68
82
sha-1 of the text of the file
71
85
size in bytes of the text of the file
73
87
(reading a version 4 tree created a text_id field.)
75
89
>>> i = Inventory()
78
92
>>> i.add(InventoryDirectory('123', 'src', ROOT_ID))
79
InventoryDirectory('123', 'src', parent_id='TREE_ROOT')
93
InventoryDirectory('123', 'src', parent_id='TREE_ROOT', revision=None)
80
94
>>> i.add(InventoryFile('2323', 'hello.c', parent_id='123'))
81
InventoryFile('2323', 'hello.c', parent_id='123')
82
>>> shouldbe = {0: 'src', 1: os.path.join('src','hello.c')}
95
InventoryFile('2323', 'hello.c', parent_id='123', sha1=None, len=None, revision=None)
96
>>> shouldbe = {0: '', 1: 'src', 2: 'src/hello.c'}
83
97
>>> for ix, j in enumerate(i.iter_entries()):
84
98
... print (j[0] == shouldbe[ix], j[1])
86
(True, InventoryDirectory('123', 'src', parent_id='TREE_ROOT'))
87
(True, InventoryFile('2323', 'hello.c', parent_id='123'))
88
>>> i.add(InventoryFile('2323', 'bye.c', '123'))
89
Traceback (most recent call last):
91
BzrError: inventory already contains entry with id {2323}
100
(True, InventoryDirectory('TREE_ROOT', u'', parent_id=None, revision=None))
101
(True, InventoryDirectory('123', 'src', parent_id='TREE_ROOT', revision=None))
102
(True, InventoryFile('2323', 'hello.c', parent_id='123', sha1=None, len=None, revision=None))
92
103
>>> i.add(InventoryFile('2324', 'bye.c', '123'))
93
InventoryFile('2324', 'bye.c', parent_id='123')
104
InventoryFile('2324', 'bye.c', parent_id='123', sha1=None, len=None, revision=None)
94
105
>>> i.add(InventoryDirectory('2325', 'wibble', '123'))
95
InventoryDirectory('2325', 'wibble', parent_id='123')
106
InventoryDirectory('2325', 'wibble', parent_id='123', revision=None)
96
107
>>> i.path2id('src/wibble')
100
109
>>> i.add(InventoryFile('2326', 'wibble.c', '2325'))
101
InventoryFile('2326', 'wibble.c', parent_id='2325')
110
InventoryFile('2326', 'wibble.c', parent_id='2325', sha1=None, len=None, revision=None)
103
InventoryFile('2326', 'wibble.c', parent_id='2325')
112
InventoryFile('2326', 'wibble.c', parent_id='2325', sha1=None, len=None, revision=None)
104
113
>>> for path, entry in i.iter_entries():
105
... print path.replace('\\\\', '/') # for win32 os.sep
106
... assert i.path2id(path)
112
121
src/wibble/wibble.c
113
>>> i.id2path('2326').replace('\\\\', '/')
122
>>> i.id2path('2326')
114
123
'src/wibble/wibble.c'
117
__slots__ = ['text_sha1', 'text_size', 'file_id', 'name', 'kind',
118
'text_id', 'parent_id', 'children', 'executable',
121
def _add_text_to_weave(self, new_lines, parents, weave_store, transaction):
122
weave_store.add_text(self.file_id, self.revision, new_lines, parents,
126
# Constants returned by describe_change()
128
# TODO: These should probably move to some kind of FileChangeDescription
129
# class; that's like what's inside a TreeDelta but we want to be able to
130
# generate them just for one file at a time.
132
MODIFIED_AND_RENAMED = 'modified and renamed'
134
__slots__ = ['file_id', 'revision', 'parent_id', 'name']
136
# Attributes that all InventoryEntry instances are expected to have, but
137
# that don't vary for all kinds of entry. (e.g. symlink_target is only
138
# relevant to InventoryLink, so there's no reason to make every
139
# InventoryFile instance allocate space to hold a value for it.)
140
# Attributes that only vary for files: executable, text_sha1, text_size,
146
# Attributes that only vary for symlinks: symlink_target
147
symlink_target = None
148
# Attributes that only vary for tree-references: reference_revision
149
reference_revision = None
125
152
def detect_changes(self, old_entry):
126
153
"""Return a (text_modified, meta_modified) from this to old_entry.
128
_read_tree_state must have been called on self and old_entry prior to
155
_read_tree_state must have been called on self and old_entry prior to
129
156
calling detect_changes.
131
158
return False, False
133
def diff(self, text_diff, from_label, tree, to_label, to_entry, to_tree,
134
output_to, reverse=False):
135
"""Perform a diff from this to to_entry.
137
text_diff will be used for textual difference calculation.
138
This is a template method, override _diff in child classes.
140
self._read_tree_state(tree.id2path(self.file_id), tree)
142
# cannot diff from one kind to another - you must do a removal
143
# and an addif they do not match.
144
assert self.kind == to_entry.kind
145
to_entry._read_tree_state(to_tree.id2path(to_entry.file_id),
147
self._diff(text_diff, from_label, tree, to_label, to_entry, to_tree,
150
160
def _diff(self, text_diff, from_label, tree, to_label, to_entry, to_tree,
151
161
output_to, reverse=False):
152
162
"""Perform a diff between two entries of the same kind."""
154
def find_previous_heads(self, previous_inventories, entry_weave):
155
"""Return the revisions and entries that directly preceed this.
157
Returned as a map from revision to inventory entry.
159
This is a map containing the file revisions in all parents
160
for which the file exists, and its revision is not a parent of
161
any other. If the file is new, the set will be empty.
164
def parent_candidates(self, previous_inventories):
165
"""Find possible per-file graph parents.
167
This is currently defined by:
168
- Select the last changed revision in the parent inventory.
169
- Do deal with a short lived bug in bzr 0.8's development two entries
170
that have the same last changed but different 'x' bit settings are
163
def get_ancestors(weave, entry):
164
return set(map(weave.idx_to_name,
165
weave.inclusions([weave.lookup(entry.revision)])))
173
# revision:ie mapping for each ie found in previous_inventories.
175
# identify candidate head revision ids.
168
176
for inv in previous_inventories:
169
if self.file_id in inv:
177
if inv.has_id(self.file_id):
170
178
ie = inv[self.file_id]
171
assert ie.file_id == self.file_id
172
if ie.revision in heads:
173
# fixup logic, there was a bug in revision updates.
174
# with x bit support.
179
if ie.revision in candidates:
180
# same revision value in two different inventories:
181
# correct possible inconsistencies:
182
# * there was a bug in revision updates with 'x' bit
176
if heads[ie.revision].executable != ie.executable:
177
heads[ie.revision].executable = False
185
if candidates[ie.revision].executable != ie.executable:
186
candidates[ie.revision].executable = False
178
187
ie.executable = False
179
188
except AttributeError:
181
assert heads[ie.revision] == ie
183
# may want to add it.
184
# may already be covered:
185
already_present = 0 != len(
186
[head for head in heads
187
if ie.revision in head_ancestors[head]])
189
# an ancestor of a known head.
192
ancestors = get_ancestors(entry_weave, ie)
193
# may knock something else out:
194
check_heads = list(heads.keys())
195
for head in check_heads:
196
if head in ancestors:
197
# this head is not really a head
199
head_ancestors[ie.revision] = ancestors
200
heads[ie.revision] = ie
203
def get_tar_item(self, root, dp, now, tree):
204
"""Get a tarfile item and a file stream for its content."""
205
item = tarfile.TarInfo(os.path.join(root, dp))
206
# TODO: would be cool to actually set it to the timestamp of the
207
# revision it was last changed
209
fileobj = self._put_in_tar(item, tree)
191
# add this revision as a candidate.
192
candidates[ie.revision] = ie
212
195
def has_text(self):
213
196
"""Return true if the object this entry represents has textual data.
235
218
Traceback (most recent call last):
236
219
InvalidEntryName: Invalid entry name: src/hello.c
238
assert isinstance(name, basestring), name
239
221
if '/' in name or '\\' in name:
240
raise InvalidEntryName(name=name)
241
self.executable = False
222
raise errors.InvalidEntryName(name=name)
223
self.file_id = file_id
242
224
self.revision = None
243
self.text_sha1 = None
244
self.text_size = None
245
self.file_id = file_id
247
self.text_id = text_id
248
226
self.parent_id = parent_id
249
self.symlink_target = None
251
228
def kind_character(self):
252
229
"""Return a short kind indicator useful for appending to names."""
253
raise BzrError('unknown kind %r' % self.kind)
255
known_kinds = ('file', 'directory', 'symlink', 'root_directory')
257
def _put_in_tar(self, item, tree):
258
"""populate item for stashing in a tar, and return the content stream.
260
If no content is available, return None.
262
raise BzrError("don't know how to export {%s} of kind %r" %
263
(self.file_id, self.kind))
265
def put_on_disk(self, dest, dp, tree):
266
"""Create a representation of self on disk in the prefix dest.
268
This is a template method - implement _put_on_disk in subclasses.
270
fullpath = appendpath(dest, dp)
271
self._put_on_disk(fullpath, tree)
272
mutter(" export {%s} kind %s to %s", self.file_id,
275
def _put_on_disk(self, fullpath, tree):
276
"""Put this entry onto disk at fullpath, from tree tree."""
277
raise BzrError("don't know how to export {%s} of kind %r" % (self.file_id, self.kind))
230
raise errors.BzrError('unknown kind %r' % self.kind)
232
known_kinds = ('file', 'directory', 'symlink')
279
234
def sorted_children(self):
280
l = self.children.items()
235
return sorted(self.children.items())
285
238
def versionable_kind(kind):
286
return kind in ('file', 'directory', 'symlink')
239
return (kind in ('file', 'directory', 'symlink', 'tree-reference'))
288
def check(self, checker, rev_id, inv, tree):
241
def check(self, checker, rev_id, inv):
289
242
"""Check this inventory entry is intact.
291
244
This is a template method, override _check for kind specific
247
:param checker: Check object providing context for the checks;
248
can be used to find out what parts of the repository have already
250
:param rev_id: Revision id from which this InventoryEntry was loaded.
251
Not necessarily the last-changed revision for this file.
252
:param inv: Inventory from which the entry was loaded.
294
if self.parent_id != None:
254
if self.parent_id is not None:
295
255
if not inv.has_id(self.parent_id):
296
raise BzrCheckError('missing parent {%s} in inventory for revision {%s}'
297
% (self.parent_id, rev_id))
298
self._check(checker, rev_id, tree)
256
raise errors.BzrCheckError(
257
'missing parent {%s} in inventory for revision {%s}' % (
258
self.parent_id, rev_id))
259
checker._add_entry_to_text_key_references(inv, self)
260
self._check(checker, rev_id)
300
def _check(self, checker, rev_id, tree):
262
def _check(self, checker, rev_id):
301
263
"""Check this inventory entry for kind specific errors."""
302
raise BzrCheckError('unknown entry kind %r in revision {%s}' %
264
checker._report_items.append(
265
'unknown entry kind %r in revision {%s}' % (self.kind, rev_id))
307
268
"""Clone this inventory entry."""
308
269
raise NotImplementedError
310
def _get_snapshot_change(self, previous_entries):
311
if len(previous_entries) > 1:
313
elif len(previous_entries) == 0:
272
def describe_change(old_entry, new_entry):
273
"""Describe the change between old_entry and this.
275
This smells of being an InterInventoryEntry situation, but as its
276
the first one, we're making it a static method for now.
278
An entry with a different parent, or different name is considered
279
to be renamed. Reparenting is an internal detail.
280
Note that renaming the parent does not trigger a rename for the
283
# TODO: Perhaps return an object rather than just a string
284
if old_entry is new_entry:
285
# also the case of both being None
287
elif old_entry is None:
316
return 'modified/renamed/reparented'
289
elif new_entry is None:
291
if old_entry.kind != new_entry.kind:
293
text_modified, meta_modified = new_entry.detect_changes(old_entry)
294
if text_modified or meta_modified:
298
# TODO 20060511 (mbp, rbc) factor out 'detect_rename' here.
299
if old_entry.parent_id != new_entry.parent_id:
301
elif old_entry.name != new_entry.name:
305
if renamed and not modified:
306
return InventoryEntry.RENAMED
307
if modified and not renamed:
309
if modified and renamed:
310
return InventoryEntry.MODIFIED_AND_RENAMED
318
313
def __repr__(self):
319
return ("%s(%r, %r, parent_id=%r)"
314
return ("%s(%r, %r, parent_id=%r, revision=%r)"
320
315
% (self.__class__.__name__,
325
def snapshot(self, revision, path, previous_entries,
326
work_tree, weave_store, transaction):
327
"""Make a snapshot of this entry which may or may not have changed.
329
This means that all its fields are populated, that it has its
330
text stored in the text store or weave.
332
mutter('new parents of %s are %r', path, previous_entries)
333
self._read_tree_state(path, work_tree)
334
if len(previous_entries) == 1:
335
# cannot be unchanged unless there is only one parent file rev.
336
parent_ie = previous_entries.values()[0]
337
if self._unchanged(parent_ie):
338
mutter("found unchanged entry")
339
self.revision = parent_ie.revision
341
return self.snapshot_revision(revision, previous_entries,
342
work_tree, weave_store, transaction)
344
def snapshot_revision(self, revision, previous_entries, work_tree,
345
weave_store, transaction):
346
"""Record this revision unconditionally."""
347
mutter('new revision for {%s}', self.file_id)
348
self.revision = revision
349
change = self._get_snapshot_change(previous_entries)
350
self._snapshot_text(previous_entries, work_tree, weave_store,
354
def _snapshot_text(self, file_parents, work_tree, weave_store, transaction):
355
"""Record the 'text' of this entry, whatever form that takes.
357
This default implementation simply adds an empty text.
359
mutter('storing file {%s} in revision {%s}',
360
self.file_id, self.revision)
361
self._add_text_to_weave([], file_parents, weave_store, transaction)
363
321
def __eq__(self, other):
323
# For the case when objects are cached
364
325
if not isinstance(other, InventoryEntry):
365
326
return NotImplemented
674
585
return compatible
677
class Inventory(object):
678
"""Inventory of versioned files in a tree.
680
This describes which file_id is present at each point in the tree,
681
and possibly the SHA-1 or other information about the file.
588
class TreeReference(InventoryEntry):
590
__slots__ = ['reference_revision']
592
kind = 'tree-reference'
594
def __init__(self, file_id, name, parent_id, revision=None,
595
reference_revision=None):
596
InventoryEntry.__init__(self, file_id, name, parent_id)
597
self.revision = revision
598
self.reference_revision = reference_revision
601
return TreeReference(self.file_id, self.name, self.parent_id,
602
self.revision, self.reference_revision)
604
def _read_tree_state(self, path, work_tree):
605
"""Populate fields in the inventory entry from the given tree.
607
self.reference_revision = work_tree.get_reference_revision(
610
def _forget_tree_state(self):
611
self.reference_revision = None
613
def _unchanged(self, previous_ie):
614
"""See InventoryEntry._unchanged."""
615
compatible = super(TreeReference, self)._unchanged(previous_ie)
616
if self.reference_revision != previous_ie.reference_revision:
621
class CommonInventory(object):
622
"""Basic inventory logic, defined in terms of primitives like has_id.
624
An inventory is the metadata about the contents of a tree.
626
This is broadly a map from file_id to entries such as directories, files,
627
symlinks and tree references. Each entry maintains its own metadata like
628
SHA1 and length for files, or children for a directory.
682
630
Entries can be looked up either by path or by file_id.
684
The inventory represents a typical unix file tree, with
685
directories containing files and subdirectories. We never store
686
the full path to a file, because renaming a directory implicitly
687
moves all of its contents. This class internally maintains a
688
lookup tree that allows the children under a directory to be
691
632
InventoryEntry objects must not be modified after they are
692
633
inserted, other than through the Inventory API.
694
>>> inv = Inventory()
695
>>> inv.add(InventoryFile('123-123', 'hello.c', ROOT_ID))
696
InventoryFile('123-123', 'hello.c', parent_id='TREE_ROOT')
697
>>> inv['123-123'].name
700
May be treated as an iterator or set to look up file ids:
636
@deprecated_method(deprecated_in((2, 4, 0)))
637
def __contains__(self, file_id):
638
"""True if this entry contains a file with given id.
640
>>> inv = Inventory()
641
>>> inv.add(InventoryFile('123', 'foo.c', ROOT_ID))
642
InventoryFile('123', 'foo.c', parent_id='TREE_ROOT', sha1=None, len=None, revision=None)
643
>>> inv.has_id('123')
645
>>> inv.has_id('456')
648
Note that this method along with __iter__ are not encouraged for use as
649
they are less clear than specific query methods - they may be rmeoved
652
return self.has_id(file_id)
654
def has_filename(self, filename):
655
return bool(self.path2id(filename))
657
def id2path(self, file_id):
658
"""Return as a string the path to file_id.
661
>>> e = i.add(InventoryDirectory('src-id', 'src', ROOT_ID))
662
>>> e = i.add(InventoryFile('foo-id', 'foo.c', parent_id='src-id'))
663
>>> print i.id2path('foo-id')
666
:raises NoSuchId: If file_id is not present in the inventory.
668
# get all names, skipping root
669
return '/'.join(reversed(
670
[parent.name for parent in
671
self._iter_file_id_parents(file_id)][:-1]))
673
def iter_entries(self, from_dir=None, recursive=True):
674
"""Return (path, entry) pairs, in order by name.
676
:param from_dir: if None, start from the root,
677
otherwise start from this directory (either file-id or entry)
678
:param recursive: recurse into directories or not
681
if self.root is None:
685
elif isinstance(from_dir, basestring):
686
from_dir = self[from_dir]
688
# unrolling the recursive called changed the time from
689
# 440ms/663ms (inline/total) to 116ms/116ms
690
children = from_dir.children.items()
693
for name, ie in children:
696
children = collections.deque(children)
697
stack = [(u'', children)]
699
from_dir_relpath, children = stack[-1]
702
name, ie = children.popleft()
704
# we know that from_dir_relpath never ends in a slash
705
# and 'f' doesn't begin with one, we can do a string op, rather
706
# than the checks of pathjoin(), though this means that all paths
708
path = from_dir_relpath + '/' + name
712
if ie.kind != 'directory':
715
# But do this child first
716
new_children = ie.children.items()
718
new_children = collections.deque(new_children)
719
stack.append((path, new_children))
720
# Break out of inner loop, so that we start outer loop with child
723
# if we finished all children, pop it off the stack
726
def _preload_cache(self):
727
"""Populate any caches, we are about to access all items.
729
The default implementation does nothing, because CommonInventory doesn't
702
>>> bool(inv.path2id('hello.c'))
707
May also look up by name:
709
>>> [x[0] for x in inv.iter_entries()]
711
>>> inv = Inventory('TREE_ROOT-12345678-12345678')
712
>>> inv.add(InventoryFile('123-123', 'hello.c', ROOT_ID))
713
InventoryFile('123-123', 'hello.c', parent_id='TREE_ROOT-12345678-12345678')
715
def __init__(self, root_id=ROOT_ID):
716
"""Create or read an inventory.
718
If a working directory is specified, the inventory is read
719
from there. If the file is specified, read from that. If not,
720
the inventory is created empty.
722
The inventory is created with a default root directory, with
734
def iter_entries_by_dir(self, from_dir=None, specific_file_ids=None,
735
yield_parents=False):
736
"""Iterate over the entries in a directory first order.
738
This returns all entries for a directory before returning
739
the entries for children of a directory. This is not
740
lexicographically sorted order, and is a hybrid between
741
depth-first and breadth-first.
743
:param yield_parents: If True, yield the parents from the root leading
744
down to specific_file_ids that have been requested. This has no
745
impact if specific_file_ids is None.
746
:return: This yields (path, entry) pairs
725
# We are letting Branch.initialize() create a unique inventory
726
# root id. Rather than generating a random one here.
728
# root_id = bzrlib.branch.gen_file_id('TREE_ROOT')
729
self.root = RootEntry(root_id)
730
self._byid = {self.root.file_id: self.root}
734
other = Inventory(self.root.file_id)
735
# copy recursively so we know directories will be added before
736
# their children. There are more efficient ways than this...
737
for path, entry in self.iter_entries():
738
if entry == self.root:
740
other.add(entry.copy())
745
return iter(self._byid)
749
"""Returns number of entries."""
750
return len(self._byid)
753
def iter_entries(self, from_dir=None):
754
"""Return (path, entry) pairs, in order by name."""
748
if specific_file_ids and not isinstance(specific_file_ids, set):
749
specific_file_ids = set(specific_file_ids)
750
# TODO? Perhaps this should return the from_dir so that the root is
751
# yielded? or maybe an option?
752
if from_dir is None and specific_file_ids is None:
753
# They are iterating from the root, and have not specified any
754
# specific entries to look at. All current callers fully consume the
755
# iterator, so we can safely assume we are accessing all entries
756
self._preload_cache()
758
if self.root is None:
760
# Optimize a common case
761
if (not yield_parents and specific_file_ids is not None and
762
len(specific_file_ids) == 1):
763
file_id = list(specific_file_ids)[0]
764
if self.has_id(file_id):
765
yield self.id2path(file_id), self[file_id]
757
767
from_dir = self.root
768
if (specific_file_ids is None or yield_parents or
769
self.root.file_id in specific_file_ids):
758
771
elif isinstance(from_dir, basestring):
759
from_dir = self._byid[from_dir]
761
kids = from_dir.children.items()
763
for name, ie in kids:
765
if ie.kind == 'directory':
766
for cn, cie in self.iter_entries(from_dir=ie.file_id):
767
yield os.path.join(name, cn), cie
772
from_dir = self[from_dir]
774
if specific_file_ids is not None:
775
# TODO: jam 20070302 This could really be done as a loop rather
776
# than a bunch of recursive calls.
779
def add_ancestors(file_id):
780
if not byid.has_id(file_id):
782
parent_id = byid[file_id].parent_id
783
if parent_id is None:
785
if parent_id not in parents:
786
parents.add(parent_id)
787
add_ancestors(parent_id)
788
for file_id in specific_file_ids:
789
add_ancestors(file_id)
793
stack = [(u'', from_dir)]
795
cur_relpath, cur_dir = stack.pop()
798
for child_name, child_ie in sorted(cur_dir.children.iteritems()):
800
child_relpath = cur_relpath + child_name
802
if (specific_file_ids is None or
803
child_ie.file_id in specific_file_ids or
804
(yield_parents and child_ie.file_id in parents)):
805
yield child_relpath, child_ie
807
if child_ie.kind == 'directory':
808
if parents is None or child_ie.file_id in parents:
809
child_dirs.append((child_relpath+'/', child_ie))
810
stack.extend(reversed(child_dirs))
812
def _make_delta(self, old):
813
"""Make an inventory delta from two inventories."""
816
adds = new_ids - old_ids
817
deletes = old_ids - new_ids
818
common = old_ids.intersection(new_ids)
820
for file_id in deletes:
821
delta.append((old.id2path(file_id), None, file_id, None))
823
delta.append((None, self.id2path(file_id), file_id, self[file_id]))
824
for file_id in common:
825
if old[file_id] != self[file_id]:
826
delta.append((old.id2path(file_id), self.id2path(file_id),
827
file_id, self[file_id]))
830
def make_entry(self, kind, name, parent_id, file_id=None):
831
"""Simple thunk to bzrlib.inventory.make_entry."""
832
return make_entry(kind, name, parent_id, file_id)
770
834
def entries(self):
771
835
"""Return list of (path, ie) for all entries except the root.
777
841
kids = dir_ie.children.items()
779
843
for name, ie in kids:
780
child_path = os.path.join(dir_path, name)
844
child_path = osutils.pathjoin(dir_path, name)
781
845
accum.append((child_path, ie))
782
846
if ie.kind == 'directory':
783
847
descend(ie, child_path)
785
descend(self.root, '')
849
if self.root is not None:
850
descend(self.root, u'')
789
853
def directories(self):
790
854
"""Return (path, entry) pairs for all directories, including the root.
793
857
def descend(parent_ie, parent_path):
794
858
accum.append((parent_path, parent_ie))
796
860
kids = [(ie.name, ie) for ie in parent_ie.children.itervalues() if ie.kind == 'directory']
799
863
for name, child_ie in kids:
800
child_path = os.path.join(parent_path, name)
864
child_path = osutils.pathjoin(parent_path, name)
801
865
descend(child_ie, child_path)
802
descend(self.root, '')
866
descend(self.root, u'')
869
def path2id(self, relpath):
870
"""Walk down through directories to return entry of last component.
872
:param relpath: may be either a list of path components, or a single
873
string, in which case it is automatically split.
875
This returns the entry of the last component in the path,
876
which may be either a file or a directory.
878
Returns None IFF the path is not found.
880
if isinstance(relpath, basestring):
881
names = osutils.splitpath(relpath)
887
except errors.NoSuchId:
888
# root doesn't exist yet so nothing else can
894
children = getattr(parent, 'children', None)
903
return parent.file_id
905
def filter(self, specific_fileids):
906
"""Get an inventory view filtered against a set of file-ids.
908
Children of directories and parents are included.
910
The result may or may not reference the underlying inventory
911
so it should be treated as immutable.
913
interesting_parents = set()
914
for fileid in specific_fileids:
916
interesting_parents.update(self.get_idpath(fileid))
917
except errors.NoSuchId:
918
# This fileid is not in the inventory - that's ok
920
entries = self.iter_entries()
921
if self.root is None:
922
return Inventory(root_id=None)
923
other = Inventory(entries.next()[1].file_id)
924
other.root.revision = self.root.revision
925
other.revision_id = self.revision_id
926
directories_to_expand = set()
927
for path, entry in entries:
928
file_id = entry.file_id
929
if (file_id in specific_fileids
930
or entry.parent_id in directories_to_expand):
931
if entry.kind == 'directory':
932
directories_to_expand.add(file_id)
933
elif file_id not in interesting_parents:
935
other.add(entry.copy())
938
def get_idpath(self, file_id):
939
"""Return a list of file_ids for the path to an entry.
941
The list contains one element for each directory followed by
942
the id of the file itself. So the length of the returned list
943
is equal to the depth of the file in the tree, counting the
944
root directory as depth 1.
947
for parent in self._iter_file_id_parents(file_id):
948
p.insert(0, parent.file_id)
952
class Inventory(CommonInventory):
953
"""Mutable dict based in-memory inventory.
955
We never store the full path to a file, because renaming a directory
956
implicitly moves all of its contents. This class internally maintains a
957
lookup tree that allows the children under a directory to be
960
>>> inv = Inventory()
961
>>> inv.add(InventoryFile('123-123', 'hello.c', ROOT_ID))
962
InventoryFile('123-123', 'hello.c', parent_id='TREE_ROOT', sha1=None, len=None, revision=None)
963
>>> inv['123-123'].name
966
Id's may be looked up from paths:
968
>>> inv.path2id('hello.c')
970
>>> inv.has_id('123-123')
973
There are iterators over the contents:
975
>>> [entry[0] for entry in inv.iter_entries()]
979
def __init__(self, root_id=ROOT_ID, revision_id=None):
980
"""Create or read an inventory.
982
If a working directory is specified, the inventory is read
983
from there. If the file is specified, read from that. If not,
984
the inventory is created empty.
986
The inventory is created with a default root directory, with
989
if root_id is not None:
990
self._set_root(InventoryDirectory(root_id, u'', None))
994
self.revision_id = revision_id
997
# More than one page of ouput is not useful anymore to debug
1000
contents = repr(self._byid)
1001
if len(contents) > max_len:
1002
contents = contents[:(max_len-len(closing))] + closing
1003
return "<Inventory object at %x, contents=%r>" % (id(self), contents)
1005
def apply_delta(self, delta):
1006
"""Apply a delta to this inventory.
1008
See the inventory developers documentation for the theory behind
1011
If delta application fails the inventory is left in an indeterminate
1012
state and must not be used.
1014
:param delta: A list of changes to apply. After all the changes are
1015
applied the final inventory must be internally consistent, but it
1016
is ok to supply changes which, if only half-applied would have an
1017
invalid result - such as supplying two changes which rename two
1018
files, 'A' and 'B' with each other : [('A', 'B', 'A-id', a_entry),
1019
('B', 'A', 'B-id', b_entry)].
1021
Each change is a tuple, of the form (old_path, new_path, file_id,
1024
When new_path is None, the change indicates the removal of an entry
1025
from the inventory and new_entry will be ignored (using None is
1026
appropriate). If new_path is not None, then new_entry must be an
1027
InventoryEntry instance, which will be incorporated into the
1028
inventory (and replace any existing entry with the same file id).
1030
When old_path is None, the change indicates the addition of
1031
a new entry to the inventory.
1033
When neither new_path nor old_path are None, the change is a
1034
modification to an entry, such as a rename, reparent, kind change
1037
The children attribute of new_entry is ignored. This is because
1038
this method preserves children automatically across alterations to
1039
the parent of the children, and cases where the parent id of a
1040
child is changing require the child to be passed in as a separate
1041
change regardless. E.g. in the recursive deletion of a directory -
1042
the directory's children must be included in the delta, or the
1043
final inventory will be invalid.
1045
Note that a file_id must only appear once within a given delta.
1046
An AssertionError is raised otherwise.
1048
# Check that the delta is legal. It would be nice if this could be
1049
# done within the loops below but it's safer to validate the delta
1050
# before starting to mutate the inventory, as there isn't a rollback
1052
list(_check_delta_unique_ids(_check_delta_unique_new_paths(
1053
_check_delta_unique_old_paths(_check_delta_ids_match_entry(
1054
_check_delta_ids_are_valid(
1055
_check_delta_new_path_entry_both_or_None(
1059
# Remove all affected items which were in the original inventory,
1060
# starting with the longest paths, thus ensuring parents are examined
1061
# after their children, which means that everything we examine has no
1062
# modified children remaining by the time we examine it.
1063
for old_path, file_id in sorted(((op, f) for op, np, f, e in delta
1064
if op is not None), reverse=True):
1065
# Preserve unaltered children of file_id for later reinsertion.
1066
file_id_children = getattr(self[file_id], 'children', {})
1067
if len(file_id_children):
1068
children[file_id] = file_id_children
1069
if self.id2path(file_id) != old_path:
1070
raise errors.InconsistentDelta(old_path, file_id,
1071
"Entry was at wrong other path %r." % self.id2path(file_id))
1072
# Remove file_id and the unaltered children. If file_id is not
1073
# being deleted it will be reinserted back later.
1074
self.remove_recursive_id(file_id)
1075
# Insert all affected which should be in the new inventory, reattaching
1076
# their children if they had any. This is done from shortest path to
1077
# longest, ensuring that items which were modified and whose parents in
1078
# the resulting inventory were also modified, are inserted after their
1080
for new_path, f, new_entry in sorted((np, f, e) for op, np, f, e in
1081
delta if np is not None):
1082
if new_entry.kind == 'directory':
1083
# Pop the child which to allow detection of children whose
1084
# parents were deleted and which were not reattached to a new
1086
replacement = InventoryDirectory(new_entry.file_id,
1087
new_entry.name, new_entry.parent_id)
1088
replacement.revision = new_entry.revision
1089
replacement.children = children.pop(replacement.file_id, {})
1090
new_entry = replacement
1093
except errors.DuplicateFileId:
1094
raise errors.InconsistentDelta(new_path, new_entry.file_id,
1095
"New id is already present in target.")
1096
except AttributeError:
1097
raise errors.InconsistentDelta(new_path, new_entry.file_id,
1098
"Parent is not a directory.")
1099
if self.id2path(new_entry.file_id) != new_path:
1100
raise errors.InconsistentDelta(new_path, new_entry.file_id,
1101
"New path is not consistent with parent path.")
1103
# Get the parent id that was deleted
1104
parent_id, children = children.popitem()
1105
raise errors.InconsistentDelta("<deleted>", parent_id,
1106
"The file id was deleted but its children were not deleted.")
1108
def create_by_apply_delta(self, inventory_delta, new_revision_id,
1109
propagate_caches=False):
1110
"""See CHKInventory.create_by_apply_delta()"""
1111
new_inv = self.copy()
1112
new_inv.apply_delta(inventory_delta)
1113
new_inv.revision_id = new_revision_id
1116
def _set_root(self, ie):
1118
self._byid = {self.root.file_id: self.root}
1121
# TODO: jam 20051218 Should copy also copy the revision_id?
1122
entries = self.iter_entries()
1123
if self.root is None:
1124
return Inventory(root_id=None)
1125
other = Inventory(entries.next()[1].file_id)
1126
other.root.revision = self.root.revision
1127
# copy recursively so we know directories will be added before
1128
# their children. There are more efficient ways than this...
1129
for path, entry in entries:
1130
other.add(entry.copy())
1134
"""Iterate over all file-ids."""
1135
return iter(self._byid)
1137
def iter_just_entries(self):
1138
"""Iterate over all entries.
807
def __contains__(self, file_id):
808
"""True if this entry contains a file with given id.
810
>>> inv = Inventory()
811
>>> inv.add(InventoryFile('123', 'foo.c', ROOT_ID))
812
InventoryFile('123', 'foo.c', parent_id='TREE_ROOT')
1140
Unlike iter_entries(), just the entries are returned (not (path, ie))
1141
and the order of entries is undefined.
1143
XXX: We may not want to merge this into bzr.dev.
818
return file_id in self._byid
1145
if self.root is None:
1147
for _, ie in self._byid.iteritems():
1151
"""Returns number of entries."""
1152
return len(self._byid)
821
1154
def __getitem__(self, file_id):
822
1155
"""Return the entry for given file_id.
824
1157
>>> inv = Inventory()
825
1158
>>> inv.add(InventoryFile('123123', 'hello.c', ROOT_ID))
826
InventoryFile('123123', 'hello.c', parent_id='TREE_ROOT')
1159
InventoryFile('123123', 'hello.c', parent_id='TREE_ROOT', sha1=None, len=None, revision=None)
827
1160
>>> inv['123123'].name
831
1164
return self._byid[file_id]
832
1165
except KeyError:
834
raise BzrError("can't look up file_id None")
836
raise BzrError("file_id {%s} not in inventory" % file_id)
1166
# really we're passing an inventory, not a tree...
1167
raise errors.NoSuchId(self, file_id)
839
1169
def get_file_kind(self, file_id):
840
1170
return self._byid[file_id].kind
1060
1368
del old_parent.children[file_ie.name]
1061
1369
new_parent.children[new_name] = file_ie
1063
1371
file_ie.name = new_name
1064
1372
file_ie.parent_id = new_parent_id
1374
def is_root(self, file_id):
1375
return self.root is not None and file_id == self.root.file_id
1378
class CHKInventory(CommonInventory):
1379
"""An inventory persisted in a CHK store.
1381
By design, a CHKInventory is immutable so many of the methods
1382
supported by Inventory - add, rename, apply_delta, etc - are *not*
1383
supported. To create a new CHKInventory, use create_by_apply_delta()
1384
or from_inventory(), say.
1386
Internally, a CHKInventory has one or two CHKMaps:
1388
* id_to_entry - a map from (file_id,) => InventoryEntry as bytes
1389
* parent_id_basename_to_file_id - a map from (parent_id, basename_utf8)
1392
The second map is optional and not present in early CHkRepository's.
1394
No caching is performed: every method call or item access will perform
1395
requests to the storage layer. As such, keep references to objects you
1399
def __init__(self, search_key_name):
1400
CommonInventory.__init__(self)
1401
self._fileid_to_entry_cache = {}
1402
self._fully_cached = False
1403
self._path_to_fileid_cache = {}
1404
self._search_key_name = search_key_name
1407
def __eq__(self, other):
1408
"""Compare two sets by comparing their contents."""
1409
if not isinstance(other, CHKInventory):
1410
return NotImplemented
1412
this_key = self.id_to_entry.key()
1413
other_key = other.id_to_entry.key()
1414
this_pid_key = self.parent_id_basename_to_file_id.key()
1415
other_pid_key = other.parent_id_basename_to_file_id.key()
1416
if None in (this_key, this_pid_key, other_key, other_pid_key):
1418
return this_key == other_key and this_pid_key == other_pid_key
1420
def _entry_to_bytes(self, entry):
1421
"""Serialise entry as a single bytestring.
1423
:param Entry: An inventory entry.
1424
:return: A bytestring for the entry.
1427
ENTRY ::= FILE | DIR | SYMLINK | TREE
1428
FILE ::= "file: " COMMON SEP SHA SEP SIZE SEP EXECUTABLE
1429
DIR ::= "dir: " COMMON
1430
SYMLINK ::= "symlink: " COMMON SEP TARGET_UTF8
1431
TREE ::= "tree: " COMMON REFERENCE_REVISION
1432
COMMON ::= FILE_ID SEP PARENT_ID SEP NAME_UTF8 SEP REVISION
1435
if entry.parent_id is not None:
1436
parent_str = entry.parent_id
1439
name_str = entry.name.encode("utf8")
1440
if entry.kind == 'file':
1441
if entry.executable:
1445
return "file: %s\n%s\n%s\n%s\n%s\n%d\n%s" % (
1446
entry.file_id, parent_str, name_str, entry.revision,
1447
entry.text_sha1, entry.text_size, exec_str)
1448
elif entry.kind == 'directory':
1449
return "dir: %s\n%s\n%s\n%s" % (
1450
entry.file_id, parent_str, name_str, entry.revision)
1451
elif entry.kind == 'symlink':
1452
return "symlink: %s\n%s\n%s\n%s\n%s" % (
1453
entry.file_id, parent_str, name_str, entry.revision,
1454
entry.symlink_target.encode("utf8"))
1455
elif entry.kind == 'tree-reference':
1456
return "tree: %s\n%s\n%s\n%s\n%s" % (
1457
entry.file_id, parent_str, name_str, entry.revision,
1458
entry.reference_revision)
1460
raise ValueError("unknown kind %r" % entry.kind)
1462
def _expand_fileids_to_parents_and_children(self, file_ids):
1463
"""Give a more wholistic view starting with the given file_ids.
1465
For any file_id which maps to a directory, we will include all children
1466
of that directory. We will also include all directories which are
1467
parents of the given file_ids, but we will not include their children.
1474
fringle # fringle-id
1478
if given [foo-id] we will include
1479
TREE_ROOT as interesting parents
1481
foo-id, baz-id, frob-id, fringle-id
1485
# TODO: Pre-pass over the list of fileids to see if anything is already
1486
# deserialized in self._fileid_to_entry_cache
1488
directories_to_expand = set()
1489
children_of_parent_id = {}
1490
# It is okay if some of the fileids are missing
1491
for entry in self._getitems(file_ids):
1492
if entry.kind == 'directory':
1493
directories_to_expand.add(entry.file_id)
1494
interesting.add(entry.parent_id)
1495
children_of_parent_id.setdefault(entry.parent_id, set()
1496
).add(entry.file_id)
1498
# Now, interesting has all of the direct parents, but not the
1499
# parents of those parents. It also may have some duplicates with
1501
remaining_parents = interesting.difference(file_ids)
1502
# When we hit the TREE_ROOT, we'll get an interesting parent of None,
1503
# but we don't actually want to recurse into that
1504
interesting.add(None) # this will auto-filter it in the loop
1505
remaining_parents.discard(None)
1506
while remaining_parents:
1507
next_parents = set()
1508
for entry in self._getitems(remaining_parents):
1509
next_parents.add(entry.parent_id)
1510
children_of_parent_id.setdefault(entry.parent_id, set()
1511
).add(entry.file_id)
1512
# Remove any search tips we've already processed
1513
remaining_parents = next_parents.difference(interesting)
1514
interesting.update(remaining_parents)
1515
# We should probably also .difference(directories_to_expand)
1516
interesting.update(file_ids)
1517
interesting.discard(None)
1518
while directories_to_expand:
1519
# Expand directories by looking in the
1520
# parent_id_basename_to_file_id map
1521
keys = [StaticTuple(f,).intern() for f in directories_to_expand]
1522
directories_to_expand = set()
1523
items = self.parent_id_basename_to_file_id.iteritems(keys)
1524
next_file_ids = set([item[1] for item in items])
1525
next_file_ids = next_file_ids.difference(interesting)
1526
interesting.update(next_file_ids)
1527
for entry in self._getitems(next_file_ids):
1528
if entry.kind == 'directory':
1529
directories_to_expand.add(entry.file_id)
1530
children_of_parent_id.setdefault(entry.parent_id, set()
1531
).add(entry.file_id)
1532
return interesting, children_of_parent_id
1534
def filter(self, specific_fileids):
1535
"""Get an inventory view filtered against a set of file-ids.
1537
Children of directories and parents are included.
1539
The result may or may not reference the underlying inventory
1540
so it should be treated as immutable.
1543
parent_to_children) = self._expand_fileids_to_parents_and_children(
1545
# There is some overlap here, but we assume that all interesting items
1546
# are in the _fileid_to_entry_cache because we had to read them to
1547
# determine if they were a dir we wanted to recurse, or just a file
1548
# This should give us all the entries we'll want to add, so start
1550
other = Inventory(self.root_id)
1551
other.root.revision = self.root.revision
1552
other.revision_id = self.revision_id
1553
if not interesting or not parent_to_children:
1554
# empty filter, or filtering entrys that don't exist
1555
# (if even 1 existed, then we would have populated
1556
# parent_to_children with at least the tree root.)
1558
cache = self._fileid_to_entry_cache
1559
remaining_children = collections.deque(parent_to_children[self.root_id])
1560
while remaining_children:
1561
file_id = remaining_children.popleft()
1563
if ie.kind == 'directory':
1564
ie = ie.copy() # We create a copy to depopulate the .children attribute
1565
# TODO: depending on the uses of 'other' we should probably alwyas
1566
# '.copy()' to prevent someone from mutating other and
1567
# invaliding our internal cache
1569
if file_id in parent_to_children:
1570
remaining_children.extend(parent_to_children[file_id])
1574
def _bytes_to_utf8name_key(bytes):
1575
"""Get the file_id, revision_id key out of bytes."""
1576
# We don't normally care about name, except for times when we want
1577
# to filter out empty names because of non rich-root...
1578
sections = bytes.split('\n')
1579
kind, file_id = sections[0].split(': ')
1580
return (sections[2], intern(file_id), intern(sections[3]))
1582
def _bytes_to_entry(self, bytes):
1583
"""Deserialise a serialised entry."""
1584
sections = bytes.split('\n')
1585
if sections[0].startswith("file: "):
1586
result = InventoryFile(sections[0][6:],
1587
sections[2].decode('utf8'),
1589
result.text_sha1 = sections[4]
1590
result.text_size = int(sections[5])
1591
result.executable = sections[6] == "Y"
1592
elif sections[0].startswith("dir: "):
1593
result = CHKInventoryDirectory(sections[0][5:],
1594
sections[2].decode('utf8'),
1596
elif sections[0].startswith("symlink: "):
1597
result = InventoryLink(sections[0][9:],
1598
sections[2].decode('utf8'),
1600
result.symlink_target = sections[4].decode('utf8')
1601
elif sections[0].startswith("tree: "):
1602
result = TreeReference(sections[0][6:],
1603
sections[2].decode('utf8'),
1605
result.reference_revision = sections[4]
1607
raise ValueError("Not a serialised entry %r" % bytes)
1608
result.file_id = intern(result.file_id)
1609
result.revision = intern(sections[3])
1610
if result.parent_id == '':
1611
result.parent_id = None
1612
self._fileid_to_entry_cache[result.file_id] = result
1615
def create_by_apply_delta(self, inventory_delta, new_revision_id,
1616
propagate_caches=False):
1617
"""Create a new CHKInventory by applying inventory_delta to this one.
1619
See the inventory developers documentation for the theory behind
1622
:param inventory_delta: The inventory delta to apply. See
1623
Inventory.apply_delta for details.
1624
:param new_revision_id: The revision id of the resulting CHKInventory.
1625
:param propagate_caches: If True, the caches for this inventory are
1626
copied to and updated for the result.
1627
:return: The new CHKInventory.
1629
split = osutils.split
1630
result = CHKInventory(self._search_key_name)
1631
if propagate_caches:
1632
# Just propagate the path-to-fileid cache for now
1633
result._path_to_fileid_cache = dict(self._path_to_fileid_cache.iteritems())
1634
search_key_func = chk_map.search_key_registry.get(self._search_key_name)
1635
self.id_to_entry._ensure_root()
1636
maximum_size = self.id_to_entry._root_node.maximum_size
1637
result.revision_id = new_revision_id
1638
result.id_to_entry = chk_map.CHKMap(
1639
self.id_to_entry._store,
1640
self.id_to_entry.key(),
1641
search_key_func=search_key_func)
1642
result.id_to_entry._ensure_root()
1643
result.id_to_entry._root_node.set_maximum_size(maximum_size)
1644
# Change to apply to the parent_id_basename delta. The dict maps
1645
# (parent_id, basename) -> (old_key, new_value). We use a dict because
1646
# when a path has its id replaced (e.g. the root is changed, or someone
1647
# does bzr mv a b, bzr mv c a, we should output a single change to this
1648
# map rather than two.
1649
parent_id_basename_delta = {}
1650
if self.parent_id_basename_to_file_id is not None:
1651
result.parent_id_basename_to_file_id = chk_map.CHKMap(
1652
self.parent_id_basename_to_file_id._store,
1653
self.parent_id_basename_to_file_id.key(),
1654
search_key_func=search_key_func)
1655
result.parent_id_basename_to_file_id._ensure_root()
1656
self.parent_id_basename_to_file_id._ensure_root()
1657
result_p_id_root = result.parent_id_basename_to_file_id._root_node
1658
p_id_root = self.parent_id_basename_to_file_id._root_node
1659
result_p_id_root.set_maximum_size(p_id_root.maximum_size)
1660
result_p_id_root._key_width = p_id_root._key_width
1662
result.parent_id_basename_to_file_id = None
1663
result.root_id = self.root_id
1664
id_to_entry_delta = []
1665
# inventory_delta is only traversed once, so we just update the
1667
# Check for repeated file ids
1668
inventory_delta = _check_delta_unique_ids(inventory_delta)
1669
# Repeated old paths
1670
inventory_delta = _check_delta_unique_old_paths(inventory_delta)
1671
# Check for repeated new paths
1672
inventory_delta = _check_delta_unique_new_paths(inventory_delta)
1673
# Check for entries that don't match the fileid
1674
inventory_delta = _check_delta_ids_match_entry(inventory_delta)
1675
# Check for nonsense fileids
1676
inventory_delta = _check_delta_ids_are_valid(inventory_delta)
1677
# Check for new_path <-> entry consistency
1678
inventory_delta = _check_delta_new_path_entry_both_or_None(
1680
# All changed entries need to have their parents be directories and be
1681
# at the right path. This set contains (path, id) tuples.
1683
# When we delete an item, all the children of it must be either deleted
1684
# or altered in their own right. As we batch process the change via
1685
# CHKMap.apply_delta, we build a set of things to use to validate the
1689
for old_path, new_path, file_id, entry in inventory_delta:
1692
result.root_id = file_id
1693
if new_path is None:
1698
if propagate_caches:
1700
del result._path_to_fileid_cache[old_path]
1703
deletes.add(file_id)
1705
new_key = StaticTuple(file_id,)
1706
new_value = result._entry_to_bytes(entry)
1707
# Update caches. It's worth doing this whether
1708
# we're propagating the old caches or not.
1709
result._path_to_fileid_cache[new_path] = file_id
1710
parents.add((split(new_path)[0], entry.parent_id))
1711
if old_path is None:
1714
old_key = StaticTuple(file_id,)
1715
if self.id2path(file_id) != old_path:
1716
raise errors.InconsistentDelta(old_path, file_id,
1717
"Entry was at wrong other path %r." %
1718
self.id2path(file_id))
1719
altered.add(file_id)
1720
id_to_entry_delta.append(StaticTuple(old_key, new_key, new_value))
1721
if result.parent_id_basename_to_file_id is not None:
1722
# parent_id, basename changes
1723
if old_path is None:
1726
old_entry = self[file_id]
1727
old_key = self._parent_id_basename_key(old_entry)
1728
if new_path is None:
1732
new_key = self._parent_id_basename_key(entry)
1734
# If the two keys are the same, the value will be unchanged
1735
# as its always the file id for this entry.
1736
if old_key != new_key:
1737
# Transform a change into explicit delete/add preserving
1738
# a possible match on the key from a different file id.
1739
if old_key is not None:
1740
parent_id_basename_delta.setdefault(
1741
old_key, [None, None])[0] = old_key
1742
if new_key is not None:
1743
parent_id_basename_delta.setdefault(
1744
new_key, [None, None])[1] = new_value
1745
# validate that deletes are complete.
1746
for file_id in deletes:
1747
entry = self[file_id]
1748
if entry.kind != 'directory':
1750
# This loop could potentially be better by using the id_basename
1751
# map to just get the child file ids.
1752
for child in entry.children.values():
1753
if child.file_id not in altered:
1754
raise errors.InconsistentDelta(self.id2path(child.file_id),
1755
child.file_id, "Child not deleted or reparented when "
1757
result.id_to_entry.apply_delta(id_to_entry_delta)
1758
if parent_id_basename_delta:
1759
# Transform the parent_id_basename delta data into a linear delta
1760
# with only one record for a given key. Optimally this would allow
1761
# re-keying, but its simpler to just output that as a delete+add
1762
# to spend less time calculating the delta.
1764
for key, (old_key, value) in parent_id_basename_delta.iteritems():
1765
if value is not None:
1766
delta_list.append((old_key, key, value))
1768
delta_list.append((old_key, None, None))
1769
result.parent_id_basename_to_file_id.apply_delta(delta_list)
1770
parents.discard(('', None))
1771
for parent_path, parent in parents:
1773
if result[parent].kind != 'directory':
1774
raise errors.InconsistentDelta(result.id2path(parent), parent,
1775
'Not a directory, but given children')
1776
except errors.NoSuchId:
1777
raise errors.InconsistentDelta("<unknown>", parent,
1778
"Parent is not present in resulting inventory.")
1779
if result.path2id(parent_path) != parent:
1780
raise errors.InconsistentDelta(parent_path, parent,
1781
"Parent has wrong path %r." % result.path2id(parent_path))
1785
def deserialise(klass, chk_store, bytes, expected_revision_id):
1786
"""Deserialise a CHKInventory.
1788
:param chk_store: A CHK capable VersionedFiles instance.
1789
:param bytes: The serialised bytes.
1790
:param expected_revision_id: The revision ID we think this inventory is
1792
:return: A CHKInventory
1794
lines = bytes.split('\n')
1796
raise AssertionError('bytes to deserialize must end with an eol')
1798
if lines[0] != 'chkinventory:':
1799
raise ValueError("not a serialised CHKInventory: %r" % bytes)
1801
allowed_keys = frozenset(['root_id', 'revision_id', 'search_key_name',
1802
'parent_id_basename_to_file_id',
1804
for line in lines[1:]:
1805
key, value = line.split(': ', 1)
1806
if key not in allowed_keys:
1807
raise errors.BzrError('Unknown key in inventory: %r\n%r'
1810
raise errors.BzrError('Duplicate key in inventory: %r\n%r'
1813
revision_id = intern(info['revision_id'])
1814
root_id = intern(info['root_id'])
1815
search_key_name = intern(info.get('search_key_name', 'plain'))
1816
parent_id_basename_to_file_id = intern(info.get(
1817
'parent_id_basename_to_file_id', None))
1818
if not parent_id_basename_to_file_id.startswith('sha1:'):
1819
raise ValueError('parent_id_basename_to_file_id should be a sha1'
1820
' key not %r' % (parent_id_basename_to_file_id,))
1821
id_to_entry = info['id_to_entry']
1822
if not id_to_entry.startswith('sha1:'):
1823
raise ValueError('id_to_entry should be a sha1'
1824
' key not %r' % (id_to_entry,))
1826
result = CHKInventory(search_key_name)
1827
result.revision_id = revision_id
1828
result.root_id = root_id
1829
search_key_func = chk_map.search_key_registry.get(
1830
result._search_key_name)
1831
if parent_id_basename_to_file_id is not None:
1832
result.parent_id_basename_to_file_id = chk_map.CHKMap(
1833
chk_store, StaticTuple(parent_id_basename_to_file_id,),
1834
search_key_func=search_key_func)
1836
result.parent_id_basename_to_file_id = None
1838
result.id_to_entry = chk_map.CHKMap(chk_store,
1839
StaticTuple(id_to_entry,),
1840
search_key_func=search_key_func)
1841
if (result.revision_id,) != expected_revision_id:
1842
raise ValueError("Mismatched revision id and expected: %r, %r" %
1843
(result.revision_id, expected_revision_id))
1847
def from_inventory(klass, chk_store, inventory, maximum_size=0, search_key_name='plain'):
1848
"""Create a CHKInventory from an existing inventory.
1850
The content of inventory is copied into the chk_store, and a
1851
CHKInventory referencing that is returned.
1853
:param chk_store: A CHK capable VersionedFiles instance.
1854
:param inventory: The inventory to copy.
1855
:param maximum_size: The CHKMap node size limit.
1856
:param search_key_name: The identifier for the search key function
1858
result = klass(search_key_name)
1859
result.revision_id = inventory.revision_id
1860
result.root_id = inventory.root.file_id
1862
entry_to_bytes = result._entry_to_bytes
1863
parent_id_basename_key = result._parent_id_basename_key
1864
id_to_entry_dict = {}
1865
parent_id_basename_dict = {}
1866
for path, entry in inventory.iter_entries():
1867
key = StaticTuple(entry.file_id,).intern()
1868
id_to_entry_dict[key] = entry_to_bytes(entry)
1869
p_id_key = parent_id_basename_key(entry)
1870
parent_id_basename_dict[p_id_key] = entry.file_id
1872
result._populate_from_dicts(chk_store, id_to_entry_dict,
1873
parent_id_basename_dict, maximum_size=maximum_size)
1876
def _populate_from_dicts(self, chk_store, id_to_entry_dict,
1877
parent_id_basename_dict, maximum_size):
1878
search_key_func = chk_map.search_key_registry.get(self._search_key_name)
1879
root_key = chk_map.CHKMap.from_dict(chk_store, id_to_entry_dict,
1880
maximum_size=maximum_size, key_width=1,
1881
search_key_func=search_key_func)
1882
self.id_to_entry = chk_map.CHKMap(chk_store, root_key,
1884
root_key = chk_map.CHKMap.from_dict(chk_store,
1885
parent_id_basename_dict,
1886
maximum_size=maximum_size, key_width=2,
1887
search_key_func=search_key_func)
1888
self.parent_id_basename_to_file_id = chk_map.CHKMap(chk_store,
1889
root_key, search_key_func)
1891
def _parent_id_basename_key(self, entry):
1892
"""Create a key for a entry in a parent_id_basename_to_file_id index."""
1893
if entry.parent_id is not None:
1894
parent_id = entry.parent_id
1897
return StaticTuple(parent_id, entry.name.encode('utf8')).intern()
1899
def __getitem__(self, file_id):
1900
"""map a single file_id -> InventoryEntry."""
1902
raise errors.NoSuchId(self, file_id)
1903
result = self._fileid_to_entry_cache.get(file_id, None)
1904
if result is not None:
1907
return self._bytes_to_entry(
1908
self.id_to_entry.iteritems([StaticTuple(file_id,)]).next()[1])
1909
except StopIteration:
1910
# really we're passing an inventory, not a tree...
1911
raise errors.NoSuchId(self, file_id)
1913
def _getitems(self, file_ids):
1914
"""Similar to __getitem__, but lets you query for multiple.
1916
The returned order is undefined. And currently if an item doesn't
1917
exist, it isn't included in the output.
1921
for file_id in file_ids:
1922
entry = self._fileid_to_entry_cache.get(file_id, None)
1924
remaining.append(file_id)
1926
result.append(entry)
1927
file_keys = [StaticTuple(f,).intern() for f in remaining]
1928
for file_key, value in self.id_to_entry.iteritems(file_keys):
1929
entry = self._bytes_to_entry(value)
1930
result.append(entry)
1931
self._fileid_to_entry_cache[entry.file_id] = entry
1934
def has_id(self, file_id):
1935
# Perhaps have an explicit 'contains' method on CHKMap ?
1936
if self._fileid_to_entry_cache.get(file_id, None) is not None:
1939
self.id_to_entry.iteritems([StaticTuple(file_id,)]))) == 1
1941
def is_root(self, file_id):
1942
return file_id == self.root_id
1944
def _iter_file_id_parents(self, file_id):
1945
"""Yield the parents of file_id up to the root."""
1946
while file_id is not None:
1950
raise errors.NoSuchId(tree=self, file_id=file_id)
1952
file_id = ie.parent_id
1955
"""Iterate over all file-ids."""
1956
for key, _ in self.id_to_entry.iteritems():
1959
def iter_just_entries(self):
1960
"""Iterate over all entries.
1962
Unlike iter_entries(), just the entries are returned (not (path, ie))
1963
and the order of entries is undefined.
1965
XXX: We may not want to merge this into bzr.dev.
1967
for key, entry in self.id_to_entry.iteritems():
1969
ie = self._fileid_to_entry_cache.get(file_id, None)
1971
ie = self._bytes_to_entry(entry)
1972
self._fileid_to_entry_cache[file_id] = ie
1975
def _preload_cache(self):
1976
"""Make sure all file-ids are in _fileid_to_entry_cache"""
1977
if self._fully_cached:
1978
return # No need to do it again
1979
# The optimal sort order is to use iteritems() directly
1980
cache = self._fileid_to_entry_cache
1981
for key, entry in self.id_to_entry.iteritems():
1983
if file_id not in cache:
1984
ie = self._bytes_to_entry(entry)
1988
last_parent_id = last_parent_ie = None
1989
pid_items = self.parent_id_basename_to_file_id.iteritems()
1990
for key, child_file_id in pid_items:
1991
if key == ('', ''): # This is the root
1992
if child_file_id != self.root_id:
1993
raise ValueError('Data inconsistency detected.'
1994
' We expected data with key ("","") to match'
1995
' the root id, but %s != %s'
1996
% (child_file_id, self.root_id))
1998
parent_id, basename = key
1999
ie = cache[child_file_id]
2000
if parent_id == last_parent_id:
2001
parent_ie = last_parent_ie
2003
parent_ie = cache[parent_id]
2004
if parent_ie.kind != 'directory':
2005
raise ValueError('Data inconsistency detected.'
2006
' An entry in the parent_id_basename_to_file_id map'
2007
' has parent_id {%s} but the kind of that object'
2008
' is %r not "directory"' % (parent_id, parent_ie.kind))
2009
if parent_ie._children is None:
2010
parent_ie._children = {}
2011
basename = basename.decode('utf-8')
2012
if basename in parent_ie._children:
2013
existing_ie = parent_ie._children[basename]
2014
if existing_ie != ie:
2015
raise ValueError('Data inconsistency detected.'
2016
' Two entries with basename %r were found'
2017
' in the parent entry {%s}'
2018
% (basename, parent_id))
2019
if basename != ie.name:
2020
raise ValueError('Data inconsistency detected.'
2021
' In the parent_id_basename_to_file_id map, file_id'
2022
' {%s} is listed as having basename %r, but in the'
2023
' id_to_entry map it is %r'
2024
% (child_file_id, basename, ie.name))
2025
parent_ie._children[basename] = ie
2026
self._fully_cached = True
2028
def iter_changes(self, basis):
2029
"""Generate a Tree.iter_changes change list between this and basis.
2031
:param basis: Another CHKInventory.
2032
:return: An iterator over the changes between self and basis, as per
2033
tree.iter_changes().
2035
# We want: (file_id, (path_in_source, path_in_target),
2036
# changed_content, versioned, parent, name, kind,
2038
for key, basis_value, self_value in \
2039
self.id_to_entry.iter_changes(basis.id_to_entry):
2041
if basis_value is not None:
2042
basis_entry = basis._bytes_to_entry(basis_value)
2043
path_in_source = basis.id2path(file_id)
2044
basis_parent = basis_entry.parent_id
2045
basis_name = basis_entry.name
2046
basis_executable = basis_entry.executable
2048
path_in_source = None
2051
basis_executable = None
2052
if self_value is not None:
2053
self_entry = self._bytes_to_entry(self_value)
2054
path_in_target = self.id2path(file_id)
2055
self_parent = self_entry.parent_id
2056
self_name = self_entry.name
2057
self_executable = self_entry.executable
2059
path_in_target = None
2062
self_executable = None
2063
if basis_value is None:
2065
kind = (None, self_entry.kind)
2066
versioned = (False, True)
2067
elif self_value is None:
2069
kind = (basis_entry.kind, None)
2070
versioned = (True, False)
2072
kind = (basis_entry.kind, self_entry.kind)
2073
versioned = (True, True)
2074
changed_content = False
2075
if kind[0] != kind[1]:
2076
changed_content = True
2077
elif kind[0] == 'file':
2078
if (self_entry.text_size != basis_entry.text_size or
2079
self_entry.text_sha1 != basis_entry.text_sha1):
2080
changed_content = True
2081
elif kind[0] == 'symlink':
2082
if self_entry.symlink_target != basis_entry.symlink_target:
2083
changed_content = True
2084
elif kind[0] == 'tree-reference':
2085
if (self_entry.reference_revision !=
2086
basis_entry.reference_revision):
2087
changed_content = True
2088
parent = (basis_parent, self_parent)
2089
name = (basis_name, self_name)
2090
executable = (basis_executable, self_executable)
2091
if (not changed_content
2092
and parent[0] == parent[1]
2093
and name[0] == name[1]
2094
and executable[0] == executable[1]):
2095
# Could happen when only the revision changed for a directory
2098
yield (file_id, (path_in_source, path_in_target), changed_content,
2099
versioned, parent, name, kind, executable)
2102
"""Return the number of entries in the inventory."""
2103
return len(self.id_to_entry)
2105
def _make_delta(self, old):
2106
"""Make an inventory delta from two inventories."""
2107
if type(old) != CHKInventory:
2108
return CommonInventory._make_delta(self, old)
2110
for key, old_value, self_value in \
2111
self.id_to_entry.iter_changes(old.id_to_entry):
2113
if old_value is not None:
2114
old_path = old.id2path(file_id)
2117
if self_value is not None:
2118
entry = self._bytes_to_entry(self_value)
2119
self._fileid_to_entry_cache[file_id] = entry
2120
new_path = self.id2path(file_id)
2124
delta.append((old_path, new_path, file_id, entry))
2127
def path2id(self, relpath):
2128
"""See CommonInventory.path2id()."""
2129
# TODO: perhaps support negative hits?
2130
result = self._path_to_fileid_cache.get(relpath, None)
2131
if result is not None:
2133
if isinstance(relpath, basestring):
2134
names = osutils.splitpath(relpath)
2137
current_id = self.root_id
2138
if current_id is None:
2140
parent_id_index = self.parent_id_basename_to_file_id
2142
for basename in names:
2143
if cur_path is None:
2146
cur_path = cur_path + '/' + basename
2147
basename_utf8 = basename.encode('utf8')
2148
file_id = self._path_to_fileid_cache.get(cur_path, None)
2150
key_filter = [StaticTuple(current_id, basename_utf8)]
2151
items = parent_id_index.iteritems(key_filter)
2152
for (parent_id, name_utf8), file_id in items:
2153
if parent_id != current_id or name_utf8 != basename_utf8:
2154
raise errors.BzrError("corrupt inventory lookup! "
2155
"%r %r %r %r" % (parent_id, current_id, name_utf8,
2160
self._path_to_fileid_cache[cur_path] = file_id
2161
current_id = file_id
2165
"""Serialise the inventory to lines."""
2166
lines = ["chkinventory:\n"]
2167
if self._search_key_name != 'plain':
2168
# custom ordering grouping things that don't change together
2169
lines.append('search_key_name: %s\n' % (self._search_key_name,))
2170
lines.append("root_id: %s\n" % self.root_id)
2171
lines.append('parent_id_basename_to_file_id: %s\n' %
2172
(self.parent_id_basename_to_file_id.key()[0],))
2173
lines.append("revision_id: %s\n" % self.revision_id)
2174
lines.append("id_to_entry: %s\n" % (self.id_to_entry.key()[0],))
2176
lines.append("revision_id: %s\n" % self.revision_id)
2177
lines.append("root_id: %s\n" % self.root_id)
2178
if self.parent_id_basename_to_file_id is not None:
2179
lines.append('parent_id_basename_to_file_id: %s\n' %
2180
(self.parent_id_basename_to_file_id.key()[0],))
2181
lines.append("id_to_entry: %s\n" % (self.id_to_entry.key()[0],))
2186
"""Get the root entry."""
2187
return self[self.root_id]
2190
class CHKInventoryDirectory(InventoryDirectory):
2191
"""A directory in an inventory."""
2193
__slots__ = ['_children', '_chk_inventory']
2195
def __init__(self, file_id, name, parent_id, chk_inventory):
2196
# Don't call InventoryDirectory.__init__ - it isn't right for this
2198
InventoryEntry.__init__(self, file_id, name, parent_id)
2199
self._children = None
2200
self._chk_inventory = chk_inventory
2204
"""Access the list of children of this directory.
2206
With a parent_id_basename_to_file_id index, loads all the children,
2207
without loads the entire index. Without is bad. A more sophisticated
2208
proxy object might be nice, to allow partial loading of children as
2209
well when specific names are accessed. (So path traversal can be
2210
written in the obvious way but not examine siblings.).
2212
if self._children is not None:
2213
return self._children
2214
# No longer supported
2215
if self._chk_inventory.parent_id_basename_to_file_id is None:
2216
raise AssertionError("Inventories without"
2217
" parent_id_basename_to_file_id are no longer supported")
2219
# XXX: Todo - use proxy objects for the children rather than loading
2220
# all when the attribute is referenced.
2221
parent_id_index = self._chk_inventory.parent_id_basename_to_file_id
2223
for (parent_id, name_utf8), file_id in parent_id_index.iteritems(
2224
key_filter=[StaticTuple(self.file_id,)]):
2225
child_keys.add(StaticTuple(file_id,))
2227
for file_id_key in child_keys:
2228
entry = self._chk_inventory._fileid_to_entry_cache.get(
2229
file_id_key[0], None)
2230
if entry is not None:
2231
result[entry.name] = entry
2232
cached.add(file_id_key)
2233
child_keys.difference_update(cached)
2234
# populate; todo: do by name
2235
id_to_entry = self._chk_inventory.id_to_entry
2236
for file_id_key, bytes in id_to_entry.iteritems(child_keys):
2237
entry = self._chk_inventory._bytes_to_entry(bytes)
2238
result[entry.name] = entry
2239
self._chk_inventory._fileid_to_entry_cache[file_id_key[0]] = entry
2240
self._children = result
2244
'directory': InventoryDirectory,
2245
'file': InventoryFile,
2246
'symlink': InventoryLink,
2247
'tree-reference': TreeReference
2250
def make_entry(kind, name, parent_id, file_id=None):
2251
"""Create an inventory entry.
2253
:param kind: the type of inventory entry to create.
2254
:param name: the basename of the entry.
2255
:param parent_id: the parent_id of the entry.
2256
:param file_id: the file_id to use. if None, one will be created.
2259
file_id = generate_ids.gen_file_id(name)
2260
name = ensure_normalized_name(name)
2262
factory = entry_factory[kind]
2264
raise errors.BadFileKindError(name, kind)
2265
return factory(file_id, name, parent_id)
2268
def ensure_normalized_name(name):
2271
:raises InvalidNormalization: When name is not normalized, and cannot be
2272
accessed on this platform by the normalized path.
2273
:return: The NFC normalised version of name.
2275
#------- This has been copied to bzrlib.dirstate.DirState.add, please
2276
# keep them synchronised.
2277
# we dont import normalized_filename directly because we want to be
2278
# able to change the implementation at runtime for tests.
2279
norm_name, can_access = osutils.normalized_filename(name)
2280
if norm_name != name:
2284
# TODO: jam 20060701 This would probably be more useful
2285
# if the error was raised with the full path
2286
raise errors.InvalidNormalization(name)
2290
_NAME_RE = lazy_regex.lazy_compile(r'^[^/\\]+$')
1071
2292
def is_valid_name(name):
1073
if _NAME_RE == None:
1074
_NAME_RE = re.compile(r'^[^/\\]+$')
1076
2293
return bool(_NAME_RE.match(name))
2296
def _check_delta_unique_ids(delta):
2297
"""Decorate a delta and check that the file ids in it are unique.
2299
:return: A generator over delta.
2303
length = len(ids) + 1
2305
if len(ids) != length:
2306
raise errors.InconsistentDelta(item[0] or item[1], item[2],
2311
def _check_delta_unique_new_paths(delta):
2312
"""Decorate a delta and check that the new paths in it are unique.
2314
:return: A generator over delta.
2318
length = len(paths) + 1
2320
if path is not None:
2322
if len(paths) != length:
2323
raise errors.InconsistentDelta(path, item[2], "repeated path")
2327
def _check_delta_unique_old_paths(delta):
2328
"""Decorate a delta and check that the old paths in it are unique.
2330
:return: A generator over delta.
2334
length = len(paths) + 1
2336
if path is not None:
2338
if len(paths) != length:
2339
raise errors.InconsistentDelta(path, item[2], "repeated path")
2343
def _check_delta_ids_are_valid(delta):
2344
"""Decorate a delta and check that the ids in it are valid.
2346
:return: A generator over delta.
2351
raise errors.InconsistentDelta(item[0] or item[1], item[2],
2352
"entry with file_id None %r" % entry)
2353
if type(item[2]) != str:
2354
raise errors.InconsistentDelta(item[0] or item[1], item[2],
2355
"entry with non bytes file_id %r" % entry)
2359
def _check_delta_ids_match_entry(delta):
2360
"""Decorate a delta and check that the ids in it match the entry.file_id.
2362
:return: A generator over delta.
2366
if entry is not None:
2367
if entry.file_id != item[2]:
2368
raise errors.InconsistentDelta(item[0] or item[1], item[2],
2369
"mismatched id with %r" % entry)
2373
def _check_delta_new_path_entry_both_or_None(delta):
2374
"""Decorate a delta and check that the new_path and entry are paired.
2376
:return: A generator over delta.
2381
if new_path is None and entry is not None:
2382
raise errors.InconsistentDelta(item[0], item[1],
2383
"Entry with no new_path")
2384
if new_path is not None and entry is None:
2385
raise errors.InconsistentDelta(new_path, item[1],
2386
"new_path with no entry")
2390
def mutable_inventory_from_tree(tree):
2391
"""Create a new inventory that has the same contents as a specified tree.
2393
:param tree: Revision tree to create inventory from
2395
entries = tree.iter_entries_by_dir()
2396
inv = Inventory(None, tree.get_revision_id())
2397
for path, inv_entry in entries:
2398
inv.add(inv_entry.copy())
added
removed
bzrlib/inventory.py
