1
# Copyright (C) 2005-2010 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,
30
revision as _mod_revision,
33
from bzrlib.decorators import needs_read_lock
34
from bzrlib.errors import BzrError, NoSuchId
35
from bzrlib import errors
36
from bzrlib.inventory import InventoryFile
37
from bzrlib.inter import InterObject
38
from bzrlib.osutils import fingerprint_file
39
from bzrlib.symbol_versioning import deprecated_function, deprecated_in
40
from bzrlib.trace import note
44
"""Abstract file tree.
46
There are several subclasses:
48
* `WorkingTree` exists as files on disk editable by the user.
50
* `RevisionTree` is a tree as recorded at some point in the past.
52
Trees contain an `Inventory` object, and also know how to retrieve
53
file texts mentioned in the inventory, either from a working
54
directory or from a store.
56
It is possible for trees to contain files that are not described
57
in their inventory or vice versa; for this use `filenames()`.
59
Trees can be compared, etc, regardless of whether they are working
60
trees or versioned trees.
63
def changes_from(self, other, want_unchanged=False, specific_files=None,
64
extra_trees=None, require_versioned=False, include_root=False,
65
want_unversioned=False):
66
"""Return a TreeDelta of the changes from other to this tree.
68
:param other: A tree to compare with.
69
:param specific_files: An optional list of file paths to restrict the
70
comparison to. When mapping filenames to ids, all matches in all
71
trees (including optional extra_trees) are used, and all children of
72
matched directories are included.
73
:param want_unchanged: An optional boolean requesting the inclusion of
74
unchanged entries in the result.
75
:param extra_trees: An optional list of additional trees to use when
76
mapping the contents of specific_files (paths) to file_ids.
77
:param require_versioned: An optional boolean (defaults to False). When
78
supplied and True all the 'specific_files' must be versioned, or
79
a PathsNotVersionedError will be thrown.
80
:param want_unversioned: Scan for unversioned paths.
82
The comparison will be performed by an InterTree object looked up on
85
# Martin observes that Tree.changes_from returns a TreeDelta and this
86
# may confuse people, because the class name of the returned object is
87
# a synonym of the object referenced in the method name.
88
return InterTree.get(other, self).compare(
89
want_unchanged=want_unchanged,
90
specific_files=specific_files,
91
extra_trees=extra_trees,
92
require_versioned=require_versioned,
93
include_root=include_root,
94
want_unversioned=want_unversioned,
97
def iter_changes(self, from_tree, include_unchanged=False,
98
specific_files=None, pb=None, extra_trees=None,
99
require_versioned=True, want_unversioned=False):
100
"""See InterTree.iter_changes"""
101
intertree = InterTree.get(from_tree, self)
102
return intertree.iter_changes(include_unchanged, specific_files, pb,
103
extra_trees, require_versioned, want_unversioned=want_unversioned)
106
"""Get a list of the conflicts in the tree.
108
Each conflict is an instance of bzrlib.conflicts.Conflict.
110
return _mod_conflicts.ConflictList()
113
"""For trees that can have unversioned files, return all such paths."""
116
def get_parent_ids(self):
117
"""Get the parent ids for this tree.
119
:return: a list of parent ids. [] is returned to indicate
120
a tree with no parents.
121
:raises: BzrError if the parents are not known.
123
raise NotImplementedError(self.get_parent_ids)
125
def has_filename(self, filename):
126
"""True if the tree has given filename."""
127
raise NotImplementedError(self.has_filename)
129
def has_id(self, file_id):
130
return self.inventory.has_id(file_id)
132
def __contains__(self, file_id):
133
return self.has_id(file_id)
135
def has_or_had_id(self, file_id):
136
return self.inventory.has_id(file_id)
138
def is_ignored(self, filename):
139
"""Check whether the filename is ignored by this tree.
141
:param filename: The relative filename within the tree.
142
:return: True if the filename is ignored.
147
return iter(self.inventory)
149
def all_file_ids(self):
150
"""Iterate through all file ids, including ids for missing files."""
151
return set(self.inventory)
153
def id2path(self, file_id):
154
"""Return the path for a file id.
158
return self.inventory.id2path(file_id)
160
def is_control_filename(self, filename):
161
"""True if filename is the name of a control file in this tree.
163
:param filename: A filename within the tree. This is a relative path
164
from the root of this tree.
166
This is true IF and ONLY IF the filename is part of the meta data
167
that bzr controls in this tree. I.E. a random .bzr directory placed
168
on disk will not be a control file for this tree.
170
return self.bzrdir.is_control_filename(filename)
173
def iter_entries_by_dir(self, specific_file_ids=None, yield_parents=False):
174
"""Walk the tree in 'by_dir' order.
176
This will yield each entry in the tree as a (path, entry) tuple.
177
The order that they are yielded is:
179
Directories are walked in a depth-first lexicographical order,
180
however, whenever a directory is reached, all of its direct child
181
nodes are yielded in lexicographical order before yielding the
184
For example, in the tree::
194
The yield order (ignoring root) would be::
195
a, f, a/b, a/d, a/b/c, a/d/e, f/g
197
:param yield_parents: If True, yield the parents from the root leading
198
down to specific_file_ids that have been requested. This has no
199
impact if specific_file_ids is None.
201
return self.inventory.iter_entries_by_dir(
202
specific_file_ids=specific_file_ids, yield_parents=yield_parents)
204
def iter_references(self):
205
if self.supports_tree_reference():
206
for path, entry in self.iter_entries_by_dir():
207
if entry.kind == 'tree-reference':
208
yield path, entry.file_id
210
def kind(self, file_id):
211
raise NotImplementedError("Tree subclass %s must implement kind"
212
% self.__class__.__name__)
214
def stored_kind(self, file_id):
215
"""File kind stored for this file_id.
217
May not match kind on disk for working trees. Always available
218
for versioned files, even when the file itself is missing.
220
return self.kind(file_id)
222
def path_content_summary(self, path):
223
"""Get a summary of the information about path.
225
All the attributes returned are for the canonical form, not the
226
convenient form (if content filters are in use.)
228
:param path: A relative path within the tree.
229
:return: A tuple containing kind, size, exec, sha1-or-link.
230
Kind is always present (see tree.kind()).
231
size is present if kind is file and the size of the
232
canonical form can be cheaply determined, None otherwise.
233
exec is None unless kind is file and the platform supports the 'x'
235
sha1-or-link is the link target if kind is symlink, or the sha1 if
236
it can be obtained without reading the file.
238
raise NotImplementedError(self.path_content_summary)
240
def get_reference_revision(self, file_id, path=None):
241
raise NotImplementedError("Tree subclass %s must implement "
242
"get_reference_revision"
243
% self.__class__.__name__)
245
def _comparison_data(self, entry, path):
246
"""Return a tuple of kind, executable, stat_value for a file.
248
entry may be None if there is no inventory entry for the file, but
249
path must always be supplied.
251
kind is None if there is no file present (even if an inventory id is
252
present). executable is False for non-file entries.
254
raise NotImplementedError(self._comparison_data)
256
def _file_size(self, entry, stat_value):
257
raise NotImplementedError(self._file_size)
259
def _get_inventory(self):
260
return self._inventory
262
def get_file(self, file_id, path=None):
263
"""Return a file object for the file file_id in the tree.
265
If both file_id and path are defined, it is implementation defined as
266
to which one is used.
268
raise NotImplementedError(self.get_file)
270
def get_file_with_stat(self, file_id, path=None):
271
"""Get a file handle and stat object for file_id.
273
The default implementation returns (self.get_file, None) for backwards
276
:param file_id: The file id to read.
277
:param path: The path of the file, if it is known.
278
:return: A tuple (file_handle, stat_value_or_None). If the tree has
279
no stat facility, or need for a stat cache feedback during commit,
280
it may return None for the second element of the tuple.
282
return (self.get_file(file_id, path), None)
284
def get_file_text(self, file_id, path=None):
285
"""Return the byte content of a file.
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
my_file = self.get_file(file_id, path)
294
return my_file.read()
298
def get_file_lines(self, file_id, path=None):
299
"""Return the content of a file, as lines.
301
:param file_id: The file_id of the file.
302
:param path: The path of the file.
303
If both file_id and path are supplied, an implementation may use
306
return osutils.split_lines(self.get_file_text(file_id, path))
308
def get_file_mtime(self, file_id, path=None):
309
"""Return the modification time for a file.
311
:param file_id: The handle for this file.
312
:param path: The path that this file can be found at.
313
These must point to the same object.
315
raise NotImplementedError(self.get_file_mtime)
317
def get_file_size(self, file_id):
318
"""Return the size of a file in bytes.
320
This applies only to regular files. If invoked on directories or
321
symlinks, it will return None.
322
:param file_id: The file-id of the file
324
raise NotImplementedError(self.get_file_size)
326
def get_file_by_path(self, path):
327
return self.get_file(self._inventory.path2id(path), path)
329
def iter_files_bytes(self, desired_files):
330
"""Iterate through file contents.
332
Files will not necessarily be returned in the order they occur in
333
desired_files. No specific order is guaranteed.
335
Yields pairs of identifier, bytes_iterator. identifier is an opaque
336
value supplied by the caller as part of desired_files. It should
337
uniquely identify the file version in the caller's context. (Examples:
338
an index number or a TreeTransform trans_id.)
340
bytes_iterator is an iterable of bytestrings for the file. The
341
kind of iterable and length of the bytestrings are unspecified, but for
342
this implementation, it is a tuple containing a single bytestring with
343
the complete text of the file.
345
:param desired_files: a list of (file_id, identifier) pairs
347
for file_id, identifier in desired_files:
348
# We wrap the string in a tuple so that we can return an iterable
349
# of bytestrings. (Technically, a bytestring is also an iterable
350
# of bytestrings, but iterating through each character is not
352
cur_file = (self.get_file_text(file_id),)
353
yield identifier, cur_file
355
def get_symlink_target(self, file_id):
356
"""Get the target for a given file_id.
358
It is assumed that the caller already knows that file_id is referencing
360
:param file_id: Handle for the symlink entry.
361
:return: The path the symlink points to.
363
raise NotImplementedError(self.get_symlink_target)
365
def get_canonical_inventory_paths(self, paths):
366
"""Like get_canonical_inventory_path() but works on multiple items.
368
:param paths: A sequence of paths relative to the root of the tree.
369
:return: A list of paths, with each item the corresponding input path
370
adjusted to account for existing elements that match case
373
return list(self._yield_canonical_inventory_paths(paths))
375
def get_canonical_inventory_path(self, path):
376
"""Returns the first inventory item that case-insensitively matches path.
378
If a path matches exactly, it is returned. If no path matches exactly
379
but more than one path matches case-insensitively, it is implementation
380
defined which is returned.
382
If no path matches case-insensitively, the input path is returned, but
383
with as many path entries that do exist changed to their canonical
386
If you need to resolve many names from the same tree, you should
387
use get_canonical_inventory_paths() to avoid O(N) behaviour.
389
:param path: A paths relative to the root of the tree.
390
:return: The input path adjusted to account for existing elements
391
that match case insensitively.
393
return self._yield_canonical_inventory_paths([path]).next()
395
def _yield_canonical_inventory_paths(self, paths):
397
# First, if the path as specified exists exactly, just use it.
398
if self.path2id(path) is not None:
402
cur_id = self.get_root_id()
404
bit_iter = iter(path.split("/"))
408
for child in self.iter_children(cur_id):
410
# XXX: it seem like if the child is known to be in the
411
# tree, we shouldn't need to go from its id back to
412
# its path -- mbp 2010-02-11
414
# XXX: it seems like we could be more efficient
415
# by just directly looking up the original name and
416
# only then searching all children; also by not
417
# chopping paths so much. -- mbp 2010-02-11
418
child_base = os.path.basename(self.id2path(child))
419
if (child_base == elt):
420
# if we found an exact match, we can stop now; if
421
# we found an approximate match we need to keep
422
# searching because there might be an exact match
425
new_path = osutils.pathjoin(cur_path, child_base)
427
elif child_base.lower() == lelt:
429
new_path = osutils.pathjoin(cur_path, child_base)
431
# before a change is committed we can see this error...
436
# got to the end of this directory and no entries matched.
437
# Return what matched so far, plus the rest as specified.
438
cur_path = osutils.pathjoin(cur_path, elt, *list(bit_iter))
443
def get_root_id(self):
444
"""Return the file_id for the root of this tree."""
445
raise NotImplementedError(self.get_root_id)
447
def annotate_iter(self, file_id,
448
default_revision=_mod_revision.CURRENT_REVISION):
449
"""Return an iterator of revision_id, line tuples.
451
For working trees (and mutable trees in general), the special
452
revision_id 'current:' will be used for lines that are new in this
453
tree, e.g. uncommitted changes.
454
:param file_id: The file to produce an annotated version from
455
:param default_revision: For lines that don't match a basis, mark them
456
with this revision id. Not all implementations will make use of
459
raise NotImplementedError(self.annotate_iter)
461
def _get_plan_merge_data(self, file_id, other, base):
462
from bzrlib import versionedfile
463
vf = versionedfile._PlanMergeVersionedFile(file_id)
464
last_revision_a = self._get_file_revision(file_id, vf, 'this:')
465
last_revision_b = other._get_file_revision(file_id, vf, 'other:')
467
last_revision_base = None
469
last_revision_base = base._get_file_revision(file_id, vf, 'base:')
470
return vf, last_revision_a, last_revision_b, last_revision_base
472
def plan_file_merge(self, file_id, other, base=None):
473
"""Generate a merge plan based on annotations.
475
If the file contains uncommitted changes in this tree, they will be
476
attributed to the 'current:' pseudo-revision. If the file contains
477
uncommitted changes in the other tree, they will be assigned to the
478
'other:' pseudo-revision.
480
data = self._get_plan_merge_data(file_id, other, base)
481
vf, last_revision_a, last_revision_b, last_revision_base = data
482
return vf.plan_merge(last_revision_a, last_revision_b,
485
def plan_file_lca_merge(self, file_id, other, base=None):
486
"""Generate a merge plan based lca-newness.
488
If the file contains uncommitted changes in this tree, they will be
489
attributed to the 'current:' pseudo-revision. If the file contains
490
uncommitted changes in the other tree, they will be assigned to the
491
'other:' pseudo-revision.
493
data = self._get_plan_merge_data(file_id, other, base)
494
vf, last_revision_a, last_revision_b, last_revision_base = data
495
return vf.plan_lca_merge(last_revision_a, last_revision_b,
498
def _iter_parent_trees(self):
499
"""Iterate through parent trees, defaulting to Tree.revision_tree."""
500
for revision_id in self.get_parent_ids():
502
yield self.revision_tree(revision_id)
503
except errors.NoSuchRevisionInTree:
504
yield self.repository.revision_tree(revision_id)
507
def _file_revision(revision_tree, file_id):
508
"""Determine the revision associated with a file in a given tree."""
509
revision_tree.lock_read()
511
return revision_tree.inventory[file_id].revision
513
revision_tree.unlock()
515
def _get_file_revision(self, file_id, vf, tree_revision):
516
"""Ensure that file_id, tree_revision is in vf to plan the merge."""
518
if getattr(self, '_repository', None) is None:
519
last_revision = tree_revision
520
parent_keys = [(file_id, self._file_revision(t, file_id)) for t in
521
self._iter_parent_trees()]
522
vf.add_lines((file_id, last_revision), parent_keys,
523
self.get_file_lines(file_id))
524
repo = self.branch.repository
527
last_revision = self._file_revision(self, file_id)
528
base_vf = self._repository.texts
529
if base_vf not in vf.fallback_versionedfiles:
530
vf.fallback_versionedfiles.append(base_vf)
533
inventory = property(_get_inventory,
534
doc="Inventory of this Tree")
536
def _check_retrieved(self, ie, f):
539
fp = fingerprint_file(f)
542
if ie.text_size is not None:
543
if ie.text_size != fp['size']:
544
raise BzrError("mismatched size for file %r in %r" % (ie.file_id, self._store),
545
["inventory expects %d bytes" % ie.text_size,
546
"file is actually %d bytes" % fp['size'],
547
"store is probably damaged/corrupt"])
549
if ie.text_sha1 != fp['sha1']:
550
raise BzrError("wrong SHA-1 for file %r in %r" % (ie.file_id, self._store),
551
["inventory expects %s" % ie.text_sha1,
552
"file is actually %s" % fp['sha1'],
553
"store is probably damaged/corrupt"])
556
def path2id(self, path):
557
"""Return the id for path in this tree."""
558
return self._inventory.path2id(path)
560
def paths2ids(self, paths, trees=[], require_versioned=True):
561
"""Return all the ids that can be reached by walking from paths.
563
Each path is looked up in this tree and any extras provided in
564
trees, and this is repeated recursively: the children in an extra tree
565
of a directory that has been renamed under a provided path in this tree
566
are all returned, even if none exist under a provided path in this
567
tree, and vice versa.
569
:param paths: An iterable of paths to start converting to ids from.
570
Alternatively, if paths is None, no ids should be calculated and None
571
will be returned. This is offered to make calling the api unconditional
572
for code that *might* take a list of files.
573
:param trees: Additional trees to consider.
574
:param require_versioned: If False, do not raise NotVersionedError if
575
an element of paths is not versioned in this tree and all of trees.
577
return find_ids_across_trees(paths, [self] + list(trees), require_versioned)
579
def iter_children(self, file_id):
580
entry = self.iter_entries_by_dir([file_id]).next()[1]
581
for child in getattr(entry, 'children', {}).itervalues():
585
"""Lock this tree for multiple read only operations.
587
:return: A bzrlib.lock.LogicalLockResult.
591
def revision_tree(self, revision_id):
592
"""Obtain a revision tree for the revision revision_id.
594
The intention of this method is to allow access to possibly cached
595
tree data. Implementors of this method should raise NoSuchRevision if
596
the tree is not locally available, even if they could obtain the
597
tree via a repository or some other means. Callers are responsible
598
for finding the ultimate source for a revision tree.
600
:param revision_id: The revision_id of the requested tree.
602
:raises: NoSuchRevision if the tree cannot be obtained.
604
raise errors.NoSuchRevisionInTree(self, revision_id)
607
"""What files are present in this tree and unknown.
609
:return: an iterator over the unknown files.
616
def filter_unversioned_files(self, paths):
617
"""Filter out paths that are versioned.
619
:return: set of paths.
621
# NB: we specifically *don't* call self.has_filename, because for
622
# WorkingTrees that can indicate files that exist on disk but that
624
pred = self.inventory.has_filename
625
return set((p for p in paths if not pred(p)))
627
def walkdirs(self, prefix=""):
628
"""Walk the contents of this tree from path down.
630
This yields all the data about the contents of a directory at a time.
631
After each directory has been yielded, if the caller has mutated the
632
list to exclude some directories, they are then not descended into.
634
The data yielded is of the form:
635
((directory-relpath, directory-path-from-root, directory-fileid),
636
[(relpath, basename, kind, lstat, path_from_tree_root, file_id,
637
versioned_kind), ...]),
638
- directory-relpath is the containing dirs relpath from prefix
639
- directory-path-from-root is the containing dirs path from /
640
- directory-fileid is the id of the directory if it is versioned.
641
- relpath is the relative path within the subtree being walked.
642
- basename is the basename
643
- kind is the kind of the file now. If unknonwn then the file is not
644
present within the tree - but it may be recorded as versioned. See
646
- lstat is the stat data *if* the file was statted.
647
- path_from_tree_root is the path from the root of the tree.
648
- file_id is the file_id if the entry is versioned.
649
- versioned_kind is the kind of the file as last recorded in the
650
versioning system. If 'unknown' the file is not versioned.
651
One of 'kind' and 'versioned_kind' must not be 'unknown'.
653
:param prefix: Start walking from prefix within the tree rather than
654
at the root. This allows one to walk a subtree but get paths that are
655
relative to a tree rooted higher up.
656
:return: an iterator over the directory data.
658
raise NotImplementedError(self.walkdirs)
660
def supports_content_filtering(self):
663
def _content_filter_stack(self, path=None, file_id=None):
664
"""The stack of content filters for a path if filtering is supported.
666
Readers will be applied in first-to-last order.
667
Writers will be applied in last-to-first order.
668
Either the path or the file-id needs to be provided.
670
:param path: path relative to the root of the tree
672
:param file_id: file_id or None if unknown
673
:return: the list of filters - [] if there are none
675
filter_pref_names = filters._get_registered_names()
676
if len(filter_pref_names) == 0:
679
path = self.id2path(file_id)
680
prefs = self.iter_search_rules([path], filter_pref_names).next()
681
stk = filters._get_filter_stack_for(prefs)
682
if 'filters' in debug.debug_flags:
683
note("*** %s content-filter: %s => %r" % (path,prefs,stk))
686
def _content_filter_stack_provider(self):
687
"""A function that returns a stack of ContentFilters.
689
The function takes a path (relative to the top of the tree) and a
690
file-id as parameters.
692
:return: None if content filtering is not supported by this tree.
694
if self.supports_content_filtering():
695
return lambda path, file_id: \
696
self._content_filter_stack(path, file_id)
700
def iter_search_rules(self, path_names, pref_names=None,
701
_default_searcher=None):
702
"""Find the preferences for filenames in a tree.
704
:param path_names: an iterable of paths to find attributes for.
705
Paths are given relative to the root of the tree.
706
:param pref_names: the list of preferences to lookup - None for all
707
:param _default_searcher: private parameter to assist testing - don't use
708
:return: an iterator of tuple sequences, one per path-name.
709
See _RulesSearcher.get_items for details on the tuple sequence.
711
if _default_searcher is None:
712
_default_searcher = rules._per_user_searcher
713
searcher = self._get_rules_searcher(_default_searcher)
714
if searcher is not None:
715
if pref_names is not None:
716
for path in path_names:
717
yield searcher.get_selected_items(path, pref_names)
719
for path in path_names:
720
yield searcher.get_items(path)
722
def _get_rules_searcher(self, default_searcher):
723
"""Get the RulesSearcher for this tree given the default one."""
724
searcher = default_searcher
728
######################################################################
731
# TODO: Merge these two functions into a single one that can operate
732
# on either a whole tree or a set of files.
734
# TODO: Return the diff in order by filename, not by category or in
735
# random order. Can probably be done by lock-stepping through the
736
# filenames from both trees.
739
def file_status(filename, old_tree, new_tree):
740
"""Return single-letter status, old and new names for a file.
742
The complexity here is in deciding how to represent renames;
743
many complex cases are possible.
745
old_inv = old_tree.inventory
746
new_inv = new_tree.inventory
747
new_id = new_inv.path2id(filename)
748
old_id = old_inv.path2id(filename)
750
if not new_id and not old_id:
751
# easy: doesn't exist in either; not versioned at all
752
if new_tree.is_ignored(filename):
753
return 'I', None, None
755
return '?', None, None
757
# There is now a file of this name, great.
760
# There is no longer a file of this name, but we can describe
761
# what happened to the file that used to have
762
# this name. There are two possibilities: either it was
763
# deleted entirely, or renamed.
764
if new_inv.has_id(old_id):
765
return 'X', old_inv.id2path(old_id), new_inv.id2path(old_id)
767
return 'D', old_inv.id2path(old_id), None
769
# if the file_id is new in this revision, it is added
770
if new_id and not old_inv.has_id(new_id):
773
# if there used to be a file of this name, but that ID has now
774
# disappeared, it is deleted
775
if old_id and not new_inv.has_id(old_id):
781
@deprecated_function(deprecated_in((1, 9, 0)))
782
def find_renames(old_inv, new_inv):
783
for file_id in old_inv:
784
if file_id not in new_inv:
786
old_name = old_inv.id2path(file_id)
787
new_name = new_inv.id2path(file_id)
788
if old_name != new_name:
789
yield (old_name, new_name)
792
def find_ids_across_trees(filenames, trees, require_versioned=True):
793
"""Find the ids corresponding to specified filenames.
795
All matches in all trees will be used, and all children of matched
796
directories will be used.
798
:param filenames: The filenames to find file_ids for (if None, returns
800
:param trees: The trees to find file_ids within
801
:param require_versioned: if true, all specified filenames must occur in
803
:return: a set of file ids for the specified filenames and their children.
807
specified_path_ids = _find_ids_across_trees(filenames, trees,
809
return _find_children_across_trees(specified_path_ids, trees)
812
def _find_ids_across_trees(filenames, trees, require_versioned):
813
"""Find the ids corresponding to specified filenames.
815
All matches in all trees will be used, but subdirectories are not scanned.
817
:param filenames: The filenames to find file_ids for
818
:param trees: The trees to find file_ids within
819
:param require_versioned: if true, all specified filenames must occur in
821
:return: a set of file ids for the specified filenames
824
interesting_ids = set()
825
for tree_path in filenames:
828
file_id = tree.path2id(tree_path)
829
if file_id is not None:
830
interesting_ids.add(file_id)
833
not_versioned.append(tree_path)
834
if len(not_versioned) > 0 and require_versioned:
835
raise errors.PathsNotVersionedError(not_versioned)
836
return interesting_ids
839
def _find_children_across_trees(specified_ids, trees):
840
"""Return a set including specified ids and their children.
842
All matches in all trees will be used.
844
:param trees: The trees to find file_ids within
845
:return: a set containing all specified ids and their children
847
interesting_ids = set(specified_ids)
848
pending = interesting_ids
849
# now handle children of interesting ids
850
# we loop so that we handle all children of each id in both trees
851
while len(pending) > 0:
853
for file_id in pending:
855
if not tree.has_or_had_id(file_id):
857
for child_id in tree.iter_children(file_id):
858
if child_id not in interesting_ids:
859
new_pending.add(child_id)
860
interesting_ids.update(new_pending)
861
pending = new_pending
862
return interesting_ids
865
class InterTree(InterObject):
866
"""This class represents operations taking place between two Trees.
868
Its instances have methods like 'compare' and contain references to the
869
source and target trees these operations are to be carried out on.
871
Clients of bzrlib should not need to use InterTree directly, rather they
872
should use the convenience methods on Tree such as 'Tree.compare()' which
873
will pass through to InterTree as appropriate.
876
# Formats that will be used to test this InterTree. If both are
877
# None, this InterTree will not be tested (e.g. because a complex
879
_matching_from_tree_format = None
880
_matching_to_tree_format = None
884
def _changes_from_entries(self, source_entry, target_entry,
885
source_path=None, target_path=None):
886
"""Generate a iter_changes tuple between source_entry and target_entry.
888
:param source_entry: An inventory entry from self.source, or None.
889
:param target_entry: An inventory entry from self.target, or None.
890
:param source_path: The path of source_entry, if known. If not known
891
it will be looked up.
892
:param target_path: The path of target_entry, if known. If not known
893
it will be looked up.
894
:return: A tuple, item 0 of which is an iter_changes result tuple, and
895
item 1 is True if there are any changes in the result tuple.
897
if source_entry is None:
898
if target_entry is None:
900
file_id = target_entry.file_id
902
file_id = source_entry.file_id
903
if source_entry is not None:
904
source_versioned = True
905
source_name = source_entry.name
906
source_parent = source_entry.parent_id
907
if source_path is None:
908
source_path = self.source.id2path(file_id)
909
source_kind, source_executable, source_stat = \
910
self.source._comparison_data(source_entry, source_path)
912
source_versioned = False
916
source_executable = None
917
if target_entry is not None:
918
target_versioned = True
919
target_name = target_entry.name
920
target_parent = target_entry.parent_id
921
if target_path is None:
922
target_path = self.target.id2path(file_id)
923
target_kind, target_executable, target_stat = \
924
self.target._comparison_data(target_entry, target_path)
926
target_versioned = False
930
target_executable = None
931
versioned = (source_versioned, target_versioned)
932
kind = (source_kind, target_kind)
933
changed_content = False
934
if source_kind != target_kind:
935
changed_content = True
936
elif source_kind == 'file':
937
if (self.source.get_file_sha1(file_id, source_path, source_stat) !=
938
self.target.get_file_sha1(file_id, target_path, target_stat)):
939
changed_content = True
940
elif source_kind == 'symlink':
941
if (self.source.get_symlink_target(file_id) !=
942
self.target.get_symlink_target(file_id)):
943
changed_content = True
944
# XXX: Yes, the indentation below is wrong. But fixing it broke
945
# test_merge.TestMergerEntriesLCAOnDisk.
946
# test_nested_tree_subtree_renamed_and_modified. We'll wait for
947
# the fix from bzr.dev -- vila 2009026
948
elif source_kind == 'tree-reference':
949
if (self.source.get_reference_revision(file_id, source_path)
950
!= self.target.get_reference_revision(file_id, target_path)):
951
changed_content = True
952
parent = (source_parent, target_parent)
953
name = (source_name, target_name)
954
executable = (source_executable, target_executable)
955
if (changed_content is not False or versioned[0] != versioned[1]
956
or parent[0] != parent[1] or name[0] != name[1] or
957
executable[0] != executable[1]):
961
return (file_id, (source_path, target_path), changed_content,
962
versioned, parent, name, kind, executable), changes
965
def compare(self, want_unchanged=False, specific_files=None,
966
extra_trees=None, require_versioned=False, include_root=False,
967
want_unversioned=False):
968
"""Return the changes from source to target.
970
:return: A TreeDelta.
971
:param specific_files: An optional list of file paths to restrict the
972
comparison to. When mapping filenames to ids, all matches in all
973
trees (including optional extra_trees) are used, and all children of
974
matched directories are included.
975
:param want_unchanged: An optional boolean requesting the inclusion of
976
unchanged entries in the result.
977
:param extra_trees: An optional list of additional trees to use when
978
mapping the contents of specific_files (paths) to file_ids.
979
:param require_versioned: An optional boolean (defaults to False). When
980
supplied and True all the 'specific_files' must be versioned, or
981
a PathsNotVersionedError will be thrown.
982
:param want_unversioned: Scan for unversioned paths.
984
trees = (self.source,)
985
if extra_trees is not None:
986
trees = trees + tuple(extra_trees)
987
# target is usually the newer tree:
988
specific_file_ids = self.target.paths2ids(specific_files, trees,
989
require_versioned=require_versioned)
990
if specific_files and not specific_file_ids:
991
# All files are unversioned, so just return an empty delta
992
# _compare_trees would think we want a complete delta
993
result = delta.TreeDelta()
994
fake_entry = InventoryFile('unused', 'unused', 'unused')
995
result.unversioned = [(path, None,
996
self.target._comparison_data(fake_entry, path)[0]) for path in
999
return delta._compare_trees(self.source, self.target, want_unchanged,
1000
specific_files, include_root, extra_trees=extra_trees,
1001
require_versioned=require_versioned,
1002
want_unversioned=want_unversioned)
1004
def iter_changes(self, include_unchanged=False,
1005
specific_files=None, pb=None, extra_trees=[],
1006
require_versioned=True, want_unversioned=False):
1007
"""Generate an iterator of changes between trees.
1009
A tuple is returned:
1010
(file_id, (path_in_source, path_in_target),
1011
changed_content, versioned, parent, name, kind,
1014
Changed_content is True if the file's content has changed. This
1015
includes changes to its kind, and to a symlink's target.
1017
versioned, parent, name, kind, executable are tuples of (from, to).
1018
If a file is missing in a tree, its kind is None.
1020
Iteration is done in parent-to-child order, relative to the target
1023
There is no guarantee that all paths are in sorted order: the
1024
requirement to expand the search due to renames may result in children
1025
that should be found early being found late in the search, after
1026
lexically later results have been returned.
1027
:param require_versioned: Raise errors.PathsNotVersionedError if a
1028
path in the specific_files list is not versioned in one of
1029
source, target or extra_trees.
1030
:param specific_files: An optional list of file paths to restrict the
1031
comparison to. When mapping filenames to ids, all matches in all
1032
trees (including optional extra_trees) are used, and all children
1033
of matched directories are included. The parents in the target tree
1034
of the specific files up to and including the root of the tree are
1035
always evaluated for changes too.
1036
:param want_unversioned: Should unversioned files be returned in the
1037
output. An unversioned file is defined as one with (False, False)
1038
for the versioned pair.
1040
lookup_trees = [self.source]
1042
lookup_trees.extend(extra_trees)
1043
# The ids of items we need to examine to insure delta consistency.
1044
precise_file_ids = set()
1045
changed_file_ids = []
1046
if specific_files == []:
1047
specific_file_ids = []
1049
specific_file_ids = self.target.paths2ids(specific_files,
1050
lookup_trees, require_versioned=require_versioned)
1051
if specific_files is not None:
1052
# reparented or added entries must have their parents included
1053
# so that valid deltas can be created. The seen_parents set
1054
# tracks the parents that we need to have.
1055
# The seen_dirs set tracks directory entries we've yielded.
1056
# After outputting version object in to_entries we set difference
1057
# the two seen sets and start checking parents.
1058
seen_parents = set()
1060
if want_unversioned:
1061
all_unversioned = sorted([(p.split('/'), p) for p in
1062
self.target.extras()
1063
if specific_files is None or
1064
osutils.is_inside_any(specific_files, p)])
1065
all_unversioned = deque(all_unversioned)
1067
all_unversioned = deque()
1069
from_entries_by_dir = list(self.source.iter_entries_by_dir(
1070
specific_file_ids=specific_file_ids))
1071
from_data = dict((e.file_id, (p, e)) for p, e in from_entries_by_dir)
1072
to_entries_by_dir = list(self.target.iter_entries_by_dir(
1073
specific_file_ids=specific_file_ids))
1074
num_entries = len(from_entries_by_dir) + len(to_entries_by_dir)
1076
# the unversioned path lookup only occurs on real trees - where there
1077
# can be extras. So the fake_entry is solely used to look up
1078
# executable it values when execute is not supported.
1079
fake_entry = InventoryFile('unused', 'unused', 'unused')
1080
for target_path, target_entry in to_entries_by_dir:
1081
while (all_unversioned and
1082
all_unversioned[0][0] < target_path.split('/')):
1083
unversioned_path = all_unversioned.popleft()
1084
target_kind, target_executable, target_stat = \
1085
self.target._comparison_data(fake_entry, unversioned_path[1])
1086
yield (None, (None, unversioned_path[1]), True, (False, False),
1088
(None, unversioned_path[0][-1]),
1089
(None, target_kind),
1090
(None, target_executable))
1091
source_path, source_entry = from_data.get(target_entry.file_id,
1093
result, changes = self._changes_from_entries(source_entry,
1094
target_entry, source_path=source_path, target_path=target_path)
1095
to_paths[result[0]] = result[1][1]
1100
pb.update('comparing files', entry_count, num_entries)
1101
if changes or include_unchanged:
1102
if specific_file_ids is not None:
1103
new_parent_id = result[4][1]
1104
precise_file_ids.add(new_parent_id)
1105
changed_file_ids.append(result[0])
1107
# Ensure correct behaviour for reparented/added specific files.
1108
if specific_files is not None:
1109
# Record output dirs
1110
if result[6][1] == 'directory':
1111
seen_dirs.add(result[0])
1112
# Record parents of reparented/added entries.
1113
versioned = result[3]
1115
if not versioned[0] or parents[0] != parents[1]:
1116
seen_parents.add(parents[1])
1117
while all_unversioned:
1118
# yield any trailing unversioned paths
1119
unversioned_path = all_unversioned.popleft()
1120
to_kind, to_executable, to_stat = \
1121
self.target._comparison_data(fake_entry, unversioned_path[1])
1122
yield (None, (None, unversioned_path[1]), True, (False, False),
1124
(None, unversioned_path[0][-1]),
1126
(None, to_executable))
1127
# Yield all remaining source paths
1128
for path, from_entry in from_entries_by_dir:
1129
file_id = from_entry.file_id
1130
if file_id in to_paths:
1133
if file_id not in self.target.all_file_ids():
1134
# common case - paths we have not emitted are not present in
1138
to_path = self.target.id2path(file_id)
1141
pb.update('comparing files', entry_count, num_entries)
1142
versioned = (True, False)
1143
parent = (from_entry.parent_id, None)
1144
name = (from_entry.name, None)
1145
from_kind, from_executable, stat_value = \
1146
self.source._comparison_data(from_entry, path)
1147
kind = (from_kind, None)
1148
executable = (from_executable, None)
1149
changed_content = from_kind is not None
1150
# the parent's path is necessarily known at this point.
1151
changed_file_ids.append(file_id)
1152
yield(file_id, (path, to_path), changed_content, versioned, parent,
1153
name, kind, executable)
1154
changed_file_ids = set(changed_file_ids)
1155
if specific_file_ids is not None:
1156
for result in self._handle_precise_ids(precise_file_ids,
1160
def _get_entry(self, tree, file_id):
1161
"""Get an inventory entry from a tree, with missing entries as None.
1163
If the tree raises NotImplementedError on accessing .inventory, then
1164
this is worked around using iter_entries_by_dir on just the file id
1167
:param tree: The tree to lookup the entry in.
1168
:param file_id: The file_id to lookup.
1171
inventory = tree.inventory
1172
except NotImplementedError:
1173
# No inventory available.
1175
iterator = tree.iter_entries_by_dir(specific_file_ids=[file_id])
1176
return iterator.next()[1]
1177
except StopIteration:
1181
return inventory[file_id]
1182
except errors.NoSuchId:
1185
def _handle_precise_ids(self, precise_file_ids, changed_file_ids,
1186
discarded_changes=None):
1187
"""Fill out a partial iter_changes to be consistent.
1189
:param precise_file_ids: The file ids of parents that were seen during
1191
:param changed_file_ids: The file ids of already emitted items.
1192
:param discarded_changes: An optional dict of precalculated
1193
iter_changes items which the partial iter_changes had not output
1195
:return: A generator of iter_changes items to output.
1197
# process parents of things that had changed under the users
1198
# requested paths to prevent incorrect paths or parent ids which
1199
# aren't in the tree.
1200
while precise_file_ids:
1201
precise_file_ids.discard(None)
1202
# Don't emit file_ids twice
1203
precise_file_ids.difference_update(changed_file_ids)
1204
if not precise_file_ids:
1206
# If the there was something at a given output path in source, we
1207
# have to include the entry from source in the delta, or we would
1208
# be putting this entry into a used path.
1210
for parent_id in precise_file_ids:
1212
paths.append(self.target.id2path(parent_id))
1213
except errors.NoSuchId:
1214
# This id has been dragged in from the source by delta
1215
# expansion and isn't present in target at all: we don't
1216
# need to check for path collisions on it.
1219
old_id = self.source.path2id(path)
1220
precise_file_ids.add(old_id)
1221
precise_file_ids.discard(None)
1222
current_ids = precise_file_ids
1223
precise_file_ids = set()
1224
# We have to emit all of precise_file_ids that have been altered.
1225
# We may have to output the children of some of those ids if any
1226
# directories have stopped being directories.
1227
for file_id in current_ids:
1229
if discarded_changes:
1230
result = discarded_changes.get(file_id)
1235
old_entry = self._get_entry(self.source, file_id)
1236
new_entry = self._get_entry(self.target, file_id)
1237
result, changes = self._changes_from_entries(
1238
old_entry, new_entry)
1241
# Get this parents parent to examine.
1242
new_parent_id = result[4][1]
1243
precise_file_ids.add(new_parent_id)
1245
if (result[6][0] == 'directory' and
1246
result[6][1] != 'directory'):
1247
# This stopped being a directory, the old children have
1249
if old_entry is None:
1250
# Reusing a discarded change.
1251
old_entry = self._get_entry(self.source, file_id)
1252
for child in old_entry.children.values():
1253
precise_file_ids.add(child.file_id)
1254
changed_file_ids.add(result[0])
1258
class MultiWalker(object):
1259
"""Walk multiple trees simultaneously, getting combined results."""
1261
# Note: This could be written to not assume you can do out-of-order
1262
# lookups. Instead any nodes that don't match in all trees could be
1263
# marked as 'deferred', and then returned in the final cleanup loop.
1264
# For now, I think it is "nicer" to return things as close to the
1265
# "master_tree" order as we can.
1267
def __init__(self, master_tree, other_trees):
1268
"""Create a new MultiWalker.
1270
All trees being walked must implement "iter_entries_by_dir()", such
1271
that they yield (path, object) tuples, where that object will have a
1272
'.file_id' member, that can be used to check equality.
1274
:param master_tree: All trees will be 'slaved' to the master_tree such
1275
that nodes in master_tree will be used as 'first-pass' sync points.
1276
Any nodes that aren't in master_tree will be merged in a second
1278
:param other_trees: A list of other trees to walk simultaneously.
1280
self._master_tree = master_tree
1281
self._other_trees = other_trees
1283
# Keep track of any nodes that were properly processed just out of
1284
# order, that way we don't return them at the end, we don't have to
1285
# track *all* processed file_ids, just the out-of-order ones
1286
self._out_of_order_processed = set()
1289
def _step_one(iterator):
1290
"""Step an iter_entries_by_dir iterator.
1292
:return: (has_more, path, ie)
1293
If has_more is False, path and ie will be None.
1296
path, ie = iterator.next()
1297
except StopIteration:
1298
return False, None, None
1300
return True, path, ie
1303
def _cmp_path_by_dirblock(path1, path2):
1304
"""Compare two paths based on what directory they are in.
1306
This generates a sort order, such that all children of a directory are
1307
sorted together, and grandchildren are in the same order as the
1308
children appear. But all grandchildren come after all children.
1310
:param path1: first path
1311
:param path2: the second path
1312
:return: negative number if ``path1`` comes first,
1313
0 if paths are equal
1314
and a positive number if ``path2`` sorts first
1316
# Shortcut this special case
1319
# This is stolen from _dirstate_helpers_py.py, only switching it to
1320
# Unicode objects. Consider using encode_utf8() and then using the
1321
# optimized versions, or maybe writing optimized unicode versions.
1322
if not isinstance(path1, unicode):
1323
raise TypeError("'path1' must be a unicode string, not %s: %r"
1324
% (type(path1), path1))
1325
if not isinstance(path2, unicode):
1326
raise TypeError("'path2' must be a unicode string, not %s: %r"
1327
% (type(path2), path2))
1328
return cmp(MultiWalker._path_to_key(path1),
1329
MultiWalker._path_to_key(path2))
1332
def _path_to_key(path):
1333
dirname, basename = osutils.split(path)
1334
return (dirname.split(u'/'), basename)
1336
def _lookup_by_file_id(self, extra_entries, other_tree, file_id):
1337
"""Lookup an inventory entry by file_id.
1339
This is called when an entry is missing in the normal order.
1340
Generally this is because a file was either renamed, or it was
1341
deleted/added. If the entry was found in the inventory and not in
1342
extra_entries, it will be added to self._out_of_order_processed
1344
:param extra_entries: A dictionary of {file_id: (path, ie)}. This
1345
should be filled with entries that were found before they were
1346
used. If file_id is present, it will be removed from the
1348
:param other_tree: The Tree to search, in case we didn't find the entry
1350
:param file_id: The file_id to look for
1351
:return: (path, ie) if found or (None, None) if not present.
1353
if file_id in extra_entries:
1354
return extra_entries.pop(file_id)
1355
# TODO: Is id2path better as the first call, or is
1356
# inventory[file_id] better as a first check?
1358
cur_path = other_tree.id2path(file_id)
1359
except errors.NoSuchId:
1361
if cur_path is None:
1364
self._out_of_order_processed.add(file_id)
1365
cur_ie = other_tree.inventory[file_id]
1366
return (cur_path, cur_ie)
1369
"""Match up the values in the different trees."""
1370
for result in self._walk_master_tree():
1372
self._finish_others()
1373
for result in self._walk_others():
1376
def _walk_master_tree(self):
1377
"""First pass, walk all trees in lock-step.
1379
When we are done, all nodes in the master_tree will have been
1380
processed. _other_walkers, _other_entries, and _others_extra will be
1381
set on 'self' for future processing.
1383
# This iterator has the most "inlining" done, because it tends to touch
1384
# every file in the tree, while the others only hit nodes that don't
1386
master_iterator = self._master_tree.iter_entries_by_dir()
1388
other_walkers = [other.iter_entries_by_dir()
1389
for other in self._other_trees]
1390
other_entries = [self._step_one(walker) for walker in other_walkers]
1391
# Track extra nodes in the other trees
1392
others_extra = [{} for i in xrange(len(self._other_trees))]
1394
master_has_more = True
1395
step_one = self._step_one
1396
lookup_by_file_id = self._lookup_by_file_id
1397
out_of_order_processed = self._out_of_order_processed
1399
while master_has_more:
1400
(master_has_more, path, master_ie) = step_one(master_iterator)
1401
if not master_has_more:
1404
file_id = master_ie.file_id
1406
other_values_append = other_values.append
1407
next_other_entries = []
1408
next_other_entries_append = next_other_entries.append
1409
for idx, (other_has_more, other_path, other_ie) in enumerate(other_entries):
1410
if not other_has_more:
1411
other_values_append(lookup_by_file_id(
1412
others_extra[idx], self._other_trees[idx], file_id))
1413
next_other_entries_append((False, None, None))
1414
elif file_id == other_ie.file_id:
1415
# This is the critical code path, as most of the entries
1416
# should match between most trees.
1417
other_values_append((other_path, other_ie))
1418
next_other_entries_append(step_one(other_walkers[idx]))
1420
# This walker did not match, step it until it either
1421
# matches, or we know we are past the current walker.
1422
other_walker = other_walkers[idx]
1423
other_extra = others_extra[idx]
1424
while (other_has_more and
1425
self._cmp_path_by_dirblock(other_path, path) < 0):
1426
other_file_id = other_ie.file_id
1427
if other_file_id not in out_of_order_processed:
1428
other_extra[other_file_id] = (other_path, other_ie)
1429
other_has_more, other_path, other_ie = \
1430
step_one(other_walker)
1431
if other_has_more and other_ie.file_id == file_id:
1432
# We ended up walking to this point, match and step
1434
other_values_append((other_path, other_ie))
1435
other_has_more, other_path, other_ie = \
1436
step_one(other_walker)
1438
# This record isn't in the normal order, see if it
1440
other_values_append(lookup_by_file_id(
1441
other_extra, self._other_trees[idx], file_id))
1442
next_other_entries_append((other_has_more, other_path,
1444
other_entries = next_other_entries
1446
# We've matched all the walkers, yield this datapoint
1447
yield path, file_id, master_ie, other_values
1448
self._other_walkers = other_walkers
1449
self._other_entries = other_entries
1450
self._others_extra = others_extra
1452
def _finish_others(self):
1453
"""Finish walking the other iterators, so we get all entries."""
1454
for idx, info in enumerate(self._other_entries):
1455
other_extra = self._others_extra[idx]
1456
(other_has_more, other_path, other_ie) = info
1457
while other_has_more:
1458
other_file_id = other_ie.file_id
1459
if other_file_id not in self._out_of_order_processed:
1460
other_extra[other_file_id] = (other_path, other_ie)
1461
other_has_more, other_path, other_ie = \
1462
self._step_one(self._other_walkers[idx])
1463
del self._other_entries
1465
def _walk_others(self):
1466
"""Finish up by walking all the 'deferred' nodes."""
1467
# TODO: One alternative would be to grab all possible unprocessed
1468
# file_ids, and then sort by path, and then yield them. That
1469
# might ensure better ordering, in case a caller strictly
1470
# requires parents before children.
1471
for idx, other_extra in enumerate(self._others_extra):
1472
others = sorted(other_extra.itervalues(),
1473
key=lambda x: self._path_to_key(x[0]))
1474
for other_path, other_ie in others:
1475
file_id = other_ie.file_id
1476
# We don't need to check out_of_order_processed here, because
1477
# the lookup_by_file_id will be removing anything processed
1478
# from the extras cache
1479
other_extra.pop(file_id)
1480
other_values = [(None, None) for i in xrange(idx)]
1481
other_values.append((other_path, other_ie))
1482
for alt_idx, alt_extra in enumerate(self._others_extra[idx+1:]):
1483
alt_idx = alt_idx + idx + 1
1484
alt_extra = self._others_extra[alt_idx]
1485
alt_tree = self._other_trees[alt_idx]
1486
other_values.append(self._lookup_by_file_id(
1487
alt_extra, alt_tree, file_id))
1488
yield other_path, file_id, None, other_values
added
removed
bzrlib/tree.py
