~bzr-pqm/bzr/bzr.dev

« back to all changes in this revision

Viewing changes to bzrlib/workingtree_4.py

  • Committer: Patch Queue Manager
  • Date: 2011-09-22 14:12:18 UTC
  • mfrom: (6155.3.1 jam)
  • Revision ID: pqm@pqm.ubuntu.com-20110922141218-86s4uu6nqvourw4f
(jameinel) Cleanup comments bzrlib/smart/__init__.py (John A Meinel)

Show diffs side-by-side

added added

removed removed

Lines of Context:
1
 
# Copyright (C) 2005, 2006, 2007 Canonical Ltd
 
1
# Copyright (C) 2007-2011 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
 
import bzrlib
42
34
from bzrlib import (
43
35
    bzrdir,
44
36
    cache_utf8,
 
37
    config,
45
38
    conflicts as _mod_conflicts,
46
 
    delta,
 
39
    debug,
47
40
    dirstate,
48
41
    errors,
 
42
    filters as _mod_filters,
49
43
    generate_ids,
50
 
    globbing,
51
 
    hashcache,
52
 
    ignores,
53
 
    merge,
54
44
    osutils,
 
45
    revision as _mod_revision,
55
46
    revisiontree,
56
 
    textui,
 
47
    trace,
57
48
    transform,
58
 
    urlutils,
59
 
    xml5,
60
 
    xml6,
 
49
    views,
61
50
    )
62
 
import bzrlib.branch
63
 
from bzrlib.transport import get_transport
64
 
import bzrlib.ui
65
51
""")
66
52
 
67
 
from bzrlib import symbol_versioning
68
53
from bzrlib.decorators import needs_read_lock, needs_write_lock
69
 
from bzrlib.inventory import InventoryEntry, Inventory, ROOT_ID, entry_factory
70
 
from bzrlib.lockable_files import LockableFiles, TransportLock
 
54
from bzrlib.inventory import Inventory, ROOT_ID, entry_factory
 
55
from bzrlib.lock import LogicalLockResult
 
56
from bzrlib.lockable_files import LockableFiles
71
57
from bzrlib.lockdir import LockDir
72
 
import bzrlib.mutabletree
73
58
from bzrlib.mutabletree import needs_tree_write_lock
74
59
from bzrlib.osutils import (
75
60
    file_kind,
76
61
    isdir,
77
 
    normpath,
78
62
    pathjoin,
79
 
    rand_chars,
80
63
    realpath,
81
64
    safe_unicode,
82
 
    splitpath,
83
65
    )
84
 
from bzrlib.trace import mutter, note
85
66
from bzrlib.transport.local import LocalTransport
86
 
from bzrlib.tree import InterTree
87
 
from bzrlib.progress import DummyProgress, ProgressPhase
88
 
from bzrlib.revision import NULL_REVISION, CURRENT_REVISION
89
 
from bzrlib.rio import RioReader, rio_file, Stanza
90
 
from bzrlib.symbol_versioning import (deprecated_passed,
91
 
        deprecated_method,
92
 
        deprecated_function,
93
 
        DEPRECATED_PARAMETER,
94
 
        )
95
 
from bzrlib.tree import Tree
96
 
from bzrlib.workingtree import WorkingTree, WorkingTree3, WorkingTreeFormat3
97
 
 
98
 
 
99
 
class WorkingTree4(WorkingTree3):
100
 
    """This is the Format 4 working tree.
101
 
 
102
 
    This differs from WorkingTree3 by:
103
 
     - Having a consolidated internal dirstate, stored in a
104
 
       randomly-accessible sorted file on disk.
105
 
     - Not having a regular inventory attribute.  One can be synthesized 
106
 
       on demand but this is expensive and should be avoided.
107
 
 
108
 
    This is new in bzr 0.15.
109
 
    """
 
67
from bzrlib.tree import (
 
68
    InterTree,
 
69
    InventoryTree,
 
70
    )
 
71
from bzrlib.workingtree import (
 
72
    InventoryWorkingTree,
 
73
    WorkingTree,
 
74
    WorkingTreeFormat,
 
75
    )
 
76
 
 
77
 
 
78
class DirStateWorkingTree(InventoryWorkingTree):
110
79
 
111
80
    def __init__(self, basedir,
112
81
                 branch,
122
91
        """
123
92
        self._format = _format
124
93
        self.bzrdir = _bzrdir
125
 
        from bzrlib.trace import note, mutter
126
 
        assert isinstance(basedir, basestring), \
127
 
            "base directory %r is not a string" % basedir
128
94
        basedir = safe_unicode(basedir)
129
 
        mutter("opening working tree %r", basedir)
 
95
        trace.mutter("opening working tree %r", basedir)
130
96
        self._branch = branch
131
 
        assert isinstance(self.branch, bzrlib.branch.Branch), \
132
 
            "branch %r is not a Branch" % self.branch
133
97
        self.basedir = realpath(basedir)
134
98
        # if branch is at our basedir and is a format 6 or less
135
99
        # assume all other formats have their own control files.
136
 
        assert isinstance(_control_files, LockableFiles), \
137
 
            "_control_files must be a LockableFiles, not %r" % _control_files
138
100
        self._control_files = _control_files
 
101
        self._transport = self._control_files._transport
139
102
        self._dirty = None
140
103
        #-------------
141
104
        # during a read or write lock these objects are set, and are
143
106
        self._dirstate = None
144
107
        self._inventory = None
145
108
        #-------------
 
109
        self._setup_directory_is_tree_reference()
 
110
        self._detect_case_handling()
 
111
        self._rules_searcher = None
 
112
        self.views = self._make_views()
 
113
        #--- allow tests to select the dirstate iter_changes implementation
 
114
        self._iter_changes = dirstate._process_entry
146
115
 
147
116
    @needs_tree_write_lock
148
117
    def _add(self, files, ids, kinds):
150
119
        state = self.current_dirstate()
151
120
        for f, file_id, kind in zip(files, ids, kinds):
152
121
            f = f.strip('/')
153
 
            assert '//' not in f
154
 
            assert '..' not in f
155
122
            if self.path2id(f):
156
123
                # special case tree root handling.
157
124
                if f == '' and self.path2id(f) == ROOT_ID:
165
132
            state.add(f, file_id, kind, None, '')
166
133
        self._make_dirty(reset_inventory=True)
167
134
 
 
135
    def _get_check_refs(self):
 
136
        """Return the references needed to perform a check of this tree."""
 
137
        return [('trees', self.last_revision())]
 
138
 
168
139
    def _make_dirty(self, reset_inventory):
169
140
        """Make the tree state dirty.
170
141
 
178
149
    @needs_tree_write_lock
179
150
    def add_reference(self, sub_tree):
180
151
        # use standard implementation, which calls back to self._add
181
 
        # 
 
152
        #
182
153
        # So we don't store the reference_revision in the working dirstate,
183
 
        # it's just recorded at the moment of commit. 
 
154
        # it's just recorded at the moment of commit.
184
155
        self._add_reference(sub_tree)
185
156
 
186
157
    def break_lock(self):
222
193
 
223
194
    def _comparison_data(self, entry, path):
224
195
        kind, executable, stat_value = \
225
 
            WorkingTree3._comparison_data(self, entry, path)
 
196
            WorkingTree._comparison_data(self, entry, path)
226
197
        # it looks like a plain directory, but it's really a reference -- see
227
198
        # also kind()
228
 
        if (self._repo_supports_tree_reference and
229
 
            kind == 'directory' and
230
 
            self._directory_is_tree_reference(path)):
 
199
        if (self._repo_supports_tree_reference and kind == 'directory'
 
200
            and entry is not None and entry.kind == 'tree-reference'):
231
201
            kind = 'tree-reference'
232
202
        return kind, executable, stat_value
233
203
 
235
205
    def commit(self, message=None, revprops=None, *args, **kwargs):
236
206
        # mark the tree as dirty post commit - commit
237
207
        # can change the current versioned list by doing deletes.
238
 
        result = WorkingTree3.commit(self, message, revprops, *args, **kwargs)
 
208
        result = WorkingTree.commit(self, message, revprops, *args, **kwargs)
239
209
        self._make_dirty(reset_inventory=True)
240
210
        return result
241
211
 
259
229
            return self._dirstate
260
230
        local_path = self.bzrdir.get_workingtree_transport(None
261
231
            ).local_abspath('dirstate')
262
 
        self._dirstate = dirstate.DirState.on_file(local_path)
 
232
        self._dirstate = dirstate.DirState.on_file(local_path,
 
233
            self._sha1_provider(), self._worth_saving_limit())
263
234
        return self._dirstate
264
235
 
265
 
    def _directory_is_tree_reference(self, relpath):
266
 
        # as a special case, if a directory contains control files then 
267
 
        # it's a tree reference, except that the root of the tree is not
268
 
        return relpath and osutils.isdir(self.abspath(relpath) + u"/.bzr")
269
 
        # TODO: We could ask all the control formats whether they
270
 
        # recognize this directory, but at the moment there's no cheap api
271
 
        # to do that.  Since we probably can only nest bzr checkouts and
272
 
        # they always use this name it's ok for now.  -- mbp 20060306
273
 
        #
274
 
        # FIXME: There is an unhandled case here of a subdirectory
275
 
        # containing .bzr but not a branch; that will probably blow up
276
 
        # when you try to commit it.  It might happen if there is a
277
 
        # checkout in a subdirectory.  This can be avoided by not adding
278
 
        # it.  mbp 20070306
 
236
    def _sha1_provider(self):
 
237
        """A function that returns a SHA1Provider suitable for this tree.
 
238
 
 
239
        :return: None if content filtering is not supported by this tree.
 
240
          Otherwise, a SHA1Provider is returned that sha's the canonical
 
241
          form of files, i.e. after read filters are applied.
 
242
        """
 
243
        if self.supports_content_filtering():
 
244
            return ContentFilterAwareSHA1Provider(self)
 
245
        else:
 
246
            return None
 
247
 
 
248
    def _worth_saving_limit(self):
 
249
        """How many hash changes are ok before we must save the dirstate.
 
250
 
 
251
        :return: an integer. -1 means never save.
 
252
        """
 
253
        # FIXME: We want a WorkingTreeStack here -- vila 20110812
 
254
        conf = config.BranchStack(self.branch)
 
255
        return conf.get('bzr.workingtree.worth_saving_limit')
279
256
 
280
257
    def filter_unversioned_files(self, paths):
281
258
        """Filter out paths that are versioned.
314
291
 
315
292
    def _generate_inventory(self):
316
293
        """Create and set self.inventory from the dirstate object.
317
 
        
 
294
 
318
295
        This is relatively expensive: we have to walk the entire dirstate.
319
296
        Ideally we would not, and can deprecate this function.
320
297
        """
324
301
        state._read_dirblocks_if_needed()
325
302
        root_key, current_entry = self._get_entry(path='')
326
303
        current_id = root_key[2]
327
 
        assert current_entry[0][0] == 'd' # directory
 
304
        if not (current_entry[0][0] == 'd'): # directory
 
305
            raise AssertionError(current_entry)
328
306
        inv = Inventory(root_id=current_id)
329
307
        # Turn some things into local variables
330
308
        minikind_to_kind = dirstate.DirState._minikind_to_kind
363
341
                    # add this entry to the parent map.
364
342
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
365
343
                elif kind == 'tree-reference':
366
 
                    assert self._repo_supports_tree_reference, \
367
 
                        "repository of %r " \
368
 
                        "doesn't support tree references " \
369
 
                        "required by entry %r" \
370
 
                        % (self, name)
 
344
                    if not self._repo_supports_tree_reference:
 
345
                        raise errors.UnsupportedOperation(
 
346
                            self._generate_inventory,
 
347
                            self.branch.repository)
371
348
                    inv_entry.reference_revision = link_or_sha1 or None
372
349
                elif kind != 'symlink':
373
350
                    raise AssertionError("unknown kind %r" % kind)
374
351
                # These checks cost us around 40ms on a 55k entry tree
375
 
                assert file_id not in inv_byid, ('file_id %s already in'
376
 
                    ' inventory as %s' % (file_id, inv_byid[file_id]))
377
 
                assert name_unicode not in parent_ie.children
 
352
                if file_id in inv_byid:
 
353
                    raise AssertionError('file_id %s already in'
 
354
                        ' inventory as %s' % (file_id, inv_byid[file_id]))
 
355
                if name_unicode in parent_ie.children:
 
356
                    raise AssertionError('name %r already in parent'
 
357
                        % (name_unicode,))
378
358
                inv_byid[file_id] = inv_entry
379
359
                parent_ie.children[name_unicode] = inv_entry
380
360
        self._inventory = inv
385
365
        If either file_id or path is supplied, it is used as the key to lookup.
386
366
        If both are supplied, the fastest lookup is used, and an error is
387
367
        raised if they do not both point at the same row.
388
 
        
 
368
 
389
369
        :param file_id: An optional unicode file_id to be looked up.
390
370
        :param path: An optional unicode path to be looked up.
391
371
        :return: The dirstate row tuple for path/file_id, or (None, None)
400
380
    def get_file_sha1(self, file_id, path=None, stat_value=None):
401
381
        # check file id is valid unconditionally.
402
382
        entry = self._get_entry(file_id=file_id, path=path)
403
 
        assert entry[0] is not None, 'what error should this raise'
 
383
        if entry[0] is None:
 
384
            raise errors.NoSuchId(self, file_id)
404
385
        if path is None:
405
386
            path = pathjoin(entry[0][0], entry[0][1]).decode('utf8')
406
387
 
407
388
        file_abspath = self.abspath(path)
408
389
        state = self.current_dirstate()
409
 
        link_or_sha1 = state.update_entry(entry, file_abspath,
410
 
                                          stat_value=stat_value)
 
390
        if stat_value is None:
 
391
            try:
 
392
                stat_value = osutils.lstat(file_abspath)
 
393
            except OSError, e:
 
394
                if e.errno == errno.ENOENT:
 
395
                    return None
 
396
                else:
 
397
                    raise
 
398
        link_or_sha1 = dirstate.update_entry(state, entry, file_abspath,
 
399
            stat_value=stat_value)
411
400
        if entry[1][0][0] == 'f':
412
 
            return link_or_sha1
 
401
            if link_or_sha1 is None:
 
402
                file_obj, statvalue = self.get_file_with_stat(file_id, path)
 
403
                try:
 
404
                    sha1 = osutils.sha_file(file_obj)
 
405
                finally:
 
406
                    file_obj.close()
 
407
                self._observed_sha1(file_id, path, (sha1, statvalue))
 
408
                return sha1
 
409
            else:
 
410
                return link_or_sha1
413
411
        return None
414
412
 
415
413
    def _get_inventory(self):
416
414
        """Get the inventory for the tree. This is only valid within a lock."""
 
415
        if 'evil' in debug.debug_flags:
 
416
            trace.mutter_callsite(2,
 
417
                "accessing .inventory forces a size of tree translation.")
417
418
        if self._inventory is not None:
418
419
            return self._inventory
419
420
        self._must_be_locked()
426
427
    @needs_read_lock
427
428
    def get_parent_ids(self):
428
429
        """See Tree.get_parent_ids.
429
 
        
 
430
 
430
431
        This implementation requests the ids list from the dirstate file.
431
432
        """
432
433
        return self.current_dirstate().get_parent_ids()
448
449
 
449
450
    def has_id(self, file_id):
450
451
        state = self.current_dirstate()
451
 
        file_id = osutils.safe_file_id(file_id)
452
452
        row, parents = self._get_entry(file_id=file_id)
453
453
        if row is None:
454
454
            return False
455
455
        return osutils.lexists(pathjoin(
456
456
                    self.basedir, row[0].decode('utf8'), row[1].decode('utf8')))
457
457
 
 
458
    def has_or_had_id(self, file_id):
 
459
        state = self.current_dirstate()
 
460
        row, parents = self._get_entry(file_id=file_id)
 
461
        return row is not None
 
462
 
458
463
    @needs_read_lock
459
464
    def id2path(self, file_id):
460
 
        file_id = osutils.safe_file_id(file_id)
 
465
        "Convert a file-id to a path."
461
466
        state = self.current_dirstate()
462
467
        entry = self._get_entry(file_id=file_id)
463
468
        if entry == (None, None):
465
470
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
466
471
        return path_utf8.decode('utf8')
467
472
 
 
473
    def _is_executable_from_path_and_stat_from_basis(self, path, stat_result):
 
474
        entry = self._get_entry(path=path)
 
475
        if entry == (None, None):
 
476
            return False # Missing entries are not executable
 
477
        return entry[1][0][3] # Executable?
 
478
 
468
479
    if not osutils.supports_executable():
469
 
        @needs_read_lock
470
480
        def is_executable(self, file_id, path=None):
471
 
            file_id = osutils.safe_file_id(file_id)
 
481
            """Test if a file is executable or not.
 
482
 
 
483
            Note: The caller is expected to take a read-lock before calling this.
 
484
            """
472
485
            entry = self._get_entry(file_id=file_id, path=path)
473
486
            if entry == (None, None):
474
487
                return False
475
488
            return entry[1][0][3]
 
489
 
 
490
        _is_executable_from_path_and_stat = \
 
491
            _is_executable_from_path_and_stat_from_basis
476
492
    else:
477
 
        @needs_read_lock
478
493
        def is_executable(self, file_id, path=None):
 
494
            """Test if a file is executable or not.
 
495
 
 
496
            Note: The caller is expected to take a read-lock before calling this.
 
497
            """
 
498
            self._must_be_locked()
479
499
            if not path:
480
 
                file_id = osutils.safe_file_id(file_id)
481
500
                path = self.id2path(file_id)
482
 
            mode = os.lstat(self.abspath(path)).st_mode
 
501
            mode = osutils.lstat(self.abspath(path)).st_mode
483
502
            return bool(stat.S_ISREG(mode) and stat.S_IEXEC & mode)
484
503
 
 
504
    def all_file_ids(self):
 
505
        """See Tree.iter_all_file_ids"""
 
506
        self._must_be_locked()
 
507
        result = set()
 
508
        for key, tree_details in self.current_dirstate()._iter_entries():
 
509
            if tree_details[0][0] in ('a', 'r'): # relocated
 
510
                continue
 
511
            result.add(key[2])
 
512
        return result
 
513
 
485
514
    @needs_read_lock
486
515
    def __iter__(self):
487
516
        """Iterate through file_ids for this tree.
500
529
        return iter(result)
501
530
 
502
531
    def iter_references(self):
 
532
        if not self._repo_supports_tree_reference:
 
533
            # When the repo doesn't support references, we will have nothing to
 
534
            # return
 
535
            return
503
536
        for key, tree_details in self.current_dirstate()._iter_entries():
504
537
            if tree_details[0][0] in ('a', 'r'): # absent, relocated
505
538
                # not relevant to the working tree
507
540
            if not key[1]:
508
541
                # the root is not a reference.
509
542
                continue
510
 
            path = pathjoin(self.basedir, key[0].decode('utf8'), key[1].decode('utf8'))
 
543
            relpath = pathjoin(key[0].decode('utf8'), key[1].decode('utf8'))
511
544
            try:
512
 
                if self._kind(path) == 'tree-reference':
513
 
                    yield path, key[2]
 
545
                if self._kind(relpath) == 'tree-reference':
 
546
                    yield relpath, key[2]
514
547
            except errors.NoSuchFile:
515
548
                # path is missing on disk.
516
549
                continue
517
550
 
518
 
    @needs_read_lock
 
551
    def _observed_sha1(self, file_id, path, (sha1, statvalue)):
 
552
        """See MutableTree._observed_sha1."""
 
553
        state = self.current_dirstate()
 
554
        entry = self._get_entry(file_id=file_id, path=path)
 
555
        state._observed_sha1(entry, sha1, statvalue)
 
556
 
519
557
    def kind(self, file_id):
520
558
        """Return the kind of a file.
521
559
 
522
560
        This is always the actual kind that's on disk, regardless of what it
523
561
        was added as.
 
562
 
 
563
        Note: The caller is expected to take a read-lock before calling this.
524
564
        """
525
565
        relpath = self.id2path(file_id)
526
 
        assert relpath != None, \
527
 
            "path for id {%s} is None!" % file_id
 
566
        if relpath is None:
 
567
            raise AssertionError(
 
568
                "path for id {%s} is None!" % file_id)
528
569
        return self._kind(relpath)
529
570
 
530
571
    def _kind(self, relpath):
531
572
        abspath = self.abspath(relpath)
532
573
        kind = file_kind(abspath)
533
 
        if (self._repo_supports_tree_reference and
534
 
            kind == 'directory' and
535
 
            self._directory_is_tree_reference(relpath)):
536
 
            kind = 'tree-reference'
 
574
        if (self._repo_supports_tree_reference and kind == 'directory'):
 
575
            entry = self._get_entry(path=relpath)
 
576
            if entry[1] is not None:
 
577
                if entry[1][0][0] == 't':
 
578
                    kind = 'tree-reference'
537
579
        return kind
538
580
 
539
581
    @needs_read_lock
543
585
        if parent_ids:
544
586
            return parent_ids[0]
545
587
        else:
546
 
            return None
 
588
            return _mod_revision.NULL_REVISION
547
589
 
548
590
    def lock_read(self):
549
 
        """See Branch.lock_read, and WorkingTree.unlock."""
 
591
        """See Branch.lock_read, and WorkingTree.unlock.
 
592
 
 
593
        :return: A bzrlib.lock.LogicalLockResult.
 
594
        """
550
595
        self.branch.lock_read()
551
596
        try:
552
597
            self._control_files.lock_read()
557
602
                # set our support for tree references from the repository in
558
603
                # use.
559
604
                self._repo_supports_tree_reference = getattr(
560
 
                    self.branch.repository._format, "support_tree_reference",
 
605
                    self.branch.repository._format, "supports_tree_reference",
561
606
                    False)
562
607
            except:
563
608
                self._control_files.unlock()
565
610
        except:
566
611
            self.branch.unlock()
567
612
            raise
 
613
        return LogicalLockResult(self.unlock)
568
614
 
569
615
    def _lock_self_write(self):
570
616
        """This should be called after the branch is locked."""
577
623
                # set our support for tree references from the repository in
578
624
                # use.
579
625
                self._repo_supports_tree_reference = getattr(
580
 
                    self.branch.repository._format, "support_tree_reference",
 
626
                    self.branch.repository._format, "supports_tree_reference",
581
627
                    False)
582
628
            except:
583
629
                self._control_files.unlock()
585
631
        except:
586
632
            self.branch.unlock()
587
633
            raise
 
634
        return LogicalLockResult(self.unlock)
588
635
 
589
636
    def lock_tree_write(self):
590
 
        """See MutableTree.lock_tree_write, and WorkingTree.unlock."""
 
637
        """See MutableTree.lock_tree_write, and WorkingTree.unlock.
 
638
 
 
639
        :return: A bzrlib.lock.LogicalLockResult.
 
640
        """
591
641
        self.branch.lock_read()
592
 
        self._lock_self_write()
 
642
        return self._lock_self_write()
593
643
 
594
644
    def lock_write(self):
595
 
        """See MutableTree.lock_write, and WorkingTree.unlock."""
 
645
        """See MutableTree.lock_write, and WorkingTree.unlock.
 
646
 
 
647
        :return: A bzrlib.lock.LogicalLockResult.
 
648
        """
596
649
        self.branch.lock_write()
597
 
        self._lock_self_write()
 
650
        return self._lock_self_write()
598
651
 
599
652
    @needs_tree_write_lock
600
653
    def move(self, from_paths, to_dir, after=False):
602
655
        result = []
603
656
        if not from_paths:
604
657
            return result
605
 
 
606
658
        state = self.current_dirstate()
607
 
 
608
 
        assert not isinstance(from_paths, basestring)
 
659
        if isinstance(from_paths, basestring):
 
660
            raise ValueError()
609
661
        to_dir_utf8 = to_dir.encode('utf8')
610
662
        to_entry_dirname, to_basename = os.path.split(to_dir_utf8)
611
663
        id_index = state._get_id_index()
633
685
        if self._inventory is not None:
634
686
            update_inventory = True
635
687
            inv = self.inventory
 
688
            to_dir_id = to_entry[0][2]
636
689
            to_dir_ie = inv[to_dir_id]
637
 
            to_dir_id = to_entry[0][2]
638
690
        else:
639
691
            update_inventory = False
640
692
 
663
715
            new_entry = to_block[1][added_entry_index]
664
716
            rollbacks.append(lambda:state._make_absent(new_entry))
665
717
 
666
 
        # create rename entries and tuples
667
718
        for from_rel in from_paths:
668
719
            # from_rel is 'pathinroot/foo/bar'
669
720
            from_rel_utf8 = from_rel.encode('utf8')
672
723
            from_entry = self._get_entry(path=from_rel)
673
724
            if from_entry == (None, None):
674
725
                raise errors.BzrMoveFailedError(from_rel,to_dir,
675
 
                    errors.NotVersionedError(path=str(from_rel)))
 
726
                    errors.NotVersionedError(path=from_rel))
676
727
 
677
728
            from_id = from_entry[0][2]
678
729
            to_rel = pathjoin(to_dir, from_tail)
705
756
                if from_missing: # implicitly just update our path mapping
706
757
                    move_file = False
707
758
                elif not after:
708
 
                    raise errors.RenameFailedFilesExist(from_rel, to_rel,
709
 
                        extra="(Use --after to update the Bazaar id)")
 
759
                    raise errors.RenameFailedFilesExist(from_rel, to_rel)
710
760
 
711
761
            rollbacks = []
712
762
            def rollback_rename():
767
817
 
768
818
                if minikind == 'd':
769
819
                    def update_dirblock(from_dir, to_key, to_dir_utf8):
770
 
                        """all entries in this block need updating.
771
 
 
772
 
                        TODO: This is pretty ugly, and doesn't support
773
 
                        reverting, but it works.
774
 
                        """
775
 
                        assert from_dir != '', "renaming root not supported"
 
820
                        """Recursively update all entries in this dirblock."""
 
821
                        if from_dir == '':
 
822
                            raise AssertionError("renaming root not supported")
776
823
                        from_key = (from_dir, '')
777
824
                        from_block_idx, present = \
778
825
                            state._find_block_index_from_key(from_key)
788
835
                        to_block_index = state._ensure_block(
789
836
                            to_block_index, to_entry_index, to_dir_utf8)
790
837
                        to_block = state._dirblocks[to_block_index]
791
 
                        for entry in from_block[1]:
792
 
                            assert entry[0][0] == from_dir
 
838
 
 
839
                        # Grab a copy since move_one may update the list.
 
840
                        for entry in from_block[1][:]:
 
841
                            if not (entry[0][0] == from_dir):
 
842
                                raise AssertionError()
793
843
                            cur_details = entry[1][0]
794
844
                            to_key = (to_dir_utf8, entry[0][1], entry[0][2])
795
845
                            from_path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
796
846
                            to_path_utf8 = osutils.pathjoin(to_dir_utf8, entry[0][1])
797
847
                            minikind = cur_details[0]
 
848
                            if minikind in 'ar':
 
849
                                # Deleted children of a renamed directory
 
850
                                # Do not need to be updated.
 
851
                                # Children that have been renamed out of this
 
852
                                # directory should also not be updated
 
853
                                continue
798
854
                            move_one(entry, from_path_utf8=from_path_utf8,
799
855
                                     minikind=minikind,
800
856
                                     executable=cur_details[3],
803
859
                                     size=cur_details[2],
804
860
                                     to_block=to_block,
805
861
                                     to_key=to_key,
806
 
                                     to_path_utf8=to_rel_utf8)
 
862
                                     to_path_utf8=to_path_utf8)
807
863
                            if minikind == 'd':
808
864
                                # We need to move all the children of this
809
865
                                # entry
814
870
                rollback_rename()
815
871
                raise
816
872
            result.append((from_rel, to_rel))
817
 
            state._dirblock_state = dirstate.DirState.IN_MEMORY_MODIFIED
 
873
            state._mark_modified()
818
874
            self._make_dirty(reset_inventory=False)
819
875
 
820
876
        return result
848
904
        for tree in trees:
849
905
            if not (isinstance(tree, DirStateRevisionTree) and tree._revision_id in
850
906
                parents):
851
 
                return super(WorkingTree4, self).paths2ids(paths, trees, require_versioned)
 
907
                return super(DirStateWorkingTree, self).paths2ids(paths,
 
908
                    trees, require_versioned)
852
909
        search_indexes = [0] + [1 + parents.index(tree._revision_id) for tree in trees]
853
910
        # -- make all paths utf8 --
854
911
        paths_utf8 = set()
914
971
            if not all_versioned:
915
972
                raise errors.PathsNotVersionedError(paths)
916
973
        # -- remove redundancy in supplied paths to prevent over-scanning --
917
 
        search_paths = set()
918
 
        for path in paths:
919
 
            other_paths = paths.difference(set([path]))
920
 
            if not osutils.is_inside_any(other_paths, path):
921
 
                # this is a top level path, we must check it.
922
 
                search_paths.add(path)
923
 
        # sketch: 
 
974
        search_paths = osutils.minimum_path_selection(paths)
 
975
        # sketch:
924
976
        # for all search_indexs in each path at or under each element of
925
977
        # search_paths, if the detail is relocated: add the id, and add the
926
978
        # relocated path as one to search if its not searched already. If the
982
1034
 
983
1035
    def read_working_inventory(self):
984
1036
        """Read the working inventory.
985
 
        
 
1037
 
986
1038
        This is a meaningless operation for dirstate, but we obey it anyhow.
987
1039
        """
988
1040
        return self.inventory
993
1045
 
994
1046
        WorkingTree4 supplies revision_trees for any basis tree.
995
1047
        """
996
 
        revision_id = osutils.safe_revision_id(revision_id)
997
1048
        dirstate = self.current_dirstate()
998
1049
        parent_ids = dirstate.get_parent_ids()
999
1050
        if revision_id not in parent_ids:
1006
1057
    @needs_tree_write_lock
1007
1058
    def set_last_revision(self, new_revision):
1008
1059
        """Change the last revision in the working tree."""
1009
 
        new_revision = osutils.safe_revision_id(new_revision)
1010
1060
        parents = self.get_parent_ids()
1011
 
        if new_revision in (NULL_REVISION, None):
1012
 
            assert len(parents) < 2, (
1013
 
                "setting the last parent to none with a pending merge is "
1014
 
                "unsupported.")
 
1061
        if new_revision in (_mod_revision.NULL_REVISION, None):
 
1062
            if len(parents) >= 2:
 
1063
                raise AssertionError(
 
1064
                    "setting the last parent to none with a pending merge is "
 
1065
                    "unsupported.")
1015
1066
            self.set_parent_ids([])
1016
1067
        else:
1017
1068
            self.set_parent_ids([new_revision] + parents[1:],
1020
1071
    @needs_tree_write_lock
1021
1072
    def set_parent_ids(self, revision_ids, allow_leftmost_as_ghost=False):
1022
1073
        """Set the parent ids to revision_ids.
1023
 
        
 
1074
 
1024
1075
        See also set_parent_trees. This api will try to retrieve the tree data
1025
1076
        for each element of revision_ids from the trees repository. If you have
1026
1077
        tree data already available, it is more efficient to use
1030
1081
        :param revision_ids: The revision_ids to set as the parent ids of this
1031
1082
            working tree. Any of these may be ghosts.
1032
1083
        """
1033
 
        revision_ids = [osutils.safe_revision_id(r) for r in revision_ids]
1034
1084
        trees = []
1035
1085
        for revision_id in revision_ids:
1036
1086
            try:
1042
1092
            except (errors.NoSuchRevision, errors.RevisionNotPresent):
1043
1093
                revtree = None
1044
1094
            trees.append((revision_id, revtree))
1045
 
        self.current_dirstate()._validate()
1046
1095
        self.set_parent_trees(trees,
1047
1096
            allow_leftmost_as_ghost=allow_leftmost_as_ghost)
1048
 
        self.current_dirstate()._validate()
1049
1097
 
1050
1098
    @needs_tree_write_lock
1051
1099
    def set_parent_trees(self, parents_list, allow_leftmost_as_ghost=False):
1056
1104
            parent tree - i.e. a ghost.
1057
1105
        """
1058
1106
        dirstate = self.current_dirstate()
1059
 
        dirstate._validate()
1060
1107
        if len(parents_list) > 0:
1061
1108
            if not allow_leftmost_as_ghost and parents_list[0][1] is None:
1062
1109
                raise errors.GhostRevisionUnusableHere(parents_list[0][0])
1063
1110
        real_trees = []
1064
1111
        ghosts = []
 
1112
 
 
1113
        parent_ids = [rev_id for rev_id, tree in parents_list]
 
1114
        graph = self.branch.repository.get_graph()
 
1115
        heads = graph.heads(parent_ids)
 
1116
        accepted_revisions = set()
 
1117
 
1065
1118
        # convert absent trees to the null tree, which we convert back to
1066
1119
        # missing on access.
1067
1120
        for rev_id, tree in parents_list:
1068
 
            rev_id = osutils.safe_revision_id(rev_id)
 
1121
            if len(accepted_revisions) > 0:
 
1122
                # we always accept the first tree
 
1123
                if rev_id in accepted_revisions or rev_id not in heads:
 
1124
                    # We have already included either this tree, or its
 
1125
                    # descendent, so we skip it.
 
1126
                    continue
 
1127
            _mod_revision.check_not_reserved_id(rev_id)
1069
1128
            if tree is not None:
1070
1129
                real_trees.append((rev_id, tree))
1071
1130
            else:
1072
1131
                real_trees.append((rev_id,
1073
 
                    self.branch.repository.revision_tree(None)))
 
1132
                    self.branch.repository.revision_tree(
 
1133
                        _mod_revision.NULL_REVISION)))
1074
1134
                ghosts.append(rev_id)
1075
 
        dirstate._validate()
1076
 
        dirstate.set_parent_trees(real_trees, ghosts=ghosts)
1077
 
        dirstate._validate()
 
1135
            accepted_revisions.add(rev_id)
 
1136
        updated = False
 
1137
        if (len(real_trees) == 1
 
1138
            and not ghosts
 
1139
            and self.branch.repository._format.fast_deltas
 
1140
            and isinstance(real_trees[0][1],
 
1141
                revisiontree.InventoryRevisionTree)
 
1142
            and self.get_parent_ids()):
 
1143
            rev_id, rev_tree = real_trees[0]
 
1144
            basis_id = self.get_parent_ids()[0]
 
1145
            # There are times when basis_tree won't be in
 
1146
            # self.branch.repository, (switch, for example)
 
1147
            try:
 
1148
                basis_tree = self.branch.repository.revision_tree(basis_id)
 
1149
            except errors.NoSuchRevision:
 
1150
                # Fall back to the set_parent_trees(), since we can't use
 
1151
                # _make_delta if we can't get the RevisionTree
 
1152
                pass
 
1153
            else:
 
1154
                delta = rev_tree.inventory._make_delta(basis_tree.inventory)
 
1155
                dirstate.update_basis_by_delta(delta, rev_id)
 
1156
                updated = True
 
1157
        if not updated:
 
1158
            dirstate.set_parent_trees(real_trees, ghosts=ghosts)
1078
1159
        self._make_dirty(reset_inventory=False)
1079
 
        dirstate._validate()
1080
1160
 
1081
1161
    def _set_root_id(self, file_id):
1082
1162
        """See WorkingTree.set_root_id."""
1085
1165
        if state._dirblock_state == dirstate.DirState.IN_MEMORY_MODIFIED:
1086
1166
            self._make_dirty(reset_inventory=True)
1087
1167
 
 
1168
    def _sha_from_stat(self, path, stat_result):
 
1169
        """Get a sha digest from the tree's stat cache.
 
1170
 
 
1171
        The default implementation assumes no stat cache is present.
 
1172
 
 
1173
        :param path: The path.
 
1174
        :param stat_result: The stat result being looked up.
 
1175
        """
 
1176
        return self.current_dirstate().sha1_from_stat(path, stat_result)
 
1177
 
1088
1178
    @needs_read_lock
1089
1179
    def supports_tree_reference(self):
1090
1180
        return self._repo_supports_tree_reference
1092
1182
    def unlock(self):
1093
1183
        """Unlock in format 4 trees needs to write the entire dirstate."""
1094
1184
        if self._control_files._lock_count == 1:
 
1185
            # do non-implementation specific cleanup
 
1186
            self._cleanup()
 
1187
 
1095
1188
            # eventually we should do signature checking during read locks for
1096
1189
            # dirstate updates.
1097
1190
            if self._control_files._lock_mode == 'w':
1127
1220
            return
1128
1221
        state = self.current_dirstate()
1129
1222
        state._read_dirblocks_if_needed()
1130
 
        ids_to_unversion = set()
1131
 
        for file_id in file_ids:
1132
 
            ids_to_unversion.add(osutils.safe_file_id(file_id))
 
1223
        ids_to_unversion = set(file_ids)
1133
1224
        paths_to_unversion = set()
1134
1225
        # sketch:
1135
1226
        # check if the root is to be unversioned, if so, assert for now.
1162
1253
                # just forget the whole block.
1163
1254
                entry_index = 0
1164
1255
                while entry_index < len(block[1]):
1165
 
                    # Mark this file id as having been removed
1166
 
                    ids_to_unversion.discard(block[1][entry_index][0][2])
1167
 
                    if not state._make_absent(block[1][entry_index]):
 
1256
                    entry = block[1][entry_index]
 
1257
                    if entry[1][0][0] in 'ar':
 
1258
                        # don't remove absent or renamed entries
1168
1259
                        entry_index += 1
 
1260
                    else:
 
1261
                        # Mark this file id as having been removed
 
1262
                        ids_to_unversion.discard(entry[0][2])
 
1263
                        if not state._make_absent(entry):
 
1264
                            # The block has not shrunk.
 
1265
                            entry_index += 1
1169
1266
                # go to the next block. (At the moment we dont delete empty
1170
1267
                # dirblocks)
1171
1268
                block_index += 1
1192
1289
        # have to change the legacy inventory too.
1193
1290
        if self._inventory is not None:
1194
1291
            for file_id in file_ids:
1195
 
                self._inventory.remove_recursive_id(file_id)
 
1292
                if self._inventory.has_id(file_id):
 
1293
                    self._inventory.remove_recursive_id(file_id)
 
1294
 
 
1295
    @needs_tree_write_lock
 
1296
    def rename_one(self, from_rel, to_rel, after=False):
 
1297
        """See WorkingTree.rename_one"""
 
1298
        self.flush()
 
1299
        super(DirStateWorkingTree, self).rename_one(from_rel, to_rel, after)
 
1300
 
 
1301
    @needs_tree_write_lock
 
1302
    def apply_inventory_delta(self, changes):
 
1303
        """See MutableTree.apply_inventory_delta"""
 
1304
        state = self.current_dirstate()
 
1305
        state.update_by_delta(changes)
 
1306
        self._make_dirty(reset_inventory=True)
 
1307
 
 
1308
    def update_basis_by_delta(self, new_revid, delta):
 
1309
        """See MutableTree.update_basis_by_delta."""
 
1310
        if self.last_revision() == new_revid:
 
1311
            raise AssertionError()
 
1312
        self.current_dirstate().update_basis_by_delta(delta, new_revid)
 
1313
 
 
1314
    @needs_read_lock
 
1315
    def _validate(self):
 
1316
        self._dirstate._validate()
1196
1317
 
1197
1318
    @needs_tree_write_lock
1198
1319
    def _write_inventory(self, inv):
1199
1320
        """Write inventory as the current inventory."""
1200
 
        assert not self._dirty, "attempting to write an inventory when the dirstate is dirty will cause data loss"
1201
 
        self.current_dirstate().set_state_from_inventory(inv)
1202
 
        self._make_dirty(reset_inventory=False)
1203
 
        if self._inventory is not None:
 
1321
        if self._dirty:
 
1322
            raise AssertionError("attempting to write an inventory when the "
 
1323
                "dirstate is dirty will lose pending changes")
 
1324
        had_inventory = self._inventory is not None
 
1325
        # Setting self._inventory = None forces the dirstate to regenerate the
 
1326
        # working inventory. We do this because self.inventory may be inv, or
 
1327
        # may have been modified, and either case would prevent a clean delta
 
1328
        # being created.
 
1329
        self._inventory = None
 
1330
        # generate a delta,
 
1331
        delta = inv._make_delta(self.inventory)
 
1332
        # and apply it.
 
1333
        self.apply_inventory_delta(delta)
 
1334
        if had_inventory:
1204
1335
            self._inventory = inv
1205
1336
        self.flush()
1206
1337
 
1207
 
 
1208
 
class WorkingTreeFormat4(WorkingTreeFormat3):
1209
 
    """The first consolidated dirstate working tree format.
1210
 
 
1211
 
    This format:
1212
 
        - exists within a metadir controlling .bzr
1213
 
        - includes an explicit version marker for the workingtree control
1214
 
          files, separate from the BzrDir format
1215
 
        - modifies the hash cache format
1216
 
        - is new in bzr 0.15
1217
 
        - uses a LockDir to guard access to it.
1218
 
    """
1219
 
 
1220
 
    def get_format_string(self):
1221
 
        """See WorkingTreeFormat.get_format_string()."""
1222
 
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
1223
 
 
1224
 
    def get_format_description(self):
1225
 
        """See WorkingTreeFormat.get_format_description()."""
1226
 
        return "Working tree format 4"
1227
 
 
1228
 
    def initialize(self, a_bzrdir, revision_id=None):
 
1338
    @needs_tree_write_lock
 
1339
    def reset_state(self, revision_ids=None):
 
1340
        """Reset the state of the working tree.
 
1341
 
 
1342
        This does a hard-reset to a last-known-good state. This is a way to
 
1343
        fix if something got corrupted (like the .bzr/checkout/dirstate file)
 
1344
        """
 
1345
        if revision_ids is None:
 
1346
            revision_ids = self.get_parent_ids()
 
1347
        if not revision_ids:
 
1348
            base_tree = self.branch.repository.revision_tree(
 
1349
                _mod_revision.NULL_REVISION)
 
1350
            trees = []
 
1351
        else:
 
1352
            trees = zip(revision_ids,
 
1353
                        self.branch.repository.revision_trees(revision_ids))
 
1354
            base_tree = trees[0][1]
 
1355
        state = self.current_dirstate()
 
1356
        # We don't support ghosts yet
 
1357
        state.set_state_from_scratch(base_tree.inventory, trees, [])
 
1358
 
 
1359
 
 
1360
class ContentFilterAwareSHA1Provider(dirstate.SHA1Provider):
 
1361
 
 
1362
    def __init__(self, tree):
 
1363
        self.tree = tree
 
1364
 
 
1365
    def sha1(self, abspath):
 
1366
        """See dirstate.SHA1Provider.sha1()."""
 
1367
        filters = self.tree._content_filter_stack(
 
1368
            self.tree.relpath(osutils.safe_unicode(abspath)))
 
1369
        return _mod_filters.internal_size_sha_file_byname(abspath, filters)[1]
 
1370
 
 
1371
    def stat_and_sha1(self, abspath):
 
1372
        """See dirstate.SHA1Provider.stat_and_sha1()."""
 
1373
        filters = self.tree._content_filter_stack(
 
1374
            self.tree.relpath(osutils.safe_unicode(abspath)))
 
1375
        file_obj = file(abspath, 'rb', 65000)
 
1376
        try:
 
1377
            statvalue = os.fstat(file_obj.fileno())
 
1378
            if filters:
 
1379
                file_obj = _mod_filters.filtered_input_file(file_obj, filters)
 
1380
            sha1 = osutils.size_sha_file(file_obj)[1]
 
1381
        finally:
 
1382
            file_obj.close()
 
1383
        return statvalue, sha1
 
1384
 
 
1385
 
 
1386
class ContentFilteringDirStateWorkingTree(DirStateWorkingTree):
 
1387
    """Dirstate working tree that supports content filtering.
 
1388
 
 
1389
    The dirstate holds the hash and size of the canonical form of the file, 
 
1390
    and most methods must return that.
 
1391
    """
 
1392
 
 
1393
    def _file_content_summary(self, path, stat_result):
 
1394
        # This is to support the somewhat obsolete path_content_summary method
 
1395
        # with content filtering: see
 
1396
        # <https://bugs.launchpad.net/bzr/+bug/415508>.
 
1397
        #
 
1398
        # If the dirstate cache is up to date and knows the hash and size,
 
1399
        # return that.
 
1400
        # Otherwise if there are no content filters, return the on-disk size
 
1401
        # and leave the hash blank.
 
1402
        # Otherwise, read and filter the on-disk file and use its size and
 
1403
        # hash.
 
1404
        #
 
1405
        # The dirstate doesn't store the size of the canonical form so we
 
1406
        # can't trust it for content-filtered trees.  We just return None.
 
1407
        dirstate_sha1 = self._dirstate.sha1_from_stat(path, stat_result)
 
1408
        executable = self._is_executable_from_path_and_stat(path, stat_result)
 
1409
        return ('file', None, executable, dirstate_sha1)
 
1410
 
 
1411
 
 
1412
class WorkingTree4(DirStateWorkingTree):
 
1413
    """This is the Format 4 working tree.
 
1414
 
 
1415
    This differs from WorkingTree by:
 
1416
     - Having a consolidated internal dirstate, stored in a
 
1417
       randomly-accessible sorted file on disk.
 
1418
     - Not having a regular inventory attribute.  One can be synthesized
 
1419
       on demand but this is expensive and should be avoided.
 
1420
 
 
1421
    This is new in bzr 0.15.
 
1422
    """
 
1423
 
 
1424
 
 
1425
class WorkingTree5(ContentFilteringDirStateWorkingTree):
 
1426
    """This is the Format 5 working tree.
 
1427
 
 
1428
    This differs from WorkingTree4 by:
 
1429
     - Supporting content filtering.
 
1430
 
 
1431
    This is new in bzr 1.11.
 
1432
    """
 
1433
 
 
1434
 
 
1435
class WorkingTree6(ContentFilteringDirStateWorkingTree):
 
1436
    """This is the Format 6 working tree.
 
1437
 
 
1438
    This differs from WorkingTree5 by:
 
1439
     - Supporting a current view that may mask the set of files in a tree
 
1440
       impacted by most user operations.
 
1441
 
 
1442
    This is new in bzr 1.14.
 
1443
    """
 
1444
 
 
1445
    def _make_views(self):
 
1446
        return views.PathBasedViews(self)
 
1447
 
 
1448
 
 
1449
class DirStateWorkingTreeFormat(WorkingTreeFormat):
 
1450
 
 
1451
    missing_parent_conflicts = True
 
1452
 
 
1453
    supports_versioned_directories = True
 
1454
 
 
1455
    _lock_class = LockDir
 
1456
    _lock_file_name = 'lock'
 
1457
 
 
1458
    def _open_control_files(self, a_bzrdir):
 
1459
        transport = a_bzrdir.get_workingtree_transport(None)
 
1460
        return LockableFiles(transport, self._lock_file_name,
 
1461
                             self._lock_class)
 
1462
 
 
1463
    def initialize(self, a_bzrdir, revision_id=None, from_branch=None,
 
1464
                   accelerator_tree=None, hardlink=False):
1229
1465
        """See WorkingTreeFormat.initialize().
1230
1466
 
1231
1467
        :param revision_id: allows creating a working tree at a different
1232
 
        revision than the branch is at.
 
1468
            revision than the branch is at.
 
1469
        :param accelerator_tree: A tree which can be used for retrieving file
 
1470
            contents more quickly than the revision tree, i.e. a workingtree.
 
1471
            The revision tree will be used for cases where accelerator_tree's
 
1472
            content is different.
 
1473
        :param hardlink: If true, hard-link files from accelerator_tree,
 
1474
            where possible.
1233
1475
 
1234
 
        These trees get an initial random root id.
 
1476
        These trees get an initial random root id, if their repository supports
 
1477
        rich root data, TREE_ROOT otherwise.
1235
1478
        """
1236
 
        revision_id = osutils.safe_revision_id(revision_id)
1237
1479
        if not isinstance(a_bzrdir.transport, LocalTransport):
1238
1480
            raise errors.NotLocalUrl(a_bzrdir.transport.base)
1239
1481
        transport = a_bzrdir.get_workingtree_transport(self)
1240
1482
        control_files = self._open_control_files(a_bzrdir)
1241
1483
        control_files.create_lock()
1242
1484
        control_files.lock_write()
1243
 
        control_files.put_utf8('format', self.get_format_string())
1244
 
        branch = a_bzrdir.open_branch()
 
1485
        transport.put_bytes('format', self.get_format_string(),
 
1486
            mode=a_bzrdir._get_file_mode())
 
1487
        if from_branch is not None:
 
1488
            branch = from_branch
 
1489
        else:
 
1490
            branch = a_bzrdir.open_branch()
1245
1491
        if revision_id is None:
1246
1492
            revision_id = branch.last_revision()
1247
1493
        local_path = transport.local_abspath('dirstate')
1248
1494
        # write out new dirstate (must exist when we create the tree)
1249
1495
        state = dirstate.DirState.initialize(local_path)
1250
1496
        state.unlock()
1251
 
        wt = WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
 
1497
        del state
 
1498
        wt = self._tree_class(a_bzrdir.root_transport.local_abspath('.'),
1252
1499
                         branch,
1253
1500
                         _format=self,
1254
1501
                         _bzrdir=a_bzrdir,
1255
1502
                         _control_files=control_files)
1256
1503
        wt._new_tree()
1257
1504
        wt.lock_tree_write()
1258
 
        state._validate()
1259
1505
        try:
1260
 
            if revision_id in (None, NULL_REVISION):
1261
 
                wt._set_root_id(generate_ids.gen_root_id())
 
1506
            self._init_custom_control_files(wt)
 
1507
            if revision_id in (None, _mod_revision.NULL_REVISION):
 
1508
                if branch.repository.supports_rich_root():
 
1509
                    wt._set_root_id(generate_ids.gen_root_id())
 
1510
                else:
 
1511
                    wt._set_root_id(ROOT_ID)
1262
1512
                wt.flush()
1263
 
                wt.current_dirstate()._validate()
1264
 
            wt.set_last_revision(revision_id)
1265
 
            wt.flush()
1266
 
            basis = wt.basis_tree()
 
1513
            basis = None
 
1514
            # frequently, we will get here due to branching.  The accelerator
 
1515
            # tree will be the tree from the branch, so the desired basis
 
1516
            # tree will often be a parent of the accelerator tree.
 
1517
            if accelerator_tree is not None:
 
1518
                try:
 
1519
                    basis = accelerator_tree.revision_tree(revision_id)
 
1520
                except errors.NoSuchRevision:
 
1521
                    pass
 
1522
            if basis is None:
 
1523
                basis = branch.repository.revision_tree(revision_id)
 
1524
            if revision_id == _mod_revision.NULL_REVISION:
 
1525
                parents_list = []
 
1526
            else:
 
1527
                parents_list = [(revision_id, basis)]
1267
1528
            basis.lock_read()
1268
 
            # if the basis has a root id we have to use that; otherwise we use
1269
 
            # a new random one
1270
 
            basis_root_id = basis.get_root_id()
1271
 
            if basis_root_id is not None:
1272
 
                wt._set_root_id(basis_root_id)
 
1529
            try:
 
1530
                wt.set_parent_trees(parents_list, allow_leftmost_as_ghost=True)
1273
1531
                wt.flush()
1274
 
            transform.build_tree(basis, wt)
1275
 
            basis.unlock()
 
1532
                # if the basis has a root id we have to use that; otherwise we
 
1533
                # use a new random one
 
1534
                basis_root_id = basis.get_root_id()
 
1535
                if basis_root_id is not None:
 
1536
                    wt._set_root_id(basis_root_id)
 
1537
                    wt.flush()
 
1538
                if wt.supports_content_filtering():
 
1539
                    # The original tree may not have the same content filters
 
1540
                    # applied so we can't safely build the inventory delta from
 
1541
                    # the source tree.
 
1542
                    delta_from_tree = False
 
1543
                else:
 
1544
                    delta_from_tree = True
 
1545
                # delta_from_tree is safe even for DirStateRevisionTrees,
 
1546
                # because wt4.apply_inventory_delta does not mutate the input
 
1547
                # inventory entries.
 
1548
                transform.build_tree(basis, wt, accelerator_tree,
 
1549
                                     hardlink=hardlink,
 
1550
                                     delta_from_tree=delta_from_tree)
 
1551
            finally:
 
1552
                basis.unlock()
1276
1553
        finally:
1277
1554
            control_files.unlock()
1278
1555
            wt.unlock()
1279
1556
        return wt
1280
1557
 
 
1558
    def _init_custom_control_files(self, wt):
 
1559
        """Subclasses with custom control files should override this method.
 
1560
 
 
1561
        The working tree and control files are locked for writing when this
 
1562
        method is called.
 
1563
 
 
1564
        :param wt: the WorkingTree object
 
1565
        """
 
1566
 
 
1567
    def open(self, a_bzrdir, _found=False):
 
1568
        """Return the WorkingTree object for a_bzrdir
 
1569
 
 
1570
        _found is a private parameter, do not use it. It is used to indicate
 
1571
               if format probing has already been done.
 
1572
        """
 
1573
        if not _found:
 
1574
            # we are being called directly and must probe.
 
1575
            raise NotImplementedError
 
1576
        if not isinstance(a_bzrdir.transport, LocalTransport):
 
1577
            raise errors.NotLocalUrl(a_bzrdir.transport.base)
 
1578
        wt = self._open(a_bzrdir, self._open_control_files(a_bzrdir))
 
1579
        return wt
 
1580
 
1281
1581
    def _open(self, a_bzrdir, control_files):
1282
1582
        """Open the tree itself.
1283
1583
 
1284
1584
        :param a_bzrdir: the dir for the tree.
1285
1585
        :param control_files: the control files for the tree.
1286
1586
        """
1287
 
        return WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
 
1587
        return self._tree_class(a_bzrdir.root_transport.local_abspath('.'),
1288
1588
                           branch=a_bzrdir.open_branch(),
1289
1589
                           _format=self,
1290
1590
                           _bzrdir=a_bzrdir,
1291
1591
                           _control_files=control_files)
1292
1592
 
1293
1593
    def __get_matchingbzrdir(self):
 
1594
        return self._get_matchingbzrdir()
 
1595
 
 
1596
    def _get_matchingbzrdir(self):
 
1597
        """Overrideable method to get a bzrdir for testing."""
1294
1598
        # please test against something that will let us do tree references
1295
1599
        return bzrdir.format_registry.make_bzrdir(
1296
1600
            'dirstate-with-subtree')
1298
1602
    _matchingbzrdir = property(__get_matchingbzrdir)
1299
1603
 
1300
1604
 
1301
 
class DirStateRevisionTree(Tree):
1302
 
    """A revision tree pulling the inventory from a dirstate."""
 
1605
class WorkingTreeFormat4(DirStateWorkingTreeFormat):
 
1606
    """The first consolidated dirstate working tree format.
 
1607
 
 
1608
    This format:
 
1609
        - exists within a metadir controlling .bzr
 
1610
        - includes an explicit version marker for the workingtree control
 
1611
          files, separate from the BzrDir format
 
1612
        - modifies the hash cache format
 
1613
        - is new in bzr 0.15
 
1614
        - uses a LockDir to guard access to it.
 
1615
    """
 
1616
 
 
1617
    upgrade_recommended = False
 
1618
 
 
1619
    _tree_class = WorkingTree4
 
1620
 
 
1621
    def get_format_string(self):
 
1622
        """See WorkingTreeFormat.get_format_string()."""
 
1623
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
 
1624
 
 
1625
    def get_format_description(self):
 
1626
        """See WorkingTreeFormat.get_format_description()."""
 
1627
        return "Working tree format 4"
 
1628
 
 
1629
 
 
1630
class WorkingTreeFormat5(DirStateWorkingTreeFormat):
 
1631
    """WorkingTree format supporting content filtering.
 
1632
    """
 
1633
 
 
1634
    upgrade_recommended = False
 
1635
 
 
1636
    _tree_class = WorkingTree5
 
1637
 
 
1638
    def get_format_string(self):
 
1639
        """See WorkingTreeFormat.get_format_string()."""
 
1640
        return "Bazaar Working Tree Format 5 (bzr 1.11)\n"
 
1641
 
 
1642
    def get_format_description(self):
 
1643
        """See WorkingTreeFormat.get_format_description()."""
 
1644
        return "Working tree format 5"
 
1645
 
 
1646
    def supports_content_filtering(self):
 
1647
        return True
 
1648
 
 
1649
 
 
1650
class WorkingTreeFormat6(DirStateWorkingTreeFormat):
 
1651
    """WorkingTree format supporting views.
 
1652
    """
 
1653
 
 
1654
    upgrade_recommended = False
 
1655
 
 
1656
    _tree_class = WorkingTree6
 
1657
 
 
1658
    def get_format_string(self):
 
1659
        """See WorkingTreeFormat.get_format_string()."""
 
1660
        return "Bazaar Working Tree Format 6 (bzr 1.14)\n"
 
1661
 
 
1662
    def get_format_description(self):
 
1663
        """See WorkingTreeFormat.get_format_description()."""
 
1664
        return "Working tree format 6"
 
1665
 
 
1666
    def _init_custom_control_files(self, wt):
 
1667
        """Subclasses with custom control files should override this method."""
 
1668
        wt._transport.put_bytes('views', '', mode=wt.bzrdir._get_file_mode())
 
1669
 
 
1670
    def supports_content_filtering(self):
 
1671
        return True
 
1672
 
 
1673
    def supports_views(self):
 
1674
        return True
 
1675
 
 
1676
 
 
1677
class DirStateRevisionTree(InventoryTree):
 
1678
    """A revision tree pulling the inventory from a dirstate.
 
1679
    
 
1680
    Note that this is one of the historical (ie revision) trees cached in the
 
1681
    dirstate for easy access, not the workingtree.
 
1682
    """
1303
1683
 
1304
1684
    def __init__(self, dirstate, revision_id, repository):
1305
1685
        self._dirstate = dirstate
1306
 
        self._revision_id = osutils.safe_revision_id(revision_id)
 
1686
        self._revision_id = revision_id
1307
1687
        self._repository = repository
1308
1688
        self._inventory = None
1309
1689
        self._locked = 0
1310
1690
        self._dirstate_locked = False
 
1691
        self._repo_supports_tree_reference = getattr(
 
1692
            repository._format, "supports_tree_reference",
 
1693
            False)
1311
1694
 
1312
1695
    def __repr__(self):
1313
1696
        return "<%s of %s in %s>" % \
1314
1697
            (self.__class__.__name__, self._revision_id, self._dirstate)
1315
1698
 
1316
 
    def annotate_iter(self, file_id):
 
1699
    def annotate_iter(self, file_id,
 
1700
                      default_revision=_mod_revision.CURRENT_REVISION):
1317
1701
        """See Tree.annotate_iter"""
1318
 
        w = self._repository.weave_store.get_weave(file_id,
1319
 
                           self._repository.get_transaction())
1320
 
        return w.annotate_iter(self.inventory[file_id].revision)
 
1702
        text_key = (file_id, self.get_file_revision(file_id))
 
1703
        annotations = self._repository.texts.annotate(text_key)
 
1704
        return [(key[-1], line) for (key, line) in annotations]
1321
1705
 
1322
1706
    def _comparison_data(self, entry, path):
1323
1707
        """See Tree._comparison_data."""
1341
1725
    def get_root_id(self):
1342
1726
        return self.path2id('')
1343
1727
 
 
1728
    def id2path(self, file_id):
 
1729
        "Convert a file-id to a path."
 
1730
        entry = self._get_entry(file_id=file_id)
 
1731
        if entry == (None, None):
 
1732
            raise errors.NoSuchId(tree=self, file_id=file_id)
 
1733
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
 
1734
        return path_utf8.decode('utf8')
 
1735
 
 
1736
    def iter_references(self):
 
1737
        if not self._repo_supports_tree_reference:
 
1738
            # When the repo doesn't support references, we will have nothing to
 
1739
            # return
 
1740
            return iter([])
 
1741
        # Otherwise, fall back to the default implementation
 
1742
        return super(DirStateRevisionTree, self).iter_references()
 
1743
 
1344
1744
    def _get_parent_index(self):
1345
1745
        """Return the index in the dirstate referenced by this tree."""
1346
1746
        return self._dirstate.get_parent_ids().index(self._revision_id) + 1
1351
1751
        If either file_id or path is supplied, it is used as the key to lookup.
1352
1752
        If both are supplied, the fastest lookup is used, and an error is
1353
1753
        raised if they do not both point at the same row.
1354
 
        
 
1754
 
1355
1755
        :param file_id: An optional unicode file_id to be looked up.
1356
1756
        :param path: An optional unicode path to be looked up.
1357
1757
        :return: The dirstate row tuple for path/file_id, or (None, None)
1358
1758
        """
1359
1759
        if file_id is None and path is None:
1360
1760
            raise errors.BzrError('must supply file_id or path')
1361
 
        file_id = osutils.safe_file_id(file_id)
1362
1761
        if path is not None:
1363
1762
            path = path.encode('utf8')
1364
1763
        parent_index = self._get_parent_index()
1372
1771
 
1373
1772
        This is relatively expensive: we have to walk the entire dirstate.
1374
1773
        """
1375
 
        assert self._locked, 'cannot generate inventory of an unlocked '\
1376
 
            'dirstate revision tree'
 
1774
        if not self._locked:
 
1775
            raise AssertionError(
 
1776
                'cannot generate inventory of an unlocked '
 
1777
                'dirstate revision tree')
1377
1778
        # separate call for profiling - makes it clear where the costs are.
1378
1779
        self._dirstate._read_dirblocks_if_needed()
1379
 
        assert self._revision_id in self._dirstate.get_parent_ids(), \
1380
 
            'parent %s has disappeared from %s' % (
1381
 
            self._revision_id, self._dirstate.get_parent_ids())
 
1780
        if self._revision_id not in self._dirstate.get_parent_ids():
 
1781
            raise AssertionError(
 
1782
                'parent %s has disappeared from %s' % (
 
1783
                self._revision_id, self._dirstate.get_parent_ids()))
1382
1784
        parent_index = self._dirstate.get_parent_ids().index(self._revision_id) + 1
1383
1785
        # This is identical now to the WorkingTree _generate_inventory except
1384
1786
        # for the tree index use.
1385
1787
        root_key, current_entry = self._dirstate._get_entry(parent_index, path_utf8='')
1386
1788
        current_id = root_key[2]
1387
 
        assert current_entry[parent_index][0] == 'd'
 
1789
        if current_entry[parent_index][0] != 'd':
 
1790
            raise AssertionError()
1388
1791
        inv = Inventory(root_id=current_id, revision_id=self._revision_id)
1389
1792
        inv.root.revision = current_entry[parent_index][4]
1390
1793
        # Turn some things into local variables
1421
1824
                elif kind == 'directory':
1422
1825
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
1423
1826
                elif kind == 'symlink':
1424
 
                    inv_entry.executable = False
1425
 
                    inv_entry.text_size = None
1426
1827
                    inv_entry.symlink_target = utf8_decode(fingerprint)[0]
1427
1828
                elif kind == 'tree-reference':
1428
1829
                    inv_entry.reference_revision = fingerprint or None
1430
1831
                    raise AssertionError("cannot convert entry %r into an InventoryEntry"
1431
1832
                            % entry)
1432
1833
                # These checks cost us around 40ms on a 55k entry tree
1433
 
                assert file_id not in inv_byid
1434
 
                assert name_unicode not in parent_ie.children
 
1834
                if file_id in inv_byid:
 
1835
                    raise AssertionError('file_id %s already in'
 
1836
                        ' inventory as %s' % (file_id, inv_byid[file_id]))
 
1837
                if name_unicode in parent_ie.children:
 
1838
                    raise AssertionError('name %r already in parent'
 
1839
                        % (name_unicode,))
1435
1840
                inv_byid[file_id] = inv_entry
1436
1841
                parent_ie.children[name_unicode] = inv_entry
1437
1842
        self._inventory = inv
1444
1849
        # Make sure the file exists
1445
1850
        entry = self._get_entry(file_id, path=path)
1446
1851
        if entry == (None, None): # do we raise?
1447
 
            return None
 
1852
            raise errors.NoSuchId(self, file_id)
1448
1853
        parent_index = self._get_parent_index()
1449
1854
        last_changed_revision = entry[1][parent_index][4]
1450
 
        return self._repository.get_revision(last_changed_revision).timestamp
 
1855
        try:
 
1856
            rev = self._repository.get_revision(last_changed_revision)
 
1857
        except errors.NoSuchRevision:
 
1858
            raise errors.FileTimestampUnavailable(self.id2path(file_id))
 
1859
        return rev.timestamp
1451
1860
 
1452
1861
    def get_file_sha1(self, file_id, path=None, stat_value=None):
1453
1862
        entry = self._get_entry(file_id=file_id, path=path)
1457
1866
            return parent_details[1]
1458
1867
        return None
1459
1868
 
1460
 
    def get_file(self, file_id):
 
1869
    @needs_read_lock
 
1870
    def get_file_revision(self, file_id):
 
1871
        return self.inventory[file_id].revision
 
1872
 
 
1873
    def get_file(self, file_id, path=None):
1461
1874
        return StringIO(self.get_file_text(file_id))
1462
1875
 
1463
 
    def get_file_lines(self, file_id):
1464
 
        ie = self.inventory[file_id]
1465
 
        return self._repository.weave_store.get_weave(file_id,
1466
 
                self._repository.get_transaction()).get_lines(ie.revision)
1467
 
 
1468
1876
    def get_file_size(self, file_id):
 
1877
        """See Tree.get_file_size"""
1469
1878
        return self.inventory[file_id].text_size
1470
1879
 
1471
 
    def get_file_text(self, file_id):
1472
 
        return ''.join(self.get_file_lines(file_id))
 
1880
    def get_file_text(self, file_id, path=None):
 
1881
        _, content = list(self.iter_files_bytes([(file_id, None)]))[0]
 
1882
        return ''.join(content)
1473
1883
 
1474
1884
    def get_reference_revision(self, file_id, path=None):
1475
1885
        return self.inventory[file_id].reference_revision
1476
1886
 
1477
 
    def get_symlink_target(self, file_id):
 
1887
    def iter_files_bytes(self, desired_files):
 
1888
        """See Tree.iter_files_bytes.
 
1889
 
 
1890
        This version is implemented on top of Repository.iter_files_bytes"""
 
1891
        parent_index = self._get_parent_index()
 
1892
        repo_desired_files = []
 
1893
        for file_id, identifier in desired_files:
 
1894
            entry = self._get_entry(file_id)
 
1895
            if entry == (None, None):
 
1896
                raise errors.NoSuchId(self, file_id)
 
1897
            repo_desired_files.append((file_id, entry[1][parent_index][4],
 
1898
                                       identifier))
 
1899
        return self._repository.iter_files_bytes(repo_desired_files)
 
1900
 
 
1901
    def get_symlink_target(self, file_id, path=None):
1478
1902
        entry = self._get_entry(file_id=file_id)
1479
1903
        parent_index = self._get_parent_index()
1480
1904
        if entry[1][parent_index][0] != 'l':
1481
1905
            return None
1482
1906
        else:
1483
 
            # At present, none of the tree implementations supports non-ascii
1484
 
            # symlink targets. So we will just assume that the dirstate path is
1485
 
            # correct.
1486
 
            return entry[1][parent_index][1]
 
1907
            target = entry[1][parent_index][1]
 
1908
            target = target.decode('utf8')
 
1909
            return target
1487
1910
 
1488
1911
    def get_revision_id(self):
1489
1912
        """Return the revision id for this tree."""
1507
1930
        return bool(self.path2id(filename))
1508
1931
 
1509
1932
    def kind(self, file_id):
1510
 
        return self.inventory[file_id].kind
 
1933
        entry = self._get_entry(file_id=file_id)[1]
 
1934
        if entry is None:
 
1935
            raise errors.NoSuchId(tree=self, file_id=file_id)
 
1936
        parent_index = self._get_parent_index()
 
1937
        return dirstate.DirState._minikind_to_kind[entry[parent_index][0]]
 
1938
 
 
1939
    def stored_kind(self, file_id):
 
1940
        """See Tree.stored_kind"""
 
1941
        return self.kind(file_id)
 
1942
 
 
1943
    def path_content_summary(self, path):
 
1944
        """See Tree.path_content_summary."""
 
1945
        id = self.inventory.path2id(path)
 
1946
        if id is None:
 
1947
            return ('missing', None, None, None)
 
1948
        entry = self._inventory[id]
 
1949
        kind = entry.kind
 
1950
        if kind == 'file':
 
1951
            return (kind, entry.text_size, entry.executable, entry.text_sha1)
 
1952
        elif kind == 'symlink':
 
1953
            return (kind, None, None, entry.symlink_target)
 
1954
        else:
 
1955
            return (kind, None, None, None)
1511
1956
 
1512
1957
    def is_executable(self, file_id, path=None):
1513
1958
        ie = self.inventory[file_id]
1514
1959
        if ie.kind != "file":
1515
 
            return None
 
1960
            return False
1516
1961
        return ie.executable
1517
1962
 
1518
 
    def list_files(self, include_root=False):
 
1963
    def is_locked(self):
 
1964
        return self._locked
 
1965
 
 
1966
    def list_files(self, include_root=False, from_dir=None, recursive=True):
1519
1967
        # We use a standard implementation, because DirStateRevisionTree is
1520
1968
        # dealing with one of the parents of the current state
1521
1969
        inv = self._get_inventory()
1522
 
        entries = inv.iter_entries()
1523
 
        if self.inventory.root is not None and not include_root:
 
1970
        if from_dir is None:
 
1971
            from_dir_id = None
 
1972
        else:
 
1973
            from_dir_id = inv.path2id(from_dir)
 
1974
            if from_dir_id is None:
 
1975
                # Directory not versioned
 
1976
                return
 
1977
        entries = inv.iter_entries(from_dir=from_dir_id, recursive=recursive)
 
1978
        if inv.root is not None and not include_root and from_dir is None:
1524
1979
            entries.next()
1525
1980
        for path, entry in entries:
1526
1981
            yield path, 'V', entry.kind, entry.file_id, entry
1527
1982
 
1528
1983
    def lock_read(self):
1529
 
        """Lock the tree for a set of operations."""
 
1984
        """Lock the tree for a set of operations.
 
1985
 
 
1986
        :return: A bzrlib.lock.LogicalLockResult.
 
1987
        """
1530
1988
        if not self._locked:
1531
1989
            self._repository.lock_read()
1532
1990
            if self._dirstate._lock_token is None:
1533
1991
                self._dirstate.lock_read()
1534
1992
                self._dirstate_locked = True
1535
1993
        self._locked += 1
 
1994
        return LogicalLockResult(self.unlock)
1536
1995
 
1537
1996
    def _must_be_locked(self):
1538
1997
        if not self._locked:
1559
2018
                self._dirstate_locked = False
1560
2019
            self._repository.unlock()
1561
2020
 
 
2021
    @needs_read_lock
 
2022
    def supports_tree_reference(self):
 
2023
        return self._repo_supports_tree_reference
 
2024
 
1562
2025
    def walkdirs(self, prefix=""):
1563
2026
        # TODO: jam 20070215 This is the lazy way by using the RevisionTree
1564
 
        # implementation based on an inventory.  
 
2027
        # implementation based on an inventory.
1565
2028
        # This should be cleaned up to use the much faster Dirstate code
1566
2029
        # So for now, we just build up the parent inventory, and extract
1567
2030
        # it the same way RevisionTree does.
1596
2059
 
1597
2060
class InterDirStateTree(InterTree):
1598
2061
    """Fast path optimiser for changes_from with dirstate trees.
1599
 
    
1600
 
    This is used only when both trees are in the dirstate working file, and 
1601
 
    the source is any parent within the dirstate, and the destination is 
 
2062
 
 
2063
    This is used only when both trees are in the dirstate working file, and
 
2064
    the source is any parent within the dirstate, and the destination is
1602
2065
    the current working tree of the same dirstate.
1603
2066
    """
1604
2067
    # this could be generalized to allow comparisons between any trees in the
1613
2076
    def make_source_parent_tree(source, target):
1614
2077
        """Change the source tree into a parent of the target."""
1615
2078
        revid = source.commit('record tree')
1616
 
        target.branch.repository.fetch(source.branch.repository, revid)
 
2079
        target.branch.fetch(source.branch, revid)
1617
2080
        target.set_parent_ids([revid])
1618
2081
        return target.basis_tree(), target
1619
2082
 
 
2083
    @classmethod
 
2084
    def make_source_parent_tree_python_dirstate(klass, test_case, source, target):
 
2085
        result = klass.make_source_parent_tree(source, target)
 
2086
        result[1]._iter_changes = dirstate.ProcessEntryPython
 
2087
        return result
 
2088
 
 
2089
    @classmethod
 
2090
    def make_source_parent_tree_compiled_dirstate(klass, test_case, source,
 
2091
                                                  target):
 
2092
        from bzrlib.tests.test__dirstate_helpers import \
 
2093
            compiled_dirstate_helpers_feature
 
2094
        test_case.requireFeature(compiled_dirstate_helpers_feature)
 
2095
        from bzrlib._dirstate_helpers_pyx import ProcessEntryC
 
2096
        result = klass.make_source_parent_tree(source, target)
 
2097
        result[1]._iter_changes = ProcessEntryC
 
2098
        return result
 
2099
 
1620
2100
    _matching_from_tree_format = WorkingTreeFormat4()
1621
2101
    _matching_to_tree_format = WorkingTreeFormat4()
1622
 
    _test_mutable_trees_to_test_trees = make_source_parent_tree
1623
 
 
1624
 
    def _iter_changes(self, include_unchanged=False,
 
2102
 
 
2103
    @classmethod
 
2104
    def _test_mutable_trees_to_test_trees(klass, test_case, source, target):
 
2105
        # This method shouldn't be called, because we have python and C
 
2106
        # specific flavours.
 
2107
        raise NotImplementedError
 
2108
 
 
2109
    def iter_changes(self, include_unchanged=False,
1625
2110
                      specific_files=None, pb=None, extra_trees=[],
1626
2111
                      require_versioned=True, want_unversioned=False):
1627
2112
        """Return the changes from source to target.
1628
2113
 
1629
 
        :return: An iterator that yields tuples. See InterTree._iter_changes
 
2114
        :return: An iterator that yields tuples. See InterTree.iter_changes
1630
2115
            for details.
1631
2116
        :param specific_files: An optional list of file paths to restrict the
1632
2117
            comparison to. When mapping filenames to ids, all matches in all
1643
2128
            output. An unversioned file is defined as one with (False, False)
1644
2129
            for the versioned pair.
1645
2130
        """
1646
 
        utf8_decode = cache_utf8._utf8_decode_with_None
1647
 
        _minikind_to_kind = dirstate.DirState._minikind_to_kind
1648
 
        # NB: show_status depends on being able to pass in non-versioned files
1649
 
        # and report them as unknown
1650
2131
        # TODO: handle extra trees in the dirstate.
1651
 
        # TODO: handle comparisons as an empty tree as a different special
1652
 
        # case? mbp 20070226
1653
 
        if extra_trees or (self.source._revision_id == NULL_REVISION):
 
2132
        if (extra_trees or specific_files == []):
1654
2133
            # we can't fast-path these cases (yet)
1655
 
            for f in super(InterDirStateTree, self)._iter_changes(
 
2134
            return super(InterDirStateTree, self).iter_changes(
1656
2135
                include_unchanged, specific_files, pb, extra_trees,
1657
 
                require_versioned, want_unversioned=want_unversioned):
1658
 
                yield f
1659
 
            return
 
2136
                require_versioned, want_unversioned=want_unversioned)
1660
2137
        parent_ids = self.target.get_parent_ids()
1661
 
        assert (self.source._revision_id in parent_ids), \
1662
 
                "revision {%s} is not stored in {%s}, but %s " \
1663
 
                "can only be used for trees stored in the dirstate" \
1664
 
                % (self.source._revision_id, self.target, self._iter_changes)
 
2138
        if not (self.source._revision_id in parent_ids
 
2139
                or self.source._revision_id == _mod_revision.NULL_REVISION):
 
2140
            raise AssertionError(
 
2141
                "revision {%s} is not stored in {%s}, but %s "
 
2142
                "can only be used for trees stored in the dirstate"
 
2143
                % (self.source._revision_id, self.target, self.iter_changes))
1665
2144
        target_index = 0
1666
 
        if self.source._revision_id == NULL_REVISION:
 
2145
        if self.source._revision_id == _mod_revision.NULL_REVISION:
1667
2146
            source_index = None
1668
2147
            indices = (target_index,)
1669
2148
        else:
1670
 
            assert (self.source._revision_id in parent_ids), \
1671
 
                "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
1672
 
                self.source._revision_id, parent_ids)
 
2149
            if not (self.source._revision_id in parent_ids):
 
2150
                raise AssertionError(
 
2151
                    "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
 
2152
                    self.source._revision_id, parent_ids))
1673
2153
            source_index = 1 + parent_ids.index(self.source._revision_id)
1674
 
            indices = (source_index,target_index)
 
2154
            indices = (source_index, target_index)
1675
2155
        # -- make all specific_files utf8 --
1676
2156
        if specific_files:
1677
2157
            specific_files_utf8 = set()
1678
2158
            for path in specific_files:
 
2159
                # Note, if there are many specific files, using cache_utf8
 
2160
                # would be good here.
1679
2161
                specific_files_utf8.add(path.encode('utf8'))
1680
2162
            specific_files = specific_files_utf8
1681
2163
        else:
1682
2164
            specific_files = set([''])
1683
2165
        # -- specific_files is now a utf8 path set --
 
2166
 
1684
2167
        # -- get the state object and prepare it.
1685
2168
        state = self.target.current_dirstate()
1686
2169
        state._read_dirblocks_if_needed()
1687
 
        def _entries_for_path(path):
1688
 
            """Return a list with all the entries that match path for all ids.
1689
 
            """
1690
 
            dirname, basename = os.path.split(path)
1691
 
            key = (dirname, basename, '')
1692
 
            block_index, present = state._find_block_index_from_key(key)
1693
 
            if not present:
1694
 
                # the block which should contain path is absent.
1695
 
                return []
1696
 
            result = []
1697
 
            block = state._dirblocks[block_index][1]
1698
 
            entry_index, _ = state._find_entry_index(key, block)
1699
 
            # we may need to look at multiple entries at this path: walk while the specific_files match.
1700
 
            while (entry_index < len(block) and
1701
 
                block[entry_index][0][0:2] == key[0:2]):
1702
 
                result.append(block[entry_index])
1703
 
                entry_index += 1
1704
 
            return result
1705
2170
        if require_versioned:
1706
2171
            # -- check all supplied paths are versioned in a search tree. --
1707
 
            all_versioned = True
 
2172
            not_versioned = []
1708
2173
            for path in specific_files:
1709
 
                path_entries = _entries_for_path(path)
 
2174
                path_entries = state._entries_for_path(path)
1710
2175
                if not path_entries:
1711
2176
                    # this specified path is not present at all: error
1712
 
                    all_versioned = False
1713
 
                    break
 
2177
                    not_versioned.append(path)
 
2178
                    continue
1714
2179
                found_versioned = False
1715
2180
                # for each id at this path
1716
2181
                for entry in path_entries:
1723
2188
                if not found_versioned:
1724
2189
                    # none of the indexes was not 'absent' at all ids for this
1725
2190
                    # path.
1726
 
                    all_versioned = False
1727
 
                    break
1728
 
            if not all_versioned:
1729
 
                raise errors.PathsNotVersionedError(specific_files)
 
2191
                    not_versioned.append(path)
 
2192
            if len(not_versioned) > 0:
 
2193
                raise errors.PathsNotVersionedError(not_versioned)
1730
2194
        # -- remove redundancy in supplied specific_files to prevent over-scanning --
1731
 
        search_specific_files = set()
1732
 
        for path in specific_files:
1733
 
            other_specific_files = specific_files.difference(set([path]))
1734
 
            if not osutils.is_inside_any(other_specific_files, path):
1735
 
                # this is a top level path, we must check it.
1736
 
                search_specific_files.add(path)
1737
 
        # sketch: 
1738
 
        # compare source_index and target_index at or under each element of search_specific_files.
1739
 
        # follow the following comparison table. Note that we only want to do diff operations when
1740
 
        # the target is fdl because thats when the walkdirs logic will have exposed the pathinfo 
1741
 
        # for the target.
1742
 
        # cases:
1743
 
        # 
1744
 
        # Source | Target | disk | action
1745
 
        #   r    | fdlt   |      | add source to search, add id path move and perform
1746
 
        #        |        |      | diff check on source-target
1747
 
        #   r    | fdlt   |  a   | dangling file that was present in the basis. 
1748
 
        #        |        |      | ???
1749
 
        #   r    |  a     |      | add source to search
1750
 
        #   r    |  a     |  a   | 
1751
 
        #   r    |  r     |      | this path is present in a non-examined tree, skip.
1752
 
        #   r    |  r     |  a   | this path is present in a non-examined tree, skip.
1753
 
        #   a    | fdlt   |      | add new id
1754
 
        #   a    | fdlt   |  a   | dangling locally added file, skip
1755
 
        #   a    |  a     |      | not present in either tree, skip
1756
 
        #   a    |  a     |  a   | not present in any tree, skip
1757
 
        #   a    |  r     |      | not present in either tree at this path, skip as it
1758
 
        #        |        |      | may not be selected by the users list of paths.
1759
 
        #   a    |  r     |  a   | not present in either tree at this path, skip as it
1760
 
        #        |        |      | may not be selected by the users list of paths.
1761
 
        #  fdlt  | fdlt   |      | content in both: diff them
1762
 
        #  fdlt  | fdlt   |  a   | deleted locally, but not unversioned - show as deleted ?
1763
 
        #  fdlt  |  a     |      | unversioned: output deleted id for now
1764
 
        #  fdlt  |  a     |  a   | unversioned and deleted: output deleted id
1765
 
        #  fdlt  |  r     |      | relocated in this tree, so add target to search.
1766
 
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
1767
 
        #        |        |      | this id at the other path.
1768
 
        #  fdlt  |  r     |  a   | relocated in this tree, so add target to search.
1769
 
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
1770
 
        #        |        |      | this id at the other path.
1771
 
 
1772
 
        # for all search_indexs in each path at or under each element of
1773
 
        # search_specific_files, if the detail is relocated: add the id, and add the
1774
 
        # relocated path as one to search if its not searched already. If the
1775
 
        # detail is not relocated, add the id.
1776
 
        searched_specific_files = set()
1777
 
        NULL_PARENT_DETAILS = dirstate.DirState.NULL_PARENT_DETAILS
1778
 
        # Using a list so that we can access the values and change them in
1779
 
        # nested scope. Each one is [path, file_id, entry]
1780
 
        last_source_parent = [None, None, None]
1781
 
        last_target_parent = [None, None, None]
 
2195
        search_specific_files = osutils.minimum_path_selection(specific_files)
1782
2196
 
1783
2197
        use_filesystem_for_exec = (sys.platform != 'win32')
1784
 
 
1785
 
        def _process_entry(entry, path_info):
1786
 
            """Compare an entry and real disk to generate delta information.
1787
 
 
1788
 
            :param path_info: top_relpath, basename, kind, lstat, abspath for
1789
 
                the path of entry. If None, then the path is considered absent.
1790
 
                (Perhaps we should pass in a concrete entry for this ?)
1791
 
                Basename is returned as a utf8 string because we expect this
1792
 
                tuple will be ignored, and don't want to take the time to
1793
 
                decode.
1794
 
            """
1795
 
            if source_index is None:
1796
 
                source_details = NULL_PARENT_DETAILS
1797
 
            else:
1798
 
                source_details = entry[1][source_index]
1799
 
            target_details = entry[1][target_index]
1800
 
            target_minikind = target_details[0]
1801
 
            if path_info is not None and target_minikind in 'fdlt':
1802
 
                assert target_index == 0
1803
 
                link_or_sha1 = state.update_entry(entry, abspath=path_info[4],
1804
 
                                                  stat_value=path_info[3])
1805
 
                # The entry may have been modified by update_entry
1806
 
                target_details = entry[1][target_index]
1807
 
                target_minikind = target_details[0]
1808
 
            else:
1809
 
                link_or_sha1 = None
1810
 
            source_minikind = source_details[0]
1811
 
            if source_minikind in 'fdltr' and target_minikind in 'fdlt':
1812
 
                # claimed content in both: diff
1813
 
                #   r    | fdlt   |      | add source to search, add id path move and perform
1814
 
                #        |        |      | diff check on source-target
1815
 
                #   r    | fdlt   |  a   | dangling file that was present in the basis.
1816
 
                #        |        |      | ???
1817
 
                if source_minikind in 'r':
1818
 
                    # add the source to the search path to find any children it
1819
 
                    # has.  TODO ? : only add if it is a container ?
1820
 
                    if not osutils.is_inside_any(searched_specific_files,
1821
 
                                                 source_details[1]):
1822
 
                        search_specific_files.add(source_details[1])
1823
 
                    # generate the old path; this is needed for stating later
1824
 
                    # as well.
1825
 
                    old_path = source_details[1]
1826
 
                    old_dirname, old_basename = os.path.split(old_path)
1827
 
                    path = pathjoin(entry[0][0], entry[0][1])
1828
 
                    old_entry = state._get_entry(source_index,
1829
 
                                                 path_utf8=old_path)
1830
 
                    # update the source details variable to be the real
1831
 
                    # location.
1832
 
                    source_details = old_entry[1][source_index]
1833
 
                    source_minikind = source_details[0]
1834
 
                else:
1835
 
                    old_dirname = entry[0][0]
1836
 
                    old_basename = entry[0][1]
1837
 
                    old_path = path = pathjoin(old_dirname, old_basename)
1838
 
                if path_info is None:
1839
 
                    # the file is missing on disk, show as removed.
1840
 
                    content_change = True
1841
 
                    target_kind = None
1842
 
                    target_exec = False
1843
 
                else:
1844
 
                    # source and target are both versioned and disk file is present.
1845
 
                    target_kind = path_info[2]
1846
 
                    if target_kind == 'directory':
1847
 
                        if source_minikind != 'd':
1848
 
                            content_change = True
1849
 
                        else:
1850
 
                            # directories have no fingerprint
1851
 
                            content_change = False
1852
 
                        target_exec = False
1853
 
                    elif target_kind == 'file':
1854
 
                        if source_minikind != 'f':
1855
 
                            content_change = True
1856
 
                        else:
1857
 
                            # We could check the size, but we already have the
1858
 
                            # sha1 hash.
1859
 
                            content_change = (link_or_sha1 != source_details[1])
1860
 
                        # Target details is updated at update_entry time
1861
 
                        if use_filesystem_for_exec:
1862
 
                            # We don't need S_ISREG here, because we are sure
1863
 
                            # we are dealing with a file.
1864
 
                            target_exec = bool(stat.S_IEXEC & path_info[3].st_mode)
1865
 
                        else:
1866
 
                            target_exec = target_details[3]
1867
 
                    elif target_kind == 'symlink':
1868
 
                        if source_minikind != 'l':
1869
 
                            content_change = True
1870
 
                        else:
1871
 
                            content_change = (link_or_sha1 != source_details[1])
1872
 
                        target_exec = False
1873
 
                    elif target_kind == 'tree-reference':
1874
 
                        if source_minikind != 't':
1875
 
                            content_change = True
1876
 
                        else:
1877
 
                            content_change = False
1878
 
                        target_exec = False
1879
 
                    else:
1880
 
                        raise Exception, "unknown kind %s" % path_info[2]
1881
 
                # parent id is the entry for the path in the target tree
1882
 
                if old_dirname == last_source_parent[0]:
1883
 
                    source_parent_id = last_source_parent[1]
1884
 
                else:
1885
 
                    source_parent_entry = state._get_entry(source_index,
1886
 
                                                           path_utf8=old_dirname)
1887
 
                    source_parent_id = source_parent_entry[0][2]
1888
 
                    if source_parent_id == entry[0][2]:
1889
 
                        # This is the root, so the parent is None
1890
 
                        source_parent_id = None
1891
 
                    else:
1892
 
                        last_source_parent[0] = old_dirname
1893
 
                        last_source_parent[1] = source_parent_id
1894
 
                        last_source_parent[2] = source_parent_entry
1895
 
                new_dirname = entry[0][0]
1896
 
                if new_dirname == last_target_parent[0]:
1897
 
                    target_parent_id = last_target_parent[1]
1898
 
                else:
1899
 
                    # TODO: We don't always need to do the lookup, because the
1900
 
                    #       parent entry will be the same as the source entry.
1901
 
                    target_parent_entry = state._get_entry(target_index,
1902
 
                                                           path_utf8=new_dirname)
1903
 
                    target_parent_id = target_parent_entry[0][2]
1904
 
                    if target_parent_id == entry[0][2]:
1905
 
                        # This is the root, so the parent is None
1906
 
                        target_parent_id = None
1907
 
                    else:
1908
 
                        last_target_parent[0] = new_dirname
1909
 
                        last_target_parent[1] = target_parent_id
1910
 
                        last_target_parent[2] = target_parent_entry
1911
 
 
1912
 
                source_exec = source_details[3]
1913
 
                return ((entry[0][2], (old_path, path), content_change,
1914
 
                        (True, True),
1915
 
                        (source_parent_id, target_parent_id),
1916
 
                        (old_basename, entry[0][1]),
1917
 
                        (_minikind_to_kind[source_minikind], target_kind),
1918
 
                        (source_exec, target_exec)),)
1919
 
            elif source_minikind in 'a' and target_minikind in 'fdlt':
1920
 
                # looks like a new file
1921
 
                if path_info is not None:
1922
 
                    path = pathjoin(entry[0][0], entry[0][1])
1923
 
                    # parent id is the entry for the path in the target tree
1924
 
                    # TODO: these are the same for an entire directory: cache em.
1925
 
                    parent_id = state._get_entry(target_index,
1926
 
                                                 path_utf8=entry[0][0])[0][2]
1927
 
                    if parent_id == entry[0][2]:
1928
 
                        parent_id = None
1929
 
                    if use_filesystem_for_exec:
1930
 
                        # We need S_ISREG here, because we aren't sure if this
1931
 
                        # is a file or not.
1932
 
                        target_exec = bool(
1933
 
                            stat.S_ISREG(path_info[3].st_mode)
1934
 
                            and stat.S_IEXEC & path_info[3].st_mode)
1935
 
                    else:
1936
 
                        target_exec = target_details[3]
1937
 
                    return ((entry[0][2], (None, path), True,
1938
 
                            (False, True),
1939
 
                            (None, parent_id),
1940
 
                            (None, entry[0][1]),
1941
 
                            (None, path_info[2]),
1942
 
                            (None, target_exec)),)
1943
 
                else:
1944
 
                    # but its not on disk: we deliberately treat this as just
1945
 
                    # never-present. (Why ?! - RBC 20070224)
1946
 
                    pass
1947
 
            elif source_minikind in 'fdlt' and target_minikind in 'a':
1948
 
                # unversioned, possibly, or possibly not deleted: we dont care.
1949
 
                # if its still on disk, *and* theres no other entry at this
1950
 
                # path [we dont know this in this routine at the moment -
1951
 
                # perhaps we should change this - then it would be an unknown.
1952
 
                old_path = pathjoin(entry[0][0], entry[0][1])
1953
 
                # parent id is the entry for the path in the target tree
1954
 
                parent_id = state._get_entry(source_index, path_utf8=entry[0][0])[0][2]
1955
 
                if parent_id == entry[0][2]:
1956
 
                    parent_id = None
1957
 
                return ((entry[0][2], (old_path, None), True,
1958
 
                        (True, False),
1959
 
                        (parent_id, None),
1960
 
                        (entry[0][1], None),
1961
 
                        (_minikind_to_kind[source_minikind], None),
1962
 
                        (source_details[3], None)),)
1963
 
            elif source_minikind in 'fdlt' and target_minikind in 'r':
1964
 
                # a rename; could be a true rename, or a rename inherited from
1965
 
                # a renamed parent. TODO: handle this efficiently. Its not
1966
 
                # common case to rename dirs though, so a correct but slow
1967
 
                # implementation will do.
1968
 
                if not osutils.is_inside_any(searched_specific_files, target_details[1]):
1969
 
                    search_specific_files.add(target_details[1])
1970
 
            elif source_minikind in 'ra' and target_minikind in 'ra':
1971
 
                # neither of the selected trees contain this file,
1972
 
                # so skip over it. This is not currently directly tested, but
1973
 
                # is indirectly via test_too_much.TestCommands.test_conflicts.
1974
 
                pass
1975
 
            else:
1976
 
                raise AssertionError("don't know how to compare "
1977
 
                    "source_minikind=%r, target_minikind=%r"
1978
 
                    % (source_minikind, target_minikind))
1979
 
                ## import pdb;pdb.set_trace()
1980
 
            return ()
1981
 
 
1982
 
        while search_specific_files:
1983
 
            # TODO: the pending list should be lexically sorted?  the
1984
 
            # interface doesn't require it.
1985
 
            current_root = search_specific_files.pop()
1986
 
            searched_specific_files.add(current_root)
1987
 
            # process the entries for this containing directory: the rest will be
1988
 
            # found by their parents recursively.
1989
 
            root_entries = _entries_for_path(current_root)
1990
 
            root_abspath = self.target.abspath(current_root)
1991
 
            try:
1992
 
                root_stat = os.lstat(root_abspath)
1993
 
            except OSError, e:
1994
 
                if e.errno == errno.ENOENT:
1995
 
                    # the path does not exist: let _process_entry know that.
1996
 
                    root_dir_info = None
1997
 
                else:
1998
 
                    # some other random error: hand it up.
1999
 
                    raise
2000
 
            else:
2001
 
                root_dir_info = ('', current_root,
2002
 
                    osutils.file_kind_from_stat_mode(root_stat.st_mode), root_stat,
2003
 
                    root_abspath)
2004
 
                if root_dir_info[2] == 'directory':
2005
 
                    if self.target._directory_is_tree_reference(
2006
 
                        current_root.decode('utf8')):
2007
 
                        root_dir_info = root_dir_info[:2] + \
2008
 
                            ('tree-reference',) + root_dir_info[3:]
2009
 
 
2010
 
            if not root_entries and not root_dir_info:
2011
 
                # this specified path is not present at all, skip it.
2012
 
                continue
2013
 
            path_handled = False
2014
 
            for entry in root_entries:
2015
 
                for result in _process_entry(entry, root_dir_info):
2016
 
                    # this check should probably be outside the loop: one
2017
 
                    # 'iterate two trees' api, and then _iter_changes filters
2018
 
                    # unchanged pairs. - RBC 20070226
2019
 
                    path_handled = True
2020
 
                    if (include_unchanged
2021
 
                        or result[2]                    # content change
2022
 
                        or result[3][0] != result[3][1] # versioned status
2023
 
                        or result[4][0] != result[4][1] # parent id
2024
 
                        or result[5][0] != result[5][1] # name
2025
 
                        or result[6][0] != result[6][1] # kind
2026
 
                        or result[7][0] != result[7][1] # executable
2027
 
                        ):
2028
 
                        result = (result[0],
2029
 
                            ((utf8_decode(result[1][0])[0]),
2030
 
                             utf8_decode(result[1][1])[0]),) + result[2:]
2031
 
                        yield result
2032
 
            if want_unversioned and not path_handled:
2033
 
                new_executable = bool(
2034
 
                    stat.S_ISREG(root_dir_info[3].st_mode)
2035
 
                    and stat.S_IEXEC & root_dir_info[3].st_mode)
2036
 
                yield (None, (None, current_root), True, (False, False),
2037
 
                    (None, None),
2038
 
                    (None, splitpath(current_root)[-1]),
2039
 
                    (None, root_dir_info[2]), (None, new_executable))
2040
 
            initial_key = (current_root, '', '')
2041
 
            block_index, _ = state._find_block_index_from_key(initial_key)
2042
 
            if block_index == 0:
2043
 
                # we have processed the total root already, but because the
2044
 
                # initial key matched it we should skip it here.
2045
 
                block_index +=1
2046
 
            if root_dir_info and root_dir_info[2] == 'tree-reference':
2047
 
                current_dir_info = None
2048
 
            else:
2049
 
                dir_iterator = osutils._walkdirs_utf8(root_abspath, prefix=current_root)
2050
 
                try:
2051
 
                    current_dir_info = dir_iterator.next()
2052
 
                except OSError, e:
2053
 
                    if e.errno in (errno.ENOENT, errno.ENOTDIR):
2054
 
                        # there may be directories in the inventory even though
2055
 
                        # this path is not a file on disk: so mark it as end of
2056
 
                        # iterator
2057
 
                        current_dir_info = None
2058
 
                    else:
2059
 
                        raise
2060
 
                else:
2061
 
                    if current_dir_info[0][0] == '':
2062
 
                        # remove .bzr from iteration
2063
 
                        bzr_index = bisect_left(current_dir_info[1], ('.bzr',))
2064
 
                        assert current_dir_info[1][bzr_index][0] == '.bzr'
2065
 
                        del current_dir_info[1][bzr_index]
2066
 
            # walk until both the directory listing and the versioned metadata
2067
 
            # are exhausted. 
2068
 
            if (block_index < len(state._dirblocks) and
2069
 
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
2070
 
                current_block = state._dirblocks[block_index]
2071
 
            else:
2072
 
                current_block = None
2073
 
            while (current_dir_info is not None or
2074
 
                   current_block is not None):
2075
 
                if (current_dir_info and current_block
2076
 
                    and current_dir_info[0][0] != current_block[0]):
2077
 
                    if current_dir_info[0][0] < current_block[0] :
2078
 
                        # filesystem data refers to paths not covered by the dirblock.
2079
 
                        # this has two possibilities:
2080
 
                        # A) it is versioned but empty, so there is no block for it
2081
 
                        # B) it is not versioned.
2082
 
                        # in either case it was processed by the containing directories walk:
2083
 
                        # if it is root/foo, when we walked root we emitted it,
2084
 
                        # or if we ere given root/foo to walk specifically, we
2085
 
                        # emitted it when checking the walk-root entries
2086
 
                        # advance the iterator and loop - we dont need to emit it.
2087
 
                        try:
2088
 
                            current_dir_info = dir_iterator.next()
2089
 
                        except StopIteration:
2090
 
                            current_dir_info = None
2091
 
                    else:
2092
 
                        # We have a dirblock entry for this location, but there
2093
 
                        # is no filesystem path for this. This is most likely
2094
 
                        # because a directory was removed from the disk.
2095
 
                        # We don't have to report the missing directory,
2096
 
                        # because that should have already been handled, but we
2097
 
                        # need to handle all of the files that are contained
2098
 
                        # within.
2099
 
                        for current_entry in current_block[1]:
2100
 
                            # entry referring to file not present on disk.
2101
 
                            # advance the entry only, after processing.
2102
 
                            for result in _process_entry(current_entry, None):
2103
 
                                # this check should probably be outside the loop: one
2104
 
                                # 'iterate two trees' api, and then _iter_changes filters
2105
 
                                # unchanged pairs. - RBC 20070226
2106
 
                                if (include_unchanged
2107
 
                                    or result[2]                    # content change
2108
 
                                    or result[3][0] != result[3][1] # versioned status
2109
 
                                    or result[4][0] != result[4][1] # parent id
2110
 
                                    or result[5][0] != result[5][1] # name
2111
 
                                    or result[6][0] != result[6][1] # kind
2112
 
                                    or result[7][0] != result[7][1] # executable
2113
 
                                    ):
2114
 
                                    result = (result[0],
2115
 
                                        ((utf8_decode(result[1][0])[0]),
2116
 
                                         utf8_decode(result[1][1])[0]),) + result[2:]
2117
 
                                    yield result
2118
 
                        block_index +=1
2119
 
                        if (block_index < len(state._dirblocks) and
2120
 
                            osutils.is_inside(current_root,
2121
 
                                              state._dirblocks[block_index][0])):
2122
 
                            current_block = state._dirblocks[block_index]
2123
 
                        else:
2124
 
                            current_block = None
2125
 
                    continue
2126
 
                entry_index = 0
2127
 
                if current_block and entry_index < len(current_block[1]):
2128
 
                    current_entry = current_block[1][entry_index]
2129
 
                else:
2130
 
                    current_entry = None
2131
 
                advance_entry = True
2132
 
                path_index = 0
2133
 
                if current_dir_info and path_index < len(current_dir_info[1]):
2134
 
                    current_path_info = current_dir_info[1][path_index]
2135
 
                    if current_path_info[2] == 'directory':
2136
 
                        if self.target._directory_is_tree_reference(
2137
 
                            current_path_info[0].decode('utf8')):
2138
 
                            current_path_info = current_path_info[:2] + \
2139
 
                                ('tree-reference',) + current_path_info[3:]
2140
 
                else:
2141
 
                    current_path_info = None
2142
 
                advance_path = True
2143
 
                path_handled = False
2144
 
                while (current_entry is not None or
2145
 
                    current_path_info is not None):
2146
 
                    if current_entry is None:
2147
 
                        # the check for path_handled when the path is adnvaced
2148
 
                        # will yield this path if needed.
2149
 
                        pass
2150
 
                    elif current_path_info is None:
2151
 
                        # no path is fine: the per entry code will handle it.
2152
 
                        for result in _process_entry(current_entry, current_path_info):
2153
 
                            # this check should probably be outside the loop: one
2154
 
                            # 'iterate two trees' api, and then _iter_changes filters
2155
 
                            # unchanged pairs. - RBC 20070226
2156
 
                            if (include_unchanged
2157
 
                                or result[2]                    # content change
2158
 
                                or result[3][0] != result[3][1] # versioned status
2159
 
                                or result[4][0] != result[4][1] # parent id
2160
 
                                or result[5][0] != result[5][1] # name
2161
 
                                or result[6][0] != result[6][1] # kind
2162
 
                                or result[7][0] != result[7][1] # executable
2163
 
                                ):
2164
 
                                result = (result[0],
2165
 
                                    ((utf8_decode(result[1][0])[0]),
2166
 
                                     utf8_decode(result[1][1])[0]),) + result[2:]
2167
 
                                yield result
2168
 
                    elif current_entry[0][1] != current_path_info[1]:
2169
 
                        if current_path_info[1] < current_entry[0][1]:
2170
 
                            # extra file on disk: pass for now, but only
2171
 
                            # increment the path, not the entry
2172
 
                            advance_entry = False
2173
 
                        else:
2174
 
                            # entry referring to file not present on disk.
2175
 
                            # advance the entry only, after processing.
2176
 
                            for result in _process_entry(current_entry, None):
2177
 
                                # this check should probably be outside the loop: one
2178
 
                                # 'iterate two trees' api, and then _iter_changes filters
2179
 
                                # unchanged pairs. - RBC 20070226
2180
 
                                path_handled = True
2181
 
                                if (include_unchanged
2182
 
                                    or result[2]                    # content change
2183
 
                                    or result[3][0] != result[3][1] # versioned status
2184
 
                                    or result[4][0] != result[4][1] # parent id
2185
 
                                    or result[5][0] != result[5][1] # name
2186
 
                                    or result[6][0] != result[6][1] # kind
2187
 
                                    or result[7][0] != result[7][1] # executable
2188
 
                                    ):
2189
 
                                    result = (result[0],
2190
 
                                        ((utf8_decode(result[1][0])[0]),
2191
 
                                         utf8_decode(result[1][1])[0]),) + result[2:]
2192
 
                                    yield result
2193
 
                            advance_path = False
2194
 
                    else:
2195
 
                        for result in _process_entry(current_entry, current_path_info):
2196
 
                            # this check should probably be outside the loop: one
2197
 
                            # 'iterate two trees' api, and then _iter_changes filters
2198
 
                            # unchanged pairs. - RBC 20070226
2199
 
                            path_handled = True
2200
 
                            if (include_unchanged
2201
 
                                or result[2]                    # content change
2202
 
                                or result[3][0] != result[3][1] # versioned status
2203
 
                                or result[4][0] != result[4][1] # parent id
2204
 
                                or result[5][0] != result[5][1] # name
2205
 
                                or result[6][0] != result[6][1] # kind
2206
 
                                or result[7][0] != result[7][1] # executable
2207
 
                                ):
2208
 
                                result = (result[0],
2209
 
                                    ((utf8_decode(result[1][0])[0]),
2210
 
                                     utf8_decode(result[1][1])[0]),) + result[2:]
2211
 
                                yield result
2212
 
                    if advance_entry and current_entry is not None:
2213
 
                        entry_index += 1
2214
 
                        if entry_index < len(current_block[1]):
2215
 
                            current_entry = current_block[1][entry_index]
2216
 
                        else:
2217
 
                            current_entry = None
2218
 
                    else:
2219
 
                        advance_entry = True # reset the advance flaga
2220
 
                    if advance_path and current_path_info is not None:
2221
 
                        if not path_handled:
2222
 
                            # unversioned in all regards
2223
 
                            if want_unversioned:
2224
 
                                new_executable = bool(
2225
 
                                    stat.S_ISREG(current_path_info[3].st_mode)
2226
 
                                    and stat.S_IEXEC & current_path_info[3].st_mode)
2227
 
                                if want_unversioned:
2228
 
                                    yield (None, (None, current_path_info[0]),
2229
 
                                        True,
2230
 
                                        (False, False),
2231
 
                                        (None, None),
2232
 
                                        (None, current_path_info[1]),
2233
 
                                        (None, current_path_info[2]),
2234
 
                                        (None, new_executable))
2235
 
                            # dont descend into this unversioned path if it is
2236
 
                            # a dir
2237
 
                            if current_path_info[2] in ('directory'):
2238
 
                                del current_dir_info[1][path_index]
2239
 
                                path_index -= 1
2240
 
                        # dont descend the disk iterator into any tree 
2241
 
                        # paths.
2242
 
                        if current_path_info[2] == 'tree-reference':
2243
 
                            del current_dir_info[1][path_index]
2244
 
                            path_index -= 1
2245
 
                        path_index += 1
2246
 
                        if path_index < len(current_dir_info[1]):
2247
 
                            current_path_info = current_dir_info[1][path_index]
2248
 
                            if current_path_info[2] == 'directory':
2249
 
                                if self.target._directory_is_tree_reference(
2250
 
                                    current_path_info[0].decode('utf8')):
2251
 
                                    current_path_info = current_path_info[:2] + \
2252
 
                                        ('tree-reference',) + current_path_info[3:]
2253
 
                        else:
2254
 
                            current_path_info = None
2255
 
                        path_handled = False
2256
 
                    else:
2257
 
                        advance_path = True # reset the advance flagg.
2258
 
                if current_block is not None:
2259
 
                    block_index += 1
2260
 
                    if (block_index < len(state._dirblocks) and
2261
 
                        osutils.is_inside(current_root, state._dirblocks[block_index][0])):
2262
 
                        current_block = state._dirblocks[block_index]
2263
 
                    else:
2264
 
                        current_block = None
2265
 
                if current_dir_info is not None:
2266
 
                    try:
2267
 
                        current_dir_info = dir_iterator.next()
2268
 
                    except StopIteration:
2269
 
                        current_dir_info = None
2270
 
 
 
2198
        iter_changes = self.target._iter_changes(include_unchanged,
 
2199
            use_filesystem_for_exec, search_specific_files, state,
 
2200
            source_index, target_index, want_unversioned, self.target)
 
2201
        return iter_changes.iter_changes()
2271
2202
 
2272
2203
    @staticmethod
2273
2204
    def is_compatible(source, target):
2274
2205
        # the target must be a dirstate working tree
2275
 
        if not isinstance(target, WorkingTree4):
 
2206
        if not isinstance(target, DirStateWorkingTree):
2276
2207
            return False
2277
 
        # the source must be a revtreee or dirstate rev tree.
 
2208
        # the source must be a revtree or dirstate rev tree.
2278
2209
        if not isinstance(source,
2279
2210
            (revisiontree.RevisionTree, DirStateRevisionTree)):
2280
2211
            return False
2281
2212
        # the source revid must be in the target dirstate
2282
 
        if not (source._revision_id == NULL_REVISION or
 
2213
        if not (source._revision_id == _mod_revision.NULL_REVISION or
2283
2214
            source._revision_id in target.get_parent_ids()):
2284
 
            # TODO: what about ghosts? it may well need to 
 
2215
            # TODO: what about ghosts? it may well need to
2285
2216
            # check for them explicitly.
2286
2217
            return False
2287
2218
        return True
2297
2228
 
2298
2229
    def convert(self, tree):
2299
2230
        # lock the control files not the tree, so that we dont get tree
2300
 
        # on-unlock behaviours, and so that noone else diddles with the 
 
2231
        # on-unlock behaviours, and so that noone else diddles with the
2301
2232
        # tree during upgrade.
2302
2233
        tree._control_files.lock_write()
2303
2234
        try:
 
2235
            tree.read_working_inventory()
2304
2236
            self.create_dirstate_data(tree)
2305
2237
            self.update_format(tree)
2306
2238
            self.remove_xml_files(tree)
2328
2260
 
2329
2261
    def update_format(self, tree):
2330
2262
        """Change the format marker."""
2331
 
        tree._control_files.put_utf8('format',
2332
 
            self.target_format.get_format_string())
 
2263
        tree._transport.put_bytes('format',
 
2264
            self.target_format.get_format_string(),
 
2265
            mode=tree.bzrdir._get_file_mode())
 
2266
 
 
2267
 
 
2268
class Converter4to5(object):
 
2269
    """Perform an in-place upgrade of format 4 to format 5 trees."""
 
2270
 
 
2271
    def __init__(self):
 
2272
        self.target_format = WorkingTreeFormat5()
 
2273
 
 
2274
    def convert(self, tree):
 
2275
        # lock the control files not the tree, so that we don't get tree
 
2276
        # on-unlock behaviours, and so that no-one else diddles with the
 
2277
        # tree during upgrade.
 
2278
        tree._control_files.lock_write()
 
2279
        try:
 
2280
            self.update_format(tree)
 
2281
        finally:
 
2282
            tree._control_files.unlock()
 
2283
 
 
2284
    def update_format(self, tree):
 
2285
        """Change the format marker."""
 
2286
        tree._transport.put_bytes('format',
 
2287
            self.target_format.get_format_string(),
 
2288
            mode=tree.bzrdir._get_file_mode())
 
2289
 
 
2290
 
 
2291
class Converter4or5to6(object):
 
2292
    """Perform an in-place upgrade of format 4 or 5 to format 6 trees."""
 
2293
 
 
2294
    def __init__(self):
 
2295
        self.target_format = WorkingTreeFormat6()
 
2296
 
 
2297
    def convert(self, tree):
 
2298
        # lock the control files not the tree, so that we don't get tree
 
2299
        # on-unlock behaviours, and so that no-one else diddles with the
 
2300
        # tree during upgrade.
 
2301
        tree._control_files.lock_write()
 
2302
        try:
 
2303
            self.init_custom_control_files(tree)
 
2304
            self.update_format(tree)
 
2305
        finally:
 
2306
            tree._control_files.unlock()
 
2307
 
 
2308
    def init_custom_control_files(self, tree):
 
2309
        """Initialize custom control files."""
 
2310
        tree._transport.put_bytes('views', '',
 
2311
            mode=tree.bzrdir._get_file_mode())
 
2312
 
 
2313
    def update_format(self, tree):
 
2314
        """Change the format marker."""
 
2315
        tree._transport.put_bytes('format',
 
2316
            self.target_format.get_format_string(),
 
2317
            mode=tree.bzrdir._get_file_mode())