1
# Copyright (C) 2005-2011 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
"""Tree classes, representing directory at point in time.
21
from collections import deque
25
conflicts as _mod_conflicts,
31
revision as _mod_revision,
34
from bzrlib.decorators import needs_read_lock
35
from bzrlib.inventory import InventoryFile
36
from bzrlib.inter import InterObject
37
from bzrlib.osutils import fingerprint_file
38
from bzrlib.trace import note
42
"""Abstract file tree.
44
There are several subclasses:
46
* `WorkingTree` exists as files on disk editable by the user.
48
* `RevisionTree` is a tree as recorded at some point in the past.
50
Trees contain an `Inventory` object, and also know how to retrieve
51
file texts mentioned in the inventory, either from a working
52
directory or from a store.
54
It is possible for trees to contain files that are not described
55
in their inventory or vice versa; for this use `filenames()`.
57
Trees can be compared, etc, regardless of whether they are working
58
trees or versioned trees.
61
def changes_from(self, other, want_unchanged=False, specific_files=None,
62
extra_trees=None, require_versioned=False, include_root=False,
63
want_unversioned=False):
64
"""Return a TreeDelta of the changes from other to this tree.
66
:param other: A tree to compare with.
67
:param specific_files: An optional list of file paths to restrict the
68
comparison to. When mapping filenames to ids, all matches in all
69
trees (including optional extra_trees) are used, and all children of
70
matched directories are included.
71
:param want_unchanged: An optional boolean requesting the inclusion of
72
unchanged entries in the result.
73
:param extra_trees: An optional list of additional trees to use when
74
mapping the contents of specific_files (paths) to file_ids.
75
:param require_versioned: An optional boolean (defaults to False). When
76
supplied and True all the 'specific_files' must be versioned, or
77
a PathsNotVersionedError will be thrown.
78
:param want_unversioned: Scan for unversioned paths.
80
The comparison will be performed by an InterTree object looked up on
83
# Martin observes that Tree.changes_from returns a TreeDelta and this
84
# may confuse people, because the class name of the returned object is
85
# a synonym of the object referenced in the method name.
86
return InterTree.get(other, self).compare(
87
want_unchanged=want_unchanged,
88
specific_files=specific_files,
89
extra_trees=extra_trees,
90
require_versioned=require_versioned,
91
include_root=include_root,
92
want_unversioned=want_unversioned,
95
def iter_changes(self, from_tree, include_unchanged=False,
96
specific_files=None, pb=None, extra_trees=None,
97
require_versioned=True, want_unversioned=False):
98
"""See InterTree.iter_changes"""
99
intertree = InterTree.get(from_tree, self)
100
return intertree.iter_changes(include_unchanged, specific_files, pb,
101
extra_trees, require_versioned, want_unversioned=want_unversioned)
104
"""Get a list of the conflicts in the tree.
106
Each conflict is an instance of bzrlib.conflicts.Conflict.
108
return _mod_conflicts.ConflictList()
111
"""For trees that can have unversioned files, return all such paths."""
114
def get_parent_ids(self):
115
"""Get the parent ids for this tree.
117
:return: a list of parent ids. [] is returned to indicate
118
a tree with no parents.
119
:raises: BzrError if the parents are not known.
121
raise NotImplementedError(self.get_parent_ids)
123
def has_filename(self, filename):
124
"""True if the tree has given filename."""
125
raise NotImplementedError(self.has_filename)
127
def has_id(self, file_id):
128
return self.inventory.has_id(file_id)
130
def __contains__(self, file_id):
131
return self.has_id(file_id)
133
def has_or_had_id(self, file_id):
134
return self.inventory.has_id(file_id)
136
def is_ignored(self, filename):
137
"""Check whether the filename is ignored by this tree.
139
:param filename: The relative filename within the tree.
140
:return: True if the filename is ignored.
145
return iter(self.inventory)
147
def all_file_ids(self):
148
"""Iterate through all file ids, including ids for missing files."""
149
return set(self.inventory)
151
def id2path(self, file_id):
152
"""Return the path for a file id.
156
return self.inventory.id2path(file_id)
159
def iter_entries_by_dir(self, specific_file_ids=None, yield_parents=False):
160
"""Walk the tree in 'by_dir' order.
162
This will yield each entry in the tree as a (path, entry) tuple.
163
The order that they are yielded is:
165
Directories are walked in a depth-first lexicographical order,
166
however, whenever a directory is reached, all of its direct child
167
nodes are yielded in lexicographical order before yielding the
170
For example, in the tree::
180
The yield order (ignoring root) would be::
181
a, f, a/b, a/d, a/b/c, a/d/e, f/g
183
:param yield_parents: If True, yield the parents from the root leading
184
down to specific_file_ids that have been requested. This has no
185
impact if specific_file_ids is None.
187
return self.inventory.iter_entries_by_dir(
188
specific_file_ids=specific_file_ids, yield_parents=yield_parents)
190
def iter_references(self):
191
if self.supports_tree_reference():
192
for path, entry in self.iter_entries_by_dir():
193
if entry.kind == 'tree-reference':
194
yield path, entry.file_id
196
def kind(self, file_id):
197
raise NotImplementedError("Tree subclass %s must implement kind"
198
% self.__class__.__name__)
200
def stored_kind(self, file_id):
201
"""File kind stored for this file_id.
203
May not match kind on disk for working trees. Always available
204
for versioned files, even when the file itself is missing.
206
return self.kind(file_id)
208
def path_content_summary(self, path):
209
"""Get a summary of the information about path.
211
All the attributes returned are for the canonical form, not the
212
convenient form (if content filters are in use.)
214
:param path: A relative path within the tree.
215
:return: A tuple containing kind, size, exec, sha1-or-link.
216
Kind is always present (see tree.kind()).
217
size is present if kind is file and the size of the
218
canonical form can be cheaply determined, None otherwise.
219
exec is None unless kind is file and the platform supports the 'x'
221
sha1-or-link is the link target if kind is symlink, or the sha1 if
222
it can be obtained without reading the file.
224
raise NotImplementedError(self.path_content_summary)
226
def get_reference_revision(self, file_id, path=None):
227
raise NotImplementedError("Tree subclass %s must implement "
228
"get_reference_revision"
229
% self.__class__.__name__)
231
def _comparison_data(self, entry, path):
232
"""Return a tuple of kind, executable, stat_value for a file.
234
entry may be None if there is no inventory entry for the file, but
235
path must always be supplied.
237
kind is None if there is no file present (even if an inventory id is
238
present). executable is False for non-file entries.
240
raise NotImplementedError(self._comparison_data)
242
def _file_size(self, entry, stat_value):
243
raise NotImplementedError(self._file_size)
245
def _get_inventory(self):
246
return self._inventory
248
def get_file(self, file_id, path=None):
249
"""Return a file object for the file file_id in the tree.
251
If both file_id and path are defined, it is implementation defined as
252
to which one is used.
254
raise NotImplementedError(self.get_file)
256
def get_file_with_stat(self, file_id, path=None):
257
"""Get a file handle and stat object for file_id.
259
The default implementation returns (self.get_file, None) for backwards
262
:param file_id: The file id to read.
263
:param path: The path of the file, if it is known.
264
:return: A tuple (file_handle, stat_value_or_None). If the tree has
265
no stat facility, or need for a stat cache feedback during commit,
266
it may return None for the second element of the tuple.
268
return (self.get_file(file_id, path), None)
270
def get_file_text(self, file_id, path=None):
271
"""Return the byte content of a file.
273
:param file_id: The file_id of the file.
274
:param path: The path of the file.
275
If both file_id and path are supplied, an implementation may use
278
my_file = self.get_file(file_id, path)
280
return my_file.read()
284
def get_file_lines(self, file_id, path=None):
285
"""Return the content of a file, as lines.
287
:param file_id: The file_id of the file.
288
:param path: The path of the file.
289
If both file_id and path are supplied, an implementation may use
292
return osutils.split_lines(self.get_file_text(file_id, path))
294
def get_file_mtime(self, file_id, path=None):
295
"""Return the modification time for a file.
297
:param file_id: The handle for this file.
298
:param path: The path that this file can be found at.
299
These must point to the same object.
301
raise NotImplementedError(self.get_file_mtime)
303
def get_file_size(self, file_id):
304
"""Return the size of a file in bytes.
306
This applies only to regular files. If invoked on directories or
307
symlinks, it will return None.
308
:param file_id: The file-id of the file
310
raise NotImplementedError(self.get_file_size)
312
def get_file_by_path(self, path):
313
return self.get_file(self._inventory.path2id(path), path)
315
def iter_files_bytes(self, desired_files):
316
"""Iterate through file contents.
318
Files will not necessarily be returned in the order they occur in
319
desired_files. No specific order is guaranteed.
321
Yields pairs of identifier, bytes_iterator. identifier is an opaque
322
value supplied by the caller as part of desired_files. It should
323
uniquely identify the file version in the caller's context. (Examples:
324
an index number or a TreeTransform trans_id.)
326
bytes_iterator is an iterable of bytestrings for the file. The
327
kind of iterable and length of the bytestrings are unspecified, but for
328
this implementation, it is a tuple containing a single bytestring with
329
the complete text of the file.
331
:param desired_files: a list of (file_id, identifier) pairs
333
for file_id, identifier in desired_files:
334
# We wrap the string in a tuple so that we can return an iterable
335
# of bytestrings. (Technically, a bytestring is also an iterable
336
# of bytestrings, but iterating through each character is not
338
cur_file = (self.get_file_text(file_id),)
339
yield identifier, cur_file
341
def get_symlink_target(self, file_id):
342
"""Get the target for a given file_id.
344
It is assumed that the caller already knows that file_id is referencing
346
:param file_id: Handle for the symlink entry.
347
:return: The path the symlink points to.
349
raise NotImplementedError(self.get_symlink_target)
351
def get_canonical_inventory_paths(self, paths):
352
"""Like get_canonical_inventory_path() but works on multiple items.
354
:param paths: A sequence of paths relative to the root of the tree.
355
:return: A list of paths, with each item the corresponding input path
356
adjusted to account for existing elements that match case
359
return list(self._yield_canonical_inventory_paths(paths))
361
def get_canonical_inventory_path(self, path):
362
"""Returns the first inventory item that case-insensitively matches path.
364
If a path matches exactly, it is returned. If no path matches exactly
365
but more than one path matches case-insensitively, it is implementation
366
defined which is returned.
368
If no path matches case-insensitively, the input path is returned, but
369
with as many path entries that do exist changed to their canonical
372
If you need to resolve many names from the same tree, you should
373
use get_canonical_inventory_paths() to avoid O(N) behaviour.
375
:param path: A paths relative to the root of the tree.
376
:return: The input path adjusted to account for existing elements
377
that match case insensitively.
379
return self._yield_canonical_inventory_paths([path]).next()
381
def _yield_canonical_inventory_paths(self, paths):
383
# First, if the path as specified exists exactly, just use it.
384
if self.path2id(path) is not None:
388
cur_id = self.get_root_id()
390
bit_iter = iter(path.split("/"))
394
for child in self.iter_children(cur_id):
396
# XXX: it seem like if the child is known to be in the
397
# tree, we shouldn't need to go from its id back to
398
# its path -- mbp 2010-02-11
400
# XXX: it seems like we could be more efficient
401
# by just directly looking up the original name and
402
# only then searching all children; also by not
403
# chopping paths so much. -- mbp 2010-02-11
404
child_base = os.path.basename(self.id2path(child))
405
if (child_base == elt):
406
# if we found an exact match, we can stop now; if
407
# we found an approximate match we need to keep
408
# searching because there might be an exact match
411
new_path = osutils.pathjoin(cur_path, child_base)
413
elif child_base.lower() == lelt:
415
new_path = osutils.pathjoin(cur_path, child_base)
416
except errors.NoSuchId:
417
# before a change is committed we can see this error...
422
# got to the end of this directory and no entries matched.
423
# Return what matched so far, plus the rest as specified.
424
cur_path = osutils.pathjoin(cur_path, elt, *list(bit_iter))
429
def get_root_id(self):
430
"""Return the file_id for the root of this tree."""
431
raise NotImplementedError(self.get_root_id)
433
def annotate_iter(self, file_id,
434
default_revision=_mod_revision.CURRENT_REVISION):
435
"""Return an iterator of revision_id, line tuples.
437
For working trees (and mutable trees in general), the special
438
revision_id 'current:' will be used for lines that are new in this
439
tree, e.g. uncommitted changes.
440
:param file_id: The file to produce an annotated version from
441
:param default_revision: For lines that don't match a basis, mark them
442
with this revision id. Not all implementations will make use of
445
raise NotImplementedError(self.annotate_iter)
447
def _get_plan_merge_data(self, file_id, other, base):
448
from bzrlib import versionedfile
449
vf = versionedfile._PlanMergeVersionedFile(file_id)
450
last_revision_a = self._get_file_revision(file_id, vf, 'this:')
451
last_revision_b = other._get_file_revision(file_id, vf, 'other:')
453
last_revision_base = None
455
last_revision_base = base._get_file_revision(file_id, vf, 'base:')
456
return vf, last_revision_a, last_revision_b, last_revision_base
458
def plan_file_merge(self, file_id, other, base=None):
459
"""Generate a merge plan based on annotations.
461
If the file contains uncommitted changes in this tree, they will be
462
attributed to the 'current:' pseudo-revision. If the file contains
463
uncommitted changes in the other tree, they will be assigned to the
464
'other:' pseudo-revision.
466
data = self._get_plan_merge_data(file_id, other, base)
467
vf, last_revision_a, last_revision_b, last_revision_base = data
468
return vf.plan_merge(last_revision_a, last_revision_b,
471
def plan_file_lca_merge(self, file_id, other, base=None):
472
"""Generate a merge plan based lca-newness.
474
If the file contains uncommitted changes in this tree, they will be
475
attributed to the 'current:' pseudo-revision. If the file contains
476
uncommitted changes in the other tree, they will be assigned to the
477
'other:' pseudo-revision.
479
data = self._get_plan_merge_data(file_id, other, base)
480
vf, last_revision_a, last_revision_b, last_revision_base = data
481
return vf.plan_lca_merge(last_revision_a, last_revision_b,
484
def _iter_parent_trees(self):
485
"""Iterate through parent trees, defaulting to Tree.revision_tree."""
486
for revision_id in self.get_parent_ids():
488
yield self.revision_tree(revision_id)
489
except errors.NoSuchRevisionInTree:
490
yield self.repository.revision_tree(revision_id)
493
def _file_revision(revision_tree, file_id):
494
"""Determine the revision associated with a file in a given tree."""
495
revision_tree.lock_read()
497
return revision_tree.inventory[file_id].revision
499
revision_tree.unlock()
501
def _get_file_revision(self, file_id, vf, tree_revision):
502
"""Ensure that file_id, tree_revision is in vf to plan the merge."""
504
if getattr(self, '_repository', None) is None:
505
last_revision = tree_revision
506
parent_keys = [(file_id, self._file_revision(t, file_id)) for t in
507
self._iter_parent_trees()]
508
vf.add_lines((file_id, last_revision), parent_keys,
509
self.get_file_lines(file_id))
510
repo = self.branch.repository
513
last_revision = self._file_revision(self, file_id)
514
base_vf = self._repository.texts
515
if base_vf not in vf.fallback_versionedfiles:
516
vf.fallback_versionedfiles.append(base_vf)
519
inventory = property(_get_inventory,
520
doc="Inventory of this Tree")
522
def _check_retrieved(self, ie, f):
525
fp = fingerprint_file(f)
528
if ie.text_size is not None:
529
if ie.text_size != fp['size']:
530
raise errors.BzrError(
531
"mismatched size for file %r in %r" %
532
(ie.file_id, self._store),
533
["inventory expects %d bytes" % ie.text_size,
534
"file is actually %d bytes" % fp['size'],
535
"store is probably damaged/corrupt"])
537
if ie.text_sha1 != fp['sha1']:
538
raise errors.BzrError("wrong SHA-1 for file %r in %r" %
539
(ie.file_id, self._store),
540
["inventory expects %s" % ie.text_sha1,
541
"file is actually %s" % fp['sha1'],
542
"store is probably damaged/corrupt"])
545
def path2id(self, path):
546
"""Return the id for path in this tree."""
547
return self._inventory.path2id(path)
549
def paths2ids(self, paths, trees=[], require_versioned=True):
550
"""Return all the ids that can be reached by walking from paths.
552
Each path is looked up in this tree and any extras provided in
553
trees, and this is repeated recursively: the children in an extra tree
554
of a directory that has been renamed under a provided path in this tree
555
are all returned, even if none exist under a provided path in this
556
tree, and vice versa.
558
:param paths: An iterable of paths to start converting to ids from.
559
Alternatively, if paths is None, no ids should be calculated and None
560
will be returned. This is offered to make calling the api unconditional
561
for code that *might* take a list of files.
562
:param trees: Additional trees to consider.
563
:param require_versioned: If False, do not raise NotVersionedError if
564
an element of paths is not versioned in this tree and all of trees.
566
return find_ids_across_trees(paths, [self] + list(trees), require_versioned)
568
def iter_children(self, file_id):
569
entry = self.iter_entries_by_dir([file_id]).next()[1]
570
for child in getattr(entry, 'children', {}).itervalues():
574
"""Lock this tree for multiple read only operations.
576
:return: A bzrlib.lock.LogicalLockResult.
580
def revision_tree(self, revision_id):
581
"""Obtain a revision tree for the revision revision_id.
583
The intention of this method is to allow access to possibly cached
584
tree data. Implementors of this method should raise NoSuchRevision if
585
the tree is not locally available, even if they could obtain the
586
tree via a repository or some other means. Callers are responsible
587
for finding the ultimate source for a revision tree.
589
:param revision_id: The revision_id of the requested tree.
591
:raises: NoSuchRevision if the tree cannot be obtained.
593
raise errors.NoSuchRevisionInTree(self, revision_id)
596
"""What files are present in this tree and unknown.
598
:return: an iterator over the unknown files.
605
def filter_unversioned_files(self, paths):
606
"""Filter out paths that are versioned.
608
:return: set of paths.
610
# NB: we specifically *don't* call self.has_filename, because for
611
# WorkingTrees that can indicate files that exist on disk but that
613
pred = self.inventory.has_filename
614
return set((p for p in paths if not pred(p)))
616
def walkdirs(self, prefix=""):
617
"""Walk the contents of this tree from path down.
619
This yields all the data about the contents of a directory at a time.
620
After each directory has been yielded, if the caller has mutated the
621
list to exclude some directories, they are then not descended into.
623
The data yielded is of the form:
624
((directory-relpath, directory-path-from-root, directory-fileid),
625
[(relpath, basename, kind, lstat, path_from_tree_root, file_id,
626
versioned_kind), ...]),
627
- directory-relpath is the containing dirs relpath from prefix
628
- directory-path-from-root is the containing dirs path from /
629
- directory-fileid is the id of the directory if it is versioned.
630
- relpath is the relative path within the subtree being walked.
631
- basename is the basename
632
- kind is the kind of the file now. If unknonwn then the file is not
633
present within the tree - but it may be recorded as versioned. See
635
- lstat is the stat data *if* the file was statted.
636
- path_from_tree_root is the path from the root of the tree.
637
- file_id is the file_id if the entry is versioned.
638
- versioned_kind is the kind of the file as last recorded in the
639
versioning system. If 'unknown' the file is not versioned.
640
One of 'kind' and 'versioned_kind' must not be 'unknown'.
642
:param prefix: Start walking from prefix within the tree rather than
643
at the root. This allows one to walk a subtree but get paths that are
644
relative to a tree rooted higher up.
645
:return: an iterator over the directory data.
647
raise NotImplementedError(self.walkdirs)
649
def supports_content_filtering(self):
652
def _content_filter_stack(self, path=None, file_id=None):
653
"""The stack of content filters for a path if filtering is supported.
655
Readers will be applied in first-to-last order.
656
Writers will be applied in last-to-first order.
657
Either the path or the file-id needs to be provided.
659
:param path: path relative to the root of the tree
661
:param file_id: file_id or None if unknown
662
:return: the list of filters - [] if there are none
664
filter_pref_names = filters._get_registered_names()
665
if len(filter_pref_names) == 0:
668
path = self.id2path(file_id)
669
prefs = self.iter_search_rules([path], filter_pref_names).next()
670
stk = filters._get_filter_stack_for(prefs)
671
if 'filters' in debug.debug_flags:
672
note("*** %s content-filter: %s => %r" % (path,prefs,stk))
675
def _content_filter_stack_provider(self):
676
"""A function that returns a stack of ContentFilters.
678
The function takes a path (relative to the top of the tree) and a
679
file-id as parameters.
681
:return: None if content filtering is not supported by this tree.
683
if self.supports_content_filtering():
684
return lambda path, file_id: \
685
self._content_filter_stack(path, file_id)
689
def iter_search_rules(self, path_names, pref_names=None,
690
_default_searcher=None):
691
"""Find the preferences for filenames in a tree.
693
:param path_names: an iterable of paths to find attributes for.
694
Paths are given relative to the root of the tree.
695
:param pref_names: the list of preferences to lookup - None for all
696
:param _default_searcher: private parameter to assist testing - don't use
697
:return: an iterator of tuple sequences, one per path-name.
698
See _RulesSearcher.get_items for details on the tuple sequence.
700
if _default_searcher is None:
701
_default_searcher = rules._per_user_searcher
702
searcher = self._get_rules_searcher(_default_searcher)
703
if searcher is not None:
704
if pref_names is not None:
705
for path in path_names:
706
yield searcher.get_selected_items(path, pref_names)
708
for path in path_names:
709
yield searcher.get_items(path)
711
def _get_rules_searcher(self, default_searcher):
712
"""Get the RulesSearcher for this tree given the default one."""
713
searcher = default_searcher
717
######################################################################
720
# TODO: Merge these two functions into a single one that can operate
721
# on either a whole tree or a set of files.
723
# TODO: Return the diff in order by filename, not by category or in
724
# random order. Can probably be done by lock-stepping through the
725
# filenames from both trees.
728
def file_status(filename, old_tree, new_tree):
729
"""Return single-letter status, old and new names for a file.
731
The complexity here is in deciding how to represent renames;
732
many complex cases are possible.
734
old_inv = old_tree.inventory
735
new_inv = new_tree.inventory
736
new_id = new_inv.path2id(filename)
737
old_id = old_inv.path2id(filename)
739
if not new_id and not old_id:
740
# easy: doesn't exist in either; not versioned at all
741
if new_tree.is_ignored(filename):
742
return 'I', None, None
744
return '?', None, None
746
# There is now a file of this name, great.
749
# There is no longer a file of this name, but we can describe
750
# what happened to the file that used to have
751
# this name. There are two possibilities: either it was
752
# deleted entirely, or renamed.
753
if new_inv.has_id(old_id):
754
return 'X', old_inv.id2path(old_id), new_inv.id2path(old_id)
756
return 'D', old_inv.id2path(old_id), None
758
# if the file_id is new in this revision, it is added
759
if new_id and not old_inv.has_id(new_id):
762
# if there used to be a file of this name, but that ID has now
763
# disappeared, it is deleted
764
if old_id and not new_inv.has_id(old_id):
770
def find_ids_across_trees(filenames, trees, require_versioned=True):
771
"""Find the ids corresponding to specified filenames.
773
All matches in all trees will be used, and all children of matched
774
directories will be used.
776
:param filenames: The filenames to find file_ids for (if None, returns
778
:param trees: The trees to find file_ids within
779
:param require_versioned: if true, all specified filenames must occur in
781
:return: a set of file ids for the specified filenames and their children.
785
specified_path_ids = _find_ids_across_trees(filenames, trees,
787
return _find_children_across_trees(specified_path_ids, trees)
790
def _find_ids_across_trees(filenames, trees, require_versioned):
791
"""Find the ids corresponding to specified filenames.
793
All matches in all trees will be used, but subdirectories are not scanned.
795
:param filenames: The filenames to find file_ids for
796
:param trees: The trees to find file_ids within
797
:param require_versioned: if true, all specified filenames must occur in
799
:return: a set of file ids for the specified filenames
802
interesting_ids = set()
803
for tree_path in filenames:
806
file_id = tree.path2id(tree_path)
807
if file_id is not None:
808
interesting_ids.add(file_id)
811
not_versioned.append(tree_path)
812
if len(not_versioned) > 0 and require_versioned:
813
raise errors.PathsNotVersionedError(not_versioned)
814
return interesting_ids
817
def _find_children_across_trees(specified_ids, trees):
818
"""Return a set including specified ids and their children.
820
All matches in all trees will be used.
822
:param trees: The trees to find file_ids within
823
:return: a set containing all specified ids and their children
825
interesting_ids = set(specified_ids)
826
pending = interesting_ids
827
# now handle children of interesting ids
828
# we loop so that we handle all children of each id in both trees
829
while len(pending) > 0:
831
for file_id in pending:
833
if not tree.has_or_had_id(file_id):
835
for child_id in tree.iter_children(file_id):
836
if child_id not in interesting_ids:
837
new_pending.add(child_id)
838
interesting_ids.update(new_pending)
839
pending = new_pending
840
return interesting_ids
843
class InterTree(InterObject):
844
"""This class represents operations taking place between two Trees.
846
Its instances have methods like 'compare' and contain references to the
847
source and target trees these operations are to be carried out on.
849
Clients of bzrlib should not need to use InterTree directly, rather they
850
should use the convenience methods on Tree such as 'Tree.compare()' which
851
will pass through to InterTree as appropriate.
854
# Formats that will be used to test this InterTree. If both are
855
# None, this InterTree will not be tested (e.g. because a complex
857
_matching_from_tree_format = None
858
_matching_to_tree_format = None
862
def _changes_from_entries(self, source_entry, target_entry,
863
source_path=None, target_path=None):
864
"""Generate a iter_changes tuple between source_entry and target_entry.
866
:param source_entry: An inventory entry from self.source, or None.
867
:param target_entry: An inventory entry from self.target, or None.
868
:param source_path: The path of source_entry, if known. If not known
869
it will be looked up.
870
:param target_path: The path of target_entry, if known. If not known
871
it will be looked up.
872
:return: A tuple, item 0 of which is an iter_changes result tuple, and
873
item 1 is True if there are any changes in the result tuple.
875
if source_entry is None:
876
if target_entry is None:
878
file_id = target_entry.file_id
880
file_id = source_entry.file_id
881
if source_entry is not None:
882
source_versioned = True
883
source_name = source_entry.name
884
source_parent = source_entry.parent_id
885
if source_path is None:
886
source_path = self.source.id2path(file_id)
887
source_kind, source_executable, source_stat = \
888
self.source._comparison_data(source_entry, source_path)
890
source_versioned = False
894
source_executable = None
895
if target_entry is not None:
896
target_versioned = True
897
target_name = target_entry.name
898
target_parent = target_entry.parent_id
899
if target_path is None:
900
target_path = self.target.id2path(file_id)
901
target_kind, target_executable, target_stat = \
902
self.target._comparison_data(target_entry, target_path)
904
target_versioned = False
908
target_executable = None
909
versioned = (source_versioned, target_versioned)
910
kind = (source_kind, target_kind)
911
changed_content = False
912
if source_kind != target_kind:
913
changed_content = True
914
elif source_kind == 'file':
915
if (self.source.get_file_sha1(file_id, source_path, source_stat) !=
916
self.target.get_file_sha1(file_id, target_path, target_stat)):
917
changed_content = True
918
elif source_kind == 'symlink':
919
if (self.source.get_symlink_target(file_id) !=
920
self.target.get_symlink_target(file_id)):
921
changed_content = True
922
# XXX: Yes, the indentation below is wrong. But fixing it broke
923
# test_merge.TestMergerEntriesLCAOnDisk.
924
# test_nested_tree_subtree_renamed_and_modified. We'll wait for
925
# the fix from bzr.dev -- vila 2009026
926
elif source_kind == 'tree-reference':
927
if (self.source.get_reference_revision(file_id, source_path)
928
!= self.target.get_reference_revision(file_id, target_path)):
929
changed_content = True
930
parent = (source_parent, target_parent)
931
name = (source_name, target_name)
932
executable = (source_executable, target_executable)
933
if (changed_content is not False or versioned[0] != versioned[1]
934
or parent[0] != parent[1] or name[0] != name[1] or
935
executable[0] != executable[1]):
939
return (file_id, (source_path, target_path), changed_content,
940
versioned, parent, name, kind, executable), changes
943
def compare(self, want_unchanged=False, specific_files=None,
944
extra_trees=None, require_versioned=False, include_root=False,
945
want_unversioned=False):
946
"""Return the changes from source to target.
948
:return: A TreeDelta.
949
:param specific_files: An optional list of file paths to restrict the
950
comparison to. When mapping filenames to ids, all matches in all
951
trees (including optional extra_trees) are used, and all children of
952
matched directories are included.
953
:param want_unchanged: An optional boolean requesting the inclusion of
954
unchanged entries in the result.
955
:param extra_trees: An optional list of additional trees to use when
956
mapping the contents of specific_files (paths) to file_ids.
957
:param require_versioned: An optional boolean (defaults to False). When
958
supplied and True all the 'specific_files' must be versioned, or
959
a PathsNotVersionedError will be thrown.
960
:param want_unversioned: Scan for unversioned paths.
962
trees = (self.source,)
963
if extra_trees is not None:
964
trees = trees + tuple(extra_trees)
965
# target is usually the newer tree:
966
specific_file_ids = self.target.paths2ids(specific_files, trees,
967
require_versioned=require_versioned)
968
if specific_files and not specific_file_ids:
969
# All files are unversioned, so just return an empty delta
970
# _compare_trees would think we want a complete delta
971
result = delta.TreeDelta()
972
fake_entry = InventoryFile('unused', 'unused', 'unused')
973
result.unversioned = [(path, None,
974
self.target._comparison_data(fake_entry, path)[0]) for path in
977
return delta._compare_trees(self.source, self.target, want_unchanged,
978
specific_files, include_root, extra_trees=extra_trees,
979
require_versioned=require_versioned,
980
want_unversioned=want_unversioned)
982
def iter_changes(self, include_unchanged=False,
983
specific_files=None, pb=None, extra_trees=[],
984
require_versioned=True, want_unversioned=False):
985
"""Generate an iterator of changes between trees.
988
(file_id, (path_in_source, path_in_target),
989
changed_content, versioned, parent, name, kind,
992
Changed_content is True if the file's content has changed. This
993
includes changes to its kind, and to a symlink's target.
995
versioned, parent, name, kind, executable are tuples of (from, to).
996
If a file is missing in a tree, its kind is None.
998
Iteration is done in parent-to-child order, relative to the target
1001
There is no guarantee that all paths are in sorted order: the
1002
requirement to expand the search due to renames may result in children
1003
that should be found early being found late in the search, after
1004
lexically later results have been returned.
1005
:param require_versioned: Raise errors.PathsNotVersionedError if a
1006
path in the specific_files list is not versioned in one of
1007
source, target or extra_trees.
1008
:param specific_files: An optional list of file paths to restrict the
1009
comparison to. When mapping filenames to ids, all matches in all
1010
trees (including optional extra_trees) are used, and all children
1011
of matched directories are included. The parents in the target tree
1012
of the specific files up to and including the root of the tree are
1013
always evaluated for changes too.
1014
:param want_unversioned: Should unversioned files be returned in the
1015
output. An unversioned file is defined as one with (False, False)
1016
for the versioned pair.
1018
lookup_trees = [self.source]
1020
lookup_trees.extend(extra_trees)
1021
# The ids of items we need to examine to insure delta consistency.
1022
precise_file_ids = set()
1023
changed_file_ids = []
1024
if specific_files == []:
1025
specific_file_ids = []
1027
specific_file_ids = self.target.paths2ids(specific_files,
1028
lookup_trees, require_versioned=require_versioned)
1029
if specific_files is not None:
1030
# reparented or added entries must have their parents included
1031
# so that valid deltas can be created. The seen_parents set
1032
# tracks the parents that we need to have.
1033
# The seen_dirs set tracks directory entries we've yielded.
1034
# After outputting version object in to_entries we set difference
1035
# the two seen sets and start checking parents.
1036
seen_parents = set()
1038
if want_unversioned:
1039
all_unversioned = sorted([(p.split('/'), p) for p in
1040
self.target.extras()
1041
if specific_files is None or
1042
osutils.is_inside_any(specific_files, p)])
1043
all_unversioned = deque(all_unversioned)
1045
all_unversioned = deque()
1047
from_entries_by_dir = list(self.source.iter_entries_by_dir(
1048
specific_file_ids=specific_file_ids))
1049
from_data = dict((e.file_id, (p, e)) for p, e in from_entries_by_dir)
1050
to_entries_by_dir = list(self.target.iter_entries_by_dir(
1051
specific_file_ids=specific_file_ids))
1052
num_entries = len(from_entries_by_dir) + len(to_entries_by_dir)
1054
# the unversioned path lookup only occurs on real trees - where there
1055
# can be extras. So the fake_entry is solely used to look up
1056
# executable it values when execute is not supported.
1057
fake_entry = InventoryFile('unused', 'unused', 'unused')
1058
for target_path, target_entry in to_entries_by_dir:
1059
while (all_unversioned and
1060
all_unversioned[0][0] < target_path.split('/')):
1061
unversioned_path = all_unversioned.popleft()
1062
target_kind, target_executable, target_stat = \
1063
self.target._comparison_data(fake_entry, unversioned_path[1])
1064
yield (None, (None, unversioned_path[1]), True, (False, False),
1066
(None, unversioned_path[0][-1]),
1067
(None, target_kind),
1068
(None, target_executable))
1069
source_path, source_entry = from_data.get(target_entry.file_id,
1071
result, changes = self._changes_from_entries(source_entry,
1072
target_entry, source_path=source_path, target_path=target_path)
1073
to_paths[result[0]] = result[1][1]
1078
pb.update('comparing files', entry_count, num_entries)
1079
if changes or include_unchanged:
1080
if specific_file_ids is not None:
1081
new_parent_id = result[4][1]
1082
precise_file_ids.add(new_parent_id)
1083
changed_file_ids.append(result[0])
1085
# Ensure correct behaviour for reparented/added specific files.
1086
if specific_files is not None:
1087
# Record output dirs
1088
if result[6][1] == 'directory':
1089
seen_dirs.add(result[0])
1090
# Record parents of reparented/added entries.
1091
versioned = result[3]
1093
if not versioned[0] or parents[0] != parents[1]:
1094
seen_parents.add(parents[1])
1095
while all_unversioned:
1096
# yield any trailing unversioned paths
1097
unversioned_path = all_unversioned.popleft()
1098
to_kind, to_executable, to_stat = \
1099
self.target._comparison_data(fake_entry, unversioned_path[1])
1100
yield (None, (None, unversioned_path[1]), True, (False, False),
1102
(None, unversioned_path[0][-1]),
1104
(None, to_executable))
1105
# Yield all remaining source paths
1106
for path, from_entry in from_entries_by_dir:
1107
file_id = from_entry.file_id
1108
if file_id in to_paths:
1111
if not self.target.has_id(file_id):
1112
# common case - paths we have not emitted are not present in
1116
to_path = self.target.id2path(file_id)
1119
pb.update('comparing files', entry_count, num_entries)
1120
versioned = (True, False)
1121
parent = (from_entry.parent_id, None)
1122
name = (from_entry.name, None)
1123
from_kind, from_executable, stat_value = \
1124
self.source._comparison_data(from_entry, path)
1125
kind = (from_kind, None)
1126
executable = (from_executable, None)
1127
changed_content = from_kind is not None
1128
# the parent's path is necessarily known at this point.
1129
changed_file_ids.append(file_id)
1130
yield(file_id, (path, to_path), changed_content, versioned, parent,
1131
name, kind, executable)
1132
changed_file_ids = set(changed_file_ids)
1133
if specific_file_ids is not None:
1134
for result in self._handle_precise_ids(precise_file_ids,
1138
def _get_entry(self, tree, file_id):
1139
"""Get an inventory entry from a tree, with missing entries as None.
1141
If the tree raises NotImplementedError on accessing .inventory, then
1142
this is worked around using iter_entries_by_dir on just the file id
1145
:param tree: The tree to lookup the entry in.
1146
:param file_id: The file_id to lookup.
1149
inventory = tree.inventory
1150
except NotImplementedError:
1151
# No inventory available.
1153
iterator = tree.iter_entries_by_dir(specific_file_ids=[file_id])
1154
return iterator.next()[1]
1155
except StopIteration:
1159
return inventory[file_id]
1160
except errors.NoSuchId:
1163
def _handle_precise_ids(self, precise_file_ids, changed_file_ids,
1164
discarded_changes=None):
1165
"""Fill out a partial iter_changes to be consistent.
1167
:param precise_file_ids: The file ids of parents that were seen during
1169
:param changed_file_ids: The file ids of already emitted items.
1170
:param discarded_changes: An optional dict of precalculated
1171
iter_changes items which the partial iter_changes had not output
1173
:return: A generator of iter_changes items to output.
1175
# process parents of things that had changed under the users
1176
# requested paths to prevent incorrect paths or parent ids which
1177
# aren't in the tree.
1178
while precise_file_ids:
1179
precise_file_ids.discard(None)
1180
# Don't emit file_ids twice
1181
precise_file_ids.difference_update(changed_file_ids)
1182
if not precise_file_ids:
1184
# If the there was something at a given output path in source, we
1185
# have to include the entry from source in the delta, or we would
1186
# be putting this entry into a used path.
1188
for parent_id in precise_file_ids:
1190
paths.append(self.target.id2path(parent_id))
1191
except errors.NoSuchId:
1192
# This id has been dragged in from the source by delta
1193
# expansion and isn't present in target at all: we don't
1194
# need to check for path collisions on it.
1197
old_id = self.source.path2id(path)
1198
precise_file_ids.add(old_id)
1199
precise_file_ids.discard(None)
1200
current_ids = precise_file_ids
1201
precise_file_ids = set()
1202
# We have to emit all of precise_file_ids that have been altered.
1203
# We may have to output the children of some of those ids if any
1204
# directories have stopped being directories.
1205
for file_id in current_ids:
1207
if discarded_changes:
1208
result = discarded_changes.get(file_id)
1213
old_entry = self._get_entry(self.source, file_id)
1214
new_entry = self._get_entry(self.target, file_id)
1215
result, changes = self._changes_from_entries(
1216
old_entry, new_entry)
1219
# Get this parents parent to examine.
1220
new_parent_id = result[4][1]
1221
precise_file_ids.add(new_parent_id)
1223
if (result[6][0] == 'directory' and
1224
result[6][1] != 'directory'):
1225
# This stopped being a directory, the old children have
1227
if old_entry is None:
1228
# Reusing a discarded change.
1229
old_entry = self._get_entry(self.source, file_id)
1230
for child in old_entry.children.values():
1231
precise_file_ids.add(child.file_id)
1232
changed_file_ids.add(result[0])
1236
class MultiWalker(object):
1237
"""Walk multiple trees simultaneously, getting combined results."""
1239
# Note: This could be written to not assume you can do out-of-order
1240
# lookups. Instead any nodes that don't match in all trees could be
1241
# marked as 'deferred', and then returned in the final cleanup loop.
1242
# For now, I think it is "nicer" to return things as close to the
1243
# "master_tree" order as we can.
1245
def __init__(self, master_tree, other_trees):
1246
"""Create a new MultiWalker.
1248
All trees being walked must implement "iter_entries_by_dir()", such
1249
that they yield (path, object) tuples, where that object will have a
1250
'.file_id' member, that can be used to check equality.
1252
:param master_tree: All trees will be 'slaved' to the master_tree such
1253
that nodes in master_tree will be used as 'first-pass' sync points.
1254
Any nodes that aren't in master_tree will be merged in a second
1256
:param other_trees: A list of other trees to walk simultaneously.
1258
self._master_tree = master_tree
1259
self._other_trees = other_trees
1261
# Keep track of any nodes that were properly processed just out of
1262
# order, that way we don't return them at the end, we don't have to
1263
# track *all* processed file_ids, just the out-of-order ones
1264
self._out_of_order_processed = set()
1267
def _step_one(iterator):
1268
"""Step an iter_entries_by_dir iterator.
1270
:return: (has_more, path, ie)
1271
If has_more is False, path and ie will be None.
1274
path, ie = iterator.next()
1275
except StopIteration:
1276
return False, None, None
1278
return True, path, ie
1281
def _cmp_path_by_dirblock(path1, path2):
1282
"""Compare two paths based on what directory they are in.
1284
This generates a sort order, such that all children of a directory are
1285
sorted together, and grandchildren are in the same order as the
1286
children appear. But all grandchildren come after all children.
1288
:param path1: first path
1289
:param path2: the second path
1290
:return: negative number if ``path1`` comes first,
1291
0 if paths are equal
1292
and a positive number if ``path2`` sorts first
1294
# Shortcut this special case
1297
# This is stolen from _dirstate_helpers_py.py, only switching it to
1298
# Unicode objects. Consider using encode_utf8() and then using the
1299
# optimized versions, or maybe writing optimized unicode versions.
1300
if not isinstance(path1, unicode):
1301
raise TypeError("'path1' must be a unicode string, not %s: %r"
1302
% (type(path1), path1))
1303
if not isinstance(path2, unicode):
1304
raise TypeError("'path2' must be a unicode string, not %s: %r"
1305
% (type(path2), path2))
1306
return cmp(MultiWalker._path_to_key(path1),
1307
MultiWalker._path_to_key(path2))
1310
def _path_to_key(path):
1311
dirname, basename = osutils.split(path)
1312
return (dirname.split(u'/'), basename)
1314
def _lookup_by_file_id(self, extra_entries, other_tree, file_id):
1315
"""Lookup an inventory entry by file_id.
1317
This is called when an entry is missing in the normal order.
1318
Generally this is because a file was either renamed, or it was
1319
deleted/added. If the entry was found in the inventory and not in
1320
extra_entries, it will be added to self._out_of_order_processed
1322
:param extra_entries: A dictionary of {file_id: (path, ie)}. This
1323
should be filled with entries that were found before they were
1324
used. If file_id is present, it will be removed from the
1326
:param other_tree: The Tree to search, in case we didn't find the entry
1328
:param file_id: The file_id to look for
1329
:return: (path, ie) if found or (None, None) if not present.
1331
if file_id in extra_entries:
1332
return extra_entries.pop(file_id)
1333
# TODO: Is id2path better as the first call, or is
1334
# inventory[file_id] better as a first check?
1336
cur_path = other_tree.id2path(file_id)
1337
except errors.NoSuchId:
1339
if cur_path is None:
1342
self._out_of_order_processed.add(file_id)
1343
cur_ie = other_tree.inventory[file_id]
1344
return (cur_path, cur_ie)
1347
"""Match up the values in the different trees."""
1348
for result in self._walk_master_tree():
1350
self._finish_others()
1351
for result in self._walk_others():
1354
def _walk_master_tree(self):
1355
"""First pass, walk all trees in lock-step.
1357
When we are done, all nodes in the master_tree will have been
1358
processed. _other_walkers, _other_entries, and _others_extra will be
1359
set on 'self' for future processing.
1361
# This iterator has the most "inlining" done, because it tends to touch
1362
# every file in the tree, while the others only hit nodes that don't
1364
master_iterator = self._master_tree.iter_entries_by_dir()
1366
other_walkers = [other.iter_entries_by_dir()
1367
for other in self._other_trees]
1368
other_entries = [self._step_one(walker) for walker in other_walkers]
1369
# Track extra nodes in the other trees
1370
others_extra = [{} for i in xrange(len(self._other_trees))]
1372
master_has_more = True
1373
step_one = self._step_one
1374
lookup_by_file_id = self._lookup_by_file_id
1375
out_of_order_processed = self._out_of_order_processed
1377
while master_has_more:
1378
(master_has_more, path, master_ie) = step_one(master_iterator)
1379
if not master_has_more:
1382
file_id = master_ie.file_id
1384
other_values_append = other_values.append
1385
next_other_entries = []
1386
next_other_entries_append = next_other_entries.append
1387
for idx, (other_has_more, other_path, other_ie) in enumerate(other_entries):
1388
if not other_has_more:
1389
other_values_append(lookup_by_file_id(
1390
others_extra[idx], self._other_trees[idx], file_id))
1391
next_other_entries_append((False, None, None))
1392
elif file_id == other_ie.file_id:
1393
# This is the critical code path, as most of the entries
1394
# should match between most trees.
1395
other_values_append((other_path, other_ie))
1396
next_other_entries_append(step_one(other_walkers[idx]))
1398
# This walker did not match, step it until it either
1399
# matches, or we know we are past the current walker.
1400
other_walker = other_walkers[idx]
1401
other_extra = others_extra[idx]
1402
while (other_has_more and
1403
self._cmp_path_by_dirblock(other_path, path) < 0):
1404
other_file_id = other_ie.file_id
1405
if other_file_id not in out_of_order_processed:
1406
other_extra[other_file_id] = (other_path, other_ie)
1407
other_has_more, other_path, other_ie = \
1408
step_one(other_walker)
1409
if other_has_more and other_ie.file_id == file_id:
1410
# We ended up walking to this point, match and step
1412
other_values_append((other_path, other_ie))
1413
other_has_more, other_path, other_ie = \
1414
step_one(other_walker)
1416
# This record isn't in the normal order, see if it
1418
other_values_append(lookup_by_file_id(
1419
other_extra, self._other_trees[idx], file_id))
1420
next_other_entries_append((other_has_more, other_path,
1422
other_entries = next_other_entries
1424
# We've matched all the walkers, yield this datapoint
1425
yield path, file_id, master_ie, other_values
1426
self._other_walkers = other_walkers
1427
self._other_entries = other_entries
1428
self._others_extra = others_extra
1430
def _finish_others(self):
1431
"""Finish walking the other iterators, so we get all entries."""
1432
for idx, info in enumerate(self._other_entries):
1433
other_extra = self._others_extra[idx]
1434
(other_has_more, other_path, other_ie) = info
1435
while other_has_more:
1436
other_file_id = other_ie.file_id
1437
if other_file_id not in self._out_of_order_processed:
1438
other_extra[other_file_id] = (other_path, other_ie)
1439
other_has_more, other_path, other_ie = \
1440
self._step_one(self._other_walkers[idx])
1441
del self._other_entries
1443
def _walk_others(self):
1444
"""Finish up by walking all the 'deferred' nodes."""
1445
# TODO: One alternative would be to grab all possible unprocessed
1446
# file_ids, and then sort by path, and then yield them. That
1447
# might ensure better ordering, in case a caller strictly
1448
# requires parents before children.
1449
for idx, other_extra in enumerate(self._others_extra):
1450
others = sorted(other_extra.itervalues(),
1451
key=lambda x: self._path_to_key(x[0]))
1452
for other_path, other_ie in others:
1453
file_id = other_ie.file_id
1454
# We don't need to check out_of_order_processed here, because
1455
# the lookup_by_file_id will be removing anything processed
1456
# from the extras cache
1457
other_extra.pop(file_id)
1458
other_values = [(None, None) for i in xrange(idx)]
1459
other_values.append((other_path, other_ie))
1460
for alt_idx, alt_extra in enumerate(self._others_extra[idx+1:]):
1461
alt_idx = alt_idx + idx + 1
1462
alt_extra = self._others_extra[alt_idx]
1463
alt_tree = self._other_trees[alt_idx]
1464
other_values.append(self._lookup_by_file_id(
1465
alt_extra, alt_tree, file_id))
1466
yield other_path, file_id, None, other_values
added
removed
bzrlib/tree.py
