~bzr-pqm/bzr/bzr.dev

« back to all changes in this revision

Viewing changes to bzrlib/workingtree_4.py

  • Committer: Ian Clatworthy
  • Date: 2009-09-09 11:43:10 UTC
  • mto: (4634.37.2 prepare-2.0)
  • mto: This revision was merged to the branch mainline in revision 4689.
  • Revision ID: ian.clatworthy@canonical.com-20090909114310-glw7tv76i5gnx9pt
put rules back in Makefile supporting plain-style docs

Show diffs side-by-side

added added

removed removed

Lines of Context:
1
 
# Copyright (C) 2005, 2006, 2007 Canonical Ltd
 
1
# Copyright (C) 2005, 2006, 2007, 2008, 2009 Canonical Ltd
2
2
#
3
3
# This program is free software; you can redistribute it and/or modify
4
4
# it under the terms of the GNU General Public License as published by
12
12
#
13
13
# You should have received a copy of the GNU General Public License
14
14
# along with this program; if not, write to the Free Software
15
 
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
 
15
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
16
16
 
17
17
"""WorkingTree4 format and implementation.
18
18
 
28
28
 
29
29
from bzrlib.lazy_import import lazy_import
30
30
lazy_import(globals(), """
31
 
from bisect import bisect_left
32
 
import collections
33
 
from copy import deepcopy
34
31
import errno
35
 
import itertools
36
 
import operator
37
32
import stat
38
 
from time import time
39
 
import warnings
40
33
 
41
34
import bzrlib
42
35
from bzrlib import (
43
36
    bzrdir,
44
37
    cache_utf8,
45
 
    conflicts as _mod_conflicts,
46
38
    debug,
47
 
    delta,
48
39
    dirstate,
49
40
    errors,
50
41
    generate_ids,
51
 
    globbing,
52
 
    hashcache,
53
 
    ignores,
54
 
    merge,
55
42
    osutils,
56
43
    revision as _mod_revision,
57
44
    revisiontree,
58
 
    textui,
59
45
    trace,
60
46
    transform,
61
 
    urlutils,
62
 
    xml5,
63
 
    xml6,
 
47
    views,
64
48
    )
65
49
import bzrlib.branch
66
 
from bzrlib.transport import get_transport
67
50
import bzrlib.ui
68
51
""")
69
52
 
70
 
from bzrlib import symbol_versioning
71
53
from bzrlib.decorators import needs_read_lock, needs_write_lock
72
 
from bzrlib.inventory import InventoryEntry, Inventory, ROOT_ID, entry_factory
73
 
from bzrlib.lockable_files import LockableFiles, TransportLock
74
 
from bzrlib.lockdir import LockDir
 
54
from bzrlib.filters import filtered_input_file, internal_size_sha_file_byname
 
55
from bzrlib.inventory import Inventory, ROOT_ID, entry_factory
75
56
import bzrlib.mutabletree
76
57
from bzrlib.mutabletree import needs_tree_write_lock
77
58
from bzrlib.osutils import (
78
59
    file_kind,
79
60
    isdir,
80
 
    normpath,
81
61
    pathjoin,
82
 
    rand_chars,
83
62
    realpath,
84
63
    safe_unicode,
85
 
    splitpath,
86
64
    )
87
 
from bzrlib.trace import mutter, note
 
65
from bzrlib.trace import mutter
88
66
from bzrlib.transport.local import LocalTransport
89
67
from bzrlib.tree import InterTree
90
 
from bzrlib.progress import DummyProgress, ProgressPhase
91
 
from bzrlib.revision import NULL_REVISION, CURRENT_REVISION
92
 
from bzrlib.rio import RioReader, rio_file, Stanza
93
 
from bzrlib.symbol_versioning import (deprecated_passed,
94
 
        deprecated_method,
95
 
        deprecated_function,
96
 
        DEPRECATED_PARAMETER,
97
 
        )
98
68
from bzrlib.tree import Tree
99
69
from bzrlib.workingtree import WorkingTree, WorkingTree3, WorkingTreeFormat3
100
70
 
101
71
 
102
 
# This is the Windows equivalent of ENOTDIR
103
 
# It is defined in pywin32.winerror, but we don't want a strong dependency for
104
 
# just an error code.
105
 
ERROR_PATH_NOT_FOUND = 3
106
 
ERROR_DIRECTORY = 267
107
 
 
108
 
 
109
 
class WorkingTree4(WorkingTree3):
110
 
    """This is the Format 4 working tree.
111
 
 
112
 
    This differs from WorkingTree3 by:
113
 
     - Having a consolidated internal dirstate, stored in a
114
 
       randomly-accessible sorted file on disk.
115
 
     - Not having a regular inventory attribute.  One can be synthesized 
116
 
       on demand but this is expensive and should be avoided.
117
 
 
118
 
    This is new in bzr 0.15.
119
 
    """
120
 
 
 
72
class DirStateWorkingTree(WorkingTree3):
121
73
    def __init__(self, basedir,
122
74
                 branch,
123
75
                 _control_files=None,
132
84
        """
133
85
        self._format = _format
134
86
        self.bzrdir = _bzrdir
135
 
        from bzrlib.trace import note, mutter
136
 
        assert isinstance(basedir, basestring), \
137
 
            "base directory %r is not a string" % basedir
138
87
        basedir = safe_unicode(basedir)
139
88
        mutter("opening working tree %r", basedir)
140
89
        self._branch = branch
141
 
        assert isinstance(self.branch, bzrlib.branch.Branch), \
142
 
            "branch %r is not a Branch" % self.branch
143
90
        self.basedir = realpath(basedir)
144
91
        # if branch is at our basedir and is a format 6 or less
145
92
        # assume all other formats have their own control files.
146
 
        assert isinstance(_control_files, LockableFiles), \
147
 
            "_control_files must be a LockableFiles, not %r" % _control_files
148
93
        self._control_files = _control_files
 
94
        self._transport = self._control_files._transport
149
95
        self._dirty = None
150
96
        #-------------
151
97
        # during a read or write lock these objects are set, and are
153
99
        self._dirstate = None
154
100
        self._inventory = None
155
101
        #-------------
 
102
        self._setup_directory_is_tree_reference()
 
103
        self._detect_case_handling()
 
104
        self._rules_searcher = None
 
105
        self.views = self._make_views()
 
106
        #--- allow tests to select the dirstate iter_changes implementation
 
107
        self._iter_changes = dirstate._process_entry
156
108
 
157
109
    @needs_tree_write_lock
158
110
    def _add(self, files, ids, kinds):
160
112
        state = self.current_dirstate()
161
113
        for f, file_id, kind in zip(files, ids, kinds):
162
114
            f = f.strip('/')
163
 
            assert '//' not in f
164
 
            assert '..' not in f
165
115
            if self.path2id(f):
166
116
                # special case tree root handling.
167
117
                if f == '' and self.path2id(f) == ROOT_ID:
188
138
    @needs_tree_write_lock
189
139
    def add_reference(self, sub_tree):
190
140
        # use standard implementation, which calls back to self._add
191
 
        # 
 
141
        #
192
142
        # So we don't store the reference_revision in the working dirstate,
193
 
        # it's just recorded at the moment of commit. 
 
143
        # it's just recorded at the moment of commit.
194
144
        self._add_reference(sub_tree)
195
145
 
196
146
    def break_lock(self):
235
185
            WorkingTree3._comparison_data(self, entry, path)
236
186
        # it looks like a plain directory, but it's really a reference -- see
237
187
        # also kind()
238
 
        if (self._repo_supports_tree_reference and
239
 
            kind == 'directory' and
240
 
            self._directory_is_tree_reference(path)):
 
188
        if (self._repo_supports_tree_reference and kind == 'directory'
 
189
            and entry is not None and entry.kind == 'tree-reference'):
241
190
            kind = 'tree-reference'
242
191
        return kind, executable, stat_value
243
192
 
269
218
            return self._dirstate
270
219
        local_path = self.bzrdir.get_workingtree_transport(None
271
220
            ).local_abspath('dirstate')
272
 
        self._dirstate = dirstate.DirState.on_file(local_path)
 
221
        self._dirstate = dirstate.DirState.on_file(local_path,
 
222
            self._sha1_provider())
273
223
        return self._dirstate
274
224
 
275
 
    def _directory_is_tree_reference(self, relpath):
276
 
        # as a special case, if a directory contains control files then 
277
 
        # it's a tree reference, except that the root of the tree is not
278
 
        return relpath and osutils.isdir(self.abspath(relpath) + u"/.bzr")
279
 
        # TODO: We could ask all the control formats whether they
280
 
        # recognize this directory, but at the moment there's no cheap api
281
 
        # to do that.  Since we probably can only nest bzr checkouts and
282
 
        # they always use this name it's ok for now.  -- mbp 20060306
283
 
        #
284
 
        # FIXME: There is an unhandled case here of a subdirectory
285
 
        # containing .bzr but not a branch; that will probably blow up
286
 
        # when you try to commit it.  It might happen if there is a
287
 
        # checkout in a subdirectory.  This can be avoided by not adding
288
 
        # it.  mbp 20070306
 
225
    def _sha1_provider(self):
 
226
        """A function that returns a SHA1Provider suitable for this tree.
 
227
 
 
228
        :return: None if content filtering is not supported by this tree.
 
229
          Otherwise, a SHA1Provider is returned that sha's the canonical
 
230
          form of files, i.e. after read filters are applied.
 
231
        """
 
232
        if self.supports_content_filtering():
 
233
            return ContentFilterAwareSHA1Provider(self)
 
234
        else:
 
235
            return None
289
236
 
290
237
    def filter_unversioned_files(self, paths):
291
238
        """Filter out paths that are versioned.
324
271
 
325
272
    def _generate_inventory(self):
326
273
        """Create and set self.inventory from the dirstate object.
327
 
        
 
274
 
328
275
        This is relatively expensive: we have to walk the entire dirstate.
329
276
        Ideally we would not, and can deprecate this function.
330
277
        """
334
281
        state._read_dirblocks_if_needed()
335
282
        root_key, current_entry = self._get_entry(path='')
336
283
        current_id = root_key[2]
337
 
        assert current_entry[0][0] == 'd' # directory
 
284
        if not (current_entry[0][0] == 'd'): # directory
 
285
            raise AssertionError(current_entry)
338
286
        inv = Inventory(root_id=current_id)
339
287
        # Turn some things into local variables
340
288
        minikind_to_kind = dirstate.DirState._minikind_to_kind
373
321
                    # add this entry to the parent map.
374
322
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
375
323
                elif kind == 'tree-reference':
376
 
                    assert self._repo_supports_tree_reference, \
377
 
                        "repository of %r " \
378
 
                        "doesn't support tree references " \
379
 
                        "required by entry %r" \
380
 
                        % (self, name)
 
324
                    if not self._repo_supports_tree_reference:
 
325
                        raise errors.UnsupportedOperation(
 
326
                            self._generate_inventory,
 
327
                            self.branch.repository)
381
328
                    inv_entry.reference_revision = link_or_sha1 or None
382
329
                elif kind != 'symlink':
383
330
                    raise AssertionError("unknown kind %r" % kind)
384
331
                # These checks cost us around 40ms on a 55k entry tree
385
 
                assert file_id not in inv_byid, ('file_id %s already in'
386
 
                    ' inventory as %s' % (file_id, inv_byid[file_id]))
387
 
                assert name_unicode not in parent_ie.children
 
332
                if file_id in inv_byid:
 
333
                    raise AssertionError('file_id %s already in'
 
334
                        ' inventory as %s' % (file_id, inv_byid[file_id]))
 
335
                if name_unicode in parent_ie.children:
 
336
                    raise AssertionError('name %r already in parent'
 
337
                        % (name_unicode,))
388
338
                inv_byid[file_id] = inv_entry
389
339
                parent_ie.children[name_unicode] = inv_entry
390
340
        self._inventory = inv
395
345
        If either file_id or path is supplied, it is used as the key to lookup.
396
346
        If both are supplied, the fastest lookup is used, and an error is
397
347
        raised if they do not both point at the same row.
398
 
        
 
348
 
399
349
        :param file_id: An optional unicode file_id to be looked up.
400
350
        :param path: An optional unicode path to be looked up.
401
351
        :return: The dirstate row tuple for path/file_id, or (None, None)
425
375
                    return None
426
376
                else:
427
377
                    raise
428
 
        link_or_sha1 = state.update_entry(entry, file_abspath,
429
 
                                          stat_value=stat_value)
 
378
        link_or_sha1 = dirstate.update_entry(state, entry, file_abspath,
 
379
            stat_value=stat_value)
430
380
        if entry[1][0][0] == 'f':
431
 
            return link_or_sha1
 
381
            if link_or_sha1 is None:
 
382
                file_obj, statvalue = self.get_file_with_stat(file_id, path)
 
383
                try:
 
384
                    sha1 = osutils.sha_file(file_obj)
 
385
                finally:
 
386
                    file_obj.close()
 
387
                self._observed_sha1(file_id, path, (sha1, statvalue))
 
388
                return sha1
 
389
            else:
 
390
                return link_or_sha1
432
391
        return None
433
392
 
434
393
    def _get_inventory(self):
448
407
    @needs_read_lock
449
408
    def get_parent_ids(self):
450
409
        """See Tree.get_parent_ids.
451
 
        
 
410
 
452
411
        This implementation requests the ids list from the dirstate file.
453
412
        """
454
413
        return self.current_dirstate().get_parent_ids()
470
429
 
471
430
    def has_id(self, file_id):
472
431
        state = self.current_dirstate()
473
 
        file_id = osutils.safe_file_id(file_id)
474
432
        row, parents = self._get_entry(file_id=file_id)
475
433
        if row is None:
476
434
            return False
477
435
        return osutils.lexists(pathjoin(
478
436
                    self.basedir, row[0].decode('utf8'), row[1].decode('utf8')))
479
437
 
 
438
    def has_or_had_id(self, file_id):
 
439
        state = self.current_dirstate()
 
440
        row, parents = self._get_entry(file_id=file_id)
 
441
        return row is not None
 
442
 
480
443
    @needs_read_lock
481
444
    def id2path(self, file_id):
482
445
        "Convert a file-id to a path."
483
 
        file_id = osutils.safe_file_id(file_id)
484
446
        state = self.current_dirstate()
485
447
        entry = self._get_entry(file_id=file_id)
486
448
        if entry == (None, None):
488
450
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
489
451
        return path_utf8.decode('utf8')
490
452
 
 
453
    def _is_executable_from_path_and_stat_from_basis(self, path, stat_result):
 
454
        entry = self._get_entry(path=path)
 
455
        if entry == (None, None):
 
456
            return False # Missing entries are not executable
 
457
        return entry[1][0][3] # Executable?
 
458
 
491
459
    if not osutils.supports_executable():
492
460
        def is_executable(self, file_id, path=None):
493
461
            """Test if a file is executable or not.
494
462
 
495
463
            Note: The caller is expected to take a read-lock before calling this.
496
464
            """
497
 
            file_id = osutils.safe_file_id(file_id)
498
465
            entry = self._get_entry(file_id=file_id, path=path)
499
466
            if entry == (None, None):
500
467
                return False
501
468
            return entry[1][0][3]
 
469
 
 
470
        _is_executable_from_path_and_stat = \
 
471
            _is_executable_from_path_and_stat_from_basis
502
472
    else:
503
473
        def is_executable(self, file_id, path=None):
504
474
            """Test if a file is executable or not.
505
475
 
506
476
            Note: The caller is expected to take a read-lock before calling this.
507
477
            """
 
478
            self._must_be_locked()
508
479
            if not path:
509
 
                file_id = osutils.safe_file_id(file_id)
510
480
                path = self.id2path(file_id)
511
481
            mode = os.lstat(self.abspath(path)).st_mode
512
482
            return bool(stat.S_ISREG(mode) and stat.S_IEXEC & mode)
513
483
 
 
484
    def all_file_ids(self):
 
485
        """See Tree.iter_all_file_ids"""
 
486
        self._must_be_locked()
 
487
        result = set()
 
488
        for key, tree_details in self.current_dirstate()._iter_entries():
 
489
            if tree_details[0][0] in ('a', 'r'): # relocated
 
490
                continue
 
491
            result.add(key[2])
 
492
        return result
 
493
 
514
494
    @needs_read_lock
515
495
    def __iter__(self):
516
496
        """Iterate through file_ids for this tree.
529
509
        return iter(result)
530
510
 
531
511
    def iter_references(self):
 
512
        if not self._repo_supports_tree_reference:
 
513
            # When the repo doesn't support references, we will have nothing to
 
514
            # return
 
515
            return
532
516
        for key, tree_details in self.current_dirstate()._iter_entries():
533
517
            if tree_details[0][0] in ('a', 'r'): # absent, relocated
534
518
                # not relevant to the working tree
536
520
            if not key[1]:
537
521
                # the root is not a reference.
538
522
                continue
539
 
            path = pathjoin(self.basedir, key[0].decode('utf8'), key[1].decode('utf8'))
 
523
            relpath = pathjoin(key[0].decode('utf8'), key[1].decode('utf8'))
540
524
            try:
541
 
                if self._kind(path) == 'tree-reference':
542
 
                    yield path, key[2]
 
525
                if self._kind(relpath) == 'tree-reference':
 
526
                    yield relpath, key[2]
543
527
            except errors.NoSuchFile:
544
528
                # path is missing on disk.
545
529
                continue
546
530
 
 
531
    def _observed_sha1(self, file_id, path, (sha1, statvalue)):
 
532
        """See MutableTree._observed_sha1."""
 
533
        state = self.current_dirstate()
 
534
        entry = self._get_entry(file_id=file_id, path=path)
 
535
        state._observed_sha1(entry, sha1, statvalue)
 
536
 
547
537
    def kind(self, file_id):
548
538
        """Return the kind of a file.
549
539
 
553
543
        Note: The caller is expected to take a read-lock before calling this.
554
544
        """
555
545
        relpath = self.id2path(file_id)
556
 
        assert relpath != None, \
557
 
            "path for id {%s} is None!" % file_id
 
546
        if relpath is None:
 
547
            raise AssertionError(
 
548
                "path for id {%s} is None!" % file_id)
558
549
        return self._kind(relpath)
559
550
 
560
551
    def _kind(self, relpath):
561
552
        abspath = self.abspath(relpath)
562
553
        kind = file_kind(abspath)
563
 
        if (self._repo_supports_tree_reference and
564
 
            kind == 'directory' and
565
 
            self._directory_is_tree_reference(relpath)):
566
 
            kind = 'tree-reference'
 
554
        if (self._repo_supports_tree_reference and kind == 'directory'):
 
555
            entry = self._get_entry(path=relpath)
 
556
            if entry[1] is not None:
 
557
                if entry[1][0][0] == 't':
 
558
                    kind = 'tree-reference'
567
559
        return kind
568
560
 
569
561
    @needs_read_lock
632
624
        result = []
633
625
        if not from_paths:
634
626
            return result
635
 
 
636
627
        state = self.current_dirstate()
637
 
 
638
 
        assert not isinstance(from_paths, basestring)
 
628
        if isinstance(from_paths, basestring):
 
629
            raise ValueError()
639
630
        to_dir_utf8 = to_dir.encode('utf8')
640
631
        to_entry_dirname, to_basename = os.path.split(to_dir_utf8)
641
632
        id_index = state._get_id_index()
663
654
        if self._inventory is not None:
664
655
            update_inventory = True
665
656
            inv = self.inventory
 
657
            to_dir_id = to_entry[0][2]
666
658
            to_dir_ie = inv[to_dir_id]
667
 
            to_dir_id = to_entry[0][2]
668
659
        else:
669
660
            update_inventory = False
670
661
 
701
692
            from_entry = self._get_entry(path=from_rel)
702
693
            if from_entry == (None, None):
703
694
                raise errors.BzrMoveFailedError(from_rel,to_dir,
704
 
                    errors.NotVersionedError(path=str(from_rel)))
 
695
                    errors.NotVersionedError(path=from_rel))
705
696
 
706
697
            from_id = from_entry[0][2]
707
698
            to_rel = pathjoin(to_dir, from_tail)
734
725
                if from_missing: # implicitly just update our path mapping
735
726
                    move_file = False
736
727
                elif not after:
737
 
                    raise errors.RenameFailedFilesExist(from_rel, to_rel,
738
 
                        extra="(Use --after to update the Bazaar id)")
 
728
                    raise errors.RenameFailedFilesExist(from_rel, to_rel)
739
729
 
740
730
            rollbacks = []
741
731
            def rollback_rename():
797
787
                if minikind == 'd':
798
788
                    def update_dirblock(from_dir, to_key, to_dir_utf8):
799
789
                        """Recursively update all entries in this dirblock."""
800
 
                        assert from_dir != '', "renaming root not supported"
 
790
                        if from_dir == '':
 
791
                            raise AssertionError("renaming root not supported")
801
792
                        from_key = (from_dir, '')
802
793
                        from_block_idx, present = \
803
794
                            state._find_block_index_from_key(from_key)
816
807
 
817
808
                        # Grab a copy since move_one may update the list.
818
809
                        for entry in from_block[1][:]:
819
 
                            assert entry[0][0] == from_dir
 
810
                            if not (entry[0][0] == from_dir):
 
811
                                raise AssertionError()
820
812
                            cur_details = entry[1][0]
821
813
                            to_key = (to_dir_utf8, entry[0][1], entry[0][2])
822
814
                            from_path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
881
873
        for tree in trees:
882
874
            if not (isinstance(tree, DirStateRevisionTree) and tree._revision_id in
883
875
                parents):
884
 
                return super(WorkingTree4, self).paths2ids(paths, trees, require_versioned)
 
876
                return super(DirStateWorkingTree, self).paths2ids(paths,
 
877
                    trees, require_versioned)
885
878
        search_indexes = [0] + [1 + parents.index(tree._revision_id) for tree in trees]
886
879
        # -- make all paths utf8 --
887
880
        paths_utf8 = set()
947
940
            if not all_versioned:
948
941
                raise errors.PathsNotVersionedError(paths)
949
942
        # -- remove redundancy in supplied paths to prevent over-scanning --
950
 
        search_paths = set()
951
 
        for path in paths:
952
 
            other_paths = paths.difference(set([path]))
953
 
            if not osutils.is_inside_any(other_paths, path):
954
 
                # this is a top level path, we must check it.
955
 
                search_paths.add(path)
956
 
        # sketch: 
 
943
        search_paths = osutils.minimum_path_selection(paths)
 
944
        # sketch:
957
945
        # for all search_indexs in each path at or under each element of
958
946
        # search_paths, if the detail is relocated: add the id, and add the
959
947
        # relocated path as one to search if its not searched already. If the
1015
1003
 
1016
1004
    def read_working_inventory(self):
1017
1005
        """Read the working inventory.
1018
 
        
 
1006
 
1019
1007
        This is a meaningless operation for dirstate, but we obey it anyhow.
1020
1008
        """
1021
1009
        return self.inventory
1026
1014
 
1027
1015
        WorkingTree4 supplies revision_trees for any basis tree.
1028
1016
        """
1029
 
        revision_id = osutils.safe_revision_id(revision_id)
1030
1017
        dirstate = self.current_dirstate()
1031
1018
        parent_ids = dirstate.get_parent_ids()
1032
1019
        if revision_id not in parent_ids:
1039
1026
    @needs_tree_write_lock
1040
1027
    def set_last_revision(self, new_revision):
1041
1028
        """Change the last revision in the working tree."""
1042
 
        new_revision = osutils.safe_revision_id(new_revision)
1043
1029
        parents = self.get_parent_ids()
1044
 
        if new_revision in (NULL_REVISION, None):
1045
 
            assert len(parents) < 2, (
1046
 
                "setting the last parent to none with a pending merge is "
1047
 
                "unsupported.")
 
1030
        if new_revision in (_mod_revision.NULL_REVISION, None):
 
1031
            if len(parents) >= 2:
 
1032
                raise AssertionError(
 
1033
                    "setting the last parent to none with a pending merge is "
 
1034
                    "unsupported.")
1048
1035
            self.set_parent_ids([])
1049
1036
        else:
1050
1037
            self.set_parent_ids([new_revision] + parents[1:],
1053
1040
    @needs_tree_write_lock
1054
1041
    def set_parent_ids(self, revision_ids, allow_leftmost_as_ghost=False):
1055
1042
        """Set the parent ids to revision_ids.
1056
 
        
 
1043
 
1057
1044
        See also set_parent_trees. This api will try to retrieve the tree data
1058
1045
        for each element of revision_ids from the trees repository. If you have
1059
1046
        tree data already available, it is more efficient to use
1063
1050
        :param revision_ids: The revision_ids to set as the parent ids of this
1064
1051
            working tree. Any of these may be ghosts.
1065
1052
        """
1066
 
        revision_ids = [osutils.safe_revision_id(r) for r in revision_ids]
1067
1053
        trees = []
1068
1054
        for revision_id in revision_ids:
1069
1055
            try:
1092
1078
                raise errors.GhostRevisionUnusableHere(parents_list[0][0])
1093
1079
        real_trees = []
1094
1080
        ghosts = []
 
1081
 
 
1082
        parent_ids = [rev_id for rev_id, tree in parents_list]
 
1083
        graph = self.branch.repository.get_graph()
 
1084
        heads = graph.heads(parent_ids)
 
1085
        accepted_revisions = set()
 
1086
 
1095
1087
        # convert absent trees to the null tree, which we convert back to
1096
1088
        # missing on access.
1097
1089
        for rev_id, tree in parents_list:
1098
 
            rev_id = osutils.safe_revision_id(rev_id)
 
1090
            if len(accepted_revisions) > 0:
 
1091
                # we always accept the first tree
 
1092
                if rev_id in accepted_revisions or rev_id not in heads:
 
1093
                    # We have already included either this tree, or its
 
1094
                    # descendent, so we skip it.
 
1095
                    continue
1099
1096
            _mod_revision.check_not_reserved_id(rev_id)
1100
1097
            if tree is not None:
1101
1098
                real_trees.append((rev_id, tree))
1102
1099
            else:
1103
1100
                real_trees.append((rev_id,
1104
 
                    self.branch.repository.revision_tree(None)))
 
1101
                    self.branch.repository.revision_tree(
 
1102
                        _mod_revision.NULL_REVISION)))
1105
1103
                ghosts.append(rev_id)
 
1104
            accepted_revisions.add(rev_id)
1106
1105
        dirstate.set_parent_trees(real_trees, ghosts=ghosts)
1107
1106
        self._make_dirty(reset_inventory=False)
1108
1107
 
1113
1112
        if state._dirblock_state == dirstate.DirState.IN_MEMORY_MODIFIED:
1114
1113
            self._make_dirty(reset_inventory=True)
1115
1114
 
 
1115
    def _sha_from_stat(self, path, stat_result):
 
1116
        """Get a sha digest from the tree's stat cache.
 
1117
 
 
1118
        The default implementation assumes no stat cache is present.
 
1119
 
 
1120
        :param path: The path.
 
1121
        :param stat_result: The stat result being looked up.
 
1122
        """
 
1123
        return self.current_dirstate().sha1_from_stat(path, stat_result)
 
1124
 
1116
1125
    @needs_read_lock
1117
1126
    def supports_tree_reference(self):
1118
1127
        return self._repo_supports_tree_reference
1158
1167
            return
1159
1168
        state = self.current_dirstate()
1160
1169
        state._read_dirblocks_if_needed()
1161
 
        ids_to_unversion = set()
1162
 
        for file_id in file_ids:
1163
 
            ids_to_unversion.add(osutils.safe_file_id(file_id))
 
1170
        ids_to_unversion = set(file_ids)
1164
1171
        paths_to_unversion = set()
1165
1172
        # sketch:
1166
1173
        # check if the root is to be unversioned, if so, assert for now.
1193
1200
                # just forget the whole block.
1194
1201
                entry_index = 0
1195
1202
                while entry_index < len(block[1]):
1196
 
                    # Mark this file id as having been removed
1197
1203
                    entry = block[1][entry_index]
1198
 
                    ids_to_unversion.discard(entry[0][2])
1199
 
                    if (entry[1][0][0] == 'a'
1200
 
                        or not state._make_absent(entry)):
 
1204
                    if entry[1][0][0] in 'ar':
 
1205
                        # don't remove absent or renamed entries
1201
1206
                        entry_index += 1
 
1207
                    else:
 
1208
                        # Mark this file id as having been removed
 
1209
                        ids_to_unversion.discard(entry[0][2])
 
1210
                        if not state._make_absent(entry):
 
1211
                            # The block has not shrunk.
 
1212
                            entry_index += 1
1202
1213
                # go to the next block. (At the moment we dont delete empty
1203
1214
                # dirblocks)
1204
1215
                block_index += 1
1227
1238
            for file_id in file_ids:
1228
1239
                self._inventory.remove_recursive_id(file_id)
1229
1240
 
 
1241
    @needs_tree_write_lock
 
1242
    def rename_one(self, from_rel, to_rel, after=False):
 
1243
        """See WorkingTree.rename_one"""
 
1244
        self.flush()
 
1245
        WorkingTree.rename_one(self, from_rel, to_rel, after)
 
1246
 
 
1247
    @needs_tree_write_lock
 
1248
    def apply_inventory_delta(self, changes):
 
1249
        """See MutableTree.apply_inventory_delta"""
 
1250
        state = self.current_dirstate()
 
1251
        state.update_by_delta(changes)
 
1252
        self._make_dirty(reset_inventory=True)
 
1253
 
 
1254
    def update_basis_by_delta(self, new_revid, delta):
 
1255
        """See MutableTree.update_basis_by_delta."""
 
1256
        if self.last_revision() == new_revid:
 
1257
            raise AssertionError()
 
1258
        self.current_dirstate().update_basis_by_delta(delta, new_revid)
 
1259
 
1230
1260
    @needs_read_lock
1231
1261
    def _validate(self):
1232
1262
        self._dirstate._validate()
1234
1264
    @needs_tree_write_lock
1235
1265
    def _write_inventory(self, inv):
1236
1266
        """Write inventory as the current inventory."""
1237
 
        assert not self._dirty, "attempting to write an inventory when the dirstate is dirty will cause data loss"
 
1267
        if self._dirty:
 
1268
            raise AssertionError("attempting to write an inventory when the "
 
1269
                "dirstate is dirty will lose pending changes")
1238
1270
        self.current_dirstate().set_state_from_inventory(inv)
1239
1271
        self._make_dirty(reset_inventory=False)
1240
1272
        if self._inventory is not None:
1242
1274
        self.flush()
1243
1275
 
1244
1276
 
1245
 
class WorkingTreeFormat4(WorkingTreeFormat3):
1246
 
    """The first consolidated dirstate working tree format.
1247
 
 
1248
 
    This format:
1249
 
        - exists within a metadir controlling .bzr
1250
 
        - includes an explicit version marker for the workingtree control
1251
 
          files, separate from the BzrDir format
1252
 
        - modifies the hash cache format
1253
 
        - is new in bzr 0.15
1254
 
        - uses a LockDir to guard access to it.
1255
 
    """
1256
 
 
1257
 
    upgrade_recommended = False
1258
 
 
1259
 
    def get_format_string(self):
1260
 
        """See WorkingTreeFormat.get_format_string()."""
1261
 
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
1262
 
 
1263
 
    def get_format_description(self):
1264
 
        """See WorkingTreeFormat.get_format_description()."""
1265
 
        return "Working tree format 4"
1266
 
 
1267
 
    def initialize(self, a_bzrdir, revision_id=None):
 
1277
class ContentFilterAwareSHA1Provider(dirstate.SHA1Provider):
 
1278
 
 
1279
    def __init__(self, tree):
 
1280
        self.tree = tree
 
1281
 
 
1282
    def sha1(self, abspath):
 
1283
        """See dirstate.SHA1Provider.sha1()."""
 
1284
        filters = self.tree._content_filter_stack(
 
1285
            self.tree.relpath(osutils.safe_unicode(abspath)))
 
1286
        return internal_size_sha_file_byname(abspath, filters)[1]
 
1287
 
 
1288
    def stat_and_sha1(self, abspath):
 
1289
        """See dirstate.SHA1Provider.stat_and_sha1()."""
 
1290
        filters = self.tree._content_filter_stack(
 
1291
            self.tree.relpath(osutils.safe_unicode(abspath)))
 
1292
        file_obj = file(abspath, 'rb', 65000)
 
1293
        try:
 
1294
            statvalue = os.fstat(file_obj.fileno())
 
1295
            if filters:
 
1296
                file_obj = filtered_input_file(file_obj, filters)
 
1297
            sha1 = osutils.size_sha_file(file_obj)[1]
 
1298
        finally:
 
1299
            file_obj.close()
 
1300
        return statvalue, sha1
 
1301
 
 
1302
 
 
1303
class ContentFilteringDirStateWorkingTree(DirStateWorkingTree):
 
1304
    """Dirstate working tree that supports content filtering.
 
1305
 
 
1306
    The dirstate holds the hash and size of the canonical form of the file, 
 
1307
    and most methods must return that.
 
1308
    """
 
1309
 
 
1310
    def _file_content_summary(self, path, stat_result):
 
1311
        # This is to support the somewhat obsolete path_content_summary method
 
1312
        # with content filtering: see
 
1313
        # <https://bugs.edge.launchpad.net/bzr/+bug/415508>.
 
1314
        #
 
1315
        # If the dirstate cache is up to date and knows the hash and size,
 
1316
        # return that.
 
1317
        # Otherwise if there are no content filters, return the on-disk size
 
1318
        # and leave the hash blank.
 
1319
        # Otherwise, read and filter the on-disk file and use its size and
 
1320
        # hash.
 
1321
        #
 
1322
        # The dirstate doesn't store the size of the canonical form so we
 
1323
        # can't trust it for content-filtered trees.  We just return None.
 
1324
        dirstate_sha1 = self._dirstate.sha1_from_stat(path, stat_result)
 
1325
        executable = self._is_executable_from_path_and_stat(path, stat_result)
 
1326
        return ('file', None, executable, dirstate_sha1)
 
1327
 
 
1328
 
 
1329
class WorkingTree4(DirStateWorkingTree):
 
1330
    """This is the Format 4 working tree.
 
1331
 
 
1332
    This differs from WorkingTree3 by:
 
1333
     - Having a consolidated internal dirstate, stored in a
 
1334
       randomly-accessible sorted file on disk.
 
1335
     - Not having a regular inventory attribute.  One can be synthesized
 
1336
       on demand but this is expensive and should be avoided.
 
1337
 
 
1338
    This is new in bzr 0.15.
 
1339
    """
 
1340
 
 
1341
 
 
1342
class WorkingTree5(ContentFilteringDirStateWorkingTree):
 
1343
    """This is the Format 5 working tree.
 
1344
 
 
1345
    This differs from WorkingTree4 by:
 
1346
     - Supporting content filtering.
 
1347
 
 
1348
    This is new in bzr 1.11.
 
1349
    """
 
1350
 
 
1351
 
 
1352
class WorkingTree6(ContentFilteringDirStateWorkingTree):
 
1353
    """This is the Format 6 working tree.
 
1354
 
 
1355
    This differs from WorkingTree5 by:
 
1356
     - Supporting a current view that may mask the set of files in a tree
 
1357
       impacted by most user operations.
 
1358
 
 
1359
    This is new in bzr 1.14.
 
1360
    """
 
1361
 
 
1362
    def _make_views(self):
 
1363
        return views.PathBasedViews(self)
 
1364
 
 
1365
 
 
1366
class DirStateWorkingTreeFormat(WorkingTreeFormat3):
 
1367
    def initialize(self, a_bzrdir, revision_id=None, from_branch=None,
 
1368
                   accelerator_tree=None, hardlink=False):
1268
1369
        """See WorkingTreeFormat.initialize().
1269
1370
 
1270
1371
        :param revision_id: allows creating a working tree at a different
1271
1372
        revision than the branch is at.
 
1373
        :param accelerator_tree: A tree which can be used for retrieving file
 
1374
            contents more quickly than the revision tree, i.e. a workingtree.
 
1375
            The revision tree will be used for cases where accelerator_tree's
 
1376
            content is different.
 
1377
        :param hardlink: If true, hard-link files from accelerator_tree,
 
1378
            where possible.
1272
1379
 
1273
1380
        These trees get an initial random root id, if their repository supports
1274
1381
        rich root data, TREE_ROOT otherwise.
1275
1382
        """
1276
 
        revision_id = osutils.safe_revision_id(revision_id)
1277
1383
        if not isinstance(a_bzrdir.transport, LocalTransport):
1278
1384
            raise errors.NotLocalUrl(a_bzrdir.transport.base)
1279
1385
        transport = a_bzrdir.get_workingtree_transport(self)
1280
1386
        control_files = self._open_control_files(a_bzrdir)
1281
1387
        control_files.create_lock()
1282
1388
        control_files.lock_write()
1283
 
        control_files.put_utf8('format', self.get_format_string())
1284
 
        branch = a_bzrdir.open_branch()
 
1389
        transport.put_bytes('format', self.get_format_string(),
 
1390
            mode=a_bzrdir._get_file_mode())
 
1391
        if from_branch is not None:
 
1392
            branch = from_branch
 
1393
        else:
 
1394
            branch = a_bzrdir.open_branch()
1285
1395
        if revision_id is None:
1286
1396
            revision_id = branch.last_revision()
1287
1397
        local_path = transport.local_abspath('dirstate')
1289
1399
        state = dirstate.DirState.initialize(local_path)
1290
1400
        state.unlock()
1291
1401
        del state
1292
 
        wt = WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
 
1402
        wt = self._tree_class(a_bzrdir.root_transport.local_abspath('.'),
1293
1403
                         branch,
1294
1404
                         _format=self,
1295
1405
                         _bzrdir=a_bzrdir,
1297
1407
        wt._new_tree()
1298
1408
        wt.lock_tree_write()
1299
1409
        try:
1300
 
            if revision_id in (None, NULL_REVISION):
 
1410
            self._init_custom_control_files(wt)
 
1411
            if revision_id in (None, _mod_revision.NULL_REVISION):
1301
1412
                if branch.repository.supports_rich_root():
1302
1413
                    wt._set_root_id(generate_ids.gen_root_id())
1303
1414
                else:
1304
1415
                    wt._set_root_id(ROOT_ID)
1305
1416
                wt.flush()
1306
 
            wt.set_last_revision(revision_id)
1307
 
            wt.flush()
1308
 
            basis = wt.basis_tree()
 
1417
            basis = None
 
1418
            # frequently, we will get here due to branching.  The accelerator
 
1419
            # tree will be the tree from the branch, so the desired basis
 
1420
            # tree will often be a parent of the accelerator tree.
 
1421
            if accelerator_tree is not None:
 
1422
                try:
 
1423
                    basis = accelerator_tree.revision_tree(revision_id)
 
1424
                except errors.NoSuchRevision:
 
1425
                    pass
 
1426
            if basis is None:
 
1427
                basis = branch.repository.revision_tree(revision_id)
 
1428
            if revision_id == _mod_revision.NULL_REVISION:
 
1429
                parents_list = []
 
1430
            else:
 
1431
                parents_list = [(revision_id, basis)]
1309
1432
            basis.lock_read()
1310
 
            # if the basis has a root id we have to use that; otherwise we use
1311
 
            # a new random one
1312
 
            basis_root_id = basis.get_root_id()
1313
 
            if basis_root_id is not None:
1314
 
                wt._set_root_id(basis_root_id)
 
1433
            try:
 
1434
                wt.set_parent_trees(parents_list, allow_leftmost_as_ghost=True)
1315
1435
                wt.flush()
1316
 
            transform.build_tree(basis, wt)
1317
 
            basis.unlock()
 
1436
                # if the basis has a root id we have to use that; otherwise we
 
1437
                # use a new random one
 
1438
                basis_root_id = basis.get_root_id()
 
1439
                if basis_root_id is not None:
 
1440
                    wt._set_root_id(basis_root_id)
 
1441
                    wt.flush()
 
1442
                # If content filtering is supported, do not use the accelerator
 
1443
                # tree - the cost of transforming the content both ways and
 
1444
                # checking for changed content can outweight the gains it gives.
 
1445
                # Note: do NOT move this logic up higher - using the basis from
 
1446
                # the accelerator tree is still desirable because that can save
 
1447
                # a minute or more of processing on large trees!
 
1448
                # The original tree may not have the same content filters
 
1449
                # applied so we can't safely build the inventory delta from
 
1450
                # the source tree.
 
1451
                if wt.supports_content_filtering():
 
1452
                    if hardlink:
 
1453
                        # see https://bugs.edge.launchpad.net/bzr/+bug/408193
 
1454
                        trace.warning("hardlinking working copy files is not currently "
 
1455
                            "supported in %r" % (wt,))
 
1456
                    accelerator_tree = None
 
1457
                    delta_from_tree = False
 
1458
                else:
 
1459
                    delta_from_tree = True
 
1460
                # delta_from_tree is safe even for DirStateRevisionTrees,
 
1461
                # because wt4.apply_inventory_delta does not mutate the input
 
1462
                # inventory entries.
 
1463
                transform.build_tree(basis, wt, accelerator_tree,
 
1464
                                     hardlink=hardlink,
 
1465
                                     delta_from_tree=delta_from_tree)
 
1466
            finally:
 
1467
                basis.unlock()
1318
1468
        finally:
1319
1469
            control_files.unlock()
1320
1470
            wt.unlock()
1321
1471
        return wt
1322
1472
 
 
1473
    def _init_custom_control_files(self, wt):
 
1474
        """Subclasses with custom control files should override this method.
 
1475
 
 
1476
        The working tree and control files are locked for writing when this
 
1477
        method is called.
 
1478
 
 
1479
        :param wt: the WorkingTree object
 
1480
        """
 
1481
 
1323
1482
    def _open(self, a_bzrdir, control_files):
1324
1483
        """Open the tree itself.
1325
1484
 
1326
1485
        :param a_bzrdir: the dir for the tree.
1327
1486
        :param control_files: the control files for the tree.
1328
1487
        """
1329
 
        return WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
 
1488
        return self._tree_class(a_bzrdir.root_transport.local_abspath('.'),
1330
1489
                           branch=a_bzrdir.open_branch(),
1331
1490
                           _format=self,
1332
1491
                           _bzrdir=a_bzrdir,
1333
1492
                           _control_files=control_files)
1334
1493
 
1335
1494
    def __get_matchingbzrdir(self):
 
1495
        return self._get_matchingbzrdir()
 
1496
 
 
1497
    def _get_matchingbzrdir(self):
 
1498
        """Overrideable method to get a bzrdir for testing."""
1336
1499
        # please test against something that will let us do tree references
1337
1500
        return bzrdir.format_registry.make_bzrdir(
1338
1501
            'dirstate-with-subtree')
1340
1503
    _matchingbzrdir = property(__get_matchingbzrdir)
1341
1504
 
1342
1505
 
 
1506
class WorkingTreeFormat4(DirStateWorkingTreeFormat):
 
1507
    """The first consolidated dirstate working tree format.
 
1508
 
 
1509
    This format:
 
1510
        - exists within a metadir controlling .bzr
 
1511
        - includes an explicit version marker for the workingtree control
 
1512
          files, separate from the BzrDir format
 
1513
        - modifies the hash cache format
 
1514
        - is new in bzr 0.15
 
1515
        - uses a LockDir to guard access to it.
 
1516
    """
 
1517
 
 
1518
    upgrade_recommended = False
 
1519
 
 
1520
    _tree_class = WorkingTree4
 
1521
 
 
1522
    def get_format_string(self):
 
1523
        """See WorkingTreeFormat.get_format_string()."""
 
1524
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
 
1525
 
 
1526
    def get_format_description(self):
 
1527
        """See WorkingTreeFormat.get_format_description()."""
 
1528
        return "Working tree format 4"
 
1529
 
 
1530
 
 
1531
class WorkingTreeFormat5(DirStateWorkingTreeFormat):
 
1532
    """WorkingTree format supporting content filtering.
 
1533
    """
 
1534
 
 
1535
    upgrade_recommended = False
 
1536
 
 
1537
    _tree_class = WorkingTree5
 
1538
 
 
1539
    def get_format_string(self):
 
1540
        """See WorkingTreeFormat.get_format_string()."""
 
1541
        return "Bazaar Working Tree Format 5 (bzr 1.11)\n"
 
1542
 
 
1543
    def get_format_description(self):
 
1544
        """See WorkingTreeFormat.get_format_description()."""
 
1545
        return "Working tree format 5"
 
1546
 
 
1547
    def supports_content_filtering(self):
 
1548
        return True
 
1549
 
 
1550
 
 
1551
class WorkingTreeFormat6(DirStateWorkingTreeFormat):
 
1552
    """WorkingTree format supporting views.
 
1553
    """
 
1554
 
 
1555
    upgrade_recommended = False
 
1556
 
 
1557
    _tree_class = WorkingTree6
 
1558
 
 
1559
    def get_format_string(self):
 
1560
        """See WorkingTreeFormat.get_format_string()."""
 
1561
        return "Bazaar Working Tree Format 6 (bzr 1.14)\n"
 
1562
 
 
1563
    def get_format_description(self):
 
1564
        """See WorkingTreeFormat.get_format_description()."""
 
1565
        return "Working tree format 6"
 
1566
 
 
1567
    def _init_custom_control_files(self, wt):
 
1568
        """Subclasses with custom control files should override this method."""
 
1569
        wt._transport.put_bytes('views', '', mode=wt.bzrdir._get_file_mode())
 
1570
 
 
1571
    def supports_content_filtering(self):
 
1572
        return True
 
1573
 
 
1574
    def supports_views(self):
 
1575
        return True
 
1576
 
 
1577
 
1343
1578
class DirStateRevisionTree(Tree):
1344
 
    """A revision tree pulling the inventory from a dirstate."""
 
1579
    """A revision tree pulling the inventory from a dirstate.
 
1580
    
 
1581
    Note that this is one of the historical (ie revision) trees cached in the
 
1582
    dirstate for easy access, not the workingtree.
 
1583
    """
1345
1584
 
1346
1585
    def __init__(self, dirstate, revision_id, repository):
1347
1586
        self._dirstate = dirstate
1348
 
        self._revision_id = osutils.safe_revision_id(revision_id)
 
1587
        self._revision_id = revision_id
1349
1588
        self._repository = repository
1350
1589
        self._inventory = None
1351
1590
        self._locked = 0
1352
1591
        self._dirstate_locked = False
 
1592
        self._repo_supports_tree_reference = getattr(
 
1593
            repository._format, "supports_tree_reference",
 
1594
            False)
1353
1595
 
1354
1596
    def __repr__(self):
1355
1597
        return "<%s of %s in %s>" % \
1358
1600
    def annotate_iter(self, file_id,
1359
1601
                      default_revision=_mod_revision.CURRENT_REVISION):
1360
1602
        """See Tree.annotate_iter"""
1361
 
        w = self._get_weave(file_id)
1362
 
        return w.annotate_iter(self.inventory[file_id].revision)
 
1603
        text_key = (file_id, self.inventory[file_id].revision)
 
1604
        annotations = self._repository.texts.annotate(text_key)
 
1605
        return [(key[-1], line) for (key, line) in annotations]
1363
1606
 
1364
1607
    def _get_ancestors(self, default_revision):
1365
1608
        return set(self._repository.get_ancestry(self._revision_id,
1386
1629
    def get_root_id(self):
1387
1630
        return self.path2id('')
1388
1631
 
 
1632
    def id2path(self, file_id):
 
1633
        "Convert a file-id to a path."
 
1634
        entry = self._get_entry(file_id=file_id)
 
1635
        if entry == (None, None):
 
1636
            raise errors.NoSuchId(tree=self, file_id=file_id)
 
1637
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
 
1638
        return path_utf8.decode('utf8')
 
1639
 
 
1640
    def iter_references(self):
 
1641
        if not self._repo_supports_tree_reference:
 
1642
            # When the repo doesn't support references, we will have nothing to
 
1643
            # return
 
1644
            return iter([])
 
1645
        # Otherwise, fall back to the default implementation
 
1646
        return super(DirStateRevisionTree, self).iter_references()
 
1647
 
1389
1648
    def _get_parent_index(self):
1390
1649
        """Return the index in the dirstate referenced by this tree."""
1391
1650
        return self._dirstate.get_parent_ids().index(self._revision_id) + 1
1396
1655
        If either file_id or path is supplied, it is used as the key to lookup.
1397
1656
        If both are supplied, the fastest lookup is used, and an error is
1398
1657
        raised if they do not both point at the same row.
1399
 
        
 
1658
 
1400
1659
        :param file_id: An optional unicode file_id to be looked up.
1401
1660
        :param path: An optional unicode path to be looked up.
1402
1661
        :return: The dirstate row tuple for path/file_id, or (None, None)
1403
1662
        """
1404
1663
        if file_id is None and path is None:
1405
1664
            raise errors.BzrError('must supply file_id or path')
1406
 
        file_id = osutils.safe_file_id(file_id)
1407
1665
        if path is not None:
1408
1666
            path = path.encode('utf8')
1409
1667
        parent_index = self._get_parent_index()
1417
1675
 
1418
1676
        This is relatively expensive: we have to walk the entire dirstate.
1419
1677
        """
1420
 
        assert self._locked, 'cannot generate inventory of an unlocked '\
1421
 
            'dirstate revision tree'
 
1678
        if not self._locked:
 
1679
            raise AssertionError(
 
1680
                'cannot generate inventory of an unlocked '
 
1681
                'dirstate revision tree')
1422
1682
        # separate call for profiling - makes it clear where the costs are.
1423
1683
        self._dirstate._read_dirblocks_if_needed()
1424
 
        assert self._revision_id in self._dirstate.get_parent_ids(), \
1425
 
            'parent %s has disappeared from %s' % (
1426
 
            self._revision_id, self._dirstate.get_parent_ids())
 
1684
        if self._revision_id not in self._dirstate.get_parent_ids():
 
1685
            raise AssertionError(
 
1686
                'parent %s has disappeared from %s' % (
 
1687
                self._revision_id, self._dirstate.get_parent_ids()))
1427
1688
        parent_index = self._dirstate.get_parent_ids().index(self._revision_id) + 1
1428
1689
        # This is identical now to the WorkingTree _generate_inventory except
1429
1690
        # for the tree index use.
1430
1691
        root_key, current_entry = self._dirstate._get_entry(parent_index, path_utf8='')
1431
1692
        current_id = root_key[2]
1432
 
        assert current_entry[parent_index][0] == 'd'
 
1693
        if current_entry[parent_index][0] != 'd':
 
1694
            raise AssertionError()
1433
1695
        inv = Inventory(root_id=current_id, revision_id=self._revision_id)
1434
1696
        inv.root.revision = current_entry[parent_index][4]
1435
1697
        # Turn some things into local variables
1475
1737
                    raise AssertionError("cannot convert entry %r into an InventoryEntry"
1476
1738
                            % entry)
1477
1739
                # These checks cost us around 40ms on a 55k entry tree
1478
 
                assert file_id not in inv_byid
1479
 
                assert name_unicode not in parent_ie.children
 
1740
                if file_id in inv_byid:
 
1741
                    raise AssertionError('file_id %s already in'
 
1742
                        ' inventory as %s' % (file_id, inv_byid[file_id]))
 
1743
                if name_unicode in parent_ie.children:
 
1744
                    raise AssertionError('name %r already in parent'
 
1745
                        % (name_unicode,))
1480
1746
                inv_byid[file_id] = inv_entry
1481
1747
                parent_ie.children[name_unicode] = inv_entry
1482
1748
        self._inventory = inv
1502
1768
            return parent_details[1]
1503
1769
        return None
1504
1770
 
1505
 
    @symbol_versioning.deprecated_method(symbol_versioning.zero_ninety)
1506
 
    def get_weave(self, file_id):
1507
 
        return self._get_weave(file_id)
1508
 
 
1509
 
    def _get_weave(self, file_id):
1510
 
        return self._repository.weave_store.get_weave(file_id,
1511
 
                self._repository.get_transaction())
1512
 
 
1513
1771
    def get_file(self, file_id, path=None):
1514
1772
        return StringIO(self.get_file_text(file_id))
1515
1773
 
1516
 
    def get_file_lines(self, file_id):
1517
 
        ie = self.inventory[file_id]
1518
 
        return self._get_weave(file_id).get_lines(ie.revision)
1519
 
 
1520
1774
    def get_file_size(self, file_id):
 
1775
        """See Tree.get_file_size"""
1521
1776
        return self.inventory[file_id].text_size
1522
1777
 
1523
 
    def get_file_text(self, file_id):
1524
 
        return ''.join(self.get_file_lines(file_id))
 
1778
    def get_file_text(self, file_id, path=None):
 
1779
        _, content = list(self.iter_files_bytes([(file_id, None)]))[0]
 
1780
        return ''.join(content)
1525
1781
 
1526
1782
    def get_reference_revision(self, file_id, path=None):
1527
1783
        return self.inventory[file_id].reference_revision
1546
1802
        if entry[1][parent_index][0] != 'l':
1547
1803
            return None
1548
1804
        else:
1549
 
            # At present, none of the tree implementations supports non-ascii
1550
 
            # symlink targets. So we will just assume that the dirstate path is
1551
 
            # correct.
1552
 
            return entry[1][parent_index][1]
 
1805
            target = entry[1][parent_index][1]
 
1806
            target = target.decode('utf8')
 
1807
            return target
1553
1808
 
1554
1809
    def get_revision_id(self):
1555
1810
        """Return the revision id for this tree."""
1573
1828
        return bool(self.path2id(filename))
1574
1829
 
1575
1830
    def kind(self, file_id):
1576
 
        return self.inventory[file_id].kind
 
1831
        entry = self._get_entry(file_id=file_id)[1]
 
1832
        if entry is None:
 
1833
            raise errors.NoSuchId(tree=self, file_id=file_id)
 
1834
        return dirstate.DirState._minikind_to_kind[entry[1][0]]
 
1835
 
 
1836
    def stored_kind(self, file_id):
 
1837
        """See Tree.stored_kind"""
 
1838
        return self.kind(file_id)
 
1839
 
 
1840
    def path_content_summary(self, path):
 
1841
        """See Tree.path_content_summary."""
 
1842
        id = self.inventory.path2id(path)
 
1843
        if id is None:
 
1844
            return ('missing', None, None, None)
 
1845
        entry = self._inventory[id]
 
1846
        kind = entry.kind
 
1847
        if kind == 'file':
 
1848
            return (kind, entry.text_size, entry.executable, entry.text_sha1)
 
1849
        elif kind == 'symlink':
 
1850
            return (kind, None, None, entry.symlink_target)
 
1851
        else:
 
1852
            return (kind, None, None, None)
1577
1853
 
1578
1854
    def is_executable(self, file_id, path=None):
1579
1855
        ie = self.inventory[file_id]
1581
1857
            return None
1582
1858
        return ie.executable
1583
1859
 
1584
 
    def list_files(self, include_root=False):
 
1860
    def list_files(self, include_root=False, from_dir=None, recursive=True):
1585
1861
        # We use a standard implementation, because DirStateRevisionTree is
1586
1862
        # dealing with one of the parents of the current state
1587
1863
        inv = self._get_inventory()
1588
 
        entries = inv.iter_entries()
1589
 
        if self.inventory.root is not None and not include_root:
 
1864
        if from_dir is None:
 
1865
            from_dir_id = None
 
1866
        else:
 
1867
            from_dir_id = inv.path2id(from_dir)
 
1868
            if from_dir_id is None:
 
1869
                # Directory not versioned
 
1870
                return
 
1871
        entries = inv.iter_entries(from_dir=from_dir_id, recursive=recursive)
 
1872
        if inv.root is not None and not include_root and from_dir is None:
1590
1873
            entries.next()
1591
1874
        for path, entry in entries:
1592
1875
            yield path, 'V', entry.kind, entry.file_id, entry
1625
1908
                self._dirstate_locked = False
1626
1909
            self._repository.unlock()
1627
1910
 
 
1911
    @needs_read_lock
 
1912
    def supports_tree_reference(self):
 
1913
        return self._repo_supports_tree_reference
 
1914
 
1628
1915
    def walkdirs(self, prefix=""):
1629
1916
        # TODO: jam 20070215 This is the lazy way by using the RevisionTree
1630
 
        # implementation based on an inventory.  
 
1917
        # implementation based on an inventory.
1631
1918
        # This should be cleaned up to use the much faster Dirstate code
1632
1919
        # So for now, we just build up the parent inventory, and extract
1633
1920
        # it the same way RevisionTree does.
1662
1949
 
1663
1950
class InterDirStateTree(InterTree):
1664
1951
    """Fast path optimiser for changes_from with dirstate trees.
1665
 
    
1666
 
    This is used only when both trees are in the dirstate working file, and 
1667
 
    the source is any parent within the dirstate, and the destination is 
 
1952
 
 
1953
    This is used only when both trees are in the dirstate working file, and
 
1954
    the source is any parent within the dirstate, and the destination is
1668
1955
    the current working tree of the same dirstate.
1669
1956
    """
1670
1957
    # this could be generalized to allow comparisons between any trees in the
1683
1970
        target.set_parent_ids([revid])
1684
1971
        return target.basis_tree(), target
1685
1972
 
 
1973
    @classmethod
 
1974
    def make_source_parent_tree_python_dirstate(klass, test_case, source, target):
 
1975
        result = klass.make_source_parent_tree(source, target)
 
1976
        result[1]._iter_changes = dirstate.ProcessEntryPython
 
1977
        return result
 
1978
 
 
1979
    @classmethod
 
1980
    def make_source_parent_tree_compiled_dirstate(klass, test_case, source, target):
 
1981
        from bzrlib.tests.test__dirstate_helpers import \
 
1982
            CompiledDirstateHelpersFeature
 
1983
        if not CompiledDirstateHelpersFeature.available():
 
1984
            from bzrlib.tests import UnavailableFeature
 
1985
            raise UnavailableFeature(CompiledDirstateHelpersFeature)
 
1986
        from bzrlib._dirstate_helpers_pyx import ProcessEntryC
 
1987
        result = klass.make_source_parent_tree(source, target)
 
1988
        result[1]._iter_changes = ProcessEntryC
 
1989
        return result
 
1990
 
1686
1991
    _matching_from_tree_format = WorkingTreeFormat4()
1687
1992
    _matching_to_tree_format = WorkingTreeFormat4()
1688
 
    _test_mutable_trees_to_test_trees = make_source_parent_tree
1689
 
 
1690
 
    def _iter_changes(self, include_unchanged=False,
 
1993
 
 
1994
    @classmethod
 
1995
    def _test_mutable_trees_to_test_trees(klass, test_case, source, target):
 
1996
        # This method shouldn't be called, because we have python and C
 
1997
        # specific flavours.
 
1998
        raise NotImplementedError
 
1999
 
 
2000
    def iter_changes(self, include_unchanged=False,
1691
2001
                      specific_files=None, pb=None, extra_trees=[],
1692
2002
                      require_versioned=True, want_unversioned=False):
1693
2003
        """Return the changes from source to target.
1694
2004
 
1695
 
        :return: An iterator that yields tuples. See InterTree._iter_changes
 
2005
        :return: An iterator that yields tuples. See InterTree.iter_changes
1696
2006
            for details.
1697
2007
        :param specific_files: An optional list of file paths to restrict the
1698
2008
            comparison to. When mapping filenames to ids, all matches in all
1709
2019
            output. An unversioned file is defined as one with (False, False)
1710
2020
            for the versioned pair.
1711
2021
        """
1712
 
        utf8_decode = cache_utf8._utf8_decode
1713
 
        _minikind_to_kind = dirstate.DirState._minikind_to_kind
1714
 
        cmp_by_dirs = dirstate.cmp_by_dirs
1715
2022
        # NB: show_status depends on being able to pass in non-versioned files
1716
2023
        # and report them as unknown
1717
2024
        # TODO: handle extra trees in the dirstate.
1718
 
        # TODO: handle comparisons as an empty tree as a different special
1719
 
        # case? mbp 20070226
1720
 
        if extra_trees or (self.source._revision_id == NULL_REVISION):
 
2025
        if (extra_trees or specific_files == []):
1721
2026
            # we can't fast-path these cases (yet)
1722
 
            for f in super(InterDirStateTree, self)._iter_changes(
 
2027
            return super(InterDirStateTree, self).iter_changes(
1723
2028
                include_unchanged, specific_files, pb, extra_trees,
1724
 
                require_versioned, want_unversioned=want_unversioned):
1725
 
                yield f
1726
 
            return
 
2029
                require_versioned, want_unversioned=want_unversioned)
1727
2030
        parent_ids = self.target.get_parent_ids()
1728
 
        assert (self.source._revision_id in parent_ids), \
1729
 
                "revision {%s} is not stored in {%s}, but %s " \
1730
 
                "can only be used for trees stored in the dirstate" \
1731
 
                % (self.source._revision_id, self.target, self._iter_changes)
 
2031
        if not (self.source._revision_id in parent_ids
 
2032
                or self.source._revision_id == _mod_revision.NULL_REVISION):
 
2033
            raise AssertionError(
 
2034
                "revision {%s} is not stored in {%s}, but %s "
 
2035
                "can only be used for trees stored in the dirstate"
 
2036
                % (self.source._revision_id, self.target, self.iter_changes))
1732
2037
        target_index = 0
1733
 
        if self.source._revision_id == NULL_REVISION:
 
2038
        if self.source._revision_id == _mod_revision.NULL_REVISION:
1734
2039
            source_index = None
1735
2040
            indices = (target_index,)
1736
2041
        else:
1737
 
            assert (self.source._revision_id in parent_ids), \
1738
 
                "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
1739
 
                self.source._revision_id, parent_ids)
 
2042
            if not (self.source._revision_id in parent_ids):
 
2043
                raise AssertionError(
 
2044
                    "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
 
2045
                    self.source._revision_id, parent_ids))
1740
2046
            source_index = 1 + parent_ids.index(self.source._revision_id)
1741
 
            indices = (source_index,target_index)
 
2047
            indices = (source_index, target_index)
1742
2048
        # -- make all specific_files utf8 --
1743
2049
        if specific_files:
1744
2050
            specific_files_utf8 = set()
1745
2051
            for path in specific_files:
 
2052
                # Note, if there are many specific files, using cache_utf8
 
2053
                # would be good here.
1746
2054
                specific_files_utf8.add(path.encode('utf8'))
1747
2055
            specific_files = specific_files_utf8
1748
2056
        else:
1749
2057
            specific_files = set([''])
1750
2058
        # -- specific_files is now a utf8 path set --
 
2059
 
1751
2060
        # -- get the state object and prepare it.
1752
2061
        state = self.target.current_dirstate()
1753
2062
        state._read_dirblocks_if_needed()
1754
 
        def _entries_for_path(path):
1755
 
            """Return a list with all the entries that match path for all ids.
1756
 
            """
1757
 
            dirname, basename = os.path.split(path)
1758
 
            key = (dirname, basename, '')
1759
 
            block_index, present = state._find_block_index_from_key(key)
1760
 
            if not present:
1761
 
                # the block which should contain path is absent.
1762
 
                return []
1763
 
            result = []
1764
 
            block = state._dirblocks[block_index][1]
1765
 
            entry_index, _ = state._find_entry_index(key, block)
1766
 
            # we may need to look at multiple entries at this path: walk while the specific_files match.
1767
 
            while (entry_index < len(block) and
1768
 
                block[entry_index][0][0:2] == key[0:2]):
1769
 
                result.append(block[entry_index])
1770
 
                entry_index += 1
1771
 
            return result
1772
2063
        if require_versioned:
1773
2064
            # -- check all supplied paths are versioned in a search tree. --
1774
 
            all_versioned = True
 
2065
            not_versioned = []
1775
2066
            for path in specific_files:
1776
 
                path_entries = _entries_for_path(path)
 
2067
                path_entries = state._entries_for_path(path)
1777
2068
                if not path_entries:
1778
2069
                    # this specified path is not present at all: error
1779
 
                    all_versioned = False
1780
 
                    break
 
2070
                    not_versioned.append(path)
 
2071
                    continue
1781
2072
                found_versioned = False
1782
2073
                # for each id at this path
1783
2074
                for entry in path_entries:
1790
2081
                if not found_versioned:
1791
2082
                    # none of the indexes was not 'absent' at all ids for this
1792
2083
                    # path.
1793
 
                    all_versioned = False
1794
 
                    break
1795
 
            if not all_versioned:
1796
 
                raise errors.PathsNotVersionedError(specific_files)
 
2084
                    not_versioned.append(path)
 
2085
            if len(not_versioned) > 0:
 
2086
                raise errors.PathsNotVersionedError(not_versioned)
1797
2087
        # -- remove redundancy in supplied specific_files to prevent over-scanning --
1798
 
        search_specific_files = set()
1799
 
        for path in specific_files:
1800
 
            other_specific_files = specific_files.difference(set([path]))
1801
 
            if not osutils.is_inside_any(other_specific_files, path):
1802
 
                # this is a top level path, we must check it.
1803
 
                search_specific_files.add(path)
1804
 
        # sketch: 
1805
 
        # compare source_index and target_index at or under each element of search_specific_files.
1806
 
        # follow the following comparison table. Note that we only want to do diff operations when
1807
 
        # the target is fdl because thats when the walkdirs logic will have exposed the pathinfo 
1808
 
        # for the target.
1809
 
        # cases:
1810
 
        # 
1811
 
        # Source | Target | disk | action
1812
 
        #   r    | fdlt   |      | add source to search, add id path move and perform
1813
 
        #        |        |      | diff check on source-target
1814
 
        #   r    | fdlt   |  a   | dangling file that was present in the basis. 
1815
 
        #        |        |      | ???
1816
 
        #   r    |  a     |      | add source to search
1817
 
        #   r    |  a     |  a   | 
1818
 
        #   r    |  r     |      | this path is present in a non-examined tree, skip.
1819
 
        #   r    |  r     |  a   | this path is present in a non-examined tree, skip.
1820
 
        #   a    | fdlt   |      | add new id
1821
 
        #   a    | fdlt   |  a   | dangling locally added file, skip
1822
 
        #   a    |  a     |      | not present in either tree, skip
1823
 
        #   a    |  a     |  a   | not present in any tree, skip
1824
 
        #   a    |  r     |      | not present in either tree at this path, skip as it
1825
 
        #        |        |      | may not be selected by the users list of paths.
1826
 
        #   a    |  r     |  a   | not present in either tree at this path, skip as it
1827
 
        #        |        |      | may not be selected by the users list of paths.
1828
 
        #  fdlt  | fdlt   |      | content in both: diff them
1829
 
        #  fdlt  | fdlt   |  a   | deleted locally, but not unversioned - show as deleted ?
1830
 
        #  fdlt  |  a     |      | unversioned: output deleted id for now
1831
 
        #  fdlt  |  a     |  a   | unversioned and deleted: output deleted id
1832
 
        #  fdlt  |  r     |      | relocated in this tree, so add target to search.
1833
 
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
1834
 
        #        |        |      | this id at the other path.
1835
 
        #  fdlt  |  r     |  a   | relocated in this tree, so add target to search.
1836
 
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
1837
 
        #        |        |      | this id at the other path.
1838
 
 
1839
 
        # for all search_indexs in each path at or under each element of
1840
 
        # search_specific_files, if the detail is relocated: add the id, and add the
1841
 
        # relocated path as one to search if its not searched already. If the
1842
 
        # detail is not relocated, add the id.
1843
 
        searched_specific_files = set()
1844
 
        NULL_PARENT_DETAILS = dirstate.DirState.NULL_PARENT_DETAILS
1845
 
        # Using a list so that we can access the values and change them in
1846
 
        # nested scope. Each one is [path, file_id, entry]
1847
 
        last_source_parent = [None, None]
1848
 
        last_target_parent = [None, None]
 
2088
        search_specific_files = osutils.minimum_path_selection(specific_files)
1849
2089
 
1850
2090
        use_filesystem_for_exec = (sys.platform != 'win32')
1851
 
 
1852
 
        # Just a sentry, so that _process_entry can say that this
1853
 
        # record is handled, but isn't interesting to process (unchanged)
1854
 
        uninteresting = object()
1855
 
 
1856
 
 
1857
 
        old_dirname_to_file_id = {}
1858
 
        new_dirname_to_file_id = {}
1859
 
        # TODO: jam 20070516 - Avoid the _get_entry lookup overhead by
1860
 
        #       keeping a cache of directories that we have seen.
1861
 
 
1862
 
        def _process_entry(entry, path_info):
1863
 
            """Compare an entry and real disk to generate delta information.
1864
 
 
1865
 
            :param path_info: top_relpath, basename, kind, lstat, abspath for
1866
 
                the path of entry. If None, then the path is considered absent.
1867
 
                (Perhaps we should pass in a concrete entry for this ?)
1868
 
                Basename is returned as a utf8 string because we expect this
1869
 
                tuple will be ignored, and don't want to take the time to
1870
 
                decode.
1871
 
            :return: None if these don't match
1872
 
                     A tuple of information about the change, or
1873
 
                     the object 'uninteresting' if these match, but are
1874
 
                     basically identical.
1875
 
            """
1876
 
            if source_index is None:
1877
 
                source_details = NULL_PARENT_DETAILS
1878
 
            else:
1879
 
                source_details = entry[1][source_index]
1880
 
            target_details = entry[1][target_index]
1881
 
            target_minikind = target_details[0]
1882
 
            if path_info is not None and target_minikind in 'fdlt':
1883
 
                assert target_index == 0
1884
 
                link_or_sha1 = state.update_entry(entry, abspath=path_info[4],
1885
 
                                                  stat_value=path_info[3])
1886
 
                # The entry may have been modified by update_entry
1887
 
                target_details = entry[1][target_index]
1888
 
                target_minikind = target_details[0]
1889
 
            else:
1890
 
                link_or_sha1 = None
1891
 
            file_id = entry[0][2]
1892
 
            source_minikind = source_details[0]
1893
 
            if source_minikind in 'fdltr' and target_minikind in 'fdlt':
1894
 
                # claimed content in both: diff
1895
 
                #   r    | fdlt   |      | add source to search, add id path move and perform
1896
 
                #        |        |      | diff check on source-target
1897
 
                #   r    | fdlt   |  a   | dangling file that was present in the basis.
1898
 
                #        |        |      | ???
1899
 
                if source_minikind in 'r':
1900
 
                    # add the source to the search path to find any children it
1901
 
                    # has.  TODO ? : only add if it is a container ?
1902
 
                    if not osutils.is_inside_any(searched_specific_files,
1903
 
                                                 source_details[1]):
1904
 
                        search_specific_files.add(source_details[1])
1905
 
                    # generate the old path; this is needed for stating later
1906
 
                    # as well.
1907
 
                    old_path = source_details[1]
1908
 
                    old_dirname, old_basename = os.path.split(old_path)
1909
 
                    path = pathjoin(entry[0][0], entry[0][1])
1910
 
                    old_entry = state._get_entry(source_index,
1911
 
                                                 path_utf8=old_path)
1912
 
                    # update the source details variable to be the real
1913
 
                    # location.
1914
 
                    source_details = old_entry[1][source_index]
1915
 
                    source_minikind = source_details[0]
1916
 
                else:
1917
 
                    old_dirname = entry[0][0]
1918
 
                    old_basename = entry[0][1]
1919
 
                    old_path = path = None
1920
 
                if path_info is None:
1921
 
                    # the file is missing on disk, show as removed.
1922
 
                    content_change = True
1923
 
                    target_kind = None
1924
 
                    target_exec = False
1925
 
                else:
1926
 
                    # source and target are both versioned and disk file is present.
1927
 
                    target_kind = path_info[2]
1928
 
                    if target_kind == 'directory':
1929
 
                        if path is None:
1930
 
                            old_path = path = pathjoin(old_dirname, old_basename)
1931
 
                        new_dirname_to_file_id[path] = file_id
1932
 
                        if source_minikind != 'd':
1933
 
                            content_change = True
1934
 
                        else:
1935
 
                            # directories have no fingerprint
1936
 
                            content_change = False
1937
 
                        target_exec = False
1938
 
                    elif target_kind == 'file':
1939
 
                        if source_minikind != 'f':
1940
 
                            content_change = True
1941
 
                        else:
1942
 
                            # We could check the size, but we already have the
1943
 
                            # sha1 hash.
1944
 
                            content_change = (link_or_sha1 != source_details[1])
1945
 
                        # Target details is updated at update_entry time
1946
 
                        if use_filesystem_for_exec:
1947
 
                            # We don't need S_ISREG here, because we are sure
1948
 
                            # we are dealing with a file.
1949
 
                            target_exec = bool(stat.S_IEXEC & path_info[3].st_mode)
1950
 
                        else:
1951
 
                            target_exec = target_details[3]
1952
 
                    elif target_kind == 'symlink':
1953
 
                        if source_minikind != 'l':
1954
 
                            content_change = True
1955
 
                        else:
1956
 
                            content_change = (link_or_sha1 != source_details[1])
1957
 
                        target_exec = False
1958
 
                    elif target_kind == 'tree-reference':
1959
 
                        if source_minikind != 't':
1960
 
                            content_change = True
1961
 
                        else:
1962
 
                            content_change = False
1963
 
                        target_exec = False
1964
 
                    else:
1965
 
                        raise Exception, "unknown kind %s" % path_info[2]
1966
 
                if source_minikind == 'd':
1967
 
                    if path is None:
1968
 
                        old_path = path = pathjoin(old_dirname, old_basename)
1969
 
                    old_dirname_to_file_id[old_path] = file_id
1970
 
                # parent id is the entry for the path in the target tree
1971
 
                if old_dirname == last_source_parent[0]:
1972
 
                    source_parent_id = last_source_parent[1]
1973
 
                else:
1974
 
                    try:
1975
 
                        source_parent_id = old_dirname_to_file_id[old_dirname]
1976
 
                    except KeyError:
1977
 
                        source_parent_entry = state._get_entry(source_index,
1978
 
                                                               path_utf8=old_dirname)
1979
 
                        source_parent_id = source_parent_entry[0][2]
1980
 
                    if source_parent_id == entry[0][2]:
1981
 
                        # This is the root, so the parent is None
1982
 
                        source_parent_id = None
1983
 
                    else:
1984
 
                        last_source_parent[0] = old_dirname
1985
 
                        last_source_parent[1] = source_parent_id
1986
 
                new_dirname = entry[0][0]
1987
 
                if new_dirname == last_target_parent[0]:
1988
 
                    target_parent_id = last_target_parent[1]
1989
 
                else:
1990
 
                    try:
1991
 
                        target_parent_id = new_dirname_to_file_id[new_dirname]
1992
 
                    except KeyError:
1993
 
                        # TODO: We don't always need to do the lookup, because the
1994
 
                        #       parent entry will be the same as the source entry.
1995
 
                        target_parent_entry = state._get_entry(target_index,
1996
 
                                                               path_utf8=new_dirname)
1997
 
                        assert target_parent_entry != (None, None), (
1998
 
                            "Could not find target parent in wt: %s\nparent of: %s"
1999
 
                            % (new_dirname, entry))
2000
 
                        target_parent_id = target_parent_entry[0][2]
2001
 
                    if target_parent_id == entry[0][2]:
2002
 
                        # This is the root, so the parent is None
2003
 
                        target_parent_id = None
2004
 
                    else:
2005
 
                        last_target_parent[0] = new_dirname
2006
 
                        last_target_parent[1] = target_parent_id
2007
 
 
2008
 
                source_exec = source_details[3]
2009
 
                if (include_unchanged
2010
 
                    or content_change
2011
 
                    or source_parent_id != target_parent_id
2012
 
                    or old_basename != entry[0][1]
2013
 
                    or source_exec != target_exec
2014
 
                    ):
2015
 
                    if old_path is None:
2016
 
                        old_path = path = pathjoin(old_dirname, old_basename)
2017
 
                        old_path_u = utf8_decode(old_path)[0]
2018
 
                        path_u = old_path_u
2019
 
                    else:
2020
 
                        old_path_u = utf8_decode(old_path)[0]
2021
 
                        if old_path == path:
2022
 
                            path_u = old_path_u
2023
 
                        else:
2024
 
                            path_u = utf8_decode(path)[0]
2025
 
                    source_kind = _minikind_to_kind[source_minikind]
2026
 
                    return (entry[0][2],
2027
 
                           (old_path_u, path_u),
2028
 
                           content_change,
2029
 
                           (True, True),
2030
 
                           (source_parent_id, target_parent_id),
2031
 
                           (utf8_decode(old_basename)[0], utf8_decode(entry[0][1])[0]),
2032
 
                           (source_kind, target_kind),
2033
 
                           (source_exec, target_exec))
2034
 
                else:
2035
 
                    return uninteresting
2036
 
            elif source_minikind in 'a' and target_minikind in 'fdlt':
2037
 
                # looks like a new file
2038
 
                if path_info is not None:
2039
 
                    path = pathjoin(entry[0][0], entry[0][1])
2040
 
                    # parent id is the entry for the path in the target tree
2041
 
                    # TODO: these are the same for an entire directory: cache em.
2042
 
                    parent_id = state._get_entry(target_index,
2043
 
                                                 path_utf8=entry[0][0])[0][2]
2044
 
                    if parent_id == entry[0][2]:
2045
 
                        parent_id = None
2046
 
                    if use_filesystem_for_exec:
2047
 
                        # We need S_ISREG here, because we aren't sure if this
2048
 
                        # is a file or not.
2049
 
                        target_exec = bool(
2050
 
                            stat.S_ISREG(path_info[3].st_mode)
2051
 
                            and stat.S_IEXEC & path_info[3].st_mode)
2052
 
                    else:
2053
 
                        target_exec = target_details[3]
2054
 
                    return (entry[0][2],
2055
 
                           (None, utf8_decode(path)[0]),
2056
 
                           True,
2057
 
                           (False, True),
2058
 
                           (None, parent_id),
2059
 
                           (None, utf8_decode(entry[0][1])[0]),
2060
 
                           (None, path_info[2]),
2061
 
                           (None, target_exec))
2062
 
                else:
2063
 
                    # but its not on disk: we deliberately treat this as just
2064
 
                    # never-present. (Why ?! - RBC 20070224)
2065
 
                    pass
2066
 
            elif source_minikind in 'fdlt' and target_minikind in 'a':
2067
 
                # unversioned, possibly, or possibly not deleted: we dont care.
2068
 
                # if its still on disk, *and* theres no other entry at this
2069
 
                # path [we dont know this in this routine at the moment -
2070
 
                # perhaps we should change this - then it would be an unknown.
2071
 
                old_path = pathjoin(entry[0][0], entry[0][1])
2072
 
                # parent id is the entry for the path in the target tree
2073
 
                parent_id = state._get_entry(source_index, path_utf8=entry[0][0])[0][2]
2074
 
                if parent_id == entry[0][2]:
2075
 
                    parent_id = None
2076
 
                return (entry[0][2],
2077
 
                       (utf8_decode(old_path)[0], None),
2078
 
                       True,
2079
 
                       (True, False),
2080
 
                       (parent_id, None),
2081
 
                       (utf8_decode(entry[0][1])[0], None),
2082
 
                       (_minikind_to_kind[source_minikind], None),
2083
 
                       (source_details[3], None))
2084
 
            elif source_minikind in 'fdlt' and target_minikind in 'r':
2085
 
                # a rename; could be a true rename, or a rename inherited from
2086
 
                # a renamed parent. TODO: handle this efficiently. Its not
2087
 
                # common case to rename dirs though, so a correct but slow
2088
 
                # implementation will do.
2089
 
                if not osutils.is_inside_any(searched_specific_files, target_details[1]):
2090
 
                    search_specific_files.add(target_details[1])
2091
 
            elif source_minikind in 'ra' and target_minikind in 'ra':
2092
 
                # neither of the selected trees contain this file,
2093
 
                # so skip over it. This is not currently directly tested, but
2094
 
                # is indirectly via test_too_much.TestCommands.test_conflicts.
2095
 
                pass
2096
 
            else:
2097
 
                raise AssertionError("don't know how to compare "
2098
 
                    "source_minikind=%r, target_minikind=%r"
2099
 
                    % (source_minikind, target_minikind))
2100
 
                ## import pdb;pdb.set_trace()
2101
 
            return None
2102
 
 
2103
 
        while search_specific_files:
2104
 
            # TODO: the pending list should be lexically sorted?  the
2105
 
            # interface doesn't require it.
2106
 
            current_root = search_specific_files.pop()
2107
 
            current_root_unicode = current_root.decode('utf8')
2108
 
            searched_specific_files.add(current_root)
2109
 
            # process the entries for this containing directory: the rest will be
2110
 
            # found by their parents recursively.
2111
 
            root_entries = _entries_for_path(current_root)
2112
 
            root_abspath = self.target.abspath(current_root_unicode)
2113
 
            try:
2114
 
                root_stat = os.lstat(root_abspath)
2115
 
            except OSError, e:
2116
 
                if e.errno == errno.ENOENT:
2117
 
                    # the path does not exist: let _process_entry know that.
2118
 
                    root_dir_info = None
2119
 
                else:
2120
 
                    # some other random error: hand it up.
2121
 
                    raise
2122
 
            else:
2123
 
                root_dir_info = ('', current_root,
2124
 
                    osutils.file_kind_from_stat_mode(root_stat.st_mode), root_stat,
2125
 
                    root_abspath)
2126
 
                if root_dir_info[2] == 'directory':
2127
 
                    if self.target._directory_is_tree_reference(
2128
 
                        current_root.decode('utf8')):
2129
 
                        root_dir_info = root_dir_info[:2] + \
2130
 
                            ('tree-reference',) + root_dir_info[3:]
2131
 
 
2132
 
            if not root_entries and not root_dir_info:
2133
 
                # this specified path is not present at all, skip it.
2134
 
                continue
2135
 
            path_handled = False
2136
 
            for entry in root_entries:
2137
 
                result = _process_entry(entry, root_dir_info)
2138
 
                if result is not None:
2139
 
                    path_handled = True
2140
 
                    if result is not uninteresting:
2141
 
                        yield result
2142
 
            if want_unversioned and not path_handled and root_dir_info:
2143
 
                new_executable = bool(
2144
 
                    stat.S_ISREG(root_dir_info[3].st_mode)
2145
 
                    and stat.S_IEXEC & root_dir_info[3].st_mode)
2146
 
                yield (None,
2147
 
                       (None, current_root_unicode),
2148
 
                       True,
2149
 
                       (False, False),
2150
 
                       (None, None),
2151
 
                       (None, splitpath(current_root_unicode)[-1]),
2152
 
                       (None, root_dir_info[2]),
2153
 
                       (None, new_executable)
2154
 
                      )
2155
 
            initial_key = (current_root, '', '')
2156
 
            block_index, _ = state._find_block_index_from_key(initial_key)
2157
 
            if block_index == 0:
2158
 
                # we have processed the total root already, but because the
2159
 
                # initial key matched it we should skip it here.
2160
 
                block_index +=1
2161
 
            if root_dir_info and root_dir_info[2] == 'tree-reference':
2162
 
                current_dir_info = None
2163
 
            else:
2164
 
                dir_iterator = osutils._walkdirs_utf8(root_abspath, prefix=current_root)
2165
 
                try:
2166
 
                    current_dir_info = dir_iterator.next()
2167
 
                except OSError, e:
2168
 
                    # on win32, python2.4 has e.errno == ERROR_DIRECTORY, but
2169
 
                    # python 2.5 has e.errno == EINVAL,
2170
 
                    #            and e.winerror == ERROR_DIRECTORY
2171
 
                    e_winerror = getattr(e, 'winerror', None)
2172
 
                    win_errors = (ERROR_DIRECTORY, ERROR_PATH_NOT_FOUND)
2173
 
                    # there may be directories in the inventory even though
2174
 
                    # this path is not a file on disk: so mark it as end of
2175
 
                    # iterator
2176
 
                    if e.errno in (errno.ENOENT, errno.ENOTDIR, errno.EINVAL):
2177
 
                        current_dir_info = None
2178
 
                    elif (sys.platform == 'win32'
2179
 
                          and (e.errno in win_errors
2180
 
                               or e_winerror in win_errors)):
2181
 
                        current_dir_info = None
2182
 
                    else:
2183
 
                        raise
2184
 
                else:
2185
 
                    if current_dir_info[0][0] == '':
2186
 
                        # remove .bzr from iteration
2187
 
                        bzr_index = bisect_left(current_dir_info[1], ('.bzr',))
2188
 
                        assert current_dir_info[1][bzr_index][0] == '.bzr'
2189
 
                        del current_dir_info[1][bzr_index]
2190
 
            # walk until both the directory listing and the versioned metadata
2191
 
            # are exhausted. 
2192
 
            if (block_index < len(state._dirblocks) and
2193
 
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
2194
 
                current_block = state._dirblocks[block_index]
2195
 
            else:
2196
 
                current_block = None
2197
 
            while (current_dir_info is not None or
2198
 
                   current_block is not None):
2199
 
                if (current_dir_info and current_block
2200
 
                    and current_dir_info[0][0] != current_block[0]):
2201
 
                    if cmp_by_dirs(current_dir_info[0][0], current_block[0]) < 0:
2202
 
                        # filesystem data refers to paths not covered by the dirblock.
2203
 
                        # this has two possibilities:
2204
 
                        # A) it is versioned but empty, so there is no block for it
2205
 
                        # B) it is not versioned.
2206
 
 
2207
 
                        # if (A) then we need to recurse into it to check for
2208
 
                        # new unknown files or directories.
2209
 
                        # if (B) then we should ignore it, because we don't
2210
 
                        # recurse into unknown directories.
2211
 
                        path_index = 0
2212
 
                        while path_index < len(current_dir_info[1]):
2213
 
                                current_path_info = current_dir_info[1][path_index]
2214
 
                                if want_unversioned:
2215
 
                                    if current_path_info[2] == 'directory':
2216
 
                                        if self.target._directory_is_tree_reference(
2217
 
                                            current_path_info[0].decode('utf8')):
2218
 
                                            current_path_info = current_path_info[:2] + \
2219
 
                                                ('tree-reference',) + current_path_info[3:]
2220
 
                                    new_executable = bool(
2221
 
                                        stat.S_ISREG(current_path_info[3].st_mode)
2222
 
                                        and stat.S_IEXEC & current_path_info[3].st_mode)
2223
 
                                    yield (None,
2224
 
                                        (None, utf8_decode(current_path_info[0])[0]),
2225
 
                                        True,
2226
 
                                        (False, False),
2227
 
                                        (None, None),
2228
 
                                        (None, utf8_decode(current_path_info[1])[0]),
2229
 
                                        (None, current_path_info[2]),
2230
 
                                        (None, new_executable))
2231
 
                                # dont descend into this unversioned path if it is
2232
 
                                # a dir
2233
 
                                if current_path_info[2] in ('directory',
2234
 
                                                            'tree-reference'):
2235
 
                                    del current_dir_info[1][path_index]
2236
 
                                    path_index -= 1
2237
 
                                path_index += 1
2238
 
 
2239
 
                        # This dir info has been handled, go to the next
2240
 
                        try:
2241
 
                            current_dir_info = dir_iterator.next()
2242
 
                        except StopIteration:
2243
 
                            current_dir_info = None
2244
 
                    else:
2245
 
                        # We have a dirblock entry for this location, but there
2246
 
                        # is no filesystem path for this. This is most likely
2247
 
                        # because a directory was removed from the disk.
2248
 
                        # We don't have to report the missing directory,
2249
 
                        # because that should have already been handled, but we
2250
 
                        # need to handle all of the files that are contained
2251
 
                        # within.
2252
 
                        for current_entry in current_block[1]:
2253
 
                            # entry referring to file not present on disk.
2254
 
                            # advance the entry only, after processing.
2255
 
                            result = _process_entry(current_entry, None)
2256
 
                            if result is not None:
2257
 
                                if result is not uninteresting:
2258
 
                                    yield result
2259
 
                        block_index +=1
2260
 
                        if (block_index < len(state._dirblocks) and
2261
 
                            osutils.is_inside(current_root,
2262
 
                                              state._dirblocks[block_index][0])):
2263
 
                            current_block = state._dirblocks[block_index]
2264
 
                        else:
2265
 
                            current_block = None
2266
 
                    continue
2267
 
                entry_index = 0
2268
 
                if current_block and entry_index < len(current_block[1]):
2269
 
                    current_entry = current_block[1][entry_index]
2270
 
                else:
2271
 
                    current_entry = None
2272
 
                advance_entry = True
2273
 
                path_index = 0
2274
 
                if current_dir_info and path_index < len(current_dir_info[1]):
2275
 
                    current_path_info = current_dir_info[1][path_index]
2276
 
                    if current_path_info[2] == 'directory':
2277
 
                        if self.target._directory_is_tree_reference(
2278
 
                            current_path_info[0].decode('utf8')):
2279
 
                            current_path_info = current_path_info[:2] + \
2280
 
                                ('tree-reference',) + current_path_info[3:]
2281
 
                else:
2282
 
                    current_path_info = None
2283
 
                advance_path = True
2284
 
                path_handled = False
2285
 
                while (current_entry is not None or
2286
 
                    current_path_info is not None):
2287
 
                    if current_entry is None:
2288
 
                        # the check for path_handled when the path is adnvaced
2289
 
                        # will yield this path if needed.
2290
 
                        pass
2291
 
                    elif current_path_info is None:
2292
 
                        # no path is fine: the per entry code will handle it.
2293
 
                        result = _process_entry(current_entry, current_path_info)
2294
 
                        if result is not None:
2295
 
                            if result is not uninteresting:
2296
 
                                yield result
2297
 
                    elif (current_entry[0][1] != current_path_info[1]
2298
 
                          or current_entry[1][target_index][0] in 'ar'):
2299
 
                        # The current path on disk doesn't match the dirblock
2300
 
                        # record. Either the dirblock is marked as absent, or
2301
 
                        # the file on disk is not present at all in the
2302
 
                        # dirblock. Either way, report about the dirblock
2303
 
                        # entry, and let other code handle the filesystem one.
2304
 
 
2305
 
                        # Compare the basename for these files to determine
2306
 
                        # which comes first
2307
 
                        if current_path_info[1] < current_entry[0][1]:
2308
 
                            # extra file on disk: pass for now, but only
2309
 
                            # increment the path, not the entry
2310
 
                            advance_entry = False
2311
 
                        else:
2312
 
                            # entry referring to file not present on disk.
2313
 
                            # advance the entry only, after processing.
2314
 
                            result = _process_entry(current_entry, None)
2315
 
                            if result is not None:
2316
 
                                if result is not uninteresting:
2317
 
                                    yield result
2318
 
                            advance_path = False
2319
 
                    else:
2320
 
                        result = _process_entry(current_entry, current_path_info)
2321
 
                        if result is not None:
2322
 
                            path_handled = True
2323
 
                            if result is not uninteresting:
2324
 
                                yield result
2325
 
                    if advance_entry and current_entry is not None:
2326
 
                        entry_index += 1
2327
 
                        if entry_index < len(current_block[1]):
2328
 
                            current_entry = current_block[1][entry_index]
2329
 
                        else:
2330
 
                            current_entry = None
2331
 
                    else:
2332
 
                        advance_entry = True # reset the advance flaga
2333
 
                    if advance_path and current_path_info is not None:
2334
 
                        if not path_handled:
2335
 
                            # unversioned in all regards
2336
 
                            if want_unversioned:
2337
 
                                new_executable = bool(
2338
 
                                    stat.S_ISREG(current_path_info[3].st_mode)
2339
 
                                    and stat.S_IEXEC & current_path_info[3].st_mode)
2340
 
                                yield (None,
2341
 
                                    (None, utf8_decode(current_path_info[0])[0]),
2342
 
                                    True,
2343
 
                                    (False, False),
2344
 
                                    (None, None),
2345
 
                                    (None, utf8_decode(current_path_info[1])[0]),
2346
 
                                    (None, current_path_info[2]),
2347
 
                                    (None, new_executable))
2348
 
                            # dont descend into this unversioned path if it is
2349
 
                            # a dir
2350
 
                            if current_path_info[2] in ('directory'):
2351
 
                                del current_dir_info[1][path_index]
2352
 
                                path_index -= 1
2353
 
                        # dont descend the disk iterator into any tree 
2354
 
                        # paths.
2355
 
                        if current_path_info[2] == 'tree-reference':
2356
 
                            del current_dir_info[1][path_index]
2357
 
                            path_index -= 1
2358
 
                        path_index += 1
2359
 
                        if path_index < len(current_dir_info[1]):
2360
 
                            current_path_info = current_dir_info[1][path_index]
2361
 
                            if current_path_info[2] == 'directory':
2362
 
                                if self.target._directory_is_tree_reference(
2363
 
                                    current_path_info[0].decode('utf8')):
2364
 
                                    current_path_info = current_path_info[:2] + \
2365
 
                                        ('tree-reference',) + current_path_info[3:]
2366
 
                        else:
2367
 
                            current_path_info = None
2368
 
                        path_handled = False
2369
 
                    else:
2370
 
                        advance_path = True # reset the advance flagg.
2371
 
                if current_block is not None:
2372
 
                    block_index += 1
2373
 
                    if (block_index < len(state._dirblocks) and
2374
 
                        osutils.is_inside(current_root, state._dirblocks[block_index][0])):
2375
 
                        current_block = state._dirblocks[block_index]
2376
 
                    else:
2377
 
                        current_block = None
2378
 
                if current_dir_info is not None:
2379
 
                    try:
2380
 
                        current_dir_info = dir_iterator.next()
2381
 
                    except StopIteration:
2382
 
                        current_dir_info = None
2383
 
 
 
2091
        iter_changes = self.target._iter_changes(include_unchanged,
 
2092
            use_filesystem_for_exec, search_specific_files, state,
 
2093
            source_index, target_index, want_unversioned, self.target)
 
2094
        return iter_changes.iter_changes()
2384
2095
 
2385
2096
    @staticmethod
2386
2097
    def is_compatible(source, target):
2387
2098
        # the target must be a dirstate working tree
2388
 
        if not isinstance(target, WorkingTree4):
 
2099
        if not isinstance(target, DirStateWorkingTree):
2389
2100
            return False
2390
 
        # the source must be a revtreee or dirstate rev tree.
 
2101
        # the source must be a revtree or dirstate rev tree.
2391
2102
        if not isinstance(source,
2392
2103
            (revisiontree.RevisionTree, DirStateRevisionTree)):
2393
2104
            return False
2394
2105
        # the source revid must be in the target dirstate
2395
 
        if not (source._revision_id == NULL_REVISION or
 
2106
        if not (source._revision_id == _mod_revision.NULL_REVISION or
2396
2107
            source._revision_id in target.get_parent_ids()):
2397
 
            # TODO: what about ghosts? it may well need to 
 
2108
            # TODO: what about ghosts? it may well need to
2398
2109
            # check for them explicitly.
2399
2110
            return False
2400
2111
        return True
2410
2121
 
2411
2122
    def convert(self, tree):
2412
2123
        # lock the control files not the tree, so that we dont get tree
2413
 
        # on-unlock behaviours, and so that noone else diddles with the 
 
2124
        # on-unlock behaviours, and so that noone else diddles with the
2414
2125
        # tree during upgrade.
2415
2126
        tree._control_files.lock_write()
2416
2127
        try:
2442
2153
 
2443
2154
    def update_format(self, tree):
2444
2155
        """Change the format marker."""
2445
 
        tree._control_files.put_utf8('format',
2446
 
            self.target_format.get_format_string())
 
2156
        tree._transport.put_bytes('format',
 
2157
            self.target_format.get_format_string(),
 
2158
            mode=tree.bzrdir._get_file_mode())
 
2159
 
 
2160
 
 
2161
class Converter4to5(object):
 
2162
    """Perform an in-place upgrade of format 4 to format 5 trees."""
 
2163
 
 
2164
    def __init__(self):
 
2165
        self.target_format = WorkingTreeFormat5()
 
2166
 
 
2167
    def convert(self, tree):
 
2168
        # lock the control files not the tree, so that we don't get tree
 
2169
        # on-unlock behaviours, and so that no-one else diddles with the
 
2170
        # tree during upgrade.
 
2171
        tree._control_files.lock_write()
 
2172
        try:
 
2173
            self.update_format(tree)
 
2174
        finally:
 
2175
            tree._control_files.unlock()
 
2176
 
 
2177
    def update_format(self, tree):
 
2178
        """Change the format marker."""
 
2179
        tree._transport.put_bytes('format',
 
2180
            self.target_format.get_format_string(),
 
2181
            mode=tree.bzrdir._get_file_mode())
 
2182
 
 
2183
 
 
2184
class Converter4or5to6(object):
 
2185
    """Perform an in-place upgrade of format 4 or 5 to format 6 trees."""
 
2186
 
 
2187
    def __init__(self):
 
2188
        self.target_format = WorkingTreeFormat6()
 
2189
 
 
2190
    def convert(self, tree):
 
2191
        # lock the control files not the tree, so that we don't get tree
 
2192
        # on-unlock behaviours, and so that no-one else diddles with the
 
2193
        # tree during upgrade.
 
2194
        tree._control_files.lock_write()
 
2195
        try:
 
2196
            self.init_custom_control_files(tree)
 
2197
            self.update_format(tree)
 
2198
        finally:
 
2199
            tree._control_files.unlock()
 
2200
 
 
2201
    def init_custom_control_files(self, tree):
 
2202
        """Initialize custom control files."""
 
2203
        tree._transport.put_bytes('views', '',
 
2204
            mode=tree.bzrdir._get_file_mode())
 
2205
 
 
2206
    def update_format(self, tree):
 
2207
        """Change the format marker."""
 
2208
        tree._transport.put_bytes('format',
 
2209
            self.target_format.get_format_string(),
 
2210
            mode=tree.bzrdir._get_file_mode())