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
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
@deprecated_method(symbol_versioning.one_zero)
146
def diff(self, text_diff, from_label, tree, to_label, to_entry, to_tree,
147
output_to, reverse=False):
148
"""Perform a diff from this to to_entry.
150
text_diff will be used for textual difference calculation.
151
This is a template method, override _diff in child classes.
153
self._read_tree_state(tree.id2path(self.file_id), tree)
155
# cannot diff from one kind to another - you must do a removal
156
# and an addif they do not match.
157
assert self.kind == to_entry.kind
158
to_entry._read_tree_state(to_tree.id2path(to_entry.file_id),
160
self._diff(text_diff, from_label, tree, to_label, to_entry, to_tree,
163
def _diff(self, text_diff, from_label, tree, to_label, to_entry, to_tree,
164
output_to, reverse=False):
165
"""Perform a diff between two entries of the same kind."""
167
def parent_candidates(self, previous_inventories):
168
"""Find possible per-file graph parents.
170
This is currently defined by:
171
- Select the last changed revision in the parent inventory.
172
- Do deal with a short lived bug in bzr 0.8's development two entries
173
that have the same last changed but different 'x' bit settings are
176
# revision:ie mapping for each ie found in previous_inventories.
178
# identify candidate head revision ids.
179
for inv in previous_inventories:
180
if self.file_id in inv:
181
ie = inv[self.file_id]
182
assert ie.file_id == self.file_id
183
if ie.revision in candidates:
184
# same revision value in two different inventories:
185
# correct possible inconsistencies:
186
# * there was a bug in revision updates with 'x' bit
189
if candidates[ie.revision].executable != ie.executable:
190
candidates[ie.revision].executable = False
191
ie.executable = False
192
except AttributeError:
194
# must now be the same.
195
assert candidates[ie.revision] == ie
197
# add this revision as a candidate.
198
candidates[ie.revision] = ie
201
@deprecated_method(symbol_versioning.zero_ninetyone)
202
def find_previous_heads(self, previous_inventories,
203
versioned_file_store,
206
"""Return the revisions and entries that directly precede this.
208
Returned as a map from revision to inventory entry.
210
This is a map containing the file revisions in all parents
211
for which the file exists, and its revision is not a parent of
212
any other. If the file is new, the set will be empty.
214
:param versioned_file_store: A store where ancestry data on this
215
file id can be queried.
216
:param transaction: The transaction that queries to the versioned
217
file store should be completed under.
218
:param entry_vf: The entry versioned file, if its already available.
220
candidates = self.parent_candidates(previous_inventories)
222
# revision:ie mapping with one revision for each head.
224
# common case optimisation
225
if len(candidates) == 1:
226
# if there is only one candidate revision found
227
# then we can avoid opening the versioned file to access ancestry:
228
# there cannot be any ancestors to eliminate when there is
229
# only one revision available.
232
# --- what follows is now encapsulated in repository.get_graph.heads(),
233
# but that is not accessible from here as we have no repository
234
# pointer. Note that the repository.get_graph.heads() call can return
235
# different results *at the moment* because of the kind-changing check
236
# we have in parent_candidates().
238
# eliminate ancestors amongst the available candidates:
239
# heads are those that are not an ancestor of any other candidate
240
# - this provides convergence at a per-file level.
241
def get_ancestors(weave, entry):
242
return set(weave.get_ancestry(entry.revision, topo_sorted=False))
243
# revision: ancestor list for each head
245
for ie in candidates.values():
246
# may be an ancestor of a known head:
247
already_present = 0 != len(
248
[head for head in heads
249
if ie.revision in head_ancestors[head]])
251
# an ancestor of an analyzed candidate.
253
# not an ancestor of a known head:
254
# load the versioned file for this file id if needed
256
entry_vf = versioned_file_store.get_weave_or_empty(
257
self.file_id, transaction)
258
ancestors = get_ancestors(entry_vf, ie)
259
# may knock something else out:
260
check_heads = list(heads.keys())
261
for head in check_heads:
262
if head in ancestors:
263
# this previously discovered 'head' is not
264
# really a head - its an ancestor of the newly
267
head_ancestors[ie.revision] = ancestors
268
heads[ie.revision] = ie
271
def get_tar_item(self, root, dp, now, tree):
272
"""Get a tarfile item and a file stream for its content."""
273
item = tarfile.TarInfo(osutils.pathjoin(root, dp).encode('utf8'))
274
# TODO: would be cool to actually set it to the timestamp of the
275
# revision it was last changed
277
fileobj = self._put_in_tar(item, tree)
281
"""Return true if the object this entry represents has textual data.
283
Note that textual data includes binary content.
285
Also note that all entries get weave files created for them.
286
This attribute is primarily used when upgrading from old trees that
287
did not have the weave index for all inventory entries.
291
def __init__(self, file_id, name, parent_id, text_id=None):
292
"""Create an InventoryEntry
294
The filename must be a single component, relative to the
295
parent directory; it cannot be a whole path or relative name.
297
>>> e = InventoryFile('123', 'hello.c', ROOT_ID)
302
>>> e = InventoryFile('123', 'src/hello.c', ROOT_ID)
303
Traceback (most recent call last):
304
InvalidEntryName: Invalid entry name: src/hello.c
306
assert isinstance(name, basestring), name
307
if '/' in name or '\\' in name:
308
raise errors.InvalidEntryName(name=name)
309
self.executable = False
311
self.text_sha1 = None
312
self.text_size = None
313
self.file_id = file_id
314
assert isinstance(file_id, (str, None.__class__)), \
315
'bad type %r for %r' % (type(file_id), file_id)
317
self.text_id = text_id
318
self.parent_id = parent_id
319
self.symlink_target = None
320
self.reference_revision = None
322
def kind_character(self):
323
"""Return a short kind indicator useful for appending to names."""
324
raise BzrError('unknown kind %r' % self.kind)
326
known_kinds = ('file', 'directory', 'symlink')
328
def _put_in_tar(self, item, tree):
329
"""populate item for stashing in a tar, and return the content stream.
331
If no content is available, return None.
333
raise BzrError("don't know how to export {%s} of kind %r" %
334
(self.file_id, self.kind))
336
def put_on_disk(self, dest, dp, tree):
337
"""Create a representation of self on disk in the prefix dest.
339
This is a template method - implement _put_on_disk in subclasses.
341
fullpath = osutils.pathjoin(dest, dp)
342
self._put_on_disk(fullpath, tree)
343
# mutter(" export {%s} kind %s to %s", self.file_id,
344
# self.kind, fullpath)
346
def _put_on_disk(self, fullpath, tree):
347
"""Put this entry onto disk at fullpath, from tree tree."""
348
raise BzrError("don't know how to export {%s} of kind %r" % (self.file_id, self.kind))
350
def sorted_children(self):
351
return sorted(self.children.items())
354
def versionable_kind(kind):
355
return (kind in ('file', 'directory', 'symlink', 'tree-reference'))
357
def check(self, checker, rev_id, inv, tree):
358
"""Check this inventory entry is intact.
360
This is a template method, override _check for kind specific
363
:param checker: Check object providing context for the checks;
364
can be used to find out what parts of the repository have already
366
:param rev_id: Revision id from which this InventoryEntry was loaded.
367
Not necessarily the last-changed revision for this file.
368
:param inv: Inventory from which the entry was loaded.
369
:param tree: RevisionTree for this entry.
371
if self.parent_id is not None:
372
if not inv.has_id(self.parent_id):
373
raise BzrCheckError('missing parent {%s} in inventory for revision {%s}'
374
% (self.parent_id, rev_id))
375
self._check(checker, rev_id, tree)
377
def _check(self, checker, rev_id, tree):
378
"""Check this inventory entry for kind specific errors."""
379
raise BzrCheckError('unknown entry kind %r in revision {%s}' %
383
"""Clone this inventory entry."""
384
raise NotImplementedError
387
def describe_change(old_entry, new_entry):
388
"""Describe the change between old_entry and this.
390
This smells of being an InterInventoryEntry situation, but as its
391
the first one, we're making it a static method for now.
393
An entry with a different parent, or different name is considered
394
to be renamed. Reparenting is an internal detail.
395
Note that renaming the parent does not trigger a rename for the
398
# TODO: Perhaps return an object rather than just a string
399
if old_entry is new_entry:
400
# also the case of both being None
402
elif old_entry is None:
404
elif new_entry is None:
406
if old_entry.kind != new_entry.kind:
408
text_modified, meta_modified = new_entry.detect_changes(old_entry)
409
if text_modified or meta_modified:
413
# TODO 20060511 (mbp, rbc) factor out 'detect_rename' here.
414
if old_entry.parent_id != new_entry.parent_id:
416
elif old_entry.name != new_entry.name:
420
if renamed and not modified:
421
return InventoryEntry.RENAMED
422
if modified and not renamed:
424
if modified and renamed:
425
return InventoryEntry.MODIFIED_AND_RENAMED
429
return ("%s(%r, %r, parent_id=%r, revision=%r)"
430
% (self.__class__.__name__,
436
def __eq__(self, other):
437
if not isinstance(other, InventoryEntry):
438
return NotImplemented
440
return ((self.file_id == other.file_id)
441
and (self.name == other.name)
442
and (other.symlink_target == self.symlink_target)
443
and (self.text_sha1 == other.text_sha1)
444
and (self.text_size == other.text_size)
445
and (self.text_id == other.text_id)
446
and (self.parent_id == other.parent_id)
447
and (self.kind == other.kind)
448
and (self.revision == other.revision)
449
and (self.executable == other.executable)
450
and (self.reference_revision == other.reference_revision)
453
def __ne__(self, other):
454
return not (self == other)
457
raise ValueError('not hashable')
459
def _unchanged(self, previous_ie):
460
"""Has this entry changed relative to previous_ie.
462
This method should be overridden in child classes.
465
# different inv parent
466
if previous_ie.parent_id != self.parent_id:
469
elif previous_ie.name != self.name:
471
elif previous_ie.kind != self.kind:
475
def _read_tree_state(self, path, work_tree):
476
"""Populate fields in the inventory entry from the given tree.
478
Note that this should be modified to be a noop on virtual trees
479
as all entries created there are prepopulated.
481
# TODO: Rather than running this manually, we should check the
482
# working sha1 and other expensive properties when they're
483
# first requested, or preload them if they're already known
484
pass # nothing to do by default
486
def _forget_tree_state(self):
490
class RootEntry(InventoryEntry):
492
__slots__ = ['text_sha1', 'text_size', 'file_id', 'name', 'kind',
493
'text_id', 'parent_id', 'children', 'executable',
494
'revision', 'symlink_target', 'reference_revision']
496
def _check(self, checker, rev_id, tree):
497
"""See InventoryEntry._check"""
499
def __init__(self, file_id):
500
self.file_id = file_id
502
self.kind = 'directory'
503
self.parent_id = None
506
symbol_versioning.warn('RootEntry is deprecated as of bzr 0.10.'
507
' Please use InventoryDirectory instead.',
508
DeprecationWarning, stacklevel=2)
510
def __eq__(self, other):
511
if not isinstance(other, RootEntry):
512
return NotImplemented
514
return (self.file_id == other.file_id) \
515
and (self.children == other.children)
518
class InventoryDirectory(InventoryEntry):
519
"""A directory in an inventory."""
521
__slots__ = ['text_sha1', 'text_size', 'file_id', 'name', 'kind',
522
'text_id', 'parent_id', 'children', 'executable',
523
'revision', 'symlink_target', 'reference_revision']
525
def _check(self, checker, rev_id, tree):
526
"""See InventoryEntry._check"""
527
if self.text_sha1 is not None or self.text_size is not None or self.text_id is not None:
528
raise BzrCheckError('directory {%s} has text in revision {%s}'
529
% (self.file_id, rev_id))
532
other = InventoryDirectory(self.file_id, self.name, self.parent_id)
533
other.revision = self.revision
534
# note that children are *not* copied; they're pulled across when
538
def __init__(self, file_id, name, parent_id):
539
super(InventoryDirectory, self).__init__(file_id, name, parent_id)
541
self.kind = 'directory'
543
def kind_character(self):
544
"""See InventoryEntry.kind_character."""
547
def _put_in_tar(self, item, tree):
548
"""See InventoryEntry._put_in_tar."""
549
item.type = tarfile.DIRTYPE
556
def _put_on_disk(self, fullpath, tree):
557
"""See InventoryEntry._put_on_disk."""
561
class InventoryFile(InventoryEntry):
562
"""A file in an inventory."""
564
__slots__ = ['text_sha1', 'text_size', 'file_id', 'name', 'kind',
565
'text_id', 'parent_id', 'children', 'executable',
566
'revision', 'symlink_target', 'reference_revision']
568
def _check(self, checker, tree_revision_id, tree):
569
"""See InventoryEntry._check"""
570
t = (self.file_id, self.revision)
571
if t in checker.checked_texts:
572
prev_sha = checker.checked_texts[t]
573
if prev_sha != self.text_sha1:
575
'mismatched sha1 on {%s} in {%s} (%s != %s) %r' %
576
(self.file_id, tree_revision_id, prev_sha, self.text_sha1,
579
checker.repeated_text_cnt += 1
582
if self.file_id not in checker.checked_weaves:
583
mutter('check weave {%s}', self.file_id)
584
w = tree._get_weave(self.file_id)
585
# Not passing a progress bar, because it creates a new
586
# progress, which overwrites the current progress,
587
# and doesn't look nice
589
checker.checked_weaves[self.file_id] = True
591
w = tree._get_weave(self.file_id)
593
mutter('check version {%s} of {%s}', tree_revision_id, self.file_id)
594
checker.checked_text_cnt += 1
595
# We can't check the length, because Weave doesn't store that
596
# information, and the whole point of looking at the weave's
597
# sha1sum is that we don't have to extract the text.
598
if self.text_sha1 != w.get_sha1(self.revision):
599
raise BzrCheckError('text {%s} version {%s} wrong sha1'
600
% (self.file_id, self.revision))
601
checker.checked_texts[t] = self.text_sha1
604
other = InventoryFile(self.file_id, self.name, self.parent_id)
605
other.executable = self.executable
606
other.text_id = self.text_id
607
other.text_sha1 = self.text_sha1
608
other.text_size = self.text_size
609
other.revision = self.revision
612
def detect_changes(self, old_entry):
613
"""See InventoryEntry.detect_changes."""
614
assert self.text_sha1 is not None
615
assert old_entry.text_sha1 is not None
616
text_modified = (self.text_sha1 != old_entry.text_sha1)
617
meta_modified = (self.executable != old_entry.executable)
618
return text_modified, meta_modified
620
def _diff(self, text_diff, from_label, tree, to_label, to_entry, to_tree,
621
output_to, reverse=False):
622
"""See InventoryEntry._diff."""
623
from bzrlib.diff import DiffText
624
from_file_id = self.file_id
626
to_file_id = to_entry.file_id
630
to_file_id, from_file_id = from_file_id, to_file_id
631
tree, to_tree = to_tree, tree
632
from_label, to_label = to_label, from_label
633
differ = DiffText(tree, to_tree, output_to, 'utf-8', '', '',
635
return differ.diff_text(from_file_id, to_file_id, from_label, to_label)
638
"""See InventoryEntry.has_text."""
641
def __init__(self, file_id, name, parent_id):
642
super(InventoryFile, self).__init__(file_id, name, parent_id)
645
def kind_character(self):
646
"""See InventoryEntry.kind_character."""
649
def _put_in_tar(self, item, tree):
650
"""See InventoryEntry._put_in_tar."""
651
item.type = tarfile.REGTYPE
652
fileobj = tree.get_file(self.file_id)
653
item.size = self.text_size
654
if tree.is_executable(self.file_id):
660
def _put_on_disk(self, fullpath, tree):
661
"""See InventoryEntry._put_on_disk."""
662
osutils.pumpfile(tree.get_file(self.file_id), file(fullpath, 'wb'))
663
if tree.is_executable(self.file_id):
664
os.chmod(fullpath, 0755)
666
def _read_tree_state(self, path, work_tree):
667
"""See InventoryEntry._read_tree_state."""
668
self.text_sha1 = work_tree.get_file_sha1(self.file_id, path=path)
669
# FIXME: 20050930 probe for the text size when getting sha1
670
# in _read_tree_state
671
self.executable = work_tree.is_executable(self.file_id, path=path)
674
return ("%s(%r, %r, parent_id=%r, sha1=%r, len=%s)"
675
% (self.__class__.__name__,
682
def _forget_tree_state(self):
683
self.text_sha1 = None
685
def _unchanged(self, previous_ie):
686
"""See InventoryEntry._unchanged."""
687
compatible = super(InventoryFile, self)._unchanged(previous_ie)
688
if self.text_sha1 != previous_ie.text_sha1:
691
# FIXME: 20050930 probe for the text size when getting sha1
692
# in _read_tree_state
693
self.text_size = previous_ie.text_size
694
if self.executable != previous_ie.executable:
699
class InventoryLink(InventoryEntry):
700
"""A file in an inventory."""
702
__slots__ = ['text_sha1', 'text_size', 'file_id', 'name', 'kind',
703
'text_id', 'parent_id', 'children', 'executable',
704
'revision', 'symlink_target', 'reference_revision']
706
def _check(self, checker, rev_id, tree):
707
"""See InventoryEntry._check"""
708
if self.text_sha1 is not None or self.text_size is not None or self.text_id is not None:
709
raise BzrCheckError('symlink {%s} has text in revision {%s}'
710
% (self.file_id, rev_id))
711
if self.symlink_target is None:
712
raise BzrCheckError('symlink {%s} has no target in revision {%s}'
713
% (self.file_id, rev_id))
716
other = InventoryLink(self.file_id, self.name, self.parent_id)
717
other.symlink_target = self.symlink_target
718
other.revision = self.revision
721
def detect_changes(self, old_entry):
722
"""See InventoryEntry.detect_changes."""
723
# FIXME: which _modified field should we use ? RBC 20051003
724
text_modified = (self.symlink_target != old_entry.symlink_target)
726
mutter(" symlink target changed")
727
meta_modified = False
728
return text_modified, meta_modified
730
def _diff(self, text_diff, from_label, tree, to_label, to_entry, to_tree,
731
output_to, reverse=False):
732
"""See InventoryEntry._diff."""
733
from bzrlib.diff import DiffSymlink
734
old_target = self.symlink_target
735
if to_entry is not None:
736
new_target = to_entry.symlink_target
745
new_target, old_target = old_target, new_target
746
differ = DiffSymlink(old_tree, new_tree, output_to)
747
return differ.diff_symlink(old_target, new_target)
749
def __init__(self, file_id, name, parent_id):
750
super(InventoryLink, self).__init__(file_id, name, parent_id)
751
self.kind = 'symlink'
753
def kind_character(self):
754
"""See InventoryEntry.kind_character."""
757
def _put_in_tar(self, item, tree):
758
"""See InventoryEntry._put_in_tar."""
759
item.type = tarfile.SYMTYPE
763
item.linkname = self.symlink_target
766
def _put_on_disk(self, fullpath, tree):
767
"""See InventoryEntry._put_on_disk."""
769
os.symlink(self.symlink_target, fullpath)
771
raise BzrError("Failed to create symlink %r -> %r, error: %s" % (fullpath, self.symlink_target, e))
773
def _read_tree_state(self, path, work_tree):
774
"""See InventoryEntry._read_tree_state."""
775
self.symlink_target = work_tree.get_symlink_target(self.file_id)
777
def _forget_tree_state(self):
778
self.symlink_target = None
780
def _unchanged(self, previous_ie):
781
"""See InventoryEntry._unchanged."""
782
compatible = super(InventoryLink, self)._unchanged(previous_ie)
783
if self.symlink_target != previous_ie.symlink_target:
788
class TreeReference(InventoryEntry):
790
kind = 'tree-reference'
792
def __init__(self, file_id, name, parent_id, revision=None,
793
reference_revision=None):
794
InventoryEntry.__init__(self, file_id, name, parent_id)
795
self.revision = revision
796
self.reference_revision = reference_revision
799
return TreeReference(self.file_id, self.name, self.parent_id,
800
self.revision, self.reference_revision)
802
def _read_tree_state(self, path, work_tree):
803
"""Populate fields in the inventory entry from the given tree.
805
self.reference_revision = work_tree.get_reference_revision(
808
def _forget_tree_state(self):
809
self.reference_revision = None
811
def _unchanged(self, previous_ie):
812
"""See InventoryEntry._unchanged."""
813
compatible = super(TreeReference, self)._unchanged(previous_ie)
814
if self.reference_revision != previous_ie.reference_revision:
819
class Inventory(object):
820
"""Inventory of versioned files in a tree.
822
This describes which file_id is present at each point in the tree,
823
and possibly the SHA-1 or other information about the file.
824
Entries can be looked up either by path or by file_id.
826
The inventory represents a typical unix file tree, with
827
directories containing files and subdirectories. We never store
828
the full path to a file, because renaming a directory implicitly
829
moves all of its contents. This class internally maintains a
830
lookup tree that allows the children under a directory to be
833
InventoryEntry objects must not be modified after they are
834
inserted, other than through the Inventory API.
836
>>> inv = Inventory()
837
>>> inv.add(InventoryFile('123-123', 'hello.c', ROOT_ID))
838
InventoryFile('123-123', 'hello.c', parent_id='TREE_ROOT', sha1=None, len=None)
839
>>> inv['123-123'].name
842
May be treated as an iterator or set to look up file ids:
844
>>> bool(inv.path2id('hello.c'))
849
May also look up by name:
851
>>> [x[0] for x in inv.iter_entries()]
853
>>> inv = Inventory('TREE_ROOT-12345678-12345678')
854
>>> inv.add(InventoryFile('123-123', 'hello.c', ROOT_ID))
855
Traceback (most recent call last):
856
BzrError: parent_id {TREE_ROOT} not in inventory
857
>>> inv.add(InventoryFile('123-123', 'hello.c', 'TREE_ROOT-12345678-12345678'))
858
InventoryFile('123-123', 'hello.c', parent_id='TREE_ROOT-12345678-12345678', sha1=None, len=None)
860
def __init__(self, root_id=ROOT_ID, revision_id=None):
861
"""Create or read an inventory.
863
If a working directory is specified, the inventory is read
864
from there. If the file is specified, read from that. If not,
865
the inventory is created empty.
867
The inventory is created with a default root directory, with
870
if root_id is not None:
871
assert root_id.__class__ == str
872
self._set_root(InventoryDirectory(root_id, u'', None))
876
self.revision_id = revision_id
879
return "<Inventory object at %x, contents=%r>" % (id(self), self._byid)
881
def apply_delta(self, delta):
882
"""Apply a delta to this inventory.
884
:param delta: A list of changes to apply. After all the changes are
885
applied the final inventory must be internally consistent, but it
886
is ok to supply changes which, if only half-applied would have an
887
invalid result - such as supplying two changes which rename two
888
files, 'A' and 'B' with each other : [('A', 'B', 'A-id', a_entry),
889
('B', 'A', 'B-id', b_entry)].
891
Each change is a tuple, of the form (old_path, new_path, file_id,
894
When new_path is None, the change indicates the removal of an entry
895
from the inventory and new_entry will be ignored (using None is
896
appropriate). If new_path is not None, then new_entry must be an
897
InventoryEntry instance, which will be incorporated into the
898
inventory (and replace any existing entry with the same file id).
900
When old_path is None, the change indicates the addition of
901
a new entry to the inventory.
903
When neither new_path nor old_path are None, the change is a
904
modification to an entry, such as a rename, reparent, kind change
907
The children attribute of new_entry is ignored. This is because
908
this method preserves children automatically across alterations to
909
the parent of the children, and cases where the parent id of a
910
child is changing require the child to be passed in as a separate
911
change regardless. E.g. in the recursive deletion of a directory -
912
the directory's children must be included in the delta, or the
913
final inventory will be invalid.
916
# Remove all affected items which were in the original inventory,
917
# starting with the longest paths, thus ensuring parents are examined
918
# after their children, which means that everything we examine has no
919
# modified children remaining by the time we examine it.
920
for old_path, file_id in sorted(((op, f) for op, np, f, e in delta
921
if op is not None), reverse=True):
922
if file_id not in self:
925
# Preserve unaltered children of file_id for later reinsertion.
926
children[file_id] = getattr(self[file_id], 'children', {})
927
# Remove file_id and the unaltered children. If file_id is not
928
# being deleted it will be reinserted back later.
929
self.remove_recursive_id(file_id)
930
# Insert all affected which should be in the new inventory, reattaching
931
# their children if they had any. This is done from shortest path to
932
# longest, ensuring that items which were modified and whose parents in
933
# the resulting inventory were also modified, are inserted after their
935
for new_path, new_entry in sorted((np, e) for op, np, f, e in
936
delta if np is not None):
937
if new_entry.kind == 'directory':
938
new_entry.children = children.get(new_entry.file_id, {})
941
def _set_root(self, ie):
943
self._byid = {self.root.file_id: self.root}
946
# TODO: jam 20051218 Should copy also copy the revision_id?
947
entries = self.iter_entries()
948
if self.root is None:
949
return Inventory(root_id=None)
950
other = Inventory(entries.next()[1].file_id)
951
# copy recursively so we know directories will be added before
952
# their children. There are more efficient ways than this...
953
for path, entry in entries:
954
other.add(entry.copy())
958
return iter(self._byid)
961
"""Returns number of entries."""
962
return len(self._byid)
964
def iter_entries(self, from_dir=None):
965
"""Return (path, entry) pairs, in order by name."""
967
if self.root is None:
971
elif isinstance(from_dir, basestring):
972
from_dir = self._byid[from_dir]
974
# unrolling the recursive called changed the time from
975
# 440ms/663ms (inline/total) to 116ms/116ms
976
children = from_dir.children.items()
978
children = collections.deque(children)
979
stack = [(u'', children)]
981
from_dir_relpath, children = stack[-1]
984
name, ie = children.popleft()
986
# we know that from_dir_relpath never ends in a slash
987
# and 'f' doesn't begin with one, we can do a string op, rather
988
# than the checks of pathjoin(), though this means that all paths
990
path = from_dir_relpath + '/' + name
994
if ie.kind != 'directory':
997
# But do this child first
998
new_children = ie.children.items()
1000
new_children = collections.deque(new_children)
1001
stack.append((path, new_children))
1002
# Break out of inner loop, so that we start outer loop with child
1005
# if we finished all children, pop it off the stack
1008
def iter_entries_by_dir(self, from_dir=None, specific_file_ids=None,
1009
yield_parents=False):
1010
"""Iterate over the entries in a directory first order.
1012
This returns all entries for a directory before returning
1013
the entries for children of a directory. This is not
1014
lexicographically sorted order, and is a hybrid between
1015
depth-first and breadth-first.
1017
:param yield_parents: If True, yield the parents from the root leading
1018
down to specific_file_ids that have been requested. This has no
1019
impact if specific_file_ids is None.
1020
:return: This yields (path, entry) pairs
1022
if specific_file_ids and not isinstance(specific_file_ids, set):
1023
specific_file_ids = set(specific_file_ids)
1024
# TODO? Perhaps this should return the from_dir so that the root is
1025
# yielded? or maybe an option?
1026
if from_dir is None:
1027
if self.root is None:
1029
# Optimize a common case
1030
if (not yield_parents and specific_file_ids is not None and
1031
len(specific_file_ids) == 1):
1032
file_id = list(specific_file_ids)[0]
1034
yield self.id2path(file_id), self[file_id]
1036
from_dir = self.root
1037
if (specific_file_ids is None or yield_parents or
1038
self.root.file_id in specific_file_ids):
1039
yield u'', self.root
1040
elif isinstance(from_dir, basestring):
1041
from_dir = self._byid[from_dir]
1043
if specific_file_ids is not None:
1044
# TODO: jam 20070302 This could really be done as a loop rather
1045
# than a bunch of recursive calls.
1048
def add_ancestors(file_id):
1049
if file_id not in byid:
1051
parent_id = byid[file_id].parent_id
1052
if parent_id is None:
1054
if parent_id not in parents:
1055
parents.add(parent_id)
1056
add_ancestors(parent_id)
1057
for file_id in specific_file_ids:
1058
add_ancestors(file_id)
1062
stack = [(u'', from_dir)]
1064
cur_relpath, cur_dir = stack.pop()
1067
for child_name, child_ie in sorted(cur_dir.children.iteritems()):
1069
child_relpath = cur_relpath + child_name
1071
if (specific_file_ids is None or
1072
child_ie.file_id in specific_file_ids or
1073
(yield_parents and child_ie.file_id in parents)):
1074
yield child_relpath, child_ie
1076
if child_ie.kind == 'directory':
1077
if parents is None or child_ie.file_id in parents:
1078
child_dirs.append((child_relpath+'/', child_ie))
1079
stack.extend(reversed(child_dirs))
1081
def make_entry(self, kind, name, parent_id, file_id=None):
1082
"""Simple thunk to bzrlib.inventory.make_entry."""
1083
return make_entry(kind, name, parent_id, file_id)
1086
"""Return list of (path, ie) for all entries except the root.
1088
This may be faster than iter_entries.
1091
def descend(dir_ie, dir_path):
1092
kids = dir_ie.children.items()
1094
for name, ie in kids:
1095
child_path = osutils.pathjoin(dir_path, name)
1096
accum.append((child_path, ie))
1097
if ie.kind == 'directory':
1098
descend(ie, child_path)
1100
descend(self.root, u'')
1103
def directories(self):
1104
"""Return (path, entry) pairs for all directories, including the root.
1107
def descend(parent_ie, parent_path):
1108
accum.append((parent_path, parent_ie))
1110
kids = [(ie.name, ie) for ie in parent_ie.children.itervalues() if ie.kind == 'directory']
1113
for name, child_ie in kids:
1114
child_path = osutils.pathjoin(parent_path, name)
1115
descend(child_ie, child_path)
1116
descend(self.root, u'')
1119
def __contains__(self, file_id):
1120
"""True if this entry contains a file with given id.
1122
>>> inv = Inventory()
1123
>>> inv.add(InventoryFile('123', 'foo.c', ROOT_ID))
1124
InventoryFile('123', 'foo.c', parent_id='TREE_ROOT', sha1=None, len=None)
1130
return (file_id in self._byid)
1132
def __getitem__(self, file_id):
1133
"""Return the entry for given file_id.
1135
>>> inv = Inventory()
1136
>>> inv.add(InventoryFile('123123', 'hello.c', ROOT_ID))
1137
InventoryFile('123123', 'hello.c', parent_id='TREE_ROOT', sha1=None, len=None)
1138
>>> inv['123123'].name
1142
return self._byid[file_id]
1144
# really we're passing an inventory, not a tree...
1145
raise errors.NoSuchId(self, file_id)
1147
def get_file_kind(self, file_id):
1148
return self._byid[file_id].kind
1150
def get_child(self, parent_id, filename):
1151
return self[parent_id].children.get(filename)
1153
def _add_child(self, entry):
1154
"""Add an entry to the inventory, without adding it to its parent"""
1155
if entry.file_id in self._byid:
1156
raise BzrError("inventory already contains entry with id {%s}" %
1158
self._byid[entry.file_id] = entry
1159
for child in getattr(entry, 'children', {}).itervalues():
1160
self._add_child(child)
1163
def add(self, entry):
1164
"""Add entry to inventory.
1166
To add a file to a branch ready to be committed, use Branch.add,
1169
Returns the new entry object.
1171
if entry.file_id in self._byid:
1172
raise errors.DuplicateFileId(entry.file_id,
1173
self._byid[entry.file_id])
1175
if entry.parent_id is None:
1176
assert self.root is None and len(self._byid) == 0
1180
parent = self._byid[entry.parent_id]
1182
raise BzrError("parent_id {%s} not in inventory" %
1185
if entry.name in parent.children:
1186
raise BzrError("%s is already versioned" %
1187
osutils.pathjoin(self.id2path(parent.file_id),
1188
entry.name).encode('utf-8'))
1189
parent.children[entry.name] = entry
1190
return self._add_child(entry)
1192
def add_path(self, relpath, kind, file_id=None, parent_id=None):
1193
"""Add entry from a path.
1195
The immediate parent must already be versioned.
1197
Returns the new entry object."""
1199
parts = osutils.splitpath(relpath)
1203
file_id = generate_ids.gen_root_id()
1204
self.root = InventoryDirectory(file_id, '', None)
1205
self._byid = {self.root.file_id: self.root}
1208
parent_path = parts[:-1]
1209
parent_id = self.path2id(parent_path)
1210
if parent_id is None:
1211
raise errors.NotVersionedError(path=parent_path)
1212
ie = make_entry(kind, parts[-1], parent_id, file_id)
1215
def __delitem__(self, file_id):
1216
"""Remove entry by id.
1218
>>> inv = Inventory()
1219
>>> inv.add(InventoryFile('123', 'foo.c', ROOT_ID))
1220
InventoryFile('123', 'foo.c', parent_id='TREE_ROOT', sha1=None, len=None)
1229
assert ie.parent_id is None or \
1230
self[ie.parent_id].children[ie.name] == ie
1232
del self._byid[file_id]
1233
if ie.parent_id is not None:
1234
del self[ie.parent_id].children[ie.name]
1236
def __eq__(self, other):
1237
"""Compare two sets by comparing their contents.
1239
>>> i1 = Inventory()
1240
>>> i2 = Inventory()
1243
>>> i1.add(InventoryFile('123', 'foo', ROOT_ID))
1244
InventoryFile('123', 'foo', parent_id='TREE_ROOT', sha1=None, len=None)
1247
>>> i2.add(InventoryFile('123', 'foo', ROOT_ID))
1248
InventoryFile('123', 'foo', parent_id='TREE_ROOT', sha1=None, len=None)
1252
if not isinstance(other, Inventory):
1253
return NotImplemented
1255
return self._byid == other._byid
1257
def __ne__(self, other):
1258
return not self.__eq__(other)
1261
raise ValueError('not hashable')
1263
def _iter_file_id_parents(self, file_id):
1264
"""Yield the parents of file_id up to the root."""
1265
while file_id is not None:
1267
ie = self._byid[file_id]
1269
raise errors.NoSuchId(tree=None, file_id=file_id)
1271
file_id = ie.parent_id
1273
def get_idpath(self, file_id):
1274
"""Return a list of file_ids for the path to an entry.
1276
The list contains one element for each directory followed by
1277
the id of the file itself. So the length of the returned list
1278
is equal to the depth of the file in the tree, counting the
1279
root directory as depth 1.
1282
for parent in self._iter_file_id_parents(file_id):
1283
p.insert(0, parent.file_id)
1286
def id2path(self, file_id):
1287
"""Return as a string the path to file_id.
1290
>>> e = i.add(InventoryDirectory('src-id', 'src', ROOT_ID))
1291
>>> e = i.add(InventoryFile('foo-id', 'foo.c', parent_id='src-id'))
1292
>>> print i.id2path('foo-id')
1295
# get all names, skipping root
1296
return '/'.join(reversed(
1297
[parent.name for parent in
1298
self._iter_file_id_parents(file_id)][:-1]))
1300
def path2id(self, name):
1301
"""Walk down through directories to return entry of last component.
1303
names may be either a list of path components, or a single
1304
string, in which case it is automatically split.
1306
This returns the entry of the last component in the path,
1307
which may be either a file or a directory.
1309
Returns None IFF the path is not found.
1311
if isinstance(name, basestring):
1312
name = osutils.splitpath(name)
1314
# mutter("lookup path %r" % name)
1321
children = getattr(parent, 'children', None)
1322
if children is None:
1325
assert cie.name == f
1326
assert cie.parent_id == parent.file_id
1329
# or raise an error?
1332
return parent.file_id
1334
def has_filename(self, names):
1335
return bool(self.path2id(names))
1337
def has_id(self, file_id):
1338
return (file_id in self._byid)
1340
def remove_recursive_id(self, file_id):
1341
"""Remove file_id, and children, from the inventory.
1343
:param file_id: A file_id to remove.
1345
to_find_delete = [self._byid[file_id]]
1347
while to_find_delete:
1348
ie = to_find_delete.pop()
1349
to_delete.append(ie.file_id)
1350
if ie.kind == 'directory':
1351
to_find_delete.extend(ie.children.values())
1352
for file_id in reversed(to_delete):
1354
del self._byid[file_id]
1355
if ie.parent_id is not None:
1356
del self[ie.parent_id].children[ie.name]
1360
def rename(self, file_id, new_parent_id, new_name):
1361
"""Move a file within the inventory.
1363
This can change either the name, or the parent, or both.
1365
This does not move the working file.
1367
new_name = ensure_normalized_name(new_name)
1368
if not is_valid_name(new_name):
1369
raise BzrError("not an acceptable filename: %r" % new_name)
1371
new_parent = self._byid[new_parent_id]
1372
if new_name in new_parent.children:
1373
raise BzrError("%r already exists in %r" % (new_name, self.id2path(new_parent_id)))
1375
new_parent_idpath = self.get_idpath(new_parent_id)
1376
if file_id in new_parent_idpath:
1377
raise BzrError("cannot move directory %r into a subdirectory of itself, %r"
1378
% (self.id2path(file_id), self.id2path(new_parent_id)))
1380
file_ie = self._byid[file_id]
1381
old_parent = self._byid[file_ie.parent_id]
1383
# TODO: Don't leave things messed up if this fails
1385
del old_parent.children[file_ie.name]
1386
new_parent.children[new_name] = file_ie
1388
file_ie.name = new_name
1389
file_ie.parent_id = new_parent_id
1391
def is_root(self, file_id):
1392
return self.root is not None and file_id == self.root.file_id
1396
'directory': InventoryDirectory,
1397
'file': InventoryFile,
1398
'symlink': InventoryLink,
1399
'tree-reference': TreeReference
1402
def make_entry(kind, name, parent_id, file_id=None):
1403
"""Create an inventory entry.
1405
:param kind: the type of inventory entry to create.
1406
:param name: the basename of the entry.
1407
:param parent_id: the parent_id of the entry.
1408
:param file_id: the file_id to use. if None, one will be created.
1411
file_id = generate_ids.gen_file_id(name)
1412
name = ensure_normalized_name(name)
1414
factory = entry_factory[kind]
1416
raise BzrError("unknown kind %r" % kind)
1417
return factory(file_id, name, parent_id)
1420
def ensure_normalized_name(name):
1423
:raises InvalidNormalization: When name is not normalized, and cannot be
1424
accessed on this platform by the normalized path.
1425
:return: The NFC normalised version of name.
1427
#------- This has been copied to bzrlib.dirstate.DirState.add, please
1428
# keep them synchronised.
1429
# we dont import normalized_filename directly because we want to be
1430
# able to change the implementation at runtime for tests.
1431
norm_name, can_access = osutils.normalized_filename(name)
1432
if norm_name != name:
1436
# TODO: jam 20060701 This would probably be more useful
1437
# if the error was raised with the full path
1438
raise errors.InvalidNormalization(name)
1444
def is_valid_name(name):
1446
if _NAME_RE is None:
1447
_NAME_RE = re.compile(r'^[^/\\]+$')
1449
return bool(_NAME_RE.match(name))
added
removed
bzrlib/inventory.py
