" %
1087
if entry.name in parent.children:
1088
raise BzrError("%s is already versioned" %
1089
osutils.pathjoin(self.id2path(parent.file_id),
1090
entry.name).encode('utf-8'))
1091
parent.children[entry.name] = entry
1092
return self._add_child(entry)
1077
1094
def add_path(self, relpath, kind, file_id=None, parent_id=None):
1078
1095
"""Add entry from a path.
1232
1248
for file_id in reversed(to_delete):
1233
1249
ie = self[file_id]
1234
1250
del self._byid[file_id]
1235
if ie.parent_id is not None:
1236
del self[ie.parent_id].children[ie.name]
1251
if ie.parent_id is not None:
1252
del self[ie.parent_id].children[ie.name]
1238
1256
def rename(self, file_id, new_parent_id, new_name):
1239
1257
"""Move a file within the inventory.
1241
1259
This can change either the name, or the parent, or both.
1243
This does not move the working file."""
1261
This does not move the working file.
1263
new_name = ensure_normalized_name(new_name)
1244
1264
if not is_valid_name(new_name):
1245
1265
raise BzrError("not an acceptable filename: %r" % new_name)
1274
1304
:param file_id: the file_id to use. if None, one will be created.
1276
1306
if file_id is None:
1277
file_id = bzrlib.workingtree.gen_file_id(name)
1307
file_id = generate_ids.gen_file_id(name)
1308
name = ensure_normalized_name(name)
1310
factory = entry_factory[kind]
1312
raise BzrError("unknown kind %r" % kind)
1313
return factory(file_id, name, parent_id)
1316
def ensure_normalized_name(name):
1319
:raises InvalidNormalization: When name is not normalized, and cannot be
1320
accessed on this platform by the normalized path.
1321
:return: The NFC normalised version of name.
1323
#------- This has been copied to bzrlib.dirstate.DirState.add, please
1324
# keep them synchronised.
1325
# we dont import normalized_filename directly because we want to be
1326
# able to change the implementation at runtime for tests.
1279
1327
norm_name, can_access = osutils.normalized_filename(name)
1280
1328
if norm_name != name:
1284
1332
# TODO: jam 20060701 This would probably be more useful
1285
1333
# if the error was raised with the full path
1286
1334
raise errors.InvalidNormalization(name)
1288
if kind == 'directory':
1289
return InventoryDirectory(file_id, name, parent_id)
1290
elif kind == 'file':
1291
return InventoryFile(file_id, name, parent_id)
1292
elif kind == 'symlink':
1293
return InventoryLink(file_id, name, parent_id)
1295
raise BzrError("unknown kind %r" % kind)
1298
1338
_NAME_RE = None