~bzr-pqm/bzr/bzr.dev

« back to all changes in this revision

Viewing changes to bzrlib/workingtree_4.py

  • Committer: Martin Pool
  • Date: 2007-04-04 06:17:31 UTC
  • mto: This revision was merged to the branch mainline in revision 2397.
  • Revision ID: mbp@sourcefrog.net-20070404061731-tt2xrzllqhbodn83
Contents of TODO file moved into bug tracker

Show diffs side-by-side

added added

removed removed

Lines of Context:
1
 
# Copyright (C) 2007-2012 Canonical Ltd
 
1
# Copyright (C) 2005, 2006, 2007 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., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
 
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
16
16
 
17
17
"""WorkingTree4 format and implementation.
18
18
 
22
22
WorkingTree.open(dir).
23
23
"""
24
24
 
25
 
from __future__ import absolute_import
26
 
 
27
25
from cStringIO import StringIO
28
26
import os
29
27
import sys
30
28
 
31
29
from bzrlib.lazy_import import lazy_import
32
30
lazy_import(globals(), """
 
31
from bisect import bisect_left
 
32
import collections
 
33
from copy import deepcopy
33
34
import errno
 
35
import itertools
 
36
import operator
34
37
import stat
 
38
from time import time
 
39
import warnings
35
40
 
 
41
import bzrlib
36
42
from bzrlib import (
37
43
    bzrdir,
38
44
    cache_utf8,
39
 
    config,
40
45
    conflicts as _mod_conflicts,
41
 
    controldir,
42
 
    debug,
 
46
    delta,
43
47
    dirstate,
44
48
    errors,
45
 
    filters as _mod_filters,
46
49
    generate_ids,
 
50
    globbing,
 
51
    hashcache,
 
52
    ignores,
 
53
    merge,
47
54
    osutils,
48
 
    revision as _mod_revision,
49
55
    revisiontree,
50
 
    trace,
 
56
    textui,
51
57
    transform,
52
 
    views,
 
58
    urlutils,
 
59
    xml5,
 
60
    xml6,
53
61
    )
 
62
import bzrlib.branch
 
63
from bzrlib.transport import get_transport
 
64
import bzrlib.ui
54
65
""")
55
66
 
 
67
from bzrlib import symbol_versioning
56
68
from bzrlib.decorators import needs_read_lock, needs_write_lock
57
 
from bzrlib.inventory import Inventory, ROOT_ID, entry_factory
58
 
from bzrlib.lock import LogicalLockResult
59
 
from bzrlib.lockable_files import LockableFiles
 
69
from bzrlib.inventory import InventoryEntry, Inventory, ROOT_ID, entry_factory
 
70
from bzrlib.lockable_files import LockableFiles, TransportLock
60
71
from bzrlib.lockdir import LockDir
61
 
from bzrlib.mutabletree import (
62
 
    MutableTree,
63
 
    needs_tree_write_lock,
64
 
    )
 
72
import bzrlib.mutabletree
 
73
from bzrlib.mutabletree import needs_tree_write_lock
65
74
from bzrlib.osutils import (
66
75
    file_kind,
67
76
    isdir,
 
77
    normpath,
68
78
    pathjoin,
 
79
    rand_chars,
69
80
    realpath,
70
81
    safe_unicode,
71
 
    )
72
 
from bzrlib.symbol_versioning import (
73
 
    deprecated_in,
74
 
    deprecated_method,
75
 
    )
 
82
    splitpath,
 
83
    )
 
84
from bzrlib.trace import mutter, note
76
85
from bzrlib.transport.local import LocalTransport
77
 
from bzrlib.tree import (
78
 
    InterTree,
79
 
    InventoryTree,
80
 
    )
81
 
from bzrlib.workingtree import (
82
 
    InventoryWorkingTree,
83
 
    WorkingTree,
84
 
    WorkingTreeFormatMetaDir,
85
 
    )
86
 
 
87
 
 
88
 
class DirStateWorkingTree(InventoryWorkingTree):
 
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
# This is the Windows equivalent of ENOTDIR
 
100
# It is defined in pywin32.winerror, but we don't want a strong dependency for
 
101
# just an error code.
 
102
ERROR_DIRECTORY = 267
 
103
 
 
104
 
 
105
class WorkingTree4(WorkingTree3):
 
106
    """This is the Format 4 working tree.
 
107
 
 
108
    This differs from WorkingTree3 by:
 
109
     - Having a consolidated internal dirstate, stored in a
 
110
       randomly-accessible sorted file on disk.
 
111
     - Not having a regular inventory attribute.  One can be synthesized 
 
112
       on demand but this is expensive and should be avoided.
 
113
 
 
114
    This is new in bzr 0.15.
 
115
    """
89
116
 
90
117
    def __init__(self, basedir,
91
118
                 branch,
101
128
        """
102
129
        self._format = _format
103
130
        self.bzrdir = _bzrdir
 
131
        from bzrlib.trace import note, mutter
 
132
        assert isinstance(basedir, basestring), \
 
133
            "base directory %r is not a string" % basedir
104
134
        basedir = safe_unicode(basedir)
105
 
        trace.mutter("opening working tree %r", basedir)
 
135
        mutter("opening working tree %r", basedir)
106
136
        self._branch = branch
 
137
        assert isinstance(self.branch, bzrlib.branch.Branch), \
 
138
            "branch %r is not a Branch" % self.branch
107
139
        self.basedir = realpath(basedir)
108
140
        # if branch is at our basedir and is a format 6 or less
109
141
        # assume all other formats have their own control files.
 
142
        assert isinstance(_control_files, LockableFiles), \
 
143
            "_control_files must be a LockableFiles, not %r" % _control_files
110
144
        self._control_files = _control_files
111
 
        self._transport = self._control_files._transport
112
145
        self._dirty = None
113
146
        #-------------
114
147
        # during a read or write lock these objects are set, and are
116
149
        self._dirstate = None
117
150
        self._inventory = None
118
151
        #-------------
119
 
        self._setup_directory_is_tree_reference()
120
 
        self._detect_case_handling()
121
 
        self._rules_searcher = None
122
 
        self.views = self._make_views()
123
 
        #--- allow tests to select the dirstate iter_changes implementation
124
 
        self._iter_changes = dirstate._process_entry
125
152
 
126
153
    @needs_tree_write_lock
127
154
    def _add(self, files, ids, kinds):
129
156
        state = self.current_dirstate()
130
157
        for f, file_id, kind in zip(files, ids, kinds):
131
158
            f = f.strip('/')
 
159
            assert '//' not in f
 
160
            assert '..' not in f
132
161
            if self.path2id(f):
133
162
                # special case tree root handling.
134
163
                if f == '' and self.path2id(f) == ROOT_ID:
142
171
            state.add(f, file_id, kind, None, '')
143
172
        self._make_dirty(reset_inventory=True)
144
173
 
145
 
    def _get_check_refs(self):
146
 
        """Return the references needed to perform a check of this tree."""
147
 
        return [('trees', self.last_revision())]
148
 
 
149
174
    def _make_dirty(self, reset_inventory):
150
175
        """Make the tree state dirty.
151
176
 
159
184
    @needs_tree_write_lock
160
185
    def add_reference(self, sub_tree):
161
186
        # use standard implementation, which calls back to self._add
162
 
        #
 
187
        # 
163
188
        # So we don't store the reference_revision in the working dirstate,
164
 
        # it's just recorded at the moment of commit.
 
189
        # it's just recorded at the moment of commit. 
165
190
        self._add_reference(sub_tree)
166
191
 
167
192
    def break_lock(self):
203
228
 
204
229
    def _comparison_data(self, entry, path):
205
230
        kind, executable, stat_value = \
206
 
            WorkingTree._comparison_data(self, entry, path)
 
231
            WorkingTree3._comparison_data(self, entry, path)
207
232
        # it looks like a plain directory, but it's really a reference -- see
208
233
        # also kind()
209
 
        if (self._repo_supports_tree_reference and kind == 'directory'
210
 
            and entry is not None and entry.kind == 'tree-reference'):
 
234
        if (self._repo_supports_tree_reference and
 
235
            kind == 'directory' and
 
236
            self._directory_is_tree_reference(path)):
211
237
            kind = 'tree-reference'
212
238
        return kind, executable, stat_value
213
239
 
215
241
    def commit(self, message=None, revprops=None, *args, **kwargs):
216
242
        # mark the tree as dirty post commit - commit
217
243
        # can change the current versioned list by doing deletes.
218
 
        result = WorkingTree.commit(self, message, revprops, *args, **kwargs)
 
244
        result = WorkingTree3.commit(self, message, revprops, *args, **kwargs)
219
245
        self._make_dirty(reset_inventory=True)
220
246
        return result
221
247
 
239
265
            return self._dirstate
240
266
        local_path = self.bzrdir.get_workingtree_transport(None
241
267
            ).local_abspath('dirstate')
242
 
        self._dirstate = dirstate.DirState.on_file(local_path,
243
 
            self._sha1_provider(), self._worth_saving_limit())
 
268
        self._dirstate = dirstate.DirState.on_file(local_path)
244
269
        return self._dirstate
245
270
 
246
 
    def _sha1_provider(self):
247
 
        """A function that returns a SHA1Provider suitable for this tree.
248
 
 
249
 
        :return: None if content filtering is not supported by this tree.
250
 
          Otherwise, a SHA1Provider is returned that sha's the canonical
251
 
          form of files, i.e. after read filters are applied.
252
 
        """
253
 
        if self.supports_content_filtering():
254
 
            return ContentFilterAwareSHA1Provider(self)
255
 
        else:
256
 
            return None
257
 
 
258
 
    def _worth_saving_limit(self):
259
 
        """How many hash changes are ok before we must save the dirstate.
260
 
 
261
 
        :return: an integer. -1 means never save.
262
 
        """
263
 
        conf = self.get_config_stack()
264
 
        return conf.get('bzr.workingtree.worth_saving_limit')
 
271
    def _directory_is_tree_reference(self, relpath):
 
272
        # as a special case, if a directory contains control files then 
 
273
        # it's a tree reference, except that the root of the tree is not
 
274
        return relpath and osutils.isdir(self.abspath(relpath) + u"/.bzr")
 
275
        # TODO: We could ask all the control formats whether they
 
276
        # recognize this directory, but at the moment there's no cheap api
 
277
        # to do that.  Since we probably can only nest bzr checkouts and
 
278
        # they always use this name it's ok for now.  -- mbp 20060306
 
279
        #
 
280
        # FIXME: There is an unhandled case here of a subdirectory
 
281
        # containing .bzr but not a branch; that will probably blow up
 
282
        # when you try to commit it.  It might happen if there is a
 
283
        # checkout in a subdirectory.  This can be avoided by not adding
 
284
        # it.  mbp 20070306
265
285
 
266
286
    def filter_unversioned_files(self, paths):
267
287
        """Filter out paths that are versioned.
300
320
 
301
321
    def _generate_inventory(self):
302
322
        """Create and set self.inventory from the dirstate object.
303
 
 
 
323
        
304
324
        This is relatively expensive: we have to walk the entire dirstate.
305
325
        Ideally we would not, and can deprecate this function.
306
326
        """
310
330
        state._read_dirblocks_if_needed()
311
331
        root_key, current_entry = self._get_entry(path='')
312
332
        current_id = root_key[2]
313
 
        if not (current_entry[0][0] == 'd'): # directory
314
 
            raise AssertionError(current_entry)
 
333
        assert current_entry[0][0] == 'd' # directory
315
334
        inv = Inventory(root_id=current_id)
316
335
        # Turn some things into local variables
317
336
        minikind_to_kind = dirstate.DirState._minikind_to_kind
350
369
                    # add this entry to the parent map.
351
370
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
352
371
                elif kind == 'tree-reference':
353
 
                    if not self._repo_supports_tree_reference:
354
 
                        raise errors.UnsupportedOperation(
355
 
                            self._generate_inventory,
356
 
                            self.branch.repository)
 
372
                    assert self._repo_supports_tree_reference, \
 
373
                        "repository of %r " \
 
374
                        "doesn't support tree references " \
 
375
                        "required by entry %r" \
 
376
                        % (self, name)
357
377
                    inv_entry.reference_revision = link_or_sha1 or None
358
378
                elif kind != 'symlink':
359
379
                    raise AssertionError("unknown kind %r" % kind)
360
380
                # These checks cost us around 40ms on a 55k entry tree
361
 
                if file_id in inv_byid:
362
 
                    raise AssertionError('file_id %s already in'
363
 
                        ' inventory as %s' % (file_id, inv_byid[file_id]))
364
 
                if name_unicode in parent_ie.children:
365
 
                    raise AssertionError('name %r already in parent'
366
 
                        % (name_unicode,))
 
381
                assert file_id not in inv_byid, ('file_id %s already in'
 
382
                    ' inventory as %s' % (file_id, inv_byid[file_id]))
 
383
                assert name_unicode not in parent_ie.children
367
384
                inv_byid[file_id] = inv_entry
368
385
                parent_ie.children[name_unicode] = inv_entry
369
386
        self._inventory = inv
374
391
        If either file_id or path is supplied, it is used as the key to lookup.
375
392
        If both are supplied, the fastest lookup is used, and an error is
376
393
        raised if they do not both point at the same row.
377
 
 
 
394
        
378
395
        :param file_id: An optional unicode file_id to be looked up.
379
396
        :param path: An optional unicode path to be looked up.
380
397
        :return: The dirstate row tuple for path/file_id, or (None, None)
389
406
    def get_file_sha1(self, file_id, path=None, stat_value=None):
390
407
        # check file id is valid unconditionally.
391
408
        entry = self._get_entry(file_id=file_id, path=path)
392
 
        if entry[0] is None:
393
 
            raise errors.NoSuchId(self, file_id)
 
409
        assert entry[0] is not None, 'what error should this raise'
394
410
        if path is None:
395
411
            path = pathjoin(entry[0][0], entry[0][1]).decode('utf8')
396
412
 
397
413
        file_abspath = self.abspath(path)
398
414
        state = self.current_dirstate()
399
 
        if stat_value is None:
400
 
            try:
401
 
                stat_value = osutils.lstat(file_abspath)
402
 
            except OSError, e:
403
 
                if e.errno == errno.ENOENT:
404
 
                    return None
405
 
                else:
406
 
                    raise
407
 
        link_or_sha1 = dirstate.update_entry(state, entry, file_abspath,
408
 
            stat_value=stat_value)
 
415
        link_or_sha1 = state.update_entry(entry, file_abspath,
 
416
                                          stat_value=stat_value)
409
417
        if entry[1][0][0] == 'f':
410
 
            if link_or_sha1 is None:
411
 
                file_obj, statvalue = self.get_file_with_stat(file_id, path)
412
 
                try:
413
 
                    sha1 = osutils.sha_file(file_obj)
414
 
                finally:
415
 
                    file_obj.close()
416
 
                self._observed_sha1(file_id, path, (sha1, statvalue))
417
 
                return sha1
418
 
            else:
419
 
                return link_or_sha1
 
418
            return link_or_sha1
420
419
        return None
421
420
 
422
 
    def _get_root_inventory(self):
 
421
    def _get_inventory(self):
423
422
        """Get the inventory for the tree. This is only valid within a lock."""
424
 
        if 'evil' in debug.debug_flags:
425
 
            trace.mutter_callsite(2,
426
 
                "accessing .inventory forces a size of tree translation.")
427
423
        if self._inventory is not None:
428
424
            return self._inventory
429
425
        self._must_be_locked()
430
426
        self._generate_inventory()
431
427
        return self._inventory
432
428
 
433
 
    @deprecated_method(deprecated_in((2, 5, 0)))
434
 
    def _get_inventory(self):
435
 
        return self.root_inventory
436
 
 
437
429
    inventory = property(_get_inventory,
438
430
                         doc="Inventory of this Tree")
439
431
 
440
 
    root_inventory = property(_get_root_inventory,
441
 
        "Root inventory of this tree")
442
 
 
443
432
    @needs_read_lock
444
433
    def get_parent_ids(self):
445
434
        """See Tree.get_parent_ids.
446
 
 
 
435
        
447
436
        This implementation requests the ids list from the dirstate file.
448
437
        """
449
438
        return self.current_dirstate().get_parent_ids()
465
454
 
466
455
    def has_id(self, file_id):
467
456
        state = self.current_dirstate()
 
457
        file_id = osutils.safe_file_id(file_id)
468
458
        row, parents = self._get_entry(file_id=file_id)
469
459
        if row is None:
470
460
            return False
471
461
        return osutils.lexists(pathjoin(
472
462
                    self.basedir, row[0].decode('utf8'), row[1].decode('utf8')))
473
463
 
474
 
    def has_or_had_id(self, file_id):
475
 
        state = self.current_dirstate()
476
 
        row, parents = self._get_entry(file_id=file_id)
477
 
        return row is not None
478
 
 
479
464
    @needs_read_lock
480
465
    def id2path(self, file_id):
481
 
        "Convert a file-id to a path."
 
466
        file_id = osutils.safe_file_id(file_id)
482
467
        state = self.current_dirstate()
483
468
        entry = self._get_entry(file_id=file_id)
484
469
        if entry == (None, None):
486
471
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
487
472
        return path_utf8.decode('utf8')
488
473
 
489
 
    def _is_executable_from_path_and_stat_from_basis(self, path, stat_result):
490
 
        entry = self._get_entry(path=path)
491
 
        if entry == (None, None):
492
 
            return False # Missing entries are not executable
493
 
        return entry[1][0][3] # Executable?
494
 
 
495
 
    def is_executable(self, file_id, path=None):
496
 
        """Test if a file is executable or not.
497
 
 
498
 
        Note: The caller is expected to take a read-lock before calling this.
499
 
        """
500
 
        if not self._supports_executable():
 
474
    if not osutils.supports_executable():
 
475
        @needs_read_lock
 
476
        def is_executable(self, file_id, path=None):
 
477
            file_id = osutils.safe_file_id(file_id)
501
478
            entry = self._get_entry(file_id=file_id, path=path)
502
479
            if entry == (None, None):
503
480
                return False
504
481
            return entry[1][0][3]
505
 
        else:
506
 
            self._must_be_locked()
 
482
    else:
 
483
        @needs_read_lock
 
484
        def is_executable(self, file_id, path=None):
507
485
            if not path:
 
486
                file_id = osutils.safe_file_id(file_id)
508
487
                path = self.id2path(file_id)
509
 
            mode = osutils.lstat(self.abspath(path)).st_mode
 
488
            mode = os.lstat(self.abspath(path)).st_mode
510
489
            return bool(stat.S_ISREG(mode) and stat.S_IEXEC & mode)
511
490
 
512
 
    def all_file_ids(self):
513
 
        """See Tree.iter_all_file_ids"""
514
 
        self._must_be_locked()
515
 
        result = set()
516
 
        for key, tree_details in self.current_dirstate()._iter_entries():
517
 
            if tree_details[0][0] in ('a', 'r'): # relocated
518
 
                continue
519
 
            result.add(key[2])
520
 
        return result
521
 
 
522
491
    @needs_read_lock
523
492
    def __iter__(self):
524
493
        """Iterate through file_ids for this tree.
537
506
        return iter(result)
538
507
 
539
508
    def iter_references(self):
540
 
        if not self._repo_supports_tree_reference:
541
 
            # When the repo doesn't support references, we will have nothing to
542
 
            # return
543
 
            return
544
509
        for key, tree_details in self.current_dirstate()._iter_entries():
545
510
            if tree_details[0][0] in ('a', 'r'): # absent, relocated
546
511
                # not relevant to the working tree
548
513
            if not key[1]:
549
514
                # the root is not a reference.
550
515
                continue
551
 
            relpath = pathjoin(key[0].decode('utf8'), key[1].decode('utf8'))
 
516
            path = pathjoin(self.basedir, key[0].decode('utf8'), key[1].decode('utf8'))
552
517
            try:
553
 
                if self._kind(relpath) == 'tree-reference':
554
 
                    yield relpath, key[2]
 
518
                if self._kind(path) == 'tree-reference':
 
519
                    yield path, key[2]
555
520
            except errors.NoSuchFile:
556
521
                # path is missing on disk.
557
522
                continue
558
523
 
559
 
    def _observed_sha1(self, file_id, path, (sha1, statvalue)):
560
 
        """See MutableTree._observed_sha1."""
561
 
        state = self.current_dirstate()
562
 
        entry = self._get_entry(file_id=file_id, path=path)
563
 
        state._observed_sha1(entry, sha1, statvalue)
564
 
 
 
524
    @needs_read_lock
565
525
    def kind(self, file_id):
566
526
        """Return the kind of a file.
567
527
 
568
528
        This is always the actual kind that's on disk, regardless of what it
569
529
        was added as.
570
 
 
571
 
        Note: The caller is expected to take a read-lock before calling this.
572
530
        """
573
531
        relpath = self.id2path(file_id)
574
 
        if relpath is None:
575
 
            raise AssertionError(
576
 
                "path for id {%s} is None!" % file_id)
 
532
        assert relpath != None, \
 
533
            "path for id {%s} is None!" % file_id
577
534
        return self._kind(relpath)
578
535
 
579
536
    def _kind(self, relpath):
580
537
        abspath = self.abspath(relpath)
581
538
        kind = file_kind(abspath)
582
 
        if (self._repo_supports_tree_reference and kind == 'directory'):
583
 
            entry = self._get_entry(path=relpath)
584
 
            if entry[1] is not None:
585
 
                if entry[1][0][0] == 't':
586
 
                    kind = 'tree-reference'
 
539
        if (self._repo_supports_tree_reference and
 
540
            kind == 'directory' and
 
541
            self._directory_is_tree_reference(relpath)):
 
542
            kind = 'tree-reference'
587
543
        return kind
588
544
 
589
545
    @needs_read_lock
593
549
        if parent_ids:
594
550
            return parent_ids[0]
595
551
        else:
596
 
            return _mod_revision.NULL_REVISION
 
552
            return None
597
553
 
598
554
    def lock_read(self):
599
 
        """See Branch.lock_read, and WorkingTree.unlock.
600
 
 
601
 
        :return: A bzrlib.lock.LogicalLockResult.
602
 
        """
 
555
        """See Branch.lock_read, and WorkingTree.unlock."""
603
556
        self.branch.lock_read()
604
557
        try:
605
558
            self._control_files.lock_read()
618
571
        except:
619
572
            self.branch.unlock()
620
573
            raise
621
 
        return LogicalLockResult(self.unlock)
622
574
 
623
575
    def _lock_self_write(self):
624
576
        """This should be called after the branch is locked."""
639
591
        except:
640
592
            self.branch.unlock()
641
593
            raise
642
 
        return LogicalLockResult(self.unlock)
643
594
 
644
595
    def lock_tree_write(self):
645
 
        """See MutableTree.lock_tree_write, and WorkingTree.unlock.
646
 
 
647
 
        :return: A bzrlib.lock.LogicalLockResult.
648
 
        """
 
596
        """See MutableTree.lock_tree_write, and WorkingTree.unlock."""
649
597
        self.branch.lock_read()
650
 
        return self._lock_self_write()
 
598
        self._lock_self_write()
651
599
 
652
600
    def lock_write(self):
653
 
        """See MutableTree.lock_write, and WorkingTree.unlock.
654
 
 
655
 
        :return: A bzrlib.lock.LogicalLockResult.
656
 
        """
 
601
        """See MutableTree.lock_write, and WorkingTree.unlock."""
657
602
        self.branch.lock_write()
658
 
        return self._lock_self_write()
 
603
        self._lock_self_write()
659
604
 
660
605
    @needs_tree_write_lock
661
606
    def move(self, from_paths, to_dir, after=False):
663
608
        result = []
664
609
        if not from_paths:
665
610
            return result
 
611
 
666
612
        state = self.current_dirstate()
667
 
        if isinstance(from_paths, basestring):
668
 
            raise ValueError()
 
613
 
 
614
        assert not isinstance(from_paths, basestring)
669
615
        to_dir_utf8 = to_dir.encode('utf8')
670
616
        to_entry_dirname, to_basename = os.path.split(to_dir_utf8)
671
617
        id_index = state._get_id_index()
692
638
 
693
639
        if self._inventory is not None:
694
640
            update_inventory = True
695
 
            inv = self.root_inventory
 
641
            inv = self.inventory
 
642
            to_dir_ie = inv[to_dir_id]
696
643
            to_dir_id = to_entry[0][2]
697
 
            to_dir_ie = inv[to_dir_id]
698
644
        else:
699
645
            update_inventory = False
700
646
 
723
669
            new_entry = to_block[1][added_entry_index]
724
670
            rollbacks.append(lambda:state._make_absent(new_entry))
725
671
 
 
672
        # create rename entries and tuples
726
673
        for from_rel in from_paths:
727
674
            # from_rel is 'pathinroot/foo/bar'
728
675
            from_rel_utf8 = from_rel.encode('utf8')
731
678
            from_entry = self._get_entry(path=from_rel)
732
679
            if from_entry == (None, None):
733
680
                raise errors.BzrMoveFailedError(from_rel,to_dir,
734
 
                    errors.NotVersionedError(path=from_rel))
 
681
                    errors.NotVersionedError(path=str(from_rel)))
735
682
 
736
683
            from_id = from_entry[0][2]
737
684
            to_rel = pathjoin(to_dir, from_tail)
764
711
                if from_missing: # implicitly just update our path mapping
765
712
                    move_file = False
766
713
                elif not after:
767
 
                    raise errors.RenameFailedFilesExist(from_rel, to_rel)
 
714
                    raise errors.RenameFailedFilesExist(from_rel, to_rel,
 
715
                        extra="(Use --after to update the Bazaar id)")
768
716
 
769
717
            rollbacks = []
770
718
            def rollback_rename():
825
773
 
826
774
                if minikind == 'd':
827
775
                    def update_dirblock(from_dir, to_key, to_dir_utf8):
828
 
                        """Recursively update all entries in this dirblock."""
829
 
                        if from_dir == '':
830
 
                            raise AssertionError("renaming root not supported")
 
776
                        """all entries in this block need updating.
 
777
 
 
778
                        TODO: This is pretty ugly, and doesn't support
 
779
                        reverting, but it works.
 
780
                        """
 
781
                        assert from_dir != '', "renaming root not supported"
831
782
                        from_key = (from_dir, '')
832
783
                        from_block_idx, present = \
833
784
                            state._find_block_index_from_key(from_key)
843
794
                        to_block_index = state._ensure_block(
844
795
                            to_block_index, to_entry_index, to_dir_utf8)
845
796
                        to_block = state._dirblocks[to_block_index]
846
 
 
847
 
                        # Grab a copy since move_one may update the list.
848
 
                        for entry in from_block[1][:]:
849
 
                            if not (entry[0][0] == from_dir):
850
 
                                raise AssertionError()
 
797
                        for entry in from_block[1]:
 
798
                            assert entry[0][0] == from_dir
851
799
                            cur_details = entry[1][0]
852
800
                            to_key = (to_dir_utf8, entry[0][1], entry[0][2])
853
801
                            from_path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
854
802
                            to_path_utf8 = osutils.pathjoin(to_dir_utf8, entry[0][1])
855
803
                            minikind = cur_details[0]
856
 
                            if minikind in 'ar':
857
 
                                # Deleted children of a renamed directory
858
 
                                # Do not need to be updated.
859
 
                                # Children that have been renamed out of this
860
 
                                # directory should also not be updated
861
 
                                continue
862
804
                            move_one(entry, from_path_utf8=from_path_utf8,
863
805
                                     minikind=minikind,
864
806
                                     executable=cur_details[3],
878
820
                rollback_rename()
879
821
                raise
880
822
            result.append((from_rel, to_rel))
881
 
            state._mark_modified()
 
823
            state._dirblock_state = dirstate.DirState.IN_MEMORY_MODIFIED
882
824
            self._make_dirty(reset_inventory=False)
883
825
 
884
826
        return result
894
836
    @needs_read_lock
895
837
    def path2id(self, path):
896
838
        """Return the id for path in this tree."""
897
 
        if isinstance(path, list):
898
 
            if path == []:
899
 
                path = [""]
900
 
            path = osutils.pathjoin(*path)
901
839
        path = path.strip('/')
902
840
        entry = self._get_entry(path=path)
903
841
        if entry == (None, None):
916
854
        for tree in trees:
917
855
            if not (isinstance(tree, DirStateRevisionTree) and tree._revision_id in
918
856
                parents):
919
 
                return super(DirStateWorkingTree, self).paths2ids(paths,
920
 
                    trees, require_versioned)
 
857
                return super(WorkingTree4, self).paths2ids(paths, trees, require_versioned)
921
858
        search_indexes = [0] + [1 + parents.index(tree._revision_id) for tree in trees]
922
859
        # -- make all paths utf8 --
923
860
        paths_utf8 = set()
981
918
                    all_versioned = False
982
919
                    break
983
920
            if not all_versioned:
984
 
                raise errors.PathsNotVersionedError(
985
 
                    [p.decode('utf-8') for p in paths])
 
921
                raise errors.PathsNotVersionedError(paths)
986
922
        # -- remove redundancy in supplied paths to prevent over-scanning --
987
 
        search_paths = osutils.minimum_path_selection(paths)
988
 
        # sketch:
 
923
        search_paths = set()
 
924
        for path in paths:
 
925
            other_paths = paths.difference(set([path]))
 
926
            if not osutils.is_inside_any(other_paths, path):
 
927
                # this is a top level path, we must check it.
 
928
                search_paths.add(path)
 
929
        # sketch: 
989
930
        # for all search_indexs in each path at or under each element of
990
931
        # search_paths, if the detail is relocated: add the id, and add the
991
932
        # relocated path as one to search if its not searched already. If the
1037
978
            found_dir_names = set(dir_name_id[:2] for dir_name_id in found)
1038
979
            for dir_name in split_paths:
1039
980
                if dir_name not in found_dir_names:
1040
 
                    raise errors.PathsNotVersionedError(
1041
 
                        [p.decode('utf-8') for p in paths])
 
981
                    raise errors.PathsNotVersionedError(paths)
1042
982
 
1043
983
        for dir_name_id, trees_info in found.iteritems():
1044
984
            for index in search_indexes:
1048
988
 
1049
989
    def read_working_inventory(self):
1050
990
        """Read the working inventory.
1051
 
 
 
991
        
1052
992
        This is a meaningless operation for dirstate, but we obey it anyhow.
1053
993
        """
1054
 
        return self.root_inventory
 
994
        return self.inventory
1055
995
 
1056
996
    @needs_read_lock
1057
997
    def revision_tree(self, revision_id):
1059
999
 
1060
1000
        WorkingTree4 supplies revision_trees for any basis tree.
1061
1001
        """
 
1002
        revision_id = osutils.safe_revision_id(revision_id)
1062
1003
        dirstate = self.current_dirstate()
1063
1004
        parent_ids = dirstate.get_parent_ids()
1064
1005
        if revision_id not in parent_ids:
1071
1012
    @needs_tree_write_lock
1072
1013
    def set_last_revision(self, new_revision):
1073
1014
        """Change the last revision in the working tree."""
 
1015
        new_revision = osutils.safe_revision_id(new_revision)
1074
1016
        parents = self.get_parent_ids()
1075
 
        if new_revision in (_mod_revision.NULL_REVISION, None):
1076
 
            if len(parents) >= 2:
1077
 
                raise AssertionError(
1078
 
                    "setting the last parent to none with a pending merge is "
1079
 
                    "unsupported.")
 
1017
        if new_revision in (NULL_REVISION, None):
 
1018
            assert len(parents) < 2, (
 
1019
                "setting the last parent to none with a pending merge is "
 
1020
                "unsupported.")
1080
1021
            self.set_parent_ids([])
1081
1022
        else:
1082
1023
            self.set_parent_ids([new_revision] + parents[1:],
1085
1026
    @needs_tree_write_lock
1086
1027
    def set_parent_ids(self, revision_ids, allow_leftmost_as_ghost=False):
1087
1028
        """Set the parent ids to revision_ids.
1088
 
 
 
1029
        
1089
1030
        See also set_parent_trees. This api will try to retrieve the tree data
1090
1031
        for each element of revision_ids from the trees repository. If you have
1091
1032
        tree data already available, it is more efficient to use
1095
1036
        :param revision_ids: The revision_ids to set as the parent ids of this
1096
1037
            working tree. Any of these may be ghosts.
1097
1038
        """
 
1039
        revision_ids = [osutils.safe_revision_id(r) for r in revision_ids]
1098
1040
        trees = []
1099
1041
        for revision_id in revision_ids:
1100
1042
            try:
1106
1048
            except (errors.NoSuchRevision, errors.RevisionNotPresent):
1107
1049
                revtree = None
1108
1050
            trees.append((revision_id, revtree))
 
1051
        self.current_dirstate()._validate()
1109
1052
        self.set_parent_trees(trees,
1110
1053
            allow_leftmost_as_ghost=allow_leftmost_as_ghost)
 
1054
        self.current_dirstate()._validate()
1111
1055
 
1112
1056
    @needs_tree_write_lock
1113
1057
    def set_parent_trees(self, parents_list, allow_leftmost_as_ghost=False):
1118
1062
            parent tree - i.e. a ghost.
1119
1063
        """
1120
1064
        dirstate = self.current_dirstate()
 
1065
        dirstate._validate()
1121
1066
        if len(parents_list) > 0:
1122
1067
            if not allow_leftmost_as_ghost and parents_list[0][1] is None:
1123
1068
                raise errors.GhostRevisionUnusableHere(parents_list[0][0])
1124
1069
        real_trees = []
1125
1070
        ghosts = []
1126
 
 
1127
 
        parent_ids = [rev_id for rev_id, tree in parents_list]
1128
 
        graph = self.branch.repository.get_graph()
1129
 
        heads = graph.heads(parent_ids)
1130
 
        accepted_revisions = set()
1131
 
 
1132
1071
        # convert absent trees to the null tree, which we convert back to
1133
1072
        # missing on access.
1134
1073
        for rev_id, tree in parents_list:
1135
 
            if len(accepted_revisions) > 0:
1136
 
                # we always accept the first tree
1137
 
                if rev_id in accepted_revisions or rev_id not in heads:
1138
 
                    # We have already included either this tree, or its
1139
 
                    # descendent, so we skip it.
1140
 
                    continue
1141
 
            _mod_revision.check_not_reserved_id(rev_id)
 
1074
            rev_id = osutils.safe_revision_id(rev_id)
1142
1075
            if tree is not None:
1143
1076
                real_trees.append((rev_id, tree))
1144
1077
            else:
1145
1078
                real_trees.append((rev_id,
1146
 
                    self.branch.repository.revision_tree(
1147
 
                        _mod_revision.NULL_REVISION)))
 
1079
                    self.branch.repository.revision_tree(None)))
1148
1080
                ghosts.append(rev_id)
1149
 
            accepted_revisions.add(rev_id)
1150
 
        updated = False
1151
 
        if (len(real_trees) == 1
1152
 
            and not ghosts
1153
 
            and self.branch.repository._format.fast_deltas
1154
 
            and isinstance(real_trees[0][1],
1155
 
                revisiontree.InventoryRevisionTree)
1156
 
            and self.get_parent_ids()):
1157
 
            rev_id, rev_tree = real_trees[0]
1158
 
            basis_id = self.get_parent_ids()[0]
1159
 
            # There are times when basis_tree won't be in
1160
 
            # self.branch.repository, (switch, for example)
1161
 
            try:
1162
 
                basis_tree = self.branch.repository.revision_tree(basis_id)
1163
 
            except errors.NoSuchRevision:
1164
 
                # Fall back to the set_parent_trees(), since we can't use
1165
 
                # _make_delta if we can't get the RevisionTree
1166
 
                pass
1167
 
            else:
1168
 
                delta = rev_tree.root_inventory._make_delta(
1169
 
                    basis_tree.root_inventory)
1170
 
                dirstate.update_basis_by_delta(delta, rev_id)
1171
 
                updated = True
1172
 
        if not updated:
1173
 
            dirstate.set_parent_trees(real_trees, ghosts=ghosts)
 
1081
        dirstate._validate()
 
1082
        dirstate.set_parent_trees(real_trees, ghosts=ghosts)
 
1083
        dirstate._validate()
1174
1084
        self._make_dirty(reset_inventory=False)
 
1085
        dirstate._validate()
1175
1086
 
1176
1087
    def _set_root_id(self, file_id):
1177
1088
        """See WorkingTree.set_root_id."""
1180
1091
        if state._dirblock_state == dirstate.DirState.IN_MEMORY_MODIFIED:
1181
1092
            self._make_dirty(reset_inventory=True)
1182
1093
 
1183
 
    def _sha_from_stat(self, path, stat_result):
1184
 
        """Get a sha digest from the tree's stat cache.
1185
 
 
1186
 
        The default implementation assumes no stat cache is present.
1187
 
 
1188
 
        :param path: The path.
1189
 
        :param stat_result: The stat result being looked up.
1190
 
        """
1191
 
        return self.current_dirstate().sha1_from_stat(path, stat_result)
1192
 
 
1193
1094
    @needs_read_lock
1194
1095
    def supports_tree_reference(self):
1195
1096
        return self._repo_supports_tree_reference
1197
1098
    def unlock(self):
1198
1099
        """Unlock in format 4 trees needs to write the entire dirstate."""
1199
1100
        if self._control_files._lock_count == 1:
1200
 
            # do non-implementation specific cleanup
1201
 
            self._cleanup()
1202
 
 
1203
1101
            # eventually we should do signature checking during read locks for
1204
1102
            # dirstate updates.
1205
1103
            if self._control_files._lock_mode == 'w':
1235
1133
            return
1236
1134
        state = self.current_dirstate()
1237
1135
        state._read_dirblocks_if_needed()
1238
 
        ids_to_unversion = set(file_ids)
 
1136
        ids_to_unversion = set()
 
1137
        for file_id in file_ids:
 
1138
            ids_to_unversion.add(osutils.safe_file_id(file_id))
1239
1139
        paths_to_unversion = set()
1240
1140
        # sketch:
1241
1141
        # check if the root is to be unversioned, if so, assert for now.
1268
1168
                # just forget the whole block.
1269
1169
                entry_index = 0
1270
1170
                while entry_index < len(block[1]):
 
1171
                    # Mark this file id as having been removed
1271
1172
                    entry = block[1][entry_index]
1272
 
                    if entry[1][0][0] in 'ar':
1273
 
                        # don't remove absent or renamed entries
 
1173
                    ids_to_unversion.discard(entry[0][2])
 
1174
                    if (entry[1][0][0] == 'a'
 
1175
                        or not state._make_absent(entry)):
1274
1176
                        entry_index += 1
1275
 
                    else:
1276
 
                        # Mark this file id as having been removed
1277
 
                        ids_to_unversion.discard(entry[0][2])
1278
 
                        if not state._make_absent(entry):
1279
 
                            # The block has not shrunk.
1280
 
                            entry_index += 1
1281
1177
                # go to the next block. (At the moment we dont delete empty
1282
1178
                # dirblocks)
1283
1179
                block_index += 1
1304
1200
        # have to change the legacy inventory too.
1305
1201
        if self._inventory is not None:
1306
1202
            for file_id in file_ids:
1307
 
                if self._inventory.has_id(file_id):
1308
 
                    self._inventory.remove_recursive_id(file_id)
1309
 
 
1310
 
    @needs_tree_write_lock
1311
 
    def rename_one(self, from_rel, to_rel, after=False):
1312
 
        """See WorkingTree.rename_one"""
1313
 
        self.flush()
1314
 
        super(DirStateWorkingTree, self).rename_one(from_rel, to_rel, after)
1315
 
 
1316
 
    @needs_tree_write_lock
1317
 
    def apply_inventory_delta(self, changes):
1318
 
        """See MutableTree.apply_inventory_delta"""
1319
 
        state = self.current_dirstate()
1320
 
        state.update_by_delta(changes)
1321
 
        self._make_dirty(reset_inventory=True)
1322
 
 
1323
 
    def update_basis_by_delta(self, new_revid, delta):
1324
 
        """See MutableTree.update_basis_by_delta."""
1325
 
        if self.last_revision() == new_revid:
1326
 
            raise AssertionError()
1327
 
        self.current_dirstate().update_basis_by_delta(delta, new_revid)
 
1203
                self._inventory.remove_recursive_id(file_id)
1328
1204
 
1329
1205
    @needs_read_lock
1330
1206
    def _validate(self):
1333
1209
    @needs_tree_write_lock
1334
1210
    def _write_inventory(self, inv):
1335
1211
        """Write inventory as the current inventory."""
1336
 
        if self._dirty:
1337
 
            raise AssertionError("attempting to write an inventory when the "
1338
 
                "dirstate is dirty will lose pending changes")
1339
 
        had_inventory = self._inventory is not None
1340
 
        # Setting self._inventory = None forces the dirstate to regenerate the
1341
 
        # working inventory. We do this because self.inventory may be inv, or
1342
 
        # may have been modified, and either case would prevent a clean delta
1343
 
        # being created.
1344
 
        self._inventory = None
1345
 
        # generate a delta,
1346
 
        delta = inv._make_delta(self.root_inventory)
1347
 
        # and apply it.
1348
 
        self.apply_inventory_delta(delta)
1349
 
        if had_inventory:
 
1212
        assert not self._dirty, "attempting to write an inventory when the dirstate is dirty will cause data loss"
 
1213
        self.current_dirstate().set_state_from_inventory(inv)
 
1214
        self._make_dirty(reset_inventory=False)
 
1215
        if self._inventory is not None:
1350
1216
            self._inventory = inv
1351
1217
        self.flush()
1352
1218
 
1353
 
    @needs_tree_write_lock
1354
 
    def reset_state(self, revision_ids=None):
1355
 
        """Reset the state of the working tree.
1356
 
 
1357
 
        This does a hard-reset to a last-known-good state. This is a way to
1358
 
        fix if something got corrupted (like the .bzr/checkout/dirstate file)
1359
 
        """
1360
 
        if revision_ids is None:
1361
 
            revision_ids = self.get_parent_ids()
1362
 
        if not revision_ids:
1363
 
            base_tree = self.branch.repository.revision_tree(
1364
 
                _mod_revision.NULL_REVISION)
1365
 
            trees = []
1366
 
        else:
1367
 
            trees = zip(revision_ids,
1368
 
                        self.branch.repository.revision_trees(revision_ids))
1369
 
            base_tree = trees[0][1]
1370
 
        state = self.current_dirstate()
1371
 
        # We don't support ghosts yet
1372
 
        state.set_state_from_scratch(base_tree.root_inventory, trees, [])
1373
 
 
1374
 
 
1375
 
class ContentFilterAwareSHA1Provider(dirstate.SHA1Provider):
1376
 
 
1377
 
    def __init__(self, tree):
1378
 
        self.tree = tree
1379
 
 
1380
 
    def sha1(self, abspath):
1381
 
        """See dirstate.SHA1Provider.sha1()."""
1382
 
        filters = self.tree._content_filter_stack(
1383
 
            self.tree.relpath(osutils.safe_unicode(abspath)))
1384
 
        return _mod_filters.internal_size_sha_file_byname(abspath, filters)[1]
1385
 
 
1386
 
    def stat_and_sha1(self, abspath):
1387
 
        """See dirstate.SHA1Provider.stat_and_sha1()."""
1388
 
        filters = self.tree._content_filter_stack(
1389
 
            self.tree.relpath(osutils.safe_unicode(abspath)))
1390
 
        file_obj = file(abspath, 'rb', 65000)
1391
 
        try:
1392
 
            statvalue = os.fstat(file_obj.fileno())
1393
 
            if filters:
1394
 
                file_obj = _mod_filters.filtered_input_file(file_obj, filters)
1395
 
            sha1 = osutils.size_sha_file(file_obj)[1]
1396
 
        finally:
1397
 
            file_obj.close()
1398
 
        return statvalue, sha1
1399
 
 
1400
 
 
1401
 
class ContentFilteringDirStateWorkingTree(DirStateWorkingTree):
1402
 
    """Dirstate working tree that supports content filtering.
1403
 
 
1404
 
    The dirstate holds the hash and size of the canonical form of the file, 
1405
 
    and most methods must return that.
1406
 
    """
1407
 
 
1408
 
    def _file_content_summary(self, path, stat_result):
1409
 
        # This is to support the somewhat obsolete path_content_summary method
1410
 
        # with content filtering: see
1411
 
        # <https://bugs.launchpad.net/bzr/+bug/415508>.
1412
 
        #
1413
 
        # If the dirstate cache is up to date and knows the hash and size,
1414
 
        # return that.
1415
 
        # Otherwise if there are no content filters, return the on-disk size
1416
 
        # and leave the hash blank.
1417
 
        # Otherwise, read and filter the on-disk file and use its size and
1418
 
        # hash.
1419
 
        #
1420
 
        # The dirstate doesn't store the size of the canonical form so we
1421
 
        # can't trust it for content-filtered trees.  We just return None.
1422
 
        dirstate_sha1 = self._dirstate.sha1_from_stat(path, stat_result)
1423
 
        executable = self._is_executable_from_path_and_stat(path, stat_result)
1424
 
        return ('file', None, executable, dirstate_sha1)
1425
 
 
1426
 
 
1427
 
class WorkingTree4(DirStateWorkingTree):
1428
 
    """This is the Format 4 working tree.
1429
 
 
1430
 
    This differs from WorkingTree by:
1431
 
     - Having a consolidated internal dirstate, stored in a
1432
 
       randomly-accessible sorted file on disk.
1433
 
     - Not having a regular inventory attribute.  One can be synthesized
1434
 
       on demand but this is expensive and should be avoided.
1435
 
 
1436
 
    This is new in bzr 0.15.
1437
 
    """
1438
 
 
1439
 
 
1440
 
class WorkingTree5(ContentFilteringDirStateWorkingTree):
1441
 
    """This is the Format 5 working tree.
1442
 
 
1443
 
    This differs from WorkingTree4 by:
1444
 
     - Supporting content filtering.
1445
 
 
1446
 
    This is new in bzr 1.11.
1447
 
    """
1448
 
 
1449
 
 
1450
 
class WorkingTree6(ContentFilteringDirStateWorkingTree):
1451
 
    """This is the Format 6 working tree.
1452
 
 
1453
 
    This differs from WorkingTree5 by:
1454
 
     - Supporting a current view that may mask the set of files in a tree
1455
 
       impacted by most user operations.
1456
 
 
1457
 
    This is new in bzr 1.14.
1458
 
    """
1459
 
 
1460
 
    def _make_views(self):
1461
 
        return views.PathBasedViews(self)
1462
 
 
1463
 
 
1464
 
class DirStateWorkingTreeFormat(WorkingTreeFormatMetaDir):
1465
 
 
1466
 
    missing_parent_conflicts = True
1467
 
 
1468
 
    supports_versioned_directories = True
1469
 
 
1470
 
    _lock_class = LockDir
1471
 
    _lock_file_name = 'lock'
1472
 
 
1473
 
    def _open_control_files(self, a_bzrdir):
1474
 
        transport = a_bzrdir.get_workingtree_transport(None)
1475
 
        return LockableFiles(transport, self._lock_file_name,
1476
 
                             self._lock_class)
1477
 
 
1478
 
    def initialize(self, a_bzrdir, revision_id=None, from_branch=None,
1479
 
                   accelerator_tree=None, hardlink=False):
 
1219
 
 
1220
class WorkingTreeFormat4(WorkingTreeFormat3):
 
1221
    """The first consolidated dirstate working tree format.
 
1222
 
 
1223
    This format:
 
1224
        - exists within a metadir controlling .bzr
 
1225
        - includes an explicit version marker for the workingtree control
 
1226
          files, separate from the BzrDir format
 
1227
        - modifies the hash cache format
 
1228
        - is new in bzr 0.15
 
1229
        - uses a LockDir to guard access to it.
 
1230
    """
 
1231
 
 
1232
    upgrade_recommended = False
 
1233
 
 
1234
    def get_format_string(self):
 
1235
        """See WorkingTreeFormat.get_format_string()."""
 
1236
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
 
1237
 
 
1238
    def get_format_description(self):
 
1239
        """See WorkingTreeFormat.get_format_description()."""
 
1240
        return "Working tree format 4"
 
1241
 
 
1242
    def initialize(self, a_bzrdir, revision_id=None):
1480
1243
        """See WorkingTreeFormat.initialize().
1481
1244
 
1482
1245
        :param revision_id: allows creating a working tree at a different
1483
 
            revision than the branch is at.
1484
 
        :param accelerator_tree: A tree which can be used for retrieving file
1485
 
            contents more quickly than the revision tree, i.e. a workingtree.
1486
 
            The revision tree will be used for cases where accelerator_tree's
1487
 
            content is different.
1488
 
        :param hardlink: If true, hard-link files from accelerator_tree,
1489
 
            where possible.
 
1246
        revision than the branch is at.
1490
1247
 
1491
 
        These trees get an initial random root id, if their repository supports
1492
 
        rich root data, TREE_ROOT otherwise.
 
1248
        These trees get an initial random root id.
1493
1249
        """
 
1250
        revision_id = osutils.safe_revision_id(revision_id)
1494
1251
        if not isinstance(a_bzrdir.transport, LocalTransport):
1495
1252
            raise errors.NotLocalUrl(a_bzrdir.transport.base)
1496
1253
        transport = a_bzrdir.get_workingtree_transport(self)
1497
1254
        control_files = self._open_control_files(a_bzrdir)
1498
1255
        control_files.create_lock()
1499
1256
        control_files.lock_write()
1500
 
        transport.put_bytes('format', self.as_string(),
1501
 
            mode=a_bzrdir._get_file_mode())
1502
 
        if from_branch is not None:
1503
 
            branch = from_branch
1504
 
        else:
1505
 
            branch = a_bzrdir.open_branch()
 
1257
        control_files.put_utf8('format', self.get_format_string())
 
1258
        branch = a_bzrdir.open_branch()
1506
1259
        if revision_id is None:
1507
1260
            revision_id = branch.last_revision()
1508
1261
        local_path = transport.local_abspath('dirstate')
1510
1263
        state = dirstate.DirState.initialize(local_path)
1511
1264
        state.unlock()
1512
1265
        del state
1513
 
        wt = self._tree_class(a_bzrdir.root_transport.local_abspath('.'),
 
1266
        wt = WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
1514
1267
                         branch,
1515
1268
                         _format=self,
1516
1269
                         _bzrdir=a_bzrdir,
1517
1270
                         _control_files=control_files)
1518
1271
        wt._new_tree()
1519
1272
        wt.lock_tree_write()
 
1273
        wt.current_dirstate()._validate()
1520
1274
        try:
1521
 
            self._init_custom_control_files(wt)
1522
 
            if revision_id in (None, _mod_revision.NULL_REVISION):
1523
 
                if branch.repository.supports_rich_root():
1524
 
                    wt._set_root_id(generate_ids.gen_root_id())
1525
 
                else:
1526
 
                    wt._set_root_id(ROOT_ID)
 
1275
            if revision_id in (None, NULL_REVISION):
 
1276
                wt._set_root_id(generate_ids.gen_root_id())
1527
1277
                wt.flush()
1528
 
            basis = None
1529
 
            # frequently, we will get here due to branching.  The accelerator
1530
 
            # tree will be the tree from the branch, so the desired basis
1531
 
            # tree will often be a parent of the accelerator tree.
1532
 
            if accelerator_tree is not None:
1533
 
                try:
1534
 
                    basis = accelerator_tree.revision_tree(revision_id)
1535
 
                except errors.NoSuchRevision:
1536
 
                    pass
1537
 
            if basis is None:
1538
 
                basis = branch.repository.revision_tree(revision_id)
1539
 
            if revision_id == _mod_revision.NULL_REVISION:
1540
 
                parents_list = []
1541
 
            else:
1542
 
                parents_list = [(revision_id, basis)]
 
1278
                wt.current_dirstate()._validate()
 
1279
            wt.set_last_revision(revision_id)
 
1280
            wt.flush()
 
1281
            basis = wt.basis_tree()
1543
1282
            basis.lock_read()
1544
 
            try:
1545
 
                wt.set_parent_trees(parents_list, allow_leftmost_as_ghost=True)
 
1283
            # if the basis has a root id we have to use that; otherwise we use
 
1284
            # a new random one
 
1285
            basis_root_id = basis.get_root_id()
 
1286
            if basis_root_id is not None:
 
1287
                wt._set_root_id(basis_root_id)
1546
1288
                wt.flush()
1547
 
                # if the basis has a root id we have to use that; otherwise we
1548
 
                # use a new random one
1549
 
                basis_root_id = basis.get_root_id()
1550
 
                if basis_root_id is not None:
1551
 
                    wt._set_root_id(basis_root_id)
1552
 
                    wt.flush()
1553
 
                if wt.supports_content_filtering():
1554
 
                    # The original tree may not have the same content filters
1555
 
                    # applied so we can't safely build the inventory delta from
1556
 
                    # the source tree.
1557
 
                    delta_from_tree = False
1558
 
                else:
1559
 
                    delta_from_tree = True
1560
 
                # delta_from_tree is safe even for DirStateRevisionTrees,
1561
 
                # because wt4.apply_inventory_delta does not mutate the input
1562
 
                # inventory entries.
1563
 
                transform.build_tree(basis, wt, accelerator_tree,
1564
 
                                     hardlink=hardlink,
1565
 
                                     delta_from_tree=delta_from_tree)
1566
 
                for hook in MutableTree.hooks['post_build_tree']:
1567
 
                    hook(wt)
1568
 
            finally:
1569
 
                basis.unlock()
 
1289
            transform.build_tree(basis, wt)
 
1290
            basis.unlock()
1570
1291
        finally:
1571
1292
            control_files.unlock()
1572
1293
            wt.unlock()
1573
1294
        return wt
1574
1295
 
1575
 
    def _init_custom_control_files(self, wt):
1576
 
        """Subclasses with custom control files should override this method.
1577
 
 
1578
 
        The working tree and control files are locked for writing when this
1579
 
        method is called.
1580
 
 
1581
 
        :param wt: the WorkingTree object
1582
 
        """
1583
 
 
1584
 
    def open(self, a_bzrdir, _found=False):
1585
 
        """Return the WorkingTree object for a_bzrdir
1586
 
 
1587
 
        _found is a private parameter, do not use it. It is used to indicate
1588
 
               if format probing has already been done.
1589
 
        """
1590
 
        if not _found:
1591
 
            # we are being called directly and must probe.
1592
 
            raise NotImplementedError
1593
 
        if not isinstance(a_bzrdir.transport, LocalTransport):
1594
 
            raise errors.NotLocalUrl(a_bzrdir.transport.base)
1595
 
        wt = self._open(a_bzrdir, self._open_control_files(a_bzrdir))
1596
 
        return wt
1597
 
 
1598
1296
    def _open(self, a_bzrdir, control_files):
1599
1297
        """Open the tree itself.
1600
1298
 
1601
1299
        :param a_bzrdir: the dir for the tree.
1602
1300
        :param control_files: the control files for the tree.
1603
1301
        """
1604
 
        return self._tree_class(a_bzrdir.root_transport.local_abspath('.'),
 
1302
        return WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
1605
1303
                           branch=a_bzrdir.open_branch(),
1606
1304
                           _format=self,
1607
1305
                           _bzrdir=a_bzrdir,
1608
1306
                           _control_files=control_files)
1609
1307
 
1610
1308
    def __get_matchingbzrdir(self):
1611
 
        return self._get_matchingbzrdir()
1612
 
 
1613
 
    def _get_matchingbzrdir(self):
1614
 
        """Overrideable method to get a bzrdir for testing."""
1615
1309
        # please test against something that will let us do tree references
1616
 
        return controldir.format_registry.make_bzrdir(
1617
 
            'development-subtree')
 
1310
        return bzrdir.format_registry.make_bzrdir(
 
1311
            'dirstate-with-subtree')
1618
1312
 
1619
1313
    _matchingbzrdir = property(__get_matchingbzrdir)
1620
1314
 
1621
1315
 
1622
 
class WorkingTreeFormat4(DirStateWorkingTreeFormat):
1623
 
    """The first consolidated dirstate working tree format.
1624
 
 
1625
 
    This format:
1626
 
        - exists within a metadir controlling .bzr
1627
 
        - includes an explicit version marker for the workingtree control
1628
 
          files, separate from the ControlDir format
1629
 
        - modifies the hash cache format
1630
 
        - is new in bzr 0.15
1631
 
        - uses a LockDir to guard access to it.
1632
 
    """
1633
 
 
1634
 
    upgrade_recommended = False
1635
 
 
1636
 
    _tree_class = WorkingTree4
1637
 
 
1638
 
    @classmethod
1639
 
    def get_format_string(cls):
1640
 
        """See WorkingTreeFormat.get_format_string()."""
1641
 
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
1642
 
 
1643
 
    def get_format_description(self):
1644
 
        """See WorkingTreeFormat.get_format_description()."""
1645
 
        return "Working tree format 4"
1646
 
 
1647
 
 
1648
 
class WorkingTreeFormat5(DirStateWorkingTreeFormat):
1649
 
    """WorkingTree format supporting content filtering.
1650
 
    """
1651
 
 
1652
 
    upgrade_recommended = False
1653
 
 
1654
 
    _tree_class = WorkingTree5
1655
 
 
1656
 
    @classmethod
1657
 
    def get_format_string(cls):
1658
 
        """See WorkingTreeFormat.get_format_string()."""
1659
 
        return "Bazaar Working Tree Format 5 (bzr 1.11)\n"
1660
 
 
1661
 
    def get_format_description(self):
1662
 
        """See WorkingTreeFormat.get_format_description()."""
1663
 
        return "Working tree format 5"
1664
 
 
1665
 
    def supports_content_filtering(self):
1666
 
        return True
1667
 
 
1668
 
 
1669
 
class WorkingTreeFormat6(DirStateWorkingTreeFormat):
1670
 
    """WorkingTree format supporting views.
1671
 
    """
1672
 
 
1673
 
    upgrade_recommended = False
1674
 
 
1675
 
    _tree_class = WorkingTree6
1676
 
 
1677
 
    @classmethod
1678
 
    def get_format_string(cls):
1679
 
        """See WorkingTreeFormat.get_format_string()."""
1680
 
        return "Bazaar Working Tree Format 6 (bzr 1.14)\n"
1681
 
 
1682
 
    def get_format_description(self):
1683
 
        """See WorkingTreeFormat.get_format_description()."""
1684
 
        return "Working tree format 6"
1685
 
 
1686
 
    def _init_custom_control_files(self, wt):
1687
 
        """Subclasses with custom control files should override this method."""
1688
 
        wt._transport.put_bytes('views', '', mode=wt.bzrdir._get_file_mode())
1689
 
 
1690
 
    def supports_content_filtering(self):
1691
 
        return True
1692
 
 
1693
 
    def supports_views(self):
1694
 
        return True
1695
 
 
1696
 
    def _get_matchingbzrdir(self):
1697
 
        """Overrideable method to get a bzrdir for testing."""
1698
 
        # We use 'development-subtree' instead of '2a', because we have a
1699
 
        # few tests that want to test tree references
1700
 
        return bzrdir.format_registry.make_bzrdir('development-subtree')
1701
 
 
1702
 
 
1703
 
class DirStateRevisionTree(InventoryTree):
1704
 
    """A revision tree pulling the inventory from a dirstate.
1705
 
    
1706
 
    Note that this is one of the historical (ie revision) trees cached in the
1707
 
    dirstate for easy access, not the workingtree.
1708
 
    """
 
1316
class DirStateRevisionTree(Tree):
 
1317
    """A revision tree pulling the inventory from a dirstate."""
1709
1318
 
1710
1319
    def __init__(self, dirstate, revision_id, repository):
1711
1320
        self._dirstate = dirstate
1712
 
        self._revision_id = revision_id
 
1321
        self._revision_id = osutils.safe_revision_id(revision_id)
1713
1322
        self._repository = repository
1714
1323
        self._inventory = None
1715
1324
        self._locked = 0
1716
1325
        self._dirstate_locked = False
1717
 
        self._repo_supports_tree_reference = getattr(
1718
 
            repository._format, "supports_tree_reference",
1719
 
            False)
1720
1326
 
1721
1327
    def __repr__(self):
1722
1328
        return "<%s of %s in %s>" % \
1723
1329
            (self.__class__.__name__, self._revision_id, self._dirstate)
1724
1330
 
1725
 
    def annotate_iter(self, file_id,
1726
 
                      default_revision=_mod_revision.CURRENT_REVISION):
 
1331
    def annotate_iter(self, file_id):
1727
1332
        """See Tree.annotate_iter"""
1728
 
        text_key = (file_id, self.get_file_revision(file_id))
1729
 
        annotations = self._repository.texts.annotate(text_key)
1730
 
        return [(key[-1], line) for (key, line) in annotations]
 
1333
        w = self._repository.weave_store.get_weave(file_id,
 
1334
                           self._repository.get_transaction())
 
1335
        return w.annotate_iter(self.inventory[file_id].revision)
1731
1336
 
1732
1337
    def _comparison_data(self, entry, path):
1733
1338
        """See Tree._comparison_data."""
1751
1356
    def get_root_id(self):
1752
1357
        return self.path2id('')
1753
1358
 
1754
 
    def id2path(self, file_id):
1755
 
        "Convert a file-id to a path."
1756
 
        entry = self._get_entry(file_id=file_id)
1757
 
        if entry == (None, None):
1758
 
            raise errors.NoSuchId(tree=self, file_id=file_id)
1759
 
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
1760
 
        return path_utf8.decode('utf8')
1761
 
 
1762
 
    def iter_references(self):
1763
 
        if not self._repo_supports_tree_reference:
1764
 
            # When the repo doesn't support references, we will have nothing to
1765
 
            # return
1766
 
            return iter([])
1767
 
        # Otherwise, fall back to the default implementation
1768
 
        return super(DirStateRevisionTree, self).iter_references()
1769
 
 
1770
1359
    def _get_parent_index(self):
1771
1360
        """Return the index in the dirstate referenced by this tree."""
1772
1361
        return self._dirstate.get_parent_ids().index(self._revision_id) + 1
1777
1366
        If either file_id or path is supplied, it is used as the key to lookup.
1778
1367
        If both are supplied, the fastest lookup is used, and an error is
1779
1368
        raised if they do not both point at the same row.
1780
 
 
 
1369
        
1781
1370
        :param file_id: An optional unicode file_id to be looked up.
1782
1371
        :param path: An optional unicode path to be looked up.
1783
1372
        :return: The dirstate row tuple for path/file_id, or (None, None)
1784
1373
        """
1785
1374
        if file_id is None and path is None:
1786
1375
            raise errors.BzrError('must supply file_id or path')
 
1376
        file_id = osutils.safe_file_id(file_id)
1787
1377
        if path is not None:
1788
1378
            path = path.encode('utf8')
1789
1379
        parent_index = self._get_parent_index()
1790
 
        return self._dirstate._get_entry(parent_index, fileid_utf8=file_id,
1791
 
            path_utf8=path)
 
1380
        return self._dirstate._get_entry(parent_index, fileid_utf8=file_id, path_utf8=path)
1792
1381
 
1793
1382
    def _generate_inventory(self):
1794
1383
        """Create and set self.inventory from the dirstate object.
1798
1387
 
1799
1388
        This is relatively expensive: we have to walk the entire dirstate.
1800
1389
        """
1801
 
        if not self._locked:
1802
 
            raise AssertionError(
1803
 
                'cannot generate inventory of an unlocked '
1804
 
                'dirstate revision tree')
 
1390
        assert self._locked, 'cannot generate inventory of an unlocked '\
 
1391
            'dirstate revision tree'
1805
1392
        # separate call for profiling - makes it clear where the costs are.
1806
1393
        self._dirstate._read_dirblocks_if_needed()
1807
 
        if self._revision_id not in self._dirstate.get_parent_ids():
1808
 
            raise AssertionError(
1809
 
                'parent %s has disappeared from %s' % (
1810
 
                self._revision_id, self._dirstate.get_parent_ids()))
 
1394
        assert self._revision_id in self._dirstate.get_parent_ids(), \
 
1395
            'parent %s has disappeared from %s' % (
 
1396
            self._revision_id, self._dirstate.get_parent_ids())
1811
1397
        parent_index = self._dirstate.get_parent_ids().index(self._revision_id) + 1
1812
1398
        # This is identical now to the WorkingTree _generate_inventory except
1813
1399
        # for the tree index use.
1814
1400
        root_key, current_entry = self._dirstate._get_entry(parent_index, path_utf8='')
1815
1401
        current_id = root_key[2]
1816
 
        if current_entry[parent_index][0] != 'd':
1817
 
            raise AssertionError()
 
1402
        assert current_entry[parent_index][0] == 'd'
1818
1403
        inv = Inventory(root_id=current_id, revision_id=self._revision_id)
1819
1404
        inv.root.revision = current_entry[parent_index][4]
1820
1405
        # Turn some things into local variables
1851
1436
                elif kind == 'directory':
1852
1437
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
1853
1438
                elif kind == 'symlink':
 
1439
                    inv_entry.executable = False
 
1440
                    inv_entry.text_size = None
1854
1441
                    inv_entry.symlink_target = utf8_decode(fingerprint)[0]
1855
1442
                elif kind == 'tree-reference':
1856
1443
                    inv_entry.reference_revision = fingerprint or None
1858
1445
                    raise AssertionError("cannot convert entry %r into an InventoryEntry"
1859
1446
                            % entry)
1860
1447
                # These checks cost us around 40ms on a 55k entry tree
1861
 
                if file_id in inv_byid:
1862
 
                    raise AssertionError('file_id %s already in'
1863
 
                        ' inventory as %s' % (file_id, inv_byid[file_id]))
1864
 
                if name_unicode in parent_ie.children:
1865
 
                    raise AssertionError('name %r already in parent'
1866
 
                        % (name_unicode,))
 
1448
                assert file_id not in inv_byid
 
1449
                assert name_unicode not in parent_ie.children
1867
1450
                inv_byid[file_id] = inv_entry
1868
1451
                parent_ie.children[name_unicode] = inv_entry
1869
1452
        self._inventory = inv
1876
1459
        # Make sure the file exists
1877
1460
        entry = self._get_entry(file_id, path=path)
1878
1461
        if entry == (None, None): # do we raise?
1879
 
            raise errors.NoSuchId(self, file_id)
 
1462
            return None
1880
1463
        parent_index = self._get_parent_index()
1881
1464
        last_changed_revision = entry[1][parent_index][4]
1882
 
        try:
1883
 
            rev = self._repository.get_revision(last_changed_revision)
1884
 
        except errors.NoSuchRevision:
1885
 
            raise errors.FileTimestampUnavailable(self.id2path(file_id))
1886
 
        return rev.timestamp
 
1465
        return self._repository.get_revision(last_changed_revision).timestamp
1887
1466
 
1888
1467
    def get_file_sha1(self, file_id, path=None, stat_value=None):
1889
1468
        entry = self._get_entry(file_id=file_id, path=path)
1893
1472
            return parent_details[1]
1894
1473
        return None
1895
1474
 
1896
 
    @needs_read_lock
1897
 
    def get_file_revision(self, file_id):
1898
 
        inv, inv_file_id = self._unpack_file_id(file_id)
1899
 
        return inv[inv_file_id].revision
1900
 
 
1901
 
    def get_file(self, file_id, path=None):
 
1475
    def get_file(self, file_id):
1902
1476
        return StringIO(self.get_file_text(file_id))
1903
1477
 
 
1478
    def get_file_lines(self, file_id):
 
1479
        ie = self.inventory[file_id]
 
1480
        return self._repository.weave_store.get_weave(file_id,
 
1481
                self._repository.get_transaction()).get_lines(ie.revision)
 
1482
 
1904
1483
    def get_file_size(self, file_id):
1905
 
        """See Tree.get_file_size"""
1906
 
        inv, inv_file_id = self._unpack_file_id(file_id)
1907
 
        return inv[inv_file_id].text_size
 
1484
        return self.inventory[file_id].text_size
1908
1485
 
1909
 
    def get_file_text(self, file_id, path=None):
1910
 
        content = None
1911
 
        for _, content_iter in self.iter_files_bytes([(file_id, None)]):
1912
 
            if content is not None:
1913
 
                raise AssertionError('iter_files_bytes returned'
1914
 
                    ' too many entries')
1915
 
            # For each entry returned by iter_files_bytes, we must consume the
1916
 
            # content_iter before we step the files iterator.
1917
 
            content = ''.join(content_iter)
1918
 
        if content is None:
1919
 
            raise AssertionError('iter_files_bytes did not return'
1920
 
                ' the requested data')
1921
 
        return content
 
1486
    def get_file_text(self, file_id):
 
1487
        return ''.join(self.get_file_lines(file_id))
1922
1488
 
1923
1489
    def get_reference_revision(self, file_id, path=None):
1924
 
        inv, inv_file_id = self._unpack_file_id(file_id)
1925
 
        return inv[inv_file_id].reference_revision
1926
 
 
1927
 
    def iter_files_bytes(self, desired_files):
1928
 
        """See Tree.iter_files_bytes.
1929
 
 
1930
 
        This version is implemented on top of Repository.iter_files_bytes"""
1931
 
        parent_index = self._get_parent_index()
1932
 
        repo_desired_files = []
1933
 
        for file_id, identifier in desired_files:
1934
 
            entry = self._get_entry(file_id)
1935
 
            if entry == (None, None):
1936
 
                raise errors.NoSuchId(self, file_id)
1937
 
            repo_desired_files.append((file_id, entry[1][parent_index][4],
1938
 
                                       identifier))
1939
 
        return self._repository.iter_files_bytes(repo_desired_files)
1940
 
 
1941
 
    def get_symlink_target(self, file_id, path=None):
 
1490
        return self.inventory[file_id].reference_revision
 
1491
 
 
1492
    def get_symlink_target(self, file_id):
1942
1493
        entry = self._get_entry(file_id=file_id)
1943
1494
        parent_index = self._get_parent_index()
1944
1495
        if entry[1][parent_index][0] != 'l':
1945
1496
            return None
1946
1497
        else:
1947
 
            target = entry[1][parent_index][1]
1948
 
            target = target.decode('utf8')
1949
 
            return target
 
1498
            # At present, none of the tree implementations supports non-ascii
 
1499
            # symlink targets. So we will just assume that the dirstate path is
 
1500
            # correct.
 
1501
            return entry[1][parent_index][1]
1950
1502
 
1951
1503
    def get_revision_id(self):
1952
1504
        """Return the revision id for this tree."""
1953
1505
        return self._revision_id
1954
1506
 
1955
 
    def _get_root_inventory(self):
 
1507
    def _get_inventory(self):
1956
1508
        if self._inventory is not None:
1957
1509
            return self._inventory
1958
1510
        self._must_be_locked()
1959
1511
        self._generate_inventory()
1960
1512
        return self._inventory
1961
1513
 
1962
 
    root_inventory = property(_get_root_inventory,
1963
 
                         doc="Inventory of this Tree")
1964
 
 
1965
 
    @deprecated_method(deprecated_in((2, 5, 0)))
1966
 
    def _get_inventory(self):
1967
 
        return self.root_inventory
1968
 
 
1969
1514
    inventory = property(_get_inventory,
1970
1515
                         doc="Inventory of this Tree")
1971
1516
 
1977
1522
        return bool(self.path2id(filename))
1978
1523
 
1979
1524
    def kind(self, file_id):
1980
 
        entry = self._get_entry(file_id=file_id)[1]
1981
 
        if entry is None:
1982
 
            raise errors.NoSuchId(tree=self, file_id=file_id)
1983
 
        parent_index = self._get_parent_index()
1984
 
        return dirstate.DirState._minikind_to_kind[entry[parent_index][0]]
1985
 
 
1986
 
    def stored_kind(self, file_id):
1987
 
        """See Tree.stored_kind"""
1988
 
        return self.kind(file_id)
1989
 
 
1990
 
    def path_content_summary(self, path):
1991
 
        """See Tree.path_content_summary."""
1992
 
        inv, inv_file_id = self._path2inv_file_id(path)
1993
 
        if inv_file_id is None:
1994
 
            return ('missing', None, None, None)
1995
 
        entry = inv[inv_file_id]
1996
 
        kind = entry.kind
1997
 
        if kind == 'file':
1998
 
            return (kind, entry.text_size, entry.executable, entry.text_sha1)
1999
 
        elif kind == 'symlink':
2000
 
            return (kind, None, None, entry.symlink_target)
2001
 
        else:
2002
 
            return (kind, None, None, None)
 
1525
        return self.inventory[file_id].kind
2003
1526
 
2004
1527
    def is_executable(self, file_id, path=None):
2005
 
        inv, inv_file_id = self._unpack_file_id(file_id)
2006
 
        ie = inv[inv_file_id]
 
1528
        ie = self.inventory[file_id]
2007
1529
        if ie.kind != "file":
2008
 
            return False
 
1530
            return None
2009
1531
        return ie.executable
2010
1532
 
2011
 
    def is_locked(self):
2012
 
        return self._locked
2013
 
 
2014
 
    def list_files(self, include_root=False, from_dir=None, recursive=True):
 
1533
    def list_files(self, include_root=False):
2015
1534
        # We use a standard implementation, because DirStateRevisionTree is
2016
1535
        # dealing with one of the parents of the current state
2017
 
        if from_dir is None:
2018
 
            inv = self.root_inventory
2019
 
            from_dir_id = None
2020
 
        else:
2021
 
            inv, from_dir_id = self._path2inv_file_id(from_dir)
2022
 
            if from_dir_id is None:
2023
 
                # Directory not versioned
2024
 
                return
2025
 
        # FIXME: Support nested trees
2026
 
        entries = inv.iter_entries(from_dir=from_dir_id, recursive=recursive)
2027
 
        if inv.root is not None and not include_root and from_dir is None:
 
1536
        inv = self._get_inventory()
 
1537
        entries = inv.iter_entries()
 
1538
        if self.inventory.root is not None and not include_root:
2028
1539
            entries.next()
2029
1540
        for path, entry in entries:
2030
1541
            yield path, 'V', entry.kind, entry.file_id, entry
2031
1542
 
2032
1543
    def lock_read(self):
2033
 
        """Lock the tree for a set of operations.
2034
 
 
2035
 
        :return: A bzrlib.lock.LogicalLockResult.
2036
 
        """
 
1544
        """Lock the tree for a set of operations."""
2037
1545
        if not self._locked:
2038
1546
            self._repository.lock_read()
2039
1547
            if self._dirstate._lock_token is None:
2040
1548
                self._dirstate.lock_read()
2041
1549
                self._dirstate_locked = True
2042
1550
        self._locked += 1
2043
 
        return LogicalLockResult(self.unlock)
2044
1551
 
2045
1552
    def _must_be_locked(self):
2046
1553
        if not self._locked:
2050
1557
    def path2id(self, path):
2051
1558
        """Return the id for path in this tree."""
2052
1559
        # lookup by path: faster than splitting and walking the ivnentory.
2053
 
        if isinstance(path, list):
2054
 
            if path == []:
2055
 
                path = [""]
2056
 
            path = osutils.pathjoin(*path)
2057
1560
        entry = self._get_entry(path=path)
2058
1561
        if entry == (None, None):
2059
1562
            return None
2071
1574
                self._dirstate_locked = False
2072
1575
            self._repository.unlock()
2073
1576
 
2074
 
    @needs_read_lock
2075
 
    def supports_tree_reference(self):
2076
 
        return self._repo_supports_tree_reference
2077
 
 
2078
1577
    def walkdirs(self, prefix=""):
2079
1578
        # TODO: jam 20070215 This is the lazy way by using the RevisionTree
2080
 
        # implementation based on an inventory.
 
1579
        # implementation based on an inventory.  
2081
1580
        # This should be cleaned up to use the much faster Dirstate code
2082
1581
        # So for now, we just build up the parent inventory, and extract
2083
1582
        # it the same way RevisionTree does.
2084
1583
        _directory = 'directory'
2085
 
        inv = self._get_root_inventory()
 
1584
        inv = self._get_inventory()
2086
1585
        top_id = inv.path2id(prefix)
2087
1586
        if top_id is None:
2088
1587
            pending = []
2112
1611
 
2113
1612
class InterDirStateTree(InterTree):
2114
1613
    """Fast path optimiser for changes_from with dirstate trees.
2115
 
 
2116
 
    This is used only when both trees are in the dirstate working file, and
2117
 
    the source is any parent within the dirstate, and the destination is
 
1614
    
 
1615
    This is used only when both trees are in the dirstate working file, and 
 
1616
    the source is any parent within the dirstate, and the destination is 
2118
1617
    the current working tree of the same dirstate.
2119
1618
    """
2120
1619
    # this could be generalized to allow comparisons between any trees in the
2129
1628
    def make_source_parent_tree(source, target):
2130
1629
        """Change the source tree into a parent of the target."""
2131
1630
        revid = source.commit('record tree')
2132
 
        target.branch.fetch(source.branch, revid)
 
1631
        target.branch.repository.fetch(source.branch.repository, revid)
2133
1632
        target.set_parent_ids([revid])
2134
1633
        return target.basis_tree(), target
2135
1634
 
2136
 
    @classmethod
2137
 
    def make_source_parent_tree_python_dirstate(klass, test_case, source, target):
2138
 
        result = klass.make_source_parent_tree(source, target)
2139
 
        result[1]._iter_changes = dirstate.ProcessEntryPython
2140
 
        return result
2141
 
 
2142
 
    @classmethod
2143
 
    def make_source_parent_tree_compiled_dirstate(klass, test_case, source,
2144
 
                                                  target):
2145
 
        from bzrlib.tests.test__dirstate_helpers import \
2146
 
            compiled_dirstate_helpers_feature
2147
 
        test_case.requireFeature(compiled_dirstate_helpers_feature)
2148
 
        from bzrlib._dirstate_helpers_pyx import ProcessEntryC
2149
 
        result = klass.make_source_parent_tree(source, target)
2150
 
        result[1]._iter_changes = ProcessEntryC
2151
 
        return result
2152
 
 
2153
1635
    _matching_from_tree_format = WorkingTreeFormat4()
2154
1636
    _matching_to_tree_format = WorkingTreeFormat4()
2155
 
 
2156
 
    @classmethod
2157
 
    def _test_mutable_trees_to_test_trees(klass, test_case, source, target):
2158
 
        # This method shouldn't be called, because we have python and C
2159
 
        # specific flavours.
2160
 
        raise NotImplementedError
2161
 
 
2162
 
    def iter_changes(self, include_unchanged=False,
 
1637
    _test_mutable_trees_to_test_trees = make_source_parent_tree
 
1638
 
 
1639
    def _iter_changes(self, include_unchanged=False,
2163
1640
                      specific_files=None, pb=None, extra_trees=[],
2164
1641
                      require_versioned=True, want_unversioned=False):
2165
1642
        """Return the changes from source to target.
2166
1643
 
2167
 
        :return: An iterator that yields tuples. See InterTree.iter_changes
 
1644
        :return: An iterator that yields tuples. See InterTree._iter_changes
2168
1645
            for details.
2169
1646
        :param specific_files: An optional list of file paths to restrict the
2170
1647
            comparison to. When mapping filenames to ids, all matches in all
2181
1658
            output. An unversioned file is defined as one with (False, False)
2182
1659
            for the versioned pair.
2183
1660
        """
 
1661
        utf8_decode_or_none = cache_utf8._utf8_decode_with_None
 
1662
        _minikind_to_kind = dirstate.DirState._minikind_to_kind
 
1663
        # NB: show_status depends on being able to pass in non-versioned files
 
1664
        # and report them as unknown
2184
1665
        # TODO: handle extra trees in the dirstate.
2185
 
        if (extra_trees or specific_files == []):
 
1666
        # TODO: handle comparisons as an empty tree as a different special
 
1667
        # case? mbp 20070226
 
1668
        if extra_trees or (self.source._revision_id == NULL_REVISION):
2186
1669
            # we can't fast-path these cases (yet)
2187
 
            return super(InterDirStateTree, self).iter_changes(
 
1670
            for f in super(InterDirStateTree, self)._iter_changes(
2188
1671
                include_unchanged, specific_files, pb, extra_trees,
2189
 
                require_versioned, want_unversioned=want_unversioned)
 
1672
                require_versioned, want_unversioned=want_unversioned):
 
1673
                yield f
 
1674
            return
2190
1675
        parent_ids = self.target.get_parent_ids()
2191
 
        if not (self.source._revision_id in parent_ids
2192
 
                or self.source._revision_id == _mod_revision.NULL_REVISION):
2193
 
            raise AssertionError(
2194
 
                "revision {%s} is not stored in {%s}, but %s "
2195
 
                "can only be used for trees stored in the dirstate"
2196
 
                % (self.source._revision_id, self.target, self.iter_changes))
 
1676
        assert (self.source._revision_id in parent_ids), \
 
1677
                "revision {%s} is not stored in {%s}, but %s " \
 
1678
                "can only be used for trees stored in the dirstate" \
 
1679
                % (self.source._revision_id, self.target, self._iter_changes)
2197
1680
        target_index = 0
2198
 
        if self.source._revision_id == _mod_revision.NULL_REVISION:
 
1681
        if self.source._revision_id == NULL_REVISION:
2199
1682
            source_index = None
2200
1683
            indices = (target_index,)
2201
1684
        else:
2202
 
            if not (self.source._revision_id in parent_ids):
2203
 
                raise AssertionError(
2204
 
                    "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
2205
 
                    self.source._revision_id, parent_ids))
 
1685
            assert (self.source._revision_id in parent_ids), \
 
1686
                "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
 
1687
                self.source._revision_id, parent_ids)
2206
1688
            source_index = 1 + parent_ids.index(self.source._revision_id)
2207
 
            indices = (source_index, target_index)
 
1689
            indices = (source_index,target_index)
2208
1690
        # -- make all specific_files utf8 --
2209
1691
        if specific_files:
2210
1692
            specific_files_utf8 = set()
2211
1693
            for path in specific_files:
2212
 
                # Note, if there are many specific files, using cache_utf8
2213
 
                # would be good here.
2214
1694
                specific_files_utf8.add(path.encode('utf8'))
2215
1695
            specific_files = specific_files_utf8
2216
1696
        else:
2217
1697
            specific_files = set([''])
2218
1698
        # -- specific_files is now a utf8 path set --
2219
 
 
2220
1699
        # -- get the state object and prepare it.
2221
1700
        state = self.target.current_dirstate()
2222
1701
        state._read_dirblocks_if_needed()
 
1702
        def _entries_for_path(path):
 
1703
            """Return a list with all the entries that match path for all ids.
 
1704
            """
 
1705
            dirname, basename = os.path.split(path)
 
1706
            key = (dirname, basename, '')
 
1707
            block_index, present = state._find_block_index_from_key(key)
 
1708
            if not present:
 
1709
                # the block which should contain path is absent.
 
1710
                return []
 
1711
            result = []
 
1712
            block = state._dirblocks[block_index][1]
 
1713
            entry_index, _ = state._find_entry_index(key, block)
 
1714
            # we may need to look at multiple entries at this path: walk while the specific_files match.
 
1715
            while (entry_index < len(block) and
 
1716
                block[entry_index][0][0:2] == key[0:2]):
 
1717
                result.append(block[entry_index])
 
1718
                entry_index += 1
 
1719
            return result
2223
1720
        if require_versioned:
2224
1721
            # -- check all supplied paths are versioned in a search tree. --
2225
 
            not_versioned = []
 
1722
            all_versioned = True
2226
1723
            for path in specific_files:
2227
 
                path_entries = state._entries_for_path(path)
 
1724
                path_entries = _entries_for_path(path)
2228
1725
                if not path_entries:
2229
1726
                    # this specified path is not present at all: error
2230
 
                    not_versioned.append(path.decode('utf-8'))
2231
 
                    continue
 
1727
                    all_versioned = False
 
1728
                    break
2232
1729
                found_versioned = False
2233
1730
                # for each id at this path
2234
1731
                for entry in path_entries:
2241
1738
                if not found_versioned:
2242
1739
                    # none of the indexes was not 'absent' at all ids for this
2243
1740
                    # path.
2244
 
                    not_versioned.append(path.decode('utf-8'))
2245
 
            if len(not_versioned) > 0:
2246
 
                raise errors.PathsNotVersionedError(not_versioned)
 
1741
                    all_versioned = False
 
1742
                    break
 
1743
            if not all_versioned:
 
1744
                raise errors.PathsNotVersionedError(specific_files)
2247
1745
        # -- remove redundancy in supplied specific_files to prevent over-scanning --
2248
 
        search_specific_files = osutils.minimum_path_selection(specific_files)
 
1746
        search_specific_files = set()
 
1747
        for path in specific_files:
 
1748
            other_specific_files = specific_files.difference(set([path]))
 
1749
            if not osutils.is_inside_any(other_specific_files, path):
 
1750
                # this is a top level path, we must check it.
 
1751
                search_specific_files.add(path)
 
1752
        # sketch: 
 
1753
        # compare source_index and target_index at or under each element of search_specific_files.
 
1754
        # follow the following comparison table. Note that we only want to do diff operations when
 
1755
        # the target is fdl because thats when the walkdirs logic will have exposed the pathinfo 
 
1756
        # for the target.
 
1757
        # cases:
 
1758
        # 
 
1759
        # Source | Target | disk | action
 
1760
        #   r    | fdlt   |      | add source to search, add id path move and perform
 
1761
        #        |        |      | diff check on source-target
 
1762
        #   r    | fdlt   |  a   | dangling file that was present in the basis. 
 
1763
        #        |        |      | ???
 
1764
        #   r    |  a     |      | add source to search
 
1765
        #   r    |  a     |  a   | 
 
1766
        #   r    |  r     |      | this path is present in a non-examined tree, skip.
 
1767
        #   r    |  r     |  a   | this path is present in a non-examined tree, skip.
 
1768
        #   a    | fdlt   |      | add new id
 
1769
        #   a    | fdlt   |  a   | dangling locally added file, skip
 
1770
        #   a    |  a     |      | not present in either tree, skip
 
1771
        #   a    |  a     |  a   | not present in any tree, skip
 
1772
        #   a    |  r     |      | not present in either tree at this path, skip as it
 
1773
        #        |        |      | may not be selected by the users list of paths.
 
1774
        #   a    |  r     |  a   | not present in either tree at this path, skip as it
 
1775
        #        |        |      | may not be selected by the users list of paths.
 
1776
        #  fdlt  | fdlt   |      | content in both: diff them
 
1777
        #  fdlt  | fdlt   |  a   | deleted locally, but not unversioned - show as deleted ?
 
1778
        #  fdlt  |  a     |      | unversioned: output deleted id for now
 
1779
        #  fdlt  |  a     |  a   | unversioned and deleted: output deleted id
 
1780
        #  fdlt  |  r     |      | relocated in this tree, so add target to search.
 
1781
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1782
        #        |        |      | this id at the other path.
 
1783
        #  fdlt  |  r     |  a   | relocated in this tree, so add target to search.
 
1784
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1785
        #        |        |      | this id at the other path.
 
1786
 
 
1787
        # for all search_indexs in each path at or under each element of
 
1788
        # search_specific_files, if the detail is relocated: add the id, and add the
 
1789
        # relocated path as one to search if its not searched already. If the
 
1790
        # detail is not relocated, add the id.
 
1791
        searched_specific_files = set()
 
1792
        NULL_PARENT_DETAILS = dirstate.DirState.NULL_PARENT_DETAILS
 
1793
        # Using a list so that we can access the values and change them in
 
1794
        # nested scope. Each one is [path, file_id, entry]
 
1795
        last_source_parent = [None, None, None]
 
1796
        last_target_parent = [None, None, None]
2249
1797
 
2250
1798
        use_filesystem_for_exec = (sys.platform != 'win32')
2251
 
        iter_changes = self.target._iter_changes(include_unchanged,
2252
 
            use_filesystem_for_exec, search_specific_files, state,
2253
 
            source_index, target_index, want_unversioned, self.target)
2254
 
        return iter_changes.iter_changes()
 
1799
 
 
1800
        def _process_entry(entry, path_info):
 
1801
            """Compare an entry and real disk to generate delta information.
 
1802
 
 
1803
            :param path_info: top_relpath, basename, kind, lstat, abspath for
 
1804
                the path of entry. If None, then the path is considered absent.
 
1805
                (Perhaps we should pass in a concrete entry for this ?)
 
1806
                Basename is returned as a utf8 string because we expect this
 
1807
                tuple will be ignored, and don't want to take the time to
 
1808
                decode.
 
1809
            """
 
1810
            if source_index is None:
 
1811
                source_details = NULL_PARENT_DETAILS
 
1812
            else:
 
1813
                source_details = entry[1][source_index]
 
1814
            target_details = entry[1][target_index]
 
1815
            target_minikind = target_details[0]
 
1816
            if path_info is not None and target_minikind in 'fdlt':
 
1817
                assert target_index == 0
 
1818
                link_or_sha1 = state.update_entry(entry, abspath=path_info[4],
 
1819
                                                  stat_value=path_info[3])
 
1820
                # The entry may have been modified by update_entry
 
1821
                target_details = entry[1][target_index]
 
1822
                target_minikind = target_details[0]
 
1823
            else:
 
1824
                link_or_sha1 = None
 
1825
            source_minikind = source_details[0]
 
1826
            if source_minikind in 'fdltr' and target_minikind in 'fdlt':
 
1827
                # claimed content in both: diff
 
1828
                #   r    | fdlt   |      | add source to search, add id path move and perform
 
1829
                #        |        |      | diff check on source-target
 
1830
                #   r    | fdlt   |  a   | dangling file that was present in the basis.
 
1831
                #        |        |      | ???
 
1832
                if source_minikind in 'r':
 
1833
                    # add the source to the search path to find any children it
 
1834
                    # has.  TODO ? : only add if it is a container ?
 
1835
                    if not osutils.is_inside_any(searched_specific_files,
 
1836
                                                 source_details[1]):
 
1837
                        search_specific_files.add(source_details[1])
 
1838
                    # generate the old path; this is needed for stating later
 
1839
                    # as well.
 
1840
                    old_path = source_details[1]
 
1841
                    old_dirname, old_basename = os.path.split(old_path)
 
1842
                    path = pathjoin(entry[0][0], entry[0][1])
 
1843
                    old_entry = state._get_entry(source_index,
 
1844
                                                 path_utf8=old_path)
 
1845
                    # update the source details variable to be the real
 
1846
                    # location.
 
1847
                    source_details = old_entry[1][source_index]
 
1848
                    source_minikind = source_details[0]
 
1849
                else:
 
1850
                    old_dirname = entry[0][0]
 
1851
                    old_basename = entry[0][1]
 
1852
                    old_path = path = pathjoin(old_dirname, old_basename)
 
1853
                if path_info is None:
 
1854
                    # the file is missing on disk, show as removed.
 
1855
                    content_change = True
 
1856
                    target_kind = None
 
1857
                    target_exec = False
 
1858
                else:
 
1859
                    # source and target are both versioned and disk file is present.
 
1860
                    target_kind = path_info[2]
 
1861
                    if target_kind == 'directory':
 
1862
                        if source_minikind != 'd':
 
1863
                            content_change = True
 
1864
                        else:
 
1865
                            # directories have no fingerprint
 
1866
                            content_change = False
 
1867
                        target_exec = False
 
1868
                    elif target_kind == 'file':
 
1869
                        if source_minikind != 'f':
 
1870
                            content_change = True
 
1871
                        else:
 
1872
                            # We could check the size, but we already have the
 
1873
                            # sha1 hash.
 
1874
                            content_change = (link_or_sha1 != source_details[1])
 
1875
                        # Target details is updated at update_entry time
 
1876
                        if use_filesystem_for_exec:
 
1877
                            # We don't need S_ISREG here, because we are sure
 
1878
                            # we are dealing with a file.
 
1879
                            target_exec = bool(stat.S_IEXEC & path_info[3].st_mode)
 
1880
                        else:
 
1881
                            target_exec = target_details[3]
 
1882
                    elif target_kind == 'symlink':
 
1883
                        if source_minikind != 'l':
 
1884
                            content_change = True
 
1885
                        else:
 
1886
                            content_change = (link_or_sha1 != source_details[1])
 
1887
                        target_exec = False
 
1888
                    elif target_kind == 'tree-reference':
 
1889
                        if source_minikind != 't':
 
1890
                            content_change = True
 
1891
                        else:
 
1892
                            content_change = False
 
1893
                        target_exec = False
 
1894
                    else:
 
1895
                        raise Exception, "unknown kind %s" % path_info[2]
 
1896
                # parent id is the entry for the path in the target tree
 
1897
                if old_dirname == last_source_parent[0]:
 
1898
                    source_parent_id = last_source_parent[1]
 
1899
                else:
 
1900
                    source_parent_entry = state._get_entry(source_index,
 
1901
                                                           path_utf8=old_dirname)
 
1902
                    source_parent_id = source_parent_entry[0][2]
 
1903
                    if source_parent_id == entry[0][2]:
 
1904
                        # This is the root, so the parent is None
 
1905
                        source_parent_id = None
 
1906
                    else:
 
1907
                        last_source_parent[0] = old_dirname
 
1908
                        last_source_parent[1] = source_parent_id
 
1909
                        last_source_parent[2] = source_parent_entry
 
1910
                new_dirname = entry[0][0]
 
1911
                if new_dirname == last_target_parent[0]:
 
1912
                    target_parent_id = last_target_parent[1]
 
1913
                else:
 
1914
                    # TODO: We don't always need to do the lookup, because the
 
1915
                    #       parent entry will be the same as the source entry.
 
1916
                    target_parent_entry = state._get_entry(target_index,
 
1917
                                                           path_utf8=new_dirname)
 
1918
                    target_parent_id = target_parent_entry[0][2]
 
1919
                    if target_parent_id == entry[0][2]:
 
1920
                        # This is the root, so the parent is None
 
1921
                        target_parent_id = None
 
1922
                    else:
 
1923
                        last_target_parent[0] = new_dirname
 
1924
                        last_target_parent[1] = target_parent_id
 
1925
                        last_target_parent[2] = target_parent_entry
 
1926
 
 
1927
                source_exec = source_details[3]
 
1928
                return ((entry[0][2], (old_path, path), content_change,
 
1929
                        (True, True),
 
1930
                        (source_parent_id, target_parent_id),
 
1931
                        (old_basename, entry[0][1]),
 
1932
                        (_minikind_to_kind[source_minikind], target_kind),
 
1933
                        (source_exec, target_exec)),)
 
1934
            elif source_minikind in 'a' and target_minikind in 'fdlt':
 
1935
                # looks like a new file
 
1936
                if path_info is not None:
 
1937
                    path = pathjoin(entry[0][0], entry[0][1])
 
1938
                    # parent id is the entry for the path in the target tree
 
1939
                    # TODO: these are the same for an entire directory: cache em.
 
1940
                    parent_id = state._get_entry(target_index,
 
1941
                                                 path_utf8=entry[0][0])[0][2]
 
1942
                    if parent_id == entry[0][2]:
 
1943
                        parent_id = None
 
1944
                    if use_filesystem_for_exec:
 
1945
                        # We need S_ISREG here, because we aren't sure if this
 
1946
                        # is a file or not.
 
1947
                        target_exec = bool(
 
1948
                            stat.S_ISREG(path_info[3].st_mode)
 
1949
                            and stat.S_IEXEC & path_info[3].st_mode)
 
1950
                    else:
 
1951
                        target_exec = target_details[3]
 
1952
                    return ((entry[0][2], (None, path), True,
 
1953
                            (False, True),
 
1954
                            (None, parent_id),
 
1955
                            (None, entry[0][1]),
 
1956
                            (None, path_info[2]),
 
1957
                            (None, target_exec)),)
 
1958
                else:
 
1959
                    # but its not on disk: we deliberately treat this as just
 
1960
                    # never-present. (Why ?! - RBC 20070224)
 
1961
                    pass
 
1962
            elif source_minikind in 'fdlt' and target_minikind in 'a':
 
1963
                # unversioned, possibly, or possibly not deleted: we dont care.
 
1964
                # if its still on disk, *and* theres no other entry at this
 
1965
                # path [we dont know this in this routine at the moment -
 
1966
                # perhaps we should change this - then it would be an unknown.
 
1967
                old_path = pathjoin(entry[0][0], entry[0][1])
 
1968
                # parent id is the entry for the path in the target tree
 
1969
                parent_id = state._get_entry(source_index, path_utf8=entry[0][0])[0][2]
 
1970
                if parent_id == entry[0][2]:
 
1971
                    parent_id = None
 
1972
                return ((entry[0][2], (old_path, None), True,
 
1973
                        (True, False),
 
1974
                        (parent_id, None),
 
1975
                        (entry[0][1], None),
 
1976
                        (_minikind_to_kind[source_minikind], None),
 
1977
                        (source_details[3], None)),)
 
1978
            elif source_minikind in 'fdlt' and target_minikind in 'r':
 
1979
                # a rename; could be a true rename, or a rename inherited from
 
1980
                # a renamed parent. TODO: handle this efficiently. Its not
 
1981
                # common case to rename dirs though, so a correct but slow
 
1982
                # implementation will do.
 
1983
                if not osutils.is_inside_any(searched_specific_files, target_details[1]):
 
1984
                    search_specific_files.add(target_details[1])
 
1985
            elif source_minikind in 'ra' and target_minikind in 'ra':
 
1986
                # neither of the selected trees contain this file,
 
1987
                # so skip over it. This is not currently directly tested, but
 
1988
                # is indirectly via test_too_much.TestCommands.test_conflicts.
 
1989
                pass
 
1990
            else:
 
1991
                raise AssertionError("don't know how to compare "
 
1992
                    "source_minikind=%r, target_minikind=%r"
 
1993
                    % (source_minikind, target_minikind))
 
1994
                ## import pdb;pdb.set_trace()
 
1995
            return ()
 
1996
 
 
1997
        while search_specific_files:
 
1998
            # TODO: the pending list should be lexically sorted?  the
 
1999
            # interface doesn't require it.
 
2000
            current_root = search_specific_files.pop()
 
2001
            current_root_unicode = current_root.decode('utf8')
 
2002
            searched_specific_files.add(current_root)
 
2003
            # process the entries for this containing directory: the rest will be
 
2004
            # found by their parents recursively.
 
2005
            root_entries = _entries_for_path(current_root)
 
2006
            root_abspath = self.target.abspath(current_root_unicode)
 
2007
            try:
 
2008
                root_stat = os.lstat(root_abspath)
 
2009
            except OSError, e:
 
2010
                if e.errno == errno.ENOENT:
 
2011
                    # the path does not exist: let _process_entry know that.
 
2012
                    root_dir_info = None
 
2013
                else:
 
2014
                    # some other random error: hand it up.
 
2015
                    raise
 
2016
            else:
 
2017
                root_dir_info = ('', current_root,
 
2018
                    osutils.file_kind_from_stat_mode(root_stat.st_mode), root_stat,
 
2019
                    root_abspath)
 
2020
                if root_dir_info[2] == 'directory':
 
2021
                    if self.target._directory_is_tree_reference(
 
2022
                        current_root.decode('utf8')):
 
2023
                        root_dir_info = root_dir_info[:2] + \
 
2024
                            ('tree-reference',) + root_dir_info[3:]
 
2025
 
 
2026
            if not root_entries and not root_dir_info:
 
2027
                # this specified path is not present at all, skip it.
 
2028
                continue
 
2029
            path_handled = False
 
2030
            for entry in root_entries:
 
2031
                for result in _process_entry(entry, root_dir_info):
 
2032
                    # this check should probably be outside the loop: one
 
2033
                    # 'iterate two trees' api, and then _iter_changes filters
 
2034
                    # unchanged pairs. - RBC 20070226
 
2035
                    path_handled = True
 
2036
                    if (include_unchanged
 
2037
                        or result[2]                    # content change
 
2038
                        or result[3][0] != result[3][1] # versioned status
 
2039
                        or result[4][0] != result[4][1] # parent id
 
2040
                        or result[5][0] != result[5][1] # name
 
2041
                        or result[6][0] != result[6][1] # kind
 
2042
                        or result[7][0] != result[7][1] # executable
 
2043
                        ):
 
2044
                        yield (result[0],
 
2045
                               (utf8_decode_or_none(result[1][0]),
 
2046
                                utf8_decode_or_none(result[1][1])),
 
2047
                               result[2],
 
2048
                               result[3],
 
2049
                               result[4],
 
2050
                               (utf8_decode_or_none(result[5][0]),
 
2051
                                utf8_decode_or_none(result[5][1])),
 
2052
                               result[6],
 
2053
                               result[7],
 
2054
                              )
 
2055
            if want_unversioned and not path_handled and root_dir_info:
 
2056
                new_executable = bool(
 
2057
                    stat.S_ISREG(root_dir_info[3].st_mode)
 
2058
                    and stat.S_IEXEC & root_dir_info[3].st_mode)
 
2059
                yield (None,
 
2060
                       (None, current_root_unicode),
 
2061
                       True,
 
2062
                       (False, False),
 
2063
                       (None, None),
 
2064
                       (None, splitpath(current_root_unicode)[-1]),
 
2065
                       (None, root_dir_info[2]),
 
2066
                       (None, new_executable)
 
2067
                      )
 
2068
            initial_key = (current_root, '', '')
 
2069
            block_index, _ = state._find_block_index_from_key(initial_key)
 
2070
            if block_index == 0:
 
2071
                # we have processed the total root already, but because the
 
2072
                # initial key matched it we should skip it here.
 
2073
                block_index +=1
 
2074
            if root_dir_info and root_dir_info[2] == 'tree-reference':
 
2075
                current_dir_info = None
 
2076
            else:
 
2077
                dir_iterator = osutils._walkdirs_utf8(root_abspath, prefix=current_root)
 
2078
                try:
 
2079
                    current_dir_info = dir_iterator.next()
 
2080
                except OSError, e:
 
2081
                    # on win32, python2.4 has e.errno == ERROR_DIRECTORY, but
 
2082
                    # python 2.5 has e.errno == EINVAL,
 
2083
                    #            and e.winerror == ERROR_DIRECTORY
 
2084
                    e_winerror = getattr(e, 'winerror', None)
 
2085
                    # there may be directories in the inventory even though
 
2086
                    # this path is not a file on disk: so mark it as end of
 
2087
                    # iterator
 
2088
                    if e.errno in (errno.ENOENT, errno.ENOTDIR, errno.EINVAL):
 
2089
                        current_dir_info = None
 
2090
                    elif (sys.platform == 'win32'
 
2091
                          and ERROR_DIRECTORY in (e.errno, e_winerror)):
 
2092
                        current_dir_info = None
 
2093
                    else:
 
2094
                        raise
 
2095
                else:
 
2096
                    if current_dir_info[0][0] == '':
 
2097
                        # remove .bzr from iteration
 
2098
                        bzr_index = bisect_left(current_dir_info[1], ('.bzr',))
 
2099
                        assert current_dir_info[1][bzr_index][0] == '.bzr'
 
2100
                        del current_dir_info[1][bzr_index]
 
2101
            # walk until both the directory listing and the versioned metadata
 
2102
            # are exhausted. 
 
2103
            if (block_index < len(state._dirblocks) and
 
2104
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2105
                current_block = state._dirblocks[block_index]
 
2106
            else:
 
2107
                current_block = None
 
2108
            while (current_dir_info is not None or
 
2109
                   current_block is not None):
 
2110
                if (current_dir_info and current_block
 
2111
                    and current_dir_info[0][0] != current_block[0]):
 
2112
                    if current_dir_info[0][0] < current_block[0] :
 
2113
                        # filesystem data refers to paths not covered by the dirblock.
 
2114
                        # this has two possibilities:
 
2115
                        # A) it is versioned but empty, so there is no block for it
 
2116
                        # B) it is not versioned.
 
2117
                        # in either case it was processed by the containing directories walk:
 
2118
                        # if it is root/foo, when we walked root we emitted it,
 
2119
                        # or if we ere given root/foo to walk specifically, we
 
2120
                        # emitted it when checking the walk-root entries
 
2121
                        # advance the iterator and loop - we dont need to emit it.
 
2122
                        try:
 
2123
                            current_dir_info = dir_iterator.next()
 
2124
                        except StopIteration:
 
2125
                            current_dir_info = None
 
2126
                    else:
 
2127
                        # We have a dirblock entry for this location, but there
 
2128
                        # is no filesystem path for this. This is most likely
 
2129
                        # because a directory was removed from the disk.
 
2130
                        # We don't have to report the missing directory,
 
2131
                        # because that should have already been handled, but we
 
2132
                        # need to handle all of the files that are contained
 
2133
                        # within.
 
2134
                        for current_entry in current_block[1]:
 
2135
                            # entry referring to file not present on disk.
 
2136
                            # advance the entry only, after processing.
 
2137
                            for result in _process_entry(current_entry, None):
 
2138
                                # this check should probably be outside the loop: one
 
2139
                                # 'iterate two trees' api, and then _iter_changes filters
 
2140
                                # unchanged pairs. - RBC 20070226
 
2141
                                if (include_unchanged
 
2142
                                    or result[2]                    # content change
 
2143
                                    or result[3][0] != result[3][1] # versioned status
 
2144
                                    or result[4][0] != result[4][1] # parent id
 
2145
                                    or result[5][0] != result[5][1] # name
 
2146
                                    or result[6][0] != result[6][1] # kind
 
2147
                                    or result[7][0] != result[7][1] # executable
 
2148
                                    ):
 
2149
                                    yield (result[0],
 
2150
                                           (utf8_decode_or_none(result[1][0]),
 
2151
                                            utf8_decode_or_none(result[1][1])),
 
2152
                                           result[2],
 
2153
                                           result[3],
 
2154
                                           result[4],
 
2155
                                           (utf8_decode_or_none(result[5][0]),
 
2156
                                            utf8_decode_or_none(result[5][1])),
 
2157
                                           result[6],
 
2158
                                           result[7],
 
2159
                                          )
 
2160
                        block_index +=1
 
2161
                        if (block_index < len(state._dirblocks) and
 
2162
                            osutils.is_inside(current_root,
 
2163
                                              state._dirblocks[block_index][0])):
 
2164
                            current_block = state._dirblocks[block_index]
 
2165
                        else:
 
2166
                            current_block = None
 
2167
                    continue
 
2168
                entry_index = 0
 
2169
                if current_block and entry_index < len(current_block[1]):
 
2170
                    current_entry = current_block[1][entry_index]
 
2171
                else:
 
2172
                    current_entry = None
 
2173
                advance_entry = True
 
2174
                path_index = 0
 
2175
                if current_dir_info and path_index < len(current_dir_info[1]):
 
2176
                    current_path_info = current_dir_info[1][path_index]
 
2177
                    if current_path_info[2] == 'directory':
 
2178
                        if self.target._directory_is_tree_reference(
 
2179
                            current_path_info[0].decode('utf8')):
 
2180
                            current_path_info = current_path_info[:2] + \
 
2181
                                ('tree-reference',) + current_path_info[3:]
 
2182
                else:
 
2183
                    current_path_info = None
 
2184
                advance_path = True
 
2185
                path_handled = False
 
2186
                while (current_entry is not None or
 
2187
                    current_path_info is not None):
 
2188
                    if current_entry is None:
 
2189
                        # the check for path_handled when the path is adnvaced
 
2190
                        # will yield this path if needed.
 
2191
                        pass
 
2192
                    elif current_path_info is None:
 
2193
                        # no path is fine: the per entry code will handle it.
 
2194
                        for result in _process_entry(current_entry, current_path_info):
 
2195
                            # this check should probably be outside the loop: one
 
2196
                            # 'iterate two trees' api, and then _iter_changes filters
 
2197
                            # unchanged pairs. - RBC 20070226
 
2198
                            if (include_unchanged
 
2199
                                or result[2]                    # content change
 
2200
                                or result[3][0] != result[3][1] # versioned status
 
2201
                                or result[4][0] != result[4][1] # parent id
 
2202
                                or result[5][0] != result[5][1] # name
 
2203
                                or result[6][0] != result[6][1] # kind
 
2204
                                or result[7][0] != result[7][1] # executable
 
2205
                                ):
 
2206
                                yield (result[0],
 
2207
                                       (utf8_decode_or_none(result[1][0]),
 
2208
                                        utf8_decode_or_none(result[1][1])),
 
2209
                                       result[2],
 
2210
                                       result[3],
 
2211
                                       result[4],
 
2212
                                       (utf8_decode_or_none(result[5][0]),
 
2213
                                        utf8_decode_or_none(result[5][1])),
 
2214
                                       result[6],
 
2215
                                       result[7],
 
2216
                                      )
 
2217
                    elif current_entry[0][1] != current_path_info[1]:
 
2218
                        if current_path_info[1] < current_entry[0][1]:
 
2219
                            # extra file on disk: pass for now, but only
 
2220
                            # increment the path, not the entry
 
2221
                            advance_entry = False
 
2222
                        else:
 
2223
                            # entry referring to file not present on disk.
 
2224
                            # advance the entry only, after processing.
 
2225
                            for result in _process_entry(current_entry, None):
 
2226
                                # this check should probably be outside the loop: one
 
2227
                                # 'iterate two trees' api, and then _iter_changes filters
 
2228
                                # unchanged pairs. - RBC 20070226
 
2229
                                path_handled = True
 
2230
                                if (include_unchanged
 
2231
                                    or result[2]                    # content change
 
2232
                                    or result[3][0] != result[3][1] # versioned status
 
2233
                                    or result[4][0] != result[4][1] # parent id
 
2234
                                    or result[5][0] != result[5][1] # name
 
2235
                                    or result[6][0] != result[6][1] # kind
 
2236
                                    or result[7][0] != result[7][1] # executable
 
2237
                                    ):
 
2238
                                    yield (result[0],
 
2239
                                           (utf8_decode_or_none(result[1][0]),
 
2240
                                            utf8_decode_or_none(result[1][1])),
 
2241
                                           result[2],
 
2242
                                           result[3],
 
2243
                                           result[4],
 
2244
                                           (utf8_decode_or_none(result[5][0]),
 
2245
                                            utf8_decode_or_none(result[5][1])),
 
2246
                                           result[6],
 
2247
                                           result[7],
 
2248
                                          )
 
2249
                            advance_path = False
 
2250
                    else:
 
2251
                        for result in _process_entry(current_entry, current_path_info):
 
2252
                            # this check should probably be outside the loop: one
 
2253
                            # 'iterate two trees' api, and then _iter_changes filters
 
2254
                            # unchanged pairs. - RBC 20070226
 
2255
                            path_handled = True
 
2256
                            if (include_unchanged
 
2257
                                or result[2]                    # content change
 
2258
                                or result[3][0] != result[3][1] # versioned status
 
2259
                                or result[4][0] != result[4][1] # parent id
 
2260
                                or result[5][0] != result[5][1] # name
 
2261
                                or result[6][0] != result[6][1] # kind
 
2262
                                or result[7][0] != result[7][1] # executable
 
2263
                                ):
 
2264
                                yield (result[0],
 
2265
                                       (utf8_decode_or_none(result[1][0]),
 
2266
                                        utf8_decode_or_none(result[1][1])),
 
2267
                                       result[2],
 
2268
                                       result[3],
 
2269
                                       result[4],
 
2270
                                       (utf8_decode_or_none(result[5][0]),
 
2271
                                        utf8_decode_or_none(result[5][1])),
 
2272
                                       result[6],
 
2273
                                       result[7],
 
2274
                                      )
 
2275
                    if advance_entry and current_entry is not None:
 
2276
                        entry_index += 1
 
2277
                        if entry_index < len(current_block[1]):
 
2278
                            current_entry = current_block[1][entry_index]
 
2279
                        else:
 
2280
                            current_entry = None
 
2281
                    else:
 
2282
                        advance_entry = True # reset the advance flaga
 
2283
                    if advance_path and current_path_info is not None:
 
2284
                        if not path_handled:
 
2285
                            # unversioned in all regards
 
2286
                            if want_unversioned:
 
2287
                                new_executable = bool(
 
2288
                                    stat.S_ISREG(current_path_info[3].st_mode)
 
2289
                                    and stat.S_IEXEC & current_path_info[3].st_mode)
 
2290
                                if want_unversioned:
 
2291
                                    yield (None,
 
2292
                                        (None, utf8_decode_or_none(current_path_info[0])),
 
2293
                                        True,
 
2294
                                        (False, False),
 
2295
                                        (None, None),
 
2296
                                        (None, utf8_decode_or_none(current_path_info[1])),
 
2297
                                        (None, current_path_info[2]),
 
2298
                                        (None, new_executable))
 
2299
                            # dont descend into this unversioned path if it is
 
2300
                            # a dir
 
2301
                            if current_path_info[2] in ('directory'):
 
2302
                                del current_dir_info[1][path_index]
 
2303
                                path_index -= 1
 
2304
                        # dont descend the disk iterator into any tree 
 
2305
                        # paths.
 
2306
                        if current_path_info[2] == 'tree-reference':
 
2307
                            del current_dir_info[1][path_index]
 
2308
                            path_index -= 1
 
2309
                        path_index += 1
 
2310
                        if path_index < len(current_dir_info[1]):
 
2311
                            current_path_info = current_dir_info[1][path_index]
 
2312
                            if current_path_info[2] == 'directory':
 
2313
                                if self.target._directory_is_tree_reference(
 
2314
                                    current_path_info[0].decode('utf8')):
 
2315
                                    current_path_info = current_path_info[:2] + \
 
2316
                                        ('tree-reference',) + current_path_info[3:]
 
2317
                        else:
 
2318
                            current_path_info = None
 
2319
                        path_handled = False
 
2320
                    else:
 
2321
                        advance_path = True # reset the advance flagg.
 
2322
                if current_block is not None:
 
2323
                    block_index += 1
 
2324
                    if (block_index < len(state._dirblocks) and
 
2325
                        osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2326
                        current_block = state._dirblocks[block_index]
 
2327
                    else:
 
2328
                        current_block = None
 
2329
                if current_dir_info is not None:
 
2330
                    try:
 
2331
                        current_dir_info = dir_iterator.next()
 
2332
                    except StopIteration:
 
2333
                        current_dir_info = None
 
2334
 
2255
2335
 
2256
2336
    @staticmethod
2257
2337
    def is_compatible(source, target):
2258
2338
        # the target must be a dirstate working tree
2259
 
        if not isinstance(target, DirStateWorkingTree):
 
2339
        if not isinstance(target, WorkingTree4):
2260
2340
            return False
2261
 
        # the source must be a revtree or dirstate rev tree.
 
2341
        # the source must be a revtreee or dirstate rev tree.
2262
2342
        if not isinstance(source,
2263
2343
            (revisiontree.RevisionTree, DirStateRevisionTree)):
2264
2344
            return False
2265
2345
        # the source revid must be in the target dirstate
2266
 
        if not (source._revision_id == _mod_revision.NULL_REVISION or
 
2346
        if not (source._revision_id == NULL_REVISION or
2267
2347
            source._revision_id in target.get_parent_ids()):
2268
 
            # TODO: what about ghosts? it may well need to
 
2348
            # TODO: what about ghosts? it may well need to 
2269
2349
            # check for them explicitly.
2270
2350
            return False
2271
2351
        return True
2281
2361
 
2282
2362
    def convert(self, tree):
2283
2363
        # lock the control files not the tree, so that we dont get tree
2284
 
        # on-unlock behaviours, and so that noone else diddles with the
 
2364
        # on-unlock behaviours, and so that noone else diddles with the 
2285
2365
        # tree during upgrade.
2286
2366
        tree._control_files.lock_write()
2287
2367
        try:
2313
2393
 
2314
2394
    def update_format(self, tree):
2315
2395
        """Change the format marker."""
2316
 
        tree._transport.put_bytes('format',
2317
 
            self.target_format.as_string(),
2318
 
            mode=tree.bzrdir._get_file_mode())
2319
 
 
2320
 
 
2321
 
class Converter4to5(object):
2322
 
    """Perform an in-place upgrade of format 4 to format 5 trees."""
2323
 
 
2324
 
    def __init__(self):
2325
 
        self.target_format = WorkingTreeFormat5()
2326
 
 
2327
 
    def convert(self, tree):
2328
 
        # lock the control files not the tree, so that we don't get tree
2329
 
        # on-unlock behaviours, and so that no-one else diddles with the
2330
 
        # tree during upgrade.
2331
 
        tree._control_files.lock_write()
2332
 
        try:
2333
 
            self.update_format(tree)
2334
 
        finally:
2335
 
            tree._control_files.unlock()
2336
 
 
2337
 
    def update_format(self, tree):
2338
 
        """Change the format marker."""
2339
 
        tree._transport.put_bytes('format',
2340
 
            self.target_format.as_string(),
2341
 
            mode=tree.bzrdir._get_file_mode())
2342
 
 
2343
 
 
2344
 
class Converter4or5to6(object):
2345
 
    """Perform an in-place upgrade of format 4 or 5 to format 6 trees."""
2346
 
 
2347
 
    def __init__(self):
2348
 
        self.target_format = WorkingTreeFormat6()
2349
 
 
2350
 
    def convert(self, tree):
2351
 
        # lock the control files not the tree, so that we don't get tree
2352
 
        # on-unlock behaviours, and so that no-one else diddles with the
2353
 
        # tree during upgrade.
2354
 
        tree._control_files.lock_write()
2355
 
        try:
2356
 
            self.init_custom_control_files(tree)
2357
 
            self.update_format(tree)
2358
 
        finally:
2359
 
            tree._control_files.unlock()
2360
 
 
2361
 
    def init_custom_control_files(self, tree):
2362
 
        """Initialize custom control files."""
2363
 
        tree._transport.put_bytes('views', '',
2364
 
            mode=tree.bzrdir._get_file_mode())
2365
 
 
2366
 
    def update_format(self, tree):
2367
 
        """Change the format marker."""
2368
 
        tree._transport.put_bytes('format',
2369
 
            self.target_format.as_string(),
2370
 
            mode=tree.bzrdir._get_file_mode())
 
2396
        tree._control_files.put_utf8('format',
 
2397
            self.target_format.get_format_string())