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:
573
raise BzrCheckError('mismatched sha1 on {%s} in {%s}' %
574
(self.file_id, tree_revision_id))
576
checker.repeated_text_cnt += 1
579
if self.file_id not in checker.checked_weaves:
580
mutter('check weave {%s}', self.file_id)
581
w = tree._get_weave(self.file_id)
582
# Not passing a progress bar, because it creates a new
583
# progress, which overwrites the current progress,
584
# and doesn't look nice
586
checker.checked_weaves[self.file_id] = True
588
w = tree._get_weave(self.file_id)
590
mutter('check version {%s} of {%s}', tree_revision_id, self.file_id)
591
checker.checked_text_cnt += 1
592
# We can't check the length, because Weave doesn't store that
593
# information, and the whole point of looking at the weave's
594
# sha1sum is that we don't have to extract the text.
595
if self.text_sha1 != w.get_sha1(self.revision):
596
raise BzrCheckError('text {%s} version {%s} wrong sha1'
597
% (self.file_id, self.revision))
598
checker.checked_texts[t] = self.text_sha1
601
other = InventoryFile(self.file_id, self.name, self.parent_id)
602
other.executable = self.executable
603
other.text_id = self.text_id
604
other.text_sha1 = self.text_sha1
605
other.text_size = self.text_size
606
other.revision = self.revision
609
def detect_changes(self, old_entry):
610
"""See InventoryEntry.detect_changes."""
611
assert self.text_sha1 is not None
612
assert old_entry.text_sha1 is not None
613
text_modified = (self.text_sha1 != old_entry.text_sha1)
614
meta_modified = (self.executable != old_entry.executable)
615
return text_modified, meta_modified
617
def _diff(self, text_diff, from_label, tree, to_label, to_entry, to_tree,
618
output_to, reverse=False):
619
"""See InventoryEntry._diff."""
621
from_text = tree.get_file(self.file_id).readlines()
623
to_text = to_tree.get_file(to_entry.file_id).readlines()
627
text_diff(from_label, from_text,
628
to_label, to_text, output_to)
630
text_diff(to_label, to_text,
631
from_label, from_text, output_to)
632
except errors.BinaryFile:
634
label_pair = (to_label, from_label)
636
label_pair = (from_label, to_label)
637
print >> output_to, \
638
("Binary files %s and %s differ" % label_pair).encode('utf8')
641
"""See InventoryEntry.has_text."""
644
def __init__(self, file_id, name, parent_id):
645
super(InventoryFile, self).__init__(file_id, name, parent_id)
648
def kind_character(self):
649
"""See InventoryEntry.kind_character."""
652
def _put_in_tar(self, item, tree):
653
"""See InventoryEntry._put_in_tar."""
654
item.type = tarfile.REGTYPE
655
fileobj = tree.get_file(self.file_id)
656
item.size = self.text_size
657
if tree.is_executable(self.file_id):
663
def _put_on_disk(self, fullpath, tree):
664
"""See InventoryEntry._put_on_disk."""
665
osutils.pumpfile(tree.get_file(self.file_id), file(fullpath, 'wb'))
666
if tree.is_executable(self.file_id):
667
os.chmod(fullpath, 0755)
669
def _read_tree_state(self, path, work_tree):
670
"""See InventoryEntry._read_tree_state."""
671
self.text_sha1 = work_tree.get_file_sha1(self.file_id, path=path)
672
# FIXME: 20050930 probe for the text size when getting sha1
673
# in _read_tree_state
674
self.executable = work_tree.is_executable(self.file_id, path=path)
677
return ("%s(%r, %r, parent_id=%r, sha1=%r, len=%s)"
678
% (self.__class__.__name__,
685
def _forget_tree_state(self):
686
self.text_sha1 = None
688
def _unchanged(self, previous_ie):
689
"""See InventoryEntry._unchanged."""
690
compatible = super(InventoryFile, self)._unchanged(previous_ie)
691
if self.text_sha1 != previous_ie.text_sha1:
694
# FIXME: 20050930 probe for the text size when getting sha1
695
# in _read_tree_state
696
self.text_size = previous_ie.text_size
697
if self.executable != previous_ie.executable:
702
class InventoryLink(InventoryEntry):
703
"""A file in an inventory."""
705
__slots__ = ['text_sha1', 'text_size', 'file_id', 'name', 'kind',
706
'text_id', 'parent_id', 'children', 'executable',
707
'revision', 'symlink_target', 'reference_revision']
709
def _check(self, checker, rev_id, tree):
710
"""See InventoryEntry._check"""
711
if self.text_sha1 is not None or self.text_size is not None or self.text_id is not None:
712
raise BzrCheckError('symlink {%s} has text in revision {%s}'
713
% (self.file_id, rev_id))
714
if self.symlink_target is None:
715
raise BzrCheckError('symlink {%s} has no target in revision {%s}'
716
% (self.file_id, rev_id))
719
other = InventoryLink(self.file_id, self.name, self.parent_id)
720
other.symlink_target = self.symlink_target
721
other.revision = self.revision
724
def detect_changes(self, old_entry):
725
"""See InventoryEntry.detect_changes."""
726
# FIXME: which _modified field should we use ? RBC 20051003
727
text_modified = (self.symlink_target != old_entry.symlink_target)
729
mutter(" symlink target changed")
730
meta_modified = False
731
return text_modified, meta_modified
733
def _diff(self, text_diff, from_label, tree, to_label, to_entry, to_tree,
734
output_to, reverse=False):
735
"""See InventoryEntry._diff."""
736
from_text = self.symlink_target
737
if to_entry is not None:
738
to_text = to_entry.symlink_target
743
print >>output_to, '=== target changed %r => %r' % (from_text, to_text)
746
print >>output_to, '=== target was %r' % self.symlink_target
748
print >>output_to, '=== target is %r' % self.symlink_target
750
def __init__(self, file_id, name, parent_id):
751
super(InventoryLink, self).__init__(file_id, name, parent_id)
752
self.kind = 'symlink'
754
def kind_character(self):
755
"""See InventoryEntry.kind_character."""
758
def _put_in_tar(self, item, tree):
759
"""See InventoryEntry._put_in_tar."""
760
item.type = tarfile.SYMTYPE
764
item.linkname = self.symlink_target
767
def _put_on_disk(self, fullpath, tree):
768
"""See InventoryEntry._put_on_disk."""
770
os.symlink(self.symlink_target, fullpath)
772
raise BzrError("Failed to create symlink %r -> %r, error: %s" % (fullpath, self.symlink_target, e))
774
def _read_tree_state(self, path, work_tree):
775
"""See InventoryEntry._read_tree_state."""
776
self.symlink_target = work_tree.get_symlink_target(self.file_id)
778
def _forget_tree_state(self):
779
self.symlink_target = None
781
def _unchanged(self, previous_ie):
782
"""See InventoryEntry._unchanged."""
783
compatible = super(InventoryLink, self)._unchanged(previous_ie)
784
if self.symlink_target != previous_ie.symlink_target:
789
class TreeReference(InventoryEntry):
791
kind = 'tree-reference'
793
def __init__(self, file_id, name, parent_id, revision=None,
794
reference_revision=None):
795
InventoryEntry.__init__(self, file_id, name, parent_id)
796
self.revision = revision
797
self.reference_revision = reference_revision
800
return TreeReference(self.file_id, self.name, self.parent_id,
801
self.revision, self.reference_revision)
803
def _read_tree_state(self, path, work_tree):
804
"""Populate fields in the inventory entry from the given tree.
806
self.reference_revision = work_tree.get_reference_revision(
809
def _forget_tree_state(self):
810
self.reference_revision = None
812
def _unchanged(self, previous_ie):
813
"""See InventoryEntry._unchanged."""
814
compatible = super(TreeReference, self)._unchanged(previous_ie)
815
if self.reference_revision != previous_ie.reference_revision:
820
class Inventory(object):
821
"""Inventory of versioned files in a tree.
823
This describes which file_id is present at each point in the tree,
824
and possibly the SHA-1 or other information about the file.
825
Entries can be looked up either by path or by file_id.
827
The inventory represents a typical unix file tree, with
828
directories containing files and subdirectories. We never store
829
the full path to a file, because renaming a directory implicitly
830
moves all of its contents. This class internally maintains a
831
lookup tree that allows the children under a directory to be
834
InventoryEntry objects must not be modified after they are
835
inserted, other than through the Inventory API.
837
>>> inv = Inventory()
838
>>> inv.add(InventoryFile('123-123', 'hello.c', ROOT_ID))
839
InventoryFile('123-123', 'hello.c', parent_id='TREE_ROOT', sha1=None, len=None)
840
>>> inv['123-123'].name
843
May be treated as an iterator or set to look up file ids:
845
>>> bool(inv.path2id('hello.c'))
850
May also look up by name:
852
>>> [x[0] for x in inv.iter_entries()]
854
>>> inv = Inventory('TREE_ROOT-12345678-12345678')
855
>>> inv.add(InventoryFile('123-123', 'hello.c', ROOT_ID))
856
Traceback (most recent call last):
857
BzrError: parent_id {TREE_ROOT} not in inventory
858
>>> inv.add(InventoryFile('123-123', 'hello.c', 'TREE_ROOT-12345678-12345678'))
859
InventoryFile('123-123', 'hello.c', parent_id='TREE_ROOT-12345678-12345678', sha1=None, len=None)
861
def __init__(self, root_id=ROOT_ID, revision_id=None):
862
"""Create or read an inventory.
864
If a working directory is specified, the inventory is read
865
from there. If the file is specified, read from that. If not,
866
the inventory is created empty.
868
The inventory is created with a default root directory, with
871
if root_id is not None:
872
assert root_id.__class__ == str
873
self._set_root(InventoryDirectory(root_id, u'', None))
877
self.revision_id = revision_id
880
return "<Inventory object at %x, contents=%r>" % (id(self), self._byid)
882
def apply_delta(self, delta):
883
"""Apply a delta to this inventory.
885
:param delta: A list of changes to apply. After all the changes are
886
applied the final inventory must be internally consistent, but it
887
is ok to supply changes which, if only half-applied would have an
888
invalid result - such as supplying two changes which rename two
889
files, 'A' and 'B' with each other : [('A', 'B', 'A-id', a_entry),
890
('B', 'A', 'B-id', b_entry)].
892
Each change is a tuple, of the form (old_path, new_path, file_id,
895
When new_path is None, the change indicates the removal of an entry
896
from the inventory and new_entry will be ignored (using None is
897
appropriate). If new_path is not None, then new_entry must be an
898
InventoryEntry instance, which will be incorporated into the
899
inventory (and replace any existing entry with the same file id).
901
When old_path is None, the change indicates the addition of
902
a new entry to the inventory.
904
When neither new_path nor old_path are None, the change is a
905
modification to an entry, such as a rename, reparent, kind change
908
The children attribute of new_entry is ignored. This is because
909
this method preserves children automatically across alterations to
910
the parent of the children, and cases where the parent id of a
911
child is changing require the child to be passed in as a separate
912
change regardless. E.g. in the recursive deletion of a directory -
913
the directory's children must be included in the delta, or the
914
final inventory will be invalid.
917
# Remove all affected items which were in the original inventory,
918
# starting with the longest paths, thus ensuring parents are examined
919
# after their children, which means that everything we examine has no
920
# modified children remaining by the time we examine it.
921
for old_path, file_id in sorted(((op, f) for op, np, f, e in delta
922
if op is not None), reverse=True):
923
if file_id not in self:
926
# Preserve unaltered children of file_id for later reinsertion.
927
children[file_id] = getattr(self[file_id], 'children', {})
928
# Remove file_id and the unaltered children. If file_id is not
929
# being deleted it will be reinserted back later.
930
self.remove_recursive_id(file_id)
931
# Insert all affected which should be in the new inventory, reattaching
932
# their children if they had any. This is done from shortest path to
933
# longest, ensuring that items which were modified and whose parents in
934
# the resulting inventory were also modified, are inserted after their
936
for new_path, new_entry in sorted((np, e) for op, np, f, e in
937
delta if np is not None):
938
if new_entry.kind == 'directory':
939
new_entry.children = children.get(new_entry.file_id, {})
942
def _set_root(self, ie):
944
self._byid = {self.root.file_id: self.root}
947
# TODO: jam 20051218 Should copy also copy the revision_id?
948
entries = self.iter_entries()
949
other = Inventory(entries.next()[1].file_id)
950
# copy recursively so we know directories will be added before
951
# their children. There are more efficient ways than this...
952
for path, entry in entries():
953
other.add(entry.copy())
957
return iter(self._byid)
960
"""Returns number of entries."""
961
return len(self._byid)
963
def iter_entries(self, from_dir=None):
964
"""Return (path, entry) pairs, in order by name."""
966
if self.root is None:
970
elif isinstance(from_dir, basestring):
971
from_dir = self._byid[from_dir]
973
# unrolling the recursive called changed the time from
974
# 440ms/663ms (inline/total) to 116ms/116ms
975
children = from_dir.children.items()
977
children = collections.deque(children)
978
stack = [(u'', children)]
980
from_dir_relpath, children = stack[-1]
983
name, ie = children.popleft()
985
# we know that from_dir_relpath never ends in a slash
986
# and 'f' doesn't begin with one, we can do a string op, rather
987
# than the checks of pathjoin(), though this means that all paths
989
path = from_dir_relpath + '/' + name
993
if ie.kind != 'directory':
996
# But do this child first
997
new_children = ie.children.items()
999
new_children = collections.deque(new_children)
1000
stack.append((path, new_children))
1001
# Break out of inner loop, so that we start outer loop with child
1004
# if we finished all children, pop it off the stack
1007
def iter_entries_by_dir(self, from_dir=None, specific_file_ids=None,
1008
yield_parents=False):
1009
"""Iterate over the entries in a directory first order.
1011
This returns all entries for a directory before returning
1012
the entries for children of a directory. This is not
1013
lexicographically sorted order, and is a hybrid between
1014
depth-first and breadth-first.
1016
:param yield_parents: If True, yield the parents from the root leading
1017
down to specific_file_ids that have been requested. This has no
1018
impact if specific_file_ids is None.
1019
:return: This yields (path, entry) pairs
1021
if specific_file_ids and not isinstance(specific_file_ids, set):
1022
specific_file_ids = set(specific_file_ids)
1023
# TODO? Perhaps this should return the from_dir so that the root is
1024
# yielded? or maybe an option?
1025
if from_dir is None:
1026
if self.root is None:
1028
# Optimize a common case
1029
if (not yield_parents and specific_file_ids is not None and
1030
len(specific_file_ids) == 1):
1031
file_id = list(specific_file_ids)[0]
1033
yield self.id2path(file_id), self[file_id]
1035
from_dir = self.root
1036
if (specific_file_ids is None or yield_parents or
1037
self.root.file_id in specific_file_ids):
1038
yield u'', self.root
1039
elif isinstance(from_dir, basestring):
1040
from_dir = self._byid[from_dir]
1042
if specific_file_ids is not None:
1043
# TODO: jam 20070302 This could really be done as a loop rather
1044
# than a bunch of recursive calls.
1047
def add_ancestors(file_id):
1048
if file_id not in byid:
1050
parent_id = byid[file_id].parent_id
1051
if parent_id is None:
1053
if parent_id not in parents:
1054
parents.add(parent_id)
1055
add_ancestors(parent_id)
1056
for file_id in specific_file_ids:
1057
add_ancestors(file_id)
1061
stack = [(u'', from_dir)]
1063
cur_relpath, cur_dir = stack.pop()
1066
for child_name, child_ie in sorted(cur_dir.children.iteritems()):
1068
child_relpath = cur_relpath + child_name
1070
if (specific_file_ids is None or
1071
child_ie.file_id in specific_file_ids or
1072
(yield_parents and child_ie.file_id in parents)):
1073
yield child_relpath, child_ie
1075
if child_ie.kind == 'directory':
1076
if parents is None or child_ie.file_id in parents:
1077
child_dirs.append((child_relpath+'/', child_ie))
1078
stack.extend(reversed(child_dirs))
1080
def make_entry(self, kind, name, parent_id, file_id=None):
1081
"""Simple thunk to bzrlib.inventory.make_entry."""
1082
return make_entry(kind, name, parent_id, file_id)
1085
"""Return list of (path, ie) for all entries except the root.
1087
This may be faster than iter_entries.
1090
def descend(dir_ie, dir_path):
1091
kids = dir_ie.children.items()
1093
for name, ie in kids:
1094
child_path = osutils.pathjoin(dir_path, name)
1095
accum.append((child_path, ie))
1096
if ie.kind == 'directory':
1097
descend(ie, child_path)
1099
descend(self.root, u'')
1102
def directories(self):
1103
"""Return (path, entry) pairs for all directories, including the root.
1106
def descend(parent_ie, parent_path):
1107
accum.append((parent_path, parent_ie))
1109
kids = [(ie.name, ie) for ie in parent_ie.children.itervalues() if ie.kind == 'directory']
1112
for name, child_ie in kids:
1113
child_path = osutils.pathjoin(parent_path, name)
1114
descend(child_ie, child_path)
1115
descend(self.root, u'')
1118
def __contains__(self, file_id):
1119
"""True if this entry contains a file with given id.
1121
>>> inv = Inventory()
1122
>>> inv.add(InventoryFile('123', 'foo.c', ROOT_ID))
1123
InventoryFile('123', 'foo.c', parent_id='TREE_ROOT', sha1=None, len=None)
1129
return (file_id in self._byid)
1131
def __getitem__(self, file_id):
1132
"""Return the entry for given file_id.
1134
>>> inv = Inventory()
1135
>>> inv.add(InventoryFile('123123', 'hello.c', ROOT_ID))
1136
InventoryFile('123123', 'hello.c', parent_id='TREE_ROOT', sha1=None, len=None)
1137
>>> inv['123123'].name
1141
return self._byid[file_id]
1143
# really we're passing an inventory, not a tree...
1144
raise errors.NoSuchId(self, file_id)
1146
def get_file_kind(self, file_id):
1147
return self._byid[file_id].kind
1149
def get_child(self, parent_id, filename):
1150
return self[parent_id].children.get(filename)
1152
def _add_child(self, entry):
1153
"""Add an entry to the inventory, without adding it to its parent"""
1154
if entry.file_id in self._byid:
1155
raise BzrError("inventory already contains entry with id {%s}" %
1157
self._byid[entry.file_id] = entry
1158
for child in getattr(entry, 'children', {}).itervalues():
1159
self._add_child(child)
1162
def add(self, entry):
1163
"""Add entry to inventory.
1165
To add a file to a branch ready to be committed, use Branch.add,
1168
Returns the new entry object.
1170
if entry.file_id in self._byid:
1171
raise errors.DuplicateFileId(entry.file_id,
1172
self._byid[entry.file_id])
1174
if entry.parent_id is None:
1175
assert self.root is None and len(self._byid) == 0
1179
parent = self._byid[entry.parent_id]
1181
raise BzrError("parent_id {%s} not in inventory" %
1184
if entry.name in parent.children:
1185
raise BzrError("%s is already versioned" %
1186
osutils.pathjoin(self.id2path(parent.file_id),
1187
entry.name).encode('utf-8'))
1188
parent.children[entry.name] = entry
1189
return self._add_child(entry)
1191
def add_path(self, relpath, kind, file_id=None, parent_id=None):
1192
"""Add entry from a path.
1194
The immediate parent must already be versioned.
1196
Returns the new entry object."""
1198
parts = osutils.splitpath(relpath)
1202
file_id = generate_ids.gen_root_id()
1203
self.root = InventoryDirectory(file_id, '', None)
1204
self._byid = {self.root.file_id: self.root}
1207
parent_path = parts[:-1]
1208
parent_id = self.path2id(parent_path)
1209
if parent_id is None:
1210
raise errors.NotVersionedError(path=parent_path)
1211
ie = make_entry(kind, parts[-1], parent_id, file_id)
1214
def __delitem__(self, file_id):
1215
"""Remove entry by id.
1217
>>> inv = Inventory()
1218
>>> inv.add(InventoryFile('123', 'foo.c', ROOT_ID))
1219
InventoryFile('123', 'foo.c', parent_id='TREE_ROOT', sha1=None, len=None)
1228
assert ie.parent_id is None or \
1229
self[ie.parent_id].children[ie.name] == ie
1231
del self._byid[file_id]
1232
if ie.parent_id is not None:
1233
del self[ie.parent_id].children[ie.name]
1235
def __eq__(self, other):
1236
"""Compare two sets by comparing their contents.
1238
>>> i1 = Inventory()
1239
>>> i2 = Inventory()
1242
>>> i1.add(InventoryFile('123', 'foo', ROOT_ID))
1243
InventoryFile('123', 'foo', parent_id='TREE_ROOT', sha1=None, len=None)
1246
>>> i2.add(InventoryFile('123', 'foo', ROOT_ID))
1247
InventoryFile('123', 'foo', parent_id='TREE_ROOT', sha1=None, len=None)
1251
if not isinstance(other, Inventory):
1252
return NotImplemented
1254
return self._byid == other._byid
1256
def __ne__(self, other):
1257
return not self.__eq__(other)
1260
raise ValueError('not hashable')
1262
def _iter_file_id_parents(self, file_id):
1263
"""Yield the parents of file_id up to the root."""
1264
while file_id is not None:
1266
ie = self._byid[file_id]
1268
raise errors.NoSuchId(tree=None, file_id=file_id)
1270
file_id = ie.parent_id
1272
def get_idpath(self, file_id):
1273
"""Return a list of file_ids for the path to an entry.
1275
The list contains one element for each directory followed by
1276
the id of the file itself. So the length of the returned list
1277
is equal to the depth of the file in the tree, counting the
1278
root directory as depth 1.
1281
for parent in self._iter_file_id_parents(file_id):
1282
p.insert(0, parent.file_id)
1285
def id2path(self, file_id):
1286
"""Return as a string the path to file_id.
1289
>>> e = i.add(InventoryDirectory('src-id', 'src', ROOT_ID))
1290
>>> e = i.add(InventoryFile('foo-id', 'foo.c', parent_id='src-id'))
1291
>>> print i.id2path('foo-id')
1294
# get all names, skipping root
1295
return '/'.join(reversed(
1296
[parent.name for parent in
1297
self._iter_file_id_parents(file_id)][:-1]))
1299
def path2id(self, name):
1300
"""Walk down through directories to return entry of last component.
1302
names may be either a list of path components, or a single
1303
string, in which case it is automatically split.
1305
This returns the entry of the last component in the path,
1306
which may be either a file or a directory.
1308
Returns None IFF the path is not found.
1310
if isinstance(name, basestring):
1311
name = osutils.splitpath(name)
1313
# mutter("lookup path %r" % name)
1320
children = getattr(parent, 'children', None)
1321
if children is None:
1324
assert cie.name == f
1325
assert cie.parent_id == parent.file_id
1328
# or raise an error?
1331
return parent.file_id
1333
def has_filename(self, names):
1334
return bool(self.path2id(names))
1336
def has_id(self, file_id):
1337
return (file_id in self._byid)
1339
def remove_recursive_id(self, file_id):
1340
"""Remove file_id, and children, from the inventory.
1342
:param file_id: A file_id to remove.
1344
to_find_delete = [self._byid[file_id]]
1346
while to_find_delete:
1347
ie = to_find_delete.pop()
1348
to_delete.append(ie.file_id)
1349
if ie.kind == 'directory':
1350
to_find_delete.extend(ie.children.values())
1351
for file_id in reversed(to_delete):
1353
del self._byid[file_id]
1354
if ie.parent_id is not None:
1355
del self[ie.parent_id].children[ie.name]
1359
def rename(self, file_id, new_parent_id, new_name):
1360
"""Move a file within the inventory.
1362
This can change either the name, or the parent, or both.
1364
This does not move the working file.
1366
new_name = ensure_normalized_name(new_name)
1367
if not is_valid_name(new_name):
1368
raise BzrError("not an acceptable filename: %r" % new_name)
1370
new_parent = self._byid[new_parent_id]
1371
if new_name in new_parent.children:
1372
raise BzrError("%r already exists in %r" % (new_name, self.id2path(new_parent_id)))
1374
new_parent_idpath = self.get_idpath(new_parent_id)
1375
if file_id in new_parent_idpath:
1376
raise BzrError("cannot move directory %r into a subdirectory of itself, %r"
1377
% (self.id2path(file_id), self.id2path(new_parent_id)))
1379
file_ie = self._byid[file_id]
1380
old_parent = self._byid[file_ie.parent_id]
1382
# TODO: Don't leave things messed up if this fails
1384
del old_parent.children[file_ie.name]
1385
new_parent.children[new_name] = file_ie
1387
file_ie.name = new_name
1388
file_ie.parent_id = new_parent_id
1390
def is_root(self, file_id):
1391
return self.root is not None and file_id == self.root.file_id
1395
'directory': InventoryDirectory,
1396
'file': InventoryFile,
1397
'symlink': InventoryLink,
1398
'tree-reference': TreeReference
1401
def make_entry(kind, name, parent_id, file_id=None):
1402
"""Create an inventory entry.
1404
:param kind: the type of inventory entry to create.
1405
:param name: the basename of the entry.
1406
:param parent_id: the parent_id of the entry.
1407
:param file_id: the file_id to use. if None, one will be created.
1410
file_id = generate_ids.gen_file_id(name)
1411
name = ensure_normalized_name(name)
1413
factory = entry_factory[kind]
1415
raise BzrError("unknown kind %r" % kind)
1416
return factory(file_id, name, parent_id)
1419
def ensure_normalized_name(name):
1422
:raises InvalidNormalization: When name is not normalized, and cannot be
1423
accessed on this platform by the normalized path.
1424
:return: The NFC/NFKC normalised version of name.
1426
#------- This has been copied to bzrlib.dirstate.DirState.add, please
1427
# keep them synchronised.
1428
# we dont import normalized_filename directly because we want to be
1429
# able to change the implementation at runtime for tests.
1430
norm_name, can_access = osutils.normalized_filename(name)
1431
if norm_name != name:
1435
# TODO: jam 20060701 This would probably be more useful
1436
# if the error was raised with the full path
1437
raise errors.InvalidNormalization(name)
1443
def is_valid_name(name):
1445
if _NAME_RE is None:
1446
_NAME_RE = re.compile(r'^[^/\\]+$')
1448
return bool(_NAME_RE.match(name))
added
removed
bzrlib/inventory.py
