1
# (C) 2005 Canonical Ltd
1
# Copyright (C) 2005, 2006 Canonical Ltd
3
3
# This program is free software; you can redistribute it and/or modify
4
4
# it under the terms of the GNU General Public License as published by
5
5
# the Free Software Foundation; either version 2 of the License, or
6
6
# (at your option) any later version.
8
8
# This program is distributed in the hope that it will be useful,
9
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
10
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11
11
# GNU General Public License for more details.
13
13
# You should have received a copy of the GNU General Public License
14
14
# along with this program; if not, write to the Free Software
15
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
17
"""Inventories map files to their name in a revision."""
19
# TODO: Maybe store inventory_id in the file? Not really needed.
21
__author__ = "Martin Pool <mbp@canonical.com>"
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.
24
26
# This should really be an id randomly assigned when the tree is
25
27
# created, but it's not for now.
26
28
ROOT_ID = "TREE_ROOT"
29
import sys, os.path, types, re
33
from cElementTree import Element, ElementTree, SubElement
35
from elementtree.ElementTree import Element, ElementTree, SubElement
37
from xml import XMLMixin
38
from errors import bailout, BzrError
34
from bzrlib.lazy_import import lazy_import
35
lazy_import(globals(), """
41
from bzrlib.osutils import uuid, quotefn, splitpath, joinpath, appendpath
49
from bzrlib.errors import (
42
53
from bzrlib.trace import mutter
44
class InventoryEntry(XMLMixin):
56
class InventoryEntry(object):
45
57
"""Description of a versioned file.
47
59
An InventoryEntry has the following fields, which are also
48
60
present in the XML inventory-entry element:
51
* *name*: (only the basename within the directory, must not
53
* *kind*: "directory" or "file"
54
* *directory_id*: (if absent/null means the branch root directory)
55
* *text_sha1*: only for files
56
* *text_size*: in bytes, only for files
57
* *text_id*: identifier for the text version, only for files
59
InventoryEntries can also exist inside a WorkingTree
60
inventory, in which case they are not yet bound to a
61
particular revision of the file. In that case the text_sha1,
62
text_size and text_id are absent.
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.)
65
86
>>> i = Inventory()
68
>>> i.add(InventoryEntry('123', 'src', 'directory', ROOT_ID))
69
>>> i.add(InventoryEntry('2323', 'hello.c', 'file', parent_id='123'))
70
>>> for j in i.iter_entries():
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])
73
('src', InventoryEntry('123', 'src', kind='directory', parent_id='TREE_ROOT'))
74
('src/hello.c', InventoryEntry('2323', 'hello.c', kind='file', parent_id='123'))
75
>>> i.add(InventoryEntry('2323', 'bye.c', 'file', '123'))
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'))
76
101
Traceback (most recent call last):
78
BzrError: ('inventory already contains entry with id {2323}', [])
79
>>> i.add(InventoryEntry('2324', 'bye.c', 'file', '123'))
80
>>> i.add(InventoryEntry('2325', 'wibble', 'directory', '123'))
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)
81
108
>>> i.path2id('src/wibble')
85
>>> i.add(InventoryEntry('2326', 'wibble.c', 'file', '2325'))
112
>>> i.add(InventoryFile('2326', 'wibble.c', '2325'))
113
InventoryFile('2326', 'wibble.c', parent_id='2325', sha1=None, len=None)
87
InventoryEntry('2326', 'wibble.c', kind='file', parent_id='2325')
88
>>> for j in i.iter_entries():
90
... assert i.path2id(j[0])
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)
96
125
src/wibble/wibble.c
97
126
>>> i.id2path('2326')
98
127
'src/wibble/wibble.c'
100
:todo: Maybe also keep the full path of the entry, and the children?
101
But those depend on its position within a particular inventory, and
102
it would be nice not to need to hold the backpointer here.
105
# TODO: split InventoryEntry into subclasses for files,
106
# directories, etc etc.
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'
108
def __init__(self, file_id, name, kind, parent_id, text_id=None):
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):
109
276
"""Create an InventoryEntry
111
278
The filename must be a single component, relative to the
112
279
parent directory; it cannot be a whole path or relative name.
114
>>> e = InventoryEntry('123', 'hello.c', 'file', ROOT_ID)
281
>>> e = InventoryFile('123', 'hello.c', ROOT_ID)
119
>>> e = InventoryEntry('123', 'src/hello.c', 'file', ROOT_ID)
286
>>> e = InventoryFile('123', 'src/hello.c', ROOT_ID)
120
287
Traceback (most recent call last):
121
BzrError: ("InventoryEntry name is not a simple filename: 'src/hello.c'", [])
288
InvalidEntryName: Invalid entry name: src/hello.c
124
if len(splitpath(name)) != 1:
125
bailout('InventoryEntry name is not a simple filename: %r'
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
128
297
self.file_id = file_id
130
assert kind in ['file', 'directory']
132
299
self.text_id = text_id
133
300
self.parent_id = parent_id
134
self.text_sha1 = None
135
self.text_size = None
136
if kind == 'directory':
139
assert kind == 'file'
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))
142
331
def sorted_children(self):
143
l = self.children.items()
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}' %
149
other = InventoryEntry(self.file_id, self.name, self.kind,
150
self.parent_id, text_id=self.text_id)
151
other.text_sha1 = self.text_sha1
152
other.text_size = self.text_size
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
156
407
def __repr__(self):
157
return ("%s(%r, %r, kind=%r, parent_id=%r)"
408
return ("%s(%r, %r, parent_id=%r, revision=%r)"
158
409
% (self.__class__.__name__,
165
def to_element(self):
166
"""Convert to XML element"""
169
e.set('name', self.name)
170
e.set('file_id', self.file_id)
171
e.set('kind', self.kind)
173
if self.text_size != None:
174
e.set('text_size', '%d' % self.text_size)
176
for f in ['text_id', 'text_sha1']:
181
# to be conservative, we don't externalize the root pointers
182
# for now, leaving them as null in the xml form. in a future
183
# version it will be implied by nested elements.
184
if self.parent_id != ROOT_ID:
185
assert isinstance(self.parent_id, basestring)
186
e.set('parent_id', self.parent_id)
193
def from_element(cls, elt):
194
assert elt.tag == 'entry'
196
## original format inventories don't have a parent_id for
197
## nodes in the root directory, but it's cleaner to use one
199
parent_id = elt.get('parent_id')
200
if parent_id == None:
203
self = cls(elt.get('file_id'), elt.get('name'), elt.get('kind'), parent_id)
204
self.text_id = elt.get('text_id')
205
self.text_sha1 = elt.get('text_sha1')
207
## mutter("read inventoryentry: %r" % (elt.attrib))
209
v = elt.get('text_size')
210
self.text_size = v and int(v)
215
from_element = classmethod(from_element)
217
def __cmp__(self, other):
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):
220
459
if not isinstance(other, InventoryEntry):
221
460
return NotImplemented
223
return cmp(self.file_id, other.file_id) \
224
or cmp(self.name, other.name) \
225
or cmp(self.text_sha1, other.text_sha1) \
226
or cmp(self.text_size, other.text_size) \
227
or cmp(self.text_id, other.text_id) \
228
or cmp(self.parent_id, other.parent_id) \
229
or cmp(self.kind, other.kind)
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):
233
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"""
234
518
def __init__(self, file_id):
235
519
self.file_id = file_id
236
520
self.children = {}
237
self.kind = 'root_directory'
521
self.kind = 'directory'
238
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)
241
def __cmp__(self, other):
529
def __eq__(self, other):
244
530
if not isinstance(other, RootEntry):
245
531
return NotImplemented
246
return cmp(self.file_id, other.file_id) \
247
or cmp(self.children, other.children)
251
class Inventory(XMLMixin):
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):
252
825
"""Inventory of versioned files in a tree.
254
An Inventory acts like a set of InventoryEntry items. You can
255
also look files up by their file_id or name.
257
May be read from and written to a metadata file in a tree. To
258
manipulate the inventory (for example to add a file), it is read
259
in, modified, and then written back out.
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.
261
831
The inventory represents a typical unix file tree, with
262
832
directories containing files and subdirectories. We never store
316
872
The inventory is created with a default root directory, with
319
self.root = RootEntry(ROOT_ID)
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):
320
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())
323
896
def __iter__(self):
324
897
return iter(self._byid)
327
899
def __len__(self):
328
900
"""Returns number of entries."""
329
901
return len(self._byid)
332
903
def iter_entries(self, from_dir=None):
333
904
"""Return (path, entry) pairs, in order by name."""
337
elif isinstance(from_dir, basestring):
338
from_dir = self._byid[from_dir]
340
kids = from_dir.children.items()
342
for name, ie in kids:
344
if ie.kind == 'directory':
345
for cn, cie in self.iter_entries(from_dir=ie.file_id):
346
yield '/'.join((name, cn)), cie
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):
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:
964
elif isinstance(from_dir, basestring):
965
from_dir = self._byid[from_dir]
967
stack = [(u'', from_dir)]
969
cur_relpath, cur_dir = stack.pop()
972
for child_name, child_ie in sorted(cur_dir.children.iteritems()):
974
child_relpath = cur_relpath + child_name
976
yield child_relpath, child_ie
978
if child_ie.kind == 'directory':
979
child_dirs.append((child_relpath+'/', child_ie))
980
stack.extend(reversed(child_dirs))
983
"""Return list of (path, ie) for all entries except the root.
985
This may be faster than iter_entries.
988
def descend(dir_ie, dir_path):
989
kids = dir_ie.children.items()
991
for name, ie in kids:
992
child_path = osutils.pathjoin(dir_path, name)
993
accum.append((child_path, ie))
994
if ie.kind == 'directory':
995
descend(ie, child_path)
997
descend(self.root, u'')
350
1000
def directories(self):
351
"""Return (path, entry) pairs for all directories.
1001
"""Return (path, entry) pairs for all directories, including the root.
353
def descend(parent_ie):
354
parent_name = parent_ie.name
355
yield parent_name, parent_ie
357
# directory children in sorted order
359
for ie in parent_ie.children.itervalues():
360
if ie.kind == 'directory':
361
dn.append((ie.name, ie))
1004
def descend(parent_ie, parent_path):
1005
accum.append((parent_path, parent_ie))
364
for name, child_ie in dn:
365
for sub_name, sub_ie in descend(child_ie):
366
yield appendpath(parent_name, sub_name), sub_ie
1007
kids = [(ie.name, ie) for ie in parent_ie.children.itervalues() if ie.kind == 'directory']
368
for name, ie in descend(self.root):
1010
for name, child_ie in kids:
1011
child_path = osutils.pathjoin(parent_path, name)
1012
descend(child_ie, child_path)
1013
descend(self.root, u'')
373
1016
def __contains__(self, file_id):
374
1017
"""True if this entry contains a file with given id.
376
1019
>>> inv = Inventory()
377
>>> inv.add(InventoryEntry('123', 'foo.c', 'file', ROOT_ID))
1020
>>> inv.add(InventoryFile('123', 'foo.c', ROOT_ID))
1021
InventoryFile('123', 'foo.c', parent_id='TREE_ROOT', sha1=None, len=None)
378
1022
>>> '123' in inv
380
1024
>>> '456' in inv
383
return file_id in self._byid
1027
return (file_id in self._byid)
386
1029
def __getitem__(self, file_id):
387
1030
"""Return the entry for given file_id.
389
1032
>>> inv = Inventory()
390
>>> inv.add(InventoryEntry('123123', 'hello.c', 'file', ROOT_ID))
1033
>>> inv.add(InventoryFile('123123', 'hello.c', ROOT_ID))
1034
InventoryFile('123123', 'hello.c', parent_id='TREE_ROOT', sha1=None, len=None)
391
1035
>>> inv['123123'].name
395
raise BzrError("can't look up file_id None")
398
1039
return self._byid[file_id]
399
1040
except KeyError:
400
raise BzrError("file_id {%s} not in inventory" % file_id)
1041
# really we're passing an inventory, not a tree...
1042
raise errors.NoSuchId(self, file_id)
1044
def get_file_kind(self, file_id):
1045
return self._byid[file_id].kind
403
1047
def get_child(self, parent_id, filename):
404
1048
return self[parent_id].children.get(filename)
407
1050
def add(self, entry):
408
1051
"""Add entry to inventory.
410
1053
To add a file to a branch ready to be committed, use Branch.add,
1056
Returns the new entry object.
412
1058
if entry.file_id in self._byid:
413
bailout("inventory already contains entry with id {%s}" % entry.file_id)
1059
raise BzrError("inventory already contains entry with id {%s}" % entry.file_id)
1061
if entry.parent_id is None:
1062
assert self.root is None and len(self._byid) == 0
1063
self._set_root(entry)
416
1066
parent = self._byid[entry.parent_id]
417
1067
except KeyError:
418
bailout("parent_id {%s} not in inventory" % entry.parent_id)
1068
raise BzrError("parent_id {%s} not in inventory" % entry.parent_id)
420
if parent.children.has_key(entry.name):
421
bailout("%s is already versioned" %
422
appendpath(self.id2path(parent.file_id), entry.name))
1070
if entry.name in parent.children:
1071
raise BzrError("%s is already versioned" %
1072
osutils.pathjoin(self.id2path(parent.file_id), entry.name))
424
1074
self._byid[entry.file_id] = entry
425
1075
parent.children[entry.name] = entry
428
def add_path(self, relpath, kind, file_id=None):
1078
def add_path(self, relpath, kind, file_id=None, parent_id=None):
429
1079
"""Add entry from a path.
431
The immediate parent must already be versioned"""
432
parts = bzrlib.osutils.splitpath(relpath)
1081
The immediate parent must already be versioned.
1083
Returns the new entry object."""
1085
parts = osutils.splitpath(relpath)
433
1087
if len(parts) == 0:
434
bailout("cannot re-add root of inventory")
437
file_id = bzrlib.branch.gen_file_id(relpath)
439
parent_id = self.path2id(parts[:-1])
440
assert parent_id != None
441
ie = InventoryEntry(file_id, parts[-1],
442
kind=kind, parent_id=parent_id)
1089
file_id = generate_ids.gen_root_id()
1090
self.root = InventoryDirectory(file_id, '', None)
1091
self._byid = {self.root.file_id: self.root}
1094
parent_path = parts[:-1]
1095
parent_id = self.path2id(parent_path)
1096
if parent_id is None:
1097
raise errors.NotVersionedError(path=parent_path)
1098
ie = make_entry(kind, parts[-1], parent_id, file_id)
443
1099
return self.add(ie)
446
1101
def __delitem__(self, file_id):
447
1102
"""Remove entry by id.
449
1104
>>> inv = Inventory()
450
>>> inv.add(InventoryEntry('123', 'foo.c', 'file', ROOT_ID))
1105
>>> inv.add(InventoryFile('123', 'foo.c', ROOT_ID))
1106
InventoryFile('123', 'foo.c', parent_id='TREE_ROOT', sha1=None, len=None)
451
1107
>>> '123' in inv
453
1109
>>> del inv['123']
added
removed
bzrlib/inventory.py
