1
# Copyright (C) 2005-2010 Canonical Ltd
3
# This program is free software; you can redistribute it and/or modify
4
# it under the terms of the GNU General Public License as published by
5
# the Free Software Foundation; either version 2 of the License, or
6
# (at your option) any later version.
8
# This program is distributed in the hope that it will be useful,
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11
# GNU General Public License for more details.
13
# You should have received a copy of the GNU General Public License
14
# along with this program; if not, write to the Free Software
15
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
17
# FIXME: This refactoring of the workingtree code doesn't seem to keep
18
# the WorkingTree's copy of the inventory in sync with the branch. The
19
# branch modifies its working inventory when it does a commit to make
20
# missing files permanently removed.
22
# TODO: Maybe also keep the full path of the entry, and the children?
23
# But those depend on its position within a particular inventory, and
24
# it would be nice not to need to hold the backpointer here.
26
# This should really be an id randomly assigned when the tree is
27
# created, but it's not for now.
30
from bzrlib.lazy_import import lazy_import
31
lazy_import(globals(), """
46
from bzrlib.errors import (
50
from bzrlib.trace import mutter
51
from bzrlib.static_tuple import StaticTuple
54
class InventoryEntry(object):
55
"""Description of a versioned file.
57
An InventoryEntry has the following fields, which are also
58
present in the XML inventory-entry element:
63
(within the parent directory)
66
file_id of the parent directory, or ROOT_ID
69
the revision_id in which this variation of this file was
73
Indicates that this file should be executable on systems
77
sha-1 of the text of the file
80
size in bytes of the text of the file
82
(reading a version 4 tree created a text_id field.)
87
>>> i.add(InventoryDirectory('123', 'src', ROOT_ID))
88
InventoryDirectory('123', 'src', parent_id='TREE_ROOT', revision=None)
89
>>> i.add(InventoryFile('2323', 'hello.c', parent_id='123'))
90
InventoryFile('2323', 'hello.c', parent_id='123', sha1=None, len=None, revision=None)
91
>>> shouldbe = {0: '', 1: 'src', 2: 'src/hello.c'}
92
>>> for ix, j in enumerate(i.iter_entries()):
93
... print (j[0] == shouldbe[ix], j[1])
95
(True, InventoryDirectory('TREE_ROOT', u'', parent_id=None, revision=None))
96
(True, InventoryDirectory('123', 'src', parent_id='TREE_ROOT', revision=None))
97
(True, InventoryFile('2323', 'hello.c', parent_id='123', sha1=None, len=None, revision=None))
98
>>> i.add(InventoryFile('2324', 'bye.c', '123'))
99
InventoryFile('2324', 'bye.c', parent_id='123', sha1=None, len=None, revision=None)
100
>>> i.add(InventoryDirectory('2325', 'wibble', '123'))
101
InventoryDirectory('2325', 'wibble', parent_id='123', revision=None)
102
>>> i.path2id('src/wibble')
106
>>> i.add(InventoryFile('2326', 'wibble.c', '2325'))
107
InventoryFile('2326', 'wibble.c', parent_id='2325', sha1=None, len=None, revision=None)
109
InventoryFile('2326', 'wibble.c', parent_id='2325', sha1=None, len=None, revision=None)
110
>>> for path, entry in i.iter_entries():
119
>>> i.id2path('2326')
120
'src/wibble/wibble.c'
123
# Constants returned by describe_change()
125
# TODO: These should probably move to some kind of FileChangeDescription
126
# class; that's like what's inside a TreeDelta but we want to be able to
127
# generate them just for one file at a time.
129
MODIFIED_AND_RENAMED = 'modified and renamed'
131
__slots__ = ['file_id', 'revision', 'parent_id', 'name']
133
# Attributes that all InventoryEntry instances are expected to have, but
134
# that don't vary for all kinds of entry. (e.g. symlink_target is only
135
# relevant to InventoryLink, so there's no reason to make every
136
# InventoryFile instance allocate space to hold a value for it.)
137
# Attributes that only vary for files: executable, text_sha1, text_size,
143
# Attributes that only vary for symlinks: symlink_target
144
symlink_target = None
145
# Attributes that only vary for tree-references: reference_revision
146
reference_revision = None
149
def detect_changes(self, old_entry):
150
"""Return a (text_modified, meta_modified) from this to old_entry.
152
_read_tree_state must have been called on self and old_entry prior to
153
calling detect_changes.
157
def _diff(self, text_diff, from_label, tree, to_label, to_entry, to_tree,
158
output_to, reverse=False):
159
"""Perform a diff between two entries of the same kind."""
161
def parent_candidates(self, previous_inventories):
162
"""Find possible per-file graph parents.
164
This is currently defined by:
165
- Select the last changed revision in the parent inventory.
166
- Do deal with a short lived bug in bzr 0.8's development two entries
167
that have the same last changed but different 'x' bit settings are
170
# revision:ie mapping for each ie found in previous_inventories.
172
# identify candidate head revision ids.
173
for inv in previous_inventories:
174
if self.file_id in inv:
175
ie = inv[self.file_id]
176
if ie.revision in candidates:
177
# same revision value in two different inventories:
178
# correct possible inconsistencies:
179
# * there was a bug in revision updates with 'x' bit
182
if candidates[ie.revision].executable != ie.executable:
183
candidates[ie.revision].executable = False
184
ie.executable = False
185
except AttributeError:
188
# add this revision as a candidate.
189
candidates[ie.revision] = ie
193
"""Return true if the object this entry represents has textual data.
195
Note that textual data includes binary content.
197
Also note that all entries get weave files created for them.
198
This attribute is primarily used when upgrading from old trees that
199
did not have the weave index for all inventory entries.
203
def __init__(self, file_id, name, parent_id):
204
"""Create an InventoryEntry
206
The filename must be a single component, relative to the
207
parent directory; it cannot be a whole path or relative name.
209
>>> e = InventoryFile('123', 'hello.c', ROOT_ID)
214
>>> e = InventoryFile('123', 'src/hello.c', ROOT_ID)
215
Traceback (most recent call last):
216
InvalidEntryName: Invalid entry name: src/hello.c
218
if '/' in name or '\\' in name:
219
raise errors.InvalidEntryName(name=name)
220
self.file_id = file_id
223
self.parent_id = parent_id
225
def kind_character(self):
226
"""Return a short kind indicator useful for appending to names."""
227
raise BzrError('unknown kind %r' % self.kind)
229
known_kinds = ('file', 'directory', 'symlink')
231
def sorted_children(self):
232
return sorted(self.children.items())
235
def versionable_kind(kind):
236
return (kind in ('file', 'directory', 'symlink', 'tree-reference'))
238
def check(self, checker, rev_id, inv):
239
"""Check this inventory entry is intact.
241
This is a template method, override _check for kind specific
244
:param checker: Check object providing context for the checks;
245
can be used to find out what parts of the repository have already
247
:param rev_id: Revision id from which this InventoryEntry was loaded.
248
Not necessarily the last-changed revision for this file.
249
:param inv: Inventory from which the entry was loaded.
251
if self.parent_id is not None:
252
if not inv.has_id(self.parent_id):
253
raise BzrCheckError('missing parent {%s} in inventory for revision {%s}'
254
% (self.parent_id, rev_id))
255
checker._add_entry_to_text_key_references(inv, self)
256
self._check(checker, rev_id)
258
def _check(self, checker, rev_id):
259
"""Check this inventory entry for kind specific errors."""
260
checker._report_items.append(
261
'unknown entry kind %r in revision {%s}' % (self.kind, rev_id))
264
"""Clone this inventory entry."""
265
raise NotImplementedError
268
def describe_change(old_entry, new_entry):
269
"""Describe the change between old_entry and this.
271
This smells of being an InterInventoryEntry situation, but as its
272
the first one, we're making it a static method for now.
274
An entry with a different parent, or different name is considered
275
to be renamed. Reparenting is an internal detail.
276
Note that renaming the parent does not trigger a rename for the
279
# TODO: Perhaps return an object rather than just a string
280
if old_entry is new_entry:
281
# also the case of both being None
283
elif old_entry is None:
285
elif new_entry is None:
287
if old_entry.kind != new_entry.kind:
289
text_modified, meta_modified = new_entry.detect_changes(old_entry)
290
if text_modified or meta_modified:
294
# TODO 20060511 (mbp, rbc) factor out 'detect_rename' here.
295
if old_entry.parent_id != new_entry.parent_id:
297
elif old_entry.name != new_entry.name:
301
if renamed and not modified:
302
return InventoryEntry.RENAMED
303
if modified and not renamed:
305
if modified and renamed:
306
return InventoryEntry.MODIFIED_AND_RENAMED
310
return ("%s(%r, %r, parent_id=%r, revision=%r)"
311
% (self.__class__.__name__,
317
def __eq__(self, other):
319
# For the case when objects are cached
321
if not isinstance(other, InventoryEntry):
322
return NotImplemented
324
return ((self.file_id == other.file_id)
325
and (self.name == other.name)
326
and (other.symlink_target == self.symlink_target)
327
and (self.text_sha1 == other.text_sha1)
328
and (self.text_size == other.text_size)
329
and (self.text_id == other.text_id)
330
and (self.parent_id == other.parent_id)
331
and (self.kind == other.kind)
332
and (self.revision == other.revision)
333
and (self.executable == other.executable)
334
and (self.reference_revision == other.reference_revision)
337
def __ne__(self, other):
338
return not (self == other)
341
raise ValueError('not hashable')
343
def _unchanged(self, previous_ie):
344
"""Has this entry changed relative to previous_ie.
346
This method should be overridden in child classes.
349
# different inv parent
350
if previous_ie.parent_id != self.parent_id:
353
elif previous_ie.name != self.name:
355
elif previous_ie.kind != self.kind:
359
def _read_tree_state(self, path, work_tree):
360
"""Populate fields in the inventory entry from the given tree.
362
Note that this should be modified to be a noop on virtual trees
363
as all entries created there are prepopulated.
365
# TODO: Rather than running this manually, we should check the
366
# working sha1 and other expensive properties when they're
367
# first requested, or preload them if they're already known
368
pass # nothing to do by default
370
def _forget_tree_state(self):
374
class InventoryDirectory(InventoryEntry):
375
"""A directory in an inventory."""
377
__slots__ = ['children']
381
def _check(self, checker, rev_id):
382
"""See InventoryEntry._check"""
383
# In non rich root repositories we do not expect a file graph for the
385
if self.name == '' and not checker.rich_roots:
387
# Directories are stored as an empty file, but the file should exist
388
# to provide a per-fileid log. The hash of every directory content is
389
# "da..." below (the sha1sum of '').
390
checker.add_pending_item(rev_id,
391
('texts', self.file_id, self.revision), 'text',
392
'da39a3ee5e6b4b0d3255bfef95601890afd80709')
395
other = InventoryDirectory(self.file_id, self.name, self.parent_id)
396
other.revision = self.revision
397
# note that children are *not* copied; they're pulled across when
401
def __init__(self, file_id, name, parent_id):
402
super(InventoryDirectory, self).__init__(file_id, name, parent_id)
405
def kind_character(self):
406
"""See InventoryEntry.kind_character."""
410
class InventoryFile(InventoryEntry):
411
"""A file in an inventory."""
413
__slots__ = ['text_sha1', 'text_size', 'text_id', 'executable']
417
def __init__(self, file_id, name, parent_id):
418
super(InventoryFile, self).__init__(file_id, name, parent_id)
419
self.text_sha1 = None
420
self.text_size = None
422
self.executable = False
424
def _check(self, checker, tree_revision_id):
425
"""See InventoryEntry._check"""
426
# TODO: check size too.
427
checker.add_pending_item(tree_revision_id,
428
('texts', self.file_id, self.revision), 'text',
430
if self.text_size is None:
431
checker._report_items.append(
432
'fileid {%s} in {%s} has None for text_size' % (self.file_id,
436
other = InventoryFile(self.file_id, self.name, self.parent_id)
437
other.executable = self.executable
438
other.text_id = self.text_id
439
other.text_sha1 = self.text_sha1
440
other.text_size = self.text_size
441
other.revision = self.revision
444
def detect_changes(self, old_entry):
445
"""See InventoryEntry.detect_changes."""
446
text_modified = (self.text_sha1 != old_entry.text_sha1)
447
meta_modified = (self.executable != old_entry.executable)
448
return text_modified, meta_modified
450
def _diff(self, text_diff, from_label, tree, to_label, to_entry, to_tree,
451
output_to, reverse=False):
452
"""See InventoryEntry._diff."""
453
from bzrlib.diff import DiffText
454
from_file_id = self.file_id
456
to_file_id = to_entry.file_id
460
to_file_id, from_file_id = from_file_id, to_file_id
461
tree, to_tree = to_tree, tree
462
from_label, to_label = to_label, from_label
463
differ = DiffText(tree, to_tree, output_to, 'utf-8', '', '',
465
return differ.diff_text(from_file_id, to_file_id, from_label, to_label)
468
"""See InventoryEntry.has_text."""
471
def kind_character(self):
472
"""See InventoryEntry.kind_character."""
475
def _read_tree_state(self, path, work_tree):
476
"""See InventoryEntry._read_tree_state."""
477
self.text_sha1 = work_tree.get_file_sha1(self.file_id, path=path)
478
# FIXME: 20050930 probe for the text size when getting sha1
479
# in _read_tree_state
480
self.executable = work_tree.is_executable(self.file_id, path=path)
483
return ("%s(%r, %r, parent_id=%r, sha1=%r, len=%s, revision=%s)"
484
% (self.__class__.__name__,
492
def _forget_tree_state(self):
493
self.text_sha1 = None
495
def _unchanged(self, previous_ie):
496
"""See InventoryEntry._unchanged."""
497
compatible = super(InventoryFile, self)._unchanged(previous_ie)
498
if self.text_sha1 != previous_ie.text_sha1:
501
# FIXME: 20050930 probe for the text size when getting sha1
502
# in _read_tree_state
503
self.text_size = previous_ie.text_size
504
if self.executable != previous_ie.executable:
509
class InventoryLink(InventoryEntry):
510
"""A file in an inventory."""
512
__slots__ = ['symlink_target']
516
def __init__(self, file_id, name, parent_id):
517
super(InventoryLink, self).__init__(file_id, name, parent_id)
518
self.symlink_target = None
520
def _check(self, checker, tree_revision_id):
521
"""See InventoryEntry._check"""
522
if self.symlink_target is None:
523
checker._report_items.append(
524
'symlink {%s} has no target in revision {%s}'
525
% (self.file_id, tree_revision_id))
526
# Symlinks are stored as ''
527
checker.add_pending_item(tree_revision_id,
528
('texts', self.file_id, self.revision), 'text',
529
'da39a3ee5e6b4b0d3255bfef95601890afd80709')
532
other = InventoryLink(self.file_id, self.name, self.parent_id)
533
other.symlink_target = self.symlink_target
534
other.revision = self.revision
537
def detect_changes(self, old_entry):
538
"""See InventoryEntry.detect_changes."""
539
# FIXME: which _modified field should we use ? RBC 20051003
540
text_modified = (self.symlink_target != old_entry.symlink_target)
542
mutter(" symlink target changed")
543
meta_modified = False
544
return text_modified, meta_modified
546
def _diff(self, text_diff, from_label, tree, to_label, to_entry, to_tree,
547
output_to, reverse=False):
548
"""See InventoryEntry._diff."""
549
from bzrlib.diff import DiffSymlink
550
old_target = self.symlink_target
551
if to_entry is not None:
552
new_target = to_entry.symlink_target
561
new_target, old_target = old_target, new_target
562
differ = DiffSymlink(old_tree, new_tree, output_to)
563
return differ.diff_symlink(old_target, new_target)
565
def kind_character(self):
566
"""See InventoryEntry.kind_character."""
569
def _read_tree_state(self, path, work_tree):
570
"""See InventoryEntry._read_tree_state."""
571
self.symlink_target = work_tree.get_symlink_target(self.file_id)
573
def _forget_tree_state(self):
574
self.symlink_target = None
576
def _unchanged(self, previous_ie):
577
"""See InventoryEntry._unchanged."""
578
compatible = super(InventoryLink, self)._unchanged(previous_ie)
579
if self.symlink_target != previous_ie.symlink_target:
584
class TreeReference(InventoryEntry):
586
__slots__ = ['reference_revision']
588
kind = 'tree-reference'
590
def __init__(self, file_id, name, parent_id, revision=None,
591
reference_revision=None):
592
InventoryEntry.__init__(self, file_id, name, parent_id)
593
self.revision = revision
594
self.reference_revision = reference_revision
597
return TreeReference(self.file_id, self.name, self.parent_id,
598
self.revision, self.reference_revision)
600
def _read_tree_state(self, path, work_tree):
601
"""Populate fields in the inventory entry from the given tree.
603
self.reference_revision = work_tree.get_reference_revision(
606
def _forget_tree_state(self):
607
self.reference_revision = None
609
def _unchanged(self, previous_ie):
610
"""See InventoryEntry._unchanged."""
611
compatible = super(TreeReference, self)._unchanged(previous_ie)
612
if self.reference_revision != previous_ie.reference_revision:
617
class CommonInventory(object):
618
"""Basic inventory logic, defined in terms of primitives like has_id.
620
An inventory is the metadata about the contents of a tree.
622
This is broadly a map from file_id to entries such as directories, files,
623
symlinks and tree references. Each entry maintains its own metadata like
624
SHA1 and length for files, or children for a directory.
626
Entries can be looked up either by path or by file_id.
628
InventoryEntry objects must not be modified after they are
629
inserted, other than through the Inventory API.
632
def __contains__(self, file_id):
633
"""True if this entry contains a file with given id.
635
>>> inv = Inventory()
636
>>> inv.add(InventoryFile('123', 'foo.c', ROOT_ID))
637
InventoryFile('123', 'foo.c', parent_id='TREE_ROOT', sha1=None, len=None, revision=None)
643
Note that this method along with __iter__ are not encouraged for use as
644
they are less clear than specific query methods - they may be rmeoved
647
return self.has_id(file_id)
649
def has_filename(self, filename):
650
return bool(self.path2id(filename))
652
def id2path(self, file_id):
653
"""Return as a string the path to file_id.
656
>>> e = i.add(InventoryDirectory('src-id', 'src', ROOT_ID))
657
>>> e = i.add(InventoryFile('foo-id', 'foo.c', parent_id='src-id'))
658
>>> print i.id2path('foo-id')
661
:raises NoSuchId: If file_id is not present in the inventory.
663
# get all names, skipping root
664
return '/'.join(reversed(
665
[parent.name for parent in
666
self._iter_file_id_parents(file_id)][:-1]))
668
def iter_entries(self, from_dir=None, recursive=True):
669
"""Return (path, entry) pairs, in order by name.
671
:param from_dir: if None, start from the root,
672
otherwise start from this directory (either file-id or entry)
673
:param recursive: recurse into directories or not
676
if self.root is None:
680
elif isinstance(from_dir, basestring):
681
from_dir = self[from_dir]
683
# unrolling the recursive called changed the time from
684
# 440ms/663ms (inline/total) to 116ms/116ms
685
children = from_dir.children.items()
688
for name, ie in children:
691
children = collections.deque(children)
692
stack = [(u'', children)]
694
from_dir_relpath, children = stack[-1]
697
name, ie = children.popleft()
699
# we know that from_dir_relpath never ends in a slash
700
# and 'f' doesn't begin with one, we can do a string op, rather
701
# than the checks of pathjoin(), though this means that all paths
703
path = from_dir_relpath + '/' + name
707
if ie.kind != 'directory':
710
# But do this child first
711
new_children = ie.children.items()
713
new_children = collections.deque(new_children)
714
stack.append((path, new_children))
715
# Break out of inner loop, so that we start outer loop with child
718
# if we finished all children, pop it off the stack
721
def iter_entries_by_dir(self, from_dir=None, specific_file_ids=None,
722
yield_parents=False):
723
"""Iterate over the entries in a directory first order.
725
This returns all entries for a directory before returning
726
the entries for children of a directory. This is not
727
lexicographically sorted order, and is a hybrid between
728
depth-first and breadth-first.
730
:param yield_parents: If True, yield the parents from the root leading
731
down to specific_file_ids that have been requested. This has no
732
impact if specific_file_ids is None.
733
:return: This yields (path, entry) pairs
735
if specific_file_ids and not isinstance(specific_file_ids, set):
736
specific_file_ids = set(specific_file_ids)
737
# TODO? Perhaps this should return the from_dir so that the root is
738
# yielded? or maybe an option?
740
if self.root is None:
742
# Optimize a common case
743
if (not yield_parents and specific_file_ids is not None and
744
len(specific_file_ids) == 1):
745
file_id = list(specific_file_ids)[0]
747
yield self.id2path(file_id), self[file_id]
750
if (specific_file_ids is None or yield_parents or
751
self.root.file_id in specific_file_ids):
753
elif isinstance(from_dir, basestring):
754
from_dir = self[from_dir]
756
if specific_file_ids is not None:
757
# TODO: jam 20070302 This could really be done as a loop rather
758
# than a bunch of recursive calls.
761
def add_ancestors(file_id):
762
if file_id not in byid:
764
parent_id = byid[file_id].parent_id
765
if parent_id is None:
767
if parent_id not in parents:
768
parents.add(parent_id)
769
add_ancestors(parent_id)
770
for file_id in specific_file_ids:
771
add_ancestors(file_id)
775
stack = [(u'', from_dir)]
777
cur_relpath, cur_dir = stack.pop()
780
for child_name, child_ie in sorted(cur_dir.children.iteritems()):
782
child_relpath = cur_relpath + child_name
784
if (specific_file_ids is None or
785
child_ie.file_id in specific_file_ids or
786
(yield_parents and child_ie.file_id in parents)):
787
yield child_relpath, child_ie
789
if child_ie.kind == 'directory':
790
if parents is None or child_ie.file_id in parents:
791
child_dirs.append((child_relpath+'/', child_ie))
792
stack.extend(reversed(child_dirs))
794
def _make_delta(self, old):
795
"""Make an inventory delta from two inventories."""
798
adds = new_ids - old_ids
799
deletes = old_ids - new_ids
800
common = old_ids.intersection(new_ids)
802
for file_id in deletes:
803
delta.append((old.id2path(file_id), None, file_id, None))
805
delta.append((None, self.id2path(file_id), file_id, self[file_id]))
806
for file_id in common:
807
if old[file_id] != self[file_id]:
808
delta.append((old.id2path(file_id), self.id2path(file_id),
809
file_id, self[file_id]))
812
def _get_mutable_inventory(self):
813
"""Returns a mutable copy of the object.
815
Some inventories are immutable, yet working trees, for example, needs
816
to mutate exisiting inventories instead of creating a new one.
818
raise NotImplementedError(self._get_mutable_inventory)
820
def make_entry(self, kind, name, parent_id, file_id=None):
821
"""Simple thunk to bzrlib.inventory.make_entry."""
822
return make_entry(kind, name, parent_id, file_id)
825
"""Return list of (path, ie) for all entries except the root.
827
This may be faster than iter_entries.
830
def descend(dir_ie, dir_path):
831
kids = dir_ie.children.items()
833
for name, ie in kids:
834
child_path = osutils.pathjoin(dir_path, name)
835
accum.append((child_path, ie))
836
if ie.kind == 'directory':
837
descend(ie, child_path)
839
descend(self.root, u'')
842
def directories(self):
843
"""Return (path, entry) pairs for all directories, including the root.
846
def descend(parent_ie, parent_path):
847
accum.append((parent_path, parent_ie))
849
kids = [(ie.name, ie) for ie in parent_ie.children.itervalues() if ie.kind == 'directory']
852
for name, child_ie in kids:
853
child_path = osutils.pathjoin(parent_path, name)
854
descend(child_ie, child_path)
855
descend(self.root, u'')
858
def path2id(self, relpath):
859
"""Walk down through directories to return entry of last component.
861
:param relpath: may be either a list of path components, or a single
862
string, in which case it is automatically split.
864
This returns the entry of the last component in the path,
865
which may be either a file or a directory.
867
Returns None IFF the path is not found.
869
if isinstance(relpath, basestring):
870
names = osutils.splitpath(relpath)
876
except errors.NoSuchId:
877
# root doesn't exist yet so nothing else can
883
children = getattr(parent, 'children', None)
892
return parent.file_id
894
def filter(self, specific_fileids):
895
"""Get an inventory view filtered against a set of file-ids.
897
Children of directories and parents are included.
899
The result may or may not reference the underlying inventory
900
so it should be treated as immutable.
902
interesting_parents = set()
903
for fileid in specific_fileids:
905
interesting_parents.update(self.get_idpath(fileid))
906
except errors.NoSuchId:
907
# This fileid is not in the inventory - that's ok
909
entries = self.iter_entries()
910
if self.root is None:
911
return Inventory(root_id=None)
912
other = Inventory(entries.next()[1].file_id)
913
other.root.revision = self.root.revision
914
other.revision_id = self.revision_id
915
directories_to_expand = set()
916
for path, entry in entries:
917
file_id = entry.file_id
918
if (file_id in specific_fileids
919
or entry.parent_id in directories_to_expand):
920
if entry.kind == 'directory':
921
directories_to_expand.add(file_id)
922
elif file_id not in interesting_parents:
924
other.add(entry.copy())
927
def get_idpath(self, file_id):
928
"""Return a list of file_ids for the path to an entry.
930
The list contains one element for each directory followed by
931
the id of the file itself. So the length of the returned list
932
is equal to the depth of the file in the tree, counting the
933
root directory as depth 1.
936
for parent in self._iter_file_id_parents(file_id):
937
p.insert(0, parent.file_id)
941
class Inventory(CommonInventory):
942
"""Mutable dict based in-memory inventory.
944
We never store the full path to a file, because renaming a directory
945
implicitly moves all of its contents. This class internally maintains a
946
lookup tree that allows the children under a directory to be
949
>>> inv = Inventory()
950
>>> inv.add(InventoryFile('123-123', 'hello.c', ROOT_ID))
951
InventoryFile('123-123', 'hello.c', parent_id='TREE_ROOT', sha1=None, len=None, revision=None)
952
>>> inv['123-123'].name
955
Id's may be looked up from paths:
957
>>> inv.path2id('hello.c')
962
There are iterators over the contents:
964
>>> [entry[0] for entry in inv.iter_entries()]
968
def __init__(self, root_id=ROOT_ID, revision_id=None):
969
"""Create or read an inventory.
971
If a working directory is specified, the inventory is read
972
from there. If the file is specified, read from that. If not,
973
the inventory is created empty.
975
The inventory is created with a default root directory, with
978
if root_id is not None:
979
self._set_root(InventoryDirectory(root_id, u'', None))
983
self.revision_id = revision_id
986
# More than one page of ouput is not useful anymore to debug
989
contents = repr(self._byid)
990
if len(contents) > max_len:
991
contents = contents[:(max_len-len(closing))] + closing
992
return "<Inventory object at %x, contents=%r>" % (id(self), contents)
994
def apply_delta(self, delta):
995
"""Apply a delta to this inventory.
997
See the inventory developers documentation for the theory behind
1000
If delta application fails the inventory is left in an indeterminate
1001
state and must not be used.
1003
:param delta: A list of changes to apply. After all the changes are
1004
applied the final inventory must be internally consistent, but it
1005
is ok to supply changes which, if only half-applied would have an
1006
invalid result - such as supplying two changes which rename two
1007
files, 'A' and 'B' with each other : [('A', 'B', 'A-id', a_entry),
1008
('B', 'A', 'B-id', b_entry)].
1010
Each change is a tuple, of the form (old_path, new_path, file_id,
1013
When new_path is None, the change indicates the removal of an entry
1014
from the inventory and new_entry will be ignored (using None is
1015
appropriate). If new_path is not None, then new_entry must be an
1016
InventoryEntry instance, which will be incorporated into the
1017
inventory (and replace any existing entry with the same file id).
1019
When old_path is None, the change indicates the addition of
1020
a new entry to the inventory.
1022
When neither new_path nor old_path are None, the change is a
1023
modification to an entry, such as a rename, reparent, kind change
1026
The children attribute of new_entry is ignored. This is because
1027
this method preserves children automatically across alterations to
1028
the parent of the children, and cases where the parent id of a
1029
child is changing require the child to be passed in as a separate
1030
change regardless. E.g. in the recursive deletion of a directory -
1031
the directory's children must be included in the delta, or the
1032
final inventory will be invalid.
1034
Note that a file_id must only appear once within a given delta.
1035
An AssertionError is raised otherwise.
1037
# Check that the delta is legal. It would be nice if this could be
1038
# done within the loops below but it's safer to validate the delta
1039
# before starting to mutate the inventory, as there isn't a rollback
1041
list(_check_delta_unique_ids(_check_delta_unique_new_paths(
1042
_check_delta_unique_old_paths(_check_delta_ids_match_entry(
1043
_check_delta_ids_are_valid(
1044
_check_delta_new_path_entry_both_or_None(
1048
# Remove all affected items which were in the original inventory,
1049
# starting with the longest paths, thus ensuring parents are examined
1050
# after their children, which means that everything we examine has no
1051
# modified children remaining by the time we examine it.
1052
for old_path, file_id in sorted(((op, f) for op, np, f, e in delta
1053
if op is not None), reverse=True):
1054
# Preserve unaltered children of file_id for later reinsertion.
1055
file_id_children = getattr(self[file_id], 'children', {})
1056
if len(file_id_children):
1057
children[file_id] = file_id_children
1058
if self.id2path(file_id) != old_path:
1059
raise errors.InconsistentDelta(old_path, file_id,
1060
"Entry was at wrong other path %r." % self.id2path(file_id))
1061
# Remove file_id and the unaltered children. If file_id is not
1062
# being deleted it will be reinserted back later.
1063
self.remove_recursive_id(file_id)
1064
# Insert all affected which should be in the new inventory, reattaching
1065
# their children if they had any. This is done from shortest path to
1066
# longest, ensuring that items which were modified and whose parents in
1067
# the resulting inventory were also modified, are inserted after their
1069
for new_path, f, new_entry in sorted((np, f, e) for op, np, f, e in
1070
delta if np is not None):
1071
if new_entry.kind == 'directory':
1072
# Pop the child which to allow detection of children whose
1073
# parents were deleted and which were not reattached to a new
1075
replacement = InventoryDirectory(new_entry.file_id,
1076
new_entry.name, new_entry.parent_id)
1077
replacement.revision = new_entry.revision
1078
replacement.children = children.pop(replacement.file_id, {})
1079
new_entry = replacement
1082
except errors.DuplicateFileId:
1083
raise errors.InconsistentDelta(new_path, new_entry.file_id,
1084
"New id is already present in target.")
1085
except AttributeError:
1086
raise errors.InconsistentDelta(new_path, new_entry.file_id,
1087
"Parent is not a directory.")
1088
if self.id2path(new_entry.file_id) != new_path:
1089
raise errors.InconsistentDelta(new_path, new_entry.file_id,
1090
"New path is not consistent with parent path.")
1092
# Get the parent id that was deleted
1093
parent_id, children = children.popitem()
1094
raise errors.InconsistentDelta("<deleted>", parent_id,
1095
"The file id was deleted but its children were not deleted.")
1097
def create_by_apply_delta(self, inventory_delta, new_revision_id,
1098
propagate_caches=False):
1099
"""See CHKInventory.create_by_apply_delta()"""
1100
new_inv = self.copy()
1101
new_inv.apply_delta(inventory_delta)
1102
new_inv.revision_id = new_revision_id
1105
def _set_root(self, ie):
1107
self._byid = {self.root.file_id: self.root}
1110
# TODO: jam 20051218 Should copy also copy the revision_id?
1111
entries = self.iter_entries()
1112
if self.root is None:
1113
return Inventory(root_id=None)
1114
other = Inventory(entries.next()[1].file_id)
1115
other.root.revision = self.root.revision
1116
# copy recursively so we know directories will be added before
1117
# their children. There are more efficient ways than this...
1118
for path, entry in entries:
1119
other.add(entry.copy())
1122
def _get_mutable_inventory(self):
1123
"""See CommonInventory._get_mutable_inventory."""
1124
return copy.deepcopy(self)
1127
"""Iterate over all file-ids."""
1128
return iter(self._byid)
1130
def iter_just_entries(self):
1131
"""Iterate over all entries.
1133
Unlike iter_entries(), just the entries are returned (not (path, ie))
1134
and the order of entries is undefined.
1136
XXX: We may not want to merge this into bzr.dev.
1138
if self.root is None:
1140
for _, ie in self._byid.iteritems():
1144
"""Returns number of entries."""
1145
return len(self._byid)
1147
def __getitem__(self, file_id):
1148
"""Return the entry for given file_id.
1150
>>> inv = Inventory()
1151
>>> inv.add(InventoryFile('123123', 'hello.c', ROOT_ID))
1152
InventoryFile('123123', 'hello.c', parent_id='TREE_ROOT', sha1=None, len=None, revision=None)
1153
>>> inv['123123'].name
1157
return self._byid[file_id]
1159
# really we're passing an inventory, not a tree...
1160
raise errors.NoSuchId(self, file_id)
1162
def get_file_kind(self, file_id):
1163
return self._byid[file_id].kind
1165
def get_child(self, parent_id, filename):
1166
return self[parent_id].children.get(filename)
1168
def _add_child(self, entry):
1169
"""Add an entry to the inventory, without adding it to its parent"""
1170
if entry.file_id in self._byid:
1171
raise BzrError("inventory already contains entry with id {%s}" %
1173
self._byid[entry.file_id] = entry
1174
for child in getattr(entry, 'children', {}).itervalues():
1175
self._add_child(child)
1178
def add(self, entry):
1179
"""Add entry to inventory.
1181
To add a file to a branch ready to be committed, use Branch.add,
1186
if entry.file_id in self._byid:
1187
raise errors.DuplicateFileId(entry.file_id,
1188
self._byid[entry.file_id])
1189
if entry.parent_id is None:
1193
parent = self._byid[entry.parent_id]
1195
raise errors.InconsistentDelta("<unknown>", entry.parent_id,
1196
"Parent not in inventory.")
1197
if entry.name in parent.children:
1198
raise errors.InconsistentDelta(
1199
self.id2path(parent.children[entry.name].file_id),
1201
"Path already versioned")
1202
parent.children[entry.name] = entry
1203
return self._add_child(entry)
1205
def add_path(self, relpath, kind, file_id=None, parent_id=None):
1206
"""Add entry from a path.
1208
The immediate parent must already be versioned.
1210
Returns the new entry object."""
1212
parts = osutils.splitpath(relpath)
1216
file_id = generate_ids.gen_root_id()
1217
self.root = InventoryDirectory(file_id, '', None)
1218
self._byid = {self.root.file_id: self.root}
1221
parent_path = parts[:-1]
1222
parent_id = self.path2id(parent_path)
1223
if parent_id is None:
1224
raise errors.NotVersionedError(path=parent_path)
1225
ie = make_entry(kind, parts[-1], parent_id, file_id)
1228
def __delitem__(self, file_id):
1229
"""Remove entry by id.
1231
>>> inv = Inventory()
1232
>>> inv.add(InventoryFile('123', 'foo.c', ROOT_ID))
1233
InventoryFile('123', 'foo.c', parent_id='TREE_ROOT', sha1=None, len=None, revision=None)
1241
del self._byid[file_id]
1242
if ie.parent_id is not None:
1243
del self[ie.parent_id].children[ie.name]
1245
def __eq__(self, other):
1246
"""Compare two sets by comparing their contents.
1248
>>> i1 = Inventory()
1249
>>> i2 = Inventory()
1252
>>> i1.add(InventoryFile('123', 'foo', ROOT_ID))
1253
InventoryFile('123', 'foo', parent_id='TREE_ROOT', sha1=None, len=None, revision=None)
1256
>>> i2.add(InventoryFile('123', 'foo', ROOT_ID))
1257
InventoryFile('123', 'foo', parent_id='TREE_ROOT', sha1=None, len=None, revision=None)
1261
if not isinstance(other, Inventory):
1262
return NotImplemented
1264
return self._byid == other._byid
1266
def __ne__(self, other):
1267
return not self.__eq__(other)
1270
raise ValueError('not hashable')
1272
def _iter_file_id_parents(self, file_id):
1273
"""Yield the parents of file_id up to the root."""
1274
while file_id is not None:
1276
ie = self._byid[file_id]
1278
raise errors.NoSuchId(tree=None, file_id=file_id)
1280
file_id = ie.parent_id
1282
def has_id(self, file_id):
1283
return (file_id in self._byid)
1285
def _make_delta(self, old):
1286
"""Make an inventory delta from two inventories."""
1287
old_getter = getattr(old, '_byid', old)
1288
new_getter = self._byid
1289
old_ids = set(old_getter)
1290
new_ids = set(new_getter)
1291
adds = new_ids - old_ids
1292
deletes = old_ids - new_ids
1293
if not adds and not deletes:
1296
common = old_ids.intersection(new_ids)
1298
for file_id in deletes:
1299
delta.append((old.id2path(file_id), None, file_id, None))
1300
for file_id in adds:
1301
delta.append((None, self.id2path(file_id), file_id, self[file_id]))
1302
for file_id in common:
1303
new_ie = new_getter[file_id]
1304
old_ie = old_getter[file_id]
1305
# If xml_serializer returns the cached InventoryEntries (rather
1306
# than always doing .copy()), inlining the 'is' check saves 2.7M
1307
# calls to __eq__. Under lsprof this saves 20s => 6s.
1308
# It is a minor improvement without lsprof.
1309
if old_ie is new_ie or old_ie == new_ie:
1312
delta.append((old.id2path(file_id), self.id2path(file_id),
1316
def remove_recursive_id(self, file_id):
1317
"""Remove file_id, and children, from the inventory.
1319
:param file_id: A file_id to remove.
1321
to_find_delete = [self._byid[file_id]]
1323
while to_find_delete:
1324
ie = to_find_delete.pop()
1325
to_delete.append(ie.file_id)
1326
if ie.kind == 'directory':
1327
to_find_delete.extend(ie.children.values())
1328
for file_id in reversed(to_delete):
1330
del self._byid[file_id]
1331
if ie.parent_id is not None:
1332
del self[ie.parent_id].children[ie.name]
1336
def rename(self, file_id, new_parent_id, new_name):
1337
"""Move a file within the inventory.
1339
This can change either the name, or the parent, or both.
1341
This does not move the working file.
1343
new_name = ensure_normalized_name(new_name)
1344
if not is_valid_name(new_name):
1345
raise BzrError("not an acceptable filename: %r" % new_name)
1347
new_parent = self._byid[new_parent_id]
1348
if new_name in new_parent.children:
1349
raise BzrError("%r already exists in %r" % (new_name, self.id2path(new_parent_id)))
1351
new_parent_idpath = self.get_idpath(new_parent_id)
1352
if file_id in new_parent_idpath:
1353
raise BzrError("cannot move directory %r into a subdirectory of itself, %r"
1354
% (self.id2path(file_id), self.id2path(new_parent_id)))
1356
file_ie = self._byid[file_id]
1357
old_parent = self._byid[file_ie.parent_id]
1359
# TODO: Don't leave things messed up if this fails
1361
del old_parent.children[file_ie.name]
1362
new_parent.children[new_name] = file_ie
1364
file_ie.name = new_name
1365
file_ie.parent_id = new_parent_id
1367
def is_root(self, file_id):
1368
return self.root is not None and file_id == self.root.file_id
1371
class CHKInventory(CommonInventory):
1372
"""An inventory persisted in a CHK store.
1374
By design, a CHKInventory is immutable so many of the methods
1375
supported by Inventory - add, rename, apply_delta, etc - are *not*
1376
supported. To create a new CHKInventory, use create_by_apply_delta()
1377
or from_inventory(), say.
1379
Internally, a CHKInventory has one or two CHKMaps:
1381
* id_to_entry - a map from (file_id,) => InventoryEntry as bytes
1382
* parent_id_basename_to_file_id - a map from (parent_id, basename_utf8)
1385
The second map is optional and not present in early CHkRepository's.
1387
No caching is performed: every method call or item access will perform
1388
requests to the storage layer. As such, keep references to objects you
1392
def __init__(self, search_key_name):
1393
CommonInventory.__init__(self)
1394
self._fileid_to_entry_cache = {}
1395
self._path_to_fileid_cache = {}
1396
self._search_key_name = search_key_name
1399
def __eq__(self, other):
1400
"""Compare two sets by comparing their contents."""
1401
if not isinstance(other, CHKInventory):
1402
return NotImplemented
1404
this_key = self.id_to_entry.key()
1405
other_key = other.id_to_entry.key()
1406
this_pid_key = self.parent_id_basename_to_file_id.key()
1407
other_pid_key = other.parent_id_basename_to_file_id.key()
1408
if None in (this_key, this_pid_key, other_key, other_pid_key):
1410
return this_key == other_key and this_pid_key == other_pid_key
1412
def _entry_to_bytes(self, entry):
1413
"""Serialise entry as a single bytestring.
1415
:param Entry: An inventory entry.
1416
:return: A bytestring for the entry.
1419
ENTRY ::= FILE | DIR | SYMLINK | TREE
1420
FILE ::= "file: " COMMON SEP SHA SEP SIZE SEP EXECUTABLE
1421
DIR ::= "dir: " COMMON
1422
SYMLINK ::= "symlink: " COMMON SEP TARGET_UTF8
1423
TREE ::= "tree: " COMMON REFERENCE_REVISION
1424
COMMON ::= FILE_ID SEP PARENT_ID SEP NAME_UTF8 SEP REVISION
1427
if entry.parent_id is not None:
1428
parent_str = entry.parent_id
1431
name_str = entry.name.encode("utf8")
1432
if entry.kind == 'file':
1433
if entry.executable:
1437
return "file: %s\n%s\n%s\n%s\n%s\n%d\n%s" % (
1438
entry.file_id, parent_str, name_str, entry.revision,
1439
entry.text_sha1, entry.text_size, exec_str)
1440
elif entry.kind == 'directory':
1441
return "dir: %s\n%s\n%s\n%s" % (
1442
entry.file_id, parent_str, name_str, entry.revision)
1443
elif entry.kind == 'symlink':
1444
return "symlink: %s\n%s\n%s\n%s\n%s" % (
1445
entry.file_id, parent_str, name_str, entry.revision,
1446
entry.symlink_target.encode("utf8"))
1447
elif entry.kind == 'tree-reference':
1448
return "tree: %s\n%s\n%s\n%s\n%s" % (
1449
entry.file_id, parent_str, name_str, entry.revision,
1450
entry.reference_revision)
1452
raise ValueError("unknown kind %r" % entry.kind)
1454
def _expand_fileids_to_parents_and_children(self, file_ids):
1455
"""Give a more wholistic view starting with the given file_ids.
1457
For any file_id which maps to a directory, we will include all children
1458
of that directory. We will also include all directories which are
1459
parents of the given file_ids, but we will not include their children.
1466
fringle # fringle-id
1470
if given [foo-id] we will include
1471
TREE_ROOT as interesting parents
1473
foo-id, baz-id, frob-id, fringle-id
1477
# TODO: Pre-pass over the list of fileids to see if anything is already
1478
# deserialized in self._fileid_to_entry_cache
1480
directories_to_expand = set()
1481
children_of_parent_id = {}
1482
# It is okay if some of the fileids are missing
1483
for entry in self._getitems(file_ids):
1484
if entry.kind == 'directory':
1485
directories_to_expand.add(entry.file_id)
1486
interesting.add(entry.parent_id)
1487
children_of_parent_id.setdefault(entry.parent_id, []
1488
).append(entry.file_id)
1490
# Now, interesting has all of the direct parents, but not the
1491
# parents of those parents. It also may have some duplicates with
1493
remaining_parents = interesting.difference(file_ids)
1494
# When we hit the TREE_ROOT, we'll get an interesting parent of None,
1495
# but we don't actually want to recurse into that
1496
interesting.add(None) # this will auto-filter it in the loop
1497
remaining_parents.discard(None)
1498
while remaining_parents:
1499
next_parents = set()
1500
for entry in self._getitems(remaining_parents):
1501
next_parents.add(entry.parent_id)
1502
children_of_parent_id.setdefault(entry.parent_id, []
1503
).append(entry.file_id)
1504
# Remove any search tips we've already processed
1505
remaining_parents = next_parents.difference(interesting)
1506
interesting.update(remaining_parents)
1507
# We should probably also .difference(directories_to_expand)
1508
interesting.update(file_ids)
1509
interesting.discard(None)
1510
while directories_to_expand:
1511
# Expand directories by looking in the
1512
# parent_id_basename_to_file_id map
1513
keys = [StaticTuple(f,).intern() for f in directories_to_expand]
1514
directories_to_expand = set()
1515
items = self.parent_id_basename_to_file_id.iteritems(keys)
1516
next_file_ids = set([item[1] for item in items])
1517
next_file_ids = next_file_ids.difference(interesting)
1518
interesting.update(next_file_ids)
1519
for entry in self._getitems(next_file_ids):
1520
if entry.kind == 'directory':
1521
directories_to_expand.add(entry.file_id)
1522
children_of_parent_id.setdefault(entry.parent_id, []
1523
).append(entry.file_id)
1524
return interesting, children_of_parent_id
1526
def filter(self, specific_fileids):
1527
"""Get an inventory view filtered against a set of file-ids.
1529
Children of directories and parents are included.
1531
The result may or may not reference the underlying inventory
1532
so it should be treated as immutable.
1535
parent_to_children) = self._expand_fileids_to_parents_and_children(
1537
# There is some overlap here, but we assume that all interesting items
1538
# are in the _fileid_to_entry_cache because we had to read them to
1539
# determine if they were a dir we wanted to recurse, or just a file
1540
# This should give us all the entries we'll want to add, so start
1542
other = Inventory(self.root_id)
1543
other.root.revision = self.root.revision
1544
other.revision_id = self.revision_id
1545
if not interesting or not parent_to_children:
1546
# empty filter, or filtering entrys that don't exist
1547
# (if even 1 existed, then we would have populated
1548
# parent_to_children with at least the tree root.)
1550
cache = self._fileid_to_entry_cache
1551
remaining_children = collections.deque(parent_to_children[self.root_id])
1552
while remaining_children:
1553
file_id = remaining_children.popleft()
1555
if ie.kind == 'directory':
1556
ie = ie.copy() # We create a copy to depopulate the .children attribute
1557
# TODO: depending on the uses of 'other' we should probably alwyas
1558
# '.copy()' to prevent someone from mutating other and
1559
# invaliding our internal cache
1561
if file_id in parent_to_children:
1562
remaining_children.extend(parent_to_children[file_id])
1566
def _bytes_to_utf8name_key(bytes):
1567
"""Get the file_id, revision_id key out of bytes."""
1568
# We don't normally care about name, except for times when we want
1569
# to filter out empty names because of non rich-root...
1570
sections = bytes.split('\n')
1571
kind, file_id = sections[0].split(': ')
1572
return (sections[2], intern(file_id), intern(sections[3]))
1574
def _bytes_to_entry(self, bytes):
1575
"""Deserialise a serialised entry."""
1576
sections = bytes.split('\n')
1577
if sections[0].startswith("file: "):
1578
result = InventoryFile(sections[0][6:],
1579
sections[2].decode('utf8'),
1581
result.text_sha1 = sections[4]
1582
result.text_size = int(sections[5])
1583
result.executable = sections[6] == "Y"
1584
elif sections[0].startswith("dir: "):
1585
result = CHKInventoryDirectory(sections[0][5:],
1586
sections[2].decode('utf8'),
1588
elif sections[0].startswith("symlink: "):
1589
result = InventoryLink(sections[0][9:],
1590
sections[2].decode('utf8'),
1592
result.symlink_target = sections[4].decode('utf8')
1593
elif sections[0].startswith("tree: "):
1594
result = TreeReference(sections[0][6:],
1595
sections[2].decode('utf8'),
1597
result.reference_revision = sections[4]
1599
raise ValueError("Not a serialised entry %r" % bytes)
1600
result.file_id = intern(result.file_id)
1601
result.revision = intern(sections[3])
1602
if result.parent_id == '':
1603
result.parent_id = None
1604
self._fileid_to_entry_cache[result.file_id] = result
1607
def _get_mutable_inventory(self):
1608
"""See CommonInventory._get_mutable_inventory."""
1609
entries = self.iter_entries()
1610
inv = Inventory(None, self.revision_id)
1611
for path, inv_entry in entries:
1612
inv.add(inv_entry.copy())
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 iter_changes(self, basis):
1976
"""Generate a Tree.iter_changes change list between this and basis.
1978
:param basis: Another CHKInventory.
1979
:return: An iterator over the changes between self and basis, as per
1980
tree.iter_changes().
1982
# We want: (file_id, (path_in_source, path_in_target),
1983
# changed_content, versioned, parent, name, kind,
1985
for key, basis_value, self_value in \
1986
self.id_to_entry.iter_changes(basis.id_to_entry):
1988
if basis_value is not None:
1989
basis_entry = basis._bytes_to_entry(basis_value)
1990
path_in_source = basis.id2path(file_id)
1991
basis_parent = basis_entry.parent_id
1992
basis_name = basis_entry.name
1993
basis_executable = basis_entry.executable
1995
path_in_source = None
1998
basis_executable = None
1999
if self_value is not None:
2000
self_entry = self._bytes_to_entry(self_value)
2001
path_in_target = self.id2path(file_id)
2002
self_parent = self_entry.parent_id
2003
self_name = self_entry.name
2004
self_executable = self_entry.executable
2006
path_in_target = None
2009
self_executable = None
2010
if basis_value is None:
2012
kind = (None, self_entry.kind)
2013
versioned = (False, True)
2014
elif self_value is None:
2016
kind = (basis_entry.kind, None)
2017
versioned = (True, False)
2019
kind = (basis_entry.kind, self_entry.kind)
2020
versioned = (True, True)
2021
changed_content = False
2022
if kind[0] != kind[1]:
2023
changed_content = True
2024
elif kind[0] == 'file':
2025
if (self_entry.text_size != basis_entry.text_size or
2026
self_entry.text_sha1 != basis_entry.text_sha1):
2027
changed_content = True
2028
elif kind[0] == 'symlink':
2029
if self_entry.symlink_target != basis_entry.symlink_target:
2030
changed_content = True
2031
elif kind[0] == 'tree-reference':
2032
if (self_entry.reference_revision !=
2033
basis_entry.reference_revision):
2034
changed_content = True
2035
parent = (basis_parent, self_parent)
2036
name = (basis_name, self_name)
2037
executable = (basis_executable, self_executable)
2038
if (not changed_content
2039
and parent[0] == parent[1]
2040
and name[0] == name[1]
2041
and executable[0] == executable[1]):
2042
# Could happen when only the revision changed for a directory
2045
yield (file_id, (path_in_source, path_in_target), changed_content,
2046
versioned, parent, name, kind, executable)
2049
"""Return the number of entries in the inventory."""
2050
return len(self.id_to_entry)
2052
def _make_delta(self, old):
2053
"""Make an inventory delta from two inventories."""
2054
if type(old) != CHKInventory:
2055
return CommonInventory._make_delta(self, old)
2057
for key, old_value, self_value in \
2058
self.id_to_entry.iter_changes(old.id_to_entry):
2060
if old_value is not None:
2061
old_path = old.id2path(file_id)
2064
if self_value is not None:
2065
entry = self._bytes_to_entry(self_value)
2066
self._fileid_to_entry_cache[file_id] = entry
2067
new_path = self.id2path(file_id)
2071
delta.append((old_path, new_path, file_id, entry))
2074
def path2id(self, relpath):
2075
"""See CommonInventory.path2id()."""
2076
# TODO: perhaps support negative hits?
2077
result = self._path_to_fileid_cache.get(relpath, None)
2078
if result is not None:
2080
if isinstance(relpath, basestring):
2081
names = osutils.splitpath(relpath)
2084
current_id = self.root_id
2085
if current_id is None:
2087
parent_id_index = self.parent_id_basename_to_file_id
2089
for basename in names:
2090
if cur_path is None:
2093
cur_path = cur_path + '/' + basename
2094
basename_utf8 = basename.encode('utf8')
2095
file_id = self._path_to_fileid_cache.get(cur_path, None)
2097
key_filter = [StaticTuple(current_id, basename_utf8)]
2098
items = parent_id_index.iteritems(key_filter)
2099
for (parent_id, name_utf8), file_id in items:
2100
if parent_id != current_id or name_utf8 != basename_utf8:
2101
raise errors.BzrError("corrupt inventory lookup! "
2102
"%r %r %r %r" % (parent_id, current_id, name_utf8,
2107
self._path_to_fileid_cache[cur_path] = file_id
2108
current_id = file_id
2112
"""Serialise the inventory to lines."""
2113
lines = ["chkinventory:\n"]
2114
if self._search_key_name != 'plain':
2115
# custom ordering grouping things that don't change together
2116
lines.append('search_key_name: %s\n' % (self._search_key_name,))
2117
lines.append("root_id: %s\n" % self.root_id)
2118
lines.append('parent_id_basename_to_file_id: %s\n' %
2119
(self.parent_id_basename_to_file_id.key()[0],))
2120
lines.append("revision_id: %s\n" % self.revision_id)
2121
lines.append("id_to_entry: %s\n" % (self.id_to_entry.key()[0],))
2123
lines.append("revision_id: %s\n" % self.revision_id)
2124
lines.append("root_id: %s\n" % self.root_id)
2125
if self.parent_id_basename_to_file_id is not None:
2126
lines.append('parent_id_basename_to_file_id: %s\n' %
2127
(self.parent_id_basename_to_file_id.key()[0],))
2128
lines.append("id_to_entry: %s\n" % (self.id_to_entry.key()[0],))
2133
"""Get the root entry."""
2134
return self[self.root_id]
2137
class CHKInventoryDirectory(InventoryDirectory):
2138
"""A directory in an inventory."""
2140
__slots__ = ['_children', '_chk_inventory']
2142
def __init__(self, file_id, name, parent_id, chk_inventory):
2143
# Don't call InventoryDirectory.__init__ - it isn't right for this
2145
InventoryEntry.__init__(self, file_id, name, parent_id)
2146
self._children = None
2147
self._chk_inventory = chk_inventory
2151
"""Access the list of children of this directory.
2153
With a parent_id_basename_to_file_id index, loads all the children,
2154
without loads the entire index. Without is bad. A more sophisticated
2155
proxy object might be nice, to allow partial loading of children as
2156
well when specific names are accessed. (So path traversal can be
2157
written in the obvious way but not examine siblings.).
2159
if self._children is not None:
2160
return self._children
2161
# No longer supported
2162
if self._chk_inventory.parent_id_basename_to_file_id is None:
2163
raise AssertionError("Inventories without"
2164
" parent_id_basename_to_file_id are no longer supported")
2166
# XXX: Todo - use proxy objects for the children rather than loading
2167
# all when the attribute is referenced.
2168
parent_id_index = self._chk_inventory.parent_id_basename_to_file_id
2170
for (parent_id, name_utf8), file_id in parent_id_index.iteritems(
2171
key_filter=[StaticTuple(self.file_id,)]):
2172
child_keys.add(StaticTuple(file_id,))
2174
for file_id_key in child_keys:
2175
entry = self._chk_inventory._fileid_to_entry_cache.get(
2176
file_id_key[0], None)
2177
if entry is not None:
2178
result[entry.name] = entry
2179
cached.add(file_id_key)
2180
child_keys.difference_update(cached)
2181
# populate; todo: do by name
2182
id_to_entry = self._chk_inventory.id_to_entry
2183
for file_id_key, bytes in id_to_entry.iteritems(child_keys):
2184
entry = self._chk_inventory._bytes_to_entry(bytes)
2185
result[entry.name] = entry
2186
self._chk_inventory._fileid_to_entry_cache[file_id_key[0]] = entry
2187
self._children = result
2191
'directory': InventoryDirectory,
2192
'file': InventoryFile,
2193
'symlink': InventoryLink,
2194
'tree-reference': TreeReference
2197
def make_entry(kind, name, parent_id, file_id=None):
2198
"""Create an inventory entry.
2200
:param kind: the type of inventory entry to create.
2201
:param name: the basename of the entry.
2202
:param parent_id: the parent_id of the entry.
2203
:param file_id: the file_id to use. if None, one will be created.
2206
file_id = generate_ids.gen_file_id(name)
2207
name = ensure_normalized_name(name)
2209
factory = entry_factory[kind]
2211
raise errors.BadFileKindError(name, kind)
2212
return factory(file_id, name, parent_id)
2215
def ensure_normalized_name(name):
2218
:raises InvalidNormalization: When name is not normalized, and cannot be
2219
accessed on this platform by the normalized path.
2220
:return: The NFC normalised version of name.
2222
#------- This has been copied to bzrlib.dirstate.DirState.add, please
2223
# keep them synchronised.
2224
# we dont import normalized_filename directly because we want to be
2225
# able to change the implementation at runtime for tests.
2226
norm_name, can_access = osutils.normalized_filename(name)
2227
if norm_name != name:
2231
# TODO: jam 20060701 This would probably be more useful
2232
# if the error was raised with the full path
2233
raise errors.InvalidNormalization(name)
2239
def is_valid_name(name):
2241
if _NAME_RE is None:
2242
_NAME_RE = re.compile(r'^[^/\\]+$')
2244
return bool(_NAME_RE.match(name))
2247
def _check_delta_unique_ids(delta):
2248
"""Decorate a delta and check that the file ids in it are unique.
2250
:return: A generator over delta.
2254
length = len(ids) + 1
2256
if len(ids) != length:
2257
raise errors.InconsistentDelta(item[0] or item[1], item[2],
2262
def _check_delta_unique_new_paths(delta):
2263
"""Decorate a delta and check that the new paths in it are unique.
2265
:return: A generator over delta.
2269
length = len(paths) + 1
2271
if path is not None:
2273
if len(paths) != length:
2274
raise errors.InconsistentDelta(path, item[2], "repeated path")
2278
def _check_delta_unique_old_paths(delta):
2279
"""Decorate a delta and check that the old paths in it are unique.
2281
:return: A generator over delta.
2285
length = len(paths) + 1
2287
if path is not None:
2289
if len(paths) != length:
2290
raise errors.InconsistentDelta(path, item[2], "repeated path")
2294
def _check_delta_ids_are_valid(delta):
2295
"""Decorate a delta and check that the ids in it are valid.
2297
:return: A generator over delta.
2302
raise errors.InconsistentDelta(item[0] or item[1], item[2],
2303
"entry with file_id None %r" % entry)
2304
if type(item[2]) != str:
2305
raise errors.InconsistentDelta(item[0] or item[1], item[2],
2306
"entry with non bytes file_id %r" % entry)
2310
def _check_delta_ids_match_entry(delta):
2311
"""Decorate a delta and check that the ids in it match the entry.file_id.
2313
:return: A generator over delta.
2317
if entry is not None:
2318
if entry.file_id != item[2]:
2319
raise errors.InconsistentDelta(item[0] or item[1], item[2],
2320
"mismatched id with %r" % entry)
2324
def _check_delta_new_path_entry_both_or_None(delta):
2325
"""Decorate a delta and check that the new_path and entry are paired.
2327
:return: A generator over delta.
2332
if new_path is None and entry is not None:
2333
raise errors.InconsistentDelta(item[0], item[1],
2334
"Entry with no new_path")
2335
if new_path is not None and entry is None:
2336
raise errors.InconsistentDelta(new_path, item[1],
2337
"new_path with no entry")
added
removed
bzrlib/inventory.py
