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 snapshot(self, revision, path, previous_entries,
436
work_tree, commit_builder):
437
"""Make a snapshot of this entry which may or may not have changed.
439
This means that all its fields are populated, that it has its
440
text stored in the text store or weave.
442
:return: True if anything was recorded
444
# cannot be unchanged unless there is only one parent file rev.
445
self._read_tree_state(path, work_tree)
446
if len(previous_entries) == 1:
447
parent_ie = previous_entries.values()[0]
448
if self._unchanged(parent_ie):
449
self.revision = parent_ie.revision
451
self.revision = revision
452
return self._snapshot_text(previous_entries, work_tree, commit_builder)
454
def _snapshot_text(self, file_parents, work_tree, commit_builder):
455
"""Record the 'text' of this entry, whatever form that takes.
457
:return: True if anything was recorded
459
raise NotImplementedError(self._snapshot_text)
461
def __eq__(self, other):
462
if not isinstance(other, InventoryEntry):
463
return NotImplemented
465
return ((self.file_id == other.file_id)
466
and (self.name == other.name)
467
and (other.symlink_target == self.symlink_target)
468
and (self.text_sha1 == other.text_sha1)
469
and (self.text_size == other.text_size)
470
and (self.text_id == other.text_id)
471
and (self.parent_id == other.parent_id)
472
and (self.kind == other.kind)
473
and (self.revision == other.revision)
474
and (self.executable == other.executable)
475
and (self.reference_revision == other.reference_revision)
478
def __ne__(self, other):
479
return not (self == other)
482
raise ValueError('not hashable')
484
def _unchanged(self, previous_ie):
485
"""Has this entry changed relative to previous_ie.
487
This method should be overridden in child classes.
490
# different inv parent
491
if previous_ie.parent_id != self.parent_id:
494
elif previous_ie.name != self.name:
496
elif previous_ie.kind != self.kind:
500
def _read_tree_state(self, path, work_tree):
501
"""Populate fields in the inventory entry from the given tree.
503
Note that this should be modified to be a noop on virtual trees
504
as all entries created there are prepopulated.
506
# TODO: Rather than running this manually, we should check the
507
# working sha1 and other expensive properties when they're
508
# first requested, or preload them if they're already known
509
pass # nothing to do by default
511
def _forget_tree_state(self):
515
class RootEntry(InventoryEntry):
517
__slots__ = ['text_sha1', 'text_size', 'file_id', 'name', 'kind',
518
'text_id', 'parent_id', 'children', 'executable',
519
'revision', 'symlink_target', 'reference_revision']
521
def _check(self, checker, rev_id, tree):
522
"""See InventoryEntry._check"""
524
def __init__(self, file_id):
525
self.file_id = file_id
527
self.kind = 'directory'
528
self.parent_id = None
531
symbol_versioning.warn('RootEntry is deprecated as of bzr 0.10.'
532
' Please use InventoryDirectory instead.',
533
DeprecationWarning, stacklevel=2)
535
def __eq__(self, other):
536
if not isinstance(other, RootEntry):
537
return NotImplemented
539
return (self.file_id == other.file_id) \
540
and (self.children == other.children)
543
class InventoryDirectory(InventoryEntry):
544
"""A directory in an inventory."""
546
__slots__ = ['text_sha1', 'text_size', 'file_id', 'name', 'kind',
547
'text_id', 'parent_id', 'children', 'executable',
548
'revision', 'symlink_target', 'reference_revision']
550
def _check(self, checker, rev_id, tree):
551
"""See InventoryEntry._check"""
552
if self.text_sha1 is not None or self.text_size is not None or self.text_id is not None:
553
raise BzrCheckError('directory {%s} has text in revision {%s}'
554
% (self.file_id, rev_id))
557
other = InventoryDirectory(self.file_id, self.name, self.parent_id)
558
other.revision = self.revision
559
# note that children are *not* copied; they're pulled across when
563
def __init__(self, file_id, name, parent_id):
564
super(InventoryDirectory, self).__init__(file_id, name, parent_id)
566
self.kind = 'directory'
568
def kind_character(self):
569
"""See InventoryEntry.kind_character."""
572
def _put_in_tar(self, item, tree):
573
"""See InventoryEntry._put_in_tar."""
574
item.type = tarfile.DIRTYPE
581
def _put_on_disk(self, fullpath, tree):
582
"""See InventoryEntry._put_on_disk."""
585
def _snapshot_text(self, file_parents, work_tree, commit_builder):
586
"""See InventoryEntry._snapshot_text."""
587
commit_builder.modified_directory(self.file_id, file_parents)
591
class InventoryFile(InventoryEntry):
592
"""A file in an inventory."""
594
__slots__ = ['text_sha1', 'text_size', 'file_id', 'name', 'kind',
595
'text_id', 'parent_id', 'children', 'executable',
596
'revision', 'symlink_target', 'reference_revision']
598
def _check(self, checker, tree_revision_id, tree):
599
"""See InventoryEntry._check"""
600
t = (self.file_id, self.revision)
601
if t in checker.checked_texts:
602
prev_sha = checker.checked_texts[t]
603
if prev_sha != self.text_sha1:
604
raise BzrCheckError('mismatched sha1 on {%s} in {%s}' %
605
(self.file_id, tree_revision_id))
607
checker.repeated_text_cnt += 1
610
if self.file_id not in checker.checked_weaves:
611
mutter('check weave {%s}', self.file_id)
612
w = tree._get_weave(self.file_id)
613
# Not passing a progress bar, because it creates a new
614
# progress, which overwrites the current progress,
615
# and doesn't look nice
617
checker.checked_weaves[self.file_id] = True
619
w = tree._get_weave(self.file_id)
621
mutter('check version {%s} of {%s}', tree_revision_id, self.file_id)
622
checker.checked_text_cnt += 1
623
# We can't check the length, because Weave doesn't store that
624
# information, and the whole point of looking at the weave's
625
# sha1sum is that we don't have to extract the text.
626
if self.text_sha1 != w.get_sha1(self.revision):
627
raise BzrCheckError('text {%s} version {%s} wrong sha1'
628
% (self.file_id, self.revision))
629
checker.checked_texts[t] = self.text_sha1
632
other = InventoryFile(self.file_id, self.name, self.parent_id)
633
other.executable = self.executable
634
other.text_id = self.text_id
635
other.text_sha1 = self.text_sha1
636
other.text_size = self.text_size
637
other.revision = self.revision
640
def detect_changes(self, old_entry):
641
"""See InventoryEntry.detect_changes."""
642
assert self.text_sha1 is not None
643
assert old_entry.text_sha1 is not None
644
text_modified = (self.text_sha1 != old_entry.text_sha1)
645
meta_modified = (self.executable != old_entry.executable)
646
return text_modified, meta_modified
648
def _diff(self, text_diff, from_label, tree, to_label, to_entry, to_tree,
649
output_to, reverse=False):
650
"""See InventoryEntry._diff."""
652
from_text = tree.get_file(self.file_id).readlines()
654
to_text = to_tree.get_file(to_entry.file_id).readlines()
658
text_diff(from_label, from_text,
659
to_label, to_text, output_to)
661
text_diff(to_label, to_text,
662
from_label, from_text, output_to)
663
except errors.BinaryFile:
665
label_pair = (to_label, from_label)
667
label_pair = (from_label, to_label)
668
print >> output_to, \
669
("Binary files %s and %s differ" % label_pair).encode('utf8')
672
"""See InventoryEntry.has_text."""
675
def __init__(self, file_id, name, parent_id):
676
super(InventoryFile, self).__init__(file_id, name, parent_id)
679
def kind_character(self):
680
"""See InventoryEntry.kind_character."""
683
def _put_in_tar(self, item, tree):
684
"""See InventoryEntry._put_in_tar."""
685
item.type = tarfile.REGTYPE
686
fileobj = tree.get_file(self.file_id)
687
item.size = self.text_size
688
if tree.is_executable(self.file_id):
694
def _put_on_disk(self, fullpath, tree):
695
"""See InventoryEntry._put_on_disk."""
696
osutils.pumpfile(tree.get_file(self.file_id), file(fullpath, 'wb'))
697
if tree.is_executable(self.file_id):
698
os.chmod(fullpath, 0755)
700
def _read_tree_state(self, path, work_tree):
701
"""See InventoryEntry._read_tree_state."""
702
self.text_sha1 = work_tree.get_file_sha1(self.file_id, path=path)
703
# FIXME: 20050930 probe for the text size when getting sha1
704
# in _read_tree_state
705
self.executable = work_tree.is_executable(self.file_id, path=path)
708
return ("%s(%r, %r, parent_id=%r, sha1=%r, len=%s)"
709
% (self.__class__.__name__,
716
def _forget_tree_state(self):
717
self.text_sha1 = None
719
def snapshot(self, revision, path, previous_entries,
720
work_tree, commit_builder):
721
"""See InventoryEntry.snapshot."""
722
# Note: We use a custom implementation of this method for files
723
# because it's a performance critical part of commit.
725
# If this is the initial commit for this file, we know the sha is
726
# coming later so skip calculating it now (in _read_tree_state())
727
if len(previous_entries) == 0:
728
self.executable = work_tree.is_executable(self.file_id, path=path)
730
self._read_tree_state(path, work_tree)
732
# If nothing is changed from the sole parent, there's nothing to do
733
if len(previous_entries) == 1:
734
parent_ie = previous_entries.values()[0]
735
if self._unchanged(parent_ie):
736
self.revision = parent_ie.revision
739
# Add the file to the repository
740
self.revision = revision
741
def get_content_byte_lines():
742
return work_tree.get_file(self.file_id, path).readlines()
743
self.text_sha1, self.text_size = commit_builder.modified_file_text(
744
self.file_id, previous_entries, get_content_byte_lines,
745
self.text_sha1, self.text_size)
748
def _unchanged(self, previous_ie):
749
"""See InventoryEntry._unchanged."""
750
compatible = super(InventoryFile, self)._unchanged(previous_ie)
751
if self.text_sha1 != previous_ie.text_sha1:
754
# FIXME: 20050930 probe for the text size when getting sha1
755
# in _read_tree_state
756
self.text_size = previous_ie.text_size
757
if self.executable != previous_ie.executable:
762
class InventoryLink(InventoryEntry):
763
"""A file in an inventory."""
765
__slots__ = ['text_sha1', 'text_size', 'file_id', 'name', 'kind',
766
'text_id', 'parent_id', 'children', 'executable',
767
'revision', 'symlink_target', 'reference_revision']
769
def _check(self, checker, rev_id, tree):
770
"""See InventoryEntry._check"""
771
if self.text_sha1 is not None or self.text_size is not None or self.text_id is not None:
772
raise BzrCheckError('symlink {%s} has text in revision {%s}'
773
% (self.file_id, rev_id))
774
if self.symlink_target is None:
775
raise BzrCheckError('symlink {%s} has no target in revision {%s}'
776
% (self.file_id, rev_id))
779
other = InventoryLink(self.file_id, self.name, self.parent_id)
780
other.symlink_target = self.symlink_target
781
other.revision = self.revision
784
def detect_changes(self, old_entry):
785
"""See InventoryEntry.detect_changes."""
786
# FIXME: which _modified field should we use ? RBC 20051003
787
text_modified = (self.symlink_target != old_entry.symlink_target)
789
mutter(" symlink target changed")
790
meta_modified = False
791
return text_modified, meta_modified
793
def _diff(self, text_diff, from_label, tree, to_label, to_entry, to_tree,
794
output_to, reverse=False):
795
"""See InventoryEntry._diff."""
796
from_text = self.symlink_target
797
if to_entry is not None:
798
to_text = to_entry.symlink_target
803
print >>output_to, '=== target changed %r => %r' % (from_text, to_text)
806
print >>output_to, '=== target was %r' % self.symlink_target
808
print >>output_to, '=== target is %r' % self.symlink_target
810
def __init__(self, file_id, name, parent_id):
811
super(InventoryLink, self).__init__(file_id, name, parent_id)
812
self.kind = 'symlink'
814
def kind_character(self):
815
"""See InventoryEntry.kind_character."""
818
def _put_in_tar(self, item, tree):
819
"""See InventoryEntry._put_in_tar."""
820
item.type = tarfile.SYMTYPE
824
item.linkname = self.symlink_target
827
def _put_on_disk(self, fullpath, tree):
828
"""See InventoryEntry._put_on_disk."""
830
os.symlink(self.symlink_target, fullpath)
832
raise BzrError("Failed to create symlink %r -> %r, error: %s" % (fullpath, self.symlink_target, e))
834
def _read_tree_state(self, path, work_tree):
835
"""See InventoryEntry._read_tree_state."""
836
self.symlink_target = work_tree.get_symlink_target(self.file_id)
838
def _forget_tree_state(self):
839
self.symlink_target = None
841
def _unchanged(self, previous_ie):
842
"""See InventoryEntry._unchanged."""
843
compatible = super(InventoryLink, self)._unchanged(previous_ie)
844
if self.symlink_target != previous_ie.symlink_target:
848
def _snapshot_text(self, file_parents, work_tree, commit_builder):
849
"""See InventoryEntry._snapshot_text."""
850
commit_builder.modified_link(
851
self.file_id, file_parents, self.symlink_target)
855
class TreeReference(InventoryEntry):
857
kind = 'tree-reference'
859
def __init__(self, file_id, name, parent_id, revision=None,
860
reference_revision=None):
861
InventoryEntry.__init__(self, file_id, name, parent_id)
862
self.revision = revision
863
self.reference_revision = reference_revision
866
return TreeReference(self.file_id, self.name, self.parent_id,
867
self.revision, self.reference_revision)
869
def _snapshot_text(self, file_parents, work_tree, commit_builder):
870
commit_builder.modified_reference(self.file_id, file_parents)
873
def _read_tree_state(self, path, work_tree):
874
"""Populate fields in the inventory entry from the given tree.
876
self.reference_revision = work_tree.get_reference_revision(
879
def _forget_tree_state(self):
880
self.reference_revision = None
882
def _unchanged(self, previous_ie):
883
"""See InventoryEntry._unchanged."""
884
compatible = super(TreeReference, self)._unchanged(previous_ie)
885
if self.reference_revision != previous_ie.reference_revision:
890
class Inventory(object):
891
"""Inventory of versioned files in a tree.
893
This describes which file_id is present at each point in the tree,
894
and possibly the SHA-1 or other information about the file.
895
Entries can be looked up either by path or by file_id.
897
The inventory represents a typical unix file tree, with
898
directories containing files and subdirectories. We never store
899
the full path to a file, because renaming a directory implicitly
900
moves all of its contents. This class internally maintains a
901
lookup tree that allows the children under a directory to be
904
InventoryEntry objects must not be modified after they are
905
inserted, other than through the Inventory API.
907
>>> inv = Inventory()
908
>>> inv.add(InventoryFile('123-123', 'hello.c', ROOT_ID))
909
InventoryFile('123-123', 'hello.c', parent_id='TREE_ROOT', sha1=None, len=None)
910
>>> inv['123-123'].name
913
May be treated as an iterator or set to look up file ids:
915
>>> bool(inv.path2id('hello.c'))
920
May also look up by name:
922
>>> [x[0] for x in inv.iter_entries()]
924
>>> inv = Inventory('TREE_ROOT-12345678-12345678')
925
>>> inv.add(InventoryFile('123-123', 'hello.c', ROOT_ID))
926
Traceback (most recent call last):
927
BzrError: parent_id {TREE_ROOT} not in inventory
928
>>> inv.add(InventoryFile('123-123', 'hello.c', 'TREE_ROOT-12345678-12345678'))
929
InventoryFile('123-123', 'hello.c', parent_id='TREE_ROOT-12345678-12345678', sha1=None, len=None)
931
def __init__(self, root_id=ROOT_ID, revision_id=None):
932
"""Create or read an inventory.
934
If a working directory is specified, the inventory is read
935
from there. If the file is specified, read from that. If not,
936
the inventory is created empty.
938
The inventory is created with a default root directory, with
941
if root_id is not None:
942
assert root_id.__class__ == str
943
self._set_root(InventoryDirectory(root_id, u'', None))
947
self.revision_id = revision_id
950
return "<Inventory object at %x, contents=%r>" % (id(self), self._byid)
952
def apply_delta(self, delta):
953
"""Apply a delta to this inventory.
955
:param delta: A list of changes to apply. After all the changes are
956
applied the final inventory must be internally consistent, but it
957
is ok to supply changes which, if only half-applied would have an
958
invalid result - such as supplying two changes which rename two
959
files, 'A' and 'B' with each other : [('A', 'B', 'A-id', a_entry),
960
('B', 'A', 'B-id', b_entry)].
962
Each change is a tuple, of the form (old_path, new_path, file_id,
965
When new_path is None, the change indicates the removal of an entry
966
from the inventory and new_entry will be ignored (using None is
967
appropriate). If new_path is not None, then new_entry must be an
968
InventoryEntry instance, which will be incorporated into the
969
inventory (and replace any existing entry with the same file id).
971
When old_path is None, the change indicates the addition of
972
a new entry to the inventory.
974
When neither new_path nor old_path are None, the change is a
975
modification to an entry, such as a rename, reparent, kind change
978
The children attribute of new_entry is ignored. This is because
979
this method preserves children automatically across alterations to
980
the parent of the children, and cases where the parent id of a
981
child is changing require the child to be passed in as a separate
982
change regardless. E.g. in the recursive deletion of a directory -
983
the directory's children must be included in the delta, or the
984
final inventory will be invalid.
987
# Remove all affected items which were in the original inventory,
988
# starting with the longest paths, thus ensuring parents are examined
989
# after their children, which means that everything we examine has no
990
# modified children remaining by the time we examine it.
991
for old_path, file_id in sorted(((op, f) for op, np, f, e in delta
992
if op is not None), reverse=True):
993
if file_id not in self:
996
# Preserve unaltered children of file_id for later reinsertion.
997
children[file_id] = getattr(self[file_id], 'children', {})
998
# Remove file_id and the unaltered children. If file_id is not
999
# being deleted it will be reinserted back later.
1000
self.remove_recursive_id(file_id)
1001
# Insert all affected which should be in the new inventory, reattaching
1002
# their children if they had any. This is done from shortest path to
1003
# longest, ensuring that items which were modified and whose parents in
1004
# the resulting inventory were also modified, are inserted after their
1006
for new_path, new_entry in sorted((np, e) for op, np, f, e in
1007
delta if np is not None):
1008
if new_entry.kind == 'directory':
1009
new_entry.children = children.get(new_entry.file_id, {})
1012
def _set_root(self, ie):
1014
self._byid = {self.root.file_id: self.root}
1017
# TODO: jam 20051218 Should copy also copy the revision_id?
1018
entries = self.iter_entries()
1019
other = Inventory(entries.next()[1].file_id)
1020
# copy recursively so we know directories will be added before
1021
# their children. There are more efficient ways than this...
1022
for path, entry in entries():
1023
other.add(entry.copy())
1027
return iter(self._byid)
1030
"""Returns number of entries."""
1031
return len(self._byid)
1033
def iter_entries(self, from_dir=None):
1034
"""Return (path, entry) pairs, in order by name."""
1035
if from_dir is None:
1036
if self.root is None:
1038
from_dir = self.root
1040
elif isinstance(from_dir, basestring):
1041
from_dir = self._byid[from_dir]
1043
# unrolling the recursive called changed the time from
1044
# 440ms/663ms (inline/total) to 116ms/116ms
1045
children = from_dir.children.items()
1047
children = collections.deque(children)
1048
stack = [(u'', children)]
1050
from_dir_relpath, children = stack[-1]
1053
name, ie = children.popleft()
1055
# we know that from_dir_relpath never ends in a slash
1056
# and 'f' doesn't begin with one, we can do a string op, rather
1057
# than the checks of pathjoin(), though this means that all paths
1058
# start with a slash
1059
path = from_dir_relpath + '/' + name
1063
if ie.kind != 'directory':
1066
# But do this child first
1067
new_children = ie.children.items()
1069
new_children = collections.deque(new_children)
1070
stack.append((path, new_children))
1071
# Break out of inner loop, so that we start outer loop with child
1074
# if we finished all children, pop it off the stack
1077
def iter_entries_by_dir(self, from_dir=None, specific_file_ids=None,
1078
yield_parents=False):
1079
"""Iterate over the entries in a directory first order.
1081
This returns all entries for a directory before returning
1082
the entries for children of a directory. This is not
1083
lexicographically sorted order, and is a hybrid between
1084
depth-first and breadth-first.
1086
:param yield_parents: If True, yield the parents from the root leading
1087
down to specific_file_ids that have been requested. This has no
1088
impact if specific_file_ids is None.
1089
:return: This yields (path, entry) pairs
1091
if specific_file_ids:
1092
safe = osutils.safe_file_id
1093
specific_file_ids = set(safe(fid) for fid in specific_file_ids)
1094
# TODO? Perhaps this should return the from_dir so that the root is
1095
# yielded? or maybe an option?
1096
if from_dir is None:
1097
if self.root is None:
1099
# Optimize a common case
1100
if (not yield_parents and specific_file_ids is not None and
1101
len(specific_file_ids) == 1):
1102
file_id = list(specific_file_ids)[0]
1104
yield self.id2path(file_id), self[file_id]
1106
from_dir = self.root
1107
if (specific_file_ids is None or yield_parents or
1108
self.root.file_id in specific_file_ids):
1109
yield u'', self.root
1110
elif isinstance(from_dir, basestring):
1111
from_dir = self._byid[from_dir]
1113
if specific_file_ids is not None:
1114
# TODO: jam 20070302 This could really be done as a loop rather
1115
# than a bunch of recursive calls.
1118
def add_ancestors(file_id):
1119
if file_id not in byid:
1121
parent_id = byid[file_id].parent_id
1122
if parent_id is None:
1124
if parent_id not in parents:
1125
parents.add(parent_id)
1126
add_ancestors(parent_id)
1127
for file_id in specific_file_ids:
1128
add_ancestors(file_id)
1132
stack = [(u'', from_dir)]
1134
cur_relpath, cur_dir = stack.pop()
1137
for child_name, child_ie in sorted(cur_dir.children.iteritems()):
1139
child_relpath = cur_relpath + child_name
1141
if (specific_file_ids is None or
1142
child_ie.file_id in specific_file_ids or
1143
(yield_parents and child_ie.file_id in parents)):
1144
yield child_relpath, child_ie
1146
if child_ie.kind == 'directory':
1147
if parents is None or child_ie.file_id in parents:
1148
child_dirs.append((child_relpath+'/', child_ie))
1149
stack.extend(reversed(child_dirs))
1151
def make_entry(self, kind, name, parent_id, file_id=None):
1152
"""Simple thunk to bzrlib.inventory.make_entry."""
1153
return make_entry(kind, name, parent_id, file_id)
1156
"""Return list of (path, ie) for all entries except the root.
1158
This may be faster than iter_entries.
1161
def descend(dir_ie, dir_path):
1162
kids = dir_ie.children.items()
1164
for name, ie in kids:
1165
child_path = osutils.pathjoin(dir_path, name)
1166
accum.append((child_path, ie))
1167
if ie.kind == 'directory':
1168
descend(ie, child_path)
1170
descend(self.root, u'')
1173
def directories(self):
1174
"""Return (path, entry) pairs for all directories, including the root.
1177
def descend(parent_ie, parent_path):
1178
accum.append((parent_path, parent_ie))
1180
kids = [(ie.name, ie) for ie in parent_ie.children.itervalues() if ie.kind == 'directory']
1183
for name, child_ie in kids:
1184
child_path = osutils.pathjoin(parent_path, name)
1185
descend(child_ie, child_path)
1186
descend(self.root, u'')
1189
def __contains__(self, file_id):
1190
"""True if this entry contains a file with given id.
1192
>>> inv = Inventory()
1193
>>> inv.add(InventoryFile('123', 'foo.c', ROOT_ID))
1194
InventoryFile('123', 'foo.c', parent_id='TREE_ROOT', sha1=None, len=None)
1200
file_id = osutils.safe_file_id(file_id)
1201
return (file_id in self._byid)
1203
def __getitem__(self, file_id):
1204
"""Return the entry for given file_id.
1206
>>> inv = Inventory()
1207
>>> inv.add(InventoryFile('123123', 'hello.c', ROOT_ID))
1208
InventoryFile('123123', 'hello.c', parent_id='TREE_ROOT', sha1=None, len=None)
1209
>>> inv['123123'].name
1212
file_id = osutils.safe_file_id(file_id)
1214
return self._byid[file_id]
1216
# really we're passing an inventory, not a tree...
1217
raise errors.NoSuchId(self, file_id)
1219
def get_file_kind(self, file_id):
1220
file_id = osutils.safe_file_id(file_id)
1221
return self._byid[file_id].kind
1223
def get_child(self, parent_id, filename):
1224
parent_id = osutils.safe_file_id(parent_id)
1225
return self[parent_id].children.get(filename)
1227
def _add_child(self, entry):
1228
"""Add an entry to the inventory, without adding it to its parent"""
1229
if entry.file_id in self._byid:
1230
raise BzrError("inventory already contains entry with id {%s}" %
1232
self._byid[entry.file_id] = entry
1233
for child in getattr(entry, 'children', {}).itervalues():
1234
self._add_child(child)
1237
def add(self, entry):
1238
"""Add entry to inventory.
1240
To add a file to a branch ready to be committed, use Branch.add,
1243
Returns the new entry object.
1245
if entry.file_id in self._byid:
1246
raise errors.DuplicateFileId(entry.file_id,
1247
self._byid[entry.file_id])
1249
if entry.parent_id is None:
1250
assert self.root is None and len(self._byid) == 0
1254
parent = self._byid[entry.parent_id]
1256
raise BzrError("parent_id {%s} not in inventory" %
1259
if entry.name in parent.children:
1260
raise BzrError("%s is already versioned" %
1261
osutils.pathjoin(self.id2path(parent.file_id),
1262
entry.name).encode('utf-8'))
1263
parent.children[entry.name] = entry
1264
return self._add_child(entry)
1266
def add_path(self, relpath, kind, file_id=None, parent_id=None):
1267
"""Add entry from a path.
1269
The immediate parent must already be versioned.
1271
Returns the new entry object."""
1273
parts = osutils.splitpath(relpath)
1277
file_id = generate_ids.gen_root_id()
1279
file_id = osutils.safe_file_id(file_id)
1280
self.root = InventoryDirectory(file_id, '', None)
1281
self._byid = {self.root.file_id: self.root}
1284
parent_path = parts[:-1]
1285
parent_id = self.path2id(parent_path)
1286
if parent_id is None:
1287
raise errors.NotVersionedError(path=parent_path)
1288
ie = make_entry(kind, parts[-1], parent_id, file_id)
1291
def __delitem__(self, file_id):
1292
"""Remove entry by id.
1294
>>> inv = Inventory()
1295
>>> inv.add(InventoryFile('123', 'foo.c', ROOT_ID))
1296
InventoryFile('123', 'foo.c', parent_id='TREE_ROOT', sha1=None, len=None)
1303
file_id = osutils.safe_file_id(file_id)
1306
assert ie.parent_id is None or \
1307
self[ie.parent_id].children[ie.name] == ie
1309
del self._byid[file_id]
1310
if ie.parent_id is not None:
1311
del self[ie.parent_id].children[ie.name]
1313
def __eq__(self, other):
1314
"""Compare two sets by comparing their contents.
1316
>>> i1 = Inventory()
1317
>>> i2 = Inventory()
1320
>>> i1.add(InventoryFile('123', 'foo', ROOT_ID))
1321
InventoryFile('123', 'foo', parent_id='TREE_ROOT', sha1=None, len=None)
1324
>>> i2.add(InventoryFile('123', 'foo', ROOT_ID))
1325
InventoryFile('123', 'foo', parent_id='TREE_ROOT', sha1=None, len=None)
1329
if not isinstance(other, Inventory):
1330
return NotImplemented
1332
return self._byid == other._byid
1334
def __ne__(self, other):
1335
return not self.__eq__(other)
1338
raise ValueError('not hashable')
1340
def _iter_file_id_parents(self, file_id):
1341
"""Yield the parents of file_id up to the root."""
1342
file_id = osutils.safe_file_id(file_id)
1343
while file_id is not None:
1345
ie = self._byid[file_id]
1347
raise errors.NoSuchId(tree=None, file_id=file_id)
1349
file_id = ie.parent_id
1351
def get_idpath(self, file_id):
1352
"""Return a list of file_ids for the path to an entry.
1354
The list contains one element for each directory followed by
1355
the id of the file itself. So the length of the returned list
1356
is equal to the depth of the file in the tree, counting the
1357
root directory as depth 1.
1359
file_id = osutils.safe_file_id(file_id)
1361
for parent in self._iter_file_id_parents(file_id):
1362
p.insert(0, parent.file_id)
1365
def id2path(self, file_id):
1366
"""Return as a string the path to file_id.
1369
>>> e = i.add(InventoryDirectory('src-id', 'src', ROOT_ID))
1370
>>> e = i.add(InventoryFile('foo-id', 'foo.c', parent_id='src-id'))
1371
>>> print i.id2path('foo-id')
1374
file_id = osutils.safe_file_id(file_id)
1375
# get all names, skipping root
1376
return '/'.join(reversed(
1377
[parent.name for parent in
1378
self._iter_file_id_parents(file_id)][:-1]))
1380
def path2id(self, name):
1381
"""Walk down through directories to return entry of last component.
1383
names may be either a list of path components, or a single
1384
string, in which case it is automatically split.
1386
This returns the entry of the last component in the path,
1387
which may be either a file or a directory.
1389
Returns None IFF the path is not found.
1391
if isinstance(name, basestring):
1392
name = osutils.splitpath(name)
1394
# mutter("lookup path %r" % name)
1401
children = getattr(parent, 'children', None)
1402
if children is None:
1405
assert cie.name == f
1406
assert cie.parent_id == parent.file_id
1409
# or raise an error?
1412
return parent.file_id
1414
def has_filename(self, names):
1415
return bool(self.path2id(names))
1417
def has_id(self, file_id):
1418
file_id = osutils.safe_file_id(file_id)
1419
return (file_id in self._byid)
1421
def remove_recursive_id(self, file_id):
1422
"""Remove file_id, and children, from the inventory.
1424
:param file_id: A file_id to remove.
1426
file_id = osutils.safe_file_id(file_id)
1427
to_find_delete = [self._byid[file_id]]
1429
while to_find_delete:
1430
ie = to_find_delete.pop()
1431
to_delete.append(ie.file_id)
1432
if ie.kind == 'directory':
1433
to_find_delete.extend(ie.children.values())
1434
for file_id in reversed(to_delete):
1436
del self._byid[file_id]
1437
if ie.parent_id is not None:
1438
del self[ie.parent_id].children[ie.name]
1442
def rename(self, file_id, new_parent_id, new_name):
1443
"""Move a file within the inventory.
1445
This can change either the name, or the parent, or both.
1447
This does not move the working file.
1449
file_id = osutils.safe_file_id(file_id)
1450
new_name = ensure_normalized_name(new_name)
1451
if not is_valid_name(new_name):
1452
raise BzrError("not an acceptable filename: %r" % new_name)
1454
new_parent = self._byid[new_parent_id]
1455
if new_name in new_parent.children:
1456
raise BzrError("%r already exists in %r" % (new_name, self.id2path(new_parent_id)))
1458
new_parent_idpath = self.get_idpath(new_parent_id)
1459
if file_id in new_parent_idpath:
1460
raise BzrError("cannot move directory %r into a subdirectory of itself, %r"
1461
% (self.id2path(file_id), self.id2path(new_parent_id)))
1463
file_ie = self._byid[file_id]
1464
old_parent = self._byid[file_ie.parent_id]
1466
# TODO: Don't leave things messed up if this fails
1468
del old_parent.children[file_ie.name]
1469
new_parent.children[new_name] = file_ie
1471
file_ie.name = new_name
1472
file_ie.parent_id = new_parent_id
1474
def is_root(self, file_id):
1475
file_id = osutils.safe_file_id(file_id)
1476
return self.root is not None and file_id == self.root.file_id
1480
'directory': InventoryDirectory,
1481
'file': InventoryFile,
1482
'symlink': InventoryLink,
1483
'tree-reference': TreeReference
1486
def make_entry(kind, name, parent_id, file_id=None):
1487
"""Create an inventory entry.
1489
:param kind: the type of inventory entry to create.
1490
:param name: the basename of the entry.
1491
:param parent_id: the parent_id of the entry.
1492
:param file_id: the file_id to use. if None, one will be created.
1495
file_id = generate_ids.gen_file_id(name)
1497
file_id = osutils.safe_file_id(file_id)
1498
name = ensure_normalized_name(name)
1500
factory = entry_factory[kind]
1502
raise BzrError("unknown kind %r" % kind)
1503
return factory(file_id, name, parent_id)
1506
def ensure_normalized_name(name):
1509
:raises InvalidNormalization: When name is not normalized, and cannot be
1510
accessed on this platform by the normalized path.
1511
:return: The NFC/NFKC normalised version of name.
1513
#------- This has been copied to bzrlib.dirstate.DirState.add, please
1514
# keep them synchronised.
1515
# we dont import normalized_filename directly because we want to be
1516
# able to change the implementation at runtime for tests.
1517
norm_name, can_access = osutils.normalized_filename(name)
1518
if norm_name != name:
1522
# TODO: jam 20060701 This would probably be more useful
1523
# if the error was raised with the full path
1524
raise errors.InvalidNormalization(name)
1530
def is_valid_name(name):
1532
if _NAME_RE is None:
1533
_NAME_RE = re.compile(r'^[^/\\]+$')
1535
return bool(_NAME_RE.match(name))
added
removed
bzrlib/inventory.py
