1
# Copyright (C) 2006, 2007 Canonical Ltd
3
# This program is free software; you can redistribute it and/or modify
4
# it under the terms of the GNU General Public License as published by
5
# the Free Software Foundation; either version 2 of the License, or
6
# (at your option) any later version.
8
# This program is distributed in the hope that it will be useful,
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11
# GNU General Public License for more details.
13
# You should have received a copy of the GNU General Public License
14
# along with this program; if not, write to the Free Software
15
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
17
"""MutableTree object.
19
See MutableTree for more details.
23
from bzrlib.lazy_import import lazy_import
24
lazy_import(globals(), """
41
from bzrlib.decorators import needs_read_lock, needs_write_lock
44
def needs_tree_write_lock(unbound):
45
"""Decorate unbound to take out and release a tree_write lock."""
46
def tree_write_locked(self, *args, **kwargs):
47
self.lock_tree_write()
49
return unbound(self, *args, **kwargs)
52
tree_write_locked.__doc__ = unbound.__doc__
53
tree_write_locked.__name__ = unbound.__name__
54
return tree_write_locked
57
class MutableTree(tree.Tree):
58
"""A MutableTree is a specialisation of Tree which is able to be mutated.
60
Generally speaking these mutations are only possible within a lock_write
61
context, and will revert if the lock is broken abnormally - but this cannot
62
be guaranteed - depending on the exact implementation of the mutable state.
64
The most common form of Mutable Tree is WorkingTree, see bzrlib.workingtree.
65
For tests we also have MemoryTree which is a MutableTree whose contents are
68
For now, we are not treating MutableTree as an interface to provide
69
conformance tests for - rather we are testing MemoryTree specifically, and
70
interface testing implementations of WorkingTree.
72
A mutable tree always has an associated Branch and BzrDir object - the
73
branch and bzrdir attributes.
75
def __init__(self, *args, **kw):
76
super(MutableTree, self).__init__(*args, **kw)
77
# Is this tree on a case-insensitive or case-preserving file-system?
78
# Sub-classes may initialize to False if they detect they are being
79
# used on media which doesn't differentiate the case of names.
80
self.case_sensitive = True
82
@needs_tree_write_lock
83
def add(self, files, ids=None, kinds=None):
84
"""Add paths to the set of versioned paths.
86
Note that the command line normally calls smart_add instead,
87
which can automatically recurse.
89
This adds the files to the inventory, so that they will be
90
recorded by the next commit.
92
:param files: List of paths to add, relative to the base of the tree.
93
:param ids: If set, use these instead of automatically generated ids.
94
Must be the same length as the list of files, but may
95
contain None for ids that are to be autogenerated.
96
:param kinds: Optional parameter to specify the kinds to be used for
99
TODO: Perhaps callback with the ids and paths as they're added.
101
if isinstance(files, basestring):
102
# XXX: Passing a single string is inconsistent and should be
104
if not (ids is None or isinstance(ids, basestring)):
105
raise AssertionError()
106
if not (kinds is None or isinstance(kinds, basestring)):
107
raise AssertionError()
111
if kinds is not None:
114
files = [path.strip('/') for path in files]
117
ids = [None] * len(files)
119
if not (len(ids) == len(files)):
120
raise AssertionError()
122
kinds = [None] * len(files)
123
elif not len(kinds) == len(files):
124
raise AssertionError()
126
# generic constraint checks:
127
if self.is_control_filename(f):
128
raise errors.ForbiddenControlFileError(filename=f)
129
fp = osutils.splitpath(f)
130
# fill out file kinds for all files [not needed when we stop
131
# caring about the instantaneous file kind within a uncommmitted tree
133
self._gather_kinds(files, kinds)
134
self._add(files, ids, kinds)
136
def add_reference(self, sub_tree):
137
"""Add a TreeReference to the tree, pointing at sub_tree"""
138
raise errors.UnsupportedOperation(self.add_reference, self)
140
def _add_reference(self, sub_tree):
141
"""Standard add_reference implementation, for use by subclasses"""
143
sub_tree_path = self.relpath(sub_tree.basedir)
144
except errors.PathNotChild:
145
raise errors.BadReferenceTarget(self, sub_tree,
146
'Target not inside tree.')
147
sub_tree_id = sub_tree.get_root_id()
148
if sub_tree_id == self.get_root_id():
149
raise errors.BadReferenceTarget(self, sub_tree,
150
'Trees have the same root id.')
151
if sub_tree_id in self.inventory:
152
raise errors.BadReferenceTarget(self, sub_tree,
153
'Root id already present in tree')
154
self._add([sub_tree_path], [sub_tree_id], ['tree-reference'])
156
def _add(self, files, ids, kinds):
157
"""Helper function for add - updates the inventory.
159
:param files: sequence of pathnames, relative to the tree root
160
:param ids: sequence of suggested ids for the files (may be None)
161
:param kinds: sequence of inventory kinds of the files (i.e. may
162
contain "tree-reference")
164
raise NotImplementedError(self._add)
166
@needs_tree_write_lock
167
def apply_inventory_delta(self, changes):
168
"""Apply changes to the inventory as an atomic operation.
170
:param changes: An inventory delta to apply to the working tree's
173
:seealso Inventory.apply_delta: For details on the changes parameter.
177
inv.apply_delta(changes)
178
self._write_inventory(inv)
181
def commit(self, message=None, revprops=None, *args,
183
# avoid circular imports
184
from bzrlib import commit
187
possible_master_transports=[]
188
if not 'branch-nick' in revprops:
189
revprops['branch-nick'] = self.branch._get_nick(
190
kwargs.get('local', False),
191
possible_master_transports)
192
authors = kwargs.pop('authors', None)
193
author = kwargs.pop('author', None)
194
if authors is not None:
195
if author is not None:
196
raise AssertionError('Specifying both author and authors '
197
'is not allowed. Specify just authors instead')
198
if 'author' in revprops or 'authors' in revprops:
199
# XXX: maybe we should just accept one of them?
200
raise AssertionError('author property given twice')
202
for individual in authors:
203
if '\n' in individual:
204
raise AssertionError('\\n is not a valid character '
205
'in an author identity')
206
revprops['authors'] = '\n'.join(authors)
207
if author is not None:
208
symbol_versioning.warn('The parameter author was deprecated'
209
' in version 1.13. Use authors instead',
211
if 'author' in revprops or 'authors' in revprops:
212
# XXX: maybe we should just accept one of them?
213
raise AssertionError('author property given twice')
215
raise AssertionError('\\n is not a valid character '
216
'in an author identity')
217
revprops['authors'] = author
218
# args for wt.commit start at message from the Commit.commit method,
219
args = (message, ) + args
220
for hook in MutableTree.hooks['start_commit']:
222
committed_id = commit.Commit().commit(working_tree=self,
224
possible_master_transports=possible_master_transports,
226
post_hook_params = PostCommitHookParams(self)
227
for hook in MutableTree.hooks['post_commit']:
228
hook(post_hook_params)
231
def _gather_kinds(self, files, kinds):
232
"""Helper function for add - sets the entries of kinds."""
233
raise NotImplementedError(self._gather_kinds)
236
def has_changes(self, from_tree):
237
"""Quickly check that the tree contains at least one change.
239
:return: True if a change is found. False otherwise
241
changes = self.iter_changes(from_tree)
243
change = changes.next()
244
# Exclude root (talk about black magic... --vila 20090629)
245
if change[4] == (None, None):
246
change = changes.next()
248
except StopIteration:
253
def last_revision(self):
254
"""Return the revision id of the last commit performed in this tree.
256
In early tree formats the result of last_revision is the same as the
257
branch last_revision, but that is no longer the case for modern tree
260
last_revision returns the left most parent id, or None if there are no
263
last_revision was deprecated as of 0.11. Please use get_parent_ids
266
raise NotImplementedError(self.last_revision)
268
def lock_tree_write(self):
269
"""Lock the working tree for write, and the branch for read.
271
This is useful for operations which only need to mutate the working
272
tree. Taking out branch write locks is a relatively expensive process
273
and may fail if the branch is on read only media. So branch write locks
274
should only be taken out when we are modifying branch data - such as in
275
operations like commit, pull, uncommit and update.
277
raise NotImplementedError(self.lock_tree_write)
279
def lock_write(self):
280
"""Lock the tree and its branch. This allows mutating calls to be made.
282
Some mutating methods will take out implicit write locks, but in
283
general you should always obtain a write lock before calling mutating
286
raise NotImplementedError(self.lock_write)
289
def mkdir(self, path, file_id=None):
290
"""Create a directory in the tree. if file_id is None, one is assigned.
292
:param path: A unicode file path.
293
:param file_id: An optional file-id.
294
:return: the file id of the new directory.
296
raise NotImplementedError(self.mkdir)
298
def _observed_sha1(self, file_id, path, (sha1, stat_value)):
299
"""Tell the tree we have observed a paths sha1.
301
The intent of this function is to allow trees that have a hashcache to
302
update the hashcache during commit. If the observed file is too new
303
(based on the stat_value) to be safely hash-cached the tree will ignore
306
The default implementation does nothing.
308
:param file_id: The file id
309
:param path: The file path
310
:param sha1: The sha 1 that was observed.
311
:param stat_value: A stat result for the file the sha1 was read from.
315
def _fix_case_of_inventory_path(self, path):
316
"""If our tree isn't case sensitive, return the canonical path"""
317
if not self.case_sensitive:
318
path = self.get_canonical_inventory_path(path)
322
def put_file_bytes_non_atomic(self, file_id, bytes):
323
"""Update the content of a file in the tree.
325
Note that the file is written in-place rather than being
326
written to a temporary location and renamed. As a consequence,
327
readers can potentially see the file half-written.
329
:param file_id: file-id of the file
330
:param bytes: the new file contents
332
raise NotImplementedError(self.put_file_bytes_non_atomic)
334
def set_parent_ids(self, revision_ids, allow_leftmost_as_ghost=False):
335
"""Set the parents ids of the working tree.
337
:param revision_ids: A list of revision_ids.
339
raise NotImplementedError(self.set_parent_ids)
341
def set_parent_trees(self, parents_list, allow_leftmost_as_ghost=False):
342
"""Set the parents of the working tree.
344
:param parents_list: A list of (revision_id, tree) tuples.
345
If tree is None, then that element is treated as an unreachable
346
parent tree - i.e. a ghost.
348
raise NotImplementedError(self.set_parent_trees)
350
@needs_tree_write_lock
351
def smart_add(self, file_list, recurse=True, action=None, save=True):
352
"""Version file_list, optionally recursing into directories.
354
This is designed more towards DWIM for humans than API clarity.
355
For the specific behaviour see the help for cmd_add().
357
:param action: A reporter to be called with the inventory, parent_ie,
358
path and kind of the path being added. It may return a file_id if
359
a specific one should be used.
360
:param save: Save the inventory after completing the adds. If False
361
this provides dry-run functionality by doing the add and not saving
363
:return: A tuple - files_added, ignored_files. files_added is the count
364
of added files, and ignored_files is a dict mapping files that were
365
ignored to the rule that caused them to be ignored.
367
# not in an inner loop; and we want to remove direct use of this,
368
# so here as a reminder for now. RBC 20070703
369
from bzrlib.inventory import InventoryEntry
371
action = add.AddAction()
374
# no paths supplied: add the entire tree.
376
# mutter("smart add of %r")
383
# validate user file paths and convert all paths to tree
384
# relative : it's cheaper to make a tree relative path an abspath
385
# than to convert an abspath to tree relative, and it's cheaper to
386
# perform the canonicalization in bulk.
387
for filepath in osutils.canonical_relpaths(self.basedir, file_list):
388
rf = _FastPath(filepath)
389
# validate user parameters. Our recursive code avoids adding new
390
# files that need such validation
391
if self.is_control_filename(rf.raw_path):
392
raise errors.ForbiddenControlFileError(filename=rf.raw_path)
394
abspath = self.abspath(rf.raw_path)
395
kind = osutils.file_kind(abspath)
396
if kind == 'directory':
397
# schedule the dir for scanning
400
if not InventoryEntry.versionable_kind(kind):
401
raise errors.BadFileKindError(filename=abspath, kind=kind)
402
# ensure the named path is added, so that ignore rules in the later
403
# directory walk dont skip it.
404
# we dont have a parent ie known yet.: use the relatively slower
405
# inventory probing method
406
versioned = inv.has_filename(rf.raw_path)
409
added.extend(_add_one_and_parent(self, inv, None, rf, kind, action))
412
# no need to walk any directories at all.
413
if len(added) > 0 and save:
414
self._write_inventory(inv)
415
return added, ignored
417
# only walk the minimal parents needed: we have user_dirs to override
421
is_inside = osutils.is_inside_or_parent_of_any
422
for path in sorted(user_dirs):
423
if (prev_dir is None or not is_inside([prev_dir], path.raw_path)):
424
dirs_to_add.append((path, None))
425
prev_dir = path.raw_path
427
illegalpath_re = re.compile(r'[\r\n]')
428
# dirs_to_add is initialised to a list of directories, but as we scan
429
# directories we append files to it.
430
# XXX: We should determine kind of files when we scan them rather than
431
# adding to this list. RBC 20070703
432
for directory, parent_ie in dirs_to_add:
433
# directory is tree-relative
434
abspath = self.abspath(directory.raw_path)
436
# get the contents of this directory.
438
# find the kind of the path being added.
439
kind = osutils.file_kind(abspath)
441
if not InventoryEntry.versionable_kind(kind):
442
trace.warning("skipping %s (can't add file of kind '%s')",
445
if illegalpath_re.search(directory.raw_path):
446
trace.warning("skipping %r (contains \\n or \\r)" % abspath)
449
if parent_ie is not None:
450
versioned = directory.base_path in parent_ie.children
452
# without the parent ie, use the relatively slower inventory
454
versioned = inv.has_filename(
455
self._fix_case_of_inventory_path(directory.raw_path))
457
if kind == 'directory':
459
sub_branch = bzrdir.BzrDir.open(abspath)
461
except errors.NotBranchError:
463
except errors.UnsupportedFormatError:
468
if directory.raw_path == '':
469
# mutter("tree root doesn't need to be added")
473
# mutter("%r is already versioned", abspath)
475
# XXX: This is wrong; people *might* reasonably be trying to
476
# add subtrees as subtrees. This should probably only be done
477
# in formats which can represent subtrees, and even then
478
# perhaps only when the user asked to add subtrees. At the
479
# moment you can add them specially through 'join --reference',
480
# which is perhaps reasonable: adding a new reference is a
481
# special operation and can have a special behaviour. mbp
483
trace.mutter("%r is a nested bzr tree", abspath)
485
_add_one(self, inv, parent_ie, directory, kind, action)
486
added.append(directory.raw_path)
488
if kind == 'directory' and not sub_tree:
489
if parent_ie is not None:
491
this_ie = parent_ie.children[directory.base_path]
493
# without the parent ie, use the relatively slower inventory
495
this_id = inv.path2id(
496
self._fix_case_of_inventory_path(directory.raw_path))
500
this_ie = inv[this_id]
502
for subf in sorted(os.listdir(abspath)):
503
# here we could use TreeDirectory rather than
504
# string concatenation.
505
subp = osutils.pathjoin(directory.raw_path, subf)
506
# TODO: is_control_filename is very slow. Make it faster.
507
# TreeDirectory.is_control_filename could also make this
508
# faster - its impossible for a non root dir to have a
510
if self.is_control_filename(subp):
511
trace.mutter("skip control directory %r", subp)
512
elif subf in this_ie.children:
513
# recurse into this already versioned subdir.
514
dirs_to_add.append((_FastPath(subp, subf), this_ie))
516
# user selection overrides ignoes
517
# ignore while selecting files - if we globbed in the
518
# outer loop we would ignore user files.
519
ignore_glob = self.is_ignored(subp)
520
if ignore_glob is not None:
521
# mutter("skip ignored sub-file %r", subp)
522
ignored.setdefault(ignore_glob, []).append(subp)
524
#mutter("queue to add sub-file %r", subp)
525
dirs_to_add.append((_FastPath(subp, subf), this_ie))
529
self._write_inventory(inv)
531
self.read_working_inventory()
532
return added, ignored
534
def update_basis_by_delta(self, new_revid, delta):
535
"""Update the parents of this tree after a commit.
537
This gives the tree one parent, with revision id new_revid. The
538
inventory delta is applied to the current basis tree to generate the
539
inventory for the parent new_revid, and all other parent trees are
542
All the changes in the delta should be changes synchronising the basis
543
tree with some or all of the working tree, with a change to a directory
544
requiring that its contents have been recursively included. That is,
545
this is not a general purpose tree modification routine, but a helper
546
for commit which is not required to handle situations that do not arise
549
See the inventory developers documentation for the theory behind
552
:param new_revid: The new revision id for the trees parent.
553
:param delta: An inventory delta (see apply_inventory_delta) describing
554
the changes from the current left most parent revision to new_revid.
556
# if the tree is updated by a pull to the branch, as happens in
557
# WorkingTree2, when there was no separation between branch and tree,
558
# then just clear merges, efficiency is not a concern for now as this
559
# is legacy environments only, and they are slow regardless.
560
if self.last_revision() == new_revid:
561
self.set_parent_ids([new_revid])
563
# generic implementation based on Inventory manipulation. See
564
# WorkingTree classes for optimised versions for specific format trees.
565
basis = self.basis_tree()
567
# TODO: Consider re-evaluating the need for this with CHKInventory
568
# we don't strictly need to mutate an inventory for this
569
# it only makes sense when apply_delta is cheaper than get_inventory()
570
inventory = basis.inventory._get_mutable_inventory()
572
inventory.apply_delta(delta)
573
rev_tree = revisiontree.RevisionTree(self.branch.repository,
574
inventory, new_revid)
575
self.set_parent_trees([(new_revid, rev_tree)])
578
class MutableTreeHooks(hooks.Hooks):
579
"""A dictionary mapping a hook name to a list of callables for mutabletree
584
"""Create the default hooks.
587
hooks.Hooks.__init__(self)
588
self.create_hook(hooks.HookPoint('start_commit',
589
"Called before a commit is performed on a tree. The start commit "
590
"hook is able to change the tree before the commit takes place. "
591
"start_commit is called with the bzrlib.mutabletree.MutableTree "
592
"that the commit is being performed on.", (1, 4), None))
593
self.create_hook(hooks.HookPoint('post_commit',
594
"Called after a commit is performed on a tree. The hook is "
595
"called with a bzrlib.mutabletree.PostCommitHookParams object. "
596
"The mutable tree the commit was performed on is available via "
597
"the mutable_tree attribute of that object.", (2, 0), None))
600
# install the default hooks into the MutableTree class.
601
MutableTree.hooks = MutableTreeHooks()
604
class PostCommitHookParams(object):
605
"""Parameters for the post_commit hook.
607
To access the parameters, use the following attributes:
609
* mutable_tree - the MutableTree object
612
def __init__(self, mutable_tree):
613
"""Create the parameters for the post_commit hook."""
614
self.mutable_tree = mutable_tree
617
class _FastPath(object):
618
"""A path object with fast accessors for things like basename."""
620
__slots__ = ['raw_path', 'base_path']
622
def __init__(self, path, base_path=None):
623
"""Construct a FastPath from path."""
624
if base_path is None:
625
self.base_path = osutils.basename(path)
627
self.base_path = base_path
630
def __cmp__(self, other):
631
return cmp(self.raw_path, other.raw_path)
634
return hash(self.raw_path)
637
def _add_one_and_parent(tree, inv, parent_ie, path, kind, action):
638
"""Add a new entry to the inventory and automatically add unversioned parents.
640
:param inv: Inventory which will receive the new entry.
641
:param parent_ie: Parent inventory entry if known, or None. If
642
None, the parent is looked up by name and used if present, otherwise it
643
is recursively added.
644
:param kind: Kind of new entry (file, directory, etc)
645
:param action: callback(inv, parent_ie, path, kind); return ignored.
646
:return: A list of paths which have been added.
648
# Nothing to do if path is already versioned.
649
# This is safe from infinite recursion because the tree root is
651
if parent_ie is not None:
652
# we have a parent ie already
655
# slower but does not need parent_ie
656
if inv.has_filename(tree._fix_case_of_inventory_path(path.raw_path)):
658
# its really not there : add the parent
659
# note that the dirname use leads to some extra str copying etc but as
660
# there are a limited number of dirs we can be nested under, it should
661
# generally find it very fast and not recurse after that.
662
added = _add_one_and_parent(tree, inv, None,
663
_FastPath(osutils.dirname(path.raw_path)), 'directory', action)
664
parent_id = inv.path2id(osutils.dirname(path.raw_path))
665
parent_ie = inv[parent_id]
666
_add_one(tree, inv, parent_ie, path, kind, action)
667
return added + [path.raw_path]
670
def _add_one(tree, inv, parent_ie, path, kind, file_id_callback):
671
"""Add a new entry to the inventory.
673
:param inv: Inventory which will receive the new entry.
674
:param parent_ie: Parent inventory entry.
675
:param kind: Kind of new entry (file, directory, etc)
676
:param file_id_callback: callback(inv, parent_ie, path, kind); return a
677
file_id or None to generate a new file id
680
file_id = file_id_callback(inv, parent_ie, path, kind)
681
entry = inv.make_entry(kind, path.base_path, parent_ie.file_id,
added
removed
bzrlib/mutabletree.py
