1
# Copyright (C) 2005, 2006, 2007 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., 59 Temple Place, Suite 330, Boston, MA 02111-1307 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.
34
from bzrlib.lazy_import import lazy_import
35
lazy_import(globals(), """
49
from bzrlib.errors import (
53
from bzrlib.symbol_versioning import deprecated_method, zero_ninetyone
54
from bzrlib.trace import mutter
57
class InventoryEntry(object):
58
"""Description of a versioned file.
60
An InventoryEntry has the following fields, which are also
61
present in the XML inventory-entry element:
66
(within the parent directory)
69
file_id of the parent directory, or ROOT_ID
72
the revision_id in which this variation of this file was
76
Indicates that this file should be executable on systems
80
sha-1 of the text of the file
83
size in bytes of the text of the file
85
(reading a version 4 tree created a text_id field.)
90
>>> i.add(InventoryDirectory('123', 'src', ROOT_ID))
91
InventoryDirectory('123', 'src', parent_id='TREE_ROOT', revision=None)
92
>>> i.add(InventoryFile('2323', 'hello.c', parent_id='123'))
93
InventoryFile('2323', 'hello.c', parent_id='123', sha1=None, len=None)
94
>>> shouldbe = {0: '', 1: 'src', 2: 'src/hello.c'}
95
>>> for ix, j in enumerate(i.iter_entries()):
96
... print (j[0] == shouldbe[ix], j[1])
98
(True, InventoryDirectory('TREE_ROOT', u'', parent_id=None, revision=None))
99
(True, InventoryDirectory('123', 'src', parent_id='TREE_ROOT', revision=None))
100
(True, InventoryFile('2323', 'hello.c', parent_id='123', sha1=None, len=None))
101
>>> i.add(InventoryFile('2324', 'bye.c', '123'))
102
InventoryFile('2324', 'bye.c', parent_id='123', sha1=None, len=None)
103
>>> i.add(InventoryDirectory('2325', 'wibble', '123'))
104
InventoryDirectory('2325', 'wibble', parent_id='123', revision=None)
105
>>> i.path2id('src/wibble')
109
>>> i.add(InventoryFile('2326', 'wibble.c', '2325'))
110
InventoryFile('2326', 'wibble.c', parent_id='2325', sha1=None, len=None)
112
InventoryFile('2326', 'wibble.c', parent_id='2325', sha1=None, len=None)
113
>>> for path, entry in i.iter_entries():
115
... assert i.path2id(path)
123
>>> i.id2path('2326')
124
'src/wibble/wibble.c'
127
# Constants returned by describe_change()
129
# TODO: These should probably move to some kind of FileChangeDescription
130
# class; that's like what's inside a TreeDelta but we want to be able to
131
# generate them just for one file at a time.
133
MODIFIED_AND_RENAMED = 'modified and renamed'
137
def detect_changes(self, old_entry):
138
"""Return a (text_modified, meta_modified) from this to old_entry.
140
_read_tree_state must have been called on self and old_entry prior to
141
calling detect_changes.
145
def diff(self, text_diff, from_label, tree, to_label, to_entry, to_tree,
146
output_to, reverse=False):
147
"""Perform a diff from this to to_entry.
149
text_diff will be used for textual difference calculation.
150
This is a template method, override _diff in child classes.
152
self._read_tree_state(tree.id2path(self.file_id), tree)
154
# cannot diff from one kind to another - you must do a removal
155
# and an addif they do not match.
156
assert self.kind == to_entry.kind
157
to_entry._read_tree_state(to_tree.id2path(to_entry.file_id),
159
self._diff(text_diff, from_label, tree, to_label, to_entry, to_tree,
162
def _diff(self, text_diff, from_label, tree, to_label, to_entry, to_tree,
163
output_to, reverse=False):
164
"""Perform a diff between two entries of the same kind."""
166
def parent_candidates(self, previous_inventories):
167
"""Find possible per-file graph parents.
169
This is currently defined by:
170
- Select the last changed revision in the parent inventory.
171
- Do deal with a short lived bug in bzr 0.8's development two entries
172
that have the same last changed but different 'x' bit settings are
175
# revision:ie mapping for each ie found in previous_inventories.
177
# identify candidate head revision ids.
178
for inv in previous_inventories:
179
if self.file_id in inv:
180
ie = inv[self.file_id]
181
assert ie.file_id == self.file_id
182
if ie.revision in candidates:
183
# same revision value in two different inventories:
184
# correct possible inconsistencies:
185
# * there was a bug in revision updates with 'x' bit
188
if candidates[ie.revision].executable != ie.executable:
189
candidates[ie.revision].executable = False
190
ie.executable = False
191
except AttributeError:
193
# must now be the same.
194
assert candidates[ie.revision] == ie
196
# add this revision as a candidate.
197
candidates[ie.revision] = ie
200
@deprecated_method(zero_ninetyone)
201
def find_previous_heads(self, previous_inventories,
202
versioned_file_store,
205
"""Return the revisions and entries that directly precede this.
207
Returned as a map from revision to inventory entry.
209
This is a map containing the file revisions in all parents
210
for which the file exists, and its revision is not a parent of
211
any other. If the file is new, the set will be empty.
213
:param versioned_file_store: A store where ancestry data on this
214
file id can be queried.
215
:param transaction: The transaction that queries to the versioned
216
file store should be completed under.
217
:param entry_vf: The entry versioned file, if its already available.
219
candidates = self.parent_candidates(previous_inventories)
221
# revision:ie mapping with one revision for each head.
223
# common case optimisation
224
if len(candidates) == 1:
225
# if there is only one candidate revision found
226
# then we can avoid opening the versioned file to access ancestry:
227
# there cannot be any ancestors to eliminate when there is
228
# only one revision available.
231
# --- what follows is now encapsulated in repository.get_graph.heads(),
232
# but that is not accessible from here as we have no repository
233
# pointer. Note that the repository.get_graph.heads() call can return
234
# different results *at the moment* because of the kind-changing check
235
# we have in parent_candidates().
237
# eliminate ancestors amongst the available candidates:
238
# heads are those that are not an ancestor of any other candidate
239
# - this provides convergence at a per-file level.
240
def get_ancestors(weave, entry):
241
return set(weave.get_ancestry(entry.revision, topo_sorted=False))
242
# revision: ancestor list for each head
244
for ie in candidates.values():
245
# may be an ancestor of a known head:
246
already_present = 0 != len(
247
[head for head in heads
248
if ie.revision in head_ancestors[head]])
250
# an ancestor of an analyzed candidate.
252
# not an ancestor of a known head:
253
# load the versioned file for this file id if needed
255
entry_vf = versioned_file_store.get_weave_or_empty(
256
self.file_id, transaction)
257
ancestors = get_ancestors(entry_vf, ie)
258
# may knock something else out:
259
check_heads = list(heads.keys())
260
for head in check_heads:
261
if head in ancestors:
262
# this previously discovered 'head' is not
263
# really a head - its an ancestor of the newly
266
head_ancestors[ie.revision] = ancestors
267
heads[ie.revision] = ie
270
def get_tar_item(self, root, dp, now, tree):
271
"""Get a tarfile item and a file stream for its content."""
272
item = tarfile.TarInfo(osutils.pathjoin(root, dp).encode('utf8'))
273
# TODO: would be cool to actually set it to the timestamp of the
274
# revision it was last changed
276
fileobj = self._put_in_tar(item, tree)
280
"""Return true if the object this entry represents has textual data.
282
Note that textual data includes binary content.
284
Also note that all entries get weave files created for them.
285
This attribute is primarily used when upgrading from old trees that
286
did not have the weave index for all inventory entries.
290
def __init__(self, file_id, name, parent_id, text_id=None):
291
"""Create an InventoryEntry
293
The filename must be a single component, relative to the
294
parent directory; it cannot be a whole path or relative name.
296
>>> e = InventoryFile('123', 'hello.c', ROOT_ID)
301
>>> e = InventoryFile('123', 'src/hello.c', ROOT_ID)
302
Traceback (most recent call last):
303
InvalidEntryName: Invalid entry name: src/hello.c
305
assert isinstance(name, basestring), name
306
if '/' in name or '\\' in name:
307
raise errors.InvalidEntryName(name=name)
308
self.executable = False
310
self.text_sha1 = None
311
self.text_size = None
312
self.file_id = file_id
313
assert isinstance(file_id, (str, None.__class__)), \
314
'bad type %r for %r' % (type(file_id), file_id)
316
self.text_id = text_id
317
self.parent_id = parent_id
318
self.symlink_target = None
319
self.reference_revision = None
321
def kind_character(self):
322
"""Return a short kind indicator useful for appending to names."""
323
raise BzrError('unknown kind %r' % self.kind)
325
known_kinds = ('file', 'directory', 'symlink')
327
def _put_in_tar(self, item, tree):
328
"""populate item for stashing in a tar, and return the content stream.
330
If no content is available, return None.
332
raise BzrError("don't know how to export {%s} of kind %r" %
333
(self.file_id, self.kind))
335
def put_on_disk(self, dest, dp, tree):
336
"""Create a representation of self on disk in the prefix dest.
338
This is a template method - implement _put_on_disk in subclasses.
340
fullpath = osutils.pathjoin(dest, dp)
341
self._put_on_disk(fullpath, tree)
342
# mutter(" export {%s} kind %s to %s", self.file_id,
343
# self.kind, fullpath)
345
def _put_on_disk(self, fullpath, tree):
346
"""Put this entry onto disk at fullpath, from tree tree."""
347
raise BzrError("don't know how to export {%s} of kind %r" % (self.file_id, self.kind))
349
def sorted_children(self):
350
return sorted(self.children.items())
353
def versionable_kind(kind):
354
return (kind in ('file', 'directory', 'symlink', 'tree-reference'))
356
def check(self, checker, rev_id, inv, tree):
357
"""Check this inventory entry is intact.
359
This is a template method, override _check for kind specific
362
:param checker: Check object providing context for the checks;
363
can be used to find out what parts of the repository have already
365
:param rev_id: Revision id from which this InventoryEntry was loaded.
366
Not necessarily the last-changed revision for this file.
367
:param inv: Inventory from which the entry was loaded.
368
:param tree: RevisionTree for this entry.
370
if self.parent_id is not None:
371
if not inv.has_id(self.parent_id):
372
raise BzrCheckError('missing parent {%s} in inventory for revision {%s}'
373
% (self.parent_id, rev_id))
374
self._check(checker, rev_id, tree)
376
def _check(self, checker, rev_id, tree):
377
"""Check this inventory entry for kind specific errors."""
378
raise BzrCheckError('unknown entry kind %r in revision {%s}' %
382
"""Clone this inventory entry."""
383
raise NotImplementedError
386
def describe_change(old_entry, new_entry):
387
"""Describe the change between old_entry and this.
389
This smells of being an InterInventoryEntry situation, but as its
390
the first one, we're making it a static method for now.
392
An entry with a different parent, or different name is considered
393
to be renamed. Reparenting is an internal detail.
394
Note that renaming the parent does not trigger a rename for the
397
# TODO: Perhaps return an object rather than just a string
398
if old_entry is new_entry:
399
# also the case of both being None
401
elif old_entry is None:
403
elif new_entry is None:
405
if old_entry.kind != new_entry.kind:
407
text_modified, meta_modified = new_entry.detect_changes(old_entry)
408
if text_modified or meta_modified:
412
# TODO 20060511 (mbp, rbc) factor out 'detect_rename' here.
413
if old_entry.parent_id != new_entry.parent_id:
415
elif old_entry.name != new_entry.name:
419
if renamed and not modified:
420
return InventoryEntry.RENAMED
421
if modified and not renamed:
423
if modified and renamed:
424
return InventoryEntry.MODIFIED_AND_RENAMED
428
return ("%s(%r, %r, parent_id=%r, revision=%r)"
429
% (self.__class__.__name__,
435
def __eq__(self, other):
436
if not isinstance(other, InventoryEntry):
437
return NotImplemented
439
return ((self.file_id == other.file_id)
440
and (self.name == other.name)
441
and (other.symlink_target == self.symlink_target)
442
and (self.text_sha1 == other.text_sha1)
443
and (self.text_size == other.text_size)
444
and (self.text_id == other.text_id)
445
and (self.parent_id == other.parent_id)
446
and (self.kind == other.kind)
447
and (self.revision == other.revision)
448
and (self.executable == other.executable)
449
and (self.reference_revision == other.reference_revision)
452
def __ne__(self, other):
453
return not (self == other)
456
raise ValueError('not hashable')
458
def _unchanged(self, previous_ie):
459
"""Has this entry changed relative to previous_ie.
461
This method should be overridden in child classes.
464
# different inv parent
465
if previous_ie.parent_id != self.parent_id:
468
elif previous_ie.name != self.name:
470
elif previous_ie.kind != self.kind:
474
def _read_tree_state(self, path, work_tree):
475
"""Populate fields in the inventory entry from the given tree.
477
Note that this should be modified to be a noop on virtual trees
478
as all entries created there are prepopulated.
480
# TODO: Rather than running this manually, we should check the
481
# working sha1 and other expensive properties when they're
482
# first requested, or preload them if they're already known
483
pass # nothing to do by default
485
def _forget_tree_state(self):
489
class RootEntry(InventoryEntry):
491
__slots__ = ['text_sha1', 'text_size', 'file_id', 'name', 'kind',
492
'text_id', 'parent_id', 'children', 'executable',
493
'revision', 'symlink_target', 'reference_revision']
495
def _check(self, checker, rev_id, tree):
496
"""See InventoryEntry._check"""
498
def __init__(self, file_id):
499
self.file_id = file_id
501
self.kind = 'directory'
502
self.parent_id = None
505
symbol_versioning.warn('RootEntry is deprecated as of bzr 0.10.'
506
' Please use InventoryDirectory instead.',
507
DeprecationWarning, stacklevel=2)
509
def __eq__(self, other):
510
if not isinstance(other, RootEntry):
511
return NotImplemented
513
return (self.file_id == other.file_id) \
514
and (self.children == other.children)
517
class InventoryDirectory(InventoryEntry):
518
"""A directory in an inventory."""
520
__slots__ = ['text_sha1', 'text_size', 'file_id', 'name', 'kind',
521
'text_id', 'parent_id', 'children', 'executable',
522
'revision', 'symlink_target', 'reference_revision']
524
def _check(self, checker, rev_id, tree):
525
"""See InventoryEntry._check"""
526
if self.text_sha1 is not None or self.text_size is not None or self.text_id is not None:
527
raise BzrCheckError('directory {%s} has text in revision {%s}'
528
% (self.file_id, rev_id))
531
other = InventoryDirectory(self.file_id, self.name, self.parent_id)
532
other.revision = self.revision
533
# note that children are *not* copied; they're pulled across when
537
def __init__(self, file_id, name, parent_id):
538
super(InventoryDirectory, self).__init__(file_id, name, parent_id)
540
self.kind = 'directory'
542
def kind_character(self):
543
"""See InventoryEntry.kind_character."""
546
def _put_in_tar(self, item, tree):
547
"""See InventoryEntry._put_in_tar."""
548
item.type = tarfile.DIRTYPE
555
def _put_on_disk(self, fullpath, tree):
556
"""See InventoryEntry._put_on_disk."""
560
class InventoryFile(InventoryEntry):
561
"""A file in an inventory."""
563
__slots__ = ['text_sha1', 'text_size', 'file_id', 'name', 'kind',
564
'text_id', 'parent_id', 'children', 'executable',
565
'revision', 'symlink_target', 'reference_revision']
567
def _check(self, checker, tree_revision_id, tree):
568
"""See InventoryEntry._check"""
569
t = (self.file_id, self.revision)
570
if t in checker.checked_texts:
571
prev_sha = checker.checked_texts[t]
572
if prev_sha != self.text_sha1:
574
'mismatched sha1 on {%s} in {%s} (%s != %s) %r' %
575
(self.file_id, tree_revision_id, prev_sha, self.text_sha1,
578
checker.repeated_text_cnt += 1
581
if self.file_id not in checker.checked_weaves:
582
mutter('check weave {%s}', self.file_id)
583
w = tree._get_weave(self.file_id)
584
# Not passing a progress bar, because it creates a new
585
# progress, which overwrites the current progress,
586
# and doesn't look nice
588
checker.checked_weaves[self.file_id] = True
590
w = tree._get_weave(self.file_id)
592
mutter('check version {%s} of {%s}', tree_revision_id, self.file_id)
593
checker.checked_text_cnt += 1
594
# We can't check the length, because Weave doesn't store that
595
# information, and the whole point of looking at the weave's
596
# sha1sum is that we don't have to extract the text.
597
if self.text_sha1 != w.get_sha1(self.revision):
598
raise BzrCheckError('text {%s} version {%s} wrong sha1'
599
% (self.file_id, self.revision))
600
checker.checked_texts[t] = self.text_sha1
603
other = InventoryFile(self.file_id, self.name, self.parent_id)
604
other.executable = self.executable
605
other.text_id = self.text_id
606
other.text_sha1 = self.text_sha1
607
other.text_size = self.text_size
608
other.revision = self.revision
611
def detect_changes(self, old_entry):
612
"""See InventoryEntry.detect_changes."""
613
assert self.text_sha1 is not None
614
assert old_entry.text_sha1 is not None
615
text_modified = (self.text_sha1 != old_entry.text_sha1)
616
meta_modified = (self.executable != old_entry.executable)
617
return text_modified, meta_modified
619
def _diff(self, text_diff, from_label, tree, to_label, to_entry, to_tree,
620
output_to, reverse=False):
621
"""See InventoryEntry._diff."""
623
from_text = tree.get_file(self.file_id).readlines()
625
to_text = to_tree.get_file(to_entry.file_id).readlines()
629
text_diff(from_label, from_text,
630
to_label, to_text, output_to)
632
text_diff(to_label, to_text,
633
from_label, from_text, output_to)
634
except errors.BinaryFile:
636
label_pair = (to_label, from_label)
638
label_pair = (from_label, to_label)
640
("Binary files %s and %s differ\n" % label_pair).encode('utf8'))
643
"""See InventoryEntry.has_text."""
646
def __init__(self, file_id, name, parent_id):
647
super(InventoryFile, self).__init__(file_id, name, parent_id)
650
def kind_character(self):
651
"""See InventoryEntry.kind_character."""
654
def _put_in_tar(self, item, tree):
655
"""See InventoryEntry._put_in_tar."""
656
item.type = tarfile.REGTYPE
657
fileobj = tree.get_file(self.file_id)
658
item.size = self.text_size
659
if tree.is_executable(self.file_id):
665
def _put_on_disk(self, fullpath, tree):
666
"""See InventoryEntry._put_on_disk."""
667
osutils.pumpfile(tree.get_file(self.file_id), file(fullpath, 'wb'))
668
if tree.is_executable(self.file_id):
669
os.chmod(fullpath, 0755)
671
def _read_tree_state(self, path, work_tree):
672
"""See InventoryEntry._read_tree_state."""
673
self.text_sha1 = work_tree.get_file_sha1(self.file_id, path=path)
674
# FIXME: 20050930 probe for the text size when getting sha1
675
# in _read_tree_state
676
self.executable = work_tree.is_executable(self.file_id, path=path)
679
return ("%s(%r, %r, parent_id=%r, sha1=%r, len=%s)"
680
% (self.__class__.__name__,
687
def _forget_tree_state(self):
688
self.text_sha1 = None
690
def _unchanged(self, previous_ie):
691
"""See InventoryEntry._unchanged."""
692
compatible = super(InventoryFile, self)._unchanged(previous_ie)
693
if self.text_sha1 != previous_ie.text_sha1:
696
# FIXME: 20050930 probe for the text size when getting sha1
697
# in _read_tree_state
698
self.text_size = previous_ie.text_size
699
if self.executable != previous_ie.executable:
704
class InventoryLink(InventoryEntry):
705
"""A file in an inventory."""
707
__slots__ = ['text_sha1', 'text_size', 'file_id', 'name', 'kind',
708
'text_id', 'parent_id', 'children', 'executable',
709
'revision', 'symlink_target', 'reference_revision']
711
def _check(self, checker, rev_id, tree):
712
"""See InventoryEntry._check"""
713
if self.text_sha1 is not None or self.text_size is not None or self.text_id is not None:
714
raise BzrCheckError('symlink {%s} has text in revision {%s}'
715
% (self.file_id, rev_id))
716
if self.symlink_target is None:
717
raise BzrCheckError('symlink {%s} has no target in revision {%s}'
718
% (self.file_id, rev_id))
721
other = InventoryLink(self.file_id, self.name, self.parent_id)
722
other.symlink_target = self.symlink_target
723
other.revision = self.revision
726
def detect_changes(self, old_entry):
727
"""See InventoryEntry.detect_changes."""
728
# FIXME: which _modified field should we use ? RBC 20051003
729
text_modified = (self.symlink_target != old_entry.symlink_target)
731
mutter(" symlink target changed")
732
meta_modified = False
733
return text_modified, meta_modified
735
def _diff(self, text_diff, from_label, tree, to_label, to_entry, to_tree,
736
output_to, reverse=False):
737
"""See InventoryEntry._diff."""
738
from_text = self.symlink_target
739
if to_entry is not None:
740
to_text = to_entry.symlink_target
745
output_to.write('=== target changed %r => %r\n' % (from_text, to_text))
748
output_to.write('=== target was %r\n' % self.symlink_target)
750
output_to.write('=== target is %r\n' % self.symlink_target)
752
def __init__(self, file_id, name, parent_id):
753
super(InventoryLink, self).__init__(file_id, name, parent_id)
754
self.kind = 'symlink'
756
def kind_character(self):
757
"""See InventoryEntry.kind_character."""
760
def _put_in_tar(self, item, tree):
761
"""See InventoryEntry._put_in_tar."""
762
item.type = tarfile.SYMTYPE
766
item.linkname = self.symlink_target
769
def _put_on_disk(self, fullpath, tree):
770
"""See InventoryEntry._put_on_disk."""
772
os.symlink(self.symlink_target, fullpath)
774
raise BzrError("Failed to create symlink %r -> %r, error: %s" % (fullpath, self.symlink_target, e))
776
def _read_tree_state(self, path, work_tree):
777
"""See InventoryEntry._read_tree_state."""
778
self.symlink_target = work_tree.get_symlink_target(self.file_id)
780
def _forget_tree_state(self):
781
self.symlink_target = None
783
def _unchanged(self, previous_ie):
784
"""See InventoryEntry._unchanged."""
785
compatible = super(InventoryLink, self)._unchanged(previous_ie)
786
if self.symlink_target != previous_ie.symlink_target:
791
class TreeReference(InventoryEntry):
793
kind = 'tree-reference'
795
def __init__(self, file_id, name, parent_id, revision=None,
796
reference_revision=None):
797
InventoryEntry.__init__(self, file_id, name, parent_id)
798
self.revision = revision
799
self.reference_revision = reference_revision
802
return TreeReference(self.file_id, self.name, self.parent_id,
803
self.revision, self.reference_revision)
805
def _read_tree_state(self, path, work_tree):
806
"""Populate fields in the inventory entry from the given tree.
808
self.reference_revision = work_tree.get_reference_revision(
811
def _forget_tree_state(self):
812
self.reference_revision = None
814
def _unchanged(self, previous_ie):
815
"""See InventoryEntry._unchanged."""
816
compatible = super(TreeReference, self)._unchanged(previous_ie)
817
if self.reference_revision != previous_ie.reference_revision:
822
class Inventory(object):
823
"""Inventory of versioned files in a tree.
825
This describes which file_id is present at each point in the tree,
826
and possibly the SHA-1 or other information about the file.
827
Entries can be looked up either by path or by file_id.
829
The inventory represents a typical unix file tree, with
830
directories containing files and subdirectories. We never store
831
the full path to a file, because renaming a directory implicitly
832
moves all of its contents. This class internally maintains a
833
lookup tree that allows the children under a directory to be
836
InventoryEntry objects must not be modified after they are
837
inserted, other than through the Inventory API.
839
>>> inv = Inventory()
840
>>> inv.add(InventoryFile('123-123', 'hello.c', ROOT_ID))
841
InventoryFile('123-123', 'hello.c', parent_id='TREE_ROOT', sha1=None, len=None)
842
>>> inv['123-123'].name
845
May be treated as an iterator or set to look up file ids:
847
>>> bool(inv.path2id('hello.c'))
852
May also look up by name:
854
>>> [x[0] for x in inv.iter_entries()]
856
>>> inv = Inventory('TREE_ROOT-12345678-12345678')
857
>>> inv.add(InventoryFile('123-123', 'hello.c', ROOT_ID))
858
Traceback (most recent call last):
859
BzrError: parent_id {TREE_ROOT} not in inventory
860
>>> inv.add(InventoryFile('123-123', 'hello.c', 'TREE_ROOT-12345678-12345678'))
861
InventoryFile('123-123', 'hello.c', parent_id='TREE_ROOT-12345678-12345678', sha1=None, len=None)
863
def __init__(self, root_id=ROOT_ID, revision_id=None):
864
"""Create or read an inventory.
866
If a working directory is specified, the inventory is read
867
from there. If the file is specified, read from that. If not,
868
the inventory is created empty.
870
The inventory is created with a default root directory, with
873
if root_id is not None:
874
assert root_id.__class__ == str
875
self._set_root(InventoryDirectory(root_id, u'', None))
879
self.revision_id = revision_id
882
return "<Inventory object at %x, contents=%r>" % (id(self), self._byid)
884
def apply_delta(self, delta):
885
"""Apply a delta to this inventory.
887
:param delta: A list of changes to apply. After all the changes are
888
applied the final inventory must be internally consistent, but it
889
is ok to supply changes which, if only half-applied would have an
890
invalid result - such as supplying two changes which rename two
891
files, 'A' and 'B' with each other : [('A', 'B', 'A-id', a_entry),
892
('B', 'A', 'B-id', b_entry)].
894
Each change is a tuple, of the form (old_path, new_path, file_id,
897
When new_path is None, the change indicates the removal of an entry
898
from the inventory and new_entry will be ignored (using None is
899
appropriate). If new_path is not None, then new_entry must be an
900
InventoryEntry instance, which will be incorporated into the
901
inventory (and replace any existing entry with the same file id).
903
When old_path is None, the change indicates the addition of
904
a new entry to the inventory.
906
When neither new_path nor old_path are None, the change is a
907
modification to an entry, such as a rename, reparent, kind change
910
The children attribute of new_entry is ignored. This is because
911
this method preserves children automatically across alterations to
912
the parent of the children, and cases where the parent id of a
913
child is changing require the child to be passed in as a separate
914
change regardless. E.g. in the recursive deletion of a directory -
915
the directory's children must be included in the delta, or the
916
final inventory will be invalid.
919
# Remove all affected items which were in the original inventory,
920
# starting with the longest paths, thus ensuring parents are examined
921
# after their children, which means that everything we examine has no
922
# modified children remaining by the time we examine it.
923
for old_path, file_id in sorted(((op, f) for op, np, f, e in delta
924
if op is not None), reverse=True):
925
if file_id not in self:
928
# Preserve unaltered children of file_id for later reinsertion.
929
children[file_id] = getattr(self[file_id], 'children', {})
930
# Remove file_id and the unaltered children. If file_id is not
931
# being deleted it will be reinserted back later.
932
self.remove_recursive_id(file_id)
933
# Insert all affected which should be in the new inventory, reattaching
934
# their children if they had any. This is done from shortest path to
935
# longest, ensuring that items which were modified and whose parents in
936
# the resulting inventory were also modified, are inserted after their
938
for new_path, new_entry in sorted((np, e) for op, np, f, e in
939
delta if np is not None):
940
if new_entry.kind == 'directory':
941
new_entry.children = children.get(new_entry.file_id, {})
944
def _set_root(self, ie):
946
self._byid = {self.root.file_id: self.root}
949
# TODO: jam 20051218 Should copy also copy the revision_id?
950
entries = self.iter_entries()
951
other = Inventory(entries.next()[1].file_id)
952
# copy recursively so we know directories will be added before
953
# their children. There are more efficient ways than this...
954
for path, entry in entries():
955
other.add(entry.copy())
959
return iter(self._byid)
962
"""Returns number of entries."""
963
return len(self._byid)
965
def iter_entries(self, from_dir=None):
966
"""Return (path, entry) pairs, in order by name."""
968
if self.root is None:
972
elif isinstance(from_dir, basestring):
973
from_dir = self._byid[from_dir]
975
# unrolling the recursive called changed the time from
976
# 440ms/663ms (inline/total) to 116ms/116ms
977
children = from_dir.children.items()
979
children = collections.deque(children)
980
stack = [(u'', children)]
982
from_dir_relpath, children = stack[-1]
985
name, ie = children.popleft()
987
# we know that from_dir_relpath never ends in a slash
988
# and 'f' doesn't begin with one, we can do a string op, rather
989
# than the checks of pathjoin(), though this means that all paths
991
path = from_dir_relpath + '/' + name
995
if ie.kind != 'directory':
998
# But do this child first
999
new_children = ie.children.items()
1001
new_children = collections.deque(new_children)
1002
stack.append((path, new_children))
1003
# Break out of inner loop, so that we start outer loop with child
1006
# if we finished all children, pop it off the stack
1009
def iter_entries_by_dir(self, from_dir=None, specific_file_ids=None,
1010
yield_parents=False):
1011
"""Iterate over the entries in a directory first order.
1013
This returns all entries for a directory before returning
1014
the entries for children of a directory. This is not
1015
lexicographically sorted order, and is a hybrid between
1016
depth-first and breadth-first.
1018
:param yield_parents: If True, yield the parents from the root leading
1019
down to specific_file_ids that have been requested. This has no
1020
impact if specific_file_ids is None.
1021
:return: This yields (path, entry) pairs
1023
if specific_file_ids and not isinstance(specific_file_ids, set):
1024
specific_file_ids = set(specific_file_ids)
1025
# TODO? Perhaps this should return the from_dir so that the root is
1026
# yielded? or maybe an option?
1027
if from_dir is None:
1028
if self.root is None:
1030
# Optimize a common case
1031
if (not yield_parents and specific_file_ids is not None and
1032
len(specific_file_ids) == 1):
1033
file_id = list(specific_file_ids)[0]
1035
yield self.id2path(file_id), self[file_id]
1037
from_dir = self.root
1038
if (specific_file_ids is None or yield_parents or
1039
self.root.file_id in specific_file_ids):
1040
yield u'', self.root
1041
elif isinstance(from_dir, basestring):
1042
from_dir = self._byid[from_dir]
1044
if specific_file_ids is not None:
1045
# TODO: jam 20070302 This could really be done as a loop rather
1046
# than a bunch of recursive calls.
1049
def add_ancestors(file_id):
1050
if file_id not in byid:
1052
parent_id = byid[file_id].parent_id
1053
if parent_id is None:
1055
if parent_id not in parents:
1056
parents.add(parent_id)
1057
add_ancestors(parent_id)
1058
for file_id in specific_file_ids:
1059
add_ancestors(file_id)
1063
stack = [(u'', from_dir)]
1065
cur_relpath, cur_dir = stack.pop()
1068
for child_name, child_ie in sorted(cur_dir.children.iteritems()):
1070
child_relpath = cur_relpath + child_name
1072
if (specific_file_ids is None or
1073
child_ie.file_id in specific_file_ids or
1074
(yield_parents and child_ie.file_id in parents)):
1075
yield child_relpath, child_ie
1077
if child_ie.kind == 'directory':
1078
if parents is None or child_ie.file_id in parents:
1079
child_dirs.append((child_relpath+'/', child_ie))
1080
stack.extend(reversed(child_dirs))
1082
def make_entry(self, kind, name, parent_id, file_id=None):
1083
"""Simple thunk to bzrlib.inventory.make_entry."""
1084
return make_entry(kind, name, parent_id, file_id)
1087
"""Return list of (path, ie) for all entries except the root.
1089
This may be faster than iter_entries.
1092
def descend(dir_ie, dir_path):
1093
kids = dir_ie.children.items()
1095
for name, ie in kids:
1096
child_path = osutils.pathjoin(dir_path, name)
1097
accum.append((child_path, ie))
1098
if ie.kind == 'directory':
1099
descend(ie, child_path)
1101
descend(self.root, u'')
1104
def directories(self):
1105
"""Return (path, entry) pairs for all directories, including the root.
1108
def descend(parent_ie, parent_path):
1109
accum.append((parent_path, parent_ie))
1111
kids = [(ie.name, ie) for ie in parent_ie.children.itervalues() if ie.kind == 'directory']
1114
for name, child_ie in kids:
1115
child_path = osutils.pathjoin(parent_path, name)
1116
descend(child_ie, child_path)
1117
descend(self.root, u'')
1120
def __contains__(self, file_id):
1121
"""True if this entry contains a file with given id.
1123
>>> inv = Inventory()
1124
>>> inv.add(InventoryFile('123', 'foo.c', ROOT_ID))
1125
InventoryFile('123', 'foo.c', parent_id='TREE_ROOT', sha1=None, len=None)
1131
return (file_id in self._byid)
1133
def __getitem__(self, file_id):
1134
"""Return the entry for given file_id.
1136
>>> inv = Inventory()
1137
>>> inv.add(InventoryFile('123123', 'hello.c', ROOT_ID))
1138
InventoryFile('123123', 'hello.c', parent_id='TREE_ROOT', sha1=None, len=None)
1139
>>> inv['123123'].name
1143
return self._byid[file_id]
1145
# really we're passing an inventory, not a tree...
1146
raise errors.NoSuchId(self, file_id)
1148
def get_file_kind(self, file_id):
1149
return self._byid[file_id].kind
1151
def get_child(self, parent_id, filename):
1152
return self[parent_id].children.get(filename)
1154
def _add_child(self, entry):
1155
"""Add an entry to the inventory, without adding it to its parent"""
1156
if entry.file_id in self._byid:
1157
raise BzrError("inventory already contains entry with id {%s}" %
1159
self._byid[entry.file_id] = entry
1160
for child in getattr(entry, 'children', {}).itervalues():
1161
self._add_child(child)
1164
def add(self, entry):
1165
"""Add entry to inventory.
1167
To add a file to a branch ready to be committed, use Branch.add,
1170
Returns the new entry object.
1172
if entry.file_id in self._byid:
1173
raise errors.DuplicateFileId(entry.file_id,
1174
self._byid[entry.file_id])
1176
if entry.parent_id is None:
1177
assert self.root is None and len(self._byid) == 0
1181
parent = self._byid[entry.parent_id]
1183
raise BzrError("parent_id {%s} not in inventory" %
1186
if entry.name in parent.children:
1187
raise BzrError("%s is already versioned" %
1188
osutils.pathjoin(self.id2path(parent.file_id),
1189
entry.name).encode('utf-8'))
1190
parent.children[entry.name] = entry
1191
return self._add_child(entry)
1193
def add_path(self, relpath, kind, file_id=None, parent_id=None):
1194
"""Add entry from a path.
1196
The immediate parent must already be versioned.
1198
Returns the new entry object."""
1200
parts = osutils.splitpath(relpath)
1204
file_id = generate_ids.gen_root_id()
1205
self.root = InventoryDirectory(file_id, '', None)
1206
self._byid = {self.root.file_id: self.root}
1209
parent_path = parts[:-1]
1210
parent_id = self.path2id(parent_path)
1211
if parent_id is None:
1212
raise errors.NotVersionedError(path=parent_path)
1213
ie = make_entry(kind, parts[-1], parent_id, file_id)
1216
def __delitem__(self, file_id):
1217
"""Remove entry by id.
1219
>>> inv = Inventory()
1220
>>> inv.add(InventoryFile('123', 'foo.c', ROOT_ID))
1221
InventoryFile('123', 'foo.c', parent_id='TREE_ROOT', sha1=None, len=None)
1230
assert ie.parent_id is None or \
1231
self[ie.parent_id].children[ie.name] == ie
1233
del self._byid[file_id]
1234
if ie.parent_id is not None:
1235
del self[ie.parent_id].children[ie.name]
1237
def __eq__(self, other):
1238
"""Compare two sets by comparing their contents.
1240
>>> i1 = Inventory()
1241
>>> i2 = Inventory()
1244
>>> i1.add(InventoryFile('123', 'foo', ROOT_ID))
1245
InventoryFile('123', 'foo', parent_id='TREE_ROOT', sha1=None, len=None)
1248
>>> i2.add(InventoryFile('123', 'foo', ROOT_ID))
1249
InventoryFile('123', 'foo', parent_id='TREE_ROOT', sha1=None, len=None)
1253
if not isinstance(other, Inventory):
1254
return NotImplemented
1256
return self._byid == other._byid
1258
def __ne__(self, other):
1259
return not self.__eq__(other)
1262
raise ValueError('not hashable')
1264
def _iter_file_id_parents(self, file_id):
1265
"""Yield the parents of file_id up to the root."""
1266
while file_id is not None:
1268
ie = self._byid[file_id]
1270
raise errors.NoSuchId(tree=None, file_id=file_id)
1272
file_id = ie.parent_id
1274
def get_idpath(self, file_id):
1275
"""Return a list of file_ids for the path to an entry.
1277
The list contains one element for each directory followed by
1278
the id of the file itself. So the length of the returned list
1279
is equal to the depth of the file in the tree, counting the
1280
root directory as depth 1.
1283
for parent in self._iter_file_id_parents(file_id):
1284
p.insert(0, parent.file_id)
1287
def id2path(self, file_id):
1288
"""Return as a string the path to file_id.
1291
>>> e = i.add(InventoryDirectory('src-id', 'src', ROOT_ID))
1292
>>> e = i.add(InventoryFile('foo-id', 'foo.c', parent_id='src-id'))
1293
>>> print i.id2path('foo-id')
1296
# get all names, skipping root
1297
return '/'.join(reversed(
1298
[parent.name for parent in
1299
self._iter_file_id_parents(file_id)][:-1]))
1301
def path2id(self, name):
1302
"""Walk down through directories to return entry of last component.
1304
names may be either a list of path components, or a single
1305
string, in which case it is automatically split.
1307
This returns the entry of the last component in the path,
1308
which may be either a file or a directory.
1310
Returns None IFF the path is not found.
1312
if isinstance(name, basestring):
1313
name = osutils.splitpath(name)
1315
# mutter("lookup path %r" % name)
1322
children = getattr(parent, 'children', None)
1323
if children is None:
1326
assert cie.name == f
1327
assert cie.parent_id == parent.file_id
1330
# or raise an error?
1333
return parent.file_id
1335
def has_filename(self, names):
1336
return bool(self.path2id(names))
1338
def has_id(self, file_id):
1339
return (file_id in self._byid)
1341
def remove_recursive_id(self, file_id):
1342
"""Remove file_id, and children, from the inventory.
1344
:param file_id: A file_id to remove.
1346
to_find_delete = [self._byid[file_id]]
1348
while to_find_delete:
1349
ie = to_find_delete.pop()
1350
to_delete.append(ie.file_id)
1351
if ie.kind == 'directory':
1352
to_find_delete.extend(ie.children.values())
1353
for file_id in reversed(to_delete):
1355
del self._byid[file_id]
1356
if ie.parent_id is not None:
1357
del self[ie.parent_id].children[ie.name]
1361
def rename(self, file_id, new_parent_id, new_name):
1362
"""Move a file within the inventory.
1364
This can change either the name, or the parent, or both.
1366
This does not move the working file.
1368
new_name = ensure_normalized_name(new_name)
1369
if not is_valid_name(new_name):
1370
raise BzrError("not an acceptable filename: %r" % new_name)
1372
new_parent = self._byid[new_parent_id]
1373
if new_name in new_parent.children:
1374
raise BzrError("%r already exists in %r" % (new_name, self.id2path(new_parent_id)))
1376
new_parent_idpath = self.get_idpath(new_parent_id)
1377
if file_id in new_parent_idpath:
1378
raise BzrError("cannot move directory %r into a subdirectory of itself, %r"
1379
% (self.id2path(file_id), self.id2path(new_parent_id)))
1381
file_ie = self._byid[file_id]
1382
old_parent = self._byid[file_ie.parent_id]
1384
# TODO: Don't leave things messed up if this fails
1386
del old_parent.children[file_ie.name]
1387
new_parent.children[new_name] = file_ie
1389
file_ie.name = new_name
1390
file_ie.parent_id = new_parent_id
1392
def is_root(self, file_id):
1393
return self.root is not None and file_id == self.root.file_id
1397
'directory': InventoryDirectory,
1398
'file': InventoryFile,
1399
'symlink': InventoryLink,
1400
'tree-reference': TreeReference
1403
def make_entry(kind, name, parent_id, file_id=None):
1404
"""Create an inventory entry.
1406
:param kind: the type of inventory entry to create.
1407
:param name: the basename of the entry.
1408
:param parent_id: the parent_id of the entry.
1409
:param file_id: the file_id to use. if None, one will be created.
1412
file_id = generate_ids.gen_file_id(name)
1413
name = ensure_normalized_name(name)
1415
factory = entry_factory[kind]
1417
raise BzrError("unknown kind %r" % kind)
1418
return factory(file_id, name, parent_id)
1421
def ensure_normalized_name(name):
1424
:raises InvalidNormalization: When name is not normalized, and cannot be
1425
accessed on this platform by the normalized path.
1426
:return: The NFC/NFKC normalised version of name.
1428
#------- This has been copied to bzrlib.dirstate.DirState.add, please
1429
# keep them synchronised.
1430
# we dont import normalized_filename directly because we want to be
1431
# able to change the implementation at runtime for tests.
1432
norm_name, can_access = osutils.normalized_filename(name)
1433
if norm_name != name:
1437
# TODO: jam 20060701 This would probably be more useful
1438
# if the error was raised with the full path
1439
raise errors.InvalidNormalization(name)
1445
def is_valid_name(name):
1447
if _NAME_RE is None:
1448
_NAME_RE = re.compile(r'^[^/\\]+$')
1450
return bool(_NAME_RE.match(name))
added
removed
bzrlib/inventory.py
