1
# Copyright (C) 2005, 2006, 2007 Canonical Ltd
2
# -*- coding: UTF-8 -*-
3
4
# This program is free software; you can redistribute it and/or modify
4
5
# it under the terms of the GNU General Public License as published by
5
6
# the Free Software Foundation; either version 2 of the License, or
6
7
# (at your option) any later version.
8
9
# This program is distributed in the hope that it will be useful,
9
10
# but WITHOUT ANY WARRANTY; without even the implied warranty of
10
11
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11
12
# GNU General Public License for more details.
13
14
# You should have received a copy of the GNU General Public License
14
15
# along with this program; if not, write to the Free Software
15
16
# 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(), """
18
"""Inventories map files to their name in a revision."""
21
__copyright__ = "Copyright (C) 2005 Canonical Ltd."
22
__author__ = "Martin Pool <mbp@canonical.com>"
24
import sys, os.path, types
28
from cElementTree import Element, ElementTree, SubElement
30
from elementtree.ElementTree import Element, ElementTree, SubElement
32
from xml import XMLMixin
33
from errors import bailout
49
from bzrlib.errors import (
36
from bzrlib.osutils import uuid, quotefn, splitpath, joinpath, appendpath
53
37
from bzrlib.trace import mutter
56
class InventoryEntry(object):
39
class InventoryEntry(XMLMixin):
57
40
"""Description of a versioned file.
59
42
An InventoryEntry has the following fields, which are also
60
43
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.)
46
* *name*: (only the basename within the directory, must not
48
* *kind*: "directory" or "file"
49
* *directory_id*: (if absent/null means the branch root directory)
50
* *text_sha1*: only for files
51
* *text_size*: in bytes, only for files
52
* *text_id*: identifier for the text version, only for files
54
InventoryEntries can also exist inside a WorkingTree
55
inventory, in which case they are not yet bound to a
56
particular revision of the file. In that case the text_sha1,
57
text_size and text_id are absent.
86
60
>>> i = Inventory()
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])
62
>>> i.add(InventoryEntry('123', 'src', kind='directory'))
63
>>> i.add(InventoryEntry('2323', 'hello.c', parent_id='123'))
64
>>> for j in i.iter_entries():
97
(True, InventoryDirectory('TREE_ROOT', u'', 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('2324', 'bye.c', '123'))
101
InventoryFile('2324', 'bye.c', parent_id='123', sha1=None, len=None)
102
>>> i.add(InventoryDirectory('2325', 'wibble', '123'))
103
InventoryDirectory('2325', 'wibble', parent_id='123', revision=None)
67
('src', InventoryEntry('123', 'src', kind='directory', parent_id=None))
68
('src/hello.c', InventoryEntry('2323', 'hello.c', kind='file', parent_id='123'))
69
>>> i.add(InventoryEntry('2323', 'bye.c', parent_id='123'))
70
Traceback (most recent call last):
72
BzrError: ('inventory already contains entry with id {2323}', [])
73
>>> i.add(InventoryEntry('2324', 'bye.c', parent_id='123'))
74
>>> i.add(InventoryEntry('2325', 'wibble', parent_id='123', kind='directory'))
104
75
>>> i.path2id('src/wibble')
108
>>> i.add(InventoryFile('2326', 'wibble.c', '2325'))
109
InventoryFile('2326', 'wibble.c', parent_id='2325', sha1=None, len=None)
79
>>> i.add(InventoryEntry('2326', 'wibble.c', parent_id='2325'))
111
InventoryFile('2326', 'wibble.c', parent_id='2325', sha1=None, len=None)
112
>>> for path, entry in i.iter_entries():
114
... assert i.path2id(path)
81
InventoryEntry('2326', 'wibble.c', kind='file', parent_id='2325')
82
>>> for j in i.iter_entries():
84
... assert i.path2id(j[0])
121
90
src/wibble/wibble.c
122
91
>>> i.id2path('2326')
123
92
'src/wibble/wibble.c'
94
:todo: Maybe also keep the full path of the entry, and the children?
95
But those depend on its position within a particular inventory, and
96
it would be nice not to need to hold the backpointer here.
126
# Constants returned by describe_change()
128
# TODO: These should probably move to some kind of FileChangeDescription
129
# class; that's like what's inside a TreeDelta but we want to be able to
130
# generate them just for one file at a time.
132
MODIFIED_AND_RENAMED = 'modified and renamed'
136
def detect_changes(self, old_entry):
137
"""Return a (text_modified, meta_modified) from this to old_entry.
139
_read_tree_state must have been called on self and old_entry prior to
140
calling detect_changes.
144
def diff(self, text_diff, from_label, tree, to_label, to_entry, to_tree,
145
output_to, reverse=False):
146
"""Perform a diff from this to to_entry.
148
text_diff will be used for textual difference calculation.
149
This is a template method, override _diff in child classes.
151
self._read_tree_state(tree.id2path(self.file_id), tree)
153
# cannot diff from one kind to another - you must do a removal
154
# and an addif they do not match.
155
assert self.kind == to_entry.kind
156
to_entry._read_tree_state(to_tree.id2path(to_entry.file_id),
158
self._diff(text_diff, from_label, tree, to_label, to_entry, to_tree,
161
def _diff(self, text_diff, from_label, tree, to_label, to_entry, to_tree,
162
output_to, reverse=False):
163
"""Perform a diff between two entries of the same kind."""
165
def find_previous_heads(self, previous_inventories,
166
versioned_file_store,
169
"""Return the revisions and entries that directly precede this.
171
Returned as a map from revision to inventory entry.
173
This is a map containing the file revisions in all parents
174
for which the file exists, and its revision is not a parent of
175
any other. If the file is new, the set will be empty.
177
:param versioned_file_store: A store where ancestry data on this
178
file id can be queried.
179
:param transaction: The transaction that queries to the versioned
180
file store should be completed under.
181
:param entry_vf: The entry versioned file, if its already available.
183
def get_ancestors(weave, entry):
184
return set(weave.get_ancestry(entry.revision, topo_sorted=False))
185
# revision:ie mapping for each ie found in previous_inventories.
187
# revision:ie mapping with one revision for each head.
189
# revision: ancestor list for each head
191
# identify candidate head revision ids.
192
for inv in previous_inventories:
193
if self.file_id in inv:
194
ie = inv[self.file_id]
195
assert ie.file_id == self.file_id
196
if ie.kind != self.kind:
197
# Can't be a candidate if the kind has changed.
199
if ie.revision in candidates:
200
# same revision value in two different inventories:
201
# correct possible inconsistencies:
202
# * there was a bug in revision updates with 'x' bit
205
if candidates[ie.revision].executable != ie.executable:
206
candidates[ie.revision].executable = False
207
ie.executable = False
208
except AttributeError:
210
# must now be the same.
211
assert candidates[ie.revision] == ie
213
# add this revision as a candidate.
214
candidates[ie.revision] = ie
216
# common case optimisation
217
if len(candidates) == 1:
218
# if there is only one candidate revision found
219
# then we can opening the versioned file to access ancestry:
220
# there cannot be any ancestors to eliminate when there is
221
# only one revision available.
222
heads[ie.revision] = ie
225
# eliminate ancestors amongst the available candidates:
226
# heads are those that are not an ancestor of any other candidate
227
# - this provides convergence at a per-file level.
228
for ie in candidates.values():
229
# may be an ancestor of a known head:
230
already_present = 0 != len(
231
[head for head in heads
232
if ie.revision in head_ancestors[head]])
234
# an ancestor of an analyzed candidate.
236
# not an ancestor of a known head:
237
# load the versioned file for this file id if needed
239
entry_vf = versioned_file_store.get_weave_or_empty(
240
self.file_id, transaction)
241
ancestors = get_ancestors(entry_vf, ie)
242
# may knock something else out:
243
check_heads = list(heads.keys())
244
for head in check_heads:
245
if head in ancestors:
246
# this previously discovered 'head' is not
247
# really a head - its an ancestor of the newly
250
head_ancestors[ie.revision] = ancestors
251
heads[ie.revision] = ie
254
def get_tar_item(self, root, dp, now, tree):
255
"""Get a tarfile item and a file stream for its content."""
256
item = tarfile.TarInfo(osutils.pathjoin(root, dp).encode('utf8'))
257
# TODO: would be cool to actually set it to the timestamp of the
258
# revision it was last changed
260
fileobj = self._put_in_tar(item, tree)
264
"""Return true if the object this entry represents has textual data.
266
Note that textual data includes binary content.
268
Also note that all entries get weave files created for them.
269
This attribute is primarily used when upgrading from old trees that
270
did not have the weave index for all inventory entries.
274
def __init__(self, file_id, name, parent_id, text_id=None):
98
def __init__(self, file_id, name, kind='file', text_id=None,
275
100
"""Create an InventoryEntry
277
102
The filename must be a single component, relative to the
278
103
parent directory; it cannot be a whole path or relative name.
280
>>> e = InventoryFile('123', 'hello.c', ROOT_ID)
105
>>> e = InventoryEntry('123', 'hello.c')
285
>>> e = InventoryFile('123', 'src/hello.c', ROOT_ID)
110
>>> e = InventoryEntry('123', 'src/hello.c')
286
111
Traceback (most recent call last):
287
InvalidEntryName: Invalid entry name: src/hello.c
112
BzrError: ("InventoryEntry name is not a simple filename: 'src/hello.c'", [])
289
assert isinstance(name, basestring), name
290
if '/' in name or '\\' in name:
291
raise errors.InvalidEntryName(name=name)
292
self.executable = False
294
self.text_sha1 = None
295
self.text_size = None
115
if len(splitpath(name)) != 1:
116
bailout('InventoryEntry name is not a simple filename: %r'
296
119
self.file_id = file_id
297
assert isinstance(file_id, (str, None.__class__)), \
298
'bad type %r for %r' % (type(file_id), file_id)
121
assert kind in ['file', 'directory']
300
123
self.text_id = text_id
301
124
self.parent_id = parent_id
302
self.symlink_target = None
303
self.reference_revision = None
305
def kind_character(self):
306
"""Return a short kind indicator useful for appending to names."""
307
raise BzrError('unknown kind %r' % self.kind)
309
known_kinds = ('file', 'directory', 'symlink')
311
def _put_in_tar(self, item, tree):
312
"""populate item for stashing in a tar, and return the content stream.
314
If no content is available, return None.
316
raise BzrError("don't know how to export {%s} of kind %r" %
317
(self.file_id, self.kind))
319
def put_on_disk(self, dest, dp, tree):
320
"""Create a representation of self on disk in the prefix dest.
322
This is a template method - implement _put_on_disk in subclasses.
324
fullpath = osutils.pathjoin(dest, dp)
325
self._put_on_disk(fullpath, tree)
326
# mutter(" export {%s} kind %s to %s", self.file_id,
327
# self.kind, fullpath)
329
def _put_on_disk(self, fullpath, tree):
330
"""Put this entry onto disk at fullpath, from tree tree."""
331
raise BzrError("don't know how to export {%s} of kind %r" % (self.file_id, self.kind))
333
def sorted_children(self):
334
return sorted(self.children.items())
337
def versionable_kind(kind):
338
return (kind in ('file', 'directory', 'symlink', 'tree-reference'))
340
def check(self, checker, rev_id, inv, tree):
341
"""Check this inventory entry is intact.
343
This is a template method, override _check for kind specific
346
:param checker: Check object providing context for the checks;
347
can be used to find out what parts of the repository have already
349
:param rev_id: Revision id from which this InventoryEntry was loaded.
350
Not necessarily the last-changed revision for this file.
351
:param inv: Inventory from which the entry was loaded.
352
:param tree: RevisionTree for this entry.
354
if self.parent_id is not None:
355
if not inv.has_id(self.parent_id):
356
raise BzrCheckError('missing parent {%s} in inventory for revision {%s}'
357
% (self.parent_id, rev_id))
358
self._check(checker, rev_id, tree)
360
def _check(self, checker, rev_id, tree):
361
"""Check this inventory entry for kind specific errors."""
362
raise BzrCheckError('unknown entry kind %r in revision {%s}' %
366
"""Clone this inventory entry."""
367
raise NotImplementedError
370
def describe_change(old_entry, new_entry):
371
"""Describe the change between old_entry and this.
373
This smells of being an InterInventoryEntry situation, but as its
374
the first one, we're making it a static method for now.
376
An entry with a different parent, or different name is considered
377
to be renamed. Reparenting is an internal detail.
378
Note that renaming the parent does not trigger a rename for the
381
# TODO: Perhaps return an object rather than just a string
382
if old_entry is new_entry:
383
# also the case of both being None
385
elif old_entry is None:
387
elif new_entry is None:
389
if old_entry.kind != new_entry.kind:
391
text_modified, meta_modified = new_entry.detect_changes(old_entry)
392
if text_modified or meta_modified:
396
# TODO 20060511 (mbp, rbc) factor out 'detect_rename' here.
397
if old_entry.parent_id != new_entry.parent_id:
399
elif old_entry.name != new_entry.name:
403
if renamed and not modified:
404
return InventoryEntry.RENAMED
405
if modified and not renamed:
407
if modified and renamed:
408
return InventoryEntry.MODIFIED_AND_RENAMED
412
return ("%s(%r, %r, parent_id=%r, revision=%r)"
413
% (self.__class__.__name__,
419
def snapshot(self, revision, path, previous_entries,
420
work_tree, commit_builder):
421
"""Make a snapshot of this entry which may or may not have changed.
423
This means that all its fields are populated, that it has its
424
text stored in the text store or weave.
426
# mutter('new parents of %s are %r', path, previous_entries)
427
self._read_tree_state(path, work_tree)
428
# TODO: Where should we determine whether to reuse a
429
# previous revision id or create a new revision? 20060606
430
if len(previous_entries) == 1:
431
# cannot be unchanged unless there is only one parent file rev.
432
parent_ie = previous_entries.values()[0]
433
if self._unchanged(parent_ie):
434
# mutter("found unchanged entry")
435
self.revision = parent_ie.revision
437
return self._snapshot_into_revision(revision, previous_entries,
438
work_tree, commit_builder)
440
def _snapshot_into_revision(self, revision, previous_entries, work_tree,
442
"""Record this revision unconditionally into a store.
444
The entry's last-changed revision property (`revision`) is updated to
445
that of the new revision.
447
:param revision: id of the new revision that is being recorded.
449
:returns: String description of the commit (e.g. "merged", "modified"), etc.
451
# mutter('new revision {%s} for {%s}', revision, self.file_id)
452
self.revision = revision
453
self._snapshot_text(previous_entries, work_tree, commit_builder)
455
def _snapshot_text(self, file_parents, work_tree, commit_builder):
456
"""Record the 'text' of this entry, whatever form that takes.
458
This default implementation simply adds an empty text.
460
raise NotImplementedError(self._snapshot_text)
462
def __eq__(self, other):
463
if not isinstance(other, InventoryEntry):
464
return NotImplemented
466
return ((self.file_id == other.file_id)
467
and (self.name == other.name)
468
and (other.symlink_target == self.symlink_target)
469
and (self.text_sha1 == other.text_sha1)
470
and (self.text_size == other.text_size)
471
and (self.text_id == other.text_id)
472
and (self.parent_id == other.parent_id)
473
and (self.kind == other.kind)
474
and (self.revision == other.revision)
475
and (self.executable == other.executable)
476
and (self.reference_revision == other.reference_revision)
479
def __ne__(self, other):
480
return not (self == other)
483
raise ValueError('not hashable')
485
def _unchanged(self, previous_ie):
486
"""Has this entry changed relative to previous_ie.
488
This method should be overridden in child classes.
491
# different inv parent
492
if previous_ie.parent_id != self.parent_id:
495
elif previous_ie.name != self.name:
497
elif previous_ie.kind != self.kind:
501
def _read_tree_state(self, path, work_tree):
502
"""Populate fields in the inventory entry from the given tree.
504
Note that this should be modified to be a noop on virtual trees
505
as all entries created there are prepopulated.
507
# TODO: Rather than running this manually, we should check the
508
# working sha1 and other expensive properties when they're
509
# first requested, or preload them if they're already known
510
pass # nothing to do by default
512
def _forget_tree_state(self):
516
class RootEntry(InventoryEntry):
518
__slots__ = ['text_sha1', 'text_size', 'file_id', 'name', 'kind',
519
'text_id', 'parent_id', 'children', 'executable',
520
'revision', 'symlink_target', 'reference_revision']
522
def _check(self, checker, rev_id, tree):
523
"""See InventoryEntry._check"""
525
def __init__(self, file_id):
526
self.file_id = file_id
528
self.kind = 'directory'
529
self.parent_id = None
532
symbol_versioning.warn('RootEntry is deprecated as of bzr 0.10.'
533
' Please use InventoryDirectory instead.',
534
DeprecationWarning, stacklevel=2)
536
def __eq__(self, other):
537
if not isinstance(other, RootEntry):
538
return NotImplemented
540
return (self.file_id == other.file_id) \
541
and (self.children == other.children)
544
class InventoryDirectory(InventoryEntry):
545
"""A directory in an inventory."""
547
__slots__ = ['text_sha1', 'text_size', 'file_id', 'name', 'kind',
548
'text_id', 'parent_id', 'children', 'executable',
549
'revision', 'symlink_target', 'reference_revision']
551
def _check(self, checker, rev_id, tree):
552
"""See InventoryEntry._check"""
553
if self.text_sha1 is not None or self.text_size is not None or self.text_id is not None:
554
raise BzrCheckError('directory {%s} has text in revision {%s}'
555
% (self.file_id, rev_id))
558
other = InventoryDirectory(self.file_id, self.name, self.parent_id)
559
other.revision = self.revision
560
# note that children are *not* copied; they're pulled across when
564
def __init__(self, file_id, name, parent_id):
565
super(InventoryDirectory, self).__init__(file_id, name, parent_id)
567
self.kind = 'directory'
569
def kind_character(self):
570
"""See InventoryEntry.kind_character."""
573
def _put_in_tar(self, item, tree):
574
"""See InventoryEntry._put_in_tar."""
575
item.type = tarfile.DIRTYPE
582
def _put_on_disk(self, fullpath, tree):
583
"""See InventoryEntry._put_on_disk."""
586
def _snapshot_text(self, file_parents, work_tree, commit_builder):
587
"""See InventoryEntry._snapshot_text."""
588
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
125
self.text_sha1 = None
126
self.text_size = None
130
other = InventoryEntry(self.file_id, self.name, self.kind,
131
self.text_id, self.parent_id)
635
132
other.text_sha1 = self.text_sha1
636
133
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)
707
137
def __repr__(self):
708
return ("%s(%r, %r, parent_id=%r, sha1=%r, len=%s)"
138
return ("%s(%r, %r, kind=%r, parent_id=%r)"
709
139
% (self.__class__.__name__,
716
def _forget_tree_state(self):
717
self.text_sha1 = None
719
def _snapshot_text(self, file_parents, work_tree, commit_builder):
720
"""See InventoryEntry._snapshot_text."""
721
def get_content_byte_lines():
722
return work_tree.get_file(self.file_id).readlines()
723
self.text_sha1, self.text_size = commit_builder.modified_file_text(
724
self.file_id, file_parents, get_content_byte_lines, self.text_sha1, self.text_size)
726
def _unchanged(self, previous_ie):
727
"""See InventoryEntry._unchanged."""
728
compatible = super(InventoryFile, self)._unchanged(previous_ie)
729
if self.text_sha1 != previous_ie.text_sha1:
732
# FIXME: 20050930 probe for the text size when getting sha1
733
# in _read_tree_state
734
self.text_size = previous_ie.text_size
735
if self.executable != previous_ie.executable:
740
class InventoryLink(InventoryEntry):
741
"""A file in an inventory."""
743
__slots__ = ['text_sha1', 'text_size', 'file_id', 'name', 'kind',
744
'text_id', 'parent_id', 'children', 'executable',
745
'revision', 'symlink_target', 'reference_revision']
747
def _check(self, checker, rev_id, tree):
748
"""See InventoryEntry._check"""
749
if self.text_sha1 is not None or self.text_size is not None or self.text_id is not None:
750
raise BzrCheckError('symlink {%s} has text in revision {%s}'
751
% (self.file_id, rev_id))
752
if self.symlink_target is None:
753
raise BzrCheckError('symlink {%s} has no target in revision {%s}'
754
% (self.file_id, rev_id))
757
other = InventoryLink(self.file_id, self.name, self.parent_id)
758
other.symlink_target = self.symlink_target
759
other.revision = self.revision
762
def detect_changes(self, old_entry):
763
"""See InventoryEntry.detect_changes."""
764
# FIXME: which _modified field should we use ? RBC 20051003
765
text_modified = (self.symlink_target != old_entry.symlink_target)
767
mutter(" symlink target changed")
768
meta_modified = False
769
return text_modified, meta_modified
771
def _diff(self, text_diff, from_label, tree, to_label, to_entry, to_tree,
772
output_to, reverse=False):
773
"""See InventoryEntry._diff."""
774
from_text = self.symlink_target
775
if to_entry is not None:
776
to_text = to_entry.symlink_target
781
print >>output_to, '=== target changed %r => %r' % (from_text, to_text)
784
print >>output_to, '=== target was %r' % self.symlink_target
786
print >>output_to, '=== target is %r' % self.symlink_target
788
def __init__(self, file_id, name, parent_id):
789
super(InventoryLink, self).__init__(file_id, name, parent_id)
790
self.kind = 'symlink'
792
def kind_character(self):
793
"""See InventoryEntry.kind_character."""
796
def _put_in_tar(self, item, tree):
797
"""See InventoryEntry._put_in_tar."""
798
item.type = tarfile.SYMTYPE
802
item.linkname = self.symlink_target
805
def _put_on_disk(self, fullpath, tree):
806
"""See InventoryEntry._put_on_disk."""
808
os.symlink(self.symlink_target, fullpath)
810
raise BzrError("Failed to create symlink %r -> %r, error: %s" % (fullpath, self.symlink_target, e))
812
def _read_tree_state(self, path, work_tree):
813
"""See InventoryEntry._read_tree_state."""
814
self.symlink_target = work_tree.get_symlink_target(self.file_id)
816
def _forget_tree_state(self):
817
self.symlink_target = None
819
def _unchanged(self, previous_ie):
820
"""See InventoryEntry._unchanged."""
821
compatible = super(InventoryLink, self)._unchanged(previous_ie)
822
if self.symlink_target != previous_ie.symlink_target:
826
def _snapshot_text(self, file_parents, work_tree, commit_builder):
827
"""See InventoryEntry._snapshot_text."""
828
commit_builder.modified_link(
829
self.file_id, file_parents, self.symlink_target)
832
class TreeReference(InventoryEntry):
834
kind = 'tree-reference'
836
def __init__(self, file_id, name, parent_id, revision=None,
837
reference_revision=None):
838
InventoryEntry.__init__(self, file_id, name, parent_id)
839
self.revision = revision
840
self.reference_revision = reference_revision
843
return TreeReference(self.file_id, self.name, self.parent_id,
844
self.revision, self.reference_revision)
846
def _snapshot_text(self, file_parents, work_tree, commit_builder):
847
commit_builder.modified_reference(self.file_id, file_parents)
849
def _read_tree_state(self, path, work_tree):
850
"""Populate fields in the inventory entry from the given tree.
852
self.reference_revision = work_tree.get_reference_revision(
855
def _forget_tree_state(self):
856
self.reference_revision = None
859
class Inventory(object):
146
def to_element(self):
147
"""Convert to XML element"""
150
e.set('name', self.name)
151
e.set('file_id', self.file_id)
152
e.set('kind', self.kind)
154
if self.text_size is not None:
155
e.set('text_size', '%d' % self.text_size)
157
for f in ['text_id', 'text_sha1', 'parent_id']:
167
def from_element(cls, elt):
168
assert elt.tag == 'entry'
169
self = cls(elt.get('file_id'), elt.get('name'), elt.get('kind'))
170
self.text_id = elt.get('text_id')
171
self.text_sha1 = elt.get('text_sha1')
172
self.parent_id = elt.get('parent_id')
174
## mutter("read inventoryentry: %r" % (elt.attrib))
176
v = elt.get('text_size')
177
self.text_size = v and int(v)
182
from_element = classmethod(from_element)
184
def __cmp__(self, other):
187
if not isinstance(other, InventoryEntry):
188
return NotImplemented
190
return cmp(self.file_id, other.file_id) \
191
or cmp(self.name, other.name) \
192
or cmp(self.text_sha1, other.text_sha1) \
193
or cmp(self.text_size, other.text_size) \
194
or cmp(self.text_id, other.text_id) \
195
or cmp(self.parent_id, other.parent_id) \
196
or cmp(self.kind, other.kind)
200
class Inventory(XMLMixin):
860
201
"""Inventory of versioned files in a tree.
862
This describes which file_id is present at each point in the tree,
863
and possibly the SHA-1 or other information about the file.
864
Entries can be looked up either by path or by file_id.
203
An Inventory acts like a set of InventoryEntry items. You can
204
also look files up by their file_id or name.
206
May be read from and written to a metadata file in a tree. To
207
manipulate the inventory (for example to add a file), it is read
208
in, modified, and then written back out.
866
210
The inventory represents a typical unix file tree, with
867
211
directories containing files and subdirectories. We never store
889
238
May also look up by name:
891
240
>>> [x[0] for x in inv.iter_entries()]
893
>>> inv = Inventory('TREE_ROOT-12345678-12345678')
894
>>> inv.add(InventoryFile('123-123', 'hello.c', ROOT_ID))
895
Traceback (most recent call last):
896
BzrError: parent_id {TREE_ROOT} not in inventory
897
>>> inv.add(InventoryFile('123-123', 'hello.c', 'TREE_ROOT-12345678-12345678'))
898
InventoryFile('123-123', 'hello.c', parent_id='TREE_ROOT-12345678-12345678', sha1=None, len=None)
243
>>> inv.write_xml(sys.stdout)
245
<entry file_id="123-123" kind="file" name="hello.c" />
900
def __init__(self, root_id=ROOT_ID, revision_id=None):
250
## TODO: Clear up handling of files in subdirectories; we probably
251
## do want to be able to just look them up by name but this
252
## probably means gradually walking down the path, looking up as we go.
254
## TODO: Make sure only canonical filenames are stored.
256
## TODO: Do something sensible about the possible collisions on
257
## case-losing filesystems. Perhaps we should just always forbid
260
## _tree should probably just be stored as
261
## InventoryEntry._children on each directory.
901
264
"""Create or read an inventory.
903
266
If a working directory is specified, the inventory is read
904
267
from there. If the file is specified, read from that. If not,
905
268
the inventory is created empty.
907
The inventory is created with a default root directory, with
910
if root_id is not None:
911
assert root_id.__class__ == str
912
self._set_root(InventoryDirectory(root_id, u'', None))
916
self.revision_id = revision_id
918
def _set_root(self, ie):
920
self._byid = {self.root.file_id: self.root}
923
# TODO: jam 20051218 Should copy also copy the revision_id?
924
entries = self.iter_entries()
925
other = Inventory(entries.next()[1].file_id)
926
# copy recursively so we know directories will be added before
927
# their children. There are more efficient ways than this...
928
for path, entry in entries():
929
other.add(entry.copy())
272
# _tree is indexed by parent_id; at each level a map from name
273
# to ie. The None entry is the root.
274
self._tree = {None: {}}
932
277
def __iter__(self):
933
278
return iter(self._byid)
935
281
def __len__(self):
936
282
"""Returns number of entries."""
937
283
return len(self._byid)
939
def iter_entries(self, from_dir=None):
286
def iter_entries(self, parent_id=None):
940
287
"""Return (path, entry) pairs, in order by name."""
942
if self.root is None:
946
elif isinstance(from_dir, basestring):
947
from_dir = self._byid[from_dir]
949
# unrolling the recursive called changed the time from
950
# 440ms/663ms (inline/total) to 116ms/116ms
951
children = from_dir.children.items()
953
children = collections.deque(children)
954
stack = [(u'', children)]
956
from_dir_relpath, children = stack[-1]
959
name, ie = children.popleft()
961
# we know that from_dir_relpath never ends in a slash
962
# and 'f' doesn't begin with one, we can do a string op, rather
963
# than the checks of pathjoin(), though this means that all paths
965
path = from_dir_relpath + '/' + name
969
if ie.kind != 'directory':
972
# But do this child first
973
new_children = ie.children.items()
975
new_children = collections.deque(new_children)
976
stack.append((path, new_children))
977
# Break out of inner loop, so that we start outer loop with child
980
# if we finished all children, pop it off the stack
983
def iter_entries_by_dir(self, from_dir=None, specific_file_ids=None):
984
"""Iterate over the entries in a directory first order.
986
This returns all entries for a directory before returning
987
the entries for children of a directory. This is not
988
lexicographically sorted order, and is a hybrid between
989
depth-first and breadth-first.
991
:return: This yields (path, entry) pairs
993
if specific_file_ids:
994
safe = osutils.safe_file_id
995
specific_file_ids = set(safe(fid) for fid in specific_file_ids)
996
# TODO? Perhaps this should return the from_dir so that the root is
997
# yielded? or maybe an option?
999
if self.root is None:
1001
# Optimize a common case
1002
if specific_file_ids is not None and len(specific_file_ids) == 1:
1003
file_id = list(specific_file_ids)[0]
1005
yield self.id2path(file_id), self[file_id]
1007
from_dir = self.root
1008
if (specific_file_ids is None or
1009
self.root.file_id in specific_file_ids):
1010
yield u'', self.root
1011
elif isinstance(from_dir, basestring):
1012
from_dir = self._byid[from_dir]
1014
if specific_file_ids is not None:
1015
# TODO: jam 20070302 This could really be done as a loop rather
1016
# than a bunch of recursive calls.
1019
def add_ancestors(file_id):
1020
if file_id not in byid:
1022
parent_id = byid[file_id].parent_id
1023
if parent_id is None:
1025
if parent_id not in parents:
1026
parents.add(parent_id)
1027
add_ancestors(parent_id)
1028
for file_id in specific_file_ids:
1029
add_ancestors(file_id)
1033
stack = [(u'', from_dir)]
1035
cur_relpath, cur_dir = stack.pop()
1038
for child_name, child_ie in sorted(cur_dir.children.iteritems()):
1040
child_relpath = cur_relpath + child_name
1042
if (specific_file_ids is None or
1043
child_ie.file_id in specific_file_ids):
1044
yield child_relpath, child_ie
1046
if child_ie.kind == 'directory':
1047
if parents is None or child_ie.file_id in parents:
1048
child_dirs.append((child_relpath+'/', child_ie))
1049
stack.extend(reversed(child_dirs))
1051
def make_entry(self, kind, name, parent_id, file_id=None):
1052
"""Simple thunk to bzrlib.inventory.make_entry."""
1053
return make_entry(kind, name, parent_id, file_id)
1056
"""Return list of (path, ie) for all entries except the root.
1058
This may be faster than iter_entries.
1061
def descend(dir_ie, dir_path):
1062
kids = dir_ie.children.items()
1064
for name, ie in kids:
1065
child_path = osutils.pathjoin(dir_path, name)
1066
accum.append((child_path, ie))
1067
if ie.kind == 'directory':
1068
descend(ie, child_path)
1070
descend(self.root, u'')
1073
def directories(self):
1074
"""Return (path, entry) pairs for all directories, including the root.
1077
def descend(parent_ie, parent_path):
1078
accum.append((parent_path, parent_ie))
1080
kids = [(ie.name, ie) for ie in parent_ie.children.itervalues() if ie.kind == 'directory']
1083
for name, child_ie in kids:
1084
child_path = osutils.pathjoin(parent_path, name)
1085
descend(child_ie, child_path)
1086
descend(self.root, u'')
288
kids = self._tree[parent_id].items()
290
for name, ie in kids:
292
if ie.kind == 'directory':
293
for cn, cie in self.iter_entries(parent_id=ie.file_id):
294
yield joinpath([name, cn]), cie
297
def directories(self, include_root=True):
298
"""Return (path, entry) pairs for all directories.
302
for path, entry in self.iter_entries():
303
if entry.kind == 'directory':
308
def children(self, parent_id):
309
"""Return entries that are direct children of parent_id."""
310
return self._tree[parent_id]
314
# TODO: return all paths and entries
1089
317
def __contains__(self, file_id):
1090
318
"""True if this entry contains a file with given id.
1092
320
>>> inv = Inventory()
1093
>>> inv.add(InventoryFile('123', 'foo.c', ROOT_ID))
1094
InventoryFile('123', 'foo.c', parent_id='TREE_ROOT', sha1=None, len=None)
321
>>> inv.add(InventoryEntry('123', 'foo.c'))
1095
322
>>> '123' in inv
1097
324
>>> '456' in inv
1100
file_id = osutils.safe_file_id(file_id)
1101
return (file_id in self._byid)
327
return file_id in self._byid
1103
330
def __getitem__(self, file_id):
1104
331
"""Return the entry for given file_id.
1106
333
>>> inv = Inventory()
1107
>>> inv.add(InventoryFile('123123', 'hello.c', ROOT_ID))
1108
InventoryFile('123123', 'hello.c', parent_id='TREE_ROOT', sha1=None, len=None)
334
>>> inv.add(InventoryEntry('123123', 'hello.c'))
1109
335
>>> inv['123123'].name
1112
file_id = osutils.safe_file_id(file_id)
1114
return self._byid[file_id]
1116
# really we're passing an inventory, not a tree...
1117
raise errors.NoSuchId(self, file_id)
1119
def get_file_kind(self, file_id):
1120
file_id = osutils.safe_file_id(file_id)
1121
return self._byid[file_id].kind
1123
def get_child(self, parent_id, filename):
1124
parent_id = osutils.safe_file_id(parent_id)
1125
return self[parent_id].children.get(filename)
1127
def _add_child(self, entry):
1128
"""Add an entry to the inventory, without adding it to its parent"""
1129
if entry.file_id in self._byid:
1130
raise BzrError("inventory already contains entry with id {%s}" %
1132
self._byid[entry.file_id] = entry
1133
for child in getattr(entry, 'children', {}).itervalues():
1134
self._add_child(child)
338
return self._byid[file_id]
1137
341
def add(self, entry):
1138
342
"""Add entry to inventory.
1140
344
To add a file to a branch ready to be committed, use Branch.add,
1143
Returns the new entry object.
1145
if entry.file_id in self._byid:
1146
raise errors.DuplicateFileId(entry.file_id,
1147
self._byid[entry.file_id])
1149
if entry.parent_id is None:
1150
assert self.root is None and len(self._byid) == 0
1154
parent = self._byid[entry.parent_id]
1156
raise BzrError("parent_id {%s} not in inventory" %
1159
if entry.name in parent.children:
1160
raise BzrError("%s is already versioned" %
1161
osutils.pathjoin(self.id2path(parent.file_id),
1162
entry.name).encode('utf-8'))
1163
parent.children[entry.name] = entry
1164
return self._add_child(entry)
1166
def add_path(self, relpath, kind, file_id=None, parent_id=None):
346
if entry.file_id in self:
347
bailout("inventory already contains entry with id {%s}" % entry.file_id)
349
if entry.parent_id != None:
350
if entry.parent_id not in self:
351
bailout("parent_id %s of new entry not found in inventory"
354
if self._tree[entry.parent_id].has_key(entry.name):
355
bailout("%s is already versioned"
356
% appendpath(self.id2path(entry.parent_id), entry.name))
358
self._byid[entry.file_id] = entry
359
self._tree[entry.parent_id][entry.name] = entry
361
if entry.kind == 'directory':
362
self._tree[entry.file_id] = {}
365
def add_path(self, relpath, kind, file_id=None):
1167
366
"""Add entry from a path.
1169
The immediate parent must already be versioned.
1171
Returns the new entry object."""
1173
parts = osutils.splitpath(relpath)
368
The immediate parent must already be versioned"""
369
parts = bzrlib.osutils.splitpath(relpath)
1175
370
if len(parts) == 0:
1177
file_id = generate_ids.gen_root_id()
1179
file_id = osutils.safe_file_id(file_id)
1180
self.root = InventoryDirectory(file_id, '', None)
1181
self._byid = {self.root.file_id: self.root}
1184
parent_path = parts[:-1]
1185
parent_id = self.path2id(parent_path)
1186
if parent_id is None:
1187
raise errors.NotVersionedError(path=parent_path)
1188
ie = make_entry(kind, parts[-1], parent_id, file_id)
371
bailout("cannot re-add root of inventory")
374
file_id = bzrlib.branch.gen_file_id(relpath)
376
parent_id = self.path2id(parts[:-1])
377
ie = InventoryEntry(file_id, parts[-1],
378
kind=kind, parent_id=parent_id)
1189
379
return self.add(ie)
1191
382
def __delitem__(self, file_id):
1192
383
"""Remove entry by id.
1194
385
>>> inv = Inventory()
1195
>>> inv.add(InventoryFile('123', 'foo.c', ROOT_ID))
1196
InventoryFile('123', 'foo.c', parent_id='TREE_ROOT', sha1=None, len=None)
386
>>> inv.add(InventoryEntry('123', 'foo.c'))
1197
387
>>> '123' in inv
1199
389
>>> del inv['123']
1200
390
>>> '123' in inv
1203
file_id = osutils.safe_file_id(file_id)
1204
393
ie = self[file_id]
1206
assert ie.parent_id is None or \
1207
self[ie.parent_id].children[ie.name] == ie
395
assert self._tree[ie.parent_id][ie.name] == ie
397
# TODO: Test deleting all children; maybe hoist to a separate
399
if ie.kind == 'directory':
400
for cie in self._tree[file_id].values():
401
del self[cie.file_id]
402
del self._tree[file_id]
1209
404
del self._byid[file_id]
1210
if ie.parent_id is not None:
1211
del self[ie.parent_id].children[ie.name]
1213
def __eq__(self, other):
405
del self._tree[ie.parent_id][ie.name]
409
return Set(self._byid)
412
def to_element(self):
413
"""Convert to XML Element"""
414
e = Element('inventory')
416
for path, ie in self.iter_entries():
417
e.append(ie.to_element())
421
def from_element(cls, elt):
422
"""Construct from XML Element
424
>>> inv = Inventory()
425
>>> inv.add(InventoryEntry('foo.c-123981239', 'foo.c'))
426
>>> elt = inv.to_element()
427
>>> inv2 = Inventory.from_element(elt)
431
assert elt.tag == 'inventory'
434
o.add(InventoryEntry.from_element(e))
437
from_element = classmethod(from_element)
440
def __cmp__(self, other):
1214
441
"""Compare two sets by comparing their contents.
1216
443
>>> i1 = Inventory()
1217
444
>>> i2 = Inventory()
1220
>>> i1.add(InventoryFile('123', 'foo', ROOT_ID))
1221
InventoryFile('123', 'foo', parent_id='TREE_ROOT', sha1=None, len=None)
447
>>> i1.add(InventoryEntry('123', 'foo'))
1224
>>> i2.add(InventoryFile('123', 'foo', ROOT_ID))
1225
InventoryFile('123', 'foo', parent_id='TREE_ROOT', sha1=None, len=None)
450
>>> i2.add(InventoryEntry('123', 'foo'))
1229
457
if not isinstance(other, Inventory):
1230
458
return NotImplemented
1232
return self._byid == other._byid
1234
def __ne__(self, other):
1235
return not self.__eq__(other)
1238
raise ValueError('not hashable')
1240
def _iter_file_id_parents(self, file_id):
1241
"""Yield the parents of file_id up to the root."""
1242
file_id = osutils.safe_file_id(file_id)
1243
while file_id is not None:
1245
ie = self._byid[file_id]
1247
raise errors.NoSuchId(tree=None, file_id=file_id)
460
if self.id_set() ^ other.id_set():
463
for file_id in self._byid:
464
c = cmp(self[file_id], other[file_id])
470
def id2path(self, file_id):
471
"""Return as a list the path to file_id."""
473
while file_id != None:
1249
476
file_id = ie.parent_id
1251
def get_idpath(self, file_id):
1252
"""Return a list of file_ids for the path to an entry.
1254
The list contains one element for each directory followed by
1255
the id of the file itself. So the length of the returned list
1256
is equal to the depth of the file in the tree, counting the
1257
root directory as depth 1.
1259
file_id = osutils.safe_file_id(file_id)
1261
for parent in self._iter_file_id_parents(file_id):
1262
p.insert(0, parent.file_id)
1265
def id2path(self, file_id):
1266
"""Return as a string the path to file_id.
1269
>>> e = i.add(InventoryDirectory('src-id', 'src', ROOT_ID))
1270
>>> e = i.add(InventoryFile('foo-id', 'foo.c', parent_id='src-id'))
1271
>>> print i.id2path('foo-id')
1274
file_id = osutils.safe_file_id(file_id)
1275
# get all names, skipping root
1276
return '/'.join(reversed(
1277
[parent.name for parent in
1278
self._iter_file_id_parents(file_id)][:-1]))
1280
481
def path2id(self, name):
1281
482
"""Walk down through directories to return entry of last component.
added
removed
bzrlib/inventory.py
