1
# Copyright (C) 2005, 2006 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.trace import mutter
56
class InventoryEntry(object):
57
"""Description of a versioned file.
59
An InventoryEntry has the following fields, which are also
60
present in the XML inventory-entry element:
65
(within the parent directory)
68
file_id of the parent directory, or ROOT_ID
71
the revision_id in which this variation of this file was
75
Indicates that this file should be executable on systems
79
sha-1 of the text of the file
82
size in bytes of the text of the file
84
(reading a version 4 tree created a text_id field.)
89
>>> i.add(InventoryDirectory('123', 'src', ROOT_ID))
90
InventoryDirectory('123', 'src', parent_id='TREE_ROOT', revision=None)
91
>>> i.add(InventoryFile('2323', 'hello.c', parent_id='123'))
92
InventoryFile('2323', 'hello.c', parent_id='123', sha1=None, len=None)
93
>>> shouldbe = {0: '', 1: 'src', 2: 'src/hello.c'}
94
>>> for ix, j in enumerate(i.iter_entries()):
95
... print (j[0] == shouldbe[ix], j[1])
97
(True, InventoryDirectory('TREE_ROOT', '', parent_id=None, revision=None))
98
(True, InventoryDirectory('123', 'src', parent_id='TREE_ROOT', revision=None))
99
(True, InventoryFile('2323', 'hello.c', parent_id='123', sha1=None, len=None))
100
>>> i.add(InventoryFile('2323', 'bye.c', '123'))
101
Traceback (most recent call last):
103
BzrError: inventory already contains entry with id {2323}
104
>>> i.add(InventoryFile('2324', 'bye.c', '123'))
105
InventoryFile('2324', 'bye.c', parent_id='123', sha1=None, len=None)
106
>>> i.add(InventoryDirectory('2325', 'wibble', '123'))
107
InventoryDirectory('2325', 'wibble', parent_id='123', revision=None)
108
>>> i.path2id('src/wibble')
112
>>> i.add(InventoryFile('2326', 'wibble.c', '2325'))
113
InventoryFile('2326', 'wibble.c', parent_id='2325', sha1=None, len=None)
115
InventoryFile('2326', 'wibble.c', parent_id='2325', sha1=None, len=None)
116
>>> for path, entry in i.iter_entries():
118
... assert i.path2id(path)
126
>>> i.id2path('2326')
127
'src/wibble/wibble.c'
130
# Constants returned by describe_change()
132
# TODO: These should probably move to some kind of FileChangeDescription
133
# class; that's like what's inside a TreeDelta but we want to be able to
134
# generate them just for one file at a time.
136
MODIFIED_AND_RENAMED = 'modified and renamed'
140
def detect_changes(self, old_entry):
141
"""Return a (text_modified, meta_modified) from this to old_entry.
143
_read_tree_state must have been called on self and old_entry prior to
144
calling detect_changes.
148
def diff(self, text_diff, from_label, tree, to_label, to_entry, to_tree,
149
output_to, reverse=False):
150
"""Perform a diff from this to to_entry.
152
text_diff will be used for textual difference calculation.
153
This is a template method, override _diff in child classes.
155
self._read_tree_state(tree.id2path(self.file_id), tree)
157
# cannot diff from one kind to another - you must do a removal
158
# and an addif they do not match.
159
assert self.kind == to_entry.kind
160
to_entry._read_tree_state(to_tree.id2path(to_entry.file_id),
162
self._diff(text_diff, from_label, tree, to_label, to_entry, to_tree,
165
def _diff(self, text_diff, from_label, tree, to_label, to_entry, to_tree,
166
output_to, reverse=False):
167
"""Perform a diff between two entries of the same kind."""
169
def find_previous_heads(self, previous_inventories,
170
versioned_file_store,
173
"""Return the revisions and entries that directly precede this.
175
Returned as a map from revision to inventory entry.
177
This is a map containing the file revisions in all parents
178
for which the file exists, and its revision is not a parent of
179
any other. If the file is new, the set will be empty.
181
:param versioned_file_store: A store where ancestry data on this
182
file id can be queried.
183
:param transaction: The transaction that queries to the versioned
184
file store should be completed under.
185
:param entry_vf: The entry versioned file, if its already available.
187
def get_ancestors(weave, entry):
188
return set(weave.get_ancestry(entry.revision))
189
# revision:ie mapping for each ie found in previous_inventories.
191
# revision:ie mapping with one revision for each head.
193
# revision: ancestor list for each head
195
# identify candidate head revision ids.
196
for inv in previous_inventories:
197
if self.file_id in inv:
198
ie = inv[self.file_id]
199
assert ie.file_id == self.file_id
200
if ie.revision in candidates:
201
# same revision value in two different inventories:
202
# correct possible inconsistencies:
203
# * there was a bug in revision updates with 'x' bit
206
if candidates[ie.revision].executable != ie.executable:
207
candidates[ie.revision].executable = False
208
ie.executable = False
209
except AttributeError:
211
# must now be the same.
212
assert candidates[ie.revision] == ie
214
# add this revision as a candidate.
215
candidates[ie.revision] = ie
217
# common case optimisation
218
if len(candidates) == 1:
219
# if there is only one candidate revision found
220
# then we can opening the versioned file to access ancestry:
221
# there cannot be any ancestors to eliminate when there is
222
# only one revision available.
223
heads[ie.revision] = ie
226
# eliminate ancestors amongst the available candidates:
227
# heads are those that are not an ancestor of any other candidate
228
# - this provides convergence at a per-file level.
229
for ie in candidates.values():
230
# may be an ancestor of a known head:
231
already_present = 0 != len(
232
[head for head in heads
233
if ie.revision in head_ancestors[head]])
235
# an ancestor of an analyzed candidate.
237
# not an ancestor of a known head:
238
# load the versioned file for this file id if needed
240
entry_vf = versioned_file_store.get_weave_or_empty(
241
self.file_id, transaction)
242
ancestors = get_ancestors(entry_vf, ie)
243
# may knock something else out:
244
check_heads = list(heads.keys())
245
for head in check_heads:
246
if head in ancestors:
247
# this previously discovered 'head' is not
248
# really a head - its an ancestor of the newly
251
head_ancestors[ie.revision] = ancestors
252
heads[ie.revision] = ie
255
def get_tar_item(self, root, dp, now, tree):
256
"""Get a tarfile item and a file stream for its content."""
257
item = tarfile.TarInfo(osutils.pathjoin(root, dp).encode('utf8'))
258
# TODO: would be cool to actually set it to the timestamp of the
259
# revision it was last changed
261
fileobj = self._put_in_tar(item, tree)
265
"""Return true if the object this entry represents has textual data.
267
Note that textual data includes binary content.
269
Also note that all entries get weave files created for them.
270
This attribute is primarily used when upgrading from old trees that
271
did not have the weave index for all inventory entries.
275
def __init__(self, file_id, name, parent_id, text_id=None):
276
"""Create an InventoryEntry
278
The filename must be a single component, relative to the
279
parent directory; it cannot be a whole path or relative name.
281
>>> e = InventoryFile('123', 'hello.c', ROOT_ID)
286
>>> e = InventoryFile('123', 'src/hello.c', ROOT_ID)
287
Traceback (most recent call last):
288
InvalidEntryName: Invalid entry name: src/hello.c
290
assert isinstance(name, basestring), name
291
if '/' in name or '\\' in name:
292
raise errors.InvalidEntryName(name=name)
293
self.executable = False
295
self.text_sha1 = None
296
self.text_size = None
297
self.file_id = file_id
299
self.text_id = text_id
300
self.parent_id = parent_id
301
self.symlink_target = None
303
def kind_character(self):
304
"""Return a short kind indicator useful for appending to names."""
305
raise BzrError('unknown kind %r' % self.kind)
307
known_kinds = ('file', 'directory', 'symlink')
309
def _put_in_tar(self, item, tree):
310
"""populate item for stashing in a tar, and return the content stream.
312
If no content is available, return None.
314
raise BzrError("don't know how to export {%s} of kind %r" %
315
(self.file_id, self.kind))
317
def put_on_disk(self, dest, dp, tree):
318
"""Create a representation of self on disk in the prefix dest.
320
This is a template method - implement _put_on_disk in subclasses.
322
fullpath = osutils.pathjoin(dest, dp)
323
self._put_on_disk(fullpath, tree)
324
# mutter(" export {%s} kind %s to %s", self.file_id,
325
# self.kind, fullpath)
327
def _put_on_disk(self, fullpath, tree):
328
"""Put this entry onto disk at fullpath, from tree tree."""
329
raise BzrError("don't know how to export {%s} of kind %r" % (self.file_id, self.kind))
331
def sorted_children(self):
332
return sorted(self.children.items())
335
def versionable_kind(kind):
336
return (kind in ('file', 'directory', 'symlink'))
338
def check(self, checker, rev_id, inv, tree):
339
"""Check this inventory entry is intact.
341
This is a template method, override _check for kind specific
344
:param checker: Check object providing context for the checks;
345
can be used to find out what parts of the repository have already
347
:param rev_id: Revision id from which this InventoryEntry was loaded.
348
Not necessarily the last-changed revision for this file.
349
:param inv: Inventory from which the entry was loaded.
350
:param tree: RevisionTree for this entry.
352
if self.parent_id is not None:
353
if not inv.has_id(self.parent_id):
354
raise BzrCheckError('missing parent {%s} in inventory for revision {%s}'
355
% (self.parent_id, rev_id))
356
self._check(checker, rev_id, tree)
358
def _check(self, checker, rev_id, tree):
359
"""Check this inventory entry for kind specific errors."""
360
raise BzrCheckError('unknown entry kind %r in revision {%s}' %
364
"""Clone this inventory entry."""
365
raise NotImplementedError
368
def describe_change(old_entry, new_entry):
369
"""Describe the change between old_entry and this.
371
This smells of being an InterInventoryEntry situation, but as its
372
the first one, we're making it a static method for now.
374
An entry with a different parent, or different name is considered
375
to be renamed. Reparenting is an internal detail.
376
Note that renaming the parent does not trigger a rename for the
379
# TODO: Perhaps return an object rather than just a string
380
if old_entry is new_entry:
381
# also the case of both being None
383
elif old_entry is None:
385
elif new_entry is None:
387
text_modified, meta_modified = new_entry.detect_changes(old_entry)
388
if text_modified or meta_modified:
392
# TODO 20060511 (mbp, rbc) factor out 'detect_rename' here.
393
if old_entry.parent_id != new_entry.parent_id:
395
elif old_entry.name != new_entry.name:
399
if renamed and not modified:
400
return InventoryEntry.RENAMED
401
if modified and not renamed:
403
if modified and renamed:
404
return InventoryEntry.MODIFIED_AND_RENAMED
408
return ("%s(%r, %r, parent_id=%r, revision=%r)"
409
% (self.__class__.__name__,
415
def snapshot(self, revision, path, previous_entries,
416
work_tree, commit_builder):
417
"""Make a snapshot of this entry which may or may not have changed.
419
This means that all its fields are populated, that it has its
420
text stored in the text store or weave.
422
# mutter('new parents of %s are %r', path, previous_entries)
423
self._read_tree_state(path, work_tree)
424
# TODO: Where should we determine whether to reuse a
425
# previous revision id or create a new revision? 20060606
426
if len(previous_entries) == 1:
427
# cannot be unchanged unless there is only one parent file rev.
428
parent_ie = previous_entries.values()[0]
429
if self._unchanged(parent_ie):
430
# mutter("found unchanged entry")
431
self.revision = parent_ie.revision
433
return self._snapshot_into_revision(revision, previous_entries,
434
work_tree, commit_builder)
436
def _snapshot_into_revision(self, revision, previous_entries, work_tree,
438
"""Record this revision unconditionally into a store.
440
The entry's last-changed revision property (`revision`) is updated to
441
that of the new revision.
443
:param revision: id of the new revision that is being recorded.
445
:returns: String description of the commit (e.g. "merged", "modified"), etc.
447
# mutter('new revision {%s} for {%s}', revision, self.file_id)
448
self.revision = revision
449
self._snapshot_text(previous_entries, work_tree, commit_builder)
451
def _snapshot_text(self, file_parents, work_tree, commit_builder):
452
"""Record the 'text' of this entry, whatever form that takes.
454
This default implementation simply adds an empty text.
456
raise NotImplementedError(self._snapshot_text)
458
def __eq__(self, other):
459
if not isinstance(other, InventoryEntry):
460
return NotImplemented
462
return ((self.file_id == other.file_id)
463
and (self.name == other.name)
464
and (other.symlink_target == self.symlink_target)
465
and (self.text_sha1 == other.text_sha1)
466
and (self.text_size == other.text_size)
467
and (self.text_id == other.text_id)
468
and (self.parent_id == other.parent_id)
469
and (self.kind == other.kind)
470
and (self.revision == other.revision)
471
and (self.executable == other.executable)
474
def __ne__(self, other):
475
return not (self == other)
478
raise ValueError('not hashable')
480
def _unchanged(self, previous_ie):
481
"""Has this entry changed relative to previous_ie.
483
This method should be overridden in child classes.
486
# different inv parent
487
if previous_ie.parent_id != self.parent_id:
490
elif previous_ie.name != self.name:
494
def _read_tree_state(self, path, work_tree):
495
"""Populate fields in the inventory entry from the given tree.
497
Note that this should be modified to be a noop on virtual trees
498
as all entries created there are prepopulated.
500
# TODO: Rather than running this manually, we should check the
501
# working sha1 and other expensive properties when they're
502
# first requested, or preload them if they're already known
503
pass # nothing to do by default
505
def _forget_tree_state(self):
509
class RootEntry(InventoryEntry):
511
__slots__ = ['text_sha1', 'text_size', 'file_id', 'name', 'kind',
512
'text_id', 'parent_id', 'children', 'executable',
513
'revision', 'symlink_target']
515
def _check(self, checker, rev_id, tree):
516
"""See InventoryEntry._check"""
518
def __init__(self, file_id):
519
self.file_id = file_id
521
self.kind = 'directory'
522
self.parent_id = None
525
symbol_versioning.warn('RootEntry is deprecated as of bzr 0.10.'
526
' Please use InventoryDirectory instead.',
527
DeprecationWarning, stacklevel=2)
529
def __eq__(self, other):
530
if not isinstance(other, RootEntry):
531
return NotImplemented
533
return (self.file_id == other.file_id) \
534
and (self.children == other.children)
537
class InventoryDirectory(InventoryEntry):
538
"""A directory in an inventory."""
540
__slots__ = ['text_sha1', 'text_size', 'file_id', 'name', 'kind',
541
'text_id', 'parent_id', 'children', 'executable',
542
'revision', 'symlink_target']
544
def _check(self, checker, rev_id, tree):
545
"""See InventoryEntry._check"""
546
if self.text_sha1 is not None or self.text_size is not None or self.text_id is not None:
547
raise BzrCheckError('directory {%s} has text in revision {%s}'
548
% (self.file_id, rev_id))
551
other = InventoryDirectory(self.file_id, self.name, self.parent_id)
552
other.revision = self.revision
553
# note that children are *not* copied; they're pulled across when
557
def __init__(self, file_id, name, parent_id):
558
super(InventoryDirectory, self).__init__(file_id, name, parent_id)
560
self.kind = 'directory'
562
def kind_character(self):
563
"""See InventoryEntry.kind_character."""
566
def _put_in_tar(self, item, tree):
567
"""See InventoryEntry._put_in_tar."""
568
item.type = tarfile.DIRTYPE
575
def _put_on_disk(self, fullpath, tree):
576
"""See InventoryEntry._put_on_disk."""
579
def _snapshot_text(self, file_parents, work_tree, commit_builder):
580
"""See InventoryEntry._snapshot_text."""
581
commit_builder.modified_directory(self.file_id, file_parents)
584
class InventoryFile(InventoryEntry):
585
"""A file in an inventory."""
587
__slots__ = ['text_sha1', 'text_size', 'file_id', 'name', 'kind',
588
'text_id', 'parent_id', 'children', 'executable',
589
'revision', 'symlink_target']
591
def _check(self, checker, tree_revision_id, tree):
592
"""See InventoryEntry._check"""
593
t = (self.file_id, self.revision)
594
if t in checker.checked_texts:
595
prev_sha = checker.checked_texts[t]
596
if prev_sha != self.text_sha1:
597
raise BzrCheckError('mismatched sha1 on {%s} in {%s}' %
598
(self.file_id, tree_revision_id))
600
checker.repeated_text_cnt += 1
603
if self.file_id not in checker.checked_weaves:
604
mutter('check weave {%s}', self.file_id)
605
w = tree.get_weave(self.file_id)
606
# Not passing a progress bar, because it creates a new
607
# progress, which overwrites the current progress,
608
# and doesn't look nice
610
checker.checked_weaves[self.file_id] = True
612
w = tree.get_weave(self.file_id)
614
mutter('check version {%s} of {%s}', tree_revision_id, self.file_id)
615
checker.checked_text_cnt += 1
616
# We can't check the length, because Weave doesn't store that
617
# information, and the whole point of looking at the weave's
618
# sha1sum is that we don't have to extract the text.
619
if self.text_sha1 != w.get_sha1(self.revision):
620
raise BzrCheckError('text {%s} version {%s} wrong sha1'
621
% (self.file_id, self.revision))
622
checker.checked_texts[t] = self.text_sha1
625
other = InventoryFile(self.file_id, self.name, self.parent_id)
626
other.executable = self.executable
627
other.text_id = self.text_id
628
other.text_sha1 = self.text_sha1
629
other.text_size = self.text_size
630
other.revision = self.revision
633
def detect_changes(self, old_entry):
634
"""See InventoryEntry.detect_changes."""
635
assert self.text_sha1 is not None
636
assert old_entry.text_sha1 is not None
637
text_modified = (self.text_sha1 != old_entry.text_sha1)
638
meta_modified = (self.executable != old_entry.executable)
639
return text_modified, meta_modified
641
def _diff(self, text_diff, from_label, tree, to_label, to_entry, to_tree,
642
output_to, reverse=False):
643
"""See InventoryEntry._diff."""
645
from_text = tree.get_file(self.file_id).readlines()
647
to_text = to_tree.get_file(to_entry.file_id).readlines()
651
text_diff(from_label, from_text,
652
to_label, to_text, output_to)
654
text_diff(to_label, to_text,
655
from_label, from_text, output_to)
656
except errors.BinaryFile:
658
label_pair = (to_label, from_label)
660
label_pair = (from_label, to_label)
661
print >> output_to, "Binary files %s and %s differ" % label_pair
664
"""See InventoryEntry.has_text."""
667
def __init__(self, file_id, name, parent_id):
668
super(InventoryFile, self).__init__(file_id, name, parent_id)
671
def kind_character(self):
672
"""See InventoryEntry.kind_character."""
675
def _put_in_tar(self, item, tree):
676
"""See InventoryEntry._put_in_tar."""
677
item.type = tarfile.REGTYPE
678
fileobj = tree.get_file(self.file_id)
679
item.size = self.text_size
680
if tree.is_executable(self.file_id):
686
def _put_on_disk(self, fullpath, tree):
687
"""See InventoryEntry._put_on_disk."""
688
osutils.pumpfile(tree.get_file(self.file_id), file(fullpath, 'wb'))
689
if tree.is_executable(self.file_id):
690
os.chmod(fullpath, 0755)
692
def _read_tree_state(self, path, work_tree):
693
"""See InventoryEntry._read_tree_state."""
694
self.text_sha1 = work_tree.get_file_sha1(self.file_id, path=path)
695
# FIXME: 20050930 probe for the text size when getting sha1
696
# in _read_tree_state
697
self.executable = work_tree.is_executable(self.file_id, path=path)
700
return ("%s(%r, %r, parent_id=%r, sha1=%r, len=%s)"
701
% (self.__class__.__name__,
708
def _forget_tree_state(self):
709
self.text_sha1 = None
711
def _snapshot_text(self, file_parents, work_tree, commit_builder):
712
"""See InventoryEntry._snapshot_text."""
713
def get_content_byte_lines():
714
return work_tree.get_file(self.file_id).readlines()
715
self.text_sha1, self.text_size = commit_builder.modified_file_text(
716
self.file_id, file_parents, get_content_byte_lines, self.text_sha1, self.text_size)
718
def _unchanged(self, previous_ie):
719
"""See InventoryEntry._unchanged."""
720
compatible = super(InventoryFile, self)._unchanged(previous_ie)
721
if self.text_sha1 != previous_ie.text_sha1:
724
# FIXME: 20050930 probe for the text size when getting sha1
725
# in _read_tree_state
726
self.text_size = previous_ie.text_size
727
if self.executable != previous_ie.executable:
732
class InventoryLink(InventoryEntry):
733
"""A file in an inventory."""
735
__slots__ = ['text_sha1', 'text_size', 'file_id', 'name', 'kind',
736
'text_id', 'parent_id', 'children', 'executable',
737
'revision', 'symlink_target']
739
def _check(self, checker, rev_id, tree):
740
"""See InventoryEntry._check"""
741
if self.text_sha1 is not None or self.text_size is not None or self.text_id is not None:
742
raise BzrCheckError('symlink {%s} has text in revision {%s}'
743
% (self.file_id, rev_id))
744
if self.symlink_target is None:
745
raise BzrCheckError('symlink {%s} has no target in revision {%s}'
746
% (self.file_id, rev_id))
749
other = InventoryLink(self.file_id, self.name, self.parent_id)
750
other.symlink_target = self.symlink_target
751
other.revision = self.revision
754
def detect_changes(self, old_entry):
755
"""See InventoryEntry.detect_changes."""
756
# FIXME: which _modified field should we use ? RBC 20051003
757
text_modified = (self.symlink_target != old_entry.symlink_target)
759
mutter(" symlink target changed")
760
meta_modified = False
761
return text_modified, meta_modified
763
def _diff(self, text_diff, from_label, tree, to_label, to_entry, to_tree,
764
output_to, reverse=False):
765
"""See InventoryEntry._diff."""
766
from_text = self.symlink_target
767
if to_entry is not None:
768
to_text = to_entry.symlink_target
773
print >>output_to, '=== target changed %r => %r' % (from_text, to_text)
776
print >>output_to, '=== target was %r' % self.symlink_target
778
print >>output_to, '=== target is %r' % self.symlink_target
780
def __init__(self, file_id, name, parent_id):
781
super(InventoryLink, self).__init__(file_id, name, parent_id)
782
self.kind = 'symlink'
784
def kind_character(self):
785
"""See InventoryEntry.kind_character."""
788
def _put_in_tar(self, item, tree):
789
"""See InventoryEntry._put_in_tar."""
790
item.type = tarfile.SYMTYPE
794
item.linkname = self.symlink_target
797
def _put_on_disk(self, fullpath, tree):
798
"""See InventoryEntry._put_on_disk."""
800
os.symlink(self.symlink_target, fullpath)
802
raise BzrError("Failed to create symlink %r -> %r, error: %s" % (fullpath, self.symlink_target, e))
804
def _read_tree_state(self, path, work_tree):
805
"""See InventoryEntry._read_tree_state."""
806
self.symlink_target = work_tree.get_symlink_target(self.file_id)
808
def _forget_tree_state(self):
809
self.symlink_target = None
811
def _unchanged(self, previous_ie):
812
"""See InventoryEntry._unchanged."""
813
compatible = super(InventoryLink, self)._unchanged(previous_ie)
814
if self.symlink_target != previous_ie.symlink_target:
818
def _snapshot_text(self, file_parents, work_tree, commit_builder):
819
"""See InventoryEntry._snapshot_text."""
820
commit_builder.modified_link(
821
self.file_id, file_parents, self.symlink_target)
824
class Inventory(object):
825
"""Inventory of versioned files in a tree.
827
This describes which file_id is present at each point in the tree,
828
and possibly the SHA-1 or other information about the file.
829
Entries can be looked up either by path or by file_id.
831
The inventory represents a typical unix file tree, with
832
directories containing files and subdirectories. We never store
833
the full path to a file, because renaming a directory implicitly
834
moves all of its contents. This class internally maintains a
835
lookup tree that allows the children under a directory to be
838
InventoryEntry objects must not be modified after they are
839
inserted, other than through the Inventory API.
841
>>> inv = Inventory()
842
>>> inv.add(InventoryFile('123-123', 'hello.c', ROOT_ID))
843
InventoryFile('123-123', 'hello.c', parent_id='TREE_ROOT', sha1=None, len=None)
844
>>> inv['123-123'].name
847
May be treated as an iterator or set to look up file ids:
849
>>> bool(inv.path2id('hello.c'))
854
May also look up by name:
856
>>> [x[0] for x in inv.iter_entries()]
858
>>> inv = Inventory('TREE_ROOT-12345678-12345678')
859
>>> inv.add(InventoryFile('123-123', 'hello.c', ROOT_ID))
860
Traceback (most recent call last):
861
BzrError: parent_id {TREE_ROOT} not in inventory
862
>>> inv.add(InventoryFile('123-123', 'hello.c', 'TREE_ROOT-12345678-12345678'))
863
InventoryFile('123-123', 'hello.c', parent_id='TREE_ROOT-12345678-12345678', sha1=None, len=None)
865
def __init__(self, root_id=ROOT_ID, revision_id=None):
866
"""Create or read an inventory.
868
If a working directory is specified, the inventory is read
869
from there. If the file is specified, read from that. If not,
870
the inventory is created empty.
872
The inventory is created with a default root directory, with
875
if root_id is not None:
876
self._set_root(InventoryDirectory(root_id, '', None))
880
self.revision_id = revision_id
882
def _set_root(self, ie):
884
self._byid = {self.root.file_id: self.root}
887
# TODO: jam 20051218 Should copy also copy the revision_id?
888
entries = self.iter_entries()
889
other = Inventory(entries.next()[1].file_id)
890
# copy recursively so we know directories will be added before
891
# their children. There are more efficient ways than this...
892
for path, entry in entries():
893
other.add(entry.copy())
897
return iter(self._byid)
900
"""Returns number of entries."""
901
return len(self._byid)
903
def iter_entries(self, from_dir=None):
904
"""Return (path, entry) pairs, in order by name."""
906
if self.root is None:
910
elif isinstance(from_dir, basestring):
911
from_dir = self._byid[from_dir]
913
# unrolling the recursive called changed the time from
914
# 440ms/663ms (inline/total) to 116ms/116ms
915
children = from_dir.children.items()
917
children = collections.deque(children)
918
stack = [(u'', children)]
920
from_dir_relpath, children = stack[-1]
923
name, ie = children.popleft()
925
# we know that from_dir_relpath never ends in a slash
926
# and 'f' doesn't begin with one, we can do a string op, rather
927
# than the checks of pathjoin(), though this means that all paths
929
path = from_dir_relpath + '/' + name
933
if ie.kind != 'directory':
936
# But do this child first
937
new_children = ie.children.items()
939
new_children = collections.deque(new_children)
940
stack.append((path, new_children))
941
# Break out of inner loop, so that we start outer loop with child
944
# if we finished all children, pop it off the stack
947
def iter_entries_by_dir(self, from_dir=None, specific_file_ids=None):
948
"""Iterate over the entries in a directory first order.
950
This returns all entries for a directory before returning
951
the entries for children of a directory. This is not
952
lexicographically sorted order, and is a hybrid between
953
depth-first and breadth-first.
955
:return: This yields (path, entry) pairs
957
# TODO? Perhaps this should return the from_dir so that the root is
958
# yielded? or maybe an option?
960
if self.root is None:
962
# Optimize a common case
963
if specific_file_ids is not None and len(specific_file_ids) == 1:
964
file_id = list(specific_file_ids)[0]
966
yield self.id2path(file_id), self[file_id]
969
if (specific_file_ids is None or
970
self.root.file_id in specific_file_ids):
972
elif isinstance(from_dir, basestring):
973
from_dir = self._byid[from_dir]
975
if specific_file_ids is not None:
977
def add_ancestors(file_id):
978
if file_id not in self:
980
parent_id = self[file_id].parent_id
981
if parent_id is None:
983
if parent_id not in parents:
984
parents.add(parent_id)
985
add_ancestors(parent_id)
986
for file_id in specific_file_ids:
987
add_ancestors(file_id)
991
stack = [(u'', from_dir)]
993
cur_relpath, cur_dir = stack.pop()
996
for child_name, child_ie in sorted(cur_dir.children.iteritems()):
998
child_relpath = cur_relpath + child_name
1000
if (specific_file_ids is None or
1001
child_ie.file_id in specific_file_ids):
1002
yield child_relpath, child_ie
1004
if child_ie.kind == 'directory':
1005
if parents is None or child_ie.file_id in parents:
1006
child_dirs.append((child_relpath+'/', child_ie))
1007
stack.extend(reversed(child_dirs))
1010
"""Return list of (path, ie) for all entries except the root.
1012
This may be faster than iter_entries.
1015
def descend(dir_ie, dir_path):
1016
kids = dir_ie.children.items()
1018
for name, ie in kids:
1019
child_path = osutils.pathjoin(dir_path, name)
1020
accum.append((child_path, ie))
1021
if ie.kind == 'directory':
1022
descend(ie, child_path)
1024
descend(self.root, u'')
1027
def directories(self):
1028
"""Return (path, entry) pairs for all directories, including the root.
1031
def descend(parent_ie, parent_path):
1032
accum.append((parent_path, parent_ie))
1034
kids = [(ie.name, ie) for ie in parent_ie.children.itervalues() if ie.kind == 'directory']
1037
for name, child_ie in kids:
1038
child_path = osutils.pathjoin(parent_path, name)
1039
descend(child_ie, child_path)
1040
descend(self.root, u'')
1043
def __contains__(self, file_id):
1044
"""True if this entry contains a file with given id.
1046
>>> inv = Inventory()
1047
>>> inv.add(InventoryFile('123', 'foo.c', ROOT_ID))
1048
InventoryFile('123', 'foo.c', parent_id='TREE_ROOT', sha1=None, len=None)
1054
return (file_id in self._byid)
1056
def __getitem__(self, file_id):
1057
"""Return the entry for given file_id.
1059
>>> inv = Inventory()
1060
>>> inv.add(InventoryFile('123123', 'hello.c', ROOT_ID))
1061
InventoryFile('123123', 'hello.c', parent_id='TREE_ROOT', sha1=None, len=None)
1062
>>> inv['123123'].name
1066
return self._byid[file_id]
1068
# really we're passing an inventory, not a tree...
1069
raise errors.NoSuchId(self, file_id)
1071
def get_file_kind(self, file_id):
1072
return self._byid[file_id].kind
1074
def get_child(self, parent_id, filename):
1075
return self[parent_id].children.get(filename)
1077
def add(self, entry):
1078
"""Add entry to inventory.
1080
To add a file to a branch ready to be committed, use Branch.add,
1083
Returns the new entry object.
1085
if entry.file_id in self._byid:
1086
raise BzrError("inventory already contains entry with id {%s}" % entry.file_id)
1088
if entry.parent_id is None:
1089
assert self.root is None and len(self._byid) == 0
1090
self._set_root(entry)
1093
parent = self._byid[entry.parent_id]
1095
raise BzrError("parent_id {%s} not in inventory" % entry.parent_id)
1097
if entry.name in parent.children:
1098
raise BzrError("%s is already versioned" %
1099
osutils.pathjoin(self.id2path(parent.file_id), entry.name))
1101
self._byid[entry.file_id] = entry
1102
parent.children[entry.name] = entry
1105
def add_path(self, relpath, kind, file_id=None, parent_id=None):
1106
"""Add entry from a path.
1108
The immediate parent must already be versioned.
1110
Returns the new entry object."""
1112
parts = osutils.splitpath(relpath)
1116
file_id = generate_ids.gen_root_id()
1117
self.root = InventoryDirectory(file_id, '', None)
1118
self._byid = {self.root.file_id: self.root}
1121
parent_path = parts[:-1]
1122
parent_id = self.path2id(parent_path)
1123
if parent_id is None:
1124
raise errors.NotVersionedError(path=parent_path)
1125
ie = make_entry(kind, parts[-1], parent_id, file_id)
1128
def __delitem__(self, file_id):
1129
"""Remove entry by id.
1131
>>> inv = Inventory()
1132
>>> inv.add(InventoryFile('123', 'foo.c', ROOT_ID))
1133
InventoryFile('123', 'foo.c', parent_id='TREE_ROOT', sha1=None, len=None)
1142
assert ie.parent_id is None or \
1143
self[ie.parent_id].children[ie.name] == ie
1145
del self._byid[file_id]
1146
if ie.parent_id is not None:
1147
del self[ie.parent_id].children[ie.name]
1149
def __eq__(self, other):
1150
"""Compare two sets by comparing their contents.
1152
>>> i1 = Inventory()
1153
>>> i2 = Inventory()
1156
>>> i1.add(InventoryFile('123', 'foo', ROOT_ID))
1157
InventoryFile('123', 'foo', parent_id='TREE_ROOT', sha1=None, len=None)
1160
>>> i2.add(InventoryFile('123', 'foo', ROOT_ID))
1161
InventoryFile('123', 'foo', parent_id='TREE_ROOT', sha1=None, len=None)
1165
if not isinstance(other, Inventory):
1166
return NotImplemented
1168
return self._byid == other._byid
1170
def __ne__(self, other):
1171
return not self.__eq__(other)
1174
raise ValueError('not hashable')
1176
def _iter_file_id_parents(self, file_id):
1177
"""Yield the parents of file_id up to the root."""
1178
while file_id is not None:
1180
ie = self._byid[file_id]
1182
raise BzrError("file_id {%s} not found in inventory" % file_id)
1184
file_id = ie.parent_id
1186
def get_idpath(self, file_id):
1187
"""Return a list of file_ids for the path to an entry.
1189
The list contains one element for each directory followed by
1190
the id of the file itself. So the length of the returned list
1191
is equal to the depth of the file in the tree, counting the
1192
root directory as depth 1.
1195
for parent in self._iter_file_id_parents(file_id):
1196
p.insert(0, parent.file_id)
1199
def id2path(self, file_id):
1200
"""Return as a string the path to file_id.
1203
>>> e = i.add(InventoryDirectory('src-id', 'src', ROOT_ID))
1204
>>> e = i.add(InventoryFile('foo-id', 'foo.c', parent_id='src-id'))
1205
>>> print i.id2path('foo-id')
1208
# get all names, skipping root
1209
return '/'.join(reversed(
1210
[parent.name for parent in
1211
self._iter_file_id_parents(file_id)][:-1]))
1213
def path2id(self, name):
1214
"""Walk down through directories to return entry of last component.
1216
names may be either a list of path components, or a single
1217
string, in which case it is automatically split.
1219
This returns the entry of the last component in the path,
1220
which may be either a file or a directory.
1222
Returns None IFF the path is not found.
1224
if isinstance(name, basestring):
1225
name = osutils.splitpath(name)
1227
# mutter("lookup path %r" % name)
1234
children = getattr(parent, 'children', None)
1235
if children is None:
1238
assert cie.name == f
1239
assert cie.parent_id == parent.file_id
1242
# or raise an error?
1245
return parent.file_id
1247
def has_filename(self, names):
1248
return bool(self.path2id(names))
1250
def has_id(self, file_id):
1251
return (file_id in self._byid)
1253
def remove_recursive_id(self, file_id):
1254
"""Remove file_id, and children, from the inventory.
1256
:param file_id: A file_id to remove.
1258
to_find_delete = [self._byid[file_id]]
1260
while to_find_delete:
1261
ie = to_find_delete.pop()
1262
to_delete.append(ie.file_id)
1263
if ie.kind == 'directory':
1264
to_find_delete.extend(ie.children.values())
1265
for file_id in reversed(to_delete):
1267
del self._byid[file_id]
1268
if ie.parent_id is not None:
1269
del self[ie.parent_id].children[ie.name]
1271
def rename(self, file_id, new_parent_id, new_name):
1272
"""Move a file within the inventory.
1274
This can change either the name, or the parent, or both.
1276
This does not move the working file."""
1277
if not is_valid_name(new_name):
1278
raise BzrError("not an acceptable filename: %r" % new_name)
1280
new_parent = self._byid[new_parent_id]
1281
if new_name in new_parent.children:
1282
raise BzrError("%r already exists in %r" % (new_name, self.id2path(new_parent_id)))
1284
new_parent_idpath = self.get_idpath(new_parent_id)
1285
if file_id in new_parent_idpath:
1286
raise BzrError("cannot move directory %r into a subdirectory of itself, %r"
1287
% (self.id2path(file_id), self.id2path(new_parent_id)))
1289
file_ie = self._byid[file_id]
1290
old_parent = self._byid[file_ie.parent_id]
1292
# TODO: Don't leave things messed up if this fails
1294
del old_parent.children[file_ie.name]
1295
new_parent.children[new_name] = file_ie
1297
file_ie.name = new_name
1298
file_ie.parent_id = new_parent_id
1300
def is_root(self, file_id):
1301
return self.root is not None and file_id == self.root.file_id
1304
def make_entry(kind, name, parent_id, file_id=None):
1305
"""Create an inventory entry.
1307
:param kind: the type of inventory entry to create.
1308
:param name: the basename of the entry.
1309
:param parent_id: the parent_id of the entry.
1310
:param file_id: the file_id to use. if None, one will be created.
1313
file_id = generate_ids.gen_file_id(name)
1315
norm_name, can_access = osutils.normalized_filename(name)
1316
if norm_name != name:
1320
# TODO: jam 20060701 This would probably be more useful
1321
# if the error was raised with the full path
1322
raise errors.InvalidNormalization(name)
1324
if kind == 'directory':
1325
return InventoryDirectory(file_id, name, parent_id)
1326
elif kind == 'file':
1327
return InventoryFile(file_id, name, parent_id)
1328
elif kind == 'symlink':
1329
return InventoryLink(file_id, name, parent_id)
1331
raise BzrError("unknown kind %r" % kind)
1336
def is_valid_name(name):
1338
if _NAME_RE is None:
1339
_NAME_RE = re.compile(r'^[^/\\]+$')
1341
return bool(_NAME_RE.match(name))
added
removed
bzrlib/inventory.py
